This , published today in the BMJ, found starting a GLP-1 drug was linked with a 14% overall reduced risk of developing new substance use disorders, including alcohol, cannabis, cocaine, nicotine and opioids. Among people with an existing substance use disorder, taking a GLP-1 was associated with a 26% reduction in substance-related hospital admissions.
Researchers examined electronic health records from more than 600,000 veterans with diabetes who were treated through the United States Department of Veterans Affairs.
Researchers compared those newly prescribed a GLP-1 with those started on a different class of diabetes medication called SGLT2 inhibitors (including empagliflozin and dapagliflozin) – a well-established treatment used as a comparison point.
The study followed participants for up to three years, asking two questions:
The researchers used a method called “target trial emulation,” which structures an observational study to resemble a randomised controlled trial as closely as possible.
In a randomised controlled trial, participants are randomly assigned to receive either the drug being tested or a comparison treatment. The two groups should be similar in every way except for the treatment they receive. If one group does better, we can be confident the drug caused it.
Observational studies work differently. No matter how carefully researchers try to account for differences such as weight, age and other health conditions, there is always the possibility that some unmeasured factor explains the results.
The target trial emulation design used here is among the best available approaches for observational data, but it cannot eliminate this problem. It can tell us that something is with better outcomes; it cannot prove that the drug those outcomes.
With that caveat in mind, the results were notable. Among people without a prior substance use disorder, those on GLP-1 drugs were less likely to develop one across every substance category examined:
This amounted to roughly 1–6 fewer cases per 1,000 people over three years.
For those who already had a substance use disorder, those prescribed GLP-1 drugs had better outcomes across every measure:
This amounted to around 1–10 fewer events per 1,000 people over three years.
That these patterns held across multiple substances and multiple outcomes makes them harder to dismiss.
But they remain associations, not proof. The ongoing randomised trials will be essential for determining whether GLP-1 drugs genuinely cause these benefits, or whether something else is at work.
The cohort was 90% male with an average age of 65, so findings may not extend to women, younger people, or those without type 2 diabetes.
The group also had significant health complexity. More than half (57%) were current or former smokers, over 40% had high cholesterol, and many had additional conditions including high blood pressure, heart disease and heart failure.
Mental health conditions were also common – more than 18% had post-traumatic stress disorder (PTSD), over 10% had depression and over 10% had anxiety.
We also don’t know whether participants were receiving any treatment for their substance use disorder, which could itself influence outcomes.
Perhaps the most important takeaway isn’t about GLP-1 drugs at all. Substance use disorders are highly treatable.
Effective, evidence-based medications already exist – naltrexone and acamprosate for alcohol, methadone and buprenorphine for opioids – alongside a wide range of psychological therapies.
These treatments are and , yet only a small fraction of people who could benefit from them ever receive them. An estimated are ever prescribed effective medication.
The biggest barrier isn’t availability: it’s stigma, shame, fear of judgment and discrimination. Society still views addiction as a moral failing rather than a health condition.
For people living with a substance use disorder, this research on GLP-1s is encouraging but the more immediate message is that effective treatments are already available.
, Addiction Psychiatrist, Associate Professor,
This article is republished from under a Creative Commons license. Read the .
