A finding that two thirds of people on these drugs have controlled blood pressure against a quarter of everyone else would be remarkable if it were solid [1]. The trial evidence on blood pressure sits in a different place, and the scale here is the first thing to establish — the same first question asked of the eating-phenotype heatmap.
The GLP-1 arm of this national estimate is 45 people. Survey weighting then expands them to represent 994,028 American adults, and the paper reports weighted counts like 659,630 with controlled blood pressure. Those figures look like population measurements and are arithmetic performed on a few dozen respondents.
The design compounds it. This is cross-sectional: drug use and blood pressure were recorded at the same moment, so nothing establishes which came first. Someone whose hypertension is well controlled is someone in regular contact with a clinician who is adjusting medication — the same person who is likely to have been started on a newer diabetes drug. The comparison may be measuring care intensity rather than pharmacology.
One contrast in the paper points that way. GLP-1 users had a substantially higher BMI than non-users, 37.25 against 31.22. Heavier people with better-controlled blood pressure is an unusual pattern, and the readier explanation is that they were receiving more medical attention rather than that the drug overcame their weight — a selection shape this desk has flagged before in the cohort past eighty.
The data window matters too. These are survey cycles running to March 2020, before the obesity indication drove mass uptake, so the 45 users are overwhelmingly people taking these drugs for diabetes. They are not the weight-management market this desk serves, and the access picture for that market is in who the price filtered out and covered and still not treated. Blood pressure does fall on these drugs in randomized trials; this analysis is not the evidence for it.