Comparative efficacy and safety of liraglutide versus metformin, naltrexone/bupropion, and phentermine-topiramate in psychiatric patients

Abstract

Background:

Psychiatric patients have a high risk of obesity, frequently due to psychotropic medication-induced weight gain. However, real-world comparative data on antiobesity medications (AOMs) in this population remain rare.

Objectives:

To compare short-term weight-loss efficacy, adverse events (AEs), and early discontinuation (ED) among psychiatric outpatients taking liraglutide (LIRA), naltrexone/bupropion, phentermine-topiramate (PT), or metformin (MET).

Design:

Retrospective observational cohort study.

Methods:

We conducted a 12-week retrospective chart review of 117 psychiatric outpatients with International Classification of Diseases, 10th Revision, F01–F99 diagnoses. Percent weight change over time was analyzed using linear mixed-effects models. AEs and ED were compared across treatment groups.

Results:

Compared with MET, LIRA was associated with a greater percent weight reduction (estimate −3.45%, 95% confidence interval (CI) −5.35 to −1.55, p < 0.001), with a significant treatment-by-time interaction at 12 weeks (p = 0.019). Female sex and full-time employment were associated with attenuated weight loss, and the number of concomitant psychotropic medications with moderate weight-gain risk showed a trend toward greater weight reduction (p = 0.066). No significant differences were observed in AE incidence across AOMs. ED rates differed by drug type (p = 0.017), being lowest in the MET group (39.1%) and highest in the PT group (72.2%).

Conclusion:

In this real-world psychiatric cohort, LIRA was associated with greater short-term weight loss than MET without an increased observed frequency of AEs. ED rates varied across AOMs. These findings should be interpreted cautiously, given the observational design and short follow-up period, and require confirmation in larger, long-term studies.

Plain language summary

Liraglutide may offer a safe and effective weight-management strategy for patients with psychiatric disorders

Many individuals taking psychotropic medications experience weight gain, which can impact both physical and mental health. In this retrospective study, we reviewed medical records of 140 psychiatric outpatients treated with one of four weight-loss medications: liraglutide, metformin, naltrexone/bupropion, or phentermine/topiramate. Our study suggests that liraglutide was associated with greater weight loss compared to the other medications. Regarding safety, the occurrence of adverse events was generally similar across all groups, and importantly, no serious psychiatric adverse events were observed in this study. These findings indicate that GLP-1 receptor agonists, including liraglutide, may be a helpful option for weight management in this population. However, because this study could not fully account for factors such as medication cost or lifestyle habits, these results should be interpreted with caution. Larger, long-term studies are needed to confirm these initial findings.

Introduction

Obesity is a major contributor to a wide range of health complications and remains a critical public health concern worldwide. ¹ Individuals with psychiatric disorders exhibit disproportionately high rates of overweight and obesity due to lifestyle factors, comorbid physical illness, genetic vulnerability, and psychotropic medications.^2–6 This comorbidity is clinically significant because obesity can worsen psychiatric outcomes, delay recovery, lead to medication discontinuation or poor adherence, and increase the risk of relapse.^7,8 It can also contribute to premature mortality through additional physical illnesses. ⁹ Therefore, weight management is an essential component of comprehensive psychiatric care and improving the quality of life among individuals with psychiatric disorders.^2,10,11

Given the clinical importance of managing excess weight in psychiatric populations, lifestyle modification can be recommended as an initial approach. ¹² However, its effectiveness in individuals with mental health disorders has generally been low,^2,13 and findings from multiple trials remain constrained by substantial heterogeneity within study populations. ¹⁴

Recent evidence suggests that metformin (MET) can be effective in preventing antipsychotic-medication-induced weight gain, and several clinical guidelines now recommend considering its use in this context.^6,12,15,16 However, the benefits of MET remain modest, and its application has been largely limited to weight management specifically associated with antipsychotic treatment. ¹⁷

In addition to MET, several centrally acting antiobesity medications (AOMs) such as naltrexone/bupropion (NB) and phentermine-topiramate (PT) are widely used for weight reduction in the general population with obesity ¹⁸ ; however, their use in psychiatric patients is often limited by concerns about psychiatric adverse events (AEs) such as mood swings, insomnia, and anxiety.^19–22 Moreover, despite the clinical need, robust evidence regarding the efficacy and safety of these agents in psychiatric populations remains insufficient. ²³

In recent years, clinical interest in glucagon-like peptide-1 (GLP-1) receptor agonists, such as liraglutide (LIRA) and semaglutide, has been increasing, supported by emerging evidence suggesting that these agents might have a relatively favorable psychiatric safety profile.^24,25 Also, several recent reports suggest that GLP-1 receptor agonists can be used relatively safely in patients receiving second-generation antipsychotics,^26–28 indicating that they could be a viable weight-management option in psychiatric populations. However, beyond antipsychotic-medication-induced weight gain, evidence about the effectiveness of GLP-1 receptor agonists in preventing or mitigating weight gain among patients receiving other psychotropic medications—such as antidepressants or anticonvulsants—remains insufficient. This gap is partly attributable to the relatively low use of GLP-1 receptor agonists in psychiatric populations because of persistent concerns about potential psychiatric AEs and economic accessibility constraints.^29,30

In this study, we address the insufficient clinical evidence about pharmacological approaches to weight management in psychiatric patients, particularly the lack of comparative data on the effectiveness and safety of AOMs for those taking antipsychotics or other psychotropic agents. To provide a clinically meaningful evaluation grounded in real-world practice, we selected four pharmacological agents for comparison: LIRA—a GLP-1 receptor agonist with a longer history of clinical use than semaglutide, which was only approved in Korea in 2025; NB and PT—two centrally acting AOMs frequently prescribed in Korea; and MET, which has the most substantial evidence for mitigating antipsychotic-medication-induced weight gain.^6,15 By examining those four agents in a real-world psychiatric outpatient setting, we here provide comparative data on their weight-related efficacy and safety profiles. Although heterogeneity has limited the effectiveness of lifestyle interventions in psychiatric populations, the inclusion of a broad range of psychiatric diagnoses in this study reflects real-world clinical practice. This diagnostic heterogeneity could contribute to variability in treatment response and should be considered when interpreting comparative effectiveness.

Method

Study design

We performed a retrospective electronic medical chart review of individuals who received a prescription for MET, LIRA, NB, or PT for weight management from the outpatient clinic of Yeouido St. Mary’s Hospital (Seoul, Korea) between January 1, 2016, and December 31, 2024.

Participants

To compare the effectiveness and safety of the four AOMs for weight control in patients diagnosed with psychiatric disorders in an outpatient setting, we included patients with a primary diagnosis of psychiatric disorder (F01–F99 in the International Classification of Diseases, 10th revision).

Data collection

We used the date of AOM initiation as the baseline and gathered data about the type of AOM used, sex, age, job status, primary psychiatric diagnosis, and comorbid psychiatric and physical diagnoses. Job type was included to indirectly capture aspects of patients’ daily life patterns. The presence of chronic, comorbid conditions that could affect metabolism (diabetes, hypertension, dyslipidemia, or thyroid disorders) was recorded. The frequency of hospital visits and follow-up assessments was not systematically available in this retrospective chart review and could not be compared across treatment groups.

To compare the weight-loss effects of each drug, clinical data were collected at baseline, 4, 8, and 12 weeks. Because this study was retrospective, an allowable time window of ±2 weeks was applied for each time point. In addition, the type and number of psychotropic medications that can affect weight (antipsychotics, antidepressants, benzodiazepines, and others) were collected at baseline. We categorized the weight gain risk of the psychotropic medications that patients were taking at the start of AOM treatment into four categories (high, moderate, low, and weight-loss potential) based on previous studies^6,31–33 and collected the number of medications in each category. A list of drugs categorized into each category can be found in Table S1.

Safety was evaluated based on AEs reported by patients or observed by clinicians throughout the 12-week period. Independently, early discontinuation (ED) was defined as the cessation of AOM treatment for any reason within 12 weeks of starting it. Reasons for ED were classified into six categories: AEs, lack of efficacy, cost, patient decision, loss to follow-up, and other (e.g., pregnancy, injection phobia, or unknown causes). For analytical clarity, loss to follow-up was strictly categorized as a reason for discontinuation, rather than as an AE.

Through a manual review of the electronic medical records, we excluded patients if the prescribed AOMs were not used for weight loss (e.g., MET for blood glucose control). We further excluded participants from the collected data if (1) their initial weight and height were not measured at baseline, (2) more than one AOM was used concurrently, or (3) a different AOM was used less than 3 months after a first AOM was used.

Data analysis

The baseline characteristics of participants were compared across the four treatment groups using analysis of variance (ANOVA) for continuous variables and the chi-square test for categorical variables.

We separated the analyses into an efficacy analysis set and a safety analysis set and then considered the real-world drug effects in the efficacy evaluation and drug ED in the safety evaluation. The efficacy analysis set included only those patients for whom initial height and weight were measured and recorded, and for whom the first dose of AOM was prescribed and used from baseline through the first 4 weeks of the observation period. This approach was intended to evaluate the treatment effect among individuals who were sufficiently exposed to the intervention. In contrast, the safety analysis set included all participants who received at least one dose of the medication, including those who discontinued early. This allowed us to detect any AEs or tolerability issues that might have occurred during the early treatment phase, regardless of treatment continuation.

Because medically significant weight loss is generally defined as a 5% or greater reduction from the pre-intervention weight, we calculated the percentage of patients who achieved a weight loss of 5% or greater for each drug at each time point.

Linear mixed-effects models (LMMs) ³⁴ were used to evaluate percent changes in body weight over time (baseline, 4, 8, and 12 weeks) across the four AOM groups. LMMs were selected to account for the repeated-measures structure of the data and inter-individual variability in treatment response. Time, drug type, and their interaction were included as fixed effects, and individual variability was modeled using random intercepts for each participant. The model additionally controlled for potential confounding variables. Given the small sample size relative to the number of candidate covariates, a backward elimination procedure based on clinical relevance and statistical significance was applied to reduce model complexity and minimize the risk of overfitting. The final model retained sex, job status, and the number of moderate weight-gain-risk medications as covariates. Model selection was completed prior to the final inference, and information criteria (Akaike information criterion and Bayes information criterion) for the final model are reported for transparency rather than as independent evidence of model fit. Random slopes for time were explored to account for heterogeneity in individual weight trajectories; however, those models resulted in singular fits and had convergence issues, likely due to the small number of repeated observations per participant. Therefore, a random intercept-only structure was retained.

Categorical variables were treated as factors, and MET was used as the reference group for drug comparisons. The outcome variable was the percent change in body weight from baseline at each time point. The model was estimated using restricted maximum likelihood, and Satterthwaite’s method was applied to approximate the degrees of freedom for the fixed effects. A Type III ANOVA with Satterthwaite’s approximation was used to assess the significance of the main and interaction effects. ³⁵ All statistical tests were two-tailed with a significance level of 0.05. Statistical analysis was performed using R software (version 4.4.2; R Foundation for Statistical Computing, Vienna, Austria). This study is reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement (see Supplemental Material).

Results

Sample description

Data for 140 patients taking AOMs and meeting the inclusion criteria were collected in this study. Of these, 117 were included in the efficacy analysis set, excluding 23 patients who either discontinued their AOMs within the first observation period or were missing height and weight data from the first visit.

Baseline characteristics

Among the 117 participants in the efficacy analysis set, significant group differences were observed in sex distribution (p = 0.044), with the MET group having the highest proportion of males (28.2%) and the PT group the lowest (5.9%). Employment status also differed significantly by drug group (p = 0.035); the LIRA group had the highest percentage of participants with part-time or full-time jobs, and the NB group had the highest proportion of unemployed individuals (66.7%). There were notable differences in primary psychiatric diagnoses across groups (p = 0.005): more than half of the MET group (56.4%) had psychotic disorders, whereas depressive disorders were most common in the PT group (47.1%). The prevalence of ED also varied significantly (p = 0.017): it was highest in the PT group (72.2%) and lowest in the MET group (39.1%). Significant group differences were also found in the number of prescribed antipsychotics (p < 0.001) and antidepressants (p = 0.004), with the MET group using the most antipsychotics (1.64 ± 0.90) and the PT group using the most antidepressants (2.00 ± 0.94). No significant differences were observed in age, physical or psychiatric comorbidities, AEs, calculated baseline body mass index, or total number of drugs prescribed (Table 1).

Table 1.

Baseline characteristics of the participants.

Variables	Category	Overall	MET	LIRA	NB	PT	p Value
N		117	39	49	12	17
Age, SD		41.05 (13.23)	41.08 (15.03)	40.80 (11.77)	43.42 (15.62)	40.06 (11.96)	0.965 a
Job, N (%)	Students, Housewife, Retired	40 (34.2)	17 (43.6)	19 (38.8)	2 (16.7)	2 (11.8)	0.035
	Unemployed	34 (29.1)	9 (23.1)	10 (20.4)	8 (66.7)	7 (41.2)
	Part-time job	19 (16.2)	6 (15.4)	9 (18.4)	0 (0.0)	4 (23.5)
	Full-time job	23 (19.7)	7 (17.9)	11 (22.4)	2 (16.7)	3 (17.6)
	Unknown	1 (0.9)	0 (0.0)	0 (0.0)	0 (0.0)	1 (5.9)
Primary psychiatric diagnosis, N (%)	Psychotic disorder	36 (30.8)	22 (56.4)	9 (18.4)	4 (33.3)	1 (5.9)	0.005
	Bipolar disorder	15 (12.8)	3 (7.7)	8 (16.3)	1 (8.3)	3 (17.6)
	Depressive disorder	39 (33.3)	8 (20.5)	19 (38.8)	4 (33.3)	8 (47.1)
	Anxiety disorder	23 (19.7)	3 (7.7)	13 (26.5)	3 (25.0)	4 (23.5)
	Other disorders	4 (3.4)	3 (7.7)	0 (0.0)	0 (0.0)	1 (5.9)
Sex, N (%)	Male	18 (15.4)	11 (28.2)	4 (8.2)	2 (16.7)	1 (5.9)	0.044
	Female	99 (84.6)	28 (71.8)	45 (91.8)	10 (83.3)	16 (94.1)
Presence of psychiatric comorbidities N (%)	No	87 (74.4)	33 (84.6)	36 (73.5)	8 (66.7)	10 (58.8)	0.195
	Yes	30 (25.6)	6 (15.4)	13 (26.5)	4 (33.3)	7 (41.2)
Presence of physical comorbidities N (%)	No	88 (75.2)	30 (76.9)	37 (75.5)	9 (75.0)	12 (70.6)	0.968
	Yes	29 (24.8)	9 (23.1)	12 (24.5)	3 (25.0)	5 (29.4)
ED, N (%)	No	53 (45.3)	25 (64.1)	18 (36.7)	5 (41.7)	5 (29.4)	0.032
	Yes	64 (54.7)	14 (35.9)	31 (63.3)	7 (58.3)	12 (70.6)
AE, N (%)	No	91 (77.8)	31 (79.5)	38 (77.6)	8 (66.7)	14 (82.4)	0.770
	Yes	26 (22.2)	8 (20.5)	11 (22.4)	4 (33.3)	3 (17.6)
Number of drugs, mean (SD)	Total	5.24 (2.25)	5.33 (2.71)	5.02 (2.02)	5.75 (2.01)	5.29 (1.99)	0.861 a
	Antipsychotics	1.15 (0.86)	1.64 (0.90)	0.88 (0.78)	1.08 (0.67)	0.82 (0.64)	<0.001 a
	Antidepressants	1.58 (1.15)	1.10 (1.12)	1.78 (1.18)	1.75 (0.97)	2.00 (0.94)	0.004 a
	Benzodiazepines	1.24 (0.87)	1.03 (0.90)	1.27 (0.81)	1.67 (0.89)	1.35 (0.86)	0.066 a
	Other drugs	1.28 (1.08)	1.59 (1.35)	1.10 (0.85)	1.25 (1.14)	1.12 (0.86)	0.128 a
Weight, kg, mean (SD)		76.31 (17.49)	76.34 (17.27)	76.10 (17.25)	77.17 (22.84)	76.21 (16.02)	0.977 a
Height, cm, mean (SD)		162.64 (7.61)	163.47 (8.87)	162.88 (6.99)	161.57 (7.52)	160.84 (6.42)	0.200 a
BMI, kg/m2 (SD)		28.72 (5.45)	28.42 (5.23)	28.56 (5.28)	29.45 (7.66)	29.33 (5.04)	0.491 a
Number of concomitant psychiatric medications, mean (SD)	High risk for weight gain	0.31 (0.52)	0.49 (0.60)	0.20 (0.46)	0.17 (0.39)	0.29 (0.47)	0.104 a
	Moderate risk for weight gain	0.91 (0.81)	1.08 (0.81)	0.88 (0.81)	0.67 (0.98)	0.76 (0.66)	0.104 a
	Low risk for weight gain	1.56 (1.01)	1.26 (0.97)	1.61 (1.00)	1.83 (0.94)	1.88 (1.11)	0.017 a
	None or weight-loss potential	0.69 (0.84)	0.69 (1.03)	0.57 (0.74)	1.00 (0.60)	0.82 (0.73)	0.372 a

^ap-Value of ANOVA.

AE, adverse event; ED, early discontinuation; LIRA, liraglutide; MET, metformin; N, number; NB, naltrexone/bupropion; PT, phentermine-topiramate; SD, standard deviation.

Proportion of responders (⩾5% weight loss) by AOM over time

The proportion of patients who achieved ⩾5% weight loss increased over time in the LIRA and PT groups, peaking at 12 weeks (22.4% and 17.6%, respectively). LIRA consistently showed the highest responder rates at all-time points, and the MET group demonstrated the lowest response, reaching only 5.1% by week 12 (Figure 1).

Figure 1.

Proportion of patients with ⩾5% weight loss at each time point.

LIRA, liraglutide; MET, metformin; NB, naltrexone/bupropion; PT, phentermine-topiramate.

Comparative weight loss by AOM at weeks 4, 8, and 12

At week 12, the LIRA group showed the greatest mean percent reduction in body weight from baseline (−6.30% ± 0.96), followed by the PT (−4.25% ± 0.98), NB (−2.87% ± 0.40), and MET (−1.18% ± 0.90) groups. Similar trends were observed at week 8, with LIRA again demonstrating the highest weight reduction (−4.88% ± 0.56), followed by PT (−2.87% ± 0.80), NB (−3.04% ± 0.75), and MET (−1.84% ± 0.68; Figure 2).

Figure 2.

Mean percent of weight reduction by drugs at each time point for each drug.

LIRA, liraglutide; MET, metformin; NB, naltrexone/bupropion; PT, phentermine-topiramate.

Linear mixed-effects analysis of weight change by AOM type

Table 2 presents the results of the simplified LMM. LIRA demonstrated significantly greater weight reduction than the MET reference group (estimate = −3.45, 95% CI: −5.35 to −1.55, p < 0.001). A significant interaction was also observed between LIRA and Time at 12 weeks (estimate = −2.83, 95% CI: −5.20 to −0.46, p = 0.019), indicating that the weight-lowering efficacy of LIRA became more pronounced over time, compared with the MET reference group. In contrast, the NB and PT groups did not show statistically significant differences from the MET group in the main analysis (p = 0.127 and p = 0.098, respectively). These findings remained consistent in sensitivity analyses. The superior weight-lowering effect of LIRA was maintained when analyzing absolute weight changes (Table S5) and when applying the Last Observation Carried Forward (LOCF) method to account for missing data (Table S6), confirming the robustness of our primary results.

Table 2.

Fixed effects from the simplified linear mixed-effects model on percentage weight change.

Fixed effects	Estimate	SE	95% CI	p Value
(Intercept)	−2.12	1.24	−4.55 to 0.31	0.088
Drug type (ref: MET)
LIRA	−3.45	0.97	−5.35 to −1.55	<0.001*
NB	−2.4	1.57	−5.48 to 0.68	0.127
PT	−2.42	1.46	−5.28 to 0.44	0.098
Time (ref: Baseline)
Time: 8 weeks	−1.69	0.8	−3.25 to −0.13	0.034*
Time: 12 weeks	−0.93	0.84	−2.57 to 0.71	0.264
Interaction
LIRA × Time (12 weeks)	−2.83	1.21	−5.20 to −0.46	0.019*
Covariates
Sex: Female	2.97	1.08	0.85 to 5.09	0.006*
Job: Full-time	3.2	1.06	1.12 to 5.28	0.003*
Moderate-risk drugs (Count)	−0.84	0.45	−1.72 to 0.05	0.066

The model was simplified by backward elimination to prevent over-parameterization. Only significant or key variables were retained. Model diagnostics: Akaike information criterion = 1034.4, Bayes information criterion = 1096.7, Random intercept variance = 6.07.

^*p < 0.05.

CI, confidence interval; LIRA, liraglutide; MET, metformin; N, number; NB, naltrexone/bupropion; PT, phentermine-topiramate; SE, standard error.

Regarding covariates, female sex was significantly associated with a smaller percentage reduction in weight, compared with male participants (estimate = 2.97, 95% CI: 0.85–5.09, p = 0.006), indicating attenuated weight loss among women. Similarly, participants with full-time jobs showed a significantly smaller reduction in percent weight loss (estimate = 3.20, 95% CI: 1.12–5.28, p = 0.003) compared to the reference group. The number of moderate weight-gain-risk medications was not statistically significant in the final model (Estimate = −0.84, p = 0.066). Other variables, such as age and comorbidities, were excluded from the final model due to nonsignificance during the backward selection process.

Comparison of safety among the four AOMs

The incidence of AEs varied across drug groups—MET (21.7%), LIRA (21.6%), NB (41.2%), and PT (22.2%)—but the differences were not statistically significant (χ² = 3.05, df = 3, p = 0.384). When psychiatric AEs were examined separately, no events were observed in the MET group. In the LIRA group, two cases of insomnia and one case of irritability were reported; in the NB group, two cases of mood swings and one case of insomnia; and in the PT group, one case of increased sexuality and one case of mood swings (Table S3).

In contrast, overall ED rates did differ across drug types (MET: 39.1%, LIRA: 64.7%, NB: 70.6%, PT: 72.2%; χ² = 10.23, df = 3, p = 0.017). An examination of the reported reasons for ED showed that AEs were the most frequently cited cause across all AOM groups (MET: 30.3%, LIRA: 30.3%, NB: 50.0%, PT: 30.8%). Lack of efficacy was particularly common in the PT group (30.8%), and it was also observed in the MET and LIRA groups (24.2% each). Patient decision accounted for 15–25% of ED cases across treatment groups. Additional reasons included loss to follow-up, cost-related factors, and unknown causes. Injection phobia (3.0%) and pregnancy (7.7%) were reported only in the LIRA and PT groups, respectively (Table 3).

Table 3.

Cause of early discontinuation by AOM type.

AOM types	Cause of ED	%, (n)
MET	Adverse events	50.0, (9)
	Lack of efficacy	22.2, (4)
	Patient decision	11.1, (2)
	Follow-up loss	11.1, (2)
	Pregnancy	5.55 (1)
LIRA	Adverse events	30.3, (10)
	Lack of efficacy	24.2, (8)
	Patient decision	21.2, (7)
	Follow-up loss	9.1, (3)
	Cost	6.1, (2)
	Unknown	6.1, (2)
	Injection phobia	3.0, (1)
NB	Adverse events	50.0, (6)
	Patient decision	25.0, (3)
	Lack of efficacy	16.7, (2)
	Unknown	8.3, (1)
PT	Adverse events	30.8, (4)
	Lack of efficacy	30.8, (4)
	Patient decision	15.4, (2)
	Cost	7.7, (1)
	Follow-up loss	7.7, (1)
	Pregnancy	7.7, (1)

AOM, antiobesity medication; ED, early discontinuation; LIRA, liraglutide; MET, metformin; NB, naltrexone/bupropion; PT, phentermine-topiramate.

Discussion

In this study, we compared the weight-loss efficacy and safety of four AOMs in psychiatric outpatients. To our knowledge, this is the first study to compare the efficacy and safety of four AOMs, including MET and LIRA, among only psychiatric patients with and without metabolic disease. Compared with MET, LIRA had a significantly greater weight-loss effect over time during the 12-week observation period. In addition, among the demographic variables, female sex and having a full-time job were associated with a lower percent weight loss, and the number of moderate weight-gain-risk drugs was associated with a higher percent weight loss. In terms of safety, the four drugs did not differ with statistical significance in the occurrence of AEs, but in the ED analysis, MET had a significantly lower ED rate than the other drugs.

Regarding weight-loss efficacy, all four AOMs were associated with reductions in body weight by week 8. Notably, LIRA showed consistently greater weight loss than MET across all observation windows (4, 8, and 12 weeks), with the magnitude of difference increasing over time. This pattern was also reflected in the gradual increase in the proportion of responders in the LIRA group. The superior weight-loss efficacy of GLP-1 receptor agonists, compared with MET, has been reported previously, including in recent studies in which semaglutide—a GLP-1 receptor agonist similar to LIRA—demonstrated greater weight reduction than MET in patients with antipsychotic-medication-induced weight gain. ³⁶ Taken together, these findings support the potential efficacy of GLP-1 receptor agonists for weight management in psychiatric populations.

The demographic characteristics of the study population might have influenced the observed outcomes. A large proportion of the participants were women (84.6%), which might reflect greater sensitivity to weight gain and higher engagement in weight-loss treatment among female patients. ³⁷ Previous studies have shown that women treated for psychiatric disorders are at increased risk of weight gain ³⁸ and metabolic syndrome, compared with men, particularly when they receive antipsychotic medications. ³⁹ Sex-related differences in body fat distribution, ⁴⁰ drug metabolism, ⁴¹ and hormonal influences on appetite and insulin resistance,^42,43 might contribute to differences in prescribing patterns and treatment response.^44,45

Full-time employment was associated with attenuated weight loss in our analysis. Rather than reflecting lifestyle differences per se, this association could be related to practical constraints associated with employment, such as limited time availability or challenges in adhering to dietary and exercise recommendations, which were not directly measured in this study. We also observed that a higher number of concomitant psychotropic medications with a moderate risk for weight gain was associated with greater weight reduction. This finding could reflect a higher baseline metabolic burden in these patients, resulting in a larger observable change following a pharmacological weight-management intervention, rather than a direct causal effect of the psychotropic medication burden itself.

We observed differences in psychiatric diagnostic composition across the AOM groups. Approximately half of the patients in the MET group had a psychotic disorder, whereas depressive disorders predominated in the other groups. This likely reflects the established evidence base supporting MET as an off-label adjunctive treatment for antipsychotic medication-induced weight gain and insulin resistance.^6,17,46,47 Consistent with this pattern, antipsychotics were most frequently used concomitantly in the MET group. Other AOMs might have been used less often in patients with psychotic disorders because of concerns about the potential exacerbation of psychiatric symptoms or the insufficient evidence base supporting their safety in this population.^29,48–50

In terms of psychiatric safety, no psychiatric AEs were observed in the MET group, whereas a few events were reported in the LIRA, NB, and PT groups. These findings suggest that MET and LIRA were not associated with a high frequency of psychiatric AEs in this cohort; however, cautious interpretation is warranted given the small sample size and observational study design. Although no significant psychiatric deterioration was observed in this study, previous studies have reported that weight loss achieved through extreme methods can exacerbate depressive symptoms ⁵¹ and that excessive weight loss can be associated with negative psychological reactions or the development of eating disorders. ⁵² Thus, the potential for such psychological changes warrants continued attention, despite our findings.

Beyond efficacy and AEs, treatment persistence is an important consideration in real-world psychiatric practice. ED was common across all AOM groups and occurred for heterogeneous reasons, including AEs, lack of efficacy, patient decision, and non-clinical factors such as cost and loss to follow-up. Consequently, differences in the overall ED rates across AOMs should not be interpreted as a single unified measure of tolerability or treatment effectiveness, but rather as descriptive indicators of treatment persistence in routine psychiatric care. High ED rates in the NB and PT groups resulted in few patients completing the 12-week observation period, which made more detailed comparisons of efficacy across these treatments impossible. Although the ED rate was higher for LIRA than for MET, the relatively large number of patients treated with LIRA allowed for meaningful longitudinal comparisons.

Medication dose and cost might also have influenced treatment outcomes and persistence. The mean dose of MET at 12 weeks (833 mg) was lower than the doses commonly recommended for antipsychotic medication-induced weight gain,^15,16 which might have attenuated its observed effect. In contrast, although higher doses of LIRA have been used in prior studies,^53–57 clinically meaningful weight loss has also been reported at doses similar to those used in our study population. ⁵⁸ Given ongoing concerns about the cost of GLP-1 receptor agonists,^59,60 the observation that weight loss was achieved at relatively low doses suggests a potential strategy to balance efficacy and economic burden. Moreover, emerging evidence indicates that GLP-1 receptor agonists can improve quality of life^61,62 and have beneficial effects on mood and cognition through direct action in the brain, without increasing serious psychiatric AEs.^62–65

Limitations

Our study has several limitations that should be acknowledged. First, the retrospective, single-center design limits the generalizability of our findings and precludes the establishment of definitive causal relationships. A significant limitation inherent to this design is confounding by indication, that is, prescribers might have selected specific AOMs based on unmeasured patient profiles. Specifically, we did not explicitly measure socioeconomic status. Given the variation in costs among antiobesity medications, economic factors could have introduced bias in treatment allocation and adherence. Therefore, the possibility of residual confounding persists despite statistical adjustments. Second, psychiatric safety was assessed based on clinical records and patient self-reports, rather than objective, structured psychiatric rating scales. This reliance might have led to the underreporting of mild psychiatric symptoms or subtle mood changes that did not reach the threshold of clinical documentation. Third, the 12-week observation period and relatively small sample size make it difficult to assess the long-term sustainability of weight loss and rare AEs. Furthermore, the doses of AOMs used were often lower than the maximum recommended doses, and lifestyle factors such as diet and physical activity, which can affect weight outcomes, were not strictly controlled. Finally, newer GLP-1 receptor agonists, such as semaglutide and tirzepatide, were not included in this analysis because they were not available in Korea during the data collection window.^66,67 Future prospective studies with larger samples and structured assessments are needed to validate these findings.

Conclusion

In conclusion, our findings suggest that LIRA might be a feasible treatment option for weight management in psychiatric patients because it showed a significant association with good short-term weight reduction without evident safety concerns in this study. However, given the study’s limitations—retrospective design, small sample size, lack of structured psychiatric assessments, and absence of data on socioeconomic status and lifestyle factors—these results should be interpreted with caution. Large-scale prospective studies are needed to confirm the long-term safety and efficacy of AOMs in this population. Future large-scale, long-term studies that include other GLP-1 receptor agonists, such as semaglutide and tirzepatide, are needed to more comprehensively assess their efficacy and safety in psychiatric populations.