Five Blood Biomarkers Predict Multimorbidity Across All Measures in Older Adults

Multimorbidity — the simultaneous presence of two or more chronic diseases — affects up to 90% of adults over 60 and drives disability, cognitive decline, and premature death. Despite its prevalence, the biological mechanisms underlying why some people accumulate diseases faster or develop specific disease clusters remain poorly understood. This study represents one of the most comprehensive biomarker investigations of multimorbidity to date, spanning inflammatory, metabolic, vascular, and neurodegenerative pathways.

The research team analyzed 54 blood biomarkers at baseline in 2,247 participants from the Swedish National Study on Aging and Care in Kungsholmen (SNAC-K), a population-based longitudinal cohort of adults aged 60 and over (mean age 72.7 years, 61.5% female). Multimorbidity was assessed using three complementary measures: total baseline disease count, five multimorbidity patterns identified through latent class analysis (LCA), and the 15-year longitudinal rate of disease accumulation. LASSO regression was applied to identify the most predictive biomarkers for each measure.

Five biomarkers emerged as consistently and positively associated with all multimorbidity measures: growth differentiation factor 15 (GDF-15), hemoglobin A1c (HbA1c), cystatin C, leptin, and insulin. These reflect cellular stress responses, glycemic dysregulation, kidney function, and metabolic signaling — underscoring metabolic disturbance as a central, cross-cutting driver of multimorbidity. The five LCA-derived disease clusters included an Unspecific pattern, a Neuropsychiatric pattern, a Psychiatric and Respiratory pattern, a Sensory Impairment and Anemia pattern, and a Cardiometabolic pattern. Beyond the shared five biomarkers, each pattern showed additional specific biomarker associations, suggesting distinct biological substrates contribute to different disease combinations. Faster longitudinal disease accumulation was independently associated with elevated gamma-glutamyl transferase (liver stress/oxidative damage) and lower albumin (nutritional status and systemic inflammation).

Longitudinal findings were externally validated in 522 participants from the Baltimore Longitudinal Study of Aging (BLSA), where predictive accuracy was comparable, strengthening confidence in the generalizability of the results across different aging populations. This dual-cohort validation is a notable methodological strength.

The findings align with the geroscience hypothesis — that targeting shared mechanisms of aging (rather than individual diseases) may prevent or delay multimorbidity. The identification of modifiable metabolic biomarkers like HbA1c, insulin, and leptin as universal drivers points toward actionable intervention targets. However, the observational design precludes causal inference, and the studied cohorts are predominantly older urban Europeans, which may limit generalizability to other ethnic or socioeconomic groups.