Link between schizophrenia and risk of type 2 diabetes

We conducted searches on PubMed, Cochrane Library, Embase, and Web of Science to identify observational studies published from the inception of the databases to September 19, 2023. The language was restricted to English, and our search strategy incorporated a combination of medical subject headings (MeSH) and keywords. The search terms encompassed a range of topics, including schizophrenia, schizophreni*, Dementia Praecox, Diabetes Mellitus, Diabetes Insipidus, Diet, Diabetic, Prediabetic State, Scleredema Adultorum, Glucose Intolerance, and Gastroparesis. Additionally, we scrutinized the reference lists of included cohort studies, case-control studies, cross-sectional studies, and other published meta-analyses to identify relevant trials.

The inclusion criteria for trials were as follows (1): observational studies were considered, with the exception of intervention studies (2); the observation group comprised patients diagnosed with schizophrenia, while the control group consisted of individuals without schizophrenia or comparisons were made with large datasets containing prevalence data on T2DM (3); the original study should accurately diagnose both schizophrenia and T2DM (4); trials that did not recruit a control group but utilized previously published general population data were considered (5); preference was given to trials that included both baseline and follow-up data, with prioritization given to the latter. Trials with low NOS or AHRQ scores were excluded. In cases where multiple studies reported data from the same cohort, priority was given to the study with the longest follow-up or the largest number of participants. Trials presenting excessively wide 95% confidence intervals (CI) were excluded. Additionally, the following types of articles were excluded: conference abstracts, study protocols, duplicate publications, and studies lacking outcomes of interest. In instances of mixed samples, efforts were made to extract data specifically related to individuals with schizophrenia. If such data extraction was not feasible, attempts were made to contact the authors up to two times within a one-month period to obtain data specifically for individuals with schizophrenia. Trials where contact was unsuccessful were excluded.

Two reviewers (KD and PS) independently screened the literature based on the eligibility and exclusion criteria. Initially, duplicate and irrelevant articles were excluded by assessing their titles and abstracts. Subsequently, the full texts of potentially eligible articles were retrieved and thoroughly reviewed to identify all suitable studies. Any discrepancies were resolved through discussion with a third reviewer (PS), serving as an arbiter.

The process of data extraction was meticulously carried out by the two aforementioned reviewers (DK, SHW,CHQ, KWZ), who referred to established guidelines for systematic reviews and meta-analysis (31). Utilizing predefined forms, they systematically extracted key information such as the first author, year of publication, country, WHO region, study type, sample size, follow-up years, year of data collection, percentage of males, age, diagnosis of schizophrenia/T2DM, and adjustments made for confounders. In instances where discrepancies arose, the reviewers engaged in thorough discussions with PS, serving as a mediator, to achieve a consensus and ensure the accuracy and reliability of the extracted data.

To gauge the methodological quality of cohort and case-control studies, the NOS was employed (32). The scoring system allocated stars on a scale of 0 to 9 for both cohort and case-control studies, with four stars designated for the selection of participants and measurement of exposure, two stars for comparability, and three stars for the assessment of outcomes and adequacy of follow-up. A higher number of stars signified a higher quality of the study. Scores falling within the ranges of 0–3, 4–6, and 7–9 were categorized as indicating low, moderate, and high quality, respectively. For the evaluation of cross-sectional studies, the AHRQ was employed (33). This scale comprises 11 items, with each item assessed using “yes”, “no”, or “unclear”. The scoring method involves assigning points for each “yes” response, resulting in a total score ranging from 0 to 11 points. Scores within the ranges of 0–3, 4–7, and 8–11 were interpreted as indicative of low, moderate, and high quality, respectively.

To assess the association between schizophrenia and the risk of diabetes, the adjusted odds ratios (OR) and their corresponding 95% confidence intervals (CI) from each trial were utilized. Heterogeneity was evaluated using the χ2-test and I2-values. In cases where P > 0.1 and I2 ≤ 50%, indicating minimal heterogeneity, a fixed-effects model was employed. However, if I2 > 50%, suggesting significant heterogeneity, a random-effects model was applied. To ensure the robustness of the overall effects, a sensitivity analysis was conducted by systematically excluding one study at a time and re-running the analysis. Publication bias was visually inspected through a funnel plot, and Egger’s regression test was employed for a statistical assessment of publication bias. Subgroup analyses were performed based on gender, study type, WHO region, year of data collection, and follow-up time to provide a more nuanced understanding of the results. All statistical analyses were executed using Stata statistical software version 14.0 (Stata Corp, College Station, Texas).

A systematic search of observational studies published up to September 19, 2023, generated a total of 2,419 results. Upon the removal of duplicate entries, the screening process involved the assessment of 1,810 abstracts and titles (Figure 1). Following this initial screening, 55 articles were identified as potentially relevant, of which 23 were subsequently excluded with detailed reasons provided. Ultimately, after a comprehensive full-text review, 32 studies (34–65) were included in the analysis. Figure 1 provides a concise summary of the search results, elucidating the rationale behind the exclusion of specific articles.

This meta-analysis aggregates findings from 32 observational studies, encompassing a substantial cohort of 2,007,168 individuals diagnosed with schizophrenia, alongside a comparison group comprising 35,883,980 individuals without schizophrenia. These studies were conducted and published between 2004 and 2023, showcasing a broad spectrum of research methodologies. Among them, 14 were cohort studies, two were case-control studies, and the remainder consisted of fifteen cross-sectional studies. The majority of participants in these investigations commenced follow-up at the age of 16 or older, with only one study focusing on individuals aged 0 to 36 years. Across all studies, diagnostic criteria for schizophrenia were consistently applied, ensuring a uniform standard across the analysis. The duration of follow-up varied across studies, ranging from 1 to 36 years, with one study exclusively focusing on a male cohort. Notably, adjusted estimates were available for nearly all studies, although adjustments for confounding variables may have differed slightly between studies. Detailed characteristics of the included trials are provided in Table 1 for reference and clarity.

Following the assessment based on the NOS for cohort and case-control studies and the AHRQ criteria for cross-sectional studies, the average NOS score for all included cohort and case-control studies was 7.12. Similarly, the average AHRQ score for cross-sectional studies was 6.73. These scores collectively affirm the high quality of all observational studies incorporated in this meta-analysis. Table 1 presents the individual scores of each included study, providing a comprehensive overview of the meticulous quality assessment conducted according to the specified criteria. The consistently high scores across these studies underscore the robustness and reliability of the evidence synthesized in this meta-analysis.

A comprehensive analysis of thirty-one observational studies (34–43, 45–65) investigated the relationship between a history of schizophrenia and the risk of T2DM. The pooled results revealed a significant association, indicating that individuals with a history of schizophrenia face a heightened risk of developing T2DM (OR = 2.15; 95% CI: 1.83–2.52; I2 = 98.9%, P < 0.001; Figure 2). The substantial heterogeneity, reflected in the I2 statistic, underscores the variability among the included studies, while the low p-value highlights the statistical significance of the observed association. To ensure the robustness of these findings, a sensitivity analysis was conducted. Encouragingly, none of the individual studies within the pool reversed the overall effect size, confirming the stability and reliability of the results (Figure 3). These insights contribute valuable knowledge to understanding the link between schizophrenia and the increased risk of T2DM, offering potential implications for clinical practice and avenues for further research.

In the examination of the included studies, a thorough subgroup analysis was conducted based on gender, study type, WHO region, and follow-up time, with detailed results presented in Table 2. For gender (Figure 4), an in-depth subgroup analysis was performed on eleven studies (35, 41, 43, 44, 46, 51, 53, 58, 60, 61, 63) within the trial comparisons. The findings indicated that females (OR=2.12; 95% CI: 1.70-2.64; I2 = 90.7%, P < 0.001) with a history of schizophrenia face a significantly higher risk of T2DM compared to males (OR=1.68; 95% CI: 1.39-2.04; I2 = 91.3%, P < 0.001). In terms of WHO region (Figure 5), a detailed subgroup analysis was conducted on twenty-nine studies (34–43, 45–47, 49–55, 57–65) within the trial comparisons. The studies were categorized into three subgroups: WPRO (35–37, 39, 55, 57, 58, 60, 62), EURO (34, 40–43, 46, 47, 49, 50, 52, 53), and AMRO (38, 45, 51, 54, 59, 61, 63–65). The within-trial comparisons revealed that EURO (OR=2.73; 95% CI: 2.23-3.35; I2 = 97.5%, P < 0.001) had a significantly higher risk of T2DM than WPRO (OR=1.72; 95% CI: 1.32-2.23; I2 = 95.2%, P < 0.001) and AMRO (OR=1.82; 95% CI: 1.40-2.37; I2 = 99.1%, P < 0.001). In the analysis of study types (Figure 6), we conducted a subgroup analysis involving thirty-one studies (34–43, 45–65) within the trial comparisons. Among these, fifteen studies (34, 35, 37, 41, 42, 45, 49, 50, 54–56, 59, 62–64) belonged to cross-sectional studies, while two studies (48, 52) were categorized as case-control studies. The remaining fourteen studies fell under the cohort studies category. Across all study types in within-trial comparisons (Figure 7), it was consistently observed that schizophrenia poses a significant risk for T2DM. The results for each study type were as follows: cohort study (OR=2.11; 95% CI: 1.83-2.67; I2 = 98.4%, P < 0.001), case-control study (OR=2.58; 95% CI: 1.34-4.97; I2 = 72.6%, P = 0.056), and cross-sectional study (OR=2.04; 95% CI: 1.47-2.83; I2 = 99.1%, P < 0.001). Regarding follow-up years, we conducted a subgroup analysis involving sixteen studies (36–40, 43, 46, 47, 51–53, 57, 58, 60, 61, 65) within the trial comparisons. These studies were further divided into three subgroups based on follow-up duration: <10 years (36, 37, 51, 53, 58, 60, 65), 10-20 years (38, 46, 47), and >20 years (39, 40, 43, 52, 57, 61). The risk of developing T2DM in patients with schizophrenia was found to be associated with the duration of the disease. Notably, the >20 years subgroup showed a significantly higher risk of T2DM (OR=3.17; 95% CI: 1.24-8.11; I2 = 99.4%, P < 0.001) compared to the 10-20 years group (OR=2.26; 95% CI: 1.76-2.90; I2 = 98.6%, P < 0.001) and the <10 years group (OR=1.68; 95% CI: 1.30-2.19; I2 = 95.4%, P < 0.001).

Upon visually examining the funnel plot, there was no discernible evidence suggesting a significant publication bias in the analysis of schizophrenia disorders and their association with the risk of T2DM (Figure 8). However, the Egger’s regression test (P = 0.010) indicated a noteworthy presence of publication bias within the scope of our meta-analysis.

This meta-analysis encompasses 32 observational studies, involving 2,007,168 individuals with schizophrenia and 35,883,980 without schizophrenia. It offers a thorough assessment of the correlation between schizophrenia and T2DM. Our findings reveal a notable escalation in the risk of T2DM among individuals with schizophrenia, demonstrating an overall 2.15-fold increase in risk compared to controls without schizophrenia. When considering recent observational studies, these results further substantiate schizophrenia as a significant risk factor for the development of T2DM.

Previous meta-analysis investigated the association between schizophrenia and T2DM (28, 29). The results showed that schizophrenia increased the risk of T2DM. However, they did not analyze subgroups for WHO region, gender, study type and follow-up time. We added more recent studies and analyzed the data according to the above subgroups, so as to provide strong evidence for the association between schizophrenia and T2DM, the previous meta-analysis did not show these meaningful conclusions.

To date, there have been limited studies investigating the association between schizophrenia and T2DM. While various mechanisms underlie the comorbidities between schizophrenia and T2DM, a consensus statement is yet to be established. In terms of genetics, schizophrenia and T2DM share numerous overlapping risk loci (12, 66), including but not limited to chromosomes 1p13, 1p36, 1q21–24, 1q25, 2q14, 2q33, and 2q36. Certain gene regions within these loci may play a role in the pathogenesis of T2DM in individuals with schizophrenia. Prior research has identified a reduction in dendritic spine density in the brains of individuals with schizophrenia (67, 68). This reduction is influenced by both Rho GTPase and the Wnt/β-Catenin pathway through distinct mechanisms (69). These pathways contribute to disruptions in insulin biosynthesis, thereby increasing susceptibility to T2DM in individuals with schizophrenia compared to the general population. Regarding inflammatory factors, a meta-analysis reported elevated levels of IL-6, IL-1β, and TNF-α in the blood and cerebrospinal fluid of individuals with schizophrenia (70). The heightened levels of these cytokines may potentially accelerate the progression of insulin resistance (71). The alterations in the immune system and inflammatory components induced by chronic stress are associated with the molecular mechanisms of T2DM in individuals with schizophrenia (72). Concerning oxidative stress, PON1 emerges as a candidate gene implicated in both schizophrenia and T2DM. The enzyme PON1 plays a crucial role in mitigating oxidative stress and exhibits an inverse relationship with cytokine levels (73). In individuals diagnosed with schizophrenia, there is a notable reduction in PON1 enzyme activity, and this diminishing trend adversely impacts the normal functioning of β-cells. Patients with schizophrenia often require prolonged use of antipsychotic medications, including olanzapine, clozapine, haloperidol, sertindole, and other commonly prescribed antipsychotics. These medications are associated with an increased susceptibility to metabolic disorders, particularly disruptions in glucose homeostasis leading to the progression from insulin resistance to T2DM (74, 75). Furthermore, the detrimental lifestyle habits, such as smoking, and cognitive dysfunction exhibited by individuals with schizophrenia can directly or indirectly influence the daily blood sugar levels of these patients (75, 76). This, in turn, contributes to a heightened risk of T2DM incidence.

In the subgroup analysis, several noteworthy results emerged that could provide valuable insights for clinicians. Notably, females with a history of schizophrenia exhibit a significantly elevated risk of developing T2DM when compared to their male counterparts. This finding underscores a significant susceptibility of females to T2DM, aligning with prior research that indicated women taking antipsychotics faced a higher likelihood of T2DM development compared to men (77). This heightened risk in females may be attributed to factors such as weight gain and the emergence of insulin resistance mediated by sex-related genes. Notably, women are more prone to developing T2DM following antipsychotic intervention, a phenomenon associated with increased body weight (78). Furthermore, the expression of specific sex-related genes appears to render women more susceptible to insulin resistance than men (79). A comprehensive meta-analysis of cross-sectional studies revealed intriguing differences between genders in the context of T2DM. In men with T2DM, significantly lower testosterone levels were observed compared to controls, while women exhibited higher testosterone levels. Prospective studies complement these findings, indicating that men with elevated testosterone levels experience a 42 percent reduction in the risk of developing T2DM compared to controls. Conversely, heightened testosterone levels in women seem to correlate with an increased risk of T2DM development (80). In the context of follow-up times subgroup analysis, our calculations align with prior research, supporting the conclusion that the prevalence of T2DM in patients with schizophrenia increases with the duration of the disease (81). This underscores the importance of considering the longitudinal aspect when evaluating the association between schizophrenia and T2DM.

Our meta-analysis examining the relationship between a history of schizophrenia and the risk of T2DM reinforces the idea that schizophrenia constitutes a risk factor for the development of T2DM. This underscores the importance of heightened awareness regarding the risk of T2DM in individuals with schizophrenia, potentially aiding early clinicians in the identification of patients at risk for T2DM. Nonetheless, it is essential to acknowledge certain limitations inherent in our study. The data included in our analysis exhibited high heterogeneity, and despite a thorough examination, the source of this heterogeneity remained unidentified. Nevertheless, a majority of the studies incorporated in our analysis meticulously controlled for numerous confounding factors, enhancing the reliability of our conclusions. To further advance the field, future research endeavors should consider incorporating additional subgroups to diversify and enrich the scope of investigation. It is worth noting that our meta-analysis did not include covariate analysis. However, the studies included in our analysis implemented control measures for adjusted confounding factors, contributing to robust confounding bias control. This strengthens the credibility of our study’s findings and facilitates seamless translation into clinical practice.

The original contributions presented in the study are included in the article. Further inquiries can be directed to the corresponding author.