Abstract
BACKGROUND: Despite their disproportionately high mortality from community-acquired pneumonia (CAP), older adults remain understudied regarding inflammasome-mediated cell death pathways. We sought to determine whether distinct pyroptosis activation patterns exist in this population and how nutritional status modifies their prognostic impact. METHODS: This retrospective cohort study enrolled 282 patients aged ≥75 years hospitalized for CAP. We quantified circulating pyroptosis effectors (gasdermin D [GSDMD], NLRP3, caspase-1) and assessed nutritional status using the Mini Nutritional Assessment-Short Form (MNA-SF). K-means clustering identified biological endotypes; generalized additive models (GAMs) characterized nonlinear biomarker-mortality relationships. The primary endpoint was 28-day all-cause mortality. RESULTS: Three pyroptosis endotypes emerged with markedly divergent outcomes: hyper-pyroptotic (n=73; mortality 57.5%), intermediate-pyroptotic (n=128; mortality 10.2%), and hypo-pyroptotic (n=81; mortality 1.2%). The hyper-pyroptotic endotype was characterized by severe malnutrition (48.2% with MNA-SF ≤7) and elevated cytokines (median IL-6: 98.4 pg/mL). GAM analysis revealed threshold-dependent, nonlinear relationships-mortality risk escalated sharply when GSDMD exceeded 3.5 ng/mL but showed attenuation at extreme values. Notably, two-dimensional analyses demonstrated supra-additive risk in patients with concurrent nutritional compromise and pyroptosis activation. An integrated prognostic model achieved AUC 0.898 (95% CI: 0.847-0.943), significantly outperforming the Pneumonia Severity Index alone (AUC 0.793; P<0.001), with superior net benefit across clinically relevant decision thresholds (10-30%). CONCLUSION: Geriatric CAP comprises biologically distinct pyroptosis endotypes. Malnutrition was associated with a stronger relationship between pyroptosis markers and short-term mortality, consistent with effect modification. These findings support integrating nutritional assessment with pyroptosis biomarker profiling for risk stratification and generate hypotheses for prospective mechanistic and interventional studies.