Comment on ‘Bempedoic Acid and Ezetimibe for the Treatment of Hypercholesterolemia: A Systematic Review and Meta-Analysis of Randomized Phase II/III trials’

Abstract

In their systematic review and meta-analysis recently published in Clinical Drug Investigation, Bhagavathula and colleagues [1] reported on their meta-analysis of randomized controlled trials of participants treated for hypercholesterolemia using a combination therapy of bempedoic acid and ezetimibe. They included three trials [2–4] comprising 388 participants and concluded that the combination was welltolerated and effective at lowering lipid parameters and inflammation marker levels. However, their meta-analysis has fundamental methodological flaws and numerical errors, leading to conclusions with no clinical relevance and incorrect statistical results. Furthermore, there also seems to be a degree of reporting bias in the article’s abstract and conclusions. As a result, when one also considers the small number of total trial participants, this systematic review hardly provides, as the authors claim, ‘robust’ evidence of a synergistic effect of this lipid-lowering combination therapy. The characteristics of the trials included in the metaanalysis were heterogeneous in terms of population characteristics and comparators. One of the included studies [3] investigated populations receiving maximum tolerated statin therapy, while participants in the other two studies [2, 4] did not receive background statin therapy. Comparator groups included placebo and ezetimibe monotherapy. The random-effects models used in the meta-analyses hardly fix the issues, neither can meta-regressions or sensitivity analyses. These approaches should not be used for obvious heterogeneity. Instead, the systematic review should have considered homogeneous subgroup meta-analyses, e.g. through stratifying by comparator groups, and should have focused on examining relationships between the clinical characteristics of the studies and the size of the intervention effects [5]. As a result, the effects for both benefits and harms calculated through the meta-analysis have little value for clinical decision making. The pooled effect measures are also mixed, as the metaanalysis pools within groups and between group measures. For example, the low-density lipoprotein (LDL)-cholesterol meta-analysis pools the percentage change from baseline within the combination therapy group from the first study [2] (− 38.80%, 95% confidence interval [CI] − 43.19 to − 34.41) with the mean difference of the percentage changes from baseline between the two combination therapy groups and the ezetimibe group from the second study [4]. The correct approach would have been, for the first study, to also calculate a similar difference by subtracting the mean of the change from baseline in the ezetimibe group from the mean of the change from baseline in the combination group, and calculating the standard deviation of the difference as the pooled standard deviation. This would have led to half of the originally estimated effect for the first study (− 19.60%, 95% CI − 22.25 to − 16.16), resulting in a much lower overall effect. However, this effect would be relative to bempedoic acid only and not to the effect of the combination therapy. The latter is impossible to estimate using a meta-analysis of these studies as the second study did not have a placebo arm. Bhagavathula and colleagues [1] also make a unit-of-analysis error in all their meta-analyses. From the study with two combination therapy groups with different bempedoic acid doses [4], they include two separate comparisons, namely low bempedoic acid combination therapy versus ezetimibe, and high bempedoic acid combination therapy versus ezetimibe. By doing this, they include the same ezetimibe group twice, effectively double-counting participants in the This comment refers to the article available online at https:// doi. org/ 10. 1007/ s4026102000989-1.