Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

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.

Https

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.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2025 Apr;47(2):2099-2118.
doi: 10.1007/s11357-024-01376-w. Epub 2024 Oct 24.

Complementary value of molecular, phenotypic, and functional aging biomarkers in dementia prediction

Andreas Engvig  1 , Karl Trygve Kalleberg  2 , Lars T Westlye  3   4 , Esten Høyland Leonardsen  3   4 ; Alzheimer’s Disease Neuroimaging Initiative Consortium
Collaborators, Affiliations

Complementary value of molecular, phenotypic, and functional aging biomarkers in dementia prediction

Andreas Engvig et al. Geroscience. 2025 Apr.

Abstract

DNA methylation age (MA), brain age (BA), and frailty index (FI) are putative aging biomarkers linked to dementia risk. We investigated their relationship and combined potential for prediction of cognitive impairment and future dementia risk using the ADNI database. Of several MA algorithms, DunedinPACE and GrimAge2, associated with memory, were combined in a composite MA alongside BA and a data-driven FI in predictive analyses. Pairwise correlations between age- and sex-adjusted measures for MA (aMA), aBA, and aFI were low. FI outperformed BA and MA in all diagnostic tasks. A model including age, sex, and aFI achieved an area under the curve (AUC) of 0.94 for differentiating cognitively normal controls (CN) from dementia patients in a held-out test set. When combined with clinical biomarkers (apolipoprotein E ε4 allele count, memory, executive function), a model including aBA and aFI predicted 5-year dementia risk among MCI patients with an out-of-sample AUC of 0.88. In the prognostic model, BA and FI offered complementary value (both βs 0.50). The tested MAs did not improve predictions. Results were consistent across FI algorithms, with data-driven health deficit selection yielding the best performance. FI had a stronger adverse effect on prognosis in males, while BA's impact was greater in females. Our findings highlight the complementary value of BA and FI in dementia prediction. The results support a multidimensional view of dementia, including an intertwined relationship between the biomarkers, sex, and prognosis. The tested MA's limited contribution suggests caution in their use for individual risk assessment of dementia.

Keywords: Biological age; Brain age; Deep learning; Dementia; Frailty index; Machine learning; Methylation age.

PubMed Disclaimer

Conflict of interest statement

Declarations. Ethical approval: The present research study was done in accordance with the ADNI data use agreement and its undertaking approved by the regional ethics committee (REK 2018/1127). Competing interests: Esten H. Leonardsen has received an honorarium for lecturing from Lundbeck AS. He is also the CTO and a major shareholder in baba.vision, a company developing clinical tools for precision diagnostics in neurological disorders. Lars T. Westlye is a shareholder in baba.vision. Karl T. Kalleberg is a shareholder of Age Labs, a company focused on developing biomarkers for age-related diseases. Consortium members and affiliations: Alzheimer’s Disease Neuroimaging Initiative Consortium investigators: Michael Weiner5, Paul Aisen6, Ronald Petersen7, Clifford R. Jack Jr7, William Jagust8, John Q. Trojanowski9, Arthur W. Toga10, Laurel Beckett11, Robert C. Green12, Andrew J. Saykin13, John C. Morris14, Leslie M. Shaw9, Enchi Liu15, Tom Montine16, Ronald G. Thomas6, Michael Donohue6, Sarah Walter6, Devon Gessert6, Tamie Sather6, Gus Jiminez6, Danielle Harvey11, Matthew Bernstein6, Nick Fox17, Paul Thompson18, Norbert Schuff19, Charles DeCarli11, Bret Borowski7, Jeff Gunter7, Matt Senjem7, Prashanthi Vemuri7, David Jones7, Kejal Kantarci7, Chad Ward7, Robert A. Koeppe20, Norm Foster21, Eric M. Reiman22, Kewei Chen22, Chet Mathis23, Susan Landau8, Nigel J. Cairns24, Erin Householder10, Lisa Taylor Reinwald24, Virginia Lee25, Magdalena Korecka25, Michal Figurski25, Karen Crawford10, Scott Neu10, Tatiana M. Foroud13, Steven Potkin26, Li Shen13, Faber Kelley13, Sungeun Kim13, Kwangsik Nho13, Zaven Kachaturian27, Richard Frank28, Peter J. Snyder29, Susan Molchan30, Jeffrey Kaye31, Joseph Quinn31, Betty Lind31, Raina Carter31, Sara Dolen31, Lon S. Schneider10, Sonia Pawluczyk10, Mauricio Becerra10, Liberty Teodoro10, Bryan M. Spann10, James Brewer6, Helen Vanderswag6, Adam Fleisher6, Judith L. Heidebrink20, Joanne L. Lord20, Ronald Petersen7, Sara S. Mason7, Colleen S. Albers7, David Knopman7, Kris Johnson7, Rachelle S. Doody32, Javier Villanueva Meyer32, Munir Chowdhury32, Susan Rountree32, Mimi Dang32, Yaakov Stern33, Lawrence S. Honig33, Karen L. Bell33, Beau Ances24, Maria Carroll24, Sue Leon24, Erin Householder24, Mark A. Mintun24, Stacy Schneider24, Angela Oliver34, Randall Griffith34, David Clark34, David Geldmacher34, John Brockington34, Erik Roberson34, Hillel Grossman35, Effie Mitsis35, Leyla deToledo-Morrell36, Raj C. Shah36, Ranjan Duara37, Daniel Varon37, Maria T. Greig37, Peggy Roberts37, Marilyn Albert38, Chiadi Onyike38, Daniel D’Agostino II38, Stephanie Kielb38, James E. Galvin39, Dana M. Pogorele39, Brittany Cerbone39, Christina A. Michel39, Henry Rusinek39, Mony J. de Leon39, Lidia Glodzik39, Susan De Santi39, P. Murali Doraiswamy40, Jeffrey R. Petrella40, Terence Z. Wong40, Steven E. Arnold9, Jason H. Karlawish9, David A. Wolk23, Charles D. Smith41, Greg Jicha41, Peter Hardy41, Partha Sinha41, Elizabeth Oates41, Gary Conrad41, Oscar L. Lopez23, MaryAnn Oakley23, Donna M. Simpson23, Anton P. Porsteinsson42, Bonnie S. Goldstein42, Kim Martin42, Kelly M. Makino42, M. Saleem Ismail42, Connie Brand42, Ruth A. Mulnard26, Gaby Thai26, Catherine Mc Adams Ortiz26, Kyle Womack43, Dana Mathews43, Mary Quiceno43, Ramon Diaz Arrastia43, Richard King43, Myron Weiner43, Kristen Martin Cook43, Michael DeVous43, Allan I. Levey44, James J. Lah44, Janet S. Cellar44, Jeffrey M. Burns45, Heather S. Anderson45, Russell H. Swerdlow45, Liana Apostolova46, Kathleen Tingus46, Ellen Woo46, Daniel H. S. Silverman46, Po H. Lu46, George Bartzokis46, Neill R. Graff Radford47, Francine Parfitt47, Tracy Kendall47, Heather Johnson47, Martin R. Farlow13, Ann Marie Hake13, Brandy R. Matthews13, Scott Herring13, Cynthia Hunt13, Christopher H. van Dyck48, Richard E. Carson48, Martha G. MacAvoy48, Howard Chertkow49, Howard Bergman49, Chris Hosein49, Sandra Black50, Bojana Stefanovic50, Curtis Caldwell50, Ging Yuek Robin Hsiung51, Howard Feldman51, Benita Mudge51, Michele Assaly Past51, Andrew Kertesz52, John Rogers52, Dick Trost52, Charles Bernick53, Donna Munic53, Diana Kerwin54, Marek Marsel Mesulam54, Kristine Lipowski54, Chuang Kuo Wu54, Nancy Johnson54, Carl Sadowsky55, Walter Martinez55, Teresa Villena55, Raymond Scott Turner56, Kathleen Johnson56, Brigid Reynolds56, Reisa A. Sperling12, Keith A. Johnson12, Gad Marshall12, Meghan Frey12, Jerome Yesavage57, Joy L. Taylor57, Barton Lane57, Allyson Rosen57, Jared Tinklenberg57, Marwan N. Sabbagh58, Christine M. Belden58, Sandra A. Jacobson58, Sherye A. Sirrel58, Neil Kowall58, Ronald Killiany59, Andrew E. Budson59, Alexander Norbash59, Patricia Lynn Johnson59, Thomas O. Obisesan60, Saba Wolday60, Joanne Allard60, Alan Lerner61, Paula Ogrocki61, Leon Hudson61, Evan Fletcher62, Owen Carmichael62, John Olichney62, Charles DeCarli62, Smita Kittur63, Michael Borrie64, T. Y. Lee64, Rob Bartha64, Sterling Johnson65, Sanjay Asthana65, Cynthia M. Carlsson65, Steven G. Potkin66, Adrian Preda66, Dana Nguyen66, Pierre Tariot22, Adam Fleisher22, Stephanie Reeder22, Vernice Bates67, Horacio Capote67, Michelle Rainka67, Douglas W. Scharre68, Maria Kataki68, Anahita Adeli68, Earl A. Zimmerman69, Dzintra Celmins69, Alice D. Brown69, Godfrey D. Pearlson70, Karen Blank70, Karen Anderson70, Robert B. Santulli71, Tamar J. Kitzmiller71, Eben S. Schwartz71, Kaycee M. Sink72, Jeff D. Williamson72, Pradeep Garg72, Franklin Watkins72, Brian R. Ott73, Henry Querfurth73, Geoffrey Tremont73, Stephen Salloway74, Paul Malloy74, Stephen Correia74, Howard J. Rosen12, Bruce L. Miller12, Jacobo Mintzer75, Kenneth Spicer75, David Bachman75, Elizabether Finger76, Stephen Pasternak76, Irina Rachinsky76, John Rogers76, Andrew Kertesz76, Dick Drost76, Nunzio Pomara77, Raymundo Hernando77, Antero Sarrael77, Susan K. Schultz78, Laura L. Boles Ponto78, Hyungsub Shim78, Karen Elizabeth Smith78, Norman Relkin79, Gloria Chaing79, Lisa Raudin79, Amanda Smith80, Kristin Fargher80 & Balebail Ashok Raj80 5UC San Francisco, San Francisco, CA, USA. 6University of California San Diego, San Diego, CA, USA. 7Mayo Clinic, Rochester, NY, USA. 8UC Berkeley, Berkeley, CA, USA. 9University of Pennsylvania, Philadelphia, PA, USA. 10University of Southern California, Los Angeles, CA, USA. 11Davis, Davis, CA, USA. 12Brigham and Women’s Hospital, Harvard Medical School, Boston, MA, USA. 13Indiana University, Bloomington, IND, USA. 14Janssen Alzheimer Immunotherapy, South San Francisco, CA, USA. 15University of Washington, Seattle, WA, USA. 16University of London, London, UK. 17USC School of Medicine, Los Angeles, CA, USA. 18UCSF MRI, San Francisco, CA, USA. 19University of Michigan, Ann Arbor, MI, USA. 20University of Utah, Salt Lake City, UT, USA. 21Banner Alzheimer’s Institute, Phoenix, AZ, USA. 22University of Pittsburgh, Pittsburgh, PA, USA. 23Washington University St. Louis, St. Louis, MO, USA. 24UPenn School of Medicine, Philadelphia, PA, USA. 25University of California, Irvine, CA, USA. 26Khachaturian, Radebaugh & Associates, Inc and Alzheimer’s Association’s Ronald and Nancy Reagan’s Research Institute, Chicago, IL, USA. 27General Electric, Boston, MA, USA. 28Brown University, Providence, RI, USA. 29National Institute on Aging/National Institutes of Health, Bethesda, MD, USA. 30Oregon Health and Science University, Portland, OR, USA. 31Baylor College of Medicine, Houston, TX, USA. 32Columbia University Medical Center, New York, NY, USA. 33University of Alabama Birmingham, Birmingham, MO, USA. 34Mount Sinai School of Medicine, New York, NY, USA. 35Rush University Medical Center, Chicago, IL, USA. 36Wien Center, Vienna, Austria. 37Johns Hopkins University, Baltimore, MD, USA. 38New York University, New York, NY, USA. 39Duke University Medical Center, Durham, NC, USA. 40University of Kentucky, Lexington, KY, USA. 41University of Rochester Medical Center, Rochester, NY, USA. 42University of Texas Southwestern Medical School, Dallas, TX, USA. 43Emory University, Atlanta, GA, USA. 44University of Kansas, Medical Center, Lawrence, KS, USA. 45University of California, Los Angeles, CA USA. 46Mayo Clinic, Jacksonville, FL, USA. 47Yale University School of Medicine, New Haven, CT, USA. 48McGill Univ., Montreal Jewish General Hospital, Montreal, WI, USA. 49Sunnybrook Health Sciences, Toronto, ON, Canada. 50U.B.C. Clinic for AD & Related Disorders, British Columbia, BC, Canada. 51Cognitive Neurology St. Joseph’s, Toronto, ON, Canada. 52Cleveland Clinic Lou Ruvo Center for Brain Health, Las Vegas, NV, USA. 53Northwestern University, Evanston, IL, USA. 54Premiere Research Inst Palm Beach Neurology, West Palm Beach, FL, USA. 55Georgetown University Medical Center, Washington, DC, USA. 56Stanford University, Santa Clara County, CA, USA. 57Banner Sun Health Research Institute, Sun City, AZ, USA. 58Boston University, Boston, MA, USA. 59Howard University, Washington, DC, USA. 60Case Western Reserve University, Cleveland, OH, USA. 61University of California, Davis Sacramento, CA, USA. 62Neurological Care of CNY, New York, NY, USA. 63Parkwood Hospital, Parkwood, CA, USA. 64University of Wisconsin, Madison, WI, USA. 65University of California, Irvine BIC, Irvine, CA, USA. 66Dent Neurologic Institute, Amherst, MA, USA. 67Ohio State University, Columbus, OH, USA. 68Albany Medical College, Albany, NY, USA. 69Hartford Hosp, Olin Neuropsychiatry Research Center, Hartford, CT, USA. 70Dartmouth Hitchcock Medical Center, Albany, NY, USA. 71Wake Forest University Health Sciences, Winston-Salem, NC, USA. 72Rhode Island Hospital, Providence, RI, USA. 73Butler Hospital, Providence, RI, USA. 74Medical University South Carolina, Charleston, SC, USA. 75St. Joseph’s Health Care, Toronto, ON, Canada. 76Nathan Kline Institute, Orangeburg, SC, USA. 77University of Iowa College of Medicine, Iowa City, IA, USA. 78Cornell University, Ithaca, NY, USA. 79University of South Florida: USF Health Byrd Alzheimer’s Institute, Tampa, FL, USA.

Figures

Fig. 1
Fig. 1
a The dataset was split into subsets to facilitate hyperparameter-tuning and obtain unbiased out-of-sample estimates of model performance. Due to various combinations of missing biomarker data, we first extracted all participants with complete frailty index (FI) and brain age (BA) data, but lacking methylation data into a stage 1 modelling training set (n = 1491). This was used to identify the optimal selection of age- and sex-adjusted BA and FI (i.e., aBA, aFI, or aBA + aFI) as predictors across all predictive tasks. The remaining 385 participants (with complete FI, BA, and DNA methylation data) were stratified using diagnosis, age, and sex and divided into another training set for a second modelling stage (stage 2) and a hold-out test set. b All training data was utilized for descriptive statistics, and we employed a data-driven variable selection procedure to determine which methylation age markers to use. After selecting the best combination of BA and FI as predictors, stage 2 modelling was done to find the optimal selection of all three aging biomarkers. The best models from stages 1 and 2 were compared in a cross-validation loop before evaluating the best model in the test set to obtain an unbiased estimate of model performance
Fig. 2
Fig. 2
Scatterplots showing individual data and model fit lines from linear regressions from the stage 2 training dataset of age- and sex-adjusted candidate methylation ages (MA) and memory (Rey Auditory Verbal Learning Test, immediate recall (RAVLT), left column) and executive function (Trail Making Test, part B (TMT-B), right column). β represents the linear regression coefficient of each MA and the corresponding nominal p-value. Significant associations are colored
Fig. 3
Fig. 3
a Training set density plots illustrate the distributions of age- and sex-adjusted aging biomarkers (standardized residuals) in the subgroups used for diagnostic and prognostic classification. The 1st row shows the distributions for subjects with normal cognition (CN) and dementia (DEM), respectively, 2nd row for CN and subjects with mild cognitive impairment (MCI), 3rd row for MCI and DEM, and 4th row for subjects with MCI progressing to DEM within 5 years (pMCI) and subjects with MCI remaining stable (sMCI). Note: The results reported here are derived from all available training data for each of the biomarkers separately. As a result, medians and p-values shown in row 4 (sMCI, pMCI) differ slightly from those in Table 1 which are based on a smaller selection of subjects with complete data for all biomarkers. The blue and red values are subgroup aging biomarker medians. p-values are false discovery rate-adjusted from Kruskal–Wallis H-tests of subgroup differences in the biomarkers. b Pearson correlations between the adjusted aging biomarkers and common clinical markers in the training set. Additional abbreviations: prefix a = standardized age- and sex-adjusted residual; APOE, apolipoprotein E ε4 allele count; BA, brain age; FI, frailty index; MA, Composite methylation age; RAVLT, Rey Auditory Verbal Learning Test, immediate recall; TMT-B, Trail Making Test, part B
Fig. 4
Fig. 4
Predictive performance––reported as the area under the receiver operating characteristic curve (AUC)––for the various models for each of the three predictive (diagnostic) tasks. For each task, a baseline model was first fit in the stage 1 dataset using age and sex as predictors. Next, models including standardized age- and sex-adjusted brain age (aBA) and frailty index (FI) residuals, both independently and in combination, were trained using the same data. The x-axis denotes these different models, and the y-axis denotes their AUC. Individual points (blue) represent performance in independent folds in the cross-validation, whereas the black line denotes their mean. The red line represents the performance of the best model in the hold-out test set. For each row, the ROC curves underlying the AUC computation are shown on the right. Additional abbreviations: CN, cognitively normal; MCI, mild cognitive impairment; DEM, dementia
Fig. 5
Fig. 5
a Comparison of the prognostic models containing different subsets of the aging biomarkers for predicting progression from mild cognitive impairment (MCI) at baseline to dementia (DEM) within 5 years. b ROC curves underlying the AUCs computed in a. c Comparison of the best prognostic aging model with a prognostic model containing tests commonly used in the clinical workup of suspected DEM: apolipoprotein E ε4 allele count (APOE), the Rey Auditory Verbal Learning Test, immediate recall (RAVLT), and the Trail Making Test, part B (TMT-B). The final model denoted aging + clinical combines predictors from these two. All models included age and sex. For a and c, the individual points (blue) denote the area under the receiver operating characteristic curve (AUC) in independent folds in the cross-validation (CV), the black line is the CV mean, and the red line is the model performance of the best model in the hold-out test set. d The regression coefficients of the best-performing model for predicting 5-year DEM progression
See this image and copyright information in PMC

References

    1. Pierce AL, Bullain SS, Kawas CH. Late-onset Alzheimer disease. Neurol Clin. 2017;35:283–93. 10.1016/j.ncl.2017.01.006. - PubMed
    1. Gonzales MM, Garbarino VR, Pollet E, Palavicini JP, Kellogg DL, Kraig E, Orr ME. Biological aging processes underlying cognitive decline and neurodegenerative disease. J Clin Invest. 2022;132(10). 10.1172/JCI158453 - PMC - PubMed
    1. Hou Y, Dan X, Babbar M, Wei Y, Hasselbalch SG, Croteau DL, Bohr VA. Ageing as a risk factor for neurodegenerative disease. Nat Rev Neurol. 2019;15:565–81. 10.1038/s41582-019-0244-7. - PubMed
    1. Biondo F, Jewell A, Pritchard M, Aarsland D, Steves CJ, Mueller C, Cole JH. Brain-age is associated with progression to dementia in memory clinic patients. NeuroImage Clin. 2022;36:103175. 10.1016/j.nicl.2022.103175. - PMC - PubMed
    1. Engvig A, Maglanoc LA, Doan NT, Westlye LT, Alzheimer’s Disease Neuroimaging Initiative. Data-driven health deficit assessment improves a frailty index’s prediction of current cognitive status and future conversion to dementia: results from ADNI. GeroScience. 2023;45:591–611. 10.1007/s11357-022-00669-2. - PMC - PubMed

LinkOut - more resources