Response to concerns with the statistical analysis in “Effect of combined GLP-1 analogue and bupropion/naltrexone on weight loss: a retrospective cohort study”

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

We refute Mokalla et al.’s concerns that our study conclusions are unsubstantiated by our data. Mokalla et al’s replication attempt yielded divergent results because the primary variables were analyzed without the required preprocessing. An independent statistical review of our data using identical datasets provided to Mokalla et al., completed by the UBC Applied Statistics and Data Science Group (ASDa), substantiated our conclusions (see supplementary file for ASDa’s statistical code and data output for our main analyses). Herein, we respond to Mokalla et al.’s concerns point by point.

Mokalla et al. were unable to reproduce our Table 2 results. We ascertained that the replication attempt incorrectly used raw, unprocessed data variables as their main exposures, as per their Table I. Mokalla et al.’s correspondence describes using “GLP” and “meds_response” raw variables to identify patients taking GLP-1 RA and combination therapy. These variables require further data processing to obtain final exposure variables. For example, the “GLP” variable indicates GLP-1 RA use at a particular visit from longitudinal data; thus, the use of this variable cross-sectionally does not include all patients, as patients did not take GLP-1 RA at every visit. The independent statistician, using identical datasets we provided Mokalla et al., substantiated our Table 2 conclusions (see supplementary files) [1].