Study importance

What is already known?

  • Incretin-based weight loss medications are highly efficacious for weight loss in the general population without diabetes. However, individuals prescribed second generation antipsychotics (SGA) were excluded from published trials.

What does this study add?

  • Effectiveness of incretin-based weight loss medications is lower among those prescribed high weight-inducing SGAs, relative to those prescribed low weight-inducing SGAs.

How might these results change the direction of research or the focus of clinical practice?

  • These findings may encourage primary care practitioners and psychiatrists to monitor weight changes, especially among those prescribed high weight-inducing SGAs, for promotion of additional weight management strategies.

Background

One in 18 adults in the United States have serious mental illnesses like schizophrenia, bipolar disorder, major depression, and schizoaffective disorder [1]. Second generation antipsychotics (SGA) are efficacious in reducing psychiatric symptoms of these illnesses with a lower risk of extrapyramidal symptoms than first generation agents [2]. However, antipsychotic induced weight gain (AIWG) from SGAs, as a consequence of changes in metabolism and appetite regulation, lead to increased cardiovascular risk [3]. Current pharmacological interventions (e.g., metformin, H2 antagonists) for management of AIWG require further evidence before widespread adoption [4].

Incretin mimetics like Glucagon-like Peptide Receptor Agonists (GLP1-RA) and dual GLP-1 RA/Gastric Inhibitory Peptide were approved for weekly dosage by the US Food and Drug Administration for weight management in June 2021 (semaglutide) and December 2022 (tirzepatide) [5, 6]. To be eligible for semaglutide/tirzepatide, patients needed to have either obesity (body mass index [BMI] ≥30 kg/m2), or overweight (BMI ≥ 27 kg/m2) and at least one weight related condition (e.g., high blood pressure, hyperlipidemia) [5, 6].

Published phase 3 trials of incretin-based medications for weight loss excluded individuals prescribed SGAs [7]. However, previous studies of liraglutide and exenatide suggest potential benefits among those prescribed SGA without worsening of psychiatric symptoms [8,9,10]. We hypothesized that incretin mimetics may be helpful in weight management for this vulnerable population. Therefore, our objective was to study whether prescription of semaglutide or tirzepatide was associated with heterogeneity in weight change, by categories of AIWG from SGA, among guideline-based eligible adults with overweight or obesity and without diabetes.

Methods and findings

Study design

We used longitudinal data from the Epic Cosmos Research Platform extracted between June 2021 and December 2023 [11]. This study included data regarding adult patients (18–99 years) from United States without type 1, type 2, or other forms of diabetes who were prescribed a second-generation antipsychotic (Supplementary Table 1) [12]. These patients were also concurrently eligible for incretin-based weight loss medications based on BMI and cardiometabolic comorbidities. We additionally restricted our analysis to those active patients who had at least one inpatient or outpatient encounter in each of the 2 years before prescription of the SGA, and not prescribed other weight loss medications (liraglutide, phentermine, phentermine/topiramate, naltrexone/bupropion, lorcaserin, orlistat) or metformin for weight management in the previous 3 months. Data extraction was conducted between 4 June 2024 and 6 July 2024. Analytic sample selection is presented in Supplementary Fig. 1.

Variables

Exposure

Under an intention-to-treat framework, the exposed group consisted of encounters from patients who were prescribed incretin-based weight loss medications in the calendar month of, or within 2 months, after being prescribed an SGA. We used the 2-month window to reduce immortal time bias between SGA prescription and assessment of weight [13]. The unexposed group consisted of patients who were not prescribed incretin-based weight loss medications. We used nearest neighbor propensity scores, matching each exposed patient to four unexposed patients.

Outcome

The primary outcomes were changes in weight, percentage weight loss, and achievement of 5% weight loss from baseline. We extracted the earliest record of weight (kilograms) in the period between 6 and 9 months following the SGA prescription. A framework for the analysis is provided in Supplementary Fig. 2.

Covariates

Effect modification was assessed after categorizing the index SGA by expected AIWG as high (+++; olanzapine, clozapine), intermediate (++; risperidone, quetiapine, paliperidone) and low (+; aripiprazole, asenapine, brexpiprazole, cariprazine, iloperidone, lumateperone, lurasidone, ziprasidone) [2].

Age, biological sex, race-ethnicity, and residential location were extracted from the latest available records of each patient. Percentile ranking for the Social Vulnerability Index 2020 data and the Rural-Urban Commuting Area 2010 primary codes from the Center for Disease Control & Prevention and US Department of Agriculture respectively, based on the zip code of residence. We extracted insurance status and history of comorbidities and prescriptions based on ICD-10-CM codes and Anatomical Therapeutical Classification codes prior to the date of inclusion (Supplementary Table 1).

Statistical analysis

Continuous variables were described using means (standard deviations) or medians (quartiles) as appropriate. Categorical variables were described using frequencies and percentages. Covariate balance after propensity score matching were assessed as standardized mean differences less than 20% [14]. Given that 35% of encounters did not have a weight measurement in the 6-to-9-month window when outcome was assessed, we constructed inverse probability weights for loss to follow-up [15]. We fitted marginal structural models to study the association of being prescribed incretin-based weight loss medications with changes in weight (kg), changes in percentage weight (percentage points), and achievement of at least 5% weight loss (risk ratio). We specify the target trial emulation design in Supplementary Table 2.

We conducted several sensitivity analyses to assess the robustness of our results. First, we studied if length of prescription of incretin-based weight loss medications was associated with differences in weight loss. Second, we studied if results differed after excluding those older than 75 years, by sex, or among patients with history of affective or psychosis disorders. Third, we studied if results were similar by incretin-based weight loss medication, relative to those prescribed neither.

Results

The matched analytic sample consisted of 66,574 encounters when SGAs with 56% low, 35% intermediate, and 8% expected AIWG were prescribed, of which 13,319 involved prescription of semaglutide or tirzepatide (Table 1). The patients were on average aged 47.1 years (SD: 15.8), 83% female, 74% Non-Hispanic White, 13% Non-Hispanic Black and 6% Hispanic, with an average BMI of 39.5 kg/m2 (SD: 7.7). Additional descriptive characteristics on comorbidities and medications are presented in Supplementary Table 4. The geographic distribution of the analytic sample is presented in Supplementary Fig. 3.

Table 1 Socio-demographic and clinical characteristics of analytic sample.
Full size table

Follow-up information on body weight was available for 43,478 encounters, with 35.1% achieving at least 5% weight loss at 6 months among those prescribed semaglutide/tirzepatide, relative to 19.4% who were not. Encounters for which follow-up were available were similar to those lost to follow-up (Supplementary Table 4).

Compared to those who did not receive a prescription for incretin-based weight loss medication, receiving a prescription was associated with −2.80 kg (95% CI: −3.02, −2.58) and −2.72 percentage points (95% CI: −2.92, −2.51) decrease in weight. The latter were also 1.91 (Risk Ratio; 95% CI: 1.84, 1.99) times more likely to lose at least 5% of baseline weight. Results did not change after adjusting for all baseline covariates. Weight loss was greatest among patients prescribed low risk SGAs (−3.02 kg, 95% CI: −3.29, −2.74), followed by intermediate risk SGAs (−2.77 kg, 95% CI: −3.05, −2.40) and was the least for those prescribed high-risk SGAs (−1.33 kg, 95% CI: −2.04, −0.62) (Fig. 1).

Fig. 1: Weight change 6 months after prescription of incretin-based weight loss medications by categories of second-generation antipsychotics.
figure 1

Associations (95% robust confidence intervals) are adjusted for baseline weight and duration of follow-up for prescription of incretin-based weight loss medications (semaglutide or tirzepatide). Second-generation antipsychotics were categorized based on their propensity for inducing weight gain as high (olanzapine, clozapine) intermediate (risperidone, quetiapine, paliperidone) and low (others). Panel A Weight (kg), Panel B Weight (percentage points), Panel C Risk ratio for 5% weight loss.

Full size image

Results did not differ by length of prescription (Supplementary Table 5), after excluding patients aged 75 years or older (Supplementary Table 6), and by sex (Supplementary Table 7). Weight loss was lower among individuals with history of psychosis disorders, compared to affective disorders (Supplementary Table 8). Prescription of semaglutide and tirzepatide were associated with −2.50 kg (95% CI: −2.73, −2.28) and −4.42 kg (95% CI: −4.97, −3.87) decrease in weight respectively (Supplementary Table 9).

Discussion

In an analysis of the largest integrated database of electronic health records nationwide, we showed that prescription of semaglutide and tirzepatide was associated with clinically meaningful weight loss among individuals prescribed SGAs in 6 months. The magnitude of weight loss was the least among those receiving clozapine and olanzapine. Expected weight loss was higher among those prescribed tirzepatide, relative to semaglutide, consistent with randomized trials that did not include individuals prescribed SGAs [16].

We note that the magnitude of weight loss in this study was less than that observed in the general population with obesity or overweight from the randomized trials of incretin-based weight loss medications with lifestyle that report roughly 8–16% average weight loss in 6 months after initiation [16, 17]. However, our results are consistent with real-world effectiveness studies of these medications among people with and without type 2 diabetes that suggest lower rates of weight loss [18, 19]. There are several possible explanations for lower rates of weight loss in the study population such as counterregulatory metabolic effect of SGAs, insufficient access, shorter length of prescription, low adherence in our study population, and decreased effectiveness of from other obesogenic medications like antidepressants [20]. It is important to recognize that patients prescribed high AIWG-risk SGAs like clozapine can achieve meaningful weight loss in the context of a controlled research setting with close monitoring and structured support, as demonstrated recently in the COaST trial (n = 31) [21]. However, translating these outcomes to real-world practice may be more challenging.

This study has several strengths such as the large sample size and geographic representation of SMI cases. However, there were limitations. First, we did not have data on filled prescriptions and therefore do not know if the study population initiated or continued the prescription for the duration of the study, which may partially explain the lower observed weight loss compared to phase 3 trials. We did not have information on diet and physical activity, which may be worse relative to the general population, and may also explain the lower observed weight loss compared to trials. We did include the information on the dosage given shortages during the period of analysis. Second, given the nature of EHR data, the results are susceptible to information bias. Less than half of healthcare systems in the United States use Epic software and patients may have received incretin-based weight loss medication prescriptions from other health providers that do not contribute data to Epic Cosmos. Third, the analysis is also susceptible to confounding by indication such that patients prescribed SGAs resulting in high AIWG may have more complex clinical presentations. However, the stratified analyses by history of affective and psychosis disorders were consistent with the main analysis. Finally, although we used appropriate statistical approaches to account for loss to follow-up, nearly 40% of the analytic sample had missing data on body weight. However, this is unlikely to bias our results given similarities in patient characteristics by follow-up.

In summary, semaglutide and tirzepatide may be viable options for ameliorating weight gain in this vulnerable population, However, the expected weight loss from these medications may vary according to specific antipsychotic therapies.