components and indicators of frailty measures a literature review

Europe PMC requires Javascript to function effectively.

Either your web browser doesn't support Javascript or it is currently turned off. In the latter case, please turn on Javascript support in your web browser and reload this page.

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings
• My Bibliography
• Collections
• Citation manager

Save citation to file

Email citation, add to collections.

• Create a new collection
• Add to an existing collection

Add to My Bibliography

Your saved search, create a file for external citation management software, your rss feed.

• Search in PubMed
• Search in NLM Catalog
• Add to Search

The identification of frailty: a systematic literature review

Affiliation.

• 1 Maccabi Healthcare Services, Jerusalem, Israel. [email protected]
• PMID: 22091630
• DOI: 10.1111/j.1532-5415.2011.03597.x

An operational definition of frailty is important for clinical care, research, and policy planning. The literature on the clinical definitions, screening tools, and severity measures of frailty were systematically reviewed as part of the Canadian Initiative on Frailty and Aging. Searches of MEDLINE from 1997 to 2009 were conducted, and reference lists of retrieved articles were pearled, to identify articles published in English and French on the identification of frailty in community-dwelling people aged 65 and older. Two independent reviewers extracted descriptive information on study populations, frailty criteria, and outcomes from the selected papers, and quality rankings were assigned. Of 4,334 articles retrieved from the searches and 70 articles retrieved from the pearling, 22 met study inclusion criteria. In the 22 articles, physical function, gait speed, and cognition were the most commonly used identifying components of frailty, and death, disability, and institutionalization were common outcomes. The prevalence of frailty ranged from 5% to 58%. Despite significant work over the past decade, a clear consensus definition of frailty does not emerge from the literature. The definition and outcomes that best suit the unique needs of the researchers, clinicians, or policy-makers conducting the screening determine the choice of a screening tool for frailty. Important areas for further research include whether disability should be considered a component or an outcome of frailty. In addition, the role of cognitive and mood elements in the frailty construct requires further clarification.

© 2011, Copyright the Authors Journal compilation © 2011, The American Geriatrics Society.

PubMed Disclaimer

Similar articles

• Clinical features of prefrail older individuals and emerging peripheral biomarkers: a systematic review. Fernández-Garrido J, Ruiz-Ros V, Buigues C, Navarro-Martinez R, Cauli O. Fernández-Garrido J, et al. Arch Gerontol Geriatr. 2014 Jul-Aug;59(1):7-17. doi: 10.1016/j.archger.2014.02.008. Epub 2014 Mar 1. Arch Gerontol Geriatr. 2014. PMID: 24679669 Review.
• Measurement of gait speed in older adults to identify complications associated with frailty: A systematic review. Pamoukdjian F, Paillaud E, Zelek L, Laurent M, Lévy V, Landre T, Sebbane G. Pamoukdjian F, et al. J Geriatr Oncol. 2015 Nov;6(6):484-96. doi: 10.1016/j.jgo.2015.08.006. Epub 2015 Sep 9. J Geriatr Oncol. 2015. PMID: 26362356 Review.
• Toward a conceptual definition of frail community dwelling older people. Gobbens RJ, Luijkx KG, Wijnen-Sponselee MT, Schols JM. Gobbens RJ, et al. Nurs Outlook. 2010 Mar-Apr;58(2):76-86. doi: 10.1016/j.outlook.2009.09.005. Nurs Outlook. 2010. PMID: 20362776 Review.
• The assessment of frailty in older adults. Abellan van Kan G, Rolland Y, Houles M, Gillette-Guyonnet S, Soto M, Vellas B. Abellan van Kan G, et al. Clin Geriatr Med. 2010 May;26(2):275-86. doi: 10.1016/j.cger.2010.02.002. Clin Geriatr Med. 2010. PMID: 20497846 Review.
• Gait speed at usual pace as a predictor of adverse outcomes in community-dwelling older people an International Academy on Nutrition and Aging (IANA) Task Force. Abellan van Kan G, Rolland Y, Andrieu S, Bauer J, Beauchet O, Bonnefoy M, Cesari M, Donini LM, Gillette Guyonnet S, Inzitari M, Nourhashemi F, Onder G, Ritz P, Salva A, Visser M, Vellas B. Abellan van Kan G, et al. J Nutr Health Aging. 2009 Dec;13(10):881-9. doi: 10.1007/s12603-009-0246-z. J Nutr Health Aging. 2009. PMID: 19924348 Review.
• Severe aspiration pneumonia in the elderly. Ocrospoma S, Restrepo MI. Ocrospoma S, et al. J Intensive Med. 2024 Apr 10;4(3):307-317. doi: 10.1016/j.jointm.2023.12.009. eCollection 2024 Jul. J Intensive Med. 2024. PMID: 39035624 Free PMC article. Review.
• Racial disparities in inpatient palliative care consultation among frail older patients undergoing high-risk elective surgical procedures in the United States: a cross-sectional study of the national inpatient sample. Kim KM, Muench U, Maki JE, Yefimova M, Oh A, Jopling JK, Rinaldo F, Shah NR, Giannitrapani KF, Williams MY, Lorenz KA. Kim KM, et al. Health Aff Sch. 2023 Jul 13;1(2):qxad026. doi: 10.1093/haschl/qxad026. eCollection 2023 Aug. Health Aff Sch. 2023. PMID: 38756238 Free PMC article.
• POSITIVE: experiences of an intervention aiming for reversing and preventing frailty using a home monitoring and communication platform within primary health care. Timm L, Guidetti S, Taloyan M. Timm L, et al. BMC Geriatr. 2024 Apr 30;24(1):382. doi: 10.1186/s12877-024-04990-7. BMC Geriatr. 2024. PMID: 38689226 Free PMC article.
• Delirium is more common and associated with worse outcomes in Parkinson's disease compared to older adult controls: results of two prospective longitudinal cohort studies. Gerakios F, Yarnall AJ, Bate G, Wright L, Davis D, Stephan BCM, Robinson L, Brayne C, Stebbins G, Taylor JP, Burn DJ, Allan LM, Richardson SJ, Lawson RA. Gerakios F, et al. Age Ageing. 2024 Mar 1;53(3):afae046. doi: 10.1093/ageing/afae046. Age Ageing. 2024. PMID: 38497236 Free PMC article.
• A Clustering-Based Approach to Functional and Biomechanical Parameters Recorded with a Pair of Smart Eyeglasses in Older Adults in Order to Determine Physical Performance Groups. Hellec J, Colson SS, Jaafar A, Guérin O, Chorin F. Hellec J, et al. Sensors (Basel). 2024 Feb 22;24(5):1427. doi: 10.3390/s24051427. Sensors (Basel). 2024. PMID: 38474963 Free PMC article.

Publication types

• Search in MeSH

Related information

• Cited in Books

Grants and funding

• Canadian Institutes of Health Research/Canada

LinkOut - more resources

Full text sources.

• Ovid Technologies, Inc.
• MedlinePlus Health Information

Miscellaneous

• NCI CPTAC Assay Portal

• Citation Manager

NCBI Literature Resources

MeSH PMC Bookshelf Disclaimer

The PubMed wordmark and PubMed logo are registered trademarks of the U.S. Department of Health and Human Services (HHS). Unauthorized use of these marks is strictly prohibited.

Frailty in Young and Middle-Aged Adults: An Integrative Review

• Physical Frailty
• Published: 14 April 2021
• Volume 10 , pages 327–333, ( 2021 )

Cite this article

• Courtney Loecker 1 ,
• M. Schmaderer 1 &
• L. Zimmerman 1  

2807 Accesses

29 Citations

20 Altmetric

Explore all metrics

Frailty is a public health priority resulting in poor health outcomes and early mortality in older adults. Early identification, management, and prevention of frailty may reduce frailty trajectory into later life. However, little is known about frailty in younger adults.

Describe frailty prevalence, definitions, study designs, and components contributing to multidimensional frailty in 18 to 65-year-olds and impart guidance for future research, practice, and policies with potential to positively impact frail individuals.

Integrative review approach was selected to explore frailty allowing for inclusion of diverse methodologies and varied persepectives while maintaining rigor and applicability to evidence-based practice initiatives. CINAHL, Embase, PsycInfo, PubMed databases were searched for studies describing frailty in adults age 18 to 65. Articles were excluded if published prior to 2010, not in English, lacked frailty focus, or non-Western culture.

Twelve descriptive correlational studies were included. No intervention or qualitative studies were identified. No standard conceptual definition of frailty was discovered. Studied in participants with health disparities (n=3) and chronic conditions (n=8); HIV was most common (n=4). Frailty prevalence ranged from 3.9% (313 of 8095) to 63% (24 of 38). Many factors associated with frailty were identified among physical ( 18 ) and social ( 14 ), and fewer among psychological ( 11 ) domains.

Conclusions

Universal frailty definition and multidimensional assessment tool is needed to generate generalizable results in future studies describing frailty in young and middle-aged adults. Early frailty identification by clinicians has potential to facilitate development and implementation of targeted interventions to prevent or mitigate frailty progression, but additional research is needed because risk factors in younger populations may be different than older adults.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Subscribe and save.

• Get 10 units per month
• Download Article/Chapter or eBook
• 1 Unit = 1 Article or 1 Chapter
• Cancel anytime

Price includes VAT (Russian Federation)

Instant access to the full article PDF.

Rent this article via DeepDyve

Institutional subscriptions

Similar content being viewed by others

Prevalence of frailty in Indonesia: a systematic review and meta-analysis

The Association Between Frailty and All-Cause Mortality in Community-Dwelling Older Individuals: An Umbrella Review

Prevalence and Risk Factors for Frailty Among Community-Dwelling Older People in China: A Systematic Review and Meta-Analysis

Cesari M, Prince M, Thiyagarajan JA, et al. Frailty: an emerging public health priority. J Am Med Dir Assoc. 2016;17(3):188–192. doi: https://doi.org/10.1016/j.jamda.2015.12.016

Article   PubMed   Google Scholar  

Fried LP, Tangen CM, Walston J, et al. Frailty in older adults: evidence for a phenotype. The J of Gerontol A Biol Sci Med Sci. 2001;56:M146–M157.

Article   CAS   Google Scholar  

Hanlon P, Nicholl Bhautesh JD, et al. Frailty and prefrailty in middle-aged and older adults and its association with multimorbidity and mortality: a prospective analysis of 493737 uk biobank participants. Lancet. 2018;3(7):e323–e332.

Google Scholar  

Shamliyan T, Talley KM, Ramakrishnan R, et al. Association of frailty with survival: a systematic literature review. Ageing Res Rev. 2012;12:719–736.

Vermeiren S, Vella-Azzopardi R, Beckwée D, et al. Frailty and the prediction of negative health outcomes: A Meta-Analysis. J Am Med Dir Assoc. 2016;17(12):1163.e1–1163.e17. doi: https://doi.org/10.1016/j.jamda.2016.09.010

Article   Google Scholar  

Rockwood K, Mitnitski A. How might deficit accumulation give rise to frailty? J Frailty Aging. 2012;1:169–74.

Gobbens, RJ, Luijkx, KG, Winjnen-Sponselee, MT, et al. Towards an integral conceptual model in frailty. J Nutr Health and Aging. 2010;14(3):175–181.

Sezgin D, O’Donovan M, Cornally N, Liew, A, O’Caoimh, R. Defining frailty for healthcare practice and research: a qualitative systematic review with thematic analysis. International J Nurs Stud, 2019;92, 16–26.

Salem BE, Nyamathi A, Phillips P, Development of a frailty framework among vulnerable populations. Adv in Nurs Sci 2014;37(1):70–81.

Brunner EJ, Shipley MJ, Ahmadi-Abhari S, et al. Midlife contributors to socioeconomic differences in frailty during later life: a prospective cohort study. Lancet Public Health. 2018;3(7):e313–e322.

Article   PubMed   PubMed Central   Google Scholar  

Chamberlain AM, Finney Rutten LJ, Manemann SM, et al. Frailty trajectories in an elderly population-based cohort. J Am Geriatr Soc. 2016;64(2):285–292.

Feng Z, Lugtenberg M, Franse C, et al. Risk factors and protective factors associated with incident or increase of frailty among community-dwelling older adults: A systematic review of longitudinal studies. PLoS One. 2017;12(6):e0178383.

Article   PubMed   PubMed Central   CAS   Google Scholar  

Buta, BJ, Walston, JD, Godino, JG, et al. Frailty assessment instruments: systematic characterization of the uses and contexts of highly-cited instruments. Ageing Res Rev. 2016;26:53–61.

de Vries NM, Staal JB, van Ravensberg CD, et al. Outcome instruments to measure frailty: a systematic review. Ageing Res Rev. 2011;10(1):104–114.

Article   CAS   PubMed   Google Scholar  

Sepehri A, Beggs T, Hassan A, et al. The impact of frailty on outcomes after cardiac surgery: a systematic review. J Thorac Cardiovasc Surg. 2014;148(6):3110–3117. doi: https://doi.org/10.1016/j.jtcvs.2014.07.087

Huisingh-Scheetz M, Marinchek M, Becker Y, Ferguson M K, Thompson K. Translating frailty research into clinical practice: insights from the successful aging frailty evaluation clinic. JAMDA. 2019;20:672–678. https://doi.org/10.1016/j.jamda.2018.12.002

PubMed   Google Scholar  

Medically Frail. Nebraska Department of Health and Human Services. 2020. hhs.ne.gov/Pages/Medically-Frail.aspx. Accessed November 12, 2020.

Musumeci M, Chidambaram P, O’Malley Watts, M. Key state policy choices about medical determinations for medicaid expansions adults. Kaiser Family Foundation. 2019. https://www.kff.org/report-section/keystate-policy-choices-about-medical-frailty-determinations-for-medicaid-expansion-adults-issue-brief/ . Accessed November 1, 2019.

Rodriguez-Artalejo, F, Rodriguez-Mañas, L. The frailty syndrome in the public health agenda. The Journal of Epidemiology and Community Health, 2014;68:703–704. https://doi.org/10.1136/jech-2014-203863

Travers J, Romero-Ortuno R, Bailey J, Cooney, M T. Delaying and reversing frailty:a systematic review of primary care interventions. BR J Gen Pract. 2019;69(678):e61–e69.

Raymond E, Reynolds CA, Dahl Aslan AK, et al. Drivers of Frailty from Adulthood into Old Age: Results from a 27-Year Longitudinal Population-Based Study in Sweden. J Gerontol A Biol Sci Med Sci. 2020;75(10):1943–1950. doi: https://doi.org/10.1093/gerona/glaa106

Whittemore R, Knafl, K. The integrative review: updated methodology. J Adv Nurs. 2005;52(5):546–553.

Moher D, Liberati A, Tetzlaff J, Altman DG. The PRISMA Group. Preferred reporting items for systematic reviews and meta-analyses: the prisma statement. PLoS Med. 2009;6(7).

Langevin, H M, Schnyer R N. Reconnecting the body in eastern and western medicine. J Altern Complement Med. 2017;23(4):238–241. https://doi.org/10.1089/acm.2017.0028

Moola S, Munn Z, Tufanaru C, Aromataris E, Sears K, Sfetcu R, Currie M, Qureshi R, Mattis P, Lisy K, Mu P-F. Chapter 7: Systematic reviews of etiology and risk. In: Aromataris E, Munn Z (Editors). Joanna Briggs Institute Reviewer’s Manual. The Joanna Briggs Institute, 2017.

Chode S, Malmstrom T, Miller D, et al. Frailty, diabetes, and mortality in middle-aged african americans. J Nutr Health Aging 2016;20(8):854–859.

Salem BE, Brecht M, Ekstrand ML, et al. Correlates of physical, psychological, and social frailty among formerly incarcerated, homeless women. Health Care Women Int. 2019;40(7–9):788–812.

Rubtsova AA, Marquine M, Depp C, et al. Psychosocial correlates of frailty among HIV infected and HIV-uninfected adults. Behav Med 2019;45(3):210–220.

Althoff KN, Jacobson LP, Cranston, RD, et al. Age, comorbidities, and AIDS predict a frailty phenotype in men who have sex with men. The J of Gerontol Sci Med. 2014;69(2):189–198.

Demircioglu, DT. The association of vitamin D levels and the frailty phenotype among non-geriatric dialysis patients: a cross- sectional study. Clinics. 2018;73:e116.

Haider S, Grabovac I, Berner C, et al. Frailty in seropositive rheumatoid arthritis patients of working age: a cross-sectional study. Clin Exp Rheumatol. 2019;37(4):585–592.

Ness KK, Krull KR, Jones KE, et al. Physiologic frailty as a sign of accelerated aging among adult survivors of childhood cancer: a report from the st jude lifetime cohort study. J Clin Oncol. 2013;31(36):4496–4503.

Onen NF, Shacham E, Stamm KE, Overton, ET. Sexual behaviors and results of bacterial sexually transmitted infections testing among frail hiv-infected individuals. JIAPAC 2010;9(1):30–33.

Paolillo E W, Saloner, R, Montoya JL, et al. Frailty in comorbid hiv and lifetime methamphetamine use disorder: associations with neurocognitive and everyday functioning. AIDS Res Hum Retrovir. 2019;35(11–12):1044–1053.

Article   CAS   PubMed   PubMed Central   Google Scholar  

Palmer KT, D’Angelo S, Harris EC, et al. Frailty, prefrailty and employment outcomes in health and employment after fifty (HEAF) study. Occup Environ Med. 2017;74(7):476–482.

Hadenfeldt C J, Darabaris M, Aufdenkamp M. Frailty assessment in patients utilizing a free clinic. J of Health Care Poor Underserved 2017;28(4):1423–1435.

Griffin, F, Mode, N, Ejiogu, N, Zonderman, A, Michele, E. Frailty in a racially and socioeconomically diverse sample of middle-aged americans in baltimore. PLOS One. 2018;13(4):1–12.

Collard RM, Boter H, Schoevers RA, et al. Prevalence of frailty in community-dwelling older persons: a systematic review. J Am Geriatr Soc. 2012;60(8):1487–92. doi: https://doi.org/10.1111/j.1532-5415.2012.04054

Xie, B, Larson, JL, Gonzalez, R. Components and indicators of frailty measures: a literature review. J Frailty Aging. 2016;1:1–6

Syddall H, Cooper C, Martin F, Briggs R, Aihie Sayer A. Is grip strength a useful single marker of frailty? Age Ageing. 2003;32(6):650–656. doi: https://doi.org/10.1093/ageing/afg111

Lee, L, Patel, T Hillier, LM, et al. Identifying frailty in primary care: a systematic review. Geriatr Gerontol Int. 2017;17:1358–1377

Song X, Mitnitski A, Rockwood K. Prevalence and 10-year outcomes of frailty in older adults in relation to deficit accumulation. J Am Geriatr Soc. 2010;58(4):681–687. doi: https://doi.org/10.1111/j.1532-5415.2010.02764

Leng SX, Margolick JB. Understanding frailty, aging, and inflammation in HIV infection. Curr HIV/AIDS Rep. 2015;12(1):25–32. doi: https://doi.org/10.1007/s11904-014-0247-3

Sinclair AJ, Rodriguez-Mañas L Diabetes and frailty: two converging conditions? Can J Diabetes. 2016 Feb;40(1):77–83

Ensrud KE, Kats AM, Schousboe JT, et al. Frailty phenotype and healthcare costs and utilization in older women. J Am Geriatr Soc. 2018;66(7):1276–1283. doi: https://doi.org/10.1111/jgs.15381

Hajek A, Bock JO, Saum KU, et al. Frailty and healthcare costslongitudinal results of a prospective cohort study. Age Ageing. 2018;47(2):233–241. doi: https://doi.org/10.1093/ageing/afx157

García-Nogueras I, Aranda-Reneo I, Peña-Longobardo LM, Oliva-Moreno J, Abizanda P. Use of Health Resources and Healthcare Costs associated with Frailty: The FRADEA Study. J Nutr Health Aging. 2017;21(2):207–214. doi: https://doi.org/10.1007/s12603-016-0727-9

Belsky DW, Caspi A, Cohen HJ, et al. Impact of early personal-history characteristics on the Pace of Aging: implications for clinical trials of therapies to slow aging and extend healthspan. Aging Cell. 2017;16(4):644–651. doi: https://doi.org/10.1111/acel.12591

Hoogendijk EO, Afilalo J, Ensrud KE, Kowal P, Onder G, Fried LP. Frailty: implications for clinical practice and public health. Lancet. 2019;394(10206):1365–1375.

Lafortune L, Martin S, Kelly S, et al. Behavioural Risk Factors in Mid-Life Associated with Successful Ageing, Disability, Dementia and Frailty in Later Life: A Rapid Systematic Review. PLoS One. 2016;11(2):e0144405. Published 2016 Feb 4. https://doi.org/10.1371/journal.pone.0144405

Serra-Prat M, Sist X, Domenich R, et al. Effectiveness of an intervention to prevent frailty in pre-frail community-dwelling older people consulting in primary care: a randomised controlled trial. Age Ageing. 2017;46(3):401–407. doi: https://doi.org/10.1093/ageing/afw242

CAS   PubMed   Google Scholar  

de Haas SCM, de Jonge, EAL, Voortman T, et al. Dietary patterns and changes in frailty status: the rotterdam study. Eur J Nutr. 2018;57(7):2365–2375.

Escourrou, E, Herault, M, Gdoura, S, et al. Becoming frail: a major turning point in patients’ life course. Fam Pract. 2019;36(2):231–236.

Lekan, DA, Hoover, E, Abrams, S. Perspectives of frailty among african american men and women. J Psychocos Nurs Ment Health Serv. 2018;56(7):20–29.

Meleis, A. Theoretical nursing: development and progress, 5th edn. 2012. Wolters Kluwer/Lippincott Williams & Wilkins. Philadelphia, PA.

Download references

Author information

Authors and affiliations.

College of Nursing, University of Nebraska Medical Center, Omaha, NE, USA

Courtney Loecker, M. Schmaderer & L. Zimmerman

You can also search for this author in PubMed   Google Scholar

Corresponding author

Correspondence to Courtney Loecker .

Ethics declarations

This article does not contain studies with human participants performed by any of the authors.

Additional information

Disclosure statement.

We have no disclosures.

Supporting information

Rights and permissions.

Reprints and permissions

About this article

Loecker, C., Schmaderer, M. & Zimmerman, L. Frailty in Young and Middle-Aged Adults: An Integrative Review. J Frailty Aging 10 , 327–333 (2021). https://doi.org/10.14283/jfa.2021.14

Download citation

Received : 20 December 2020

Accepted : 09 March 2021

Published : 14 April 2021

Issue Date : October 2021

DOI : https://doi.org/10.14283/jfa.2021.14

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• medical frailty
• young adult
• Find a journal
• Publish with us
• Track your research

• Geriatric Medicine
• Internal Medicine

The Identification of Frailty: A Systematic Literature Review

• September 2011
• Journal of the American Geriatrics Society 59(11):2129-38
• 59(11):2129-38

• Maccabi Healthcare Services
• This person is not on ResearchGate, or hasn't claimed this research yet.

• Ottawa Hospital Research Institute

• McGill University

Abstract and Figures

Discover the world's research

• 25+ million members
• 160+ million publication pages
• 2.3+ billion citations
• Hongfei Yang
• Jiangeng Chang
• Mengling Feng
• Jed Montayre
• Ka Man Carman Leung
• BMC Geriatr
• Haijian Guo

• Emmanuel Andres

• Matthew Stamets

• Amirah Abumismar
• Saulat Jahan

• Erica Twardzik
• Laura Samuel
• Qian-Li Xue
• Feifei Chong

• Taleen Haddad

• Patrick P. Coll

• J Am Med Assoc
• Stephanie A Studenski
• Subashan Perera

• Jack Guralnik
• J AM COLL SURGEONS

• Dorry L. Segev

• BMC MED RES METHODOL

• Marc D Rothman

• Laura S. Ho

• Emily A.W. Hardwick
• J AM GERIATR SOC
• Elizabeth A Freiheit
• Misha Eliasziw
• Colleen J. Maxwell
• CLIN GERIATR MED
• Gabor Abellan van Kan
• Yves Rolland
• Mathieu Houles
• Bruno Vellas

• Johanne Monette
• Véronique Girre
• Recruit researchers
• Join for free
• Login Email Tip: Most researchers use their institutional email address as their ResearchGate login Password Forgot password? Keep me logged in Log in or Continue with Google Welcome back! Please log in. Email · Hint Tip: Most researchers use their institutional email address as their ResearchGate login Password Forgot password? Keep me logged in Log in or Continue with Google No account? Sign up

ROBERT ALLISON, II, DO, SHAUNA ASSADZANDI, MD, AND MEGAN ADELMAN, PharmD

This is a corrected version of the article that appeared in print.

Am Fam Physician. 2021;103(4):219-226

Related Lown Right Care: Preoperative Evaluation and Frailty Assessment in Older Patients

Author disclosure: No relevant financial affiliations.

Frailty, which is a geriatric syndrome that affects 5% to 17% of older adults, is a state of increased vulnerability across multiple health domains that leads to adverse health outcomes. Frail older adults are at increased risk of falls, disability, hospitalizations, and death. Frailty may initially be overlooked or incorrectly identified as part of the normal aging process because of the variable nature of the presentation and diagnosis. Symptoms include generalized weakness, exhaustion, slow gait, poor balance, decreased physical activity, cognitive impairment, and weight loss. There is no current recommendation for routine screening. A comprehensive geriatric assessment can identify risk factors and symptoms that suggest frailty. Several validated frailty assessment tools can evaluate a patient for frailty. Patients are diagnosed as not-frail, prefrail, or frail. Patients with a larger number of frail attributes are at higher risk of poor outcomes. The management of frail patients must be individualized and tailored to each patient's goals of care and life expectancy. Physical activity and balance exercises may be suitable for patients who are less frail. Palliative care and symptom control may be appropriate for those who are more frail.

Frailty is a syndrome of growing importance among the geriatric population, occurring in 5% to 17% of older adults. 1 , 2 Clinicians need to recognize the signs and symptoms of frailty as the average life expectancy of the population continues to increase. Several studies have demonstrated that frail adults are at increased risk of adverse health outcomes. 3 – 5 However, there is no unifying definition for frailty. It is a syndrome recognized primarily in older adults that affects health, energy, and physical abilities by increasing a patient's vulnerability to stressors (e.g., falls, infection) and risk of further decline. 1 , 6 , 7 Recognizing frailty and understanding its progression will help physicians develop treatment plans and better discuss prognosis with patients and their families.

Pathophysiology

The normal aging process is a cumulative result of molecular and cellular damage that leads to a loss of physiologic reserve. A patient's physiologic reserve provides the ability to compensate for disease-related changes and maintain a homeostatic balance in the natural aging process. There is no one specific organ system that causes frailty but rather an aggregate loss across multiple systems. 8 – 11 On the cellular level, altered processes in mitochondria and protein processing, increased free radical concentrations, and amplified sensitivity to apoptosis contribute to systemic changes. 12 Key age-related changes include hormonal dysregulation (e.g., increased cortisol levels), sarcopenia (i.e., loss of skeletal and muscle mass), and increased immune system activation and proinflammatory cytokines. 10 , 11 These changes cause a loss of homeostasis and a significantly decreased physiologic reserve, making patients who are frail more vulnerable to functional decline. As a result, these patients do not adapt to the stress of systemic disease as well as patients who are not frail.

Risk Factors

The estimated prevalence of frailty in the community setting is inconsistent in the literature, and incidence is likely underreported. 2 , 13 , 14 Prevalence is hard to estimate because frailty is multifactorial, with older age, female sex, unhealthy lifestyle, and lower economic status identified as potential risk factors. 2 , 13 – 16 Social factors such as marital status, smoking history, social isolation, and lower levels of education also put people at risk. 17 – 20 Research suggests that frailty increases with the number of health deficits and presence of multiple comorbidities. 2 Patients diagnosed with diabetes mellitus, respiratory disease, stroke, dementia, multiple sclerosis, connective tissue disease, osteoarthritis, and chronic fatigue syndrome have higher documented frailty rates. 15 , 21 , 22

The relationship between medication use and frailty is not well defined. Some studies demonstrated that polypharmacy (taking five or more medications) was associated with frailty; however, potentially inappropriate medication use as defined by the Beers Criteria for Potentially Inappropriate Medication Use in Older Adults was not associated. 23 – 26 One study indicated that frailty can be a risk factor for polypharmacy. 27

Presentation

Frailty is a dynamic state of well-being involving multiple health domains that are influenced by a range of variables. 28 Patients who are frail share many common physical traits, but there is no hallmark sign or symptom that is pathognomonic to confirm the diagnosis. Early frailty symptoms often involve generalized weakness and exhaustion. 7 , 29 Other symptoms include slow gait, poor balance, decreased physical activity, decreased strength, and cognitive impairment. Weight loss is typically a sign of late-stage frailty. 7 , 29

Assessing Frailty

There are currently no recommendations to routinely screen for frailty. However, to optimize care and communicate appropriately with families, clinicians may need to identify patients at risk and those who may already be frail. One of the most effective screening approaches for frailty is a comprehensive geriatric assessment, a systematic, multidimensional assessment conducted by an interprofessional team of geriatric health professionals, including primary care physicians familiar with the components of a geriatric assessment. 25 , 30 This assessment addresses social, environmental, and medical determinants of health rather than focusing on individual disease states. Clinicians assess a patient's medical comorbidities, polypharmacy, functional abilities (activities of daily living), fall risk, hearing, vision, mental health, and cognition. 25 , 30 , 31 Impairment in any one of these areas can be a risk factor for frailty and requires further evaluation.

There is a growing interest among specialists—including cardiologists, surgeons, and oncologists—about screening for frailty as a predictor for health care outcomes. 32 – 35 During the coronavirus disease 2019 (COVID-19) pandemic, some critical care physicians have screened for frailty to assist in medical decision-making for patients hospitalized with COVID-19. 36 – 38 Primary care physicians and specialists may wish to collaborate with a geriatrician if a diagnosis is uncertain or if time limitations are a barrier. 25 Specific aspects of the assessment can span multiple visits, making it more feasible to implement into a busy family medicine practice. Medicare Annual Wellness Visits also provide opportunities to introduce some of the screening tools into a patient's visit 3 ( https://www.aafp.org/fpm/2019/0300/p25.html ).

Frailty is not defined by a single patient-reported symptom or physical examination finding, and no laboratory tests or imaging studies can diagnose frailty. Rather, frailty is assessed through a comprehensive history and physical examination, focusing on several key elements. A patient thought to be frail should be evaluated using validated frailty assessment tools. 39 Two of the most common assessment tools are the Fried frailty phenotype and the Rockwood frailty index. 40 Both tools use the history and physical examination components to assess a patient's degree of frailty based on predetermined variables that have been studied and used in research and clinical practice. 1 , 6

The Fried frailty phenotype uses five predefined variables to assess patients assumed to be frail. Patients are evaluated for unintentional weight loss, weakness, slowness, poor endurance, and low physical activity. 1 Table 1 includes the minimum values and further definitions and resources needed to score each variable. 1 [corrected]

Grip strength is often evaluated by using a dynamometer. 1 This device is held in the dominant hand with the arm flexed to 90 degrees, the elbow at the side of the body, and the handle adjusted so the distal interphalangeal joint wraps around the device. The patient squeezes the dynamometer with a maximum isometric grip for five seconds, followed by a 15-second relaxation period. The best of the three measurements is compared with minimum grip strength values based on sex and body mass index. A value below the minimum cutoff meets the criteria for frailty. A dynamometer is not commonly found in outpatient offices but may be available in some geriatric assessment clinics. Most physical therapy or occupational hand therapy departments will have dynamometers.

Slowness is measured by the time it takes a patient to walk 15 ft (4.5 m). Maximum time allotments are defined by sex and height ( Table 1 ) . 1 Time values slower than expected meet the criteria for frailty.

The questions “I felt that everything I did was an effort in the past week” and “I could not get going in the past week” from the Center for Epidemiological Studies Depression Scale are used to measure endurance. 41

A modified version of the Minnesota Leisure Time Physical Activity Questionnaire is used to determine physical activity by calculating kilocalorie expenditure per week based on a patient's sex. Energy expenditure below the predefined minimum values meets the criteria for frailty. Table 2 lists the Minnesota leisure activities, associated metabolic equivalents, and kcal expenditure formula. 42 , 43

The Rockwood frailty index is an assessment tool that includes a larger number of health indices in patients suspected to be frail. This frailty index uses predefined variables to assess a patient's level of independence, evaluate health history, review physical examination findings, assess cognitive function, and review abnormal laboratory results. 6 Each item is scored based on the variable being present or absent. The total number of deficits present are divided by the total number of variables, generating a frailty score between 0 and 1, with a value of 0.25 or greater suggesting frailty. 39

The Fried frailty phenotype and the Rockwood frailty index have advanced the field of frailty research. However, their use may prove cumbersome and impractical in certain clinical settings. Several other validated frailty assessment tools, which include elements similar to those used in the frailty phenotype and the frailty index, may be easier to apply in practice. No frailty tool has been proven superior to another. 39 Physicians must choose a frailty assessment tool that is appropriate for their patient demographic and practice setting.

Older adults who are frail have a higher likelihood of poor health outcomes that include falls, hospitalizations, institutionalization, disability, and death. 1 , 3 – 6 , 44 Compared with nonfrail adults, being diagnosed as prefrail or frail is predictive of a 1.3- to 2.6-fold worsening mobility, decreased activities of daily living, and an overall increased rate of falls, disability, hospitalization, and death. 1 , 45 , 46 Women are more often diagnosed as frail, but men diagnosed with frailty have a higher mortality rate. 45

Several studies have evaluated the dynamic nature of frailty. Over time, transitioning between frailty stages (not-frail, prefrail, frail) can occur, with patients worsening, improving, or maintaining their current degree of frailty. 3 , 22 Research shows that women and patients living in better socioeconomic conditions have a higher likelihood of improving their frailty status. 21 , 45 One study suggested that older adults who are prefrail have a better chance than their frail cohorts to improve their frailty diagnosis. 47 However, dementia and cancer limit the chances of improving frailty status. 46

When considering management options, it is important to recognize that patients diagnosed with frailty vary in their presentation and treatment needs. Evidence for an effective, comprehensive care plan is emerging, but there is a lack of quality evidence-based literature to support any specific frailty treatment plan fully. In 2019, the task force of the International Conference on Frailty and Sarcopenia Research (ICFSR) developed clinical practice guidelines for the identification and management of frailty. 48 Table 3 outlines the ICFSR frailty management recommendations, based on moderate to very low quality of evidence. 48 Clinical judgment is critical when choosing a treatment option. Some older adults may not be capable of initiating or completing certain treatments, and in certain situations, the risk of treatment may outweigh the benefit. In later stages of frailty, palliative care options may be appropriate to discuss.

Physicians are encouraged to develop a patient-specific treatment plan based on shared decision-making. Management should align with each patient's goals of care and life expectancy. Goals of care should be reviewed periodically, especially when there is a change in frailty status, to ensure that the physician, patient, and patient's family have a clear understanding of management options based on life expectancy. The Palliative and Therapeutic Harmonization model is a standardized system for frail patients to help with health care management. The model focuses on obtaining a comprehensive history, ensuring that the patient and family comprehend the patient's vulnerability and potentially shortened life span, assisting with health care decisions that consider frailty, and improving the ability of the team's emergency response system to prepare for appropriate support in the setting of sudden deterioration while paralleling goals of care. 49 , 50

Research focused on early interventions to prevent or reduce the level of frailty in community-dwelling older adults identified physical activity, nutritional support, and psychosocial engagement as possible areas of benefit. 51 – 57 Studies have evaluated these variables independently and in combination as part of an interdisciplinary approach. 51 , 52 Physical activity aims to improve strength and balance based on the American College of Sports Medicine guidelines for older adults. 56 Table 4 and Table 5 outline strength training and balance training prescriptions for patients who are frail. 55 – 57 Several European studies demonstrated a reduction in frailty and prevention of frailty progression when nutritional education was added to a physical activity routine. 58 – 60 However, one study involving intensive, multidisciplinary care demonstrated no improvement in overall frailty measures or functional decline. 61 Although there is no approved medication to treat any aspect of frailty, addressing polypharmacy may reduce the risk of becoming frail. 23 , 24 Ongoing research is necessary to help better identify optimal treatment strategies.

Data Sources: PubMed and Google Scholar were searched using the key terms (alone and in combination) geriatric, frailty, syndrome, older adult, elderly, frailty index, frailty phenotype, classification, criteria, comprehensive geriatric assessment, fitness, management, and polypharmacy. The search included meta-analyses, randomized controlled trials, clinical trials, and reviews. Also searched were the Cochrane database, Essential Evidence Plus, the Institute for Clinical Systems Improvement, and DynaMed. Search dates: January 15 to 27, 2020, and November 2, 2020.

Fried LP, Tangen CM, Walston J, et al.; Cardiovascular Health Study Collaborative Research Group. Frailty in older adults: evidence for a phenotype. J Gerontol A Biol Sci Med Sci. 2001;56(3):M146-M156.

Collard RM, Boter H, Schoevers RA, et al. Prevalence of frailty in community-dwelling older persons: a systematic review. J Am Geriatr Soc. 2012;60(8):1487-1492.

Gill TM, Gahbauer EA, Allore HG, et al. Transitions between frailty states among community-living older persons. Arch Intern Med. 2006;166(4):418-423.

Graham JE, Snih SA, Berges IM, et al. Frailty and 10-year mortality in community-living Mexican American older adults. Gerontology. 2009;55(6):644-651.

Ensrud KE, Ewing SK, Cawthon PM, et al.; Osteoporotic Fractures in Men Research Group. A comparison of frailty indexes for the prediction of falls, disability, fractures, and mortality in older men. J Am Geriatr Soc. 2009;57(3):492-498.

Rockwood K, Song X, MacKnight C, et al. A global clinical measure of fitness and frailty in elderly people. CMAJ. 2005;173(5):489-495.

Xue QL, Bandeen-Roche K, Varadhan R, et al. Initial manifestations of frailty criteria and the development of frailty phenotype in the Women's Health and Aging Study II. J Gerontol A Biol Sci Med Sci. 2008;63(9):984-990.

Lipsitz LA. Dynamics of stability: the physiologic basis of functional health and frailty. J Gerontol A Biol Sci Med Sci. 2002;57(3):B115-B125.

Fried LP, Xue QL, Cappola AR, et al. Nonlinear multisystem physiological dysregulation associated with frailty in older women: implications for etiology and treatment. J Gerontol A Biol Sci Med Sci. 2009;64(10):1049-1057.

Clegg A, Young J. The frailty syndrome. Clin Med (Lond). 2011;11(1):72-75.

Chen X, Mao G, Leng SX. Frailty syndrome: an overview. Clin Interv Aging. 2014;9:433-441.

Walston JD. Connecting age-related biological decline to frailty and late-life vulnerability. Nestle Nutr Inst Workshop Ser. 2015;83:1-10.

Siriwardhana DD, Hardoon S, Rait G, et al. Prevalence of frailty and prefrailty among community-dwelling older adults in low-income and middle-income countries: a systematic review and meta-analysis. BMJ Open. 2018;8(3):e018195.

Ofori-Asenso R, Chin KL, Mazidi M, et al. Global incidence of frailty and pre-frailty among community-dwelling older adults: a systematic review and meta-analysis. JAMA Netw Open. 2019;2(8):e198398.

Hanlon P, Nicholl BI, Jani BD, et al. Frailty and pre-frailty in middle-aged and older adults and its association with multimorbidity and mortality: a prospective analysis of 493 737 UK Biobank participants. Lancet Public Health. 2018;3(7):e323-e332.

Puts MTE, Lips P, Deeg DJH. Sex differences in the risk of frailty for mortality independent of disability and chronic diseases. J Am Geriatr Soc. 2005;53(1):40-47.

Chamberlain AM, St Sauver JL, Jacobson DJ, et al. Social and behavioural factors associated with frailty trajectories in a population-based cohort of older adults. BMJ Open. 2016;6(5):e011410.

de Labra C, Maseda A, Lorenzo-López L, et al. Social factors and quality of life aspects on frailty syndrome in community-dwelling older adults: the VERISAÚDE study. BMC Geriatr. 2018;18(1):66.

Fhon JRS, Rodrigues RAP, Santos JLF, et al. Factors associated with frailty in older adults: a longitudinal study. Rev Saude Publica. 2018;52:74.

Gobbens RJJ, van Assen MALM, Luijkx KG, et al. Determinants of frailty. J Am Med Dir Assoc. 2010;11(5):356-364.

Lee JSW, Auyeung T-W, Leung J, et al. Transitions in frailty states among community-living older adults and their associated factors. J Am Med Dir Assoc. 2014;15(4):281-286.

Espinoza SE, Jung I, Hazuda H. Frailty transitions in the San Antonio Longitudinal Study of Aging. J Am Geriatr Soc. 2012;60(4):652-660.

Aprahamian I, Biella MM, Vano Aricó de Almeida G, et al. Polypharmacy but not potential inappropriate prescription was associated with frailty in older adults from a middle-income country outpatient clinic. J Frailty Aging. 2018;7(2):108-112.

Saum KU, Schöttker B, Meid AD, et al. Is polypharmacy associated with frailty in older people? Results from the ESTHER cohort study. J Am Geriatr Soc. 2017;65(2):e27-e32.

Tatum PE, Talebreza S, Ross JS. Geriatric assessment: an office-based approach. Am Fam Physician. 2018;97(12):776-784. Accessed August 10, 2020. https://www.aafp.org/afp/2018/0615/p776.html

2019 American Geriatrics Society Beers Criteria Update Expert Panel. American Geriatrics Society 2019 updated AGS Beers Criteria for potentially inappropriate medication use in older adults. J Am Geriatr Soc. 2019;67(4):674-694.

Rieckert A, Trampisch US, Klaaßen-Mielke R, et al. Polypharmacy in older patients with chronic diseases: a cross-sectional analysis of factors associated with excessive polypharmacy. BMC Fam Pract. 2018;19(1):113.

Gobbens RJJ, Luijkx KG, Wijnen-Sponselee MT, et al. In search of an integral conceptual definition of frailty: opinions of experts. J Am Med Dir Assoc. 2010;11(5):338-343.

Stenholm S, Ferrucci L, Vahtera J, et al. Natural course of frailty components in people who develop frailty syndrome: evidence from two cohort studies. J Gerontol A Biol Sci Med Sci. 2019;74(5):667-674.

Welsh TJ, Gordon AL, Gladman JR. Comprehensive geriatric assessment—a guide for the non-specialist. Int J Clin Pract. 2014;68(3):290-293.

Hamaker ME, Jonker JM, de Rooij SE, et al. Frailty screening methods for predicting outcome of a comprehensive geriatric assessment in elderly patients with cancer: a systematic review. Lancet Oncol. 2012;13(10):e437-e444.

Partridge JSL, Harari D, Dhesi JK. Frailty in the older surgical patient: a review. Age Ageing. 2012;41(2):142-147.

Freiheit EA, Hogan DB, Patten SB, et al. Frailty trajectories after treatment for coronary artery disease in older patients. Circ Cardiovasc Qual Outcomes. 2016;9(3):230-238.

McIsaac DI, Taljaard M, Bryson GL, et al. Frailty as a predictor of death or new disability after surgery: a prospective cohort study. Ann Surg. 2020;271(2):283-289.

van Kempen JAL, Melis RJF, Perry M, et al. Diagnosis of frailty after a Comprehensive Geriatric Assessment: differences between family physicians and geriatricians. J Am Board Fam Med. 2015;28(2):240-248.

National Institute for Health and Care Excellence. COVID-19 rapid guideline: critical care in adults. March 20, 2020. Accessed August 3, 2020. https://www.nice.org.uk/guidance/ng159

Bellelli G, Rebora P, Valsecchi MG, et al. Frailty index predicts poor outcome in COVID-19 patients. Intensive Care Med. 2020;46(8):1634-1636.

Van de Voorde P, Monsieurs K, Renier W, et al. Ethical principles and guidance with regard to ethical decisions in pre-hospital and emergency medicine in Belgium during the COVID-19 pandemic. A joint statement of the Belgian Society of Emergency and Disaster Medicine and the Belgian Resuscitation Council. March 22, 2020. Accessed August 10, 2020. https://www.brc-rea.be/wp-content/uploads/2020/03/Ethical-decision-making-in-emergencies_COVID19_22032020_final.pdf

Dent E, Martin FC, Bergman H, et al. Management of frailty: opportunities, challenges, and future directions. Lancet. 2019;394(10206):1376-1386.

Hoogendijk EO, van Kan GA, Guyonnet S, et al. Components of the frailty phenotype in relation to the frailty index: results from the Toulouse frailty platform. J Am Med Dir Assoc. 2015;16(10):855-859.

Radloff LS. The CES-D scale: a self-report depression scale for research in the general population. Appl Psychol Meas. 1977;1:385-401.

Ainsworth BE, Haskell WL, Leon AS, et al. Compendium of physical activities: classification of energy costs of human physical activities. Med Sci Sports Exerc. 1993;25(1):71-80.

Taylor HL, Jacobs DR, Schucker B, et al. A questionnaire for the assessment of leisure time physical activities. J Chronic Dis. 1978;31(12):741-755.

Woods NF, LaCroix AZ, Gray SL, et al.; Women's Health Initiative. Frailty: emergence and consequences in women aged 65 and older in the Women's Health Initiative Observational Study [published correction appears in J Am Geriatr Soc . 2017;65(7):1631–1632]. J Am Geriatr Soc. 2005;53(8):1321-1330.

Vermeiren S, Vella-Azzopardi R, Beckwée D, et al. Frailty and the prediction of negative health outcomes: a meta-analysis. J Am Med Dir Assoc. 2016;17(12):1163e1-1163e17.

Trevisan C, Veronese N, Maggi S, et al. Factors influencing transitions between frailty states in elderly adults: The Progetto Veneto Anziani Longitudinal Study. J Am Geriatr Soc. 2017;65(1):179-184.

Dent E, Lien C, Lim WS, et al. The Asia-Pacific clinical practice guidelines for the management of frailty [published correction appears in J Am Med Dir Assoc . 2018;19(1):94]. J Am Med Dir Assoc. 2017;18(7):564-575.

Dent E, Morley JE, Cruz-Jentoft AJ, et al. Physical frailty: ICFSR international clinical practice guidelines for identification and management. J Nutr Health Aging. 2019;23(9):771-787.

Moorhouse P, Mallery LH. Palliative and therapeutic harmonization: a model for appropriate decision-making in frail older adults. J Am Geriatr Soc. 2012;60(12):2326-2332.

Palliative and Therapeutic Harmonization. Implementation: frailty within the health system. Accessed June 4, 2020. https://pathclinic.ca/implementation/

Liu X, Hui-Min Ng D, Wei-Ting Seah J, et al. Update on interventions to prevent or reduce frailty in community-dwelling older adults: a scoping review and community translation. Curr Geriatr Rep. 2019;8:72-86.

Puts MTE, Toubasi S, Andrew MK, et al. Interventions to prevent or reduce the level of frailty in community-dwelling older adults: a scoping review of the literature and international policies. Age Ageing. 2017;46(3):383-392.

Hsieh TJ, Su SC, Chen CW, et al. Individualized home-based exercise and nutrition interventions improve frailty in older adults: a randomized controlled trial [published correction appears in Int J Behav Nutr Phys Act . 2019;16(1):136]. Int J Behav Nutr Phys Act. 2019;16(1):119.

Arrieta H, Rezola-Pardo C, Gil SM, et al. Effects of multicomponent exercise on frailty in long-term nursing homes: a randomized controlled trial. J Am Geriatr Soc. 2019;67(6):1145-1151.

Ng TP, Feng L, Nyunt MSZ, et al. Nutritional, physical, cognitive, and combination interventions and frailty reversal among older adults: a randomized controlled trial. Am J Med. 2015;128(11):1225-1236.e1.

Chodzko-Zajko WJ, Proctor DN, Fiatarone Singh MA, et al.; American College of Sports Medicine. American College of Sports Medicine position stand. Exercise and physical activity for older adults. Med Sci Sports Exerc. 2009;41(7):1510-1530.

Lee PG, Jackson EA, Richardson CR. Exercise prescriptions in older adults. Am Fam Physician. 2017;95(7):425-432. Accessed June 5, 2020. https://www.aafp.org/afp/2017/0401/p425.html

Huguet LG, González MN, Kostov B, et al. Pre frail 80: multifactorial intervention to prevent progression of pre-frailty to frailty in the elderly. J Nutr Health Aging. 2018;22(10):1266-1274.

Serra-Prat M, Sist X, Domenich R, et al. Effectiveness of an intervention to prevent frailty in pre-frail community-dwelling older people consulting in primary care: a randomised controlled trial. Age Ageing. 2017;46(3):401-407.

Luger E, Dorner TE, Haider S, et al. Effects of a home-based and volunteer-administered physical training, nutritional, and social support program on malnutrition and frailty in older persons: a randomized controlled trial. J Am Med Dir Assoc. 2016;17(7):671e9-671e16.

Ruikes FGH, Zuidema SU, Akkermans RP, et al. Multicomponent program to reduce functional decline in frail elderly people: a cluster controlled trial. J Am Board Fam Med. 2016;29(2):209-217.

Continue Reading

More in AFP

More in pubmed.

Copyright © 2021 by the American Academy of Family Physicians.

This content is owned by the AAFP. A person viewing it online may make one printout of the material and may use that printout only for his or her personal, non-commercial reference. This material may not otherwise be downloaded, copied, printed, stored, transmitted or reproduced in any medium, whether now known or later invented, except as authorized in writing by the AAFP.  See permissions  for copyright questions and/or permission requests.

Copyright © 2024 American Academy of Family Physicians. All Rights Reserved.

Clinical Pearls & Morning Reports

Frailty in Older Adults

Posted by Carla Rothaus, MD

Published August 7, 2024

What are some specific measures that aid in the management of frailty in older adults?

Frailty is a clinically identifiable state of diminished physiological reserve and increased vulnerability to a broad range of adverse health outcomes. Read the NEJM Review Article here .

Clinical Pearls

Q: Is screening for frailty in older adults effective?

A: The benefit of routine frailty screening has been shown in high-risk clinical contexts (e.g., oncology and surgery); its benefit in primary care remains to be established.

Q: How strong is the evidence that interventions for frailty in older adults are effective?

A: The current evidence with respect to frailty screening and interventions is limited. Most of the clinical trials that have evaluated frailty interventions have been small, with heterogeneous trial populations and nonuniform screening tools, interventions, and outcome measures, all of which have contributed to low-quality evidence. Despite these limitations, certain interventions have been shown to ameliorate frailty and associated outcomes.

Morning Report Questions

Q: name some of the principles underlying the management of frailty in older adults..

A: The aim of management is to preserve physiological reserve and prevent stressors in order to maximize functioning and quality of life, guided by the patient’s goals and degree of frailty. Frailty makes older people more vulnerable to the risks associated with treatment. An important part of management is making routine care less hazardous for patients with frailty. The presence of frailty should not be used as a convenient reason to withhold potentially effective treatments but rather as an opportunity to facilitate patient-centered care. Aligning treatment with the patient’s health priorities may reduce the burden of treatment and unwanted care. Although it is necessary to minimize polypharmacy and avoid potentially inappropriate medications for patients who are frail, some treatments (e.g., exercise) may be of great benefit to such patients.

Q: What are some specific measures that aid in the management of frailty in older adults?

A: Assessment of an older person’s degree of frailty on a spectrum from fit to severely frail can provide a framework for applying evidence and principles of geriatric care. If frailty is suspected, a careful medical evaluation or comprehensive geriatric assessment should be performed to identify precipitants and exacerbating factors and to determine targets for interventions. Potentially high-yield clinical targets are depression, anemia, hypotension, hypothyroidism, vitamin B12 deficiency, unstable medical conditions, and adverse drug events. Personalized, adaptive coping strategies, such as keeping daily routines in familiar surroundings, maintaining social connections, and mobilizing resources, can help patients perform self-care and uphold social roles, despite the limitations imposed by frailty. Incorporating frailty into a prognostic model improves the estimation of life expectancy, which in turn helps guide decisions about cancer screening.

Browse more Clinical Pearls & Morning Reports »

• DOI: 10.14283/jfa.2017.11
• Corpus ID: 3912426

Components and Indicators of Frailty Measures: A Literature Review.

• B. Xie , J. Larson , +3 authors Cynthia Arslanian-Engoren
• Published in The Journal of frailty… 2017

Figures from this paper

24 Citations

Definition and assessment of psychological frailty in older adults: a scoping review.

• Highly Influenced

Frailty in Young and Middle-Aged Adults: An Integrative Review

Defining and assessing psychological frailty in older adults: a scoping review protocol, exploring the efficiency of the tilburg frailty indicator: a review, multidimensional instruments with an integral approach to identify frailty in community-dwelling people: protocol for a systematic psychometric review, different cognitive frailty models and health- and cognitive-related outcomes in older age: from epidemiology to prevention, frailty in community-dwelling older people: comparing screening instruments, measuring the effects of a nurse-led intervention on frailty status of older people living in the community in ethiopia: a protocol for a quasi-experimental study, comparison of frailty in oldest-old people using the clinical-functional vulnerability index-20 (ivcf-20) and edmonton frail scale (efs), 71 references, the identification of frailty: a systematic literature review, a new operational definition of frailty: the frailty trait scale..

• Highly Influential
• 10 Excerpts

Development of a Frailty Measure for Older Adults: The Frailty Index for Elders

A reliable measure of frailty for a community dwelling older population, development of an easy prognostic score for frailty outcomes in the aged., conceptualizations of frailty in relation to older adults., a global clinical measure of fitness and frailty in elderly people, quality of life in sarcopenia and frailty, validation of the comprehensive frailty assessment instrument against the tilburg frailty indicator, the i.a.n.a. task force on frailty assessment of older people in clinical practice, related papers.

Showing 1 through 3 of 0 Related Papers

Published in The Journal of frailty & aging 2017

B. Xie J. Larson Richard Gonzalez S. Pressler Cindy Lustig Cynthia Arslanian-Engoren

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List
• v.58(2); 2019 Mar

Language: English | Slovene

Frailty Assessment Scales for the Elderly and their Application in Primary Care: A Systematic Literature Review

Ocenjevalne lestvice krhkosti starostnika in njihova raba na primarni ravni: sistematični pregled literature.

1 University of Ljubljana, Faculty of Health Sciences, Zdravstvena pot 5, 1000, Ljubljana, Slovenia

Danica Rotar-Pavlič

2 University of Ljubljana, Faculty of Medicine, Department of Family Medicine, Poljanski nasip 58, 1000, Ljubljana, Slovenia

The increase in the elderly population is causing changes and challenges that demand a comprehensive public health response. A specific characteristic of the elderly is their frailty. Today’s problems with identifying levels of frailty are being resolved by numerous tools in the form of frailty assessment scales. This systematic review establishes which frailty assessment scales for the elderly are being used and what their applicability in primary care is like in Slovenia and around the world.

Documents published after 2010 were searched for in the PubMed database using keywords and other specific criteria.

A total of 177 search hits were obtained based on various search strings. The final analysis included 28 articles, of which three were systematic literature reviews. These three covered quantitative studies, mainly consisting of observational cross-sectional surveys or cohort studies. Three other studies featured non-systematic literature reviews. Quantitative studies (mainly cross-sectional surveys or cohort studies) prevailed among the remaining 22 articles. One study had a qualitative design (Delphi method). The main outcome measures observed by all studies were frailty assessment scales for the elderly, the majority of which were evaluated on a sample of the elderly.

Conclusions

None of the assessment scales examined are used as the gold standard for primary care. A variety of tools are being used in clinical practice to assess frailty in elderly patients, highlighting the need for standardization and guidelines. This requires evaluating the current assessment scales in terms of validity and reliability, and suitably improving them.

Izvleček

Povečan delež starejšega prebivalstva povzroča spremembe in prinaša izzive, kar zahteva celovit odziv na področju javnega zdravja. Specifičnost starostnikov je tudi njihova krhkost. Ta za posameznika pomeni večje tveganje za negativne rezultate, povezane z zdravjem. Ugotavljanje krhkosti daje teoretični okvir, v katerem lahko zdravnik primarnega zdravstvenega varstva oblikuje celovit pristop ocenjevanja in zdravljenja starejšega bolnika s kompleksno multimorbidnostjo na preprost in strukturiran način. Težave določanja stopnje krhkosti danes rešujejo številna orodja v obliki ocenjevalnih lestvic krhkosti. Slovenija se je v letu 2017 pridružila Evropski komisiji pri Skupnem evropskem ukrepanju za preprečevanje starostne krhkosti in oslabljenosti »Joint Action«. Eden izmed predlogov ukrepov in aktivnosti je tudi razviti, implementirati in spremljati sistem presejanja na krhkosti po posameznih področjih. Sicer z merjenjem krhkosti lahko pridobimo uporabne podatke, a je za oblikovanje informacij pomemben izbor ustreznega, veljavnega instrumenta. Pojavlja se vprašanje o količini in kakovosti uporabe ocenjevalnih lestvic krhkosti starostnikov. Namen sistematičnega pregleda literature je ugotoviti, katere ocenjevalne lestvice merjenja krhkosti starostnika se uporabljajo in kakšna je domnevna uporabnost na primarni ravni v svetu in v Sloveniji .

Sistematično je bila pregledana literatura, objavljena po letu 2010, o ocenjevalnih lestvicah krhkosti starostnika. Iskanje dokumentov je potekalo v bibliografski bazi PubMed po določenih kriterijih s ključnimi besedami: frailty, elderly, evaluation scale, primary, frailty scale, frailty screening in primary care .

Vseh zadetkov glede na različne iskalne nize je bilo 177. V končno analizo se je uvrstilo 28 člankov, od tega trije sistematični pregledi literature. Ti vključujejo kvantitativne raziskave, v večini opazovalne presečno pregledne ali kohortne študije. Tri raziskave so nesistematični pregledi literature. Med 22 drugimi raziskavami prevladujejo raziskave s kvantitativnimi zasnovami, v večini so presečno pregledne ali kohortne študije. Ena študija ima kvantitativno zasnovo, zbiranje podatkov pa je potekalo z delfsko metodo. Opazovani izidi vseh študij so ocenjevalne lestvice starostnikov. V večini so jih raziskovalci vrednotili na vzorcu starostnikov .

Zaključki

Zaradi starajočega se prebivalstva je potreba po ureditvi področja merjenja krhkosti starostnikov s pomočjo ocenjevalnih lestvic vse večja. Za ugotavljanje krhkosti starejših se v praksi uporablja toliko orodij, da je potreba po standardizaciji in smernicah velika. Nobena izmed ocenjevalnih lestvic nima vloge zlatega standarda uporabe za primarno raven. Pred implementacijo v slovenski prostor je potrebno obstoječe ocenjevalne lestvice vrednotiti po kriterijih veljavnosti in zanesljivosti ter jih primerno izboljšati .

1. Introduction

The population’s age structure has been changing greatly over the past decades, with the population becoming increasingly older, including in Slovenia ( 1 , 2 ). This causes many changes and challenges that demand a comprehensive public health response ( 3 , 4 ).

A specific characteristic of the elderly is their frailty. It is defined as “a condition or syndrome which results from a multi-system reduction in reserve capacity to the extent that a number of physiological systems are close to, or past, the threshold of symptomatic clinical failure.” As a consequence, the frail person is at increased risk of disability and death from minor external stresses ( 5 ). Identifying the level of frailty is a useful clinical concept for predicting and preventing frailty ( 6 , 7 , 8 ). Frailty in the elderly entails a changed perspective on age by replacing the outdated term “chronological age” with the more accurate and personalized parameter of “biological age,” and it can be measured in individuals ( 9 ). Problems with identifying the level of frailty, which were common in the past ( 5 ), are now being solved by numerous tools that can also be applied to the elderly ( 10 , 11 ).

Frailty assessment thus provides a theoretical framework that primary care physicians can use to develop a comprehensive approach to assessing and treating elderly patients with complex multimorbidity in a simple and structured way ( 7 ). In Slovenia, an important role in this regard is also played by family doctors and their teams ( 12 ). The importance of using frailty measurement tools is supported by the global lack of key information and evidence on the health of the elderly, which hinders the development and evaluation of suitable policies and programs for them ( 13 ). Frailty measurements can generally provide useful information, but that requires selecting an appropriate valid instrument ( 9 ). In agreement with the Ministry of Health, in 2017, Slovenia joined the EU Commission’s Joint Action on the Prevention of Frailty. The main outcome of Joint Action will be a common European model to approach frailty, leading to the development of improved strategies for diagnosis care and education for frailty, disability and multi-morbidity. The Joint Action outcomes are expected to contribute to the prevention of the growing burden of disability and chronic diseases and to a more effective response to older people’s needs of care delivery, a central priority for the EU and its MS. One of the measures and activities proposed was to develop, implement, and monitor a frailty screening system by individual area ( 14 ).

The question is how many frailty assessment scales are available and what their quality is like. In Slovenia, there is a need for the knowledge of frailty assessment scales for the application at the primary level. They established the subject of Geriatrics and subject Elderly, dying patient, palliative at the Faculty of Medicine at the University of Ljubljana. In Slovenia, payment models for multimorbidity and elderly are also changing. This literature review identifies research on frailty assessment scales for the elderly published after 2010. Its goals were to determine which frailty assessment scales are available, what they measure, and whether they are used in primary care. The fundamental research question is whether the knowledge on frailty assessment scales provides a selection of assessment scales that could be applied to primary care in Slovenia in order to assess the frailty of the elderly.

Literature on frailty assessment scales for the elderly was systematically reviewed. The data was collected in February 2018.

2.1. Document Sources

Documents were searched for in the online bibliographical database PubMed ( 15 ).

2.2. Document Identification Methods

Documents were searched for using the following keywords: frailty, elderly, evaluation scale, primary, frailty scale, frailty screening, and primary care. Searches were performed using Boolean operators for PubMed: (((frailty) AND elderly) AND evaluation scale); (((frailty) AND elderly) AND rating scale); (((frailty) AND elderly) AND measuring); ((frailty) AND screening) AND primary care). The search was limited to full-text open-access English articles published after 2010.

2.3. Methods of Selecting Documents to be Included in the Analysis

The selection in PubMed was narrowed down to full-text research articles. The keywords selected had to be included in the article’s title or abstract, the articles had to refer to the elderly, and they had to be written in English and published in the past 8 years. An article was deemed appropriate if it featured a study connected with the frailty assessment scales used for the elderly. Studies containing clinical frailty scales or scales used for populations other than the elderly and clinical frailty scales were not included. After selecting the relevant articles, an open discussion took place in a heterogeneous group of experts with diplomas from the Faculty of Medicine and Faculty of Health Sciences at the University of Ljubljana and head lecturer of subject Determinants of health and disease on Interdisciplinary doctoral programme in Biomedicine, field Public Health. Another discussion took place in a group of students specialized in Family Medicine from Faculty of Medicine at the subject Elderly, dying patient, palliative. Their suggestions and comments found a place in the final selection of articles and frailty assessment scales for eventual application in primary care.

2.4. Selection of Relevant Data for the Systematic Review

The data collected included year, country, research design, units observed, number of participants, and main conclusions.

2.5. Methods for Assessing Study Quality

The suitability of the studies included was evaluated in terms of their agreement with the search string.

Twenty-eight articles meeting the criteria set were selected for final analysis ( Table 1 ).

Main characteristics and results on frailty assessment scales for the elderly.

3.1. Selecting Documents for Systematic Review

The procedure of selecting documents for inclusion in the systematic review of literature on frailty assessment scales for the elderly and their application in primary care.

Twenty-two studies from various countries, published after 2010, are dominated by quantitative, mostly cross-sectional or cohort studies. One study ( 22 ) has a qualitative design and data for it was collected using the Delphi method. The number of subjects included in the study depends on the research design, ranging from 100 to 5,000 in the majority of the studies; the age criteria used vary. Four studies include geriatric specialists: GPs, specialist physicians, and so on ( 22 , 23 , 24 , 25 ). Four studies ( 26 , 27 , 28 , 29 ) are based on databases that already exist. The main outcome measures observed by all studies are frailty assessment scales, indexes, or indicators analysed from various perspectives ( Table 2 ).

Frailty assessment scales that were identified for eventual application in primary care.

3.2. Main Characteristics of the Research Studies Reviewed

This analysis includes three systematic literature reviews that together cover more than 70 quantitative studies, consisting largely of observational cross-sectional surveys or cohort studies. Three studies included in the final analysis are non-systematic literature reviews ( Table 1 ).

4. Discussion

4.1. systematic review results.

Considering that frailty is a common feature of the elderly, it is also important to obtain information on this area. Veninšek and Gabrovec ( 45 ) identified four main areas essential for the clinical management of frailty: definition of frailty, epidemiology of frailty, tools for screening and diagnosis frailty and successful interventions for decreasing frailty. The priority objective of the WHO Global Strategy and Action Plan on Aging and Health ( 13 ) to fill information gaps at the global level is thus well grounded. This is also confirmed by the results of this systematic review. The international survey conducted by Bruyère et al. ( 25 ), which included 44 countries, shows that frailty assessment is becoming a routine daily practice in treating elderly patients. According to this study, 205 (52.8%) clinicians, of whom the majority are geriatric specialists, always assess frailty in their daily practice and 38.1% report measuring it sometimes ( 25 ). All international consensus groups recommended all persons older than 70 years should be screened for frailty ( 22 ).

Factors, such as age and malnutrition, increase the risk of frailty ( 36 ), but individual deviations may be great, and the level of frailty may vary. Physical frailty in the elderly is a complex condition and the musculoskeletal aging phenotype comprises four key elements: osteoporosis, osteoarthritis, sarcopenia, and frailty ( 21 ). On the other hand, measuring the grades of frailty in the elderly can assist in assessment, management, and decision-making for osteoporosis and osteoporotic fractures ( 19 ). Fried et al. ( 44 ) proposed five frailty criteria: weakness, slow walking speed, low physical activity, self-reported exhaustion, and unintentional weight loss. The majority of physicians (64.9%) generally measure and diagnose frailty using more than one instrument ( 25 ). The most widely used tool is the gait speed test, which is performed by 43.8% of physicians ( 25 ) and is a simple yet efficient indicator for diagnosing frailty in primary care ( 17 ). This is followed by the clinical frailty scale (34.3%), the SPPB test (30.2%), the frailty phenotype test (26.8%), and the frailty index (16.8%) ( 25 ). Examples of some commonly used and validated frailty tools include the FRAIL, the Cardiovascular Health Study Frailty Screening Measure, the Clinical Frailty Scale, and the Gérontopôle Frailty Screening Tool ( 22 ). The Phenotype of Frailty is the most evaluated and frequently-used measure ( 16 ). The results of ADVANTAGE JA research ( 46 ) showed that there are multiple measurements used to screen and diagnose frailty. They have considered the most relevant, the recommended tools of frailty would be: Clinical Frailty Scale, Edmonton Frailty Scale, FRAIL Index, frailty phenotype, Inter-Frail, Prisma-7, Sherbrooke Postal Questionnaire, Short Physical performance Battery (SPPB), Study of Osteoporotic Fractures Index (SOF) and gait speed.

Other researchers ( 16 , 20 , 43 ) report a great variety of frailty scales, but their reliability and validity have rarely been examined ( 16 ). Bouillon et al. ( 16 ) highlight that only a few studies have evaluated frailty scales in terms of reliability and validity or following specific standards. An acceptable reliability coefficient and predictive validity has been confirmed for the CSHA Clinical Frailty Scale and the Edmonton Frail Scale. The frailty index and the Fried scale have been tested for validity, but not reliability ( 16 ). Specific anomalies (terminological and professional anomalies or plagiarism) occur with many assessment scales ( 16 ).

The majority of studies positively conclude that the scales examined are efficient for identifying the level of frailty ( 18 , 26 , 27 , 28 , 31 , 32 , 33 , 34 , 37 , 42 ). Other studies determine that different instruments result in different estimates of frailty and that the gains in the tests’ predictive accuracy are often modest ( 29 , 30 ). The level of frailty assessed by geriatricians and GPs may differ ( 23 , 24 ). Among other things, frailty can also be related to self-perceived health ( 38 ).

Bruyère et al. ( 25 ) report that a variety of tools are being used, highlighting the need for standardization and guidelines. None of the assessment scales are used as the gold standard in primary care ( 18 , 27 , 34 , 42 , 43 ). Widely used scales – a good example of which is the frailty scales developed by Fried et al. ( 44 ) – must be based on strict criteria. In addition, improvements and consensus of everyone involved in the healthcare for the elderly are required ( 16 ).

4.2. Research Limitations and Strengths

Conclusions can be drawn regarding the possible application of existing scales in Slovenia. It would make sense to expand the literature review by including search strings that also identify psychological frailty (e.g. “mental” frailty scales). This is the first review of literature which investigates frailty scales for use at primary level and in terms of reliability and validity.

4.3. Relevance of the Systematic Review Results for the Discipline

This systematic review provides insight into which frailty assessments scales are used for the elderly, who assesses frailty of the elderly, and the importance of primary care in assessing elderly people’s frailty.

4.4. Potential for Further Research

There is a need for more research that assesses the validity, reliability, user-friendliness, comparability, etc., of different frailty scales.

5. Conclusion

Due to population ageing, there is an increasingly greater need for standardizing the measurement of geriatric frailty using frailty assessment scales. According to the situation (resource constraints) we estimate that the most appropriate scales for primary care in Slovenia are Frailty phenotype ( 44 ), Short Physical performance Battery (SPPB) ( 25 ) and Edmonton frail scale ( 25 ). Implementing such scales in Slovenia requires further research and discussions by leading specialists in this area on extended professional college of doctors of family medicine. Also, nurses from modal practices should be included. Consensus between various healthcare levels should be reached.

Acknowledgements

Special thanks to Prof. Dr. Lijana Zaletel Kragelj, dr. med for mentoring and group of doctoral students of the Interdisciplinary Doctoral Programme in Biomedicine, Field Public Health for their suggestions and comments in the process of the final selection of articles.

Conflict of interest The authors declare that no conflicts of interest exist.

Funding This work was funded by the University of Ljubljana’s Faculty of Medicine.

Ethical Approval The method used in this systematic review involves no ethical issues and therefore no ethical approval was necessary.

• Open access
• Published: 03 August 2024

Association of the dietary inflammatory index with sarcopenic obesity and frailty in older adults

• Sukyoung Jung   ORCID: orcid.org/0000-0001-9200-2158 1 ,
• Yunhwan Lee   ORCID: orcid.org/0000-0001-8484-4750 2 ,
• Kirang Kim   ORCID: orcid.org/0000-0003-3054-8758 3   na1 &
• Sohyun Park   ORCID: orcid.org/0000-0001-6009-1002 4 , 5   na1  

BMC Geriatrics volume  24 , Article number:  654 ( 2024 ) Cite this article

270 Accesses

2 Altmetric

Metrics details

This study examined whether a higher dietary inflammatory index (DII ® ) is associated with the risk of sarcopenic obesity (SO) and frailty among Korean older adults.

A total of 950 participants aged 70–84 years, who completed the baseline nutrition survey of the Korean Frailty and Aging Cohort Study, were included in the analysis. The DII, quantifying the dietary inflammatory potential, was calculated using 23 foods and nutrients as assessed by a 24-h dietary recall. SO was defined as the coexistence of sarcopenia (dual-energy X-ray absorptiometry-measured appendicular skeletal muscle mass index of < 7.0 for males; <5.4 for females) and abdominal obesity (waist circumference of ≥ 90 cm for males; ≥85 cm for females). Frailty status was assessed using the Fried frailty index (range, 0–5), a simple tool for defining frailty that consists of three or more of five frailty items. Multinomial logistic regression models were used to estimate odds ratios (ORs) with 95% confidence intervals (CIs), adjusting for confounders.

The prevalence of SO and frailty was 9.8% and 10.8%, respectively. The DII was significantly higher in the frail group (2.7) compared to the robust and SO groups (2.0 vs. 1.8) ( P  < 0.001). Among nutrients and foods included in the DII, the frail group exhibited lower vitamin E, niacin, vitamin B 6 , energy, and protein intakes than the robust and SO groups. Multivariable-adjusted OR (95% CI) for frailty versus robust (comparing DII tertile 3 to tertile 1) was 2.3 (1.1–4.8; P -trend = 0.02). However, no significant association was observed between the DII and SO (OR, 1.1; 95% CI, 0.5–2.1; P -trend = 0.6).

Conclusions

A higher DII score was associated with increased odds of frailty but not with SO in Korean older adults, suggesting that proinflammatory diets have a greater impact on frailty than that on SO in the older population.

Peer Review reports

Introduction

The proportion of the population aged 65 years and older has increased over the past six decades, both on a global scale (from 5.5% in 1960 to 9.6% in 2021) and in South Korea (from 3.3% in 1960 to 16.7% in 2021) [ 1 ]. Notably, South Korea has experienced a remarkably rapid transition to an aged society [ 2 ]. Aging is known to result in impairments across various musculoskeletal systems, including the joints, bones, muscles, and multiple body areas or systems [ 3 ]. Additionally, aging is associated with an elevation in total body fat mass and increase in visceral fat depots may occur redistributing fat into the abdominal region [ 4 ]. These age-related changes in body composition may be linked to reduced mobility, decreased functional capacity [ 3 ], and an increased risk of cardiovascular diseases, diabetes, and metabolic syndrome [ 4 ].

To assess age-related changes in body composition, several phenotypes have been identified. Sarcopenic obesity (SO) is characterized by the coexistence of sarcopenia, which refers to the “age-related loss of skeletal muscle mass plus loss of muscle strength and/or reduced physical performance” [ 5 ], along with obesity [ 6 ]. Frailty has been defined as “a state of vulnerability to poor homeostasis resolution following a stressor event and is a consequence of a cumulative decline in several physiological systems during a lifetime [ 7 ]”. Frailty can be considered an umbrella term that encompasses sarcopenia. Globally, frailty affects 12–24% of older adults [ 8 ], whereas in South Korea, its prevalence stands at 8% [ 9 ]. The global SO prevalence is approximately 11% [ 10 ], whereas in South Korea, it is approximately 4% [ 11 ]. Although these conditions are distinct, low-grade inflammation, as indicated by elevated inflammatory markers including interleukin-6, high-sensitivity C-reactive protein (hs-CRP), and tumor necrosis factor alpha, is believed to be a common biological mechanism underlying both phenotypes [ 7 , 12 ].

Diet plays a significant role in modulating low-grade inflammation [ 13 ], and the dietary inflammatory index (DII) has been widely used for assessing the inflammatory potential of diets [ 14 ]. Extensive evidence supports a positive association between the DII and several health outcomes, including metabolic risk markers, cancer risk, cardiovascular diseases, and mortality [ 15 ], as well as phenotypes including sarcopenia [ 16 ], abdominal obesity [ 17 ], and frailty [ 18 ]. However, owing to the scarcity of studies conducted in South Korea, generalizing these findings to the Korean population is challenging. Although two studies have been conducted in the Korean context, one is limited to postmenopausal women [ 19 ], and the other has a relatively small sample size [ 20 ]. Furthermore, to date, no study has directly investigated the association between the DII and SO. Frailty and SO are common conditions among older adults, and previous studies indicated that several characteristics, including metabolic, inflammatory, and hematologic markers, are shared between the two conditions [ 21 , 22 , 23 ]. It would be more meaningful to understand these two conditions in relation to the DII rather than examining each condition separately.

Considering these gaps, the present study aimed to examine the association of the dietary inflammatory potential, as measured by the DII, with SO and frailty, among Korean older adults. We hypothesized that Korean older adults with higher DII scores would exhibit increased odds of having both SO and frailty compared with those with lower DII scores.

Study Population

The Korean Frailty and Aging Cohort Study (KFACS) is a population-based prospective cohort study that aims to investigate frailty status and changes in frailty states over time among community-dwelling older adults in Korea [ 24 ]. From May to November 2016, a total of 1,559 community dwellers aged 70–84 years were recruited using quota sampling stratified by age and sex from 10 study centers located in different regions [ 24 ].

Among 1,002 (64%) participants who completed nutrition surveys, we excluded those with the following conditions: total energy intake of < 400 kcal ( n  = 3), missing data on dual-energy X-ray absorptiometry (DEXA) ( n  = 5), and missing frailty component data ( n  = 44). Due to the limited sample size, we did not exclude participants with missing covariates ( n  = 17). The final analytical sample included 950 participants (459 males and 491 females) (Supplemental Fig.  1 ).

Data Collection

Face-to-face interviews were conducted to gather information on demographics (e.g., age, sex, education level, monthly household income, and family structure [living alone or not]), health status (e.g., comorbidity and number of prescription drugs), and health behaviors (e.g., chewing status, smoking status, alcohol consumption, and physical activity level). Trained staff performed measurements of anthropometrics (e.g., height, weight, and waist circumference), body composition (e.g., muscle mass), and physical function (e.g., hand grip strength, chair-stand time, and walking speed). Body mass index (BMI) was calculated as the ratio of weight (kg) to height squared (m 2 ). Waist circumference was measured at the midpoint between the lowest rib margin and the upper ridge of the iliac crest using an inelastic tape. Muscle mass (kg) was measured using DEXA (GE Healthcare Lunar, Madison, WI, USA; and Hologic DXA Systems, Hologic Inc., Bedford, MA, USA). Grip strength (kg) was measured using a hand grip dynamometer (T.K.K.5401; Takei Scientific Instruments Co., Tokyo, Japan). The handgrip strength was measured twice for each hand over a 3-min interval. We used the highest value among the averages of each measurement for further analysis. Walking speed over a 4-m distance was assessed using an automatic timer (Gait Speedometer; Dynamic Physiology, Daejeon, Korea). Blood samples were collected following an 8-h fasting and transported to a commercial laboratory for analysis. This study utilized serum total cholesterol (mg/dL), triglycerides (mg/dL), high-density lipoprotein (HDL) cholesterol (mg/dL), low-density lipoprotein (LDL) cholesterol (mg/dL), fasting blood glucose (mg/dL), hemoglobin A1C (HbA1C) (%), and hs-CRP (mg/dL) [ 24 ].

To collect detailed dietary information, trained interviewers administered a 24-h dietary recall in the participant’s home. Participants reported the description, quantity, and time and place of consumption for all foods and beverages consumed within the previous 24 h, with the assistance of visual aids developed by the Korea Disease Control and Prevention Agency (KDCA) [ 25 ]. Nutrient intakes were estimated using the 24-h dietary recall assessment system of the National Institute of Health and the KDCA [ 25 ].

As the main exposure, the DII was calculated to comprehensively assess the inflammatory potential of diets [ 14 ]. This composite index was developed and validated using a comprehensive review of 1,943 articles published from 1950 to 2010. Briefly, the inflammatory potential of 45 food, nutrient, and bioactive compound parameters was scored on the basis of their effects on inflammatory biomarkers (interleukin-1β, interleukin-4, interleukin-6, interleukin-10, tumor necrosis factor-α, and hs-CRP). The range for overall DII score in the DII development study was − 8.87 to + 7.98 [ 14 ]. The DII calculation in this study followed the same approach as that described in the original DII development study [ 14 ]. Originally, 45 parameters were included in the DII calculation, but the following 23 foods and nutrients , which were only available to use , were included in this study : beta-carotene, carbohydrate, cholesterol, energy, fiber, folic acid, garlic, ginger, green/black tea, vitamin A, vitamin B 1 , vitamin B 2 , vitamin B 6 , vitamin C, vitamin D, vitamin E, iron, niacin, vitamin B 12 , onion, pepper, protein, and total fat. Owing to insufficient data , the following 22 foods and nutrients were not included in the calculation : caffeine, alcohol, eugenol, magnesium, MUFA, n-3 fatty acid, n-6 fatty acid, PUFA, saffron, saturated fat, selenium, trans fat, turmeric, zinc, flavan-3-ol, flavones, flavonols, flavonones, anthocyanidins, isoflavones, thyme/oregano, and rosemary.

First, Z-scores were computed for each of the 23 parameters by subtracting the standard global mean (derived from the representative global diet database) from the actual consumption and subsequently dividing by the global standard deviation. Second, the estimated Z-scores were converted into percentiles to minimize the effect of skewness or outliers. These percentiles were centered on 0 (yielding a symmetrical distribution) by doubling each percentile value and subtracting 1. Lastly, parameter-specific DII scores were determined by multiplying the centered percentile values by the corresponding overall food parameter-specific inflammatory effect score, and the overall DII score was obtained by summing across all parameter-specific DII scores. A higher DII score indicates a more proinflammatory diet, while a lower score suggests a more anti-inflammatory diet. In this study, DII scores ranged from − 3.07 to 4.39.

Assessment of outcomes

We used the appendicular skeletal muscle mass index (ASMI) for sarcopenia diagnosis. The appendicular skeletal muscle mass (ASM) was defined as the sum of lean muscle mass in both the arms and legs, and the ASMI was obtained by dividing the ASM by the square of the height (kg/m 2 ). Sarcopenia was defined as an ASMI of < 7.0 and < 5.4 for males and females, respectively, which is a diagnostic criterion for sarcopenia proposed by the Asian Working Group for Sarcopenia (AWGS) [ 5 ]. Abdominal obesity was defined as a waist circumference of ≥ 90 and ≥ 85 cm for males and females, respectively, according to the criteria set by the Korean Society for the Study of Obesity [ 26 ]. Finally, SO was defined as the presence of both sarcopenia and abdominal obesity.

To assess frailty status, a modified version of the Fried frailty index was used [ 27 ]. The Fried frailty index included the following five components: unintended weight loss, weakness, self-assessed exhaustion, slow walking speed, and low physical activity. In the modified Fried frailty index, the physical activity component was assessed using the Korean version questionnaire to better represent the physical activity levels of the Korean population, and the remaining components were assessed using the same criteria as the original Fried frailty index. (1) Unintended weight loss component was defined as an affirmative answer to “Have you experienced unintended weight loss of 4.5 kg or more during the last year?” (2) Weakness component was defined as a grip strength of < 26 and < 18 kg for males and females, respectively [ 28 ]. (3) The exhaustion component was assessed on the basis of responses to questions from the Center for Epidemiological Studies-Depression scale: “I felt that everything I did was an effort” or “I could not get going.” Exhaustion was defined as an affirmative answer to the above mentioned questions for three or more days in a week. (4) Slow walking speed component was defined as < 1 m/s after walking 4 m at a normal rhythm. (5) The physical activity component was evaluated using the International Physical Activity Questionnaire–Short Form (Korean version) [ 29 ]. The metabolic equivalent of task (MET)-minutes was quantified by multiplying the frequency, duration, and intensity of physical activities engaged during a week. “Low physical activity” was defined as < 494.65 and < 283.50 MET-min/week for males and females, respectively (Supplemental Table 1 ).

Each frailty index component received a score of 1 if the criteria were met; otherwise, it received a score of 0. The final modified frailty index score was calculated by summing the scores of each component (range, 0–5), with a score of 3–5 defining frailty.

Assessment of covariates

Covariates included age (years), sex (male or female), education level (< 7 or ≥ 7 years of education), monthly household income (unknown, < 1, 1–2, or ≥ 2 million Korean won), family structure (living alone or living with a partner), number of chronic diseases (counts), number of prescribed drugs (< 4 or ≥ 4), chewing status (uncomfortable or comfortable), smoking status (everyday, sometimes, or none), alcohol consumption (g/day), physical activity level (MET-min/week), and total energy intake.

Statistical analysis

The normality of all continuous variables was evaluated both visually using histograms and Q–Q plots and using skewness and kurtosis values. Variables that did not follow a normal distribution were log-transformed for the statistical test (physical activity level, repeated five-chair stands, triglyceride, fasting blood glucose, HbA1C, and hs-CRP). The characteristics of study participants by SO and frailty status were described using means and standard deviations for continuous variables and frequencies and proportions for categorical variables. The significance of differences in characteristics between the robust (neither have SO nor frailty), SO, and frailty groups was examined using analysis of variance and a chi-square test for continuous and categorical variables, respectively. Age- and sex-adjusted anthropometric and metabolic characteristics between the robust, SO, and frailty groups were presented as means and standard errors using the general linear model. Except for energy and cholesterol intake, all nutrient intakes used in the DII calculation were expressed as percentages of the age- and sex-specific recommendations based on the Dietary Reference Intakes for Koreans (KDRI) 2020 [ 30 ]. Total energy intake was expressed as a percentage of the estimated energy requirement, calculated using the following formula {male: 662 − 9.53 × age + value of physical activity level (PA) [15.91 × weight (kg) + 539.6 × height (m)]; female: 354 − 6.91 × age + PA [9.36 × weight + 726 × height (m)]} [ 30 ]. PA values of 1.11 and 1.12 for males and females, respectively, were assigned [ 30 ]. The dietary characteristics between the robust, SO, and frailty groups were tested using the general linear model after adjusting for age, sex, and total energy intake. To determine group differences, we performed Tukey’s multiple comparison test.

For categorical analysis, the DII was classified into tertiles, with tertile 1 serving as the reference (range: -3.07–1.46 for tertile 1; 1.46–2.83 for tertile 2; 2.84–4.39 for tertile 3). Multinomial logistic regression models were used to estimate the odds ratios (ORs) and 95% confidence intervals (CIs) for the presence of SO and frailty versus robust by comparing tertiles 2 and 3 with tertile 1 of the DII as the exposure variables. We presented the following two adjustment models: (1) an age- and sex-adjusted model; and (2) model 1, with the inclusion of age, sex, education, monthly household income, family structure, number of chronic diseases, number of prescription drugs, chewing status, smoking status, alcohol consumption, physical activity, and total energy intake as covariates. We tested for the presence of multicollinearity using the variance inflation factor (VIF) and found no evidence of multicollinearity among the covariates (VIF < 10). The potential linear trends across increasing DII tertiles were tested by assigning the medians to each DII tertile as a continuous variable. All statistical tests were two-sided with a statistical significance level of 0.05. All statistical analyses were performed using SAS software version 9.4 (SAS Institute, Cary, NC, USA).

Participant characteristics

The characteristics of study participants in the robust, SO, and frailty groups are presented in Table  1 . Among 950 participants, 93 (9.8%) and 103 (10.8%) had SO and frailty, respectively. Participants in the frailty group were more likely to be older females with lower education levels, household income, and physical activity levels than those in the robust and SO groups. Furthermore, they were more likely to live alone, have a higher number of chronic diseases and prescribed drugs, and refer uncomfortable chewing status.

Participants in the SO group had lower ASM measures (ASM, ASM/height 2 , and ASM/weight) and HDL-cholesterol levels as well as higher waist circumference and BMI than those in the frailty and robust groups. Participants with frailty had the lowest hand grip strength, the longest time for repeated five-chair stands, and the highest triglycerides and hs-CRP levels (Table  2 ).

DII and Individual DII component characteristics

The total mean DII scores were 2.67, 1.96, and 1.82 in the frailty, robust, and SO groups, respectively. Most nutrient intakes, except for vitamin A, vitamin C, vitamin B 1 , niacin, carbohydrate, vitamin B 12 , and iron, were below the recommended nutrient intake (RNI) or adequate intake (AI) levels based on the KDRI 2020. Among individual nutrients and foods included in the DII, vitamin E, niacin, vitamin B 6 , energy, and protein intakes were lower in the frailty group than those in the robust and SO groups (Table  3 ).

DII in relation to SO and frailty

The associations between the DII and SO as well as frailty are shown in Fig.  1 . Participants with higher DII scores had a higher frailty prevalence (4.4%, 8.8%, and 19.2% in tertiles 1 [T1], 2 [T2], and 3 [T3], respectively). Participants in DII T2 had higher SO prevalence than those in T1 and T3 (9.5%, 13.6%, and 6.3% in T1, T2, and T3, respectively). After adjusting for age and sex, OR (95% CI) for frailty versus robust (comparing DII tertile 3 to tertile 1) was 3.35 (1.78–6.29; P -trend < 0.0001), whereas no association between DII and SO was observed. A positive association between the DII and frailty remained consistent in multivariable-adjusted models (T3 vs. T1, OR, 2.26; 95% CI, 1.07–4.80; P -trend = 0.02) (Fig.  1 ) and further adjustment for blood markers (T3 vs. T1, OR, 2.43; 95% CI, 1.13–5.23; P -trend = 0.02) (Supplemental Fig.  2 ).

Odds ratios (95% confidence intervals) for sarcopenic obesity and frailty by tertiles of the dietary inflammatory index score Abbreviations: T, tertile; DII, dietary inflammatory index; OR, odds ratio; CI, confidence interval

Data source: Korean Frailty and Aging Cohort Survey (KFACS)

Note: Multinomial logistic regression models are used to estimate odds ratios and their corresponding 95% confidence intervals for the presence of SO and frailty versus robust by comparing tertile 2 and 3 with tertile 1 of the DII as the exposure variables. The number of each SO and frailty cases and their percentages are presented as No. cases (%) according to the DII tertile. P for trends is determined by treating the median value of the DII score as a continuous variable using multinomial logistic regression models. The multivariable-adjusted model is adjusted for age, sex, education level, monthly household income level, family structure, number of chronic diseases, number of prescribed drugs, chewing status, smoking status, alcohol consumption, physical activity level, and total energy intake

In this cross-sectional study of 950 Korean older participants, we observed that individuals with frailty had the highest DII score, indicating a higher proinflammatory diet consumption, whereas those with SO had the lowest DII score, indicating a lower proinflammatory diet consumption. Our findings showed a significant positive association between the DII and frailty, which remained statistically significant even after controlling for demographic and lifestyle variables. However, no meaningful association was observed between the DII and SO.

In our study, the positive association between the DII and frailty is consistent with previous studies. Five cross-sectional studies conducted on individuals aged 60 years and older demonstrated that those with higher DII scores had a 1.7–3.6 times greater likelihood of being frail than those with lower DII scores [ 20 , 31 , 32 , 33 , 34 ]. Furthermore, a prospective study conducted in US adults with or at a high risk of knee osteoarthritis (age range: 45–85 years) showed that individuals with higher DII scores had a 1.4 times higher risk of developing frailty [ 35 ]. Collectively, these findings suggest that diets with a higher proinflammatory potential, as indicated by higher DII scores, have negative impacts on frail status, particularly in older adult populations.

Although our study identified a positive association between the DII and frailty, the results did not support our hypothesis regarding the association between the DII and SO. In the older adult population, SO may be a more significant predictor of age-related changes in body composition than sarcopenia or obesity alone [ 36 ]. This is because these changes frequently occur simultaneously and have a synergistic adverse effect on cardiometabolic health and mortality risk with aging [ 36 ]. Several factors might explain our findings. First, variations in body composition measurement techniques and the definition of sarcopenia and obesity across studies may contribute to inconsistent results [ 37 ]. Owing to the lack of the use of different equipment, accurately and consistently assessing both conditions is challenging [ 37 ]. We used the ASMI for sarcopenia definition as recommended by the AWGS [ 5 ] and waist circumference for obesity definition to better reflect visceral adiposity [ 38 ]. However, neither of these measures was based on gold standard measures of body composition. Furthermore, we observed no significant differences when comparing different definitions of sarcopenic obesity based on BMI or percent body fat (Supplemental Fig.  3 ). To address this issue, future studies should strive for the use of more precise body composition assessment techniques including computed tomography (CT) or magnetic resonance imaging (MRI) is encouraged.

Alternatively, it is plausible that frailty may be an inclusive concept that encompasses both sarcopenia and obesity. The definitions and diagnostic criteria for sarcopenia and frailty share several similarities [ 39 ], and sarcopenia, obesity, and frailty share common underlying mechanisms, particularly involving low-grade inflammation [ 7 , 12 , 21 ]. However, notably, sarcopenia or obesity may precede frailty development but not vice versa [ 39 , 40 ]. Moreover, considering that diet represents a long-term habitual exposure, detecting associations with relatively short-term outcomes including body composition may be insufficient. Also, diet could reflect possible impairments in the overall health status besides the SO of frailty.

Another potential explanation is that the SO group in our study exhibited better oral health, healthier lifestyle habits, and higher education and income levels. These characteristics might also influence the consumption of healthy food, which could have a positive impact on anti-inflammatory effects [ 41 , 42 ]. Although confounding variables were adjusted in the analysis model, it is possible that unmeasured positive factors in this group may have diluted the association between the DII and SO.

In our study, participants in the SO group showed more favorable dietary intake profiles overall than those in the frailty and robust groups (Tables  3 and Supplemental Table 2 ). Among those in the SO group, the proportion of participants who did not meet the Dietary Reference Intakes for Koreans recommendations for vitamins E, C, B 1 , B 2 , B 3 , B 6 , and folate was the lowest, whereas the proportion with inadequate protein intake was the highest. This finding contradicts a previous study that reported negative associations between several nutrient intakes and SO [ 43 ]. Conversely, the frailty group had lower vitamin E, B 3 , B 6 , energy, and protein intakes than the robust group, which aligns with previous knowledge [ 44 ]. Collectively, it is possible that the pathophysiology of SO may be inadequately explained by dietary intake. To further investigate this possibility, future studies using repeated dietary measures are required.

This study had several strengths. To our knowledge, this is the first study to simultaneously examine the association between dietary inflammatory potential and two age-related conditions. Second, the KFACS provided a wide range of data necessary for assessing multiple phenotypes. Body composition data from DEXA and comprehensive physical examination data allowed us to assess two frequently examined phenotypes, including SO and frailty, respectively. However, our study also had several limitations. First, owing to its cross-sectional design, we cannot establish a strong causal relationship between proinflammatory diets and frailty incidence. Second, as our study focused on Korean adults aged 70–84 years, our findings may not be applicable to other populations. Third, a single 24-h dietary recall may not fully capture individuals’ dietary habits and may poorly represent usual individual intake owing to day-to-day variations in nutrients or foods consumed [ 45 ]; however, compared with other methods, the 24-h recall method is considered to have the least bias. Fourth, while the original DII included 45 components, only 23 of them were included in the DII calculation owing to data availability, leaving behind 22 components. It is possible that the results might differ if these remaining components were added. However, of note, most of the missing components are not commonly consumed in the Korean diet. Additional studies are needed to explore this issue by including a broader range of dietary components in the DII calculation. Finally, obesity misclassification may have occurred as waist circumference may not fully capture adiposity compared with direct measures, including CT or MRI [ 38 ].

In conclusion, our findings indicate that a higher DII score is associated with increased odds of having frailty but not with SO in Korean older adults. These results suggest that proinflammatory diets have a greater impact on frailty than that on SO in the older adult population. This finding may be explained by the fact that frailty could be a more comprehensive condition that is primarily related to advanced age, whereas sarcopenia and obesity are not exclusively related to advanced age [ 39 ]. To confirm the replicability of these results, further large-scale studies based on a prospective cohort study design and randomized controlled trials are needed. Furthermore, to gain a better understanding of the observed associations, studies exploring the underlying mechanisms should be conducted.

Data availability

The data used in the study is not publicly available, but the data used and/or analyzed during the current study are available from the corresponding author on reasonable request.

United Nations Population Division. World Population Prospects. 2022 Revision. Population ages 65 and above (% of total population). Available online: https://data.worldbank.org/indicator/SP.POP.65UP.TO.ZS (accessed April 24 2023).

Baek JY, Lee E, Jung HW, Jang IY. Geriatrics fact sheet in Korea 2021. Ann Geriatr Med Res. 2021;25:65–71.

Article   PubMed   PubMed Central   Google Scholar  

WHO. Fact Sheet Musculoskeletal Health Published 14 July 2022. Available online: https://www.who.int/news-room/fact-sheets/detail/musculoskeletal-conditions (accessed Apr 24, 2023).

Ponti F, Santoro A, Mercatelli D, Gasperini C, Conte M, Martucci M, Sangiorgi L, Franceschi C, Bazzocchi A. Aging and Imaging Assessment of body composition: from Fat to facts. Front Endocrinol (Lausanne). 2019;10:861.

Article   PubMed   Google Scholar  

Chen LK, Woo J, Assantachai P, Auyeung TW, Chou MY, Iijima K, Jang HC, Kang L, Kim M, Kim S, Kojima T, Kuzuya M, Lee JSW, Lee SY, Lee WJ, Lee Y, Liang CK, Lim JY, Lim WS, Peng LN, Sugimoto K, Tanaka T, Won CW, Yamada M, Zhang T, Akishita M, Arai H. Asian Working Group for Sarcopenia: 2019 Consensus Update on Sarcopenia diagnosis and treatment. J Am Med Dir Assoc. 2020;21:300–e307302.

Roubenoff R. Sarcopenic obesity: the confluence of two epidemics. Obes Res. 2004;12:887–8.

Clegg A, Young J, Iliffe S, Rikkert MO, Rockwood K. Frailty in elderly people. Lancet. 2013;381:752–62.

O’Caoimh R, Sezgin D, O’Donovan MR, Molloy DW, Clegg A, Rockwood K, Liew A. Prevalence of frailty in 62 countries across the world: a systematic review and meta-analysis of population-level studies. Age Ageing. 2021;50:96–104.

Jung H, Kim M, Lee Y, Won CW. Prevalence of physical Frailty and its multidimensional risk factors in Korean Community-Dwelling older adults: findings from Korean Frailty and Aging Cohort Study. Int J Environ Res Public Health 2020;17.

Gao Q, Mei F, Shang Y, Hu K, Chen F, Zhao L, Ma B. Global prevalence of sarcopenic obesity in older adults: a systematic review and meta-analysis. Clin Nutr. 2021;40:4633–41.

Park JE, Lee S, Kim K. The effect of combining nutrient intake and physical activity levels on central obesity, Sarcopenia, and sarcopenic obesity: a population-based cross-sectional study in South Korea. BMC Geriatr. 2023;23:119.

Article   CAS   PubMed   PubMed Central   Google Scholar  

Koliaki C, Liatis S, Dalamaga M, Kokkinos A. Sarcopenic obesity: epidemiologic evidence, pathophysiology, and therapeutic perspectives. Curr Obes Rep. 2019;8:458–71.

Galland L. Diet and inflammation. Nutr Clin Pract. 2010;25:634–40.

Shivappa N, Steck SE, Hurley TG, Hussey JR, Hébert JR. Designing and developing a literature-derived, population-based dietary inflammatory index. Public Health Nutr. 2014;17:1689–96.

Marx W, Veronese N, Kelly JT, Smith L, Hockey M, Collins S, Trakman GL, Hoare E, Teasdale SB, Wade A, Lane M, Aslam H, Davis JA, O’Neil A, Shivappa N, Hebert JR, Blekkenhorst LC, Berk M, Segasby T, Jacka F. The Dietary Inflammatory Index and Human Health: an Umbrella Review of Meta-analyses of Observational studies. Adv Nutr. 2021;12:1681–90.

Diao H, Yan F, He Q, Li M, Zheng Q, Zhu Q, Fang F, Cui W. Association between Dietary Inflammatory Index and Sarcopenia: a Meta-analysis. Nutrients 2023;15.

Farhangi MA, Vajdi M. The association between dietary inflammatory index and risk of central obesity in adults: an updated systematic review and meta-analysis. Int J Vitam Nutr Res. 2020;90:535–52.

Article   CAS   PubMed   Google Scholar  

Moradi S, Hadi A, Mohammadi H, Asbaghi O, Zobeiri M, Marx W, Entezari MH. Dietary inflammatory index and the risk of Frailty among older adults: a systematic review and Meta-analysis. Res Aging. 2021;43:323–31.

Park S, Na W, Sohn C. Relationship between osteosarcopenic obesity and dietary inflammatory index in postmenopausal Korean women: 2009 to 2011 Korea National Health and Nutrition Examination Surveys. J Clin Biochem Nutr. 2018;63:211–6.

Kim D, Park Y. Association between the Dietary Inflammatory Index and Risk of Frailty in older individuals with Poor Nutritional Status. Nutrients 2018;10.

Picca A, Coelho-Junior HJ, Calvani R, Marzetti E, Vetrano DL. Biomarkers shared by frailty and sarcopenia in older adults: a systematic review and meta-analysis. Ageing Res Rev. 2022;73:101530.

Lee D, Kim M, Won CW. Common and different characteristics among combinations of physical frailty and sarcopenia in community-dwelling older adults: the Korean Frailty and Aging Cohort Study. Geriatr Gerontol Int. 2022;22:42–9.

Yang M, Hu M, Zhang Y, Jia S, Sun X, Zhao W, Ge M, Dong B. Sarcopenic obesity is associated with frailty among community-dwelling older adults: findings from the WCHAT study. BMC Geriatr. 2022;22:863.

Won CW, Lee S, Kim J, Chon D, Kim S, Kim CO, Kim MK, Cho B, Choi KM, Roh E, Jang HC, Son SJ, Lee JH, Park YS, Lee SG, Kim BJ, Kim HJ, Choi J, Ga H, Lee KJ, Lee Y, Kim M. Korean frailty and aging cohort study (KFACS): cohort profile. BMJ Open. 2020;10:e035573.

Kim S-Y, Kang M-S, Kim S-N, Kim J-B, Cho Y-S, Park H-J, Kim J-H. Food composition tables and national information network for food nutrition in Korea. Food Sci Ind. 2011;44:2–20.

Google Scholar  

Seo MH, Lee WY, Kim SS, Kang JH, Kang JH, Kim KK, Kim BY, Kim YH, Kim WJ, Kim EM, Kim HS, Shin YA, Shin HJ, Lee KR, Lee KY, Lee SY, Lee SK, Lee JH, Lee CB, Chung S, Cho YH, Choi KM, Han JS, Yoo SJ. 2018 Korean Society for the Study of Obesity Guideline for the management of obesity in Korea. J Obes Metab Syndr. 2019;28:40–5.

Fried LP, Tangen CM, Walston J, Newman AB, Hirsch C, Gottdiener J, Seeman T, Tracy R, Kop WJ, Burke G, McBurnie MA. Frailty in older adults: evidence for a phenotype. J Gerontol Biol Sci Med Sci. 2001;56:M146–156.

Article   CAS   Google Scholar  

Chen LK, Liu LK, Woo J, Assantachai P, Auyeung TW, Bahyah KS, Chou MY, Chen LY, Hsu PS, Krairit O, Lee JS, Lee WJ, Lee Y, Liang CK, Limpawattana P, Lin CS, Peng LN, Satake S, Suzuki T, Won CW, Wu CH, Wu SN, Zhang T, Zeng P, Akishita M, Arai H. Sarcopenia in Asia: consensus report of the Asian Working Group for Sarcopenia. J Am Med Dir Assoc. 2014;15:95–101.

Oh JY, Yang YJ, Kim BS, Kang JH. Validity and reliability of Korean version of International Physical Activity Questionnaire (IPAQ) short form. J Korean Acad Family Med. 2007;28:532–41.

Ministry of Health and Welfare. The Korean Nutrition Society. Dietary reference intakes for koreans 2020. Sejong: Ministry of Health and Welfare; 2020.

Lohman MC, Resciniti NV, Wirth MD, Shivappa N, Hébert JR, Obesity. Dietary inflammation, and Frailty among older adults: evidence from the National Health and Nutrition Examination Survey. J Nutr Gerontol Geriatr. 2019;38:18–32.

Resciniti NV, Lohman MC, Wirth MD, Shivappa N, Hebert JR. Dietary inflammatory Index, Pre-frailty and Frailty among older US adults: evidence from the National Health and Nutrition Examination Survey, 2007–2014. J Nutr Health Aging. 2019;23:323–9.

Laclaustra M, Rodriguez-Artalejo F, Guallar-Castillon P, Banegas JR, Graciani A, Garcia-Esquinas E, Lopez-Garcia E. The inflammatory potential of diet is related to incident frailty and slow walking in older adults. Clin Nutr. 2020;39:185–91.

de Oliveira GB, Yogi CM, de Carvalho Vidigal F, Lima DB, de Souza Paulino AH, de Brito TRP. Pro-inflammatory diet, frailty and sarcopenia: a study with older people in outpatient care. Res Soc Dev. 2021;10:e103101724488–103101724488.

Article   Google Scholar  

Shivappa N, Stubbs B, Hébert JR, Cesari M, Schofield P, Soysal P, Maggi S, Veronese N. The Relationship between the Dietary Inflammatory Index and Incident Frailty: a longitudinal cohort study. J Am Med Dir Assoc. 2018;19:77–82.

Atkins JL, Wannamathee SG. Sarcopenic obesity in ageing: cardiovascular outcomes and mortality. Br J Nutr. 2020;124:1102–13.

Batsis JA, Villareal DT. Sarcopenic obesity in older adults: aetiology, epidemiology and treatment strategies. Nat Rev Endocrinol. 2018;14:513–37.

Neeland IJ, Ross R, Després JP, Matsuzawa Y, Yamashita S, Shai I, Seidell J, Magni P, Santos RD, Arsenault B, Cuevas A, Hu FB, Griffin B, Zambon A, Barter P, Fruchart JC, Eckel RH. Visceral and ectopic fat, atherosclerosis, and cardiometabolic disease: a position statement. Lancet Diabetes Endocrinol. 2019;7:715–25.

Cederholm T. Overlaps between Frailty and Sarcopenia definitions. Nestle Nutr Inst Workshop Ser. 2015;83:65–9.

Yuan L, Chang M, Wang J. Abdominal obesity, body mass index and the risk of frailty in community-dwelling older adults: a systematic review and meta-analysis. Age Ageing. 2021;50:1118–28.

Abete I, Konieczna J, Zulet MA, Galmés-Panades AM, Ibero-Baraibar I, Babio N, Estruch R, Vidal J, Toledo E, Razquin C, Bartolomé R, Díaz-Lopez A, Fiol M, Casas R, Vera J, Buil-Cosiales P, Pintó X, Corbella E, Portillo MP, de Paz JA, Martín V, Daimiel L, Goday A, Rosique-Esteban N, Salas-Salvadó J, Romaguera D, Martínez JA. Association of lifestyle factors and inflammation with sarcopenic obesity: data from the PREDIMED-Plus trial. J Cachexia Sarcopenia Muscle. 2019;10:974–84.

Schoufour JD, Tieland M, Barazzoni R, Ben Allouch S, van der Bie J, Boirie Y, Cruz-Jentoft AJ, Eglseer D, Topinková E, Visser B, Voortman T, Tsagari A, Weijs PJM. The relevance of Diet, Physical Activity, Exercise, and Persuasive Technology in the Prevention and Treatment of Sarcopenic Obesity in older adults. Front Nutr. 2021;8:661449.

Son J, Yu Q, Seo JS. Sarcopenic obesity can be negatively associated with active physical activity and adequate intake of some nutrients in Korean elderly: findings from the Korea National Health and Nutrition Examination Survey (2008–2011). Nutr Res Pract. 2019;13:47–57.

Yannakoulia M, Ntanasi E, Anastasiou CA, Scarmeas N. Frailty and nutrition: from epidemiological and clinical evidence to potential mechanisms. Metabolism. 2017;68:64–76.

National Cancer Institute. Dietary Assessment Primer. 24-hour Dietary Recall (24HR) At a Glance [Internet]. Available: https://dietassessmentprimer.cancer.gov/profiles/recall (accessed May 3, 2023).

Download references

Yunhwan Lee’s work was supported by a grant from the Korea Health Technology R&D Project through the Korean Health Industry Development Institute, which is funded by the Ministry of Health & Welfare, Republic of Korea (grant number: HI15C3153). Kirang Kim’s work was supported by a grant from National Research Foundation of Korea funded by the Ministry of Education (NRF-2021R1I1A3049883). Sohyun Park’s work was supported by the Basic Science Research Program through the NRF funded by the Ministry of Science (NRF-2021R1A6A1A03044501) and the funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Author information

Kirang Kim and Sohyun Park contributed equally to this work.

Authors and Affiliations

Department of Health Care Policy Research, Korea Institute for Health and Social Affairs, Sejong, South Korea

Sukyoung Jung

Department of Preventive Medicine and Public Health, Ajou University School of Medicine, Suwon, South Korea

Yunhwan Lee

Department of Food Science and Nutrition, Dankook University, 119 Dandae-ro, Dongnam-gu, Cheonan, 31116, South Korea

Department of Food Science and Nutrition, Hallym University, Chuncheon, Gangwon, 24252, South Korea

Sohyun Park

The Korean Institute of Nutrition, Hallym University, Chuncheon, South Korea

You can also search for this author in PubMed   Google Scholar

Contributions

S.J. contributed to data curation, formal analysis, visualization, original draft preparation and Y.L. contributed to research funding, original draft preparation, reviewing and editing. K.K. and S.P. contributed to conceptualization, writing, reviewing and editing. All authors gave final approval and agreed to be accountable for all aspects of ensuring integrity and accuracy.

Corresponding authors

Correspondence to Kirang Kim or Sohyun Park .

Ethics declarations

Ethical approval and consent to participate.

The study was reviewed and approved by the Institutional Review Boards of Dankook University, Hanyang University and Kyung Hee University, and was conducted in accordance with the Declaration of Helsinki. All participants provided written informed consent.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Supplementary Material 1

Rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it.The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/ .

Reprints and permissions

About this article

Cite this article.

Jung, S., Lee, Y., Kim, K. et al. Association of the dietary inflammatory index with sarcopenic obesity and frailty in older adults. BMC Geriatr 24 , 654 (2024). https://doi.org/10.1186/s12877-024-05239-z

Download citation

Received : 18 December 2023

Accepted : 23 July 2024

Published : 03 August 2024

DOI : https://doi.org/10.1186/s12877-024-05239-z

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Dietary inflammatory index
• Sarcopenic obesity
• Korean older adults

BMC Geriatrics

ISSN: 1471-2318

• Submission enquiries: [email protected]
• General enquiries: [email protected]

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• Open access
• Published: 08 August 2024

Proteomic aging clock predicts mortality and risk of common age-related diseases in diverse populations

• M. Austin Argentieri   ORCID: orcid.org/0000-0003-0242-853X 1 , 2 , 3 ,
• Sihao Xiao 1 , 4 ,
• Derrick Bennett   ORCID: orcid.org/0000-0002-9170-8447 1 ,
• Laura Winchester 4 , 5 ,
• Alejo J. Nevado-Holgado 4 , 5 ,
• Upamanyu Ghose 4 , 5 ,
• Ashwag Albukhari   ORCID: orcid.org/0000-0003-3669-1184 4 , 6 ,
• Pang Yao 1 ,
• Mohsen Mazidi 1 ,
• Jun Lv   ORCID: orcid.org/0000-0001-7916-3870 7 , 8 , 9 ,
• Iona Millwood   ORCID: orcid.org/0000-0002-0807-0682 1 ,
• Hannah Fry   ORCID: orcid.org/0009-0006-1584-2042 1 ,
• Rodosthenis S. Rodosthenous 10 ,
• Jukka Partanen   ORCID: orcid.org/0000-0001-6681-4734 11 ,
• Zhili Zheng 2 ,
• Mitja Kurki 2 ,
• Mark J. Daly   ORCID: orcid.org/0000-0002-0949-8752 2 , 3 ,
• Aarno Palotie   ORCID: orcid.org/0000-0002-2527-5874 10 ,
• Cassandra J. Adams   ORCID: orcid.org/0000-0001-9589-4526 4 , 12 ,
• Liming Li 7 , 8 , 9 ,
• Robert Clarke   ORCID: orcid.org/0000-0002-9802-8241 1 ,
• Najaf Amin 1 ,
• Zhengming Chen   ORCID: orcid.org/0000-0001-6423-105X 1   na1 &
• Cornelia M. van Duijn   ORCID: orcid.org/0000-0002-2374-9204 1 , 4   na1  

Nature Medicine ( 2024 ) Cite this article

Metrics details

• Computational models
• Machine learning
• Predictive markers
• Preventive medicine
• Proteome informatics

Circulating plasma proteins play key roles in human health and can potentially be used to measure biological age, allowing risk prediction for age-related diseases, multimorbidity and mortality. Here we developed a proteomic age clock in the UK Biobank ( n  = 45,441) using a proteomic platform comprising 2,897 plasma proteins and explored its utility to predict major disease morbidity and mortality in diverse populations. We identified 204 proteins that accurately predict chronological age (Pearson r  = 0.94) and found that proteomic aging was associated with the incidence of 18 major chronic diseases (including diseases of the heart, liver, kidney and lung, diabetes, neurodegeneration and cancer), as well as with multimorbidity and all-cause mortality risk. Proteomic aging was also associated with age-related measures of biological, physical and cognitive function, including telomere length, frailty index and reaction time. Proteins contributing most substantially to the proteomic age clock are involved in numerous biological functions, including extracellular matrix interactions, immune response and inflammation, hormone regulation and reproduction, neuronal structure and function and development and differentiation. In a validation study involving biobanks in China ( n  = 3,977) and Finland ( n  = 1,990), the proteomic age clock showed similar age prediction accuracy (Pearson r  = 0.92 and r  = 0.94, respectively) compared to its performance in the UK Biobank. Our results demonstrate that proteomic aging involves proteins spanning multiple functional categories and can be used to predict age-related functional status, multimorbidity and mortality risk across geographically and genetically diverse populations.

Similar content being viewed by others

Undulating changes in human plasma proteome profiles across the lifespan

Organ aging signatures in the plasma proteome track health and disease

Blood protein assessment of leading incident diseases and mortality in the UK Biobank

Age is a major determinant for most common chronic diseases and causes of death 1 , 2 . Aging involves a progressive loss of physiological integrity and function over time, which ultimately leads to the development, and often co-occurrence, of major diseases and death. Rates of major chronic diseases such as ischemic heart disease (IHD), stroke, diabetes, liver and kidney diseases, neurodegenerative diseases and various cancers (for example, lung and colorectal) all increase with age 2 , 3 , 4 , although there is substantial variation across individuals in the timing and severity of age-related disorders. Chronological age is a strong but imperfect surrogate measure of ‘biological’ aging, which can be estimated more precisely by using ‘omics data to capture the level of biological functioning of an individual in comparison to an expected level of functioning for a given chronological age 5 .

How fast we age not only determines individual risk of major chronic diseases and premature death, but also shapes the extent of morbidity and disability in the population, which has a major impact on healthcare systems 2 . The ability to quantify, and possibly intervene upon, biological aging may therefore have important consequences for prevention of multimorbidity and premature death 6 . Some of the earliest and most successful biological aging clocks developed to date use DNA methylation (DNAm) 5 , 7 . Although DNAm can provide a window into epigenetic changes resulting from environmental exposures (for example, smoking 8 ), protein levels may provide a more direct mechanistic and functional insight into aging biology 5 . Loss of proteostasis is a primary hallmark of aging 9 and several previous studies have identified aging-related proteins (APs) and used these to develop proteomic age clocks 10 , 11 , 12 , 13 , 14 , 15 , 16 , 17 to predict risk of certain diseases and mortality; however, none of these previous studies has developed a proteomic age clock in a large and well-powered general population sample that would allow a comprehensive assessment of the predictive performance of a proteomic clock across all major chronic diseases and age-related functional traits. Furthermore, to our knowledge none of these previous proteomic age clocks has been validated independently across diverse ancestry populations.

To address these gaps in knowledge, we used blood proteomic information to develop a proteomic age clock in a large sample of participants from the UK Biobank (UKB; n  = 45,441) and validated this model in the China Kadoorie Biobank (CKB; n  = 3,977) and FinnGen ( n  = 1,990), both of which are geographically and genetically distinct from the UK population. We then systematically assessed the influence of proteomic age gap (ProtAgeGap; defined as the difference between protein-predicted age and chronological age) on 27 aging-related phenotypes related to biological, functional and cognitive status; all-cause mortality; and incidence of 26 common age-related diseases that are either major causes of death or are highly prevalent in aging populations (rheumatoid arthritis, macular degeneration, osteoarthritis and osteoporosis).

Proteomic age clock

A schematic representation of the study design and main analytic approaches is shown in Fig. 1 . Characteristics of participants across the discovery (UKB) and two validation cohorts (CKB and FinnGen) are shown in Table 1 . We used plasma proteomic expression data from the subset of 45,441 randomly selected UKB participants (54% female, age range 39–71 years), 3,977 CKB participants in an IHD case–cohort study (54% female, age range 30–78 years) and 1,990 Finnish (FinnGen) participants (52% female, age range 19–78 years). Across 11–16 years of follow-up in the UKB and 11–14 years of follow-up in the CKB, there were 4,828 (10.6%) and 1,426 (36%) deaths, respectively. Proteomic profiling was conducted among mostly healthy participants in FinnGen without major diseases and only 1% ( n  = 22) of FinnGen participants with proteomic data died during follow-up.

a , UKB participants were split into training and test sets at a 70:30 ratio. In the training set, a LightGBM model was trained to predict chronological age using 2,897 plasma proteins and fivefold cross-validation. We identified 204 proteins relevant for predicting chronological age using the Boruta feature selection algorithm and retrained a refined LightGBM model using these 204 proteins, which was then evaluated in the UKB test set. b , Independent data from the CKB and FinnGen were used for further independent validation of the proteomic age clock model. c , Protein-predicted age (ProtAge) was calculated in the full UKB sample using fivefold cross-validation and LightGBM. ProtAgeGap was calculated as the difference between ProtAge and chronological age. We used linear and logistic regression to test associations between ProtAgeGap and a comprehensive panel of biological aging markers and measures of frailty and physical/cognitive status. Further, we used Cox proportional hazards models to test associations between ProtAgeGap and mortality, 14 common diseases and 12 cancers. Most association analyses were carried out only in the UKB, due to the smaller sample size in the CKB and the lack of disease cases in FinnGen. Figure created with BioRender.com .

We randomly split the UKB cohort into 70% training and 30% test sets to develop the proteomic age clock. In the training phase ( n = 31,808), we compared six machine-learning methods (LASSO, elastic net, gradient boosting and three neural networks) to train proteomic age clock models to predict chronological age using normalized expression of 2,897 proteins from the Olink Explore 3072 panel. We found that gradient boosting (LightGBM 18 ) showed the second-best age prediction accuracy in the UKB test set ( n  = 13,633) but the highest accuracy in the independent samples from the CKB and FinnGen (Supplementary Fig. 1 ). Based on its superior generalizability, we selected LightGBM as our final model and used the Boruta 19 feature selection algorithm and SHAP (SHapley Additive exPlanations) 20 values to identify the subset of all proteins relevant for predicting chronological age ( Methods ). This process resulted in the identification of 204 APs in our dataset (Supplementary Table 1 ). The correlation structure among these proteins is shown in Supplementary Fig. 2 . Protein-predicted age (ProtAge) from this 204-protein model explained a similar degree of variation in chronological age compared to the 2,897-protein model (Supplementary Fig. 3a,b ), with similar model error across different age groups (Supplementary Fig. 4 ). Our gradient-boosting ProtAge model explained a high degree of variation in chronological age in the UKB test set ( R 2  = 0.88; Pearson r  = 0.94) and in the independent validation sets from the CKB ( R 2  = 0.82; Pearson r  = 0.92) and FinnGen ( R 2  = 0.87; Pearson r  = 0.94) (Fig. 2d–f ).

a , Density plot of age at recruitment in the UKB, CKB and FinnGen. b , Density plot of age at death in the UKB (4,784 deaths; 10.6%) and CKB (1,426 deaths; 36%). c , Counts of prevalent and incident cases of all common diseases studied in the UKB sample ( n  = 45,441). d , Performance of the trained proteomic aging model in the UKB holdout test set ( n  = 13,633). e , Performance of the trained proteomic aging model in the CKB ( n  = 3,977). f , Performance of the trained proteomic aging model in FinnGen ( n  = 1,990). g , Sex distributions of ProtAgeGap in the UKB (female n  = 24,579; male n  = 20,862), CKB (female n  = 2,137; male n  = 1,840) and FinnGen (female n  = 1,032; male n  = 958). h , Distributions of ProtAgeGap according to self-reported ethnicity in the UKB (white n  = 42,320; Black n  = 1,114; Asian n  = 1,016; other n  = 554; mixed n  = 293). i , Distributions of ProtAgeGap according to geographic region of residence in the CKB (Gansu n  = 397; Henan n  = 493; Hunan n  = 462; Sichuan n  = 341; Zhejiang n  = 342; Haikou n  = 298; Harbin n  = 598; Liuzhou n  = 379; Qingdao n  = 415; Suzhou n  = 252). Correlation coefficients shown in d – f are Pearson correlation coefficients. Violin plots in g – i , with center line, box limits and whiskers representing the median, interquartile range and minima/maxima within each group, respectively. RMSE, root mean squared error; MAE, mean absolute error.

To assess whether each of our AP’s association with age was stable over time, we used repeat protein expression measurements available for a subset of 149 ProtAge proteins among 1,085 UKB participants who had proteomic data measured at three time points (baseline (2006 to 2011), imaging study visit (2014+) and repeat imaging visit (2019+)). For each of these 149 APs, we assessed their association with age at each study visit using linear regression. Beta coefficients for the associations of these APs with age across all three time points were strongly correlated with each other (Pearson r  = 0.90–0.97), suggesting good stability of associations between APs and age across repeat visits spanning at least 9–13 years (Extended Data Fig. 1 ).

Using 204 APs in the final model, we calculated participants’ ProtAgeGap as the difference between ProtAge and chronological age in all three cohorts. In the UKB, the average years of ProtAgeGap among the top 5% and bottom 5% of ProtAgeGap was 6.3 and −6 years, respectively, resulting in a mean difference of approximately 12.3 years in biological aging between them. ProtAgeGap showed similar distributions across all three cohorts in females and males, across self-reported ethnicities in the UKB and across geographical regions in the CKB (Fig. 2g–i ).

As a final feature selection step, we employed recursive feature elimination using SHAP values ( Methods ) to identify a model of 20 proteins (ProtAge20) that achieved 95% of the age prediction performance of the 204-protein model (Pearson r  = 0.89, R 2  = 0.78; Supplementary Fig. 3c,d and Supplementary Table 2 ). Correlation among these 20 proteins is shown in Supplementary Fig. 5 . We further calculated the proteomic age gap according to these top 20 proteins (ProtAgeGap20) in the UKB using the same approach as above. In models with fewer than 20 proteins, model performance dropped precipitously (Supplementary Fig. 3d ).

Proteomic aging predicts frailty and aging phenotypes

To understand how proteomic aging may influence aging-related physiological and cognitive function, we examined associations in the UKB of ProtAgeGap with (1) a comprehensive frailty index 21 ( Methods ); (2) 16 individual measures of physical (for example, slow walking pace and grip strength) and cognitive function (reaction time and fluid intelligence); and (3) ten measures of biological aging (for example, telomere length and insulin-like growth factor 1 (IGF-1)) and clinical blood biochemistry (for example, albumin and creatinine). After adjustment for chronological age, sex and major sociodemographic and lifestyle confounders, ProtAgeGap was significantly associated with all measures investigated except for two liver biomarkers (alanine aminotransferase (ALT) and total bilirubin; Fig. 3a,b ). Among the biological aging mechanisms investigated (Fig. 3a ), increasing ProtAgeGap was associated with increasing levels of two kidney function biomarkers (cystatin C and creatinine), two liver enzymes (aspartate aminotransferase (AST), γ-glutamyl transferase (GGT)) and C-reactive protein; and was associated with decreased levels of albumin, IGF-1 and telomere length. Among physical measures (Fig. 3b ), increasing ProtAgeGap was associated with poor self-rated health, slow walking pace, self-rating one’s face as older than average, sleeping ≥10 h per day, feeling tired every day and having frequent insomnia. Increasing ProtAgeGap was also associated with higher values of a frailty index, systolic and diastolic blood pressure, reaction time, arterial stiffness and body mass index (BMI); and with lower values of bone mineral density, fluid intelligence, lung function and hand grip strength.

a , Associations between ProtAgeGap and biological aging mechanisms in the full UKB sample ( n  = 45,441). b , Associations between ProtAgeGap and measures of physiological and cognitive (reaction time and fluid intelligence) function in the full UKB sample ( n  = 45,441). c , Associations between ProtAgeGap and biological aging mechanisms in the subsample of UKB participants with no lifetime diagnosis of any of the 26 diseases studied ( n  = 20,315). d , Associations between ProtAgeGap and measures of physiological and cognitive function in the subsample of UKB participants with no lifetime diagnosis of any of the 26 diseases studied ( n  = 20,315). All models used linear or logistic regression and were adjusted for age, sex, Townsend deprivation index, recruitment center, ethnicity, IPAQ activity group and smoking status. In all plots, beta estimates (and 95% confidence intervals) for the association between ProtAgeGap and each outcome are shown on the x axis. Beta estimates in red are from the full 204-protein model (ProtAgeGap), whereas beta estimates in blue are from the smaller proteomic age clock model with 20 proteins (ProtAgeGap20). FEV 1 , forced expiratory volume in 1 s; IPAQ, International Physical Activity Questionnaire; FDR, false discovery rate.

To explore whether these associations are explained by reverse causation (resulting from a nondetected pathology), we also restricted the analyses to a subset of UKB participants who had no lifetime diagnoses (according to hospital inpatient, cancer registry and primary care records) of any of the 26 diseases studied ( n  = 20,315). Among these participants (Fig. 3c,d ), we found that ProtAgeGap remained significantly associated with nearly all markers except for albumin (which is a typical protein marker of end-stage morbidity), total bilirubin, self-rated facial aging, sleeping for 10+ hours per day and feeling tired every day (Fig. 3d ).

ProtAgeGap20 was associated with all aging functional phenotypes except for diastolic blood pressure. Compared to the 204-protein model, ProtAgeGap20 showed stronger effect estimates in relation to biological measures of aging (for example, telomeres and albumin) (Fig. 3a ) but somewhat smaller effect estimates for measures of frailty and physiological/cognitive function (Fig. 3b ). ProtAgeGap20 was significantly associated with all biological aging markers (Fig. 3c ) in the subset of UKB participants without lifetime disease diagnoses and was associated with all physiological measures except sleeping for 10+ hours per day, diastolic blood pressure and BMI (Fig. 3d ). Summary statistics from all models are shown in Supplementary Tables 3 – 6 .

Proteomic aging is a strong predictor of common diseases

UKB participants in the top, median and bottom deciles of ProtAgeGap showed divergent age-specific incidence rates of all-cause mortality and the 14 common noncancer diseases studied (Fig. 4a and Supplementary Table 16 ). Cumulative incidence trajectories according to these deciles of ProtAgeGap were similar in females and males (Supplementary Figs. 6 and 7 ). For those aged 65 years at recruitment, the highest cumulative incidence rates (equivalent to absolute risk) across the study follow-up period of 11–16 years for the top decile of ProtAgeGap were observed for osteoarthritis (59.4%), all-cause mortality (55.2%), IHD (50.6%), type 2 diabetes (35.3%) and chronic kidney disease (CKD; 33.6%). Neurodegenerative diseases (Parkinson’s disease, all-cause dementia and Alzheimer’s disease (AD)) all showed cumulative incidence rates below 1% in the bottom decile of ProtAgeGap across all recruitment ages, which may in part be explained by the fact that onset is typically at older ages for these diseases.

a , b , Cumulative incidence plots for the indicated diseases and mortality for the top, median and bottom deciles of ProtAgeGap in the UKB ( n  = 45,441) ( a ) and CKB ( n  = 3,977) ( b ). The number of incident cases is shown for each disease, indicating the total number of incident cases only among individuals in the three deciles shown, not the full dataset. Incidence rates are shown for the subsequent 11–16 years (UKB) or 11–14 years (CKB) of follow-up after recruitment for each given age at recruitment (for example, in a the cumulative incidence rate shown at age 65 years is the rate of incident cases in the 11–16 years of follow-up for those aged 65 years at recruitment). All plots show the cumulative density of events at a given timepoint based on the Kaplan–Meier survival function, with 95% confidence intervals shown in lighter shading. Diseases shown here for the CKB are those with more than 10 cases across the three deciles of ProtAgeGap.

In the CKB, we also calculated cumulative incidence rates according to deciles of ProtAgeGap for diseases with >10 incident cases across the three deciles of ProtAgeGap (Fig. 4b and Supplementary Table 17 ). We observed clear differences for IHD, all-cause mortality, all stroke and ischemic stroke. Differences were also observed for type 2 diabetes, chronic obstructive pulmonary disease (COPD), chronic liver diseases and CKD; however, confidence intervals were much wider due to a smaller number of incident cases.

We further used multivariable Cox proportional hazards models to investigate whether associations of ProtAgeGap with mortality and the 14 common noncancer diseases studied persisted after adjustment for chronological age, sex, smoking, physical activity, sociodemographic factors and clinical risk factors. ProtAgeGap showed a significant association with mortality and all noncancer incident disease outcomes except Parkinson’s disease across all models in the UKB (Fig. 5 ). In the fully adjusted model that also included covariates for BMI and prevalent hypertension (model 3), the largest effect size per 1-year increase of ProtAgeGap was observed for AD (hazard ratio (HR) 1.16; 95% confidence interval (CI) 1.12–1.20), all-cause dementia (HR 1.12; 95% CI 1.10–1.15) and CKD (HR 1.10; 95% CI 1.08–1.11). ProtAgeGap20 was associated with all diseases investigated, including Parkinson’s disease. Summary statistics from all models are shown in Supplementary Tables 7 – 12 .

Associations between ProtAgeGap and mortality and disease incidence using Cox proportional hazards models are shown for models with increasing levels of covariate adjustment. Shown on the x axis are HRs (and 95% CIs) for the effect of ProtAgeGap on the outcomes shown. Events listed are the total number of incident cases for each outcome. All models were run using the full UKB sample ( n  = 45,441). a , Model 1 was adjusted for age and sex. b , Model 2 was adjusted for age, sex, ethnicity, Townsend deprivation index, recruitment center, IPAQ activity group and smoking status. c , Model 3 was adjusted for age, sex, ethnicity, Townsend deprivation index, recruitment center, IPAQ activity group, smoking status, BMI and prevalent hypertension. HR estimates in red are from the full 204-protein model (ProtAgeGap), whereas estimates in blue are from the smaller proteomic age clock model with 20 proteins (ProtAgeGap20).

Based on these HRs (reported per 1-year increase of ProtAgeGap), we estimated that those in the top 5% of ProtAgeGap had on average a 2.6-fold higher risk of AD compared to those with no difference between ProtAge and chronological age and a 5.8-fold higher risk of AD compared to those in the bottom 5% of ProtAgeGap. For CKD, the increases in risk were 1.8-fold (top 5% versus 0) and 3.1-fold (top 5% versus bottom 5%) and for mortality the increases in risk were 1.9-fold (top 5% versus 0) and 3.6-fold (top 5% versus bottom 5%).

In the UKB, we also investigated the cumulative incidence of cancer diagnoses according to deciles of ProtAgeGap (Extended Data Fig. 2 ), with clear differences observed for eight cancers (breast, lung, prostate, colorectal, non-Hodgkin lymphoma, esophageal, ovarian and liver). In Cox multivariable models, ProtAgeGap was associated with four cancers (esophageal, lung, non-Hodgkin lymphoma and prostate) after adjustment for age, sex, sociodemographic and lifestyle factors, BMI and prevalent hypertension (Extended Data Fig. 3 ). Summary statistics are shown in Supplementary Tables 13 – 15 .

Although the analyses described above were adjusted for smoking status, we conducted further sensitivity analyses in never smokers. Among never smokers, ProtAgeGap remained significantly associated with mortality and all noncancer outcomes, except Parkinson’s disease (Extended Data Fig. 4a ). In a similar sensitivity analysis restricted to those within a normal weight range (BMI ≥ 18.5 and BMI < 25), ProtAgeGap remained significantly associated with all outcomes except Parkinson’s disease, macular degeneration and rheumatoid arthritis (Extended Data Fig. 4b ).

For the 20 proteins included in ProtAge20, we used Cox models to further assess associations of the individual proteins with all ProtAgeGap20-associated diseases. Notably, GDF15 was associated with 16 of 18 diseases studied, whereas ACRV1 (a testis-specific protein involved in spermatogenesis) was only associated with prostate cancer (Extended Data Fig. 5a ). In Extended Data Fig. 5b , the relative weight of each protein’s association with each disease is shown as a scaled z -score. As expected from an aging signature, many of the proteins that make up the proteomic age clock are associated with multiple major chronic diseases.

Proteomic aging increases with increasing multimorbidity

We defined multimorbidity as the number of lifetime diagnoses of any of the 26 diseases examined in the UKB and categorized participants according to having 0, 1, 2, 3 or 4+ lifetime diagnoses. We found that the average years of ProtAgeGap increased with number of lifetime conditions (Extended Data Fig. 6 ). We also found that this effect was more pronounced for younger participants at recruitment (aged 40–50 years; Extended Data Fig. 6a ), among whom the presence of disease was less common (Extended Data Fig. 6c ). On average, 1.5 more years of ProtAgeGap was observed in those with 4+ lifetime diagnoses compared to those with 0 diagnoses in participants aged 40–50 years at recruitment (Extended Data Fig. 6a ), whereas in those aged 51–65 years at recruitment we observed 0.8 more years of ProtAgeGap (Extended Data Fig. 6b ). The relationship between ProtAgeGap and multimorbidity status derived from health records was also reflected in self-reported health information. On average, 0.9 fewer years of ProtAgeGap was observed in those reporting excellent health (likely no diseases present) compared to those with poor self-reported health (Extended Data Fig. 6d ).

Biological functions and protein interaction networks

Testing for functional enrichment among the 204 APs revealed that these APs were enriched for one Gene Ontology (GO) biological process: anatomical structure development and developmental process. No enrichments were found using GO molecular function, Kyoto Encyclopedia of Genes and Genomes (KEGG) or Reactome; however, these 204 APs showed a highly interconnected subnetwork of 66 proteins with at least two node connections in a protein–protein interaction (PPI) network using coexpression information from the STRING database (Extended Data Fig. 7 ). Individual proteins with the greatest numbers of connections to other proteins were EGFR (involved in cancer drug resistance, brain structure and platelet count), CXCL12 (an immune-related chemokine involved in immune surveillance, inflammation response, tissue homeostasis and tumor growth and metastasis), ITGAV (an integrin protein implicated in body height, handedness, dyslexia and albumin/creatinine metabolism), CXCL9 (implicated in T cell function and inflammation) and CD8A (a CD8 antigen implicated in the innate immune system).

We also used SHAP interaction values from our trained ProtAge model to calculate a second PPI network that represents the interactions of proteins together in the model to predict age (Extended Data Fig. 8 ). Individual proteins with the largest numbers of connections to other proteins according to SHAP interaction values were ELN (an elastic fiber protein that makes up part of the extracellular matrix and confers elasticity to organs and tissues, including the heart, skin, lungs, ligaments and blood vessels), EDA2R (involved in the NF-κB and innate immune pathways and implicated in baldness, estradiol, testosterone and HDL metabolism), LTPB2 (a protein involved in BMI, blood pressure, neuroticism and anxiety, glaucoma and retina pathology, lung function and mortality), CXCL17 (a chemokine interacting with CXCL9 that plays a role in tumor genesis, antimicrobial defense through monocytes, macrophages and dendritic cells) and GDF15 (implicated in BMI, liver function, systemic lupus erythematosus and COVID-19). Overall, we found distinct results when using a data-driven approach to model PPIs using interactions from our machine-learning models versus using the most up-to-date experimental biological knowledge from the STRING database.

We further examined the roles and functions of the 20 proteins comprising the ProtAge20 score, which together capture ~95% of the 204-protein model’s ability to predict age. These key APs are involved in (1) cell adhesion and extracellular matrix (ECM) interactions (ELN, COL6A3, CDCP1, PODXL2, LTBP2, SCARF2 and ENG); (2) immune response and inflammation (CXCL17, LECT2, SCARF2 and GDF15); (3) hormone regulation and reproduction (FSHB, AGRP and ACRV1); (4) cell signaling (EDA2R, SCARF2 and PTPRR); (5) protease activity and enzymatic function (KLK3 and KLK7:); (6) regulation of body weight and energy balance (GDF15 and AGRP); (7) neuronal structure and function (GFAP and NEFL); and (8) development and differentiation (EDA2R, LTBP2 and ENG).

Comparisons with existing DNAm and proteomic age clocks

Proteins selected by our model showed very little overlap with corresponding genes from leading DNAm clocks, including the Horvath clock 22 , DNAm PhenoAge 23 and DunedinPACE 24 (Extended Data Fig. 9a ). Five corresponding genes from ProtAge overlapped with genes mapped to CpGs (by proximity) in the Horvath clock ( CSPG5 , CXADR , DKK3 , ENPP2 and POMC ) and 11 ProtAge genes overlapped with proximity genes mapped to DNAm PhenoAge ( AMANTSL5 , CALB1 , CTSF , CXADR , CYTL1 , DPEP2 , KLK8 , LHB , LMOD1 , MATN3 and NPL ). Only three ProtAge genes overlapped with proximity genes mapped to DunedinPACE ( ADAMTS13 , SORCS2 and TNXB ).

We also compared our findings with three of the largest published studies on proteomic aging, including (1) a systematic review of studies ( n  = 32) reporting protein associations with age (Johnson et al. 11 ), in which the authors developed a proteomic age clock using 85 proteins associated with age in at least three previous studies and validated it in the INTERVAL study ( n  = 3,301); (2) a recent study that identified 273 APs across several cohorts (Coenen et al. 10 ) ( n  = 37,650); and (3) a clock consisting of 373 APs developed in the INTERVAL and LonGenity cohorts (Lehallier et al. 12 ) ( n  = 4,263).

While the proteins selected by our model showed a greater overlap with those found in existing proteomic clocks, 134 of our ProtAge APs (64%) were not identified in any of these major previous studies on proteomic aging (Extended Data Fig. 9b ). Despite representing a largely novel set of APs, ProtAge also includes 15 APs present in the Johnson et al., Coenen et al. and Lehallier et al. analyses (Extended Data Fig. 9b and Extended Data Table 1 ). Notably, none of these 15 proteins overlap with corresponding genes from any of the DNAm clocks, suggesting that DNAm and proteomic clocks seem to converge on different gene sets. The overlap between the 204 ProtAge APs with all previous studies described here is shown in Supplementary Table 1 . Last, the top protein identified in a previously published protein inflammation clock 15 (CXCL9, an inflammatory cytokine) was identified as an AP in our model, albeit not within the top 20 proteins in our model. None of the ProtAge20 proteins or the top 20 proteins in the full ProtAge model were identified in this previous inflammation clock paper, including several inflammation-related proteins from our ProtAge model (GDF15 and CXCL17).

Our analyses of proteomic data from UK, Chinese and Finnish populations show that proteomic age signatures estimated in the UKB capture information that is highly generalizable across populations of diverse genetic ancestry and diverse levels of morbidity. Our study provides new evidence that proteomic aging is a common feature underlying a large and diverse range of traits related to physical function, frailty and cognitive status, in addition to established aging biomarkers (for example, telomeres and IGF-1). While previous studies have reported that DNAm age is not related to telomere length 25 , we show that proteomic aging is strongly inversely associated with telomere length, a key cellular hallmark of aging 9 . Of note, our study provides comprehensive and well-powered evidence demonstrating that proteomic aging is a reliable predictor of mortality and multimorbidity, and is associated with future risk of all 14 noncancer diseases studied and four common cancers (esophageal, prostate, lung and non-Hodgkin lymphoma). Our study also demonstrates that our proteomic age clock generalizes well to age groups and morbidity profiles beyond those represented in the UKB population. Of particular note is that our proteomic age model performed well in FinnGen participants (age range 20–80 years) who were largely free of disease and mortality and who were up to 20 years younger than participants in the UKB (age range 40–70 years).

This study has directly tested associations of proteomic aging with disease, multimorbidity and age-related functional status in a comprehensive manner within a large and well-powered sample. While one previous study did systematically investigate different biological clocks against 27 major disease outcomes, the proteomic age clock tested was associated with only one disease at nominal significance, likely due to a lack of statistical power in this relatively small study ( n  = 805) 17 . Moreover, we have reduced the impact of reverse causation bias in our study by restricting the analyses to participants without any major lifetime disease diagnoses and demonstrate consistent findings with the overall analyses.

When compared to the limited number of physiological and biological function measures that were analyzed in the previous proteomic clock analyses described above, our ProtAge clock demonstrated improved performance. Coenen et al. 10 reported only marginally significant partial correlations between their proteomic age clock and blood uric acid and magnesium, but nonsignificant correlations with the other 57 blood markers tested, including albumin, creatinine, γ-glutamyl transferase and C-reactive protein, all of which were strongly associated with ProtAgeGap in our study. The Johnson et al. 11 clock study did not investigate mortality, morbidity or any frailty or cognitive measures. As these previous papers used SOMAscan proteomics, it is unclear whether their lack of overlap with ProtAge APs and differences in aging phenotype associations may reflect differences between the Olink and SOMAscan platforms. Previous research comparing SOMAscan and Olink data directly demonstrated substantial discrepancies in protein–phenotype associations and genetic protein quantitative trait loci mapping between the two platforms 26 ; however, an alternative explanation may be small sample sizes in these previous proteomic age clock studies.

Our ProtAge clock also provides an advantage over many state-of-the-art DNAm clocks. It has been observed that the so-called ‘first generation’ DNAm clocks trained to predict age (for example, the Horvath clock) are only weakly associated with mortality risk and aging-related function 5 , 7 . To capture stronger associations with morbidity and mortality, newer DNAm clocks were subsequently developed that are not trained on chronological age itself, but rather are trained to predict composite phenotypic age variables that are usually weighted combinations of age and biological markers of morbidity (for example, albumin, creatinine and C-reactive protein) 23 , 24 , 27 . An advantage of our proteomic age clock is that it can be constructed by training to predict only age and still remain strongly associated with mortality and morbidity, and that this approach can be still translated to smaller independent samples (CKB and FinnGen) with very different study designs, participant genetic backgrounds and participant morbidity profiles.

One reason that DNAm-based clocks built to predict age may not predict functional and disease outcomes as well is the lack of strong correlation between age-related changes in gene expression (the functional consequence of DNAm) and age-related protein expression. Recent research using kidney and heart tissue in mice 28 , 29 , in addition to brain tissue from humans and rhesus macaques 30 , demonstrated that age-related changes in messenger RNA and protein levels were not strongly correlated. A key example of highly important proteins in our models whose abundance is not well correlated with mRNA are ECM proteins such as elastin (ELN) and collagens (COL6A3). These proteins have long half-lives that make them particularly susceptible to aging-related degradation and post-translational modifications, which contributes to the structural tissue damage encountered during aging 31 . Elastin fragmentation is a key contributor to vascular aging and is implicated in hypertension and cardiovascular outcomes 32 , although recent research also suggests a role of elastin degradation in CKD 33 and aging of cerebral arteries 34 . Additionally, functional research in Caenorhabditis   elegans suggests that ECM remodeling is required to promote longevity and that known genetic and pharmacological longevity interventions slow age-related collagen stiffening 35 . Our work provides new evidence from a human population study that ECM dynamics warrant consideration as an emerging hallmark of aging 36 , which may be difficult to capture with DNAm information.

Further, recent research in blood-derived human CD8 + T cell populations reported weak correlation between mRNA and protein abundance in immune cells 37 . Given the key role of several immune and inflammation proteins in our ProtAge model (CXCL17, LECT2, SCARF2 and GDF15) and the larger literature demonstrating the importance of immune-related inflammation in aging (inflammaging 38 ), this important axis of proteomic aging is also unlikely to be captured via DNAm or transcriptional age clocks.

As 50% of individuals with incident AD in our analysis were younger than 75 years at diagnosis, one of the advantages of our proteomic age clock may be detection or risk prediction of early onset AD. Notably, although recent reviews report that APOE (a gene strongly linked to AD risk) and FOXO3 (a transcription factor involved in apoptosis and DNA repair) are the only two genes whose associations with lifespan and longevity are consistently replicated across studies 2 , 39 , neither were found to be APs in our study. This highlights a possible disconnect between genetic determinants of lifespan and proteomic signals of aging, although it is also possible that tissue-specific expression of these proteins is not conducive to capturing a blood-based aging signal.

Our analyses also showed consistent associations between ProtAgeGap and four cancers (lung, prostate, esophageal and non-Hodgkin lymphoma) after covariate adjustment. Furthermore, various aspects of cancer development emerged in our analysis of ProtAge protein pathways. Some of the remaining nonassociated cancers had low or insufficient numbers of cases, indicating that we may have lacked statistical power to detect associations. For other cancers with adequate sample size (for example, colorectal cancer), it is possible that the proteomics panel used did not contain the relevant proteins for that cancer or that there is no reliable signature for these proteins in plasma.

Our study approach has many strengths, including our use of gradient-boosting (LightGBM) models that allow for nonlinear associations and account for interactions between all proteins. Further, our model benchmarking process shows that our gradient-boosting model provides substantially greater generalizability for estimating proteomic age in independent data compared to LASSO, elastic net or neural networks; however, our study also had several limitations. First, our model only used the Olink Explore 3072 assay, currently available in the UKB, and therefore did not capture all proteins covered in other platforms and panels, including the larger Olink HT (~5,000 proteins) or SOMAscan (>10,000 proteins) panels. Second, our datasets did not have DNAm data that would allow direct comparisons between proteomic and DNAm age clocks.

In summary, our study provides evidence that plasma proteomics is a powerful tool for measuring biological age and can be used to quantify a biological aging signature that is involved in most common age-related diseases in adult populations. Our work demonstrates that development of proteomic aging clocks can be used as a reliable tool to identify biological mechanisms involved in disease multimorbidity, and may serve as useful tools for identification of protein targets for possible drug treatment or lifestyle modification to reduce premature mortality and reduce or delay the onset of major age-related diseases and multimorbidity.

Study participants

The UKB is a prospective cohort study with extensive genetic and phenotype data available for 502,505 individuals resident in the United Kingdom who were recruited between 2006 and 2010 40 . The full UKB protocol is available online ( https://www.ukbiobank.ac.uk/media/gnkeyh2q/study-rationale.pdf ). We restricted our UKB sample to those participants with Olink Explore data available at baseline who were randomly sampled from the main UKB population ( n  = 45,441).

The CKB is a prospective cohort study of 512,724 adults aged 30–79 years who were recruited from ten geographically diverse (five rural and five urban) areas across China between 2004 and 2008. Details on the CKB study design and methods have been previously reported 41 . We restricted our CKB sample to those participants with Olink Explore data available at baseline in a nested case–cohort study of IHD and who were genetically unrelated to each other ( n  = 3,977).

The FinnGen study is a public–private partnership research project that has collected and analyzed genome and health data from 500,000 Finnish biobank donors to understand the genetic basis of diseases 42 . FinnGen includes nine Finnish biobanks, research institutes, universities and university hospitals, 13 international pharmaceutical industry partners and the Finnish Biobank Cooperative (FINBB). The project utilizes data from the nationwide longitudinal health register collected since 1969 from every resident in Finland. In FinnGen, we restricted our analyses to those participants with Olink Explore data available and passing proteomic data quality control ( n  = 1,990).

Proteomic profiling

Proteomic profiling in the UKB, CKB and FinnGen was carried out for protein analytes measured via the Olink Explore 3072 platform that links four Olink panels (Cardiometabolic, Inflammation, Neurology and Oncology). For all cohorts, the preprocessed Olink data were provided in the arbitrary NPX unit on a log 2 scale. In the UKB, the random subsample of proteomics participants ( n  = 45,441) were selected by removing those in batches 0 and 7. Randomized participants selected for proteomic profiling in the UKB have been shown previously to be highly representative of the wider UKB population 43 . UKB Olink data are provided as Normalized Protein eXpression (NPX) values on a log 2 scale, with details on sample selection, processing and quality control documented online.

In the CKB, stored baseline plasma samples from participants were retrieved, thawed and subaliquoted into multiple aliquots, with one (100 µl) aliquot used to make two sets of 96-well plates (40 µl per well). Both sets of plates were shipped on dry ice, one to the Olink Bioscience Laboratory at Uppsala (batch one, 1,463 unique proteins) and the other shipped to the Olink Laboratory in Boston (batch two, 1,460 unique proteins), for proteomic analysis using a multiplex proximity extension assay, with each batch covering all 3,977 samples. Samples were plated in the order they were retrieved from long-term storage at the Wolfson Laboratory in Oxford and normalized using both an internal control (extension control) and an inter-plate control and then transformed using a predetermined correction factor. The limit of detection (LOD) was determined using negative control samples (buffer without antigen). A sample was flagged as having a quality control warning if the incubation control deviated more than a predetermined value (±0.3) from the median value of all samples on the plate (but values below LOD were included in the analyses).

In the FinnGen study, blood samples were collected from healthy individuals and EDTA-plasma aliquots (230 µl) were processed and stored at −80 °C within 4 h. Plasma aliquots were subsequently thawed and plated in 96-well plates (120 µl per well) as per Olink’s instructions. Samples were shipped on dry ice to the Olink Bioscience Laboratory (Uppsala) for proteomic analysis using the 3,072 multiplex proximity extension assay. Samples were sent in three batches and to minimize any batch effects, bridging samples were added according to Olink’s recommendations. In addition, plates were normalized using both an internal control (extension control) and an inter-plate control and then transformed using a predetermined correction factor. The LOD was determined using negative control samples (buffer without antigen). A sample was flagged as having a quality control warning if the incubation control deviated more than a predetermined value (±0.3) from the median value of all samples on the plate (but values below LOD were included in the analyses).

We excluded from analysis any proteins not available in all three cohorts, as well as an additional three proteins that were missing in over 10% of the UKB sample (CTSS, PCOLCE and NPM1), leaving a total of 2,897 proteins for analysis. After missing data imputation (see below), proteomic data were normalized separately within each cohort by first rescaling values to be between 0 and 1 using MinMaxScaler() from scikit-learn and then centering on the median.

UKB aging biomarkers were measured using baseline nonfasting blood serum samples as previously described 44 . Biomarkers were previously adjusted for technical variation by the UKB, with sample processing ( https://biobank.ndph.ox.ac.uk/showcase/showcase/docs/serum_biochemistry.pdf ) and quality control ( https://biobank.ndph.ox.ac.uk/showcase/ukb/docs/biomarker_issues.pdf ) procedures described on the UKB website. Field IDs for all biomarkers and measures of physical and cognitive function are shown in Supplementary Table 18 . Poor self-rated health, slow walking pace, self-rated facial aging, feeling tired/lethargic every day and frequent insomnia were all binary dummy variables coded as all other responses versus responses for ‘Poor’ (overall health rating; field ID 2178), ‘Slow pace’ (usual walking pace; field ID 924), ‘Older than you are’ (facial aging; field ID 1757), ‘Nearly every day’ (frequency of tiredness/lethargy in last 2 weeks; field ID 2080) and ‘Usually’ (sleeplessness/insomnia; field ID 1200), respectively. Sleeping 10+ hours per day was coded as a binary variable using the continuous measure of self-reported sleep duration (field ID 160). Systolic and diastolic blood pressure were averaged across both automated readings. Standardized lung function (FEV 1 ) was calculated by dividing the FEV 1 best measure (field ID 20150) by standing height squared (field ID 50). Hand grip strength variables (field ID 46,47) were divided by weight (field ID 21002) to normalize according to body mass. Frailty index was calculated using the algorithm previously developed for UKB data by Williams et al. 21 . Components of the frailty index are shown in Supplementary Table 19 . Leukocyte telomere length was measured as the ratio of telomere repeat copy number (T) relative to that of a single copy gene (S; HBB , which encodes human hemoglobin subunit β) 45 . This T:S ratio was adjusted for technical variation and then both log-transformed and z-standardized using the distribution of all individuals with a telomere length measurement.

Detailed information about the linkage procedure ( https://biobank.ctsu.ox.ac.uk/crystal/refer.cgi?id=115559 ) with national registries for mortality and cause of death information in the UKB is available online. Mortality data were accessed from the UKB data portal on 23 May 2023, with a censoring date of 30 November 2022 for all participants (12–16 years of follow-up).

Data used to define prevalent and incident chronic diseases in the UKB are outlined in Supplementary Table 20 . In the UKB, incident cancer diagnoses were ascertained using International Classification of Diseases (ICD) diagnosis codes and corresponding dates of diagnosis from linked cancer and mortality register data. Incident diagnoses for all other diseases were ascertained using ICD diagnosis codes and corresponding dates of diagnosis taken from linked hospital inpatient, primary care and death register data. Primary care read codes were converted to corresponding ICD diagnosis codes using the lookup table provided by the UKB. Linked hospital inpatient, primary care and cancer register data were accessed from the UKB data portal on 23 May 2023, with a censoring date of 31 October 2022; 31 July 2021 or 28 February 2018 for participants recruited in England, Scotland or Wales, respectively (8–16 years of follow-up).

In the CKB, information about incident disease and cause-specific mortality was obtained by electronic linkage, via the unique national identification number, to established local mortality (cause-specific) and morbidity (for stroke, IHD, cancer and diabetes) registries and to the health insurance system that records any hospitalization episodes and procedures 41 , 46 . All disease diagnoses were coded using the ICD-10, blinded to any baseline information, and participants were followed up to death, loss-to-follow-up or 1 January 2019. ICD-10 codes used to define diseases studied in the CKB are shown in Supplementary Table 21 .

Missing data imputation

Missing values for all nonproteomics UKB data were imputed using the R package missRanger 47 , which combines random forest imputation with predictive mean matching. We imputed a single dataset using a maximum of ten iterations and 200 trees. All other random forest hyperparameters were left at default values. The imputation dataset included all baseline variables available in the UKB as predictors for imputation, excluding variables with any nested response patterns. Responses of ‘do not know’ were set to ‘NA’ and imputed. Responses of ‘prefer not to answer’ were not imputed and set to NA in the final analysis dataset. Age and incident health outcomes were not imputed in the UKB. CKB data had no missing values to impute.

Protein expression values were imputed in the UKB and FinnGen cohort using the miceforest package in Python. All proteins except those missing in >30% of participants were used as predictors for imputation of each protein. We imputed a single dataset using a maximum of five iterations. All other parameters were left at default values.

Calculation of chronological age measures

In the UKB, age at recruitment (field ID 21022) is only provided as a whole integer value. We derived a more accurate estimate by taking month of birth (field ID 52) and year of birth (field ID 34) and creating an approximate date of birth for each participant as the first day of their birth month and year. Age at recruitment as a decimal value was then calculated as the number of days between each participant’s recruitment date (field ID 53) and approximate birth date divided by 365.25. Age at the first imaging follow-up (2014+) and the repeat imaging follow-up (2019+) were then calculated by taking the number of days between the date of each participant’s follow-up visit and their initial recruitment date divided by 365.25 and adding this to age at recruitment as a decimal value. Recruitment age in the CKB is already provided as a decimal value.

Model benchmarking

We compared the performance of six different machine-learning models (LASSO, elastic net, LightGBM and three neural network architectures: multilayer perceptron, a residual feedforward network (ResNet) and a retrieval-augmented neural network for tabular data (TabR)) for using plasma proteomic data to predict age. For each model, we trained a regression model using all 2,897 Olink protein expression variables as input to predict chronological age. All models were trained using fivefold cross-validation in the UKB training data ( n  = 31,808) and were tested against the UKB holdout test set ( n  = 13,633), as well as independent validation sets from the CKB and FinnGen cohorts. We found that LightGBM provided the second-best model accuracy among the UKB test set, but showed markedly better performance in the independent validation sets (Supplementary Fig. 1 ).

LASSO and elastic net models were calculated using the scikit-learn package in Python. For the LASSO model, we tuned the alpha parameter using the LassoCV function and an alpha parameter space of [1 × 10 −15 , 1 × 10 −10 , 1 × 10 −8 , 1 × 10 −5 , 1 × 10 −4 , 1 × 10 −3 , 1 × 10 −2 , 1, 5, 10, 50 and 100]. Elastic net models were tuned for both alpha (using the same parameter space) and L1 ratio drawn from the following possible values: [0.1, 0.5, 0.7, 0.9, 0.95, 0.99 and 1].

The LightGBM model hyperparameters were tuned via fivefold cross-validation using the Optuna module in Python 48 , with parameters tested across 200 trials and optimized to maximize the average R 2 of the models across all folds.

The neural network architectures tested in this analysis were selected from a list of architectures that performed well on a variety of tabular datasets. The architectures considered were (1) a multilayer perceptron; (2) ResNet; and (3) TabR. All neural network model hyperparameters were tuned via fivefold cross-validation using Optuna across 100 trials and optimized to maximize the average R 2 of the models across all folds.

Calculation of ProtAge

Using gradient boosting (LightGBM) as our selected model type, we initially ran models trained separately on males and females; however, the male- and female-only models showed similar age prediction performance to a model with both sexes (Supplementary Fig. 8a–c ) and protein-predicted age from the sex-specific models were nearly perfectly correlated with protein-predicted age from the model using both sexes (Supplementary Fig. 8d,e ). We further found that when looking at the most important proteins in each sex-specific model, there was a large consistency across males and females. Specifically, 11 of the top 20 most important proteins for predicting age according to SHAP values were shared across males and females and all 11 shared proteins showed consistent directions of effect for males and females (Supplementary Fig. 9a,b ; ELN, EDA2R, LTBP2, NEFL, CXCL17, SCARF2, CDCP1, GFAP, GDF15, PODXL2 and PTPRR). We therefore calculated our proteomic age clock in both sexes combined to improve the generalizability of the findings.

To calculate proteomic age, we first split all UKB participants ( n  = 45,441) into 70:30 train–test splits. In the training data ( n  = 31,808), we trained a model to predict age at recruitment using all 2,897 proteins in a single LightGBM 18 model. First, model hyperparameters were tuned via fivefold cross-validation using the Optuna module in Python 48 , with parameters tested across 200 trials and optimized to maximize the average R 2 of the models across all folds. We then carried out Boruta feature selection via the SHAP-hypetune module. Boruta feature selection works by making random permutations of all features in the model (called shadow features), which are essentially random noise 19 . In our use of Boruta, at each iterative step these shadow features were generated and a model was run with all features and all shadow features. We then removed all features that did not have a mean of the absolute SHAP value that was higher than all random shadow features. The selection processes ended when there were no features remaining that did not perform better than all shadow features. This procedure identifies all features relevant to the outcome that have a greater influence on prediction than random noise. When running Boruta, we used 200 trials and a threshold of 100% to compare shadow and real features (meaning that a real feature is selected if it performs better than 100% of shadow features). Third, we re-tuned model hyperparameters for a new model with the subset of selected proteins using the same procedure as before. Both tuned LightGBM models before and after feature selection were checked for overfitting and validated by performing fivefold cross-validation in the combined train set and testing the performance of the model against the holdout UKB test set. Across all analysis steps, LightGBM models were run with 5,000 estimators, 20 early stopping rounds and using R 2 as a custom evaluation metric to identify the model that explained the maximum variation in age (according to R 2 ).

Once the final model with Boruta-selected APs was trained in the UKB, we calculated protein-predicted age (ProtAge) for the entire UKB cohort ( n  = 45,441) using fivefold cross-validation. Within each fold, a LightGBM model was trained using the final hyperparameters and predicted age values were generated for the test set of that fold. We then combined the predicted age values from each of the folds to create a measure of ProtAge for the entire sample. ProtAge was calculated in the CKB and FinnGen by using the trained UKB model to predict values in those datasets. Finally, we calculated proteomic aging gap (ProtAgeGap) separately in each cohort by taking the difference of ProtAge minus chronological age at recruitment separately in each cohort.

Recursive feature elimination using SHAP

For our recursive feature elimination analysis, we started from the 204 Boruta-selected proteins. In each step, we trained a model using fivefold cross-validation in the UKB training data and then within each fold calculated the model R 2 and the contribution of each protein to the model as the mean of the absolute SHAP values across all participants for that protein. R 2 values were averaged across all five folds for each model. We then removed the protein with the smallest mean of the absolute SHAP values across the folds and computed a new model, eliminating features recursively using this method until we reached a model with only five proteins. If at any step of this process a different protein was identified as the least important in the different cross-validation folds, we chose the protein ranked the lowest across the greatest number of folds to remove. We identified 20 proteins as the smallest number of proteins that provide adequate prediction of chronological age, as fewer than 20 proteins resulted in a dramatic drop in model performance (Supplementary Fig. 3d ). We re-tuned hyperparameters for this 20-protein model (ProtAge20) using Optuna according to the methods described above, and we also calculated the proteomic age gap according to these top 20 proteins (ProtAgeGap20) using fivefold cross-validation in the entire UKB cohort ( n  = 45,441) using the methods described above.

Statistical analysis

All statistical analyses were carried out using Python v.3.6 and R v.4.2.2. All associations between ProtAgeGap and aging biomarkers and physical/cognitive function measures in the UKB were tested using linear/logistic regression using the statsmodels module 49 . All models were adjusted for age, sex, Townsend deprivation index, assessment center, self-reported ethnicity (Black, white, Asian, mixed and other), IPAQ activity group (low, moderate and high) and smoking status (never, previous and current). P values were corrected for multiple comparisons via the FDR using the Benjamini–Hochberg method 50 .

All associations between ProtAgeGap and incident outcomes (mortality and 26 diseases) were tested using Cox proportional hazards models using the lifelines module 51 . Survival outcomes were defined using follow-up time to event and the binary incident event indicator. For all incident disease outcomes, prevalent cases were excluded from the dataset before models were run. For all incident outcome Cox modeling in the UKB, three successive models were tested with increasing numbers of covariates. Model 1 included adjustment for age at recruitment and sex. Model 2 included all model 1 covariates, plus Townsend deprivation index (field ID 22189), assessment center (field ID 54), physical activity (IPAQ activity group; field ID 22032) and smoking status (field ID 20116). Model 3 included all model 3 covariates plus BMI (field ID 21001) and prevalent hypertension (defined in Supplementary Table 20 ). P values were corrected for multiple comparisons via FDR.

Functional enrichments (GO biological processes, GO molecular function, KEGG and Reactome) and PPI networks were downloaded from STRING (v.12) using the STRING API in Python. For functional enrichment analyses, we used all proteins included in the Olink Explore 3072 platform as the statistical background (except for 19 Olink proteins that could not be mapped to STRING IDs. None of the proteins that could not be mapped were included in our final Boruta-selected proteins). We only considered PPIs from STRING at a high level of confidence (>0.7) from the coexpression data.

SHAP interaction values from the trained LightGBM ProtAge model were retrieved using the SHAP module 20 , 52 . SHAP-based PPI networks were generated by first taking the mean of the absolute value of each protein–protein SHAP interaction score across all samples. We then used an interaction threshold of 0.0083 and removed all interactions below this threshold, which yielded a subset of variables similar in number to the node degree >2 threshold used for the STRING PPI network. Both SHAP-based and STRING 53 -based PPI networks were visualized and plotted using the NetworkX module 54 .

Cumulative incidence curves and survival tables for deciles of ProtAgeGap were calculated using KaplanMeierFitter from the lifelines module. As our data were right-censored, we plotted cumulative events against age at recruitment on the x axis. All plots were generated using matplotlib 55 and seaborn 56 . The total fold risk of disease according to the top and bottom 5% of the ProtAgeGap was calculated by raising the HR for the disease by the total number of years comparison (12.3 years average ProtAgeGap difference between the top versus bottom 5% and 6.3 years average ProtAgeGap between the top 5% versus those with 0 years of ProtAgeGap).

Ethics approval

UKB data use (project application no. 61054) was approved by the UKB according to their established access procedures. UKB has approval from the North West Multi-centre Research Ethics Committee as a research tissue bank and as such researchers using UKB data do not require separate ethical clearance and can operate under the research tissue bank approval. The CKB complies with all the required ethical standards for medical research on human participants. Ethical approvals were granted and have been maintained by the relevant institutional ethical research committees in the United Kingdom and China. Study participants in FinnGen provided informed consent for biobank research, based on the Finnish Biobank Act. The FinnGen study is approved by the Finnish Institute for Health and Welfare (permit nos. THL/2031/6.02.00/2017, THL/1101/5.05.00/2017, THL/341/6.02.00/2018, THL/2222/6.02.00/2018, THL/283/6.02.00/2019, THL/1721/5.05.00/2019 and THL/1524/5.05.00/2020), Digital and Population Data Service Agency (permit nos. VRK43431/2017-3, VRK/6909/2018-3 and VRK/4415/2019-3), the Social Insurance Institution (permit nos. KELA 58/522/2017, KELA 131/522/2018, KELA 70/522/2019, KELA 98/522/2019, KELA 134/522/2019, KELA 138/522/2019, KELA 2/522/2020 and KELA 16/522/2020), Findata (permit nos. THL/2364/14.02/2020, THL/4055/14.06.00/2020, THL/3433/14.06.00/2020, THL/4432/14.06/2020, THL/5189/14.06/2020, THL/5894/14.06.00/2020, THL/6619/14.06.00/2020, THL/209/14.06.00/2021, THL/688/14.06.00/2021, THL/1284/14.06.00/2021, THL/1965/14.06.00/2021, THL/5546/14.02.00/2020, THL/2658/14.06.00/2021 and THL/4235/14.06.00/2021), Statistics Finland (permit nos. TK-53-1041-17 and TK/143/07.03.00/2020 (previously TK-53-90-20) TK/1735/07.03.00/2021 and TK/3112/07.03.00/2021) and Finnish Registry for Kidney Diseases permission/extract from the meeting minutes on 4 July 2019.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

UKB data are available through a procedure described at https://www.ukbiobank.ac.uk/enable-your-research . The CKB is a global resource for the investigation of lifestyle, environmental, blood biochemical and genetic factors as determinants of common diseases. The CKB study group is committed to making the cohort data available to the scientific community in China, the United Kingdom and worldwide to advance knowledge about the causes, prevention and treatment of disease. For detailed information on what data are currently available to open-access users and how to apply for them, please visit https://www.ckbiobank.org/data-access . A research proposal will be requested to ensure that any analysis is performed by bona fide researchers. Researchers who are interested in obtaining additional information or data that underline this paper should contact [email protected]. For any data that are not currently available via open access, researchers may need to develop a formal collaboration with the CKB study group. FinnGen data can be accessed through Fingenious services ( https://site.fingenious.fi/en/ ) managed by FINBB. Finnish Health Register data can be applied for from Findata ( https://findata.fi/en/data/ ). Experimental protein–protein interaction information used from the STRING database (v.12) can be accessed programmatically using the STRING API ( https://string-db.org/ ) or can be downloaded directly from the STRING website ( https://string-db.org/cgi/download.pl ).

Code availability

R and Python code used for data preparation and analysis can be found at https://github.com/miargentieri/proteomic-age-ukb .

Niccoli, T. & Partridge, L. Ageing as a risk factor for disease. Curr. Biol. 22 , R741–R752 (2012).

CAS   PubMed   Google Scholar  

Partridge, L., Deelen, J. & Slagboom, P. E. Facing up to the global challenges of ageing. Nature 561 , 45–56 (2018).

Chang, A. Y., Skirbekk, V. F., Tyrovolas, S., Kassebaum, N. J. & Dieleman, J. L. Measuring population ageing: an analysis of the Global Burden of Disease Study 2017. Lancet Public Health 4 , e159–e167 (2019).

PubMed   PubMed Central   Google Scholar  

Laconi, E., Marongiu, F. & DeGregori, J. Cancer as a disease of old age: changing mutational and microenvironmental landscapes. Br. J. Cancer 122 , 943–952 (2020).

Rutledge, J., Oh, H. & Wyss-Coray, T. Measuring biological age using omics data. Nat. Rev. Genet. 23 , 715–727 (2022).

CAS   PubMed   PubMed Central   Google Scholar  

Langenberg, C., Hingorani, A. D. & Whitty, C. J. M. Biological and functional multimorbidity-from mechanisms to management. Nat. Med. 29 , 1649–1657 (2023).

Horvath, S. & Raj, K. DNA methylation-based biomarkers and the epigenetic clock theory of ageing. Nat. Rev. Genet. 19 , 371–384 (2018).

Joehanes, R. et al. Epigenetic signatures of cigarette smoking. Circ. Cardiovasc. Genet 9 , 436–447 (2016).

López-Otín, C., Blasco, M. A., Partridge, L., Serrano, M. & Kroemer, G. Hallmarks of aging: an expanding universe. Cell 186 , 243–278 (2023).

PubMed   Google Scholar  

Coenen, L., Lehallier, B., de Vries, H. E. & Middeldorp, J. Markers of aging: unsupervised integrated analyses of the human plasma proteome. Front. Aging 4 , 1112109 (2023).

Johnson, A. A., Shokhirev, M. N., Wyss-Coray, T. & Lehallier, B. Systematic review and analysis of human proteomics aging studies unveils a novel proteomic aging clock and identifies key processes that change with age. Ageing Res. Rev. 60 , 101070 (2020).

Lehallier, B. et al. Undulating changes in human plasma proteome profiles across the lifespan. Nat. Med. 25 , 1843–1850 (2019).

Lehallier, B., Shokhirev, M. N., Wyss-Coray, T. & Johnson, A. A. Data mining of human plasma proteins generates a multitude of highly predictive aging clocks that reflect different aspects of aging. Aging Cell 19 , e13256 (2020).

Oh, H. S. et al. Organ aging signatures in the plasma proteome track health and disease. Nature 624 , 164–172 (2023).

Sayed, N. et al. An inflammatory aging clock (iAge) based on deep learning tracks multimorbidity, immunosenescence, frailty and cardiovascular aging. Nat. Aging 1 , 598–615 (2021).

Tanaka, T. et al. Plasma proteomic biomarker signature of age predicts health and life span. eLife https://doi.org/10.7554/eLife.61073 (2020).

Macdonald-Dunlop, E. et al. A catalogue of omics biological ageing clocks reveals substantial commonality and associations with disease risk. Aging 14 , 623–659 (2022).

Ke, G. et al. LightGBM: A Highly Efficient Gradient Boosting Decision Tree. In Advances in Neural Information Processing Systems Vol. 30 (eds Guyon, I. et al.) 3149–3157 (NeurIPS, 2017).

Kursa, M. B., Jankowski, A., Rudnicki, W. R. & Boruta, A. System for feature selection. Fundam. Inform. 101 , 271–285 (2010).

Google Scholar  

Lundberg, S. M. et al. From local explanations to global understanding with explainable AI for trees. Nat. Mach. Intell. 2 , 56–67 (2020).

Williams, D. M., Jylhävä, J., Pedersen, N. L. & Hägg, S. A frailty index for UK Biobank participants. J. Gerontol. A Biol. Sci. Med. Sci. 74 , 582–587 (2019).

Horvath, S. DNA methylation age of human tissues and cell types. Genome Biol. 14 , R115 (2013).

Levine, M. E. et al. An epigenetic biomarker of aging for lifespan and healthspan. Aging 10 , 573–591 (2018).

Belsky, D. W. et al. DunedinPACE, a DNA methylation biomarker of the pace of aging. eLife https://doi.org/10.7554/eLife.73420 (2022).

Kabacik, S. et al. The relationship between epigenetic age and the hallmarks of aging in human cells. Nat. Aging 2 , 484–493 (2022).

Pietzner, M. et al. Synergistic insights into human health from aptamer- and antibody-based proteomic profiling. Nat. Commun. 12 , 6822 (2021).

Lu, A. T. et al. DNA methylation GrimAge strongly predicts lifespan and healthspan. Aging 11 , 303–327 (2019).

Takemon, Y. et al. Proteomic and transcriptomic profiling reveal different aspects of aging in the kidney. eLife https://doi.org/10.7554/eLife.62585 (2021).

Gerdes Gyuricza, I. et al. Genome-wide transcript and protein analysis highlights the role of protein homeostasis in the aging mouse heart. Genome Res. 32 , 838–852 (2022).

Wei, Y. N. et al. Transcript and protein expression decoupling reveals RNA binding proteins and miRNAs as potential modulators of human aging. Genome Biol. 16 , 41 (2015).

Gorisse, L. et al. Protein carbamylation is a hallmark of aging. Proc. Natl Acad. Sci. USA 113 , 1191–1196 (2016).

Heinz, A. Elastic fibers during aging and disease. Ageing Res. Rev. 66 , 101255 (2021).

Sun, Q. et al. Elastin imaging enables noninvasive staging and treatment monitoring of kidney fibrosis. Sci. Transl. Med . https://doi.org/10.1126/scitranslmed.aat4865 (2019).

Fonck, E. et al. Effect of aging on elastin functionality in human cerebral arteries. Stroke 40 , 2552–2556 (2009).

Teuscher, A. C. et al. Longevity interventions modulate mechanotransduction and extracellular matrix homeostasis in C. elegans . Nat. Commun. 15 , 276 (2024).

Statzer, C., Park, J. Y. C. & Ewald, C. Y. Extracellular matrix dynamics as an emerging yet understudied hallmark of aging and longevity. Aging Dis. 14 , 670–693 (2023).

Nicolet, B. P. & Wolkers, M. C. The relationship of mRNA with protein expression in CD8 + T cells associates with gene class and gene characteristics. PLoS ONE 17 , e0276294 (2022).

Franceschi, C., Garagnani, P., Parini, P., Giuliani, C. & Santoro, A. Inflammaging: a new immune-metabolic viewpoint for age-related diseases. Nat. Rev. Endocrinol. 14 , 576–590 (2018).

Morris, B. J., Willcox, D. C., Donlon, T. A. & Willcox, B. J. FOXO3: a major gene for human longevity–a mini-review. Gerontology 61 , 515–525 (2015).

Sudlow, C. et al. UK Biobank: an open access resource for identifying the causes of a wide range of complex diseases of middle and old age. PLoS Med. 12 , e1001779 (2015).

Chen, Z. et al. China Kadoorie Biobank of 0.5 million people: survey methods, baseline characteristics and long-term follow-up. Int J. Epidemiol. 40 , 1652–1666 (2011).

Kurki, M. I. et al. FinnGen provides genetic insights from a well-phenotyped isolated population. Nature 613 , 508–518 (2023).

Sun, B. B. et al. Plasma proteomic associations with genetics and health in the UK Biobank. Nature 622 , 329–338 (2023).

Elliott, P. & Peakman, T. C. The UK Biobank sample handling and storage protocol for the collection, processing and archiving of human blood and urine. Int. J. Epidemiol. 37 , 234–244 (2008).

Codd, V. et al. Measurement and initial characterization of leukocyte telomere length in 474,074 participants in UK Biobank. Nat. Aging 2 , 170–179 (2022).

Chen, Z. et al. Cohort profile: the Kadoorie Study of Chronic Disease in China (KSCDC). Int. J. Epidemiol. 34 , 1243–1249 (2005).

Mayer, M. missRanger: Fast Imputation of Missing Values. R package version 2.1.0. https://doi.org/10.32614/CRAN.package.missRanger (2019).

Akiba, T., Sano, S., Yanase, T., Ohta, T. & Koyama, M. Optuna: a next-generation hyperparameter optimization framework. In Proc. 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining 2623–2631 (2019); https://doi.org/10.1145/3292500.3330701

Skipper, S. & Perktold, J. Statsmodels: econometric and statistical modeling with Python. In Proc. 9th Python in Science Conference (eds van der Walt, S. & Millman, J.) 92–96 (2010).

Benjamini, Y. & Hochberg, Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. R. Stat. Soc. Ser. B 57 , 289–300 (1995).

Davidson-Pilon, C. lifelines, survival analysis in Python. Zenodo https://doi.org/10.5281/zenodo.7883870 (2023).

Lundberg, S. M. & Lee, S.-I. A unified approach to interpreting model predictions. Adv. Neural Inf. Process. Syst. 30 , 4765–4774 (2017).

Szklarczyk, D. et al. STRING v10: protein–protein interaction networks, integrated over the tree of life. Nucleic Acids Res. 43 , D447–D452 (2015).

Hagberg, A. A., Schult, D. A. & Swart, P. J. Exploring network structure, dynamics, and function using NetworkX. In Proc. 7th Python in Science Conference (SciPy 2008) (eds Varoquaux, G., Vaught, T. & Millman, J.) 11–15 (2008).

Hunter, J. D. Matplotlib: a 2D graphics environment. Comput. Sci. Eng. 9 , 90–95 (2007).

Waskom, M. L. seaborn: statistical data visualization. J. Open Source Softw. 6 , 3021 (2021).

Download references

Acknowledgements

We acknowledge the UKB, CKB and FinnGen participants for their dedication to participating in ongoing research and electronic health record linkage. All UKB data were accessed under UKB application no. 61054. This work was supported by King Abdulaziz University and the University of Oxford Centre for Artificial Intelligence in Precision Medicine (KO-CAIPM). We also acknowledge the CKB project staff and the China CDC and its regional offices for assisting with CKB fieldwork. We thank J. Mackay in Hong Kong; Y. Wang, G. Yang, Z. Qiang, L. Feng, M. Zhou, W. Zhao and Y. Zhang in the China CDC; L. Kong, X. Yu and K. Li in the Chinese Ministry of Health; and S. Clark, M. Radley and M. Hill in the CTSU, Oxford, for assisting with the planning, conduct and organization of the CKB study. We also thank J. Juvila and J. Honkanen from the Finnish Red Cross Blood Service for their work carrying out blood sample collection. Finally, we thank P. Block for his clinical input on liver enzymes and biomarkers. S.X., A.J.N.-H., A.A., C.J.A. and C.M.vD. are funded by the KO-CAIPM. A.J.N.-H. receives research funding from Novo Nordisk, GSK and Ono Pharma. C.M.vD. is further supported by Oxford-GSK Institute of Molecular and Computational Medicine and the Novo Nordisk – Oxford Fellowship Programme, the common mechanisms and pathways in Stroke and Alzheimer’s disease (CoSTREAM) project ( www.costream.eu , grant agreement no. 667375) and ZonMW Memorabel program (project no. 733050814). L.W. is supported by Alzheimer’s Research UK. C.J.A. is supported by the Oxford National Institute for Health and Care Research (NIHR) Biomedical Research Centre (BRC). The CKB baseline survey and the first re-survey were supported by the Kadoorie Charitable Foundation in Hong Kong. The long-term follow-up and subsequent CKB resurveys were supported by Wellcome grants to Oxford University (212946/Z/18/Z, 202922/Z/16/Z, 104085/Z/14/Z and 088158/Z/09/Z) and grants from the National Natural Science Foundation of China (82192901, 82192904 and 82192900) and from the National Key Research and Development Program of China (2016YFC0900500). The UK Medical Research Council (MC_UU_00017/1, MC_UU_12026/2 and MC_U137686851), Cancer Research UK (C16077/A29186 and C500/A16896) and the British Heart Foundation (CH/1996001/9454) provide core funding to the Clinical Trial Service Unit and Epidemiological Studies Unit at Oxford University for the project. The CKB proteomic assays were supported by the British Heart Fooundation (18/23/33512), Novo Nordisk and OLINK. CKB DNA extraction and genotyping were supported by GlaxoSmithKline and the UK Medical Research Council (MC-PC-13049 and MC-PC-14135). The FinnGen project is funded by two grants from Business Finland (HUS 4685/31/2016 and UH 4386/31/2016) and the following industry partners: AbbVie, AstraZeneca UK, Biogen MA, Bristol Myers Squibb (and Celgene Corporation & Celgene International II), Genentech, Merck Sharp & Dohme, Pfizer, GlaxoSmithKline Intellectual Property Development, Sanofi US Services, Maze Therapeutics, Janssen Biotech, Novartis Pharma and Boehringer Ingelheim International. The following biobanks are acknowledged for delivering biobank samples to FinnGen: Auria Biobank ( www.auria.fi/biopankki ), THL Biobank ( www.thl.fi/biobank ), Helsinki Biobank ( www.helsinginbiopankki.fi ), Biobank Borealis of Northern Finland ( https://www.ppshp.fi/Tutkimus-ja-opetus/Biopankki/Pages/Biobank-Borealis-briefly-in-English.aspx ), Finnish Clinical Biobank Tampere ( www.tays.fi/en-US/Research_and_development/Finnish_Clinical_Biobank_Tampere ), Biobank of Eastern Finland ( www.ita-suomenbiopankki.fi/en ), Central Finland Biobank ( www.ksshp.fi/fi-FI/Potilaalle/Biopankki ), Finnish Red Cross Blood Service Biobank ( www.veripalvelu.fi/verenluovutus/biopankkitoiminta ), Terveystalo Biobank ( www.terveystalo.com/fi/Yritystietoa/Terveystalo-Biopankki/Biopankki/ ) and Arctic Biobank ( https://www.oulu.fi/en/university/faculties-and-units/faculty-medicine/northern-finland-birth-cohorts-and-arctic-biobank ). All Finnish Biobanks are members of BBMRI.fi infrastructure ( http://www.bbmri.fi/ ). FINBB ( https://finbb.fi/ ) is the coordinator of BBMRI-ERIC operations in Finland. The computational aspects of this research were supported by the Wellcome Trust Core Award grant no. 203141/Z/16/Z and the Oxford NIHR BRC. The views expressed are those of the authors and not necessarily those of the UK NHS, NIHR or Department of Health.

Author information

These authors jointly supervised this work: Zhengming Chen, Cornelia M. van Duijn.

Authors and Affiliations

Nuffield Department of Population Health, University of Oxford, Oxford, UK

M. Austin Argentieri, Sihao Xiao, Derrick Bennett, Pang Yao, Mohsen Mazidi, Iona Millwood, Hannah Fry, Robert Clarke, Najaf Amin, Zhengming Chen & Cornelia M. van Duijn

Analytic and Translational Genetics Unit, Massachusetts General Hospital, Boston, MA, USA

M. Austin Argentieri, Zhili Zheng, Mitja Kurki & Mark J. Daly

Program in Medical and Population Genetics, Broad Institute of MIT and Harvard, Boston, MA, USA

M. Austin Argentieri & Mark J. Daly

King Abdulaziz University and the University of Oxford Centre for Artificial Intelligence in Precision Medicine (KO-CAIPM), Jeddah, Saudi Arabia

Sihao Xiao, Laura Winchester, Alejo J. Nevado-Holgado, Upamanyu Ghose, Ashwag Albukhari, Cassandra J. Adams & Cornelia M. van Duijn

Department of Psychiatry, University of Oxford, Oxford, UK

Laura Winchester, Alejo J. Nevado-Holgado & Upamanyu Ghose

Biochemistry Department, Faculty of Science, King Abdulaziz University, Jeddah, Saudi Arabia

Ashwag Albukhari

Department of Epidemiology and Biostatistics, School of Public Health, Peking University, Beijing, China

Jun Lv & Liming Li

Peking University Center for Public Health and Epidemic Preparedness and Response, Beijing, China

Key Laboratory of Epidemiology of Major Diseases (Peking University), Ministry of Education, Beijing, China

Institute for Molecular Medicine Finland (FIMM), HiLIFE, University of Helsinki, Helsinki, Finland

Rodosthenis S. Rodosthenous & Aarno Palotie

Research and Development, Finnish Red Cross Blood Service, Helsinki, Finland

Jukka Partanen

Centre for Medicines Discovery, Nuffield Department of Medicine, University of Oxford, Oxford, UK

Cassandra J. Adams

You can also search for this author in PubMed   Google Scholar

Contributions

M.A.A., C.M.vD., N.A., S.X., D.B. and Z.C. conceptualized the study. M.A.A. performed all analysis and data visualization, as well as all UKB data curation. U.G. coded and analyzed neural network models. Additional CKB data curation was performed by P.Y., M.M., I.M., H.F. and D.B. and collection of CKB data was facilitated by L.L., J.L. and Z.C. FinnGen blood data collection was coordinated by J.K. and R.S.R. FinnGen proteomic data curation was performed by Z.Z., R.S.R. and M.K. and overseen by M.J.D. and A.P. Data curation and formal analyses were supervised by C.M.vD., N.A., D.B. and Z.C. Analytical and interpretational input for all analyses was provided by L.W., A.J.N.-H., P.Y., M.M. and R.C. M.A.A. prepared the paper, figures, tables and supplementary files, with edits and revisions provided by all other authors. The GitHub code repository was created and is maintained by M.A.A.

Corresponding authors

Correspondence to M. Austin Argentieri or Cornelia M. van Duijn .

Ethics declarations

Competing interests.

The authors declare no competing interests.

Peer review

Peer review information.

Nature Medicine thanks David Furman, Valur Emilsson, Xun Xu and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editor: Michael Basson, in collaboration with the Nature Medicine team.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data

Extended data fig. 1 stability of protage protein associations with age across three time points..

Comparison of betas for the association between age and each of the 149 ProtAge APs with repeat measurements available during baseline and two follow-up imaging visits (n = 1,085). a) Comparison of betas for the association between age and each of the 149 ProtAge APs during baseline and the 2014+ follow-up imaging visit. b) Comparison of betas for the association between each of these 149 ProtAge APs and age during baseline and the 2019+ imaging visit. c) Comparison of betas for the association between each of the 149 ProtAge APs and age during the 2014+ imaging visit and during the 2019+ imaging visit. Shown in each plot are the Pearson correlation coefficient (r), p-value for the correlation, and the model slope (λ). APs: aging-related proteins.

Extended Data Fig. 2 ProtAgeGap and age-specific cancer risk trajectories in the UKB.

Cumulative incidence plots for the top, median, and bottom deciles of ProtAgeGap in the UK Biobank (UKB; n = 45,441). Number of incident cases are shown for each cancer – these numbers reflect the total number of incident cases present only among those in the 3 deciles shown, not the full dataset. Incidence rates are for the 11-16 years after recruitment. Incidence rates are shown for the subsequent 11-16 years of follow up after recruitment for each given age at recruitment (for example, the cumulative incidence rate shown at age 65 is the rate of incident cases in the 11-16 years of follow up for those aged 65 at recruitment). All plots show the cumulative density of events at a given timepoint based on the Kaplan-Meier survival function, with 95% confidence intervals in lighter shading. Certain plots show multiple 0s on the y-axis because these represent decimal values < 0.5 and the y-axis values are rounded to a single digit. ProtAgeGap: proteomic age gap (in years).

Extended Data Fig. 3 Associations between ProtAgeGap and cancers in the UKB.

Associations between ProtAgeGap and and incident cancer diagnoses in Cox proportional hazards models with increasing levels of covariate adjustment. Shown on the x-axis are hazard ratios (and 95% confidence intervals) for the effect of ProtAgeGap on the outcomes shown. Events listed are the total number of incident cases for each outcome. Within each model, p-values across tests for all outcomes were corrected for multiple comparisons using the false discovery rate (FDR). All models were run in the UK Biobank (UKB; n = 45,441). a) . Model 1 is adjusted for age and sex. b) Model 2 is adjusted for age, sex, Townsend deprivation index, recruitment centre, IPAQ activity group, and smoking status. c) Model 3 is adjusted for age, sex, Townsend deprivation index, recruitment centre, IPAQ activity group, smoking status, BMI, and prevalent hypertension. ProtAgeGap: proteomic age gap (in years).

Extended Data Fig. 4 Effect size of ProtAgeGap on mortality and disease among non-smokers and those within normal weight range.

Associations between ProtAgeGap and mortality or diseases among UK Biobank participants who report being never smokers (n = 24,528) (a) and with a BMI ≥ 18.5 and < 25 kg/m 2 (n = 14,555) (b) . Shown on the x-axis are hazard ratios (and 95% confidence intervals) for the effect of ProtAgeGap on the outcomes shown. Events listed are the total number of incident cases for each outcome. All models are Cox proportional hazards models using model 2 (adjusted for age, sex, Townsend deprivation index, recruitment centre, and IPAQ activity group). No adjustment was made for multiple comparisons. ProtAgeGap: proteomic age gap (in years).

Extended Data Fig. 5 Associations between individual ProtAge APs and each disease studied.

For each outcome associated with ProtAgeGap20, a Cox proportional hazards model (n = 45,441) was calculated with all 20 proteins from the ProtAgeGap20 score, adjusted for age, sex (except prostate cancer), ethnicity, Townsend deprivation index, recruitment center, IPAQ activity group, and smoking status. No adjustment was made for multiple comparisons. In a) , the association between each protein and incident disease is colored by z-score, with z-scores for associations with p-value ≥ 0.05 set to 0. In b) , the importance of each protein with p < 0.05 is shown as a relative contribution. Relative contribution for each disease is calculated by scaling z-score for significant proteins such that they add to 1. APs: aging-related proteins; ProtAgeGap20: proteomic age gap from the 20-protein model.

Extended Data Fig. 6 ProtAgeGap increases linearly with increasing disease multimorbidity.

a) Years of ProtAgeGap in those with 0 (n = 6,826), 1 (n = 2,056), 2 (n = 605), 3 (n = 206), and 4+ (n = 116) comorbid conditions among UK Biobank (UKB) participants 40-50 years old at recruitment (total n = 9,809). b) Years of ProtAgeGap in those with 0 (n = 10,665), 1 (n = 6,903), 2 (n = 3,765), 3 (n = 1,702), and 4+ (n = 1,410) comorbid conditions among UKB participants aged 51-65 years old at recruitment (total n = 24,445). c) Percentages of the UKB population with 0, 1, 2, 3, and 4+ lifetime disease diagnoses. d) Years of ProtAgeGap according to levels of self-rated health in the UKB (total n = 43,393; Poor n = 2,249; Fair n = 9,355; Good n = 24,752; Excellent n = 7,004). Multimorbidity is defined as the number of lifetime diagnoses of any of the 26 diseases analyzed in this study. In a , b , and d , violin plots with center line, box limits, and whiskers represent the median, interquartile range, and minima/maxima within each group. For violin plots only, outliers were trimmed that were more than 2 standard deviations from total mean across all groups in the population subgroup plotted. Tests for significant differences between the means of groups were performed using a two-sided t-test. n.s.: not statistically significant; *** p-value < 0.001; ProtAgeGap: proteomic age gap (in years).

Extended Data Fig. 7 Protein–protein interaction network of ProtAge APs from the STRING database.

Protein–protein interaction (PPI) network of a highly interconnected subset of APs in the ProtAge model with at least 2 node connections using experimental PPI information from the STRING database. Proteins are sized and colored by number of connections, with those showing a greater number of connections with other proteins displayed larger and more yellow.

Extended Data Fig. 8 Protein–protein interaction network of ProtAge APs using SHAP interaction values.

Protein–protein interaction (PPI) network using SHAP values from the trained ProtAge model. Proteins shown are only those that are highly interconnected using a cutoff of 0.0083 for mean absolute SHAP interaction values. Proteins are sized and colored by number of connections, with those showing a greater number of connections with other proteins displayed larger and more yellow.

Extended Data Fig. 9 Overlap of ProtAge APs with existing DNAm and proteomic clock publications.

a) Overlap between genes coding for the 204 ProtAge APs versus genes mapped by proximity to CpGs from common DNAm clocks. b) Overlap between 204 ProtAge APs versus a recent systematic review of APs (Johnson et al. 2020), a recent comprehensive analysis of SOMAscan proteins associated with age (Coenen et al. 2023), and a recent proteomic aging clock created using SOMAscan data (Lehallier et al. 2019). APs: aging-related proteins, DNAm: DNA methylation.

Supplementary information

Supplementary information.

Supplementary Figs. 1–9 and supplementary table descriptions.

Reporting Summary

Supplementary tables 1–21., rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Argentieri, M.A., Xiao, S., Bennett, D. et al. Proteomic aging clock predicts mortality and risk of common age-related diseases in diverse populations. Nat Med (2024). https://doi.org/10.1038/s41591-024-03164-7

Download citation

Received : 14 September 2023

Accepted : 27 June 2024

Published : 08 August 2024

DOI : https://doi.org/10.1038/s41591-024-03164-7

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.