Data Source and Searches
We followed the PRISMA guidelines [10] and have summarized our search process in Figure 1. Adhering to the PRISMA checklist, we ensured comprehensive and transparent reporting of our systematic review, as seen in the supplementary file. Two investigators (JC.RM, and M.S.) developed the search strategy, which was revised and approved by the other investigators. We searched the following databases until September 1, 2024: PubMed-MEDLINE and EMBASE-OVID. The search strategy is shown in the supplementary file. There was no language limitation.
Study Selection
We included only RCTs in English reporting cardiovascular outcomes of Finerenone as treatment versus placebo. We excluded studies that did not match our population, intervention, and outcomes. Three investigators (JC.RM, M.S., and A.K.) independently screened each record title and abstract for potential inclusion. They then assessed full texts of selected abstracts. Discrepancies were resolved through discussion or by another investigator (A.B.).
Outcomes
The primary outcomes included death from cardiovascular causes, hospitalization for heart failure, death from any cause, and renal failure. The secondary outcomes focused on safety and adverse events, which included any serious adverse event, acute kidney injury, investigator-reported hyperkalemia, hyperkalemia related to the trial regimen, urinary tract infections, and pneumonia.
Data Extraction
Three investigators (JC.RM., M.S., A.K.) independently extracted data on the following variables: study design, study population, intervention group size, control group size, primary outcomes, secondary outcomes, and follow-up duration. Discrepancies in data extraction were resolved by discussion or by another investigator (A.B.). Baseline characteristics included demographics, clinical history, laboratory values, and medication use. Specifically, we summarized age, sex distribution, and racial composition (White, Black, Asian, Other, with missing data noted). Clinical history included duration of diabetes, systolic blood pressure, and the presence of cardiovascular disease. Kidney function was assessed using estimated glomerular filtration rate (eGFR), categorized into ≥60, 45 to <60, 25 to <45, and <25 ml/min/1.73 m².The urinary albumin-to-creatinine ratio was reported as a median with interquartile range (IQR) and categorized into <30, 30 to <300, and ≥300, including missing data. Baseline serum potassium levels were measured in mmol/liter. Medication use at baseline was recorded, including the use of angiotensin-converting enzyme inhibitors, angiotensin II receptor blockers, other renin-angiotensin system inhibitors, diuretics, statins, potassium-lowering agents, glucose-lowering therapies (including insulin, glucagon-like-peptide-1 receptor agonists and sodium-glucose cotransporter 2 inhibitors). Further details on these characteristics are provided in the baseline characteristics section.
Risk of Bias Assessment
Two investigators (M.S. and A.K.), independently assessed the risk of bias (RoB) by using the Cochrane RoB 2.0 tool for RCTs [11]; disagreements were resolved by discussion with a third investigator (A.B.). RoB 2.0 evaluates five bias domains: randomization process, deviations from intended interventions, missing outcome data, measurement of the outcome and selection of the reported result. The risk of bias assessment can be viewed in supplementary file.
Statistical Analysis
Our systematic review followed the 2020 PRISMA guidelines [10], and the analyses were conducted using the meta package in RevMan [12]. For the primary outcomes, we used a generic inverse variance random effects model, analyzing logged hazard ratios (HR) and confidence intervals (CI) for the two outcomes: death from cardiovascular causes and hospitalization for heart failure, as all three studies provided HRs for these measures. For death from any cause and renal failure, we employed a dichotomous random effects model, reporting outcomes as risk ratios (RR) because the FINEARTS-HF study did not provide HRs for these outcomes [13]. Adverse events were also analyzed as dichotomous outcomes using a random effects model and reported as RR.
We assessed heterogeneity among studies using the I² statistic, with I² values greater than 60% indicating substantial heterogeneity. The quality of evidence was determined using the GRADE methodology, which examines five key domains: risk of bias, inconsistency, indirectness, imprecision, and publication bias [14]. Publication bias was not assessed with Egger’s test, given a likely unreliable result with only three studies [15].