Episode Details

In Part 1 we started discussing the study Two Years After Stopping GLP-1s, Most Patients Sustain at Least Some Weight Loss by Bartelt et al. which I identified as possibly the worst, and definitely among the top 3 worst, studies I’ve ever analyzed.

In part 1 we talked about the platform (Epic Research), the researchers, and the basic methodology for reporting. Today we’re going to get into the analysis. Remember that there are two sources, the first is the actual article, the second is what I am calling the supplementary materials that is a pdf that is linked (though not at all clearly) in the article.

We’ll compare the claims made in the article to the data shown in the supplementary materials. I think as we get into this it may be helpful (and maybe just a bit comforting) to remember that this was not published in a peer reviewed journal - it’s Epic publishing Epic’s research on Epic’s website.

We’ll begin with the most basic claim: “we studied 188,722 patients who stopped using a GLP-1 medication after being on it for at least 90 days and who lost at least 5 pounds while on it.”

You might be thinking “wait - in Part 1 you said the study population was 323,782, what happened to the other 135,060 people? Let’s go on this journey together. We’ll start by looking at their data. In the supplementary materials, Table 1: Characteristics of the Study Population looks like this:

So they say in the article that they studied 188,722 patients but the very first line of the supplementary materials chart about the study population says that the total number of patients was 323,782. But it gets a bit weirder, if you add up all the participants in the different age categories you get 323,750, 32 short. If you add up the totals across the drug groups you do get 188,722, if you add up the participants in the weight loss category you get 323,782. If you add up the diabetic status you get 323,782. What is going on here? Maybe they originally looked at 323,782 records of which only 188,722 were not excluded for some reason, and maybe 32 of the participants did not have an age on their electronic health record? At least the number of people across the drug groups in the supplementary materials chart matches the number of people they claim to have studied but this is, in the absolute more charitable description, an extremely confusing way to present this data.

In part 1 we discussed the fact that the inclusion criteria were not particularly stringent but the biggest issue, to me, is that the researchers appear to have just done calculations with whatever data they happened to have, but reported it and drew conclusions as if they had consistent data on all participants.

For each of the three drug groups (semaglutide, liraglutide, tirzepatide), they present an interactive graph that shows “the proportion of patients by amount of weight regained or lost after stopping [liraglutide/semaglutide/tirzepatide].”

Let’s dig in.

In the article we find Figure 1 - Proportion of Patients by Weight Change After Stopping Semaglutide we can see in the lower left “n=139,972 patients.”

But if you look at the supplementary materials you find Table 2 Proportion of Patients by Weight Change After Stopping Semaglutide. This is the table from which they are drawing the data for Figure 1.

The final column of the supplemental table is “patients” and the number in that column represents the number of the original 139,972 patients who had taken (and subsequently stopped taking) semaglutide for which the researchers had a weight at each month of their calculation. At month 1, the researchers had a we