Abstract
Randomized Controlled Trials (RCTs) assessing adjuvant corticosteroids for hospitalized Community-Acquired Pneumonia (CAP) show mixed results, suggesting Heterogeneity of Treatment Effect (HTE). Current guidelines conflict: some restrict use to septic shock, while others advocate for broader application in severe CAP. Among three recent, large RCTs focused on severe CAP, however, only one demonstrated significant benefit, raising the question: should severity-based treatment guidance really be preferred?. The debate surrounding corticosteroid use in CAP is complicated by two issues: the lack of a unified definition for ‘severe CAP’, with definitions either based on the American Thoracic Society criteria or severity scores like the Pneumonia Severity Index, and the reliance on Aggregate Data Meta-Analyses (ADMAs). ADMAs synthesize trial results by categorizing entire RCTs as severe or non-severe, while most RCT populations included patients with varying disease severities. Individual Patient Data Meta-Analyses (IPDMAs) enable stratification on a patient-level. Our recent IPDMA utilized multivariate predictive HTE modelling in six RCT datasets to identify patient characteristics predicting corticosteroid benefit, which suggested C-reactive protein (CRP) as a predictor of benefit. Therefore, this finding was externally validated in two other RCTs, revealing statistically significant HTE found across baseline CRP subgroups (i.e., ≤ 204 mg/L vs. >204 mg/L), while no such HTE was found across severity-based subgroups. These findings may advocate for corticosteroid treatment guided by CRP instead of disease severity. CRP, however, is unlikely the sole driver of corticosteroids benefit, as other potential sources of HTE like imaging or cytokines (‘known unknowns’) or other unmeasured variables (‘ unknown unknowns’) may exist, and moreover, our IPDMA post-hoc suggested HTE across aetiology and treatment timing. Moving forward, the efficacy of precision medicine in CAP hinges on rigorous replication of HTE findings —being notoriously susceptible to false positives— in independent datasets, and harmonized and standardized data collection in future trials. Based on current evidence, CRP may serve as a pragmatic tool to guide corticosteroid treatment, but given to the drug’s “pendulum swings” from broad use to avoidance over the last decades, caution remains essential.