When insulin needs drop on one of these drugs, the obvious explanation is the weight. This analysis tested that directly and found the obvious explanation accounts for a minority of the effect early on [1]. It is the same question this site asked about liver outcomes in the liver repair that was not the weight.
Total daily insulin fell 22.6% over 26 weeks, with a confidence interval from -28.3 to -17.0. The reduction was not spread evenly: bolus insulin fell 30.5% and basal insulin 15.6%, so the basal share of the total rose from 0.56 to 0.62. Insulin per kilogram per day fell from 0.72 to 0.60, and daily carbohydrate intake fell from 137 g to 107 g.
The bolus-versus-basal pattern fits what the drug does. Bolus insulin covers meals, and the carbohydrate intake in this cohort fell by 30 g a day. Basal insulin covers the background requirement, which tracks body mass more slowly. That ordering is consistent with the appetite mechanism this site covered in the circuit that links nausea and appetite.
The population is narrow and the caveat matters. These were adults with type 1 diabetes and obesity, using automated insulin delivery systems, in a randomized trial with close supervision. Type 1 diabetes is not an approved indication for this drug, insulin reduction in an unsupervised setting carries obvious hazard, and this site has covered the stopping side of the same question in what happened when type 1 patients stopped.
What makes the analysis worth reading beyond that population is the method. Most reports of a drug’s secondary benefits assume weight is the mechanism and never test it. This one separated the two and found both present, in a ratio that changes over time for arithmetic reasons rather than pharmacological ones — the distinction this site keeps drawing in sixteen percent lower, six hundredths of a point.