Introduction
Severe mental disorders (SMI) characterized by extreme disturbances in cognition, emotional regulation, and behavior create significant burdens on psychosocial and occupational functioning. It is often exacerbated by neurobiological and genetic factors that shape its course and treatment response [1, Reference Marder and Cannon2]. These conditions lead to a mortality gap of 10–20 years compared to the general population, in part, contributing to premature cardiovascular disease (CVD), which accounts for 14.3% of global annual deaths [Reference Correll, Solmi, Veronese, Bortolato, Rosson and Santonastaso3–Reference Hjorthøj, Stürup, McGrath and Nordentoft6]. Emerging evidence underscores metabolic syndrome, a cluster of central obesity, hypertension, dyslipidemia, and hyperglycemia, as critical mediators of CVD risk in this population [Reference Polcwiartek, O’Gallagher, Friedman, Correll, Solmi and Jensen7, Reference Correll, Solmi, Croatto, Schneider, Rohani-Montez and Fairley8].
The pathophysiology of SMI-related metabolic syndrome is multifactorial and involves chronic inflammation, immune dysregulation, and genetic susceptibility [Reference Henderson, Vincenzi, Andrea, Ulloa and Copeland9–Reference Garrido-Torres, Rocha-Gonzalez, Alameda, Rodriguez-Gangoso, Vilches and Canal-Rivero11]. Second-generation antipsychotics worsen these risks owing to weight gain, insulin resistance, and lipid abnormalities [Reference Mottillo, Filion, Genest, Joseph, Pilote and Poirier12]. At the same time, both antidepressants and mood stabilizers have milder metabolic effects, and polypharmacy and high-dose regimens lead to further compound adverse outcomes [Reference McIntyre, Soczynska, Konarski and Kennedy13–Reference Peralta, de Jalón, Moreno-Izco, Peralta, Janda and Sánchez-Torres16]. Schizophrenia, affecting 1% of the global population, underlies these challenges: prolonged antipsychotic treatment and chronicity of the disease combined to increase metabolic morbidity [Reference Correll, Detraux, De Lepeleire and De Hert17–Reference Pillinger, McCutcheon, Vano, Mizuno, Arumuham and Hindley20].
Nonpharmacological interventions, primarily structured exercises, have great potential to reduce risks [Reference Stubbs, Vancampfort, Hallgren, Firth, Veronese and Solmi21]. Meta-analyses have shown that physical activity improves psychiatric symptoms, functional capacity, and quality of life in SMI [Reference Korman, Stanton, Vecchio, Chapman, Parker and Martland22–Reference Brokmeier, Firth, Vancampfort, Smith, Deenik and Rosenbaum26]. Exercise also enhances cardiorespiratory fitness and reduces obesity, diabetes, and CVD incidence, which is vital for a population with twice the occurrence of metabolic syndrome compared to the general population [Reference Korman, Stanton, Vecchio, Chapman, Parker and Martland22, Reference Heissel, Heinen, Brokmeier, Skarabis, Kangas and Vancampfort27–Reference Yang, Yuan, Zhang, Fu, Wang and Wang29]. However, there are still some important gaps. First, exercise does not directly ameliorate the components of metabolic syndrome (e.g., waist circumference and blood pressure) in patients with psychotic disorders. Second, potential synergies or antagonisms between exercise and antipsychotic pharmacodynamics have not been examined despite evidence that physical activity may modulate drug metabolism and receptor sensitivity [Reference Hassan, Shannon, Tully, McCartan, Davidson and Bunn30–Reference Vancampfort, Firth, Schuch, Rosenbaum, Mugisha and Hallgren32]. Such multidimensional advancements enhance global health and address the leading health-related issues in this group, thereby minimizing the likelihood of severe long-term repercussions [Reference Vancampfort, Rosenbaum, Schuch, Ward, Richards and Mugisha33, Reference Cuesta, Sánchez-Torres, Moreno-Izco, García de Jalón, Gil-Berrozpe and Peralta34].
This systematic review and meta-analysis aimed to (1) evaluate the efficacy of physical exercise on metabolic syndrome components in psychotic disorders and (2) explore the interactions between exercise and antipsychotic medications on metabolic outcomes. By elucidating these mechanisms, our findings aim to inform integrative treatment paradigms that optimize the mental and physical health.
Methods
Study design
This systematic review and meta-analysis followed the guidelines outlined in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [Reference Moher, Liberati, Tetzlaff and Altman35] and was registered in PROSPERO (CRD42025635273).
Search strategy
The search for studies was conducted on October 10, 2024. Articles were gathered from the Medline (PubMed), CINAHL, Web of Science, and APA PsycINFO databases. Two researchers independently screened the articles by evaluating their titles and abstracts to assess the eligibility criteria, and disagreements were resolved through discussion with a third author. The articles were screened using the Sysrev software. The search used a defined set of keywords related to physical exercise, metabolic syndrome, schizophrenia, psychotic disorders, and randomized controlled trials. The complete search strategy is available in the Supplementary Materials.
Selection criteria
The PICOS model defined the inclusion criteria (Population, Intervention, Comparison, Outcome, Study type) [Reference Methley, Campbell, Chew-Graham, McNally and Cheraghi-Sohi S36]. The inclusion of studies was based on the following criteria: (a) studies focused mainly on people over 18 years with a diagnosis of psychotic disorder based on the Diagnostic and Statistical Manual of Mental Disorders (DSM-V) [1]; (b) randomized controlled trial (RCT); (c) assessed components of metabolic syndrome including waist circumference, blood pressure, fasting glucose, triglycerides, and HDL cholesterol; and (d) evaluation of exercise intervention. Systematic reviews, meta-analyses, single-group trials, practice guidelines, recommendations, editorials, letters, conference abstracts, reports, protocols, and studies with undefined interventions or procedures were excluded. We also excluded RCTs based on other specific SMI (i.e., bipolar disorder, anxiety, and major depressive disorder).
While schizophrenia was the primary diagnosis, we included related DSM-V-defined psychotic disorders (e.g., schizoaffective disorder and schizophreniform disorder). These conditions share core clinical and neurobiological features with schizophrenia, including antipsychotic treatment patterns and an elevated cardiometabolic risk. This broader “psychotic disorders” grouping enabled a more comprehensive evidence synthesis and increased statistical power.
Screening
The search strategy gathered the results from four databases and eliminated duplicates. The titles and abstracts of the studies were screened according to eligibility criteria. Full-text articles were then evaluated, and those that met the eligibility criteria were selected for inclusion.
Data extraction
Two authors independently extracted summary data based on key observations. They used a standardized form to collect relevant information, such as authors, publication year, study population, design, sample size, interventions, and outcomes. The third author resolved any disagreements during the selection or extraction process.
Risk of bias assessment
Two authors independently evaluated the risk of bias for each included trial using the Cochrane Risk of Bias 2 (RoB-2) tool [Reference Sterne, Savović, Page, Elbers, Blencowe and Boutron37]. This tool examines domains such as the randomization process, adherence to interventions, handling of missing outcome data, measurement of the outcome, and selection of reported results. Each domain and trial was adjudicated based on a low-, some concerns, or high-risk level. Any disagreement between the two assessors was resolved through discussion with a third investigator. The overall quality of evidence was assessed using the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) approach.
Statistical analysis and data synthesis
Statistical analysis for this meta-analysis was conducted using the Comprehensive Meta-Analysis and JASP statistical software [38, 39]. Effect sizes were computed using standardized mean differences (SMD) and 95% confidence intervals (CI) to measure the impact of physical exercise on the components of metabolic syndrome. A random-effects model was used to account for potential heterogeneity between the studies [Reference Higgins, Thompson, Deeks and Altman40]. Heterogeneity was assessed using the I 2 statistic, with thresholds of 25, 50, and 75% representing low, moderate, and high heterogeneity, respectively [Reference Higgins and Thompson41]. Additionally, Q statistics were used to determine statistical significance. Data synthesis focused on pooled estimates for each metabolic component. Publication bias was evaluated using selection models and Egger’s test. The significance level was set at p < 0.05.
Results
Study selection
Figure 1 shows the flow diagram of the database search process. A total of 3904 articles were retrieved from these four databases. After removing 1124 duplicate articles, 2780 were screened by title and abstract, and 2528 were excluded. Finally, 252 articles remained for full-text screening, and 242 were excluded based on the inclusion and exclusion criteria. Finally, 10 articles were included in the review [Reference Zhou, Guan, Xiu and Wu42–Reference Masa-Font, Fernández-San-Martín, Martín López, Alba Muñoz, Oller Canet and Martín Royo51].
Figure 1. PRISMA flowchart of the study.
Risk of bias assessment
Figure 2 shows the risk of bias across included studies. Five studies were classified as having a low risk of bias, four showed some concerns, and one was rated as having a high risk of bias. The primary source of bias was the blinding of the participants and personnel. However, randomization is mentioned in the studies; the lack of detailed information regarding the allocation process and whether concealment was properly implemented raise concerns about the procedure’s validity. Furthermore, some studies have reported challenges in participant adherence to interventions. However, the methods employed to manage these cases were not sufficiently detailed in the analysis, raising concerns regarding their potential impact on the overall risk of bias.
Figure 2. Risk of bias assessment. (A) Quality of methodology in the included studies. (B) Distribution of methodological quality across included studies.
Characteristics of the participants
Ten RCTs were included in this meta-analysis. The characteristics of the included RCTs are shown in Table 1. Two studies were performed in Spain [Reference Fernández-Abascal, Suárez-Pinilla, Cobo-Corrales, Crespo-Facorro and Suárez-Pinilla48, Reference Masa-Font, Fernández-San-Martín, Martín López, Alba Muñoz, Oller Canet and Martín Royo51], two in Norway [Reference Heggelund, Nilsberg, Hoff, Morken and Helgerud43, Reference Brobakken, Nygård, Güzey, Morken, Reitan and Heggelund49], and one in Japan [Reference Ikai, Suzuki, Uchida, Saruta, Tsukinoki and Fujii45], the Netherlands [Reference Scheewe, Backx, Takken, Jörg, van Strater and Kroes46], Canada [Reference Romain, Fankam, Karelis, Letendre, Mikolajczak and Stip44], Taiwan [Reference Wu, Wang, Bai, Huang and Lee50], China [Reference Zhou, Guan, Xiu and Wu42], and the Republic of Korea [Reference Jae, Kil, So, Lee, Song and Kim47]. A total of 773 participants were randomized into exercise (405 participants) and control groups (368 participants). In all, 71.54% had schizophrenia; 13.86% had schizoaffective disorder; and 2.12%, 1.25%, and 11.23% had other SMI, schizophreniform disorders, psychotic disorders, and other mental disorders (DSM-V) [1]. The mean age of the patients was 39.9 years (standard deviation [SD]: 7.36). Of the 10 RCTs included, 9 reported the participants’ sex [Reference Zhou, Guan, Xiu and Wu42–Reference Scheewe, Backx, Takken, Jörg, van Strater and Kroes46, Reference Fernández-Abascal, Suárez-Pinilla, Cobo-Corrales, Crespo-Facorro and Suárez-Pinilla48–Reference Masa-Font, Fernández-San-Martín, Martín López, Alba Muñoz, Oller Canet and Martín Royo51], of which 38.7% were female.
Table 1. Qualitative analysis of included studies analyzing the effects of exercise on the components of metabolic syndrome
Interventions’ characteristics
Wide heterogeneity was observed during the intervention period. Exercise interventions lasted between 8 weeks and 12 months. Twelve weeks was the most frequently observed duration [Reference Zhou, Guan, Xiu and Wu42, Reference Jae, Kil, So, Lee, Song and Kim47, Reference Fernández-Abascal, Suárez-Pinilla, Cobo-Corrales, Crespo-Facorro and Suárez-Pinilla48, Reference Masa-Font, Fernández-San-Martín, Martín López, Alba Muñoz, Oller Canet and Martín Royo51], and two studies reported follow-up periods of 8 weeks and 18 months postexercise [Reference Ikai, Suzuki, Uchida, Saruta, Tsukinoki and Fujii45, Reference Fernández-Abascal, Suárez-Pinilla, Cobo-Corrales, Crespo-Facorro and Suárez-Pinilla48]. Physical exercise programs involved one to three sessions per week, with each session lasting between 30 and 120 min.
Most of the intervention groups focused on aerobic exercise, including high-intensity interval training (HIIT) (n = 2) [Reference Heggelund, Nilsberg, Hoff, Morken and Helgerud43, Reference Romain, Fankam, Karelis, Letendre, Mikolajczak and Stip44], aerobic interval training (n = 1) [Reference Brobakken, Nygård, Güzey, Morken, Reitan and Heggelund49], aerobic exercise including fast walking and stair climbing (n = 2) [Reference Fernández-Abascal, Suárez-Pinilla, Cobo-Corrales, Crespo-Facorro and Suárez-Pinilla48, Reference Wu, Wang, Bai, Huang and Lee50], Hatha yoga (n = 1) [Reference Ikai, Suzuki, Uchida, Saruta, Tsukinoki and Fujii45], dance movements (n = 1) [Reference Zhou, Guan, Xiu and Wu42], and structured walking sessions (n = 1) [Reference Masa-Font, Fernández-San-Martín, Martín López, Alba Muñoz, Oller Canet and Martín Royo51]. Additionally, many studies combined aerobic and resistance training (n = 2) [Reference Scheewe, Backx, Takken, Jörg, van Strater and Kroes46, Reference Jae, Kil, So, Lee, Song and Kim47].
Most studies provided usual medical care treatment, with ongoing psychiatric follow-ups (n = 6) [Reference Zhou, Guan, Xiu and Wu42, Reference Romain, Fankam, Karelis, Letendre, Mikolajczak and Stip44, Reference Ikai, Suzuki, Uchida, Saruta, Tsukinoki and Fujii45, Reference Fernández-Abascal, Suárez-Pinilla, Cobo-Corrales, Crespo-Facorro and Suárez-Pinilla48, Reference Wu, Wang, Bai, Huang and Lee50, Reference Masa-Font, Fernández-San-Martín, Martín López, Alba Muñoz, Oller Canet and Martín Royo51]. In contrast, Brobakken et al. [Reference Brobakken, Nygård, Güzey, Morken, Reitan and Heggelund49] performed two supervised aerobic interval training (AIT) sessions. The participants were instructed to practice the exercises at home (n = 1) [Reference Brobakken, Nygård, Güzey, Morken, Reitan and Heggelund49]. Other studies included occupational therapy sessions (n = 1) [Reference Scheewe, Backx, Takken, Jörg, van Strater and Kroes46], weekly recreational sessions (n = 1) [Reference Jae, Kil, So, Lee, Song and Kim47], and computer-based training sessions (n = 1) [Reference Heggelund, Nilsberg, Hoff, Morken and Helgerud43] for control participants.
The intensity of exercise interventions varied across the included studies. Two studies prescribed combined aerobic and resistance exercises at 45–75% of the heart rate reserve (HRR) [Reference Scheewe, Backx, Takken, Jörg, van Strater and Kroes46, Reference Jae, Kil, So, Lee, Song and Kim47]. Other studies included HIIT sessions conducted at intensities ranging from 50% to 90% HRmax [Reference Romain, Fankam, Karelis, Letendre, Mikolajczak and Stip44] and 70% to 95% HRpeak [Reference Heggelund, Nilsberg, Hoff, Morken and Helgerud43, Reference Romain, Fankam, Karelis, Letendre, Mikolajczak and Stip44], whereas AIT sessions were undertaken in a range of 70% to 75% HRpeak [Reference Brobakken, Nygård, Güzey, Morken, Reitan and Heggelund49]. Another study included aerobic exercise sessions with increasing intensity up to 75% maximum HR [Reference Fernández-Abascal, Suárez-Pinilla, Cobo-Corrales, Crespo-Facorro and Suárez-Pinilla48]. Many studies did not prescribe exercise considering the intensity due to the nature of the interventions [Reference Zhou, Guan, Xiu and Wu42, Reference Ikai, Suzuki, Uchida, Saruta, Tsukinoki and Fujii45, Reference Wu, Wang, Bai, Huang and Lee50, Reference Masa-Font, Fernández-San-Martín, Martín López, Alba Muñoz, Oller Canet and Martín Royo51]. These intensity variations were tailored to the specific demands of each intervention.
Components of metabolic syndrome
Waist circumference
Seven studies assessed the waist circumference [Reference Romain, Fankam, Karelis, Letendre, Mikolajczak and Stip44, Reference Scheewe, Backx, Takken, Jörg, van Strater and Kroes46–Reference Masa-Font, Fernández-San-Martín, Martín López, Alba Muñoz, Oller Canet and Martín Royo51]. The pooled results indicated that exercise interventions did not significantly affect waist circumference (SMD = 0.206, 95% CI [−0.118, 0.530], p = 0.171). Moderate heterogeneity was observed among the included studies (Q = 13.054, I 2 = 51.6%, p = 0.042) (Figure 3). The selection models reported no evidence of publication bias (p = 0.07), which was in line with the results of Egger’s regression (p = 0.18). The mean model estimates are presented in Supplementary Figure S1.
Figure 3. Forest plot of the effect of physical exercise on waist circumference.
Blood pressure
Seven studies examined systolic and diastolic blood pressure [Reference Heggelund, Nilsberg, Hoff, Morken and Helgerud43, Reference Romain, Fankam, Karelis, Letendre, Mikolajczak and Stip44, Reference Scheewe, Backx, Takken, Jörg, van Strater and Kroes46–Reference Brobakken, Nygård, Güzey, Morken, Reitan and Heggelund49, Reference Masa-Font, Fernández-San-Martín, Martín López, Alba Muñoz, Oller Canet and Martín Royo51]. For systolic blood pressure, the pooled data indicated no statistically significant effect of structured physical exercise interventions (SMD = 0.194, 95% CI [−0.115, 0.504], p = 0.219), with moderate heterogeneity observed (Q = 13.886, I 2 = 61.258%, p = 0.031) (Figure 4). Similarly, diastolic blood pressure showed no significant changes (SMD = −0.21, 95% CI [−0.854, 0.434], p = 0.522), though the heterogeneity was substantial (Q = 42.724, I 2 = 91.127%, p < 0.001) (Figure 5). The selection models showed no bias for either systolic (p = 0.294) or diastolic blood pressure (p = 0.31). These results were further validated by Egger’s regression tests (systolic: p = 0.589, diastolic: p = 0.773) [39, Reference Higgins, Thompson, Deeks and Altman52]. Detailed mean model estimates are provided in Supplementary Figure S2 for systolic blood pressure and Supplementary Figure S3 for diastolic blood pressure.
Figure 4. Forest plot showing the effect of physical exercise on systolic blood pressure.
Figure 5. Forest plot showing the effect of physical exercise on diastolic blood pressure.
HDL cholesterol
Seven studies evaluated the effects of exercise on high-density lipoprotein (HDL) cholesterol levels. Our findings revealed that exercise interventions did not significantly affect HDL cholesterol levels (SMD = 0.157, 95% CI [−0.36, 0.674], p = 0.551). However, substantial heterogeneity was observed across the included studies (Q = 30.617, I 2 = 80.684%, p < 0.001) (Figure 6). No evidence of publication bias was detected (p = 0.421). This finding was further validated through Egger’s regression, confirming the absence of publication bias (p = 0.824) [39, Reference Higgins, Thompson, Deeks and Altman52]. The detailed estimates of the mean model are shown in Supplementary Figure S4.
Figure 6. Forest plot of the effect of physical exercise on HDL cholesterol levels.
Triglycerides
Nine studies assessed triglyceride levels [Reference Zhou, Guan, Xiu and Wu42–Reference Scheewe, Backx, Takken, Jörg, van Strater and Kroes46, Reference Fernández-Abascal, Suárez-Pinilla, Cobo-Corrales, Crespo-Facorro and Suárez-Pinilla48–Reference Masa-Font, Fernández-San-Martín, Martín López, Alba Muñoz, Oller Canet and Martín Royo51]. Our results indicated that physical exercise interventions did not significantly affect triglyceride levels (SMD = -0.041, 95% CI [−0.461, 0.38], p = 0.849). The included studies showed significant heterogeneity (Q = 33.432, I 2 = 94.865%, p < 0.001) (Figure 7). The selection models reported no evidence of publication bias (p = 0.871), which was in line with the results of Egger’s regression (p = 0.413). The mean model estimates are presented in Supplementary Figure S5.
Figure 7. Forest plot showing the effect of physical exercise on triglyceride levels.
Glucose
Ten studies evaluated the glucose levels [Reference Zhou, Guan, Xiu and Wu42–Reference Masa-Font, Fernández-San-Martín, Martín López, Alba Muñoz, Oller Canet and Martín Royo51]. Exercise intervention had no significant effect on glucose levels (SMD = −0.071, 95% CI [−0.213, 0.071], p = 0.326). No heterogeneity was observed among the included studies (Q = 5.323, I 2 = 0%, p = 0.805) (Figure 8). The selection model analysis revealed no evidence of publication bias (p = 0.244). Egger’s regression supported this finding (p = 0.454) [39, Reference Higgins, Thompson, Deeks and Altman52]. The mean model estimates are presented in Supplementary Figure S6.
Figure 8. Forest plot showing the effect of physical exercise on glucose levels.
Exercise interactions with medication
No studies have examined the possible interactions between physical exercise and medication (e.g., antipsychotics, antidepressants, or anxiolytics). Therefore, the secondary aim of this meta-analysis could not be addressed.
Discussion
This meta-analysis evaluated whether structured physical exercise can improve metabolic syndrome components in individuals with psychotic disorders. Contrary to our hypothesis, we observed no significant benefits of exercise on metabolic syndrome. However, differences across studies make it difficult to draw definitive conclusions. Variability in exercise programs, how outcomes were measured, and the characteristics of study participants underscored the complexity of addressing metabolic dysregulation in this high-risk group. Critically, none of the included studies explored the interaction between exercise and psychotropic medications, which is a key factor that could influence the results. This gap prevented us from evaluating secondary objectives.
According to GRADE assessments, the overall certainty of evidence supporting the effects of physical exercise on metabolic outcomes in individuals with psychotic disorders was rated as low to moderate. This was mainly attributed to the presence of moderate-to-high heterogeneity across trials, risk of bias in several domains, and imprecision of the pooled estimates.
Heterogeneity in exercise interventions and metabolic outcomes
Patients with severe mental illnesses, especially those with psychotic disorders such as schizophrenia, face an elevated risk of developing metabolic syndrome due to intertwined biological, inflammatory, and iatrogenic factors [Reference Polcwiartek, O’Gallagher, Friedman, Correll, Solmi and Jensen7, Reference Correll, Detraux, De Lepeleire and De Hert17, Reference Schmitt, Maurus, Rossner, Röh, Lembeck and von Wilmsdorff18, Reference DE Hert, Schreurs, Vancampfort and VAN Winkel53]. Antipsychotics promote central obesity and insulin resistance through weight gain and adipocyte dysfunction [Reference Correll, Detraux, De Lepeleire and De Hert17, Reference Schmitt, Maurus, Rossner, Röh, Lembeck and von Wilmsdorff18], while psychotic disorders related inflammation independently worsens cardiometabolic risk [Reference Henderson, Vincenzi, Andrea, Ulloa and Copeland9, Reference Lambert, Copeland, Sampson and Duffy10, Reference Pillinger, McCutcheon, Vano, Mizuno, Arumuham and Hindley20].
Despite the well-established link between metabolic syndrome and cardiovascular disease [Reference Mottillo, Filion, Genest, Joseph, Pilote and Poirier12], the included trials assessed its components inconsistently (e.g., prioritizing isolated markers such as waist circumference or triglycerides over syndromic criteria). This methodological inconsistency underscores the need for standardized protocols in future studies [Reference Zhou, Guan, Xiu and Wu42–Reference Masa-Font, Fernández-San-Martín, Martín López, Alba Muñoz, Oller Canet and Martín Royo51].
Methodological heterogeneity in defining and measuring metabolic syndrome components significantly hampered cross-study comparability. Key examples include waist circumference measurement, protocols between the WHO-standardized midpoint (iliac crest to lower rib margin), umbilical positioning, and inconsistent subject posture or respiratory phase documentation. Despite universal reporting of triglyceride and HDL, lipid assessments were similarly compromised by variable fasting durations (8–12 h), sample processing methods, and assay techniques. This reflects nonadherence to the established diagnostic criteria (e.g., IDF and NCEP-ATP III). Future trials should implement standardized protocols (WHO STEPPS for anthropometry and CDC-NHLBI guidelines for lipids) with explicit methodological reporting. The development of harmonized COS for SMI metabolic research would enable meaningful evidence synthesis.
Notably, multimodal interventions combining aerobic and resistance training demonstrated modest benefits [Reference Scheewe, Backx, Takken, Jörg, van Strater and Kroes46, Reference Jae, Kil, So, Lee, Song and Kim47], supporting that varied exercise regimens enhance metabolic flexibility [Reference Heissel, Heinen, Brokmeier, Skarabis, Kangas and Vancampfort27, Reference Yang, Yuan, Zhang, Fu, Wang and Wang29, Reference Vancampfort, Stubbs, Mitchell, De Hert, Wampers and Ward54]. On the other hand, trials using only aerobic or low-intensity workouts reported no benefits, suggesting that both exercise type and intensity play a critical role [Reference Fernández-Abascal, Suárez-Pinilla, Cobo-Corrales, Crespo-Facorro and Suárez-Pinilla48, Reference Wu, Wang, Bai, Huang and Lee50, Reference Masa-Font, Fernández-San-Martín, Martín López, Alba Muñoz, Oller Canet and Martín Royo51]. While high-intensity interval training (HIIT) has potential advantages, it is difficult to implement in this population owing to practical challenges, including motivation and tolerance issues. A key limitation of this study is the paucity of trials comparing exercise modalities. We could not determine the optimal exercise type for metabolic improvement with only two HIIT studies, one yoga trial, and two combined training regimens. This heterogeneity prevents robust recommendations regarding specific protocols (e.g., HIIT vs. resistance training) and underscores the need for head-to-head comparative trials.
We aggregated data across schizophrenia, schizoaffective disorder, and schizophreniform disorder owing to their shared cardiometabolic risk profiles and universal exposure to second-generation antipsychotics (SGAs), which frequently induce weight gain, dyslipidemia, and glucose intolerance. While the illness course and symptomatology show some heterogeneity, this grouping aligns with established psychiatric research paradigms and reflects clinical practice realities. Future studies with larger sample sizes may facilitate diagnostic subgroup analysis.
Blood pressure and exercise intensity
Exercise interventions were not associated with significant reductions in systolic or diastolic blood pressure compared to nonexercise controls, with moderate to high heterogeneity across studies. These findings are consistent with prior evidence suggesting that exercise alone, without adjunctive dietary or pharmacological strategies, may be insufficient to treat hypertension in individuals with psychotic disorders [Reference Polcwiartek, O’Gallagher, Friedman, Correll, Solmi and Jensen7]. Notably, antipsychotic medications have been associated with hypertension, especially second-generation medications such as clozapine and olanzapine, through mechanisms such as weight gain, sodium retention, and sympathetic activation [Reference Correll, Detraux, De Lepeleire and De Hert17, Reference Schmitt, Maurus, Rossner, Röh, Lembeck and von Wilmsdorff18, Reference Newcomer55]. While Bredin et al. [Reference Bredin, Warburton and Lang56] reported blood pressure improvements after 12 weeks of aerobic or resistance training, their cohort lacked severe psychiatric comorbidities, potentially inflating generalizability. Exercise intensity is a major moderator of training outcomes; studies involving high-intensity interval training (HIIT; 70–95% HR peak) or vigorous aerobic protocols (50–90% HRmax) [Reference Heggelund, Nilsberg, Hoff, Morken and Helgerud43, Reference Romain, Fankam, Karelis, Letendre, Mikolajczak and Stip44] demonstrated greater metabolic adaptations than moderate-intensity regimens [Reference Fernández-Abascal, Suárez-Pinilla, Cobo-Corrales, Crespo-Facorro and Suárez-Pinilla48, Reference Brobakken, Nygård, Güzey, Morken, Reitan and Heggelund49], in line with the dose–response relationships observed in nonpsychiatric populations [Reference Hwang, Yoo, Kim, Hwang, Handberg and Petersen57–Reference Ramos, Dalleck, Tjonna, Beetham and Coombes59]. However, HIIT may not be feasible in psychosis, where motivational deficits, sedation with antipsychotics, and cognitive impairments pose significant barriers. Large trials stratifying patients according to illness severity and medication regimens are needed to clarify optimal intensity thresholds.
Lipid profiles and medication interactions
Despite the role of aerobic exercise in improving lipid metabolism [Reference Vancampfort, Rosenbaum, Schuch, Ward, Richards and Mugisha33], our meta-analysis did not find any significant effects on triglyceride or HDL cholesterol levels. This discrepancy likely reflects pronounced metabolic dysregulation in the SMI. Antipsychotics directly exacerbate lipid abnormalities via hepatic lipogenesis (endogenous hepatic synthesis of fatty acids from carbohydrates) and adipocyte dysfunction (impaired function of fat-storing cells) [Reference McIntyre, Soczynska, Konarski and Kennedy14, Reference Buhagiar and Jabbar60]. For instance, clozapine and olanzapine elevate triglyceride levels by up to 40% and reduce HDL levels by 15–20% in psychotic disorders [Reference Correll, Detraux, De Lepeleire and De Hert17, Reference Schmitt, Maurus, Rossner, Röh, Lembeck and von Wilmsdorff18, Reference Newcomer55], which may overwhelm the modest lipid-modulating capacity of exercise. A pilot study demonstrating HDL reduction at week 12 of training [Reference Bredin, Warburton and Lang56] further highlights the complexity of the lipid response in medicated populations. Although exercise may mitigate antipsychotic-induced hypertriglyceridemia without compromising psychiatric stability [Reference Kim, Lang, Warburton, Barr, Smith and Thornton61], the mechanisms involved remain unclear. The proposed pathways include enhanced lipoprotein lipase activity and β oxidation [Reference Mann, Beedie and Jimenez62], though inflammation and oxidative stress, which are hallmarks of psychosis, may blunt these adaptations. Future studies must incorporate direct biomarkers (e.g., apolipoprotein B and LDL particle size) and control for medication dose, duration, and polypharmacy to disentangle these interactions.
Standardization extends beyond measurement techniques to encompass the diagnostic criteria. Only 3 of 10 included studies applied established metabolic syndrome definitions (e.g., IDF or NCEP-ATP III), while others reported isolated components [Reference Romain, Fankam, Karelis, Letendre, Mikolajczak and Stip44, Reference Scheewe, Backx, Takken, Jörg, van Strater and Kroes46, Reference Fernández-Abascal, Suárez-Pinilla, Cobo-Corrales, Crespo-Facorro and Suárez-Pinilla48]. These risks misclassify outcomes and dilute actual intervention effects. Future work should mandate adherence to internationally recognized criteria and report individual components along with composite syndrome status.
Glucose metabolism and adherence challenges
Data from all studies in this meta-analysis assessed glucose levels after exercise interventions, but limited improvements were observed despite exercise is established to enhance insulin sensitivity [Reference Correll, Solmi, Veronese, Bortolato, Rosson and Santonastaso3, Reference Lambert, Copeland, Sampson and Duffy10, Reference Henderson, Cagliero, Copeland, Borba, Evins and Hayden63]. Previous studies using an intervention based on aerobic training alone [Reference Fernández-Abascal, Suárez-Pinilla, Cobo-Corrales, Crespo-Facorro and Suárez-Pinilla48, Reference Wu, Wang, Bai, Huang and Lee50] did not improve hyperglycemia, perhaps because of suboptimal intensity, insufficient duration, or poor adherence. Of note, the only adherence report – a critical determinant of exercise efficacy – reflected data at point, FFD 30 months and FFD 40 months in five trials with an adherence range of 60–75% [Reference Romain, Fankam, Karelis, Letendre, Mikolajczak and Stip44, Reference Fernández-Abascal, Suárez-Pinilla, Cobo-Corrales, Crespo-Facorro and Suárez-Pinilla48, Reference Brobakken, Nygård, Güzey, Morken, Reitan and Heggelund49]. However, one trial reported a 50% dropout rate in the intervention arm, which was primarily attributed to motivational deficits and program duration [Reference Romain, Fankam, Karelis, Letendre, Mikolajczak and Stip44].
Attrition of this type is consistent with larger evidence associating poor adherence in SMI populations with psychosocial obstacles (e.g., cognitive impairment and social isolation) and adverse consequences of medications (e.g., sedation and fatigue) [Reference Korman, Stanton, Vecchio, Chapman, Parker and Martland22]. The lack of standardized adherence metrics across studies – half omitted attendance thresholds or compliance criteria – further complicates the interpretation. These findings underscore the necessity of integrating behavioral strategies (e.g., motivational interviewing and peer support) to sustain engagement in this population.
The lack of glucose-lowering effects may also indicate profound metabolic dysregulation inherent to psychotic disorders. Antipsychotics, including olanzapine and clozapine, modulate insulin signaling via direct β-cell toxicity and adipose tissue inflammation [Reference Mendelson64, Reference Nielsen, Skadhede and Correll65], which could cancel out exercise-induced benefits. Furthermore, exercise prescriptions in the included trials rarely accounted for medication specific metabolic risks or individualized glycemic targets. Future interventions should prioritize stratified designs that control for antipsychotic class, dose, and illness chronicity while incorporating continuous glucose monitoring to capture dynamic responses.
Antipsychotic medications: A critical confounder
All participants were prescribed antipsychotics, which are known to promote weight gain, dyslipidemia, and insulin resistance [Reference Mendelson64, Reference Fleischhacker, Siu, Bodén, Pappadopulos, Karayal and Kahn66, Reference Allison, Mentore, Heo, Chandler, Cappelleri and Infante67]. One trial exclusively included patients treated with clozapine [Reference Pillinger, McCutcheon, Vano, Mizuno, Arumuham and Hindley20, Reference Wu, Wang, Bai, Huang and Lee50], a subgroup with a greater metabolic risk [Reference Mendelson64, Reference Nielsen, Skadhede and Correll65]. This timing is further challenged by the progressive metabolic burden in chronic (vs. first-episode) psychosis [Reference Nielsen, Skadhede and Correll65, Reference Fleischhacker, Siu, Bodén, Pappadopulos, Karayal and Kahn66, Reference Kessler, Sisson and Short68, Reference Izquierdo, De Souto Barreto, Arai, Bischoff-Ferrari, Cadore and Cesari69].
Although structured exercise interventions offer multiple health benefits, this meta-analysis found no evidence that exercise improves metabolic syndrome in individuals with SMI receiving antipsychotic treatment. Therefore, although physical activity should remain a safe and broadly beneficial component, it appears insufficient to overcome the metabolic burden associated with antipsychotics. Pharmacodynamic interactions, such as how exercise alters drug metabolism or receptor sensitivity, are understudied, but should be urgently investigated [Reference Izquierdo, De Souto Barreto, Arai, Bischoff-Ferrari, Cadore and Cesari69].
In all trials selected for inclusion, participants were administered antipsychotics [Reference Zhou, Guan, Xiu and Wu42–Reference Masa-Font, Fernández-San-Martín, Martín López, Alba Muñoz, Oller Canet and Martín Royo51], known independent agents of metabolic dysregulation via mechanisms of weight gain, hypertriglyceridemia, and insulin resistance [Reference Correll, Detraux, De Lepeleire and De Hert17, Reference Schmitt, Maurus, Rossner, Röh, Lembeck and von Wilmsdorff18, Reference Mendelson64]. Reversible with dose reduction, second-generation antipsychotics (SGAs) such as clozapine, olanzapine, risperidone, and quetiapine worsen cardiovascular risk through direct actions on lipid metabolism (elevating triglycerides, reducing HDL) and glucose homeostasis [Reference Henderson, Cagliero, Copeland, Borba, Evins and Hayden63, Reference Samaras, Correll, Mitchell and De Hert70, Reference Yood, DeLorenze, Quesenberry, Oliveria, Tsai and Willey71]. Clozapine, in particular, demonstrates the most severe metabolic liabilities, with studies reporting a fourfold increased risk of type 2 diabetes and 20–25% weight gain within the first year of treatment [Reference Mendelson64, Reference Nielsen, Skadhede and Correll65].
Critically, none of the trials examined exercise–psychotropic medication interactions, despite the possibility that physical activity may modulate drug pharmacokinetics (e.g., hepatic CYP450 enzyme activity) or counteract receptor-mediated metabolic harm (e.g., 5-HT2C antagonism-induced hyperphagia) [Reference Haddad and Sharma72, Reference Chaouloff73]. This limitation greatly hinders our ability to separate the independence of exercise as a therapeutic modality from its potential ameliorative effect as a treatment adjunct (e.g., metabolic salvage therapy) in drug-treated populations.
Chronicity and the case for early intervention
The higher proportion of cases classified as chronic psychosis in our analysis [Reference Zhou, Guan, Xiu and Wu42–Reference Masa-Font, Fernández-San-Martín, Martín López, Alba Muñoz, Oller Canet and Martín Royo51] is consistent with the progressive nature of metabolic syndrome in SMI, whereby long exposure to antipsychotic medications and illness-related determinants (e.g., chronic inflammation, sedentariness) contribute to increased risk over time [Reference Nielsen, Skadhede and Correll65, Reference Fleischhacker, Siu, Bodén, Pappadopulos, Karayal and Kahn66, Reference Kessler, Sisson and Short68]. In contrast, patients with first-episode disease have a lower metabolic burden of morbidity, leaving a crucial time window for early intervention [Reference Kessler, Sisson and Short68, Reference Fitzgerald, Sahm, Byrne, O’Connell, Ensor and Ní Dhubhlaing74]. Structured exercise may be protective when significant weight gain and insulin resistance begin, and exercise prevention studies initiated upon illness onset may help counter eating-induced weight gain and insulin resistance before irreversible pathophysiological derangements occur [Reference Fitzgerald, Sahm, Byrne, O’Connell, Ensor and Ní Dhubhlaing74]. However, pragmatic barriers, including motivational deficits, medication induced fatigue, and limited access to tailored programs, hinder adherence in this population [Reference Firth, Solmi, Wootton, Vancampfort, Schuch and Hoare75]. Adding exercise-coordinated specialty care models, nutritional counseling, and peer support may enhance engagement and sustainability. Our findings must be interpreted in the light of several limitations. The small, heterogeneous sample size of trials and inconsistent outcome report constrain generalizability. Although statistical tests did not detect publication bias, this remains plausible, given the niche focus of metabolic interventions in psychosis [Reference Zhou, Guan, Xiu and Wu42–Reference Masa-Font, Fernández-San-Martín, Martín López, Alba Muñoz, Oller Canet and Martín Royo51]. Critically, intervention heterogeneity, marked by disparate exercise modalities, intensities, and durations, precluded dose–response insights, while the pervasive metabolic effects of second-generation antipsychotics (SGAs) likely obscure exercise-specific benefits.
Furthermore, adherence variability, compounded by poor reporting and high attrition rates, introduces potential bias. To address these gaps, future studies should standardize metabolic syndrome criteria (e.g., incorporating direct biomarkers such as insulin sensitivity), stratified analyses by antipsychotic class, dose, and illness chronicity, and integrate evidence-based behavioral strategies to enhance adherence. Such refinements are essential for isolating the therapeutic potential of exercise within the complex biopsychosocial landscape of severe mental illnesses.
Future studies should explicitly evaluate the interactions between specific antipsychotic agents and exercise-induced metabolic responses. This includes stratifying participants by antipsychotic type (e.g., clozapine vs. aripiprazole) and dose and exploring potential mechanisms such as altered receptor binding, hepatic metabolism (e.g., CYP450 modulation), and downstream effects on lipid and glucose regulation. A deeper understanding of these interactions will enable more tailored and clinically relevant exercise prescriptions for this population.
Conclusions
While structured exercise alone yields disappointing metabolic outcomes in chronic psychosis, targeted interventions are promising for precise clinical scenarios. Early intervention cohorts, particularly first-episode patients before irreversible metabolic derangement, and those on lower risk antipsychotics (e.g., aripiprazole/lurasidone rather than olanzapine/clozapine) may derive robust benefits from multimodal regimens combining aerobic and resistance training ≥3x/week at 60–80% HRmax. Consequently, we advocate integrating exercise into a broader cardiometabolic risk mitigation strategy: (1) optimizing pharmacotherapy to minimize obesogenic agents, (2) codelivering structured dietary interventions, and (3) employing metformin for antipsychotic-induced hyperglycemia, with continuous metabolic monitoring. Future studies must stratify by illness chronicity, medication risk profiles, and exercise modalities to identify responders.
Future trials should prioritize stratification by antipsychotic type and dosage and rigorously investigate medication specific interactions with exercise. Such approaches are essential to improve the precision of clinical recommendations and develop effective, individualized interventions that mitigate cardiometabolic risk in individuals with psychotic disorders.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1192/j.eurpsy.2025.10064.
Data availability statement
Data available from the authors upon reasonable request.
Financial support
Prof. Mikel Izquierdo was supported in part by “Proyectos de I + D + i”» de los Programas Estatales de Generación de Conocimiento y Fortalecimiento Científico y Tecnológico del Sistema de I + D + i Orientada a los Retos de la Sociedad, en el marco del Plan Estatal de Investigación Científica y Técnica y de Innovación 2017–2020 (PID2020-113098RBI00) and by the European Union and Ministerio de Ciencia e Innovación, Agencia Estatal de Investigación (AEI), Unión Europea NextGenerationEU/PRTR with number PLEC2021-0080171 and by EU Funded H2020-DIABFRAIL-LATAM (H2020-SC1-BHC-825546).
Competing interests
The authors declare none.
Open access
Comments
No Comments have been published for this article.