Document resource
Introduction SGLT2 inhibitors reduce heart failure hospitalisations and cardiovascular death in patients with HFpEF/HFmrEF. However, atrial fibrillation (AF) is highly prevalent in this population. While post-hoc analyses suggest a neutral safety profile, reliance on aggregate data may obscure rare but serious safety signals. We assessed AF risk reported exclusively as a ‘Serious Adverse Event’ (SAE) using regulatory definitions. Furthermore, Trial Sequential Analysis (TSA) was employed to distinguish ‘absence of evidence of harm’ from ‘evidence of absence of harm’, determining if current data statistically rule out arrhythmic risk.Methods We conducted a systematic review (PROSPERO ID: CRD420201276042) of RCTs comparing SGLT2 inhibitors (dapagliflozin or empagliflozin) against placebo in HFpEF (LVEF >40%). To minimise reporting bias, AF data were manually extracted from ‘Serious Adverse Events’ modules of ClinicalTrials.gov by two independent reviewers, bypassing primary manuscripts. The primary outcome was SAE-defined AF incidence. Data were pooled using a random-effects model with the robust Hartung-Knapp-Sidik-Jonkman (HKSJ) method. TSA calculated the Required Information Size (RIS) to definitively refute a 20% relative risk increase (alpha=5%, power=80%).Results Four pivotal trials (DELIVER, EMPEROR-Preserved, PRESERVED-HF, EMPERIAL-Preserved) comprising 12,870 patients were included; all had low bias risk (Cochrane RoB 2). Pooled meta-analysis demonstrated a non-significant trend towards increased serious AF risk with SGLT2i vs placebo (RR 1.18; 95% CI 0.92-1.52; P=0.13). No heterogeneity was observed (I2=0%). In TSA, the cumulative Z-score was -1.45, crossing neither significance boundaries (-1.96) nor monitoring boundaries for harm. The calculated RIS to definitively rule out risk increase was 32,384 patients. Current information (n=12,870) represents only 39.7% of the RIS.Conclusions Analysis of regulatory-grade data showed SGLT2i were not associated with a significant increase in serious AF in HFpEF. However, the trend (RR 1.18) and TSA highlight that while there is an ‘absence of evidence of harm’, current data lack power for ‘evidence of absence’. Data remain underpowered to definitively exclude modest arrhythmic risk. Continued pharmacovigilance is warranted until optimal information size is met.Abstract 161 Figure 1Risk of serious AF with SGLT2i vs placebo in patients with HFpEF. Forest plot stratified by trial scale. Events derived exclusively from SAE reports in ClinicalTrials.gov. Pooled RR: 1.18 (95% CI 0.92-1.52) using a Hartung-Knapp-Sidik-Jonkman random-effects model, no heterogeneity (I2=0%)Abstract 161 Figure 2TSA of serious AF events (SGLT2i vs placebo). The cumulative Z-curve (blue) did not cross the conventional significance boundaries (Z = ±1.96) or the trial sequential monitoring boundaries for harm. The Required Information Size (RIS) was calculated as 32,384 patients, assuming a 20% relative risk increase (α = 5%, power = 80%)