E302: Do GLP-1 drugs reduce overall healthcare spending?

For many people, the FDA-approved drug class called GLP-1s has been a game changer for managing type 2 diabetes and weight loss. An estimated one in eight US adults is using this type of drug, which mimics a natural gut hormone to regulate blood sugar, slow stomach emptying, and suppress appetite. Researchers and the public have been scrutinizing the pros and cons of taking these drugs for many years. One critical question that remains is does the use of GLP-1s translate into lower medical expenses for consumers, healthcare providers, insurance companies, and the government? Today, we're speaking with Duke University health economist Jonathan Zhang on the answer to these questions from his multi-year study of veterans who use GLP-1s. The results are surprising.

Interview Summary

Jonathan, first, let's talk broadly. Would you briefly describe the kinds of health benefits people are experiencing with GLP-1s, particularly in terms of changes in behavior and overall health?

Yes. So GLP-1s are medications that manage body weight and improve metabolic health. The primary health benefits are via better blood sugar regulation and weight loss via reducing appetite. But there's larger health benefits that are much broader. So, trials have also shown that they lower blood pressure, reduce strokes and heart attacks, among other cardiovascular benefits. Overall also improves, so things like obstructive sleep apnea, improvements in inflammation-related measures. And recently it was FDA approved to treat MASH, which is a form of fatty liver disease. Finally, people also report feeling less hungry. They get full faster, fewer cravings, and reduction in what is often called food noise, which can potentially help people make a range of better behavior decisions and changes relating to food and perhaps even beyond food. However, the evidence on that seems to be promising, but new and less strong.

Thank you for sharing those. I know a lot of our listeners have heard some of those issues and some of those points about how GLP-1s could change the way we behave, and some of the health outcomes. So, let's dig into your particular study. In your working paper for the National Bureau of Economic Research (NBER) you and your co-authors describe a natural experiment involving patients and doctors in the Department of Veteran Affairs, and their access to and use of GLP-1s. Would you lay out the general parameters of your study for our listeners?

Yes, absolutely. So, like many medications, what we know about GLP-1s primarily comes from randomized controlled trials, which have small sample sizes and often select for quite sick patients. As GLP-1s reach a broader population, the population becomes less pro-comparable to those in trials, so the real-world impacts become less obvious. Moreover, there are some outcomes such as healthcare utilization or spending costs that are incredibly important for policy but not directly studied in trials. We were interested in studying in a very large healthcare system, that is the VA, the Veterans Health Administration, the real-world impacts of GLP-1s using readily collected data from the healthcare system and a quasi-experiment. So that is we try to emulate a real experiment where some patients get GLP-1s from their primary care doctor and others do not. And the reason one patient might get it while the other doesn't is because their doctors differ in their likelihood or propensity to prescribe these drugs. And we focus on the Veterans Health Administration, the VHA, because the veteran population is quite diverse, at least in terms of comorbidities relative to the trials per se and also in age, say, compared to Medicare. And GLP-1s are provided in the VA to eligible patients at a pretty affordable cost. So, sort of a nice policy playground for us to think about what outcomes might look like when GLP-1s become readily available and affordable to all Americans.

Great. Thank you for sharing that because now we have a good sense of the quasi experiment that you all were able to exploit. That there were providers who had a high tendency to prescribe GLP-1s versus those who didn't. And looking at that difference, you were able to see how GLP-1s may have affected the outcomes that you were interested in. So now tell us a bit about the findings, and I think there's some surprising results here.

So, first we replicate many of the findings from the randomized trials for this diverse, veteran population. We see reductions in body weight, we see hemoglobin A1C blood glucose levels improve and blood pressure improvements as well. Even a reduction in heart attacks and strokes for patients with existing cardiovascular conditions, and even a small reduction in moderate drinking risk as well. The trial evidence seems to hold up in this large and diverse population of we have 1.4 million patients that we're studying. Second, despite these improvements though, we do not see a reduction in healthcare spending. This is actually excluding the cost of the medication, the GLP-1. It's not because GLP-1s are expensive. We're taking that out of the equation. Over four years, this is how long we're able to track these patients for, we're able to rule out reductions of more than 7% in healthcare spending. And we see no reductions in whether that's outpatient clinic or in inpatient hospitalization. We also don't see any reductions in emergency department visits more generally, despite seeing a reduction in heart attack and strokes for those patients with existing cardiovascular condition.

Jonathan, that sounds great, but it's a little hard to understand. So why is it that you're seeing no real changes in the consumer healthcare spending? What's going on?

Yes, we were a little surprised as well. We think there could be a myriad of potential reasons. But first, I want to note that there's also at least three other studies that use real-world data to look at a similar question, and all the studies find no statistically significant spending reductions. And in fact, all of us, we all find a slight positive increase. I think the first explanation is that while heart attacks and strokes are very expensive and obviously, very serious events, they are still thankfully quite rare and thus are not very common in our three to five-year period in which these studies are looking at. But perhaps over ten or twenty years, you could start seeing some healthcare spending materialize. I think that's totally possible. Most cost projection models, whether that's from the CBO or other organizations, they project healthcare spending reductions to accumulate over time. But these models also project immediate spending reduction, which is not yet supported by the data.

Another potential explanation is what I would call these behavioral responses. So, you can imagine that a patient sees some improvement in their health, and that changes their expectation of, "Oh, wow, I could actually make some changes to my health." And they actually go and get some more care. This can also be driven by the physician as well. So, sort of, "Look, things are working. This patient's seeing improvements. Maybe we should treat some of your other health problems and issues." This is what I would call sort of a behavioral response. A third explanation could be that some of these patients have conditions that have progressed to be irreversible even with metabolic health gains. So, for example, late-stage kidney disease would be a good example. We find some suggestive evidence that this might be happening in our study. In other words, for these patients where they have some potentially irreversible health conditions, the metabolic health improvements - that's not a primary reason for their healthcare spending.

This is helpful. That explains why we're seeing some of these challenges with healthcare spending. And I appreciate your point of taking a longer view may actually reveal some differences. And also, the responses of doctors and patients to the benefits that they gain and how that might expand opportunities for healthcare. So, thank you for sharing that. I want to shift focus a little bit. And it is taking a step beyond the paper, but I think it's a natural extension of the work. As you know, I'm an agricultural economist, and I've been interested in obviously food policy issues. And the fact that we're seeing weight loss, the fact that we're seeing better management of A1C suggests that there's something probably happening with food consumption and food patterns. I'd be interested to learn from you if you have any ideas about what this work suggests about the impact of GLP-1s on consumer food choices.

Yes, I think there's a fair amount of both trial and real-world grocery purchase evidence that GLP-1s are impacting consumer food choices. In the randomized trials, GLP-1s reduce, calories consumed, reduce the desire to eat sweet, salty, and savory foods and high fat fast foods in their meals. For example, these trials will look at what people eat for lunch. In real world data, there's a study out of Denmark that found that after patients' first GLP-1 prescription, their grocery store purchases went on to have fewer calories, sugars, saturated fats, carbohydrates, and more protein over the one-year period after they initiate GLP-1s. The share of ultra-processed foods also decreased. The effects were actually quite modest, so there was only about a one percent reduction in calories and a four percent reduction in sugar and a five percent increase in protein in the year after versus the year before in this Danish study. In the US, there's a similar study using kind of this transactional level data as well. And they find that households after they start GLP-1s also reduce caloric intake, fast food consumption, caloric dense processed foods. And there it's a little larger in the US by about five to ten percent. Then the question is, you know, are they shifting towards healthier foods or is it sort of a uniform reduction in calories across all different food classes? And there the evidence is a little less strong. There's some evidence that consumers are shifting towards healthier foods, but if that is the case, the shift is quite small. In both the Danish and the US study, it's roughly looking like a level shift for all different categories. And this is also in terms of not food so much anymore, but there's also been studies looking at alcohol consumption, which has been very interesting and promising. There's a phase two study pilot trial with only forty-eight participants, but the randomized control trial found that among patients with existing alcohol use disorder GLP-1s led to a reduction in the average number of drinks per day. No change in the number of days drinking, but conditional on drinking, the average number of drinks per day declines in the treatment group that get GLP-1s. And then for out of the 48 patients that are also smokers, there's also a reduction in smoking as well in that population. So, very promising early evidence there as well.

Jonathan, this is critical information, and I know some colleagues in the Ag econ profession, folks like Brian Roe, Tim Richards and Glenn Townsend have also worked in this space, and some of the work is showing some issues around the quality of protein or overall food spend is going down. And the choices or the responsiveness of people to prices when it comes to proteins. There's some evidence, some causal evidence to suggest GLP-1s are changing the way people are even looking at food markets. And so that raises a really important issue. If consumers are changing when they're exposed to GLP-1s, I have got to imagine that the food industry, grocery stores, food manufacturers, restaurants are making some adjustments. What have you to say on that particular point?

So, we're both economists, right? We both, of course, believe that producers and grocery stores are adjusting to demand. We do know a lot less about how the supply side is adjusting compared to the demand, and that's because there's just less data and fewer studies on patients. There's more data and more studies on patients and consumers than on grocery stores and restaurants. I've seen a lot of anecdotal evidence, often reported in the mainstream media or reports from various market research agencies that GLP-1s are having this large impact on the food industry. I've seen claims like GLP-1s are hurting Walmart sales or why are beer company sales down, like AB InBev, and range into things like why restaurants are reducing their portions or recently I actually also saw over the past couple weekends claims of why Coachella portions are smaller because everyone's on GLP-1s. So, the best estimates from last year are that one in eight Americans are using GLP-1s. And if each of these users are cutting back in consumption by 5% to 10%, as in those studies, it's totally plausible this leads to a large population reduction that the food industry just cannot ignore, right? I think it's important to note a couple things. First is that the long-term adherence of GLP-1s is generally quite low, with more than 50% discontinuing within a year in most of the studies I've seen in the US. And after discontinuation, most of these patients will gain back a large fraction of their weight. And in that US groceries transactional study I told you about, they find that after discontinuation, purchases at fast food, grocery, restaurants, et cetera, they go up beyond baseline, in fact. So, beyond six months prior to when they were on GLP-1s. It's still an open question to whether these 5% to 10% individual consumption reductions will become long-term and manifest in the population level, especially if adherence is not very high. Second is that up until very recently, there's been a very strong demographic and socioeconomic profile of these GLP-1 users, right? So, they tend to skew near elderly and also a little bit more female, and generally fall in the middle- or higher-income brackets because previously Medicaid, for example, did not cover GLP-1s, right? Until earlier this year. And it still does not cover it for many states. So, any food industry changes will likely be concentrated in the restaurants and the food chains that sort of have these consumer demographic and economic base. That's something to keep in mind. Certainly, there are examples like, you know, Smoothie King has introduced a GLP-1 smoothie. But even in these clear-cut cases where the producer is clearly trying to appeal to this demand shock, it's still very difficult to disentangle the producer's stated intent from general industry market trends, right? So, general trends like rising cost of living. There's been more demand for protein. You can see protein in everything now, right? Even water. Rise in sort of small plates or family-style tapas at restaurants. Feels like every restaurant's doing this now. Or for example, a decline in alcohol consumption or interest in alcohol consumption, especially among young people around the world. I think some of this is certainly happening. Producers are listening, but it's hard to disentangle the general trends from a specific GLP-1 impact.

I'm grateful for you taking some of the hype out of this and thinking through carefully what the economic implications are of this change and recognizing that there's some real important context that people don't stay on these drugs for long. Or there are certain types of folks who are more likely to use them, and therefore we can't assume that there's this massive wave of change that's going to happen, but that there are some responses. These are important ways of framing this discussion. I have one final question. Based on what you're seeing now with the use of GLP-1s and considering their use over time, what would you say is the key takeaway for policymakers?

I think there are a few takeaways. Most importantly, I think policy makers should think about the rapid change in profile of patients who are using this drug. Just because it's beneficial for sick patients doesn't necessarily mean it's going to be equally as beneficial for patients who are less sick. One of the findings in our paper is that as the veteran population that starts to initiate GLP-1s in the more recent years, these patients actually are healthier, so their treatment effects are not quite as strong as those who were initiated in the very early period who in our setting were among the sicker veterans. Policy makers shouldn't extrapolate the potential benefits to a broader population since there's so much demand for this drug, right? This also means that some of the downstream improvements like cost savings or even improved labor outcomes like absenteeism or being able to return to work, while these things might seem logical, it's not fair to assume that it's going to be automatic, especially at a population level. Related to that last point, we should encourage additional research on important socioeconomic outcome that might completely change the calculus of the cost benefit of these drugs, right? So, labor force participation, earnings, use of government transfer programs like disability insurance, even some interesting patterns of household formation and fertility. These are all fascinating and important questions. However, it is imperative to note that GLP-1s already have incredible health benefits for many people. While it's possible that these benefits might spill over to broader societal benefits, these benefits should not be presumed. And at the same time, we shouldn't be holding GLP-1s to a higher standard, say, to reduce healthcare spending or increase earnings just because we know that it does a bunch of wonderful things for a range of conditions. So, maybe it does this and this as well, and suddenly we're less excited if there's no evidence of that and that, right? These things have incredible health benefits. Let's appreciate them for that, and we shouldn't be holding them up to a higher standard just because it appears to do so many wonderful things.

BIO

Jonathan Zhang is a health economist at the Sanford School of Public Policy at Duke University and the Margolis Institute for Health Policy. He is a Faculty Research Fellow at the National Bureau of Economic Research and an evaluator at the Department of Veterans Affairs. He researches how policy can improve mental health and substance use outcomes using quasi-experimental methods.