Severe Infection Outcomes With SGLT2 Versus DPP ‐4 Inhibitors in Type 2 Diabetes: A Systematic Review and Active‐Comparator Meta‐Analysis of Real‐World Cohorts

Introduction To update and methodologically refine the synthesis of real‐world evidence on severe infection outcomes with sodium‐glucose cotransporter 2 inhibitors (SGLT2i) versus dipeptidyl peptidase‐4 inhibitors (DPP‐4i), emphasising limits of causal interpretation. Methods PubMed, Embase, Scopus and CENTRAL were searched from inception through 29 June 2026. Quantitative synthesis was restricted to DPP‐4i active‐comparator estimates. Incident pneumonia and sepsis were prespecified co‐primary outcomes. Hazard ratios and incidence rate ratios were pooled as related but non‐identical estimands using restricted maximum likelihood with modified Hartung‐Knapp inference, with leave‐one‐out, prediction‐interval and effect‐measure‐stratified analyses. ROBINS‐I and GRADE were applied. Results Eighteen cohorts were included; the 11 contributing quantitatively were all at moderate risk of bias. For incident pneumonia (five studies), the pooled ratio was 0.65 (95% CI: 0.55–0.75; prediction interval 0.45–0.93; I 2 = 89.8%; very low certainty); omitting one multinational cohort whose follow‐up began 2 years after drug initiation yielded 0.61 (0.56–0.67; I 2 = 0%). For incident sepsis (five studies), the prediction interval crossed the null (0.38–1.13); the exploratory pooled ratio was 0.65 (0.53–0.80; I 2 = 90.6%, and 88.1%–92.4% after every omission; very low certainty). Hazard‐ratio‐only point estimates were similar, although the sepsis estimate was imprecise (95% CI including 1.0). Two cohorts reported identical, unpooled estimates for infection‐related hospitalisation (0.74 each; low certainty). Comparisons with GLP‐1 receptor agonists did not show a consistent lower‐risk pattern across four cohorts, two in selected populations, and randomised‐trial syntheses have not shown a consistent class‐wide preventive effect. Conclusions The available evidence did not suggest an increased average risk of severe infection with SGLT2i compared with DPP‐4i. Given low to very low certainty and substantial heterogeneity, the lower pooled estimates should not be interpreted as a preventive effect. Review Registration: PROSPERO: CRD420261390509.