Association Between SGLT2 Inhibitor Use and Hepatocellular Carcinoma Risk in Type 2 Diabetes: A Systematic Review and Meta-Analysis

Background and Aims: Type 2 diabetes mellitus is a recognized risk factor for hepatocellular carcinoma (HCC), particularly in the setting of metabolic dysfunction-associated steatotic liver disease (MASLD), chronic viral hepatitis, advanced fibrosis, and cirrhosis. Beyond hyperglycemia and insulin resistance, diabetic hepatocarcinogenesis is shaped by metabolic inflammation, lipotoxicity, oxidative stress, fibrogenic remodeling, and the cirrhosis-dysplasia-HCC continuum. Sodium-glucose cotransporter-2 inhibitors (SGLT2i) may influence several hepatometabolic pathways, but the epidemiologic evidence linking SGLT2i use to HCC risk remains heterogeneous. Methods: We conducted a systematic review and meta-analysis of observational studies evaluating SGLT2i exposure and incident HCC in adults with type 2 diabetes. PubMed, Embase, and the Cochrane Library were searched up to 15 March 2026. Adjusted time-to-event estimates were pooled using a restricted maximum likelihood (REML) random-effects model. The certainty of evidence was assessed using the GRADE framework and judged to be very low. Results: Six observational studies including 526,446 participants were included. SGLT2i exposure was associated with a lower observed risk of incident HCC (pooled HR 0.59, 95% CI 0.45–0.77), but between-study heterogeneity was substantial (I2 = 75.2%, τ2 = 0.074). The association remained directionally similar after exclusion of Huynh et al. (HR 0.61, 95% CI 0.45–0.81) and in a DPP-4 inhibitor-restricted active-comparator analysis (HR 0.60, 95% CI 0.39–0.92). However, the 95% prediction interval crossed the null (0.25–1.37), indicating that future comparable studies may plausibly show no protective association. Conclusions: SGLT2i exposure was associated with a lower observed risk of incident HCC across available observational studies. However, the certainty of evidence was judged to be very low, and substantial heterogeneity, comparator variation, mixed time-to-event estimands, residual confounding, and a prediction interval crossing the null preclude causal interpretation. These findings should be considered hypothesis-generating rather than practice-changing evidence and support further hepatology-oriented validation.