We have read with great interest the post hoc analysis of the Liraglutide Effect and Action in Diabetes: Evaluation of Cardiovascular Outcome Results (LEADER) randomized trial that investigated effects of liraglutide on incidence of acute gallbladder or biliary disease (uncomplicated gallbladder stones, complicated gallbladder stones, cholecystitis, and biliary obstruction) (1). In this study, time-to-first-event analysis was performed by Cox regression, and it documented increased risk of acute gallbladder or biliary disease with the use of liraglutide compared with placebo (hazard ratio 1.60; 95% CI 1.23, 2.09; P < 0.001). In addition, trends toward the increased incidence were observed for each of the four categories of gallbladder- or biliary tract–related events. Although results of this study univocally associate the use of liraglutide with acute gallbladder or biliary disease, we believe that more appropriate analysis is mandatory for this type of research.
The LEADER trial documented the positive effect of liraglutide on cardiovascular risk, especially mortality, accounting for 66 deaths less in the liraglutide arm. However, its post hoc analysis revealed an excess of 53 cases of acute gallbladder or biliary disease with the use of liraglutide. Considering the fact that the patient who died could not experience acute gallbladder or biliary disease, together with the similar difference between beneficial and detrimental effects of liraglutide, it is obvious that the statistical methodology used deserves additional attention from the clinicians.
The pitfall of performing statistical analysis in the article by Nauck et al. (1) was that the investigators should have considered an issue of competing risks. Competing risk appears when one of the events can prevent or modify the chance of occurrence of the event of interest. The researchers used the standard method of survival analysis (Cox regression) for their time-to-event analysis, which is suboptimal due to the fact that it wrongly presumes random and noninformative censoring (independent of the event of interest), but censoring due to death is informative as it precludes the event of interest from occurring. As a result, we can expect falsely higher estimates of the event of interest in the group with less censoring, in this case the liraglutide group. The method for competing risk analysis, such as the method developed by Fine and Gray (2), takes into consideration informative censoring together with all the observed events. Consequently, it interprets the effect of interest depending on the magnitude of the risk of other competing events.
Therefore, in order to make results of this trial more informative and robust for both the clinicians and the patients, additional analysis using the Fine and Gray method should be performed accounting for death as a competing risk for development of gallbladder/biliary disease.
Article Information
Duality of Interest. M.S. has received honoraria for lectures from Roche and Sandoz. D.R. has served as principal investigator or co-investigator in clinical trials of AstraZeneca, Eli Lilly, Merck Sharp & Dohme, Novo Nordisk, Sanofi, Solvay, and Trophos. D.R. has received honoraria for speaking or advisory board engagements and consulting fees from Abbott, Amgen, AstraZeneca, Bayer, Boehringer Ingelheim, Eli Lilly, LifeScan–Johnson & Johnson, Novartis, Novo Nordisk, Merck Sharp & Dohme, Pfizer, Pliva, Roche, Salvus, Sanofi, and Takeda. M.L. has received honoraria for lectures from Roche and Janssen. No other potential conflicts of interest relevant to this article were reported.
