Trials of these drugs report percentage weight loss, which is convenient and quietly consequential. This meta-analysis shows what happens when the same comparison is expressed both ways [1]. The pooled figures across the whole literature are in the weight-loss averages.
Measured as a percentage, the Asian and non-Asian trial populations lost essentially the same amount: 18.16% against 17.92%, with a p value of 0.94. Measured in kilograms they did not: 16.72 kg against 23.46 kg, with a p value of 0.03. Both statements describe the same participants in the same trials.
The authors are unusually careful about what their own null means. Only 148 Asian participants received tirzepatide, across two trials, and they state plainly that the comparison is underpowered and cannot confirm equivalence. A p value of 0.94 on a sample that small is absence of evidence rather than evidence of sameness — the distinction that governs the frailty analysis, where the sample was large enough for a null to mean something.
The design carries a further limit. This compares across trials rather than within one, so the Asian and non-Asian groups differ in every way two separate trials differ — sites, era, baseline characteristics, background care — and ethnicity is only one of them. Heterogeneity on the headline effect was also substantial, at I² = 77%.
For a reader the transferable lesson is about units rather than ethnicity. Any comparison between populations with different baseline weights will look different depending on whether it is reported as a percentage or an absolute, and trials almost always choose the percentage. The same trap sits behind cross-trial comparisons like the ecnoglutide trial, where entry criteria set a lighter starting population, and behind the subgroup ranges in the sleep apnea subgroups.