Efficacy and Safety of SGLT2 Inhibitors and Sotagliflozin in Heart Failure: A Systematic Review and Meta-Analysis of 59 Randomized Controlled Trials

Figure 1. PRISMA 2020 flow diagram of study selection Flow of records through identification, screening, eligibility assessment, and inclusion in the systematic review and meta-analysis. Generated with the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) 2020 R package [22].

Figure 2. Risk of bias 2.0 assessment: traffic light grid (Panel A) and domain summary (Panel B) Panel A: Per-study risk-of-bias judgments across five Cochrane RoB 2.0 domains plus overall judgment for all 114 included studies, sorted as Low → Some concerns → High, then alphabetically within group. Reference numbers in [N] format are shown for every record: the 59 trials in the quantitative synthesis plus the 21 citable excluded reports (observational reports, conference abstracts, and protocol papers with their own digital object identifier) carry their own number, while records that duplicate or sub-reference a trial already in the quantitative synthesis carry that parent trial's reference number. Panel B: Domain-level summary across all 114 included trials.

Figure 3. Forest plot: all-cause mortality Random-effects pooled risk ratio for all-cause mortality across 26 randomized controlled trials with at least one event. Studies sorted by reference number; reference numbers in [N] format inline with each study label.

Figure 4. Forest plot: heart failure hospitalization Random-effects pooled risk ratio for heart failure hospitalization across 15 randomized controlled trials. Reference numbers in [N] format inline with each study label.

Figure 5. Subgroup forest plot: all-cause mortality by heart failure phenotype Subgroup analysis of all-cause mortality by HF phenotype (HFrEF, HFpEF, and HFmrEF). Reference numbers in [N] format inline with each study label.

Figure 6. Sensitivity analyses: composite of leave-one-out, filter-based sensitivity, and alternative models Panel A: Leave-one-out sensitivity for all-cause mortality, with each row showing the pooled risk ratio when the corresponding trial is omitted (reference numbers inline with study labels). Panel B. Filter-based sensitivity for all-cause mortality across 12 pre-specified subsets (excluding high risk of bias (RoB), low RoB only, excluding open-label, HFrEF only, HFpEF only, follow-up ≥12 months, sample size thresholds, drug-class restrictions, excluding imputed data, excluding sotagliflozin). Panel C. Alternative statistical models for all-cause mortality (fixed-effect Mantel-Haenszel risk ratio (RR), random-effects odds ratio (OR), fixed-effect OR, Peto OR, Paule-Mandel RR).

Figure 7. Publication bias and trial sequential analysis: composite Panel A: Funnel plot for all-cause mortality across the contributing trials. Panel B: Trim-and-fill funnel plot for all-cause mortality with six imputed studies. Panel C: Funnel plot for HF hospitalization. Panel D: Trial sequential analysis Z-curve for HF hospitalization at 15% relative risk reduction threshold, with O'Brien-Fleming conservative monitoring boundaries (α = 0.05, β = 0.20); the cumulative Z-statistic crosses the lower boundary well before the required information size, confirming benefit. Funnel plots are aggregate.