The GLP-1 receptor agonists (GLP-1-RAs—drugs like semaglutide/Ozempic®/Wegovy® and the older liraglutide) have repeatedly surprised us. They were originally developed to treat type 2 diabetes (T2D) on the basis that they would wake up a diabetic patient’s recalcitrant pancreas to produce insulin quickly in response to a meal. But researchers soon discovered that even the early GLP-1-RAs led to weight loss. Through intentional engineering, researchers developed newer drugs with enhanced weight-loss effects, to the point that they’re now best known for their ability to curb appetite, reduce “food noise,” and help people shed pounds.
But that was only the beginning of the surprises. Recent trials have shown that semaglutide reduces the risk of heart attack, stroke, and death from atherosclerotic cardiovascular disease (ASCVD) in nondiabetic obese people1, and it slows the worsening of kidney function in nondiabetic obese people with ASCVD2 and in people with obesity, diabetes, and existing kidney disease.3 We were doubly surprised when, after having first ventured that the effects of semaglutide on ASCVD were a function of the weight loss, a subsequent analysis found that weight loss or even change in waist circumference could only explain a small fraction of the effect (though the investigators didn’t have direct measures of how much total or visceral adipose tissue the subjects lost).
Even more unexpectedly, there is now evidence suggesting that GLP-1-RAs can reduce addictive behavior, most notably alcohol use disorder19,20 but with more preliminary evidence for tobacco use and anecdotal accounts of gambling.
And then came the hints that these drugs might have an even more impactful and equally unanticipated benefit. Observational studies, followup studies in animal models, and above all, post hoc analyses of the earlier randomized controlled trials (RCTs) of semaglutide and other GLP-1-RAs all suggested that these drugs might keep people from developing dementia.
The most powerful of these lines of evidence was the RCT analyses. One analysis pooled together three major trials (LEADER, SUSTAIN 6, and PIONEER 6) that were originally designed to test whether GLP-1-RAs would prevent ASCVD outcomes in people with T2D. They did—but when researchers came back to the data and looked to see which patients developed some form of dementia, they found that people who were treated with GLP-1-RAs were less than half as likely to develop some form of dementia as were those who had received the placebo (hazard ratio [HR]: 0.47 (95% confidence interval [CI]: 0.25-0.86)).4
Because these patients were randomly assigned to receive a GLP-1-RA drug or a placebo, these results are much more compelling than those of a typical observational study, where pre-existing characteristics might lead some people to receive one class of drug or another, thereby biasing the eventual difference in dementia risk. For instance, people with more education might be more informed about GLP-1-RAs and opt for them, or people of higher socioeconomic status might have better insurance that would cover the drugs or be able to pay for them out of pocket. Since higher education and socioeconomic status are both themselves associated with lower risk of dementia, it would be very difficult to disentangle these effects from the pure effect of the drugs in a typical observational study of dementia risk in GLP-1-RA users versus nonusers. That’s what made the analyses of RCTs powerful.
If these results are real—if GLP-1-RAs really are potent prophylaxes against what for many people is the most fearsome of the Four Horsemen—what could explain them? Researchers formulated multiple hypotheses and tested several of them in animal models of AD. The good news is that the scramble for explanations led to the identification of a large number of potential mechanisms by which GLP-1-RAs might prevent dementia. Different studies support benefits driven by reduced inflammation and oxidative stress (either in the brain or in the periphery), reduced aggregation of beta-amyloid and tau (the signature molecular damage in AD), stimulating the proliferation of neural stem cells, enhancing brain glucose metabolism (which falls in the AD brain, starting decades before dementia sets in), and/or by preserving the integrity of the blood vessels, the neurons, the connections between neurons, or the protective blood-brain barrier (BBB) (reviewed in references11,23). Some of these mechanisms were even supported in preliminary clinical trials.8,9
The bad news is that when different research groups collectively finger almost every single mechanism you can think of, it’s hard to take any of them very seriously, or to choose which among them is most likely a key mechanism on which investigators should focus further research.
And whichever of these mechanisms is actually driving the human observations (assuming that the observations discussed above are real, causal effects), they could be mediated by the direct effects of the drugs on these tissues, since there are receptors for GLP-1 in much of the brain, or they might be indirect effects mediated by improving the animals’ metabolism outside of the brain, since semaglutide5,6 and liraglutide5,7 have access to only a limited number of areas of the brain, most of which are related to hunger and other primitive drives that are not obviously related to AD or dementia.
