A new study leverages millions of patients to find the effects - and side-effects -  no one is talking about.

I’m really the last person to call something a miracle drug, but when it comes to the new weight loss drugs – formally the GLP-1 receptor agonists like Ozempic – I’m hard-pressed to think of a different term.

Remember, these were designed as diabetes drugs. They were supposed to lower blood sugar. And they do. But it turned out they had profound affects on appetite, leading to substantial weight loss in the range of 10-20% of body weight. And I feel a bit like a late-night infomercial host here – because whenever I talk about these drugs I’m like “and that’s not all”.  Studies have also shown that they curb other appetites. People taking them are more likely to quit smoking, drinking, and gambling. They do less compulsive shopping even.  I’ve talked about this before, referring to these really as “anti-consumption” drugs, and considering them more or less the antidote to the central problem of our current culture – over-consumption.

But I’m not an unabashed optimist. When I teach interpretation of the medical literature to students, I remind them that if you think something sounds too good to be true – it probably is. What other effects do these drugs have that we don’t know about?  The problem is – lots of medical research is hypothesis-driven.  You come up with a hypothesis – “GLP-1 drugs increase the risk of pancreatitis” and then you design a study to prove that hypothesis false or true. But these drugs have so many diverse effects. Is there a way we could possibly check for all of them? At once? 

Actually, you can.  By shifting the frame from hypothesis-driven studies to something called “discovery” studies. The idea being to look for any signal of benefit or harm across a huge range of datapoints. It’s sort of like searching for extra-terrestrial intelligence. You can either point your telescope at some interesting nebula or planet to see if there is anything out there – or you can just sort of listen broadly to the entire sky at once. 

Of course, to do discovery research, you need a huge number of people to study. Ideally, those people need to have all their data collected in some large, centralized medical record.  More ideally, they should all have equal access to the drug of interest so socioeconomic issues don’t wash out all the other signals. Basically, you need a socialized medicine cohort. But the United States doesn’t have socialized medicine, right?

Actually, we do.

It’s called the VA.

The Department of Veterans Affairs is the largest integrated healthcare network in the United States, serving more than nine million veterans a year. That’s a lot of data to mine.

And mined it was, by Ziad Al-Aly and colleagues in this paper, appearing in Nature Medicine.

The researchers wanted to see what effects GLP1-receptor agonists like Ozempic and Mounjaro might have – good and bad – using what amounts to a shotgun approach. Here’s how it worked.

Until recently, these drugs were only approved for people with type 2 diabetes so they restricted their analysis to that population. They identified 215,970 individuals with diabetes who started taking the GLP-1 drugs. They used people who started taking other diabetes drugs – like sulfonylureas, dpp4 inhibitors, and sglt-2 inhibitors as controls. They also had a group of over a million people who didn’t start taking any new drugs. If you’re keeping track, that’s actually 4 distinct control groups.  I’ll try to integrate the results for you though.

What about outcomes? They used that amazing VA electronic health record to code 175 distinct health outcomes across 12 different broad categories ranging from “blood and blood forming organs” to “mental” to “symptoms” as you can see here.

But before we dig into the interesting results, it’s worth noting that the authors put in a couple of proof-of-concept outcomes just to show the whole thing is working.

As a positive control?  Weight loss.  If the data structure works as expected, we should clearly see that people started on the GLP-1 drugs lose more weight than those started on other drugs. And indeed that is the case: those started on GLP-1s were 36% more likely to lose 5 points of BMI and 25% more likely to lose 10 points. They also had about a 20% reduced risk of cardiovascular and kidney events – findings which have been shown in prior studies.

As for a negative control – they cleverly picked “trauma due to traffic injury”.  There is no reason to think that GLP1s would increase or reduce the risk of traffic accidents, and indeed they saw no association there.  So – data structure seems fine – performing as expected.

What about the 173 other outcomes? This is where it gets interesting. Obviously, I can’t report on every single one in the time I have so I will highlight a few.

Do you want the good news or bad news?

Let’s start with the good news. Putting aside the known reductions in weight, cardiovascular, and kidney disease risks.  Other benefits? Protean.

There was a reduction in septic shock, liver failure, schizophrenia, opioid use disorders, alcohol use disorders, hemorrhagic stroke, suicidal ideation, pulmonary embolism. The list goes on and on.  It’s hard to even find a unifying principle for the things these drugs improved – a variety of psychiatric disorders, infectious processes, respiratory diseases.

Of the 12 broad systems I mentioned, all had at least one outcome that was improved. The one least effected seems to be cancer – with a lower rate of liver cancer and no differences across the other types of cancer.  But broadly – impressive results.

Now the bad news.  It’s not that bad.

People started on GLP-1 drugs had increased risks of nausea and vomiting (we knew that one), acid reflux, and abdominal pain.  But there were some surprises here as well – a significantly higher rate of musculoskeletal problems like tenosynovitis, a higher rate of kidney stones, headaches, and bone pain.

But overall there’s nothing terribly concerning here – at least not when weighed against the benefits of less stroke, shock, suicide, and respiratory problems. Broadly speaking, the digestive system saw the most adverse effects, and after that the musculoskeletal system, which will certainly bear some closer monitoring. But basically I’m not too worried.

So let’s return to the start here. Are these miracle drugs?  Honestly? It’s hard for me to look at data like this and try to refute that. The primary drivers of chronic disease in the US are obesity, alcohol, smoking, and lack of physical activity. These drugs seem to meaningfully improve three out of four of those. And honestly, I wouldn’t be surprised if people become a bit more active after losing weight on Ozempic too.

Is this naïve? Should we be waiting for the other shoe to drop?  Look, I can find things to criticize in this study. It only looked at people with diabetes – and that’s not necessarily the population who will be taking these drugs in the future. The study population, being from the VA, is almost entirely male – so perhaps we can’t be as confident in the effects seen here among women. The average duration of follow-up was just three and a half years or so.  So maybe some bad things will pop up in the future – but it’s worth noting that while these drugs are relatively new in the weight loss space, they are not new for the treatment of diabetes. Exenatide was approved 20 years ago, after all – and the other shoe still isn’t dropping.

The real mystery to me is exactly how these drugs have all these beneficial effects. Clearly, they are affecting some kind of reward system in the brain, and by changing that system you end up changing all of these behaviors – from food intake to smoking to drug use and so on. They are, as I said before, fundamentally anti-consumption agents in an age of over-consumption. Perhaps in some truly idealized sense – they are the cure for what ails us.

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