Largely the companies that sell the drugs, and in the better papers it says so. On the biggest recent pooling of obesity-drug trials, ten of twelve authors declare a financial relationship with a manufacturer [1]. On a cost-effectiveness model quoting a very favorable price per quality-adjusted life year, two of three authors work for the company that sets the price the model used [2]. Neither fact makes a finding false. Both change how much weight a fine distinction can carry.
Somebody deciding between semaglutide and tirzepatide is relying on a synthesis, whether they realize it or not. There are not many such papers, and the best of them are genuinely good. Reading one carefully means reading both what it found and who was in a position to find it, a habit that also applies to the premium one of those two molecules commands.
What the big review found
It searched Medline and Embase to the end of January 2025 for randomized trials of obesity medications against placebo or an active comparator. It pooled 56 of them, enrolling 60,307 patients: 32,598 on a drug and 27,709 on placebo. The primary endpoint was percentage of total body weight lost by the end of each study.
The trial counts are lopsided in a way the headline result hides. Orlistat, the oldest drug in the set, carries 22 of the 56 trials. Semaglutide has 14 and liraglutide 11. Tirzepatide, the one most often described as the strongest, rests on six. That is not a criticism of tirzepatide, which is newer, and it does mean the comparison between the two headline molecules is built on fourteen trials against six.
Every drug in the set beat placebo on weight loss at P < 0.0001, and only semaglutide and tirzepatide passed 10% of total body weight. Both restored normoglycemia, produced remission of type 2 diabetes, and reduced heart failure hospitalization. Semaglutide alone showed a reduction in major adverse cardiovascular events and in knee osteoarthritis pain; tirzepatide alone showed remission of obstructive sleep apnea and of metabolic dysfunction-associated steatohepatitis.
The declaration
Twelve authors are listed. Ten declare a financial relationship with a company that manufactures one of the drugs under review: speaker fees, advisory board fees, consulting fees and research grants. One of those is a research award funded by a manufacturer’s charitable foundation, which is a weaker tie and worth separating rather than lumping in. Two authors declare no conflicts relevant to the article.
Where it matters most: the money models
A cost-effectiveness analysis put semaglutide at $19,911 per quality-adjusted life year in metabolic dysfunction-associated steatohepatitis, which is a very good number by the usual thresholds. Two of its three authors are employed by the company that sets the drug price the model runs on, and the price is the assumption that decides the answer.
A second model found tirzepatide saves $41,688 a patient against semaglutide over a lifetime [3]. It draws its effect sizes from the manufacturer’s own head-to-head trial, and the saving is avoided complications across decades rather than money anyone has yet. What that means for a purchase made this month is set out in cheaper over a lifetime nobody has.
An economic model is far more sensitive to its inputs than a pooled weight-loss estimate is. Change the assumed price, the assumed persistence or the assumed time horizon and the answer moves. That is why a disclosure list matters more on a model than on a meta-analysis.
Why independent synthesis is scarce
Obesity pharmacology is a field with two dominant manufacturers. The clinicians qualified to synthesize its trials are largely the same people those manufacturers recruit to advise, to speak, and to run the trials. Expertise and proximity grow together.
A buyer looking for a comparison written by somebody with no commercial connection to either drug will find very few, and the ones they find will usually be smaller, older or narrower. That is a structural problem with no clean consumer-side fix. What it has is a reading practice. Check the declaration, note which molecule the funding clusters around, and treat single-molecule superiority claims with more caution than shared findings. That is exactly the situation with the cardiovascular result.
What this changes about a purchase
Very little on the headline numbers, and saying so is more useful than manufacturing an implication. The pooled weight-loss estimates for both molecules are large, consistent across trials, and supported by placebo arms of tens of thousands of people. A declared speaker fee does not move a 10% figure that fourteen randomized trials produced.
Where it should change something is confidence in the fine distinctions. The claim that one molecule is definitively better. The ranking of secondary benefits. The extrapolation of a six-trial evidence base to a twelve-month purchase. Those are the claims that sell the more expensive option, and the price difference on this roster is real money. The same test applies to any very large effect, for the reasons in a 77% mortality reduction nobody can use.