It’s not quite a randomized trial, but the results are good news for the underdog.

It’s July, which means my hospital is filled with new interns, residents and fellows all eager to be embarking on a new stage of their career. It’s an exciting time – a bit of a scary time – but it’s also the time when the medical strategies I’ve been taking for granted get called into question. At this point in the year, I tend to get a lot of “why” questions.  Why did you order that test?  Why did you suspect that diagnosis? Why did you choose that medication?

Meds are the hardest, I find.  Sure, I can explain that I prescribed a GLP-1 receptor agonist because the patient had diabetes and was overweight, and multiple studies show that this class of drug leads to weight loss and reduced mortality risk. But then I get the follow-up.  Sure, but why THAT GLP-1 drug?  Why do I pick semaglutide (Ozempic) over tirzepatide (Mounjaro)? Here’s where I run out of good answers.  Sometimes I’m choosing a drug because that’s what the patient’s insurance has on their formulary. Sometimes it’s because it’s cheaper in general. Sometimes, it’s just force of habit. I know the dose, I have experience with the side effects – it’s comfortable.

What I can’t say is that I have solid evidence that one drug is superior to another, say from a randomized trial of semaglutide vs. tirzepatide. I don’t have that evidence because that trial has never happened, and, as I’ll explain in a minute – may never happen at all.

But we might have the next best thing. And the results may surprise you.

Why don’t we see more head-to-head trials of competitor drugs? The answer is pretty simple, honestly: risk management. For drugs that are on patent, like the GLP-1s, conducting a trial without the buy-in of the pharmaceutical company is simply too expensive – we can’t run the trial unless someone provides the drug for free. That gives the companies a lot of say in what trials get done, and it seems that most pharma companies have reached the same conclusion: a head-to-head trial is too risky. Be happy with the market share you have and try to nibble away at the edges through good old-fashioned marketing.

But if you look at the data that’s out there, you might wonder why Ozempic is the market leader. I mean, sure, it’s a heck of a weight loss drug. But the weight loss in the trials of Mounjaro was actually a bit higher. It’s worth noting here that tirzepatide (mounjaro) is not just a GLP-1 receptor agonist it is also a gastric inhibitory polypetide agonist.

But it’s very hard to compare the results of a trial pitting Ozempic against placebo with a totally different trial pitting Mounjaro against placebo. You can always argue that the patients studied were just too different at baseline – an apples and oranges situation.

Newly published this week, a study appearing in JAMA Internal Medicine uses real-world data and propensity-score matching to turn oranges back into apples. I’ll walk you through it.

The data and analysis here comes from Truveta – a collective of various US healthcare systems that share a broad swath of electronic health record data. Researchers identified 41,222 adults with overweight or obesity who were prescribed semaglutide or tirzepatide between May 2022 and September 2023.

You’d be tempted to just see which group lost more weight over time, but that is the apples and oranges problem. People prescribed Mounjaro were different than people who were prescribed Ozempic. There are a variety of factors to look at here, but the vibe is that the Mounjaro group seems healthier at baseline: They were younger, had less kidney disease, less hypertension, less hyperlipidemia. They had higher income, and were more likely to be white. They were also dramatically LESS likely to have diabetes.

To account for this, the researchers used a statistical technique called propensity score matching. Briefly, you create a model based on a variety of patient factors to predict who would be prescribed Ozempic and who would be prescribed Mounjaro.  You then identify pairs of patients with similar probability (or propensity) of receiving, say, Ozempic, where one member of the pair got Ozempic and one got Mounjaro.  Any unmatched individuals simply get dropped from the analysis.

Thus the researchers took the 41,222 individuals who started the analysis, of whom 9,193 received Mounjaro, and identified the 9,193 patients who got Ozempic that most closely matched the Mounjaro crowd. I know, it sounds confusing.  But as an example, in the original dataset, 51.9% of those who got Mounjaro had diabetes compared to 71.5% of those who got Ozempic. Among the 9,193 individuals who remained in the Ozempic group after matching, 52.1% had diabetes. By matching in this way, you balance your baseline characteristics. Turning apples into oranges. Or, maybe the better metaphor would be plucking the oranges out of a big pile of mostly apples.

Once that’s done, we can go back to do what we wanted to do in the beginning which is to look at the weight loss between the groups.

What I’m showing you here is the average percent change in body weight at 3, 6, and 12-months across the two drugs in the matched cohort. By a year out, you have basically a 15% weight loss in the Mounjaro group compared to a 8% or so in the Ozempic group.

We can slice this a different way as well – asking what percent of people in each group achieve, say, 10% weight loss?  This graph examines the percent of each treatment group who hits that weight loss target over time. Mounjaro gets there faster.

I should point out that this was a so-called “on treatment” analysis. If people stopped taking either of the drugs, they were no longer included in the study. That tends to make drugs like this appear better than they are – since as time goes on you may weed out the people who stop the drug due to lack of efficacy or due to side effects.  But in a sensitivity analysis, the authors see what happens if they just treat people as if they were taking the drug for the entire year once they had it prescribed and the results, while not as dramatic, were broadly similar. Mounjaro still came out on top.

Adverse event rates – stuff like gastroparesis and pancreatitis – were rare, but similar between the two groups.

It’s great to see studies like this that leverage real world data and a solid statistical underpinning to give us providers actionable information. Is it 100% definitive? No. But, especially considering the clinical trial data, I don’t think I’m going out on a limb to say that Mounjaro seems to be the more effective weight loss agent. That said, we don’t actually live in a world where we can prescribe medications based on a silly little thing like which is the most effective. Especially given the cost of these agents – the patient’s insurance status is going to guide our prescription pen more than this study ever could. And of course, given the demand for this class of agents, and the fact that both are actually quite effective – you may be best off prescribing whatever you can get your hands on.

But I’d like to see more of this. When I do have a choice of a medication, when costs and availability are similar, I’d like to be able to answer that question of “why did you choose that one?” with an evidence-based answer: “it’s better.”

A version of this commentary first appeared on Medscape.com.