Identifying proteins and pathways associated with multimorbidity in 53,026 adults.
Level 3 - non-randomized controlled study
Prospective observational cohort study
PubMed 39740741 · doi:10.1016/j.metabol.2024.156126
What was done
This prospective cohort study analyzed plasma proteomic data from 53,026 UK Biobank participants followed for a median of 13.3 years (baseline 2006–2010). Investigators used Cox proportional hazards regression models to evaluate associations between baseline proteins and 38 incident outcomes (31 chronic conditions, 6 system-specific mortality endpoints, and all-cause mortality). Ordinal regression assessed associations with multimorbidity status (0–1, 2, 3–4, or ≥5 diseases), followed by functional enrichment, tissue enrichment, and upstream regulator analyses.
What was found
After Bonferroni correction (p < 3.42 × 10^-7), 972 proteins (33.3%) were shared across at least two incident chronic diseases, with 93.3% demonstrating consistent effect directions across conditions. In contrast, 345 proteins (11.8%) were uniquely linked to a single condition. Four proteins (GDF15, PLAUR, WFDC2, and AREG) showed positive associations with 20 to 24 incident diseases (hazard ratios ranging from 1.21 to 3.77) and strong associations with multimorbidity severity (odds ratios ranging from 1.33 to 1.89). Protein levels were primarily explained by renal function, liver function, inflammation, and obesity, with pathway analyses highlighting immune-response pathways regulated by transcription factors NFKB1, JUN, and RELA.
Why it matters
The study maps the shared proteomic architecture across diverse chronic diseases, demonstrating that multimorbidity is largely driven by common inflammatory and immune-mediated pathways rather than isolated disease-specific mechanisms. Identifying broadly shared drivers like GDF15 and PLAUR provides potential systemic biomarkers and shared targets for multimorbidity intervention.
Limits
The UK Biobank is predominantly composed of participants of European ancestry who are generally healthier than the broader population, limiting generalizability. Proteomic profiling was conducted at a single baseline timepoint, precluding analysis of protein trajectory over time. The observational design cannot prove whether identified proteins are causal drivers or downstream markers of underlying systemic pathology.