Two drugs, a dozen headlines, and a difference of a few kilos that depends entirely on which paper you read. This is the cleanest case I know of for separating “the number in the trial” from “the number in the article about the trial”, because the two largest analyses of the same question do not agree, and they disagree in a way you can trace to their methods.
The big pooled analysis
The first is a network meta-analysis from 2024 in Diabetologia: 28 randomised trials, 23,622 participants, 44.2 percent of them women, searching PubMed and Cochrane up to November 2023 for trials of at least twelve weeks comparing subcutaneous tirzepatide or semaglutide against each other, against placebo or against other glucose-lowering drugs. Primary outcomes were change in HbA1c and body weight; secondary outcomes included reaching an HbA1c target, losing at least ten percent of body weight, and safety.
Against placebo, the highest of the three tirzepatide amounts was the most efficacious treatment for HbA1c at a mean difference of −21.61 mmol/mol, followed by the middle tirzepatide amount at −20.19, the highest semaglutide amount at −17.74, the lowest tirzepatide amount at −17.60, the middle semaglutide amount at −15.25, and the lowest semaglutide amount at −12.00. For body weight, tirzepatide produced reductions from 9.57 kg at the highest amount to 5.27 kg at the lowest, against 4.97 kg down to 2.52 kg for semaglutide, and every tirzepatide amount beat the matching semaglutide amount in the between-drug comparisons. Both drugs increased gastrointestinal adverse events against placebo; neither increased serious adverse events or severe hypoglycaemia.
Two things to notice before you carry those numbers anywhere. This is a comparison in type 2 diabetes, not a trial in people who want to lose weight. And “kg of body weight” is a scale measurement, not a body composition measurement — which is the subject of the next article in this series, because the same trials have DXA substudies, and those substudies have something different to say.
The review of the reviews
The second is a Health Technology Assessment from 2025 by an independent group in Exeter, funded by the NIHR Evidence Synthesis programme. It did not run new trials; it appraised the network meta-analyses themselves. Of 22 systematic reviews identified, 14 were prioritised for analysis, because the other eight were rated low or critically low quality.
Its findings are the interesting part. At six months, subcutaneous tirzepatide was the most effective drug, associated with 9 to 12 kg of weight loss. But the largest effects were seen for subcutaneous semaglutide at the amount registered for obesity, at between 11.5 and 12.5 kg — and that estimate came from two network meta-analyses, each informed by six trials, both merging findings across multiple time points. No network meta-analysis had compared tirzepatide with that amount registered for obesity at all.
Then the sentence that should travel with every number you read in this field: the authors write that the evidence is in general agreement with the wider literature, but that the data on tirzepatide were not as resounding as reported in some meta-analyses, and that the habit of combining multiple amounts and merging findings from multiple time points limits our understanding of amount and time effects. They call for head-to-head trials of the two drugs at the higher amounts, and for longer trials to establish effectiveness and safety beyond 72 weeks.
What the data do not show
They do not show that either drug produces a given number of kilos in you. The biggest analysis is a population with type 2 diabetes; the review of the reviews says the headline ranking depends on whether amounts are kept separate and whether six-month and twelve-month data are merged. They do not show the effect on muscle, on waist circumference, or on what happens after the drug stops — the longest follow-up the assessment is willing to endorse is beyond 72 weeks, and there is no head-to-head trial at the two highest amounts. And they do not show that one drug is “better for you” without knowing which endpoint you care about: HbA1c and body weight are the endpoints these trials were built to measure, and neither is a measure of how long you live.
