What is the evidence behind new I-A and III-A recommendations in European Society of Cardiology guidelines?

Abstract Background European Society of Cardiology (ESC) guidelines divide recommendations into various classes and Levels of Evidence (LOE) to convey their overall strength to readers. In ESC guidelines, Class I or III recommendations with LOE A represent strong recommendations (either for benefit, lack of efficacy, or harm) on the basis of multiple randomized controlled trials (RCTs) or a meta-analysis of RCTs. Previous work has examined the evidence base supporting ESC recommendations, suggesting a relatively low proportion of Class I recommendations supported by LOE A (I-A) and Class III recommendations supported by LOE A (III-A). In this cross-sectional, meta-epidemiological study, we examined the evidence cited in support of new I-A and III-A recommendations to quantify the strength of the cited evidence. Methods We identified all current ESC guidelines through the ESC website, from which new Class I and III recommendations supported by LOE A were extracted. We then identified the studies cited to support each recommendation. Next, we determined whether there was a clear outcome indicated in the recommendation, and matched this to the cited studies. If no clear outcome was specified, we searched studies for the outcome that appeared most relevant for the recommendation. If there were multiple possible outcomes, we prioritized primary over secondary outcomes, more objective over subjective outcomes, and disease-specific over generic outcomes. If all these criteria were equivalent, then we selected the first relevant outcome reported in the results section of the cited study. Outcome data were extracted in duplicate, and all outcome measures were converted to odds ratios using established methods. These estimates were then meta-analyzed using an inverse-variance weighted fixed-effects meta-analysis to produce a summary odds ratio with 95% confidence interval, standard error and z-value, representing the strength of the evidence for each recommendation. Z-values of 1.96, 2.81, and 3.29 correspond to two-sided p-values of 0.05, 0.005, and 0.001, respectively. Results We included 26 guidelines overall. We present here a preliminary subset of 28 class I-A recommendations from 9 guidelines. Each guideline contributed a median of 1 (IQR 1-4) recommendations, though 1 guideline (2023 Cardiovascular Disease and Diabetes) accounted for 13 recommendations. 22 (78.5%) were new recommendations. 22 (64.3%) recommendations were related to pharmaceutical interventions. The median odds ratio was 0.73 (IQR 0.55-0.80), with a range of 0.26 to 0.93. The median Z value was 4.82 (IQR 3.15-6.47), with a range of 2.31 to 9.88. Discussion Our preliminary results indicate that I-A recommendations are backed by strong statistical evidence. To inform future guideline changes, ongoing and future trials should aim for a high degree of statistical certainty and avoid stopping too early for benefit.