Oh Oh Oh Ozempic-is there anything GLP-1 drugs can’t do?

A new study maps wide-ranging benefits (and risks) of GLP-1 drugs.

TL;DR: In a large sample of US Veterans, GLP-1 drugs were associated with reduced risk of substance use and psychotic disorders, seizures, neurocognitive disorders (including Alzheimer’s disease and dementia), coagulation disorders, cardiometabolic disorders, infectious illnesses and several respiratory conditions, but increased risk of gastrointestinal issues, hypotension (low blood pressure), and some kidney-related conditions. Overall this is good news. But these findings are “observational” rather than from a randomized trial and thus need to be confirmed in other studies.

Every day seems to bring more promising news about GLP-1 receptor agonist drugs like Ozempic, Wegovy, and Mounjaro. I’ve been following the data on these new drugs with fascination and cautious optimism about their potential for both individual and population health. Beyond their original design for diabetes and the notable side effect of losing significant amounts of weight, clinical trials have shown that GLP-1s reduce the risk of heart attacks, strokes, kidney disease progression, heart failure, and even COVID-19 mortality. While some of these benefits come indirectly via weight loss, GLP-1s seem to also directly impact different organs by reducing inflammation or via other biological pathways.

And you know the hype may be real if the folks who have money on the table predicting people’s lifespans are paying close attention:

Link to article

It’s not fishing, it’s “discovery”

A new study in the journal Nature Medicine has given us the most comprehensive look yet at all the different ways in which GLP-1 drugs might be beneficial beyond their original intention of treating diabetes.

In this study, researchers used a “discovery” approach to look for any associations they could find between GLP-1 receptor agonist drugs and 175 health outcomes. If that sounds like a “fishy” approach, you are right. For most scientific analyses, it is frowned upon to dive into the data searching for any significant association-such an approach may be derided with scathing-for-academia names such as ‘fishing expeditions,” “data mining,” and “p-hacking.”

But why is fishing for “significant” results a bad idea? Basically, about 5% of associations identified as “statistically significant” in conventional analyses will be due to random chance rather than a “real” underlying association. For a more detailed take on statistical significance, please read this post by Dr.

Chana Davis @FueledbyScience

. If you run 175 statistical tests, on average 5% (or about 8 or 9) of these will show up as “significant” associations just by random chance. The real “fishing” problem is when someone runs lots of these tests and then only reports the one “significant” result (like “green jelly beans linked to acne!” below). The results from p-hacking usually fail to replicate in other samples…and then we have all avoided green jelly beans needlessly.

Source: https://xkcd.com/882/

BUT--there are times when this sort of “fishing” can be helpful (and thus rebranded as “discovery”). This is true especially when we know very little about a topic and are looking to “discover” new things that can be further explored. A good example are Genome-wide association studies (GWAS), where researchers search the whole genome for any association (between single nucleotide polymorphisms, or SNPs) and an outcome of interest (height, depression, high cholesterol, cancer, etc). When you are looking across the whole genome and millions of SNPs, a lot of associations will pop up that may or may not be “real” (like looking across a million colors of jellybeans). Genetics researchers know this and set a much stricter threshold for what associations get attention as potentially significant (way lower than a 5% significance threshold), and they also need to replicate their findings in other independent samples before they believe them. But, since the mapping of the human genome is relatively new and we would have no a priori way of knowing which genes are associated with which outcomes, this GWAS “discovery” strategy is hugely valuable.

We’ve studied GLP-1 drugs carefully in the lab and in clinical trials for many years, but only recently have they been used by millions of people, so we still have a lot to learn. So this paper decided to go full-on “discovery” mode, chucking everything at the wall to see what sticks. Besides some surprising new benefits, this approach can also discover rarer negative effects of the new drugs that might not be evident in clinical trials with much smaller sample sizes.

The analysis used data on over 2 million people from the US Veterans Study. Advantages of this dataset include its size and linkage to people’s medical records to follow them over time (a median of 3.7 years follow-up in this study). The downside is that the population may not be representative of the overall population since it is made of veterans (for example, 95% male). The study compared people taking GLP-1s to people prescribed a different diabetes drug. This makes for a more “apples to apples” comparison than just comparing people taking GLP-1s to everyone else, since patients were at least similar enough to need a diabetes drug.

Good News

Overall, using a GLP-1 compared to another diabetes drug was associated with a whole slew of beneficial outcomes. This included reduced risk of substance use and psychotic disorders, seizures, neurocognitive disorders (including Alzheimer’s disease and dementia), coagulation disorders, cardiometabolic disorders, infectious illnesses and several respiratory conditions. All of the conditions in blue below indicate signficant reductions in risk for those taking GLP-1 drugs compared to usual care (indicated by hazard ratios that are less than 1, so to the left of the dotted line).

Source: Xie, Y., Choi, T. & Al-Aly, Z. Mapping the effectiveness and risks of GLP-1 receptor agonists. Nat Med (2025). Figure 6.

Bad News

GLP-1 drugs were associated with increased risks of gastrointestinal disorders and a lot of other things that make sense like nausea and vomiting, headaches, sleep disturbances, hypotension (low blood pressure) and syncope (passing out). Some other conditions that popped up in the negative direction included musculoskeletal disorders like arthritis, and interstitial nephritis (inflammation within the kidneys). These associations are shown in red above, with increased risk reflected in hazard ratios greater than one.

On balance, the beneficial associations for GLP-1 drugs were stronger, more numerous, and generally related to more serious health outcomes than the negative associations. For me, the beneficial links between GLP-1s and reduced-risk of substance use disorders, suicidal ideation, and infections like sepsis and pneumonia were particularly intriguing. Clinical trials to test for effects of GLP-1s on a range of novel outcomes including alcohol use disorder and Alzheimer’s disease are already underway.

Too close to home… Source: The Onion

So overall, these results support the idea that GLP-1 drugs may have multi-system benefits beyond glycemic control, including immune and neuroprotection pathways. Again, this is all very new science. We are still likely discover new downsides of these drugs along the way (not to mention the more practical side of access and cost). But the potential is very exciting.

More than a diabetes drug: Source: Why do obesity drugs seem to treat so many other ailments? Nature

The negative associations “discovered” in this study should be followed up and confirmed in other studies. Besides helping people weigh their own risks and benefits, this could help better tailor new versions of these drugs in the pipeline (and there are many) to hopefully minimize the negative impacts as best possible. But so far it’s good news that nothing very serious or surprising on the negative side is cropping up.

It’s important to keep in mind that this “discovery” means that we can’t have full confidence that these results are “real” effects until they are replicated in other data. Because this study was “observational” rather than a randomized clinical trial, we’re a lot more worried that what we are seeing might be “confounding” rather than a true causal effect. This could happen if the “treatment” and “control” groups are not comparable- for example if doctors prescribed GLP-1s vs another diabetes medication for systematic reasons such as the health profile of the patient. The study authors did a commendable job bolstering the case that this was a good “apples to apples” comparison by first comparing only people getting new diabetes medication, then statistically controlling for a whole range of other sociodemographic factors (age, education, race/ethnicity, area-level income) and MANY baseline health characteristics and health behaviors. The authors also ran something called a “negative” control—checking whether the “treatment” was associated with something it should NOT be, like traffic accidents. If such an association were found, it would be a red flag that the treatment and comparison groups are different for reasons that don’t have to do with GLP-1 drugs (but they found NO association with traffic accidents, so no red flag). All these steps are the researchers’ best attempt to mimic a randomized control trial and make any results more believable as real effects of the drugs. But we can and should still hold on to some healthy skepticism based on this design. Finally, similar to GWAS studies, the authors used a stricter threshold for significance to account for the many statistical tests being done that could find things just by chance (to minimize the green jelly bean problem).

Bottom Line:

Personally, I was excited by the wide-ranging positive “signals” for GLP-1 drugs uncovered by this “discovery” analysis. Many of these positive and negative associations make biological sense based on more basic science, lending credibility to their plausibility. Since these drugs have only been widely prescribed for a few years, longer term follow-up will be important. I remain fascinated and cautiously optimistic about the future of this class of drugs for population health. Stay tuned and watch this space!

Jenn