Machine learning reveals pre-treatment deficit accumulation as a predictor of persistent cancer-related fatigue: The Thinking and Living with Cancer longitudinal cohort study

INTRODUCTION: Cancer-related fatigue is a long-term, distressing symptom among older survivors. However, predictors of fatigue are not well defined. We used machine learning to identify pre-treatment factors that predict persistent fatigue among older breast cancer survivors. MATERIALS AND METHODS: Participants were women ages ≥60 with non-metastatic breast cancer enrolled pre-systemic treatment (n = 543) and age, race, and site-matched controls without a cancer history (n = 496). Fatigue was assessed using the Functional Assessment of Chronic Illness Therapy-Fatigue (FACIT-F) fatigue subscale at baseline, one year, and two years; persistent fatigue was defined as ≤34 points at year one and/or two. Baseline variables included demographics, medical history, lifestyle habits, symptoms, a 46-item deficit accumulation index (DAI; an indicator of frailty) score, and cancer and treatment details (survivors only), yielding 90 variables for survivors and 82 for controls. Logistic regression with Lasso variable selection, Stochastic Gradient Boosting, Random Forest, and Boruta machine learning models were applied to assess factors that predicted persistent fatigue. RESULTS: Participants were 68.4 ± 6.1 years; 14.4% of survivors and 8.1% of controls reported persistent fatigue (p = 0.002). Model performance was strong across algorithms (corrected AUROC range: 0.74-0.83). For all participants, predictors of persistent fatigue included higher baseline fatigue, as measured using the FACIT-F total score or FACIT-F fatigue subscale score. Notably, higher pre-treatment DAI score was a strong predictor among survivors across all models (p < 0.001) but was selected by only one model for controls. DISCUSSION: Greater pre-treatment DAI score was predictive of persistent fatigue in older breast cancer survivors. Thus, a higher DAI score may be a useful clinical indicator of fatigue risk, helping target intervention efforts for these patients. CLINICALTRIALS: gov identifier:NCT03451383.

Authors

Institutions

Publication Details

Journal
Journal of Geriatric Oncology
Published
2026-11-13
DOI
https://doi.org/10.1016/j.jgo.2026.104693
Primary Topic
Cancer survivorship and care
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Machine learning reveals pre-treatment deficit accumulation as a predictor of persistent cancer-related fatigue: The Thinking and Living with Cancer longitudinal cohort study

Amber S. Kleckner, Brent J. Small, Judith Carroll, Cynthia Kusters et al.
2 links
Journal of Geriatric Oncology
Cancer survivorship and care
article

Machine learning reveals pre-treatment deficit accumulation as a predictor of persistent cancer-related fatigue: The Thinking and Living with Cancer longitudinal cohort study

Amber S. Kleckner, Brent J. Small, Judith Carroll, Cynthia Kusters, Tim A. Ahles, Brenna C. McDonald, Andrew J. Saykin, Jeanne S. Mandelblatt, Heather S.L. Jim, Sunita K. Patel, Martine Extermann, Xingtao Zhou, James C. Root, Jaeil Ahn, Alice S. Ryan, Harvey Jay Cohen
article en

Abstract

INTRODUCTION: Cancer-related fatigue is a long-term, distressing symptom among older survivors. However, predictors of fatigue are not well defined. We used machine learning to identify pre-treatment factors that predict persistent fatigue among older breast cancer survivors. MATERIALS AND METHODS: Participants were women ages ≥60 with non-metastatic breast cancer enrolled pre-systemic treatment (n = 543) and age, race, and site-matched controls without a cancer history (n = 496). Fatigue was assessed using the Functional Assessment of Chronic Illness Therapy-Fatigue (FACIT-F) fatigue subscale at baseline, one year, and two years; persistent fatigue was defined as ≤34 points at year one and/or two. Baseline variables included demographics, medical history, lifestyle habits, symptoms, a 46-item deficit accumulation index (DAI; an indicator of frailty) score, and cancer and treatment details (survivors only), yielding 90 variables for survivors and 82 for controls. Logistic regression with Lasso variable selection, Stochastic Gradient Boosting, Random Forest, and Boruta machine learning models were applied to assess factors that predicted persistent fatigue. RESULTS: Participants were 68.4 ± 6.1 years; 14.4% of survivors and 8.1% of controls reported persistent fatigue (p = 0.002). Model performance was strong across algorithms (corrected AUROC range: 0.74-0.83). For all participants, predictors of persistent fatigue included higher baseline fatigue, as measured using the FACIT-F total score or FACIT-F fatigue subscale score. Notably, higher pre-treatment DAI score was a strong predictor among survivors across all models (p < 0.001) but was selected by only one model for controls. DISCUSSION: Greater pre-treatment DAI score was predictive of persistent fatigue in older breast cancer survivors. Thus, a higher DAI score may be a useful clinical indicator of fatigue risk, helping target intervention efforts for these patients. CLINICALTRIALS: gov identifier:NCT03451383.

Journal of Geriatric OncologyVol. 18(3)
University of North Carolina at Chapel Hill (US), University of Maryland, Baltimore (US), Indiana University Health (US), City Of Hope National Medical Center (US), Memorial Sloan Kettering Cancer Center (US), University of California, Los Angeles (US), Duke University (US), Georgetown University (US), Georgetown University Medical Center (US), Moffitt Cancer Center (US), Indiana Cancer Consortium (US), Baltimore VA Medical Center (US), University of Maryland Marlene and Stewart Greenebaum Comprehensive Cancer Center (US), Indiana University – Purdue University Indianapolis (US), University of Baltimore (US)
Division of Cancer Prevention, National Cancer Institute, National Institutes of Health, Lombardi Comprehensive Cancer Center, Georgetown University, National Institute on Aging, National Cancer Institute
Good health and well-being
Cancer survivorship and care
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon