Six obesity drugs have randomized evidence behind them. Storefronts sell two. Across 56 trials and 60,307 patients, all six beat placebo on weight loss, and only semaglutide and tirzepatide exceeded 10% [1]. The other four are orlistat, liraglutide, naltrexone with bupropion, and phentermine with topiramate. A larger network meta-analysis of 262 trials and 99,791 participants puts phentermine-topiramate at −8.1% against lifestyle alone [2].
The drugs this site tracks are two of six with randomized evidence behind them, and the other four are almost never mentioned on a storefront. This review places all six in one frame. That is the only honest way to answer whether the expensive option is the right one. The trial-by-trial reading in what the semaglutide trials are worth can only half answer it.
What was pooled
A search to 31 January 2025 identified 56 randomized trials comparing obesity medications against placebo or an active comparator, enrolling 60,307 patients [1]. Orlistat appeared in 22 trials, semaglutide in 14, liraglutide in 11, tirzepatide in 6, naltrexone with bupropion in 5, and phentermine with topiramate in 2. All six produced significantly greater total body weight loss than placebo. Only semaglutide and tirzepatide exceeded 10%.
The ranking, from the largest comparison
The biggest synthesis so far pooled 262 randomized trials and 99,791 participants across 19 drugs [2]. At one year against lifestyle modification alone, tirzepatide led at −14.9%, 95% CI −16.0% to −13.9%. Cagrilintide-semaglutide came next at −14.8%, oral semaglutide at −10.9%, orforglipron at −9.9%, injected semaglutide at −9.8%, and phentermine-topiramate at −8.1%. All of those carry moderate to high certainty.
So the cheapest drug with a respectable result is phentermine with topiramate, at a little over half what tirzepatide achieves. That is the trade a reader is actually being asked to make, and it is never framed that way on a storefront. The tolerability side of the same analysis is in the discontinuation ranking.
What the older drugs cost in side effects
A Cochrane review of long-term weight drugs in people with hypertension reports the harms in a way the efficacy reviews do not [3]. Orlistat probably increases serious adverse events at RR 1.45, 95% CI 1.10 to 1.91, across three trials and 1,476 participants. Naltrexone with bupropion probably increases all adverse events at RR 1.69, 95% CI 1.58 to 1.80. Phentermine with topiramate probably increases them at RR 1.13, 95% CI 1.08 to 1.20.
None of that makes the older drugs unusable. It means the cheaper option is cheaper partly because it is older and partly because it does less, and its side-effect record is not blank. The same review found the semaglutide and tirzepatide trials reported no adverse-event results at all in that population, which is a different problem, set out in the Cochrane update.
What none of this establishes
Which drug is better than which, for a particular person. Fifty-six trials with different populations, durations and comparators can establish that each beats placebo. They cannot substitute for putting two drugs in one protocol. The review’s own conclusion is that the choice should be individualized rather than ranked, and the one head-to-head that does exist is read in the tirzepatide against semaglutide comparison.
The drugs still in trials are a separate category again. Ecnoglutide, mazdutide and retatrutide may produce reductions of 13.1% to 14.6%, with certainty graded very low to low, and none of them can be bought. What is coming and what is buyable are separated in the pipeline reading.
Why the other four matter to a buyer
Because they are cheaper, and because nobody selling a GLP-1 will mention them. Four drugs with randomized evidence of real weight loss exist outside this market entirely, at a fraction of what this roster charges. They have smaller effects and their own side-effect profiles.
That does not make them the right choice. The effect sizes really are smaller, and for many people the difference between 5% and 15% is the difference between a result and none. It does mean a reader comparing only the two drugs a storefront sells is comparing a shortlist somebody else wrote.