Research desk · Medications your clients take
Semaglutide and tirzepatide are now part of many clients’ lives, and the questions are landing in your office. Does it cause depression? Is “food noise” real? What about drinking? Here is what the newest evidence says, and where it stops.
GLP-1 receptor agonists started as diabetes medications and became the most talked-about weight-loss drugs in a generation. In practice that means a growing share of your caseload is on one, thinking about one, or living with someone who is. Online, two phrases keep trending: “food noise,” the constant mental chatter about eating that many users say goes quiet, and the “Ozempic personality,” the worry that the drug flattens pleasure along with appetite. Neither is a clinical term. Both are things clients will say to you.
For a while the worry ran the other way. In 2023, postmarketing reports of suicidal thoughts led regulators to investigate. That story has now largely turned. This page lays out the safety evidence first, then the more surprising findings on mood, substance use, and motivation, then what all of it means for your sessions.
On January 13, 2026, the FDA asked the makers of Saxenda (liraglutide), Wegovy (semaglutide), and Zepbound (tirzepatide) to remove the suicidal behavior and ideation language from the Warnings and Precautions section of their labels. The agency described a meta-analysis of 91 placebo-controlled trials with 107,910 patients that showed no increased risk of suicidal ideation or behavior. It also showed no increased risk of anxiety, depression, irritability, or psychosis. A separate Sentinel claims study of more than 2.2 million new users with type 2 diabetes found no increased risk of intentional self-harm compared with SGLT2 inhibitors. The FDA also told clinicians that anyone who discloses suicidal thoughts should be referred for mental health evaluation. That’s you.
Independent work points the same way. A 2025 JAMA Psychiatry meta-analysis pooled the 27 RCTs that systematically recorded suicide and self-harm events, covering more than 59,000 participants. [1] Event rates were very low in both groups: 0.047 per 100 person-years with a GLP-1 drug and 0.042 with placebo. The difference was not significant (rate ratio 0.76, 95% CI 0.48–1.21). A 2026 review by NIH researchers summarized the field this way: despite early pharmacovigilance alarms, GLP-1 drugs “do not seem to cause or increase the risk of psychopathology.” [2]
Two cautions keep this honest. These trials enrolled people for diabetes and weight studies, and those participants may not look like your caseload. And “no increased risk on average” never rules out an individual reaction. That is why the FDA’s own advice to patients still says to report new or worsening depression, suicidal thoughts, or unusual mood changes.
Several large 2026 studies go further than “not harmful.” Read them carefully: they are observational, so they show association, not proof.
Depression and anxiety. A Swedish national cohort followed 95,490 adults with depression or anxiety who were using diabetes medications. It used a within-individual design that compares the same person’s periods on and off a drug. [3] Compared with non-use, semaglutide was associated with a lower risk of worsening mental illness (aHR 0.58, 95% CI 0.51–0.65), and liraglutide with a smaller reduction (0.82). Exenatide and dulaglutide showed no association (both about 1.01). For semaglutide specifically, risks of worsening depression (0.56), anxiety (0.62), and substance use disorder (0.53) were all lower. As a class, GLP-1 drugs were associated with lower self-harm risk (0.56). The drug-by-drug differences are a useful reminder that “GLP-1” is not one thing.
Substance use. A 2026 BMJ cohort of US veterans with type 2 diabetes compared people starting a GLP-1 drug with people starting an SGLT2 inhibitor. [4] Starting a GLP-1 drug was associated with lower risk of new alcohol use disorder (HR 0.82), opioid use disorder (0.75), and the composite of all incident substance use disorders (0.86). Among people who already had a substance use disorder, it was also associated with fewer drug overdoses (0.61) and fewer instances of suicidal ideation or attempt (0.75). The NIH review notes that randomized trials for addiction so far have been “mixed” but promising overall, with more underway. [2]
Serious mental illness. In a US cohort of 70,257 adults with schizophrenia and type 2 diabetes, starting a GLP-1 drug was associated with fewer psychiatric hospitalizations than not starting one (weighted HR 0.76). [5] A JAMA Psychiatry target-trial emulation found lower 4-year mortality among people with serious mental illness who started a GLP-1 drug instead of an SGLT2 inhibitor (HR 0.76; absolute risk difference −1.54 percentage points). [6] For a population whose life expectancy is shortened by cardiometabolic disease, that matters.
The “Ozempic personality” worry is that a drug that dampens reward around food might dampen reward everywhere. The best direct evidence so far comes from one small randomized trial, and it points the other way on lab measures.
In a double-blind RCT of 72 adults with major depressive disorder and a BMI of 25 or higher, participants received oral semaglutide 14 mg or placebo for 16 weeks. [7] In a secondary analysis using an effort-for-reward task, semaglutide reduced effort discounting: people became more willing to work for larger rewards (effort sensitivity β = −1.737, P = .03). The authors described this as improved measures of motivation. A second analysis of the same trial found shifts in how participants balanced exploring new options against exploiting known ones. [8]
That is genuinely interesting, but it is a lab task in 72 people. It is not proof that clients will feel more motivated, and it does not mean nobody feels flat. Treat the client’s report as data. If someone says their appetite for food went quiet and so did their appetite for everything else, explore it, track it, and loop in the prescriber.
“Food noise” deserves the same respect. Many clients describe the relief of a quieter mind as life-changing. For some, it also exposes what eating was doing for them emotionally, and that gap shows up in therapy.
A way to open the topic without steering it:
“Since you started the medication, what has changed for you besides your weight: mood, sleep, cravings, how you feel about food, anything else?”
And when a client asks you directly whether it causes depression:
“The large reviews so far haven’t found that, and the FDA removed its warning this year. But you’re the expert on you. If your mood shifts, let’s track it together and make sure your prescriber knows.”
Some physical effects can show up in session looking like mood or anxiety problems. A 2026 umbrella review of 60 meta-analyses found the most consistent signals were gastrointestinal: higher odds of nausea (OR 2.47), vomiting (2.78), and diarrhea (1.94). [9] Persistent nausea and poor intake can show up as fatigue, irritability, low mood, or social withdrawal. Before you build a formulation around a new low mood, ask about the body. The same review found most other non-cardiometabolic outcomes rested on lower-certainty evidence.
Each card links to the PubMed record. Numbers are reported the way the authors reported them. Read the abstract before you quote any of them.
Ebrahimi P, Batlle JC, Ayati A et al.
27 RCTs that recorded suicide and self-harm events. Very low event rates: 0.047 vs 0.042 per 100 person-years (GLP-1 vs placebo); rate ratio 0.76 (95% CI 0.48–1.21), not significant.
Farokhnia M, Leggio L
GLP-1 therapies for addiction and mental health. Preclinical and observational evidence supports study for alcohol and substance use disorders; RCTs mixed but promising. The authors write that GLP-1 drugs “do not seem to cause or increase the risk of psychopathology.”
Taipale H, Taylor M, Lähteenvuo M et al.
95,490 people in Sweden with depression or anxiety. Worsening mental illness: semaglutide aHR 0.58 (0.51–0.65), liraglutide 0.82; exenatide and dulaglutide about 1.01. Semaglutide: depression 0.56, anxiety 0.62, SUD 0.53. Class self-harm 0.56.
Cai M, Choi T, Xie Y et al.
Type 2 diabetes, GLP-1 drug vs SGLT2 inhibitor. Incident alcohol use disorder HR 0.82; opioid 0.75; all SUDs 0.86. In people with existing SUD: overdose 0.61; suicidal ideation or attempt 0.75.
Davoodian N, Williams LJ, Berk M et al.
70,257 adults with schizophrenia and type 2 diabetes. GLP-1 initiation was associated with lower psychiatric hospitalization than non-initiation (weighted HR 0.76, 0.71–0.82). The authors note possible residual confounding.
McIntyre RS, Zhang-James Y, Kwan ATH
Serious mental illness, GLP-1 drug vs SGLT2 inhibitor. 4-year mortality HR 0.76 (0.74–0.78); absolute risk difference −1.54 percentage points.
Gill H, Badulescu S, Shah H et al.
72 adults with MDD and BMI ≥ 25; oral semaglutide 14 mg vs placebo for 16 weeks. Reduced effort discounting on an effort-for-reward task (effort sensitivity β −1.737, P = .03), which the authors interpret as improved motivation.
Zhang Z, Chen N, Wang D et al.
Same trial. Semaglutide changed explore/exploit decision-making (more exploitation and directed exploration). These effects were independent of weight loss.
Yang K, Liu C, Guo Q et al.
60 meta-analyses, 1,751 RCTs. Most consistent signals were GI: nausea OR 2.47, vomiting 2.78, diarrhea 1.94. Most other non-cardiometabolic outcomes were lower certainty.
Education for clinicians, not medical or legal advice. This page summarizes published research for licensed mental health professionals. It is not a treatment recommendation for any individual, and it does not replace your own clinical judgment, your licensing board’s rules, or consultation with a prescriber or attorney. Evidence reviewed through October 24, 2026. Spot an error? Tell us and we will fix it.
Keep reading