Inflammatory blood markers sharpen machine learning detection of mild frailty

Older adults living with frailty show a distinctive pattern of inflammation in their blood, and those peripheral immune signals can improve machine learning models that identify the syndrome, according to a new study published in BMC Geriatrics. The research, part of the Gut-brain Alzheimer’s disease Inflammation and Neurocognitive Study (GAINS), found that frail participants had higher serum levels of the immune signaling molecule IL-2, and that adding inflammatory markers to clinical variables made classification models more accurate.

The researchers enrolled community-dwelling adults from the GAINS cohort and scored each participant for frailty using the Clinical Frailty Scale (CFS), defining frailty as a score of 4 or higher. From each participant’s serum, the team measured a panel of 27 inflammatory markers using the Bio-Plex Pro Human Cytokine Assay. In the primary analysis of 165 participants with data from the index study visit, frailty was present in 31 individuals, or 18.7 percent of the cohort. Frail participants had a median CFS of 4, placing them in the mild frailty range rather than at severe levels.

The condition correlated with age, malnutrition, and cognitive assessments in this cohort. When the researchers adjusted their statistical analysis for clinical covariates including age, gender, ADAS-COG 13 cognitive scores, and Mini Nutritional Assessment (MNA) scores, frail participants still showed higher serum levels of IL-2. The presence of IL-2 was associated with an odds ratio of 3.96 for frailty, with a confidence interval of 1.76 to 8.92, meaning participants with detectable IL-2 had nearly four times the odds of being classified as frail in this analysis.

Beyond the conventional statistical findings, the team turned to machine learning to test whether inflammatory markers added predictive value. The models revealed that inflammatory markers enhanced classification accuracy of frailty presence above clinical variables alone. To improve model accuracy further, a second machine learning analysis was performed on a larger group of 189 participants that included data from follow-up visits in addition to the index visit.

The authors conclude that increased IL-2 levels are associated with mild frailty and that peripheral inflammatory markers improve classification models of the syndrome. They suggest peripheral inflammation may be an important component of mild frailty, consistent with the idea that immune dysregulation and inflammaging can affect multiple organ systems. Because the study is cross-sectional and observational, it establishes association rather than causation, and the findings come from a single cohort of community-dwelling adults scored with the Clinical Frailty Scale. The work is published as open access research, with the article shared early as a citable accepted manuscript carrying a permanent DOI ahead of the final Version of Record.

Subject of Research: Peripheral inflammatory markers and machine learning classification of frailty in older adults

Article Title: Peripheral cytokine profiles contribute to machine learning model of frailty

Article References: Akie, T. E., Loew, E., Huang, Z., Neff, H. A., Michaels, O. P., & Haran, J. P. (2026). Peripheral cytokine profiles contribute to machine learning model of frailty. BMC Geriatrics. https://doi.org/10.1186/s12877-026-08497-1

Image Credits: AI Generated

DOI: 10.1186/s12877-026-08497-1

Keywords: frailty, IL-2, cytokines, inflammaging, machine learning, BMC Geriatrics, GAINS study, older adults, Clinical Frailty Scale, immune dysregulation, serum biomarkers, Peripheral