Introduction
The life expectancy of people with schizophrenia is around 20 years shorter than that of the general population, and 60% of the causes of premature death of people with schizophrenia are related to cardiovascular diseases (Pillinger, D'Ambrosio, McCutcheon, & Howes, Reference Pillinger, D'Ambrosio, McCutcheon and Howes2019). One of the most studied cardiovascular risk indicators is metabolic syndrome (MetS), which consists of a group of parameters that indicate the risk of developing cardiovascular disease and diabetes (Eckel, Grundy, & Zimmet, Reference Eckel, Grundy and Zimmet2005). The criteria most used for the diagnosis of MetS are those of IDF (International Diabetes Federation, 2006) and ATPIII (Adult Treatment Panel III, 2001). These differ from each other in the cut-off point of the parameters that are considered pathological.
The increased prevalence of MetS in patients with schizophrenia compared to the general population is widely recognised (Kraemer, Minarzyk, Forst, Kopf, & Hundemer, Reference Kraemer, Minarzyk, Forst, Kopf and Hundemer2011), and has been mainly attributed to the use of atypical antipsychotics (Newcomer et al., Reference Newcomer, Haupt, Fucetola, Melson, Schweiger, Cooper and Selke2002; Vancampfort et al., Reference Vancampfort, Stubbs, Mitchell, De Hert, Wampers, Ward and Correll2015), as well as other risk factors that accumulate during the disease period, such as sedentary lifestyles, poor nutrition, tobacco consumption and the lack of self-care due to the negative symptoms of the disease themselves (Bobes et al., Reference Bobes, Arango, Aranda, Carmena, Garcia-Garcia and Rejas2007). In the last decade, studies have been published with patients who had not received pharmacological treatment and who show that the metabolic alterations could not be exclusively due to antipsychotics (Kirkpatrick, Garcia-Rizo, Fernandez-Egea, Miller, & Bernardo, Reference Kirkpatrick, Garcia-Rizo, Fernandez-Egea, Miller and Bernardo2011; Pillinger et al., Reference Pillinger, Beck, Gobjila, Donocik, Jauhar and Howes2017; Pillinger, Beck, Stubbs, & Howes, Reference Pillinger, Beck, Stubbs and Howes2017). Taking into account these findings, various pieces of research (Chadda, Ramshankar, Deb, & Sood, Reference Chadda, Ramshankar, Deb and Sood2013; Cordes et al., Reference Cordes, Bechdolf, Engelke, Kahl, Balijepalli, Lösch and Moebus2017; Ryan, Sharifi, Condren, & Thakore, Reference Ryan, Sharifi, Condren and Thakore2004) propose a vulnerability hypothesis for the development of metabolic disorders that is independent of the use of antipsychotics in patients with schizophrenia. Along these lines, a recent systematic meta-review (Pillinger et al., Reference Pillinger, D'Ambrosio, McCutcheon and Howes2019) found, in addition to alterations in the central nervous system, significant associations between schizophrenia and alterations in other systems such as the endocrine, immune and cardio-metabolic systems. Likewise, there are studies that frame the symptoms of schizophrenia within a systemic disease that also has basal metabolic manifestations (Kirkpatrick, Miller, García-Rizo, & Fernandez-Egea, Reference Kirkpatrick, Miller, García-Rizo and Fernandez-Egea2014).
Despite these advances in the understanding of the deleterious effects of MetS and its possible causes, it is still not clear what the prevalence of MetS in drug-naïve individuals with psychotic disorder is. This is an important limitation as most of the risk factors associated to MetS may play a role and tend to accumulate during the first years of disease (such as tobacco, sedentarism or the use of medication). This may be reflected by the important variation of MetS in patients under medication, ranging from 35.3% (Mitchell, Vancampfort, De Herdt, Yu, & De Hert, Reference Mitchell, Vancampfort, De Herdt, Yu and De Hert2013; Vancampfort et al., Reference Vancampfort, Stubbs, Mitchell, De Hert, Wampers, Ward and Correll2015) to 49% (Kraemer et al., Reference Kraemer, Minarzyk, Forst, Kopf and Hundemer2011). To date, only two meta-analyses in drug-naïve patients with psychotic disorders have been conducted: Vancampfort et al. (Reference Vancampfort, Vansteelandt, Correll, Mitchell, De Herdt, Sienaert and De Hert2013) conducted research on cardio-metabolic abnormalities in drug-naïve, first-episode and multi-episode patients with schizophrenia. One of their findings was that there was no significant difference between untreated (10%) and first-episode (15.9%) patients. A second meta-analysis (Mitchell et al., Reference Mitchell, Vancampfort, De Herdt, Yu and De Hert2013) showed the prevalence in untreated patients was 9.8%. These two works report the most solid data; however, the authors of both papers highlight several limitations, such as the difficulty in independently analysing naïve patients with a first psychotic episode, since they included studies with first episodes exposed to antipsychotics for an indeterminate time. In addition, the ‘untreated’ patient group included patients in any phase of the disease, thus the prevalence in this group may be confounded by the presence of other risk factors that develop during the disease.
Taking into account those limitations and the need to clarify the prevalence of MetS in drug-naïve patients with psychosis, we conducted a meta-analysis of studies that strictly included first psychotic episodes with 0-day exposure to antipsychotic treatment, including a population aged above 18. This will lead to a clearer understanding of the prevalence of MetS in this population, allowing a better detection of such syndrome, and helping the development of specific interventions.
Methods
The systematic review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement (Moher, Liberati, Tetzlaff, & Altman, Reference Moher, Liberati, Tetzlaff and Altman2009) and the Meta-Analysis of Observational Studies in Epidemiology (MOOSE) (Stroup et al., Reference Stroup, Berlin, Morton, Olkin, Williamson, Rennie and Thacker2000). It also followed a protocol registered in PROSPERO (CRD42020180930).
Search strategy
We searched the Web of Science Core Collection, Embase and Medline via Embase and PubMed platforms from inception until November 2020. Our queries combined natural and controlled terms related to: (first-episode psychosis or first-episode schizophrenia or FEP or FES or psychosis or schizophrenia) AND (antipsychotic-naïve or antipsychotic-free or drug-naïve or drug-free or neuroleptic-naïve or neuroleptic-free or never-medicated or untreated) AND (cholesterol or high-density lipoprotein (HDL) or low-density lipoprotein (LDL) or triglycerides or lipids or lipoproteins or MetS or metabolic or blood pressure or metabolic dysregulation) (online Supplementary Table S3). We manually screened all the references from the previous reviews in the field and extracted relevant articles from the citations of the included manuscripts. Articles identified were screened as abstracts, and after the exclusion of those which did not meet our inclusion criteria, the full texts of the remaining articles were assessed for eligibility. Then, final decisions were made regarding their inclusion in the review. We completed our search by manually reviewing the references of the included articles and extracting additional titles. Authors were contacted for missing data and to clarify overlaps. We also searched grey literature, and conducted a cross-reference search of relevant included studies and previous reviews. More details are provided in online Supplementary Tables S6 and S12.
Study selection
Two independent co-authors (NGT and ARG) screened titles and abstracts to identify studies that met the inclusion criteria outlined above using Rayyan (Ouzzani, Hammady, Fedorowicz, & Elmagarmid, Reference Ouzzani, Hammady, Fedorowicz and Elmagarmid2016) software. The same two co-authors then considered eligible full texts among these articles and the final list of included articles was reached through consensus. The κ index was 0.931. Discrepancies over the eligibility of studies were resolved through discussion with additional co-authors (MRV and BCF).
Eligibility
Inclusion criteria: (i) studies on FEP patients; (ii) studies in which psychosis diagnosis was determined according to either DSM-IV, DSM IV-TR17, DSM-5 (American Psychiatric Association, 2013) or International Classification of Diseases, Ninth or Ten Revision (ICD-9 or ICD-10); (iii) studies on individuals with FEP defined by the study authors as either drug-naïve (0 days) or minimal exposure regardless of the duration to antipsychotics will be considered for systematic review and studies on individuals with FEP and drug-naïve (0-day exposure to antipsychotic treatment) will be included in prevalence meta-analysis; (iv) cross-sectional studies or baseline assessment of prospective and retrospective cohort studies; (v) studies in which MetS diagnosis was confirmed or rejected based on current endocrinal criteria; i.e. it was defined according to any of these four sets of criteria: ATPIII-A, IDF, JIS 2009 (Alberti et al., Reference Alberti, Eckel, Grundy, Zimmet, Cleeman, Donato and Smith2009), World Health Organization (Alberti & Zimmet, Reference Alberti and Zimmet1998); and (vi) subjects aged above 18.
Exclusion criteria: (i) studies on chronic patients (⩾5 years after the FEP), despite being naïve; (ii) studies on animals or in vitro; (iii) studies not designed to calculate prevalence: quasi-experimental studies as they are unsuitable for measuring prevalence, case and control studies as they are unsuitable for measuring prevalence, randomised clinical trials as they are not designed to calculate prevalence because their inclusion/exclusion criteria are often restrictive, and subjects are not representative of the general population (Munn, Moola, Lisy, Riitano, & Tufanaru, Reference Munn, Moola, Lisy, Riitano and Tufanaru2015); (iv) studies presenting data on MetS that did not fully meet any of the above four sets of criteria; and (v) subjects aged above 65 [if a small proportion (<5% of the sample is aged >65), the studies could be considered].
Data extraction
DistillerSR (Evidence Partners, Canada) was used for data extraction, full text and quality assessment. Variables on data collection forms included age, sex, country, ethnic origin, diagnosis, study design, MetS criteria and samples. Data were collected independently by two co-authors (NGT and IRG). Two other independent co-authors (MRV and BCF) were available for mediation when inconsistencies arose.
Quality assessment
The Joanna Briggs Institute (Munn et al., Reference Munn, Moola, Lisy, Riitano and Tufanaru2015) for observational studies was used. This scale assesses observational studies and data needed to obtain prevalence. Total scores range from 0 to 10. For the total score grouping, risk of bias in studies was judged as low (⩾7 points), moderate (4–6 points) and high (<4 points). We used two versions, one for cross-sectional (Munn et al., Reference Munn, Moola, Lisy, Riitano and Tufanaru2015) and another for cohort (Moola et al., Reference Moola, Moon, Tufanaru, Aromataris, Sears, Sfectcu and Currie2020) studies (online Supplementary Table S13).
Statistical analysis
We performed a meta-analysis of proportions for the prevalence of MetS in antipsychotic-naïve patients. Prevalence estimates and 95% CI were calculated using a random-effect model due to heterogeneity between the populations and characteristics of the included studies (Barendregt, Doi, Lee, Norman, & Vos, Reference Barendregt, Doi, Lee, Norman and Vos2013). When prevalence estimates tend towards 0% or 100%, it overestimates the weight of individual studies in the meta-analysis (Barendregt et al., Reference Barendregt, Doi, Lee, Norman and Vos2013). We generated Forest plots for the prevalence estimates and their 95% CI of the individual studies and pooled estimates. Forest plots were examined visually looking for potential outliers. We assessed heterogeneity between studies using the I 2 statistic, with an I 2 >50% indicating substantial heterogeneity according to others (Davies et al., Reference Davies, Segre, Estrade, Radua, De Micheli, Provenzani and Fusar-Poli2020; Higgins, Thompson, Deeks, & Altman, Reference Higgins, Thompson, Deeks and Altman2003). We assessed the publication bias graphically using a funnel plot and the Egger's test (Egger, Smith, Schneider, & Minder, Reference Egger, Smith, Schneider and Minder1997). We explored sources of heterogeneity presence of potential outliers that could explain the heterogeneity [e.g. one individual study going in a different direction to all the others according to others (Davies et al., Reference Davies, Segre, Estrade, Radua, De Micheli, Provenzani and Fusar-Poli2020; Higgins et al., Reference Higgins, Thompson, Deeks and Altman2003)] and with sensitivity, subgroup analyses and meta-regressions. For sensitivity analyses, we excluded studies with sample sizes smaller than 50 participants, and studies with either moderate or high risk of bias. We also conducted subanalyses in those studies that despite defining their studies as drug-naïve, included FEP participants with minimal exposure. We also performed an analysis of influence and outliers according to the methods proposed by Viechtbauer and Cheung (Reference Viechtbauer and Cheung2010), and separate meta-analyses according to ATP-IIIA criteria and IDF. Lastly, we compared the results with a sex and age-matched control group. All analyses were performed using Comprehensive Meta-Analyses software (Borenstein, Hedges, Higgins, & Rothstein, Reference Borenstein, Hedges, Higgins and Rothstein2005).
Results
Search results
The search yielded 4143 articles. After the removal of duplicates, 2473 abstracts and titles were screened. At the full-text stage, 112 were screened, 18 articles were included in a systematic review, 13 (De Hert et al., Reference De Hert, Schreurs, Sweers, Van Eyck, Hanssens, Sinko and van Winkel2008; Effat et al., Reference Effat, Elsamei, Elghonemy and Roushdy2012; Enez Darcin, Yalcin Cavus, Dilbaz, Kaya, & Dogan, Reference Enez Darcin, Yalcin Cavus, Dilbaz, Kaya and Dogan2015; Garcia-Rizo et al., Reference Garcia-Rizo, Fernandez-Egea, Oliveira, Meseguer, Cabrera, Mezquida and Kirkpatrick2017; Grover, Nebhinani, Chakrabarti, Parakh, & Ghormode, Reference Grover, Nebhinani, Chakrabarti, Parakh and Ghormode2012; Kraemer et al., Reference Kraemer, Minarzyk, Forst, Kopf and Hundemer2011; Martín Otaño, Barbadillo Izquierdo, Galdeano Mondragón, Alonso Pinedo, & Querejeta Ayerdi, Reference Martín Otaño, Barbadillo Izquierdo, Galdeano Mondragón, Alonso Pinedo and Querejeta Ayerdi2013; Medved, Kuzman, Jovanovic, Grubisin, & Kuzman, Reference Medved, Kuzman, Jovanovic, Grubisin and Kuzman2009; Owiredu, Osei, Amidu, Appiah-Poku, & Osei, Reference Owiredu, Osei, Amidu, Appiah-Poku and Osei2012; Saddicha, Ameen, & Akhtar, Reference Saddicha, Ameen and Akhtar2007; Sahpolat & Ari, Reference Sahpolat and Ari2021; Saloojee, Burns, & Motala, Reference Saloojee, Burns and Motala2018; Srivastava, Bhatia, & Sharma, Reference Srivastava, Bhatia and Sharma2018) articles were included in the main statistical analyses (prevalence of MetS in drug naïve, 0 days of antipsychotic medication) (Fig. 1) and an additional five studies that included up to 47 days were considered for the supplementary sensitivity analysis (see below).
Fig. 1. PRISMA flow diagram.
Study and participant characteristics
We found 18 studies that reported patients with FEP and a drug-naïve condition. As expected, the definition of naïve was not defined exactly the same way in all the studies ranging between 0 and 47 days. The length of antipsychotic exposure was reported as 0 days in the majority of studies (Tables 1 and 2) (k = 13, n = 1009), up to 14 days in one study (k = 1, n = 76) (Srihari et al., Reference Srihari, Phutane, Ozkan, Chwastiak, Ratliff, Woods and Tek2013), and up to 47 days in four studies (k = 4, n = 711) (Chiliza et al., Reference Chiliza, Asmal, Oosthuizen, van Niekerk, Erasmus, Kidd and Emsley2015; Correll et al., Reference Correll, Robinson, Schooler, Brunette, Mueser, Rosenheck and Kane2014; Fleischhacker et al., Reference Fleischhacker, Siu, Bodén, Pappadopulos, Karayal, Kahn and EUFEST study group2013; Pallava, Chadda, Sood, & Lakshmy, Reference Pallava, Chadda, Sood and Lakshmy2012) (Tables 3 and 4). For the sake of accuracy, to calculate the prevalence in our meta-analysis, only the 13 studies with strictly naïve patients (0-day exposure) were included, but we decided to keep the five studies that included medication use up to 47 days in order to provide a comparison in sensitivity analysis.
Table 1. Characteristics of the studies included in the meta-analysis
Table 2. MetS prevalence of studies included in the meta-analysis
Table 3. Antipsychotic exposure and findings in not strictly naïve FEP
Table 4. Prevalence of metabolic syndrome in not strictly naïve patients with FEP
Across these 13 included studies, 1009 individuals with FEP and strictly naïve (0-day exposure) were included. Additionally, one study (n = 76) with minimally treated subjects (0–14 days) and four studies (n = 711) with subjects treated up to 47 days are available as post-hoc analyses in online Supplementary Figs. S2–S5. The age of participants ranged from 22 to 43 years and the percentage of female participants was 47.15% (n = 471). In most studies, diagnosis was confirmed after the FEP, schizophrenia being the most frequent (Table 1). All studies used validated criteria for the diagnosis of the MetS: ATP-IIIA (N = 9), IDF (N = 3), JIS (N = 1), both ATP-IIIA and IDF (N = 5). In the studies reporting data with ATP III and IDF, the former was chosen to calculate the overall prevalence. More details are provided in online Supplementary Table S8. Participants' ethnic origins were: Caucasian (N = 5), Indian (N = 3), Middle East (N = 3), Afro-descendants (N = 2) (online Supplementary Figs. S7a–e). Geographical location was Europe (N = 5), Africa (N = 3), Asia (N = 5) (Tables 1 and 2).
Pooled MetS prevalence
The total cases of MetS were 131 out of 1009 FEP subjects. The prevalence of MetS in strictly naïve patients with FEP is 13.2% (95% CI 8.7–19.0) (n = 1009, k = 13) (Fig. 2). Some studies did not fall within the pooled prevalence estimate (Effat et al., Reference Effat, Elsamei, Elghonemy and Roushdy2012; Enez Darcin et al., Reference Enez Darcin, Yalcin Cavus, Dilbaz, Kaya and Dogan2015; Owiredu et al., Reference Owiredu, Osei, Amidu, Appiah-Poku and Osei2012). Three studies reported a high prevalence of 40% (Effat et al., Reference Effat, Elsamei, Elghonemy and Roushdy2012), 32% (Enez Darcin et al., Reference Enez Darcin, Yalcin Cavus, Dilbaz, Kaya and Dogan2015) and 31.5% (Sahpolat & Ari, Reference Sahpolat and Ari2021). The study visually furthest from the pooled prevalence estimate (k = 1, n = 20) (Effat et al., Reference Effat, Elsamei, Elghonemy and Roushdy2012) considered ‘naïve’ as either never treated (0-day exposure) or drug-free for at least 6 months before the commencement of the study (Table 1). The graphical funnel plot and Eggers test (Fig. S1 online supplemental) showed there is no evidence of publication bias, so no trim and fill adjustment was needed (p = 0.4507). Only four studies reported the prevalence of MetS among women (Effat et al., Reference Effat, Elsamei, Elghonemy and Roushdy2012; Garcia-Rizo et al., Reference Garcia-Rizo, Fernandez-Egea, Oliveira, Meseguer, Cabrera, Mezquida and Kirkpatrick2017; Grover et al., Reference Grover, Nebhinani, Chakrabarti, Parakh and Ghormode2012; Medved et al., Reference Medved, Kuzman, Jovanovic, Grubisin and Kuzman2009). The total cases of MetS among women were 19 out of 173. Overall prevalence estimate was 9.6% (95% CI 3–14; I 2 57.02%, p = 0.06). The total cases of MetS among men were 14 out of 165. Overall prevalence estimate in men was 12.5% (95% CI 3–39).
Fig. 2. Forest plot showing MetS prevalence in strictly naïve patients (0 days).
Sensitivity and subgroup analyses
Ethnicity and geographical location
The subgroup analysis of the 13 papers based on geographical location showed that the prevalence of MetS in studies performed in Europe was 9.7% (95% CI 5–18), in Africa 8.3% (95% CI 10–44) and in Asia 20% (95% CI 12–30). Studies in Asia showed the highest prevalence. We found that studies in Africa have the highest variability between their prevalence. Only two studies were performed in the Afro-descendant population, from Ghana and South Africa (Owiredu et al., Reference Owiredu, Osei, Amidu, Appiah-Poku and Osei2012; Saloojee et al., Reference Saloojee, Burns and Motala2018). We conducted sensitivity analysis removing those studies and we found changes in the overall prevalence from 13.2% to 16%. Separate meta-analysis based on subjects' ethnic origin showed that the prevalence of MetS in studies with Caucasian patients was 9.7% (95% CI 4.7–18), Afro-descendants 3.2% (95% CI 1.4–7.5). We pooled studies conducted in the Middle East and in India and found a MetS prevalence of 32.8% (95% CI 24–42) and of 14.3% (95% CI 9.2–21), respectively (online Supplementary Figs. S7a–e).
Antipsychotic exposure and diagnostic MetS criteria
Although our meta-analysis includes drug-naïve (0-day exposure) patients only, we additionally performed post hoc sensitivity analyses through one-study-removed analysis on five studies without a strictly naïve definition, one that included minimal exposure (0–14 days) and four that included up to 47-day exposure (Figs. S2–S4 online supplementary material). The use of the 0–14 range is based on the recent evidence about the time considered as minimal exposure (Pillinger, Beck, Gobjila, et al., Reference Pillinger, Beck, Gobjila, Donocik, Jauhar and Howes2017; Pillinger, Beck, Stubbs, et al., Reference Pillinger, Beck, Stubbs and Howes2017) and the observed large changes in metabolic parameters in a median time of 6 weeks with some antipsychotics (Pillinger et al., Reference Pillinger, McCutcheon, Vano, Mizuno, Arumuham, Hindley and Howes2020).
Our post hoc analysis shows no significant changes in prevalence after removing studies. The prevalence of MetS was 12.2% (studies with 0 and 0–14 days of exposure, n = 1085, k = 14) and 12.2% (47 days of exposure, n = 711, k = 4), while the prevalence of MetS patients reported as naïve in the eighteen studies was 12.3% (95% CI 0.8–17.0) (n = 1796, k = 18). All in all, these sensitivity analyses show that the prevalence in strictly naïve (0 days of exposure) is 13.2% (95% CI 8.7–19.0) (online Supplementary Figs. S2–S4). From the excluded studies observed in Table 4, the minimal exposure study (Srihari et al., Reference Srihari, Phutane, Ozkan, Chwastiak, Ratliff, Woods and Tek2013) has the lowest prevalence of MetS. More details are provided in Fig. 2 and in online Supplementary Figs. S2–S4.
Sensitivity analyses based on diagnostic MetS criteria were also conducted in 13 studies. Although it seems all of them yield different prevalence estimates, there are no statistically significant differences between them. However, it is worth flagging that MetS prevalence is higher when diagnosed according to IDF v. ATP-IIIA criteria (online Supplementary Figs. S9–S11). We found that although the prevalence is more than double the prevalence with IDF, the confidence intervals of the prevalence in both subgroups ATP III 10% (95% CI 6–15) and IDF 21.8% (95% CI 12–34) match the confidence intervals of the global prevalence estimator 12.9% (95% CI 8–18). This result can be clearly observed by visual inspection of the forest plot figure (online Supplementary Figs. S9–S11). Additionally, in the four studies where both IDF and ATP-IIIA criteria were used to diagnose MetS, we performed individual meta-analyses for IDF and for ATP-IIIA showing that MetS prevalence in the same population is higher when diagnosed according to IDF than ATP-IIIA (online Supplementary Figs. S9–S11).
Other sensitivity and subgroup analysis
Sensitivity analyses based on sample size and one-study-removed analysis were also conducted (online Supplementary Fig. S2). One study (Effat et al., Reference Effat, Elsamei, Elghonemy and Roushdy2012) may be an outlier based on visual inspection. However, when excluding it from the analysis, the overall prevalence estimate just changed from 13.2% (95% CI 8.7–19.5) to 12% (95% CI 8–18). This change is not statistically significant. The influence analysis of Effat's study is visually striking, but not significant because it has a low weight (w = 1.34%).
Heterogeneity, quality assessment and meta-regressions
Heterogeneity was high for the primary analysis evaluating the pooled prevalence of MetS [I 2 = 81.03%, Q = 63, df(12), p = 0.00]. Also heterogeneity between subgroups was observed in the stratified analysis by criteria used for MetS [I 2 = 83.0%, Q = 7.57, df(2), p = 0.023]. We conducted meta-regressions using as covariates diagnostic criteria used for MetS, risk of bias, geographical location, ethnic origin of participants and patient settings. Geographical location is not a source of heterogeneity (R 2 0.00). Ethnicity accounted for 3% of the heterogeneity between studies, and diagnostic criteria used for MetS accounted for 7%. An additional meta-regression was performed using the MetS parameters of each study. The individual parameters for diagnosing MetS do not represent a source of heterogeneity for the prevalence estimates. The means of systolic blood pressure, diastolic blood pressure, serum glucose, HDL cholesterol and triglycerides are not significantly related to the estimated prevalence of MetS. The quality check agreement between the two raters was 81.8%. The risk of bias was graded as low (⩾7 points) for nine studies (De Hert et al., Reference De Hert, Schreurs, Sweers, Van Eyck, Hanssens, Sinko and van Winkel2008; Enez Darcin et al., Reference Enez Darcin, Yalcin Cavus, Dilbaz, Kaya and Dogan2015; Garcia-Rizo et al., Reference Garcia-Rizo, Fernandez-Egea, Oliveira, Meseguer, Cabrera, Mezquida and Kirkpatrick2017; Grover et al., Reference Grover, Nebhinani, Chakrabarti, Parakh and Ghormode2012; Kraemer et al., Reference Kraemer, Minarzyk, Forst, Kopf and Hundemer2011; Martín Otaño et al., Reference Martín Otaño, Barbadillo Izquierdo, Galdeano Mondragón, Alonso Pinedo and Querejeta Ayerdi2013; Medved et al., Reference Medved, Kuzman, Jovanovic, Grubisin and Kuzman2009; Saddicha et al., Reference Saddicha, Ameen and Akhtar2007; Sahpolat & Ari, Reference Sahpolat and Ari2021) and graded as medium (4–6 points) for four studies (Effat et al., Reference Effat, Elsamei, Elghonemy and Roushdy2012; Owiredu et al., Reference Owiredu, Osei, Amidu, Appiah-Poku and Osei2012; Saloojee et al., Reference Saloojee, Burns and Motala2018; Srivastava et al., Reference Srivastava, Bhatia and Sharma2018). Studies with a medium risk of bias (k = 4) reported a lower prevalence of 11% (95% CI 3.0–31.0) than studies with low risk of bias, which reported a prevalence of 13.9% (95% CI 8.0–21.0) but the difference between them is not statistically significant. Study quality scores of the 18 full-text selected studies may be found in online Supplementary Tables S6, S13 and Fig. S9.
Waist circumference
As for central obesity, nine of 13 studies reported data on waist circumference (De Hert et al., Reference De Hert, Schreurs, Sweers, Van Eyck, Hanssens, Sinko and van Winkel2008; Effat et al., Reference Effat, Elsamei, Elghonemy and Roushdy2012; Enez Darcin et al., Reference Enez Darcin, Yalcin Cavus, Dilbaz, Kaya and Dogan2015; Kraemer et al., Reference Kraemer, Minarzyk, Forst, Kopf and Hundemer2011; Medved et al., Reference Medved, Kuzman, Jovanovic, Grubisin and Kuzman2009; Owiredu et al., Reference Owiredu, Osei, Amidu, Appiah-Poku and Osei2012; Saddicha et al., Reference Saddicha, Ameen and Akhtar2007; Saloojee et al., Reference Saloojee, Burns and Motala2018). Patients in studies that reported a waist circumference larger than 90 cm had higher MetS prevalence than those with smaller than 90 cm (21% v. 7%; p < 0.001).
Control comparison
Of the 13 studies included, only six had control groups (348 cases and 347 controls) being all of them matched by sex and age (Effat et al., Reference Effat, Elsamei, Elghonemy and Roushdy2012; Enez Darcin et al., Reference Enez Darcin, Yalcin Cavus, Dilbaz, Kaya and Dogan2015; Garcia-Rizo et al., Reference Garcia-Rizo, Fernandez-Egea, Oliveira, Meseguer, Cabrera, Mezquida and Kirkpatrick2017; Saddicha et al., Reference Saddicha, Ameen and Akhtar2007; Sahpolat & Ari, Reference Sahpolat and Ari2021; Saloojee et al., Reference Saloojee, Burns and Motala2018). In this context, three of them used ATP-III criteria, two used IDF criteria and one used JIS-2009 criteria. Following the analysis of these studies, we found that the odds of having MetS in naïve FEP individuals was double than in controls (OR 2.52, p = 0.007) (Fig. 3). Of note is that we used studies with naïve (0 days of exposure) patients to control comparison.
Fig. 3. Forest plot showing comparison between naïve (0 days) first-episode psychosis patients v. healthy controls.
Discussion
This is the first meta-analysis of studies that strictly included patients with FEP with 0-day exposure to antipsychotic treatment. The prevalence of MetS in strictly naive patients with FEP is 13.2%. Our results are consistent with the most solid published meta-analysis on MetS in early stages of psychosis, including patients under medication and untreated patients at any stage of the disease, where a prevalence of 9.8% was found (Mitchell et al., Reference Mitchell, Vancampfort, De Herdt, Yu and De Hert2013). In contrast to this research, we specifically analysed patients with a first psychotic episode with no exposure to antipsychotics.
Naïve patients have double the amount of risk of MetS than general population
Our meta-analysis reports a higher risk of MetS in naïve patients with FEP compared to age-matched and sex-matched controls. We used studies with naïve patients (0 days of exposure) to control comparison, being all of them sex- and age-matched. The similar rates of MetS found in our study and in a previous meta-analysis (Vancampfort et al., Reference Vancampfort, Vansteelandt, Correll, Mitchell, De Herdt, Sienaert and De Hert2013) conducted in chronic populations (OR = 2.52 against OR = 2.35) is an intriguing finding that requires further exploration. It could also mean that antipsychotic use is not the only factor that can explain MetS; and that other factors, for which we have not accounted in this work, and that can account for it are already present early in the disease. In addition to antipsychotics, diet and a sedentary lifestyle, the tendency towards obesity in a group of patients with schizophrenia may also be influenced by genetic factors (Hasnain, Reference Hasnain2015) and by the impact of social adversity (Aas et al., Reference Aas, Dieset, Hope, Hoseth, Mørch, Reponen and Melle2017; Alameda et al., Reference Alameda, Levier, Gholam-Rezaee, Golay, Vandenberghe, Delacretaz and Conus2020). In this regard, one aspect to consider in future research could be the possible pathway linking social stress with obesity-related outcomes in people with psychosis, exploring the role of inflammation, stress hormones and the genetic and epigenetic underpinnings (Coleman, Krapohl, Eley, & Breen, Reference Coleman, Krapohl, Eley and Breen2018). Furthermore, current research suggests genetic vulnerability that specifically predisposes a subgroup of individuals to present metabolic alterations that are triggered by the use of antipsychotics (Crespo-Facorro, Prieto, & Sainz, Reference Crespo-Facorro, Prieto and Sainz2019; Tomasik et al., Reference Tomasik, Lago, Vazquez-Bourgon, Papiol, Suarez-Pinilla, Crespo-Facorro and Bahn2019).
Two studies included in our systematic review but not included in our OR calculation (Correll et al., Reference Correll, Robinson, Schooler, Brunette, Mueser, Rosenheck and Kane2014; Fleischhacker et al., Reference Fleischhacker, Siu, Bodén, Pappadopulos, Karayal, Kahn and EUFEST study group2013) did not use age- and sex-matched controls, but compared their MetS prevalence results in naïve patients with the general population based on findings from the Third National Health and Nutrition Examination Survey USA (Ford, Giles, & Dietz, Reference Ford, Giles and Dietz2002). The EUFEST study (Fleischhacker et al., Reference Fleischhacker, Siu, Bodén, Pappadopulos, Karayal, Kahn and EUFEST study group2013) found a 5.6% prevalence of MetS in naïve patients, which is similar to the 6% MetS prevalence reported for men and women in the USA aged 20–29 years old in an analysis of 8814 adults aged >20 years from the NHANES-III (1988–1994) survey (Ford et al., Reference Ford, Giles and Dietz2002). However, the MetS rate observed in the FEP patients in EUFEST (Fleischhacker et al., Reference Fleischhacker, Siu, Bodén, Pappadopulos, Karayal, Kahn and EUFEST study group2013) appears to be no higher than that of a general population of similar age. In the RAISE-ETP study (Correll et al., Reference Correll, Robinson, Schooler, Brunette, Mueser, Rosenheck and Kane2014), a slightly higher prevalence of MetS was found in naïve patients compared to the general population (Ford et al., Reference Ford, Giles and Dietz2002) of the same age (8.6% v. 6.0%).
A recent study (Moore, Chaudhary, & Akinyemiju, Reference Moore, Chaudhary and Akinyemiju2017) reported that rates of MetS in the general US population (all ethnicities combined, 1988–2012) in the age range of 18–29 was approximately 10%, increasing to approximately 20% in the 30–49 age bracket. No studies included in our meta-analysis used the recent published data (Moore et al., Reference Moore, Chaudhary and Akinyemiju2017) as control groups. One study (Grover et al., Reference Grover, Nebhinani, Chakrabarti, Parakh and Ghormode2012) found that the prevalence of MetS in naïve patients was lower than that of the general population (13% v. 39.5%). However, the population used as a control consisted mainly of women, with sedentary habits and with first-degree relatives who had a history of diabetes: all of these are cardiovascular risk factors. Therefore, the higher prevalence of MetS in naïve psychotics could be due to the fact that the latter were younger. For this reason, the Grover study was not used for the OR calculation.
Current criteria for MetS may not characterise risk in non-Caucasian populations
In our results, we identify that ethnic origin is a source of heterogeneity, which coincides with the majority of previous studies where ethnic differences have been described in the prevalence of MetS in patients with FEP (McEvoy et al., Reference McEvoy, Meyer, Goff, Nasrallah, Davis, Sullivan and Lieberman2005; Tek et al., Reference Tek, Kucukgoncu, Guloksuz, Woods, Srihari and Annamalai2016). In this context, the slight complexion of the Asian population discourages the use of the same circumference criteria as for the population of European descent (Lear, James, Ko, & Kumanyika, Reference Lear, James, Ko and Kumanyika2010). Asian populations have a lower prevalence of obesity (32.3% Asians v. 38.6% Westerners) (Arai et al., Reference Arai, Yamamoto, Matsuzawa, Saito, Yamada, Oikawa and Kita2006), lower HDL cholesterol (8.2% v. 37.1%), higher triglyceride (23.0% v. 30.0%) and abnormal glucose levels (11.3% v. 12.6%) compared to Western populations (Ford et al., Reference Ford, Giles and Dietz2002). The prevalence of MetS in the general population might also be lower than that of the Western population.
In four of the included studies, the systematic review (Chiliza et al., Reference Chiliza, Asmal, Oosthuizen, van Niekerk, Erasmus, Kidd and Emsley2015; Correll et al., Reference Correll, Robinson, Schooler, Brunette, Mueser, Rosenheck and Kane2014; Owiredu et al., Reference Owiredu, Osei, Amidu, Appiah-Poku and Osei2012; Saloojee et al., Reference Saloojee, Burns and Motala2018) mentioned ethnic differences as a possible element of confusion when determining results, and in two of them (Chiliza et al., Reference Chiliza, Asmal, Oosthuizen, van Niekerk, Erasmus, Kidd and Emsley2015; Saloojee et al., Reference Saloojee, Burns and Motala2018), it is suggested that this is the main source of variability in the prevalence of MetS in patients with FEP. Additionally, the low prevalence of MetS in naïve patients could be explained because they include a high proportion of Afro-descendant patients (97%) and a high prevalence of cannabis use (49.3%), both of which are factors that can modify the risk of MetS. In the same line, it has been described (Patel et al., Reference Patel, Buckley, Woolson, Hamer, McEvoy, Perkins and Lieberman2009) how for other ethnic groups the prevalence of MetS at 52 weeks of treatment is almost double that of Afro-descendants. On the contrary, several epidemiological studies reported that Afro-descendants have a higher risk of metabolic disorders such as insulin resistance and high blood pressure (Chaturvedi, Reference Chaturvedi2003). The explanation for this contradiction could be the underestimation of the risk of MetS in Afro-descendants within the current definitions of MetS according to the IDF and ATP-IIIA criteria, since these were initially created for Caucasian populations and there are factors not duly taken into account, such as body fat distribution and risk of insulin resistance (De Lucia Rolfe, Ong, Sleigh, Dunger, & Norris, Reference De Lucia Rolfe, Ong, Sleigh, Dunger and Norris2015). Hence, based on ATP-IIIA and IDF, various scientific societies in Asian and Latin American countries have adapted their own MetS criteria.
Other potential predictors of cardiovascular risks in FEP
MetS is a predictor of cardiovascular risk. Within 5–10 years, risk is best calculated with classic scales (Framingham or SCORE), which include age, gender, total cholesterol, LDL and tobacco use (Grundy, Reference Grundy2006). Our study found that the prevalence of tobacco use was 40%. Bearing in mind that a large percentage of patients with schizophrenia are smokers, it would be useful to include the influence of tobacco on future predictors of cardiovascular risk.
The alteration of individual metabolic parameters in naive FEP, such as glycaemic or lipid alterations, is widely described in the existing literature. In a recent meta-analysis, Pillinger et al. (Reference Pillinger, McCutcheon, Vano, Mizuno, Arumuham, Hindley and Howes2020) found increased insulin resistance in drug-naïve FEP compared with controls. For this reason, the use of other markers like insulin resistance as predictors of cardiovascular risk has been proposed (Garcia-Rizo et al., Reference Garcia-Rizo, Fernandez-Egea, Oliveira, Meseguer, Cabrera, Mezquida and Kirkpatrick2017). Several cardiovascular risk prediction algorithms have been developed, but only three are validated on psychiatric patients (QRISK3, QDiabetes and PRIMROSE) and these are validated with samples from only elderlies (Perry et al., Reference Perry, Upthegrove, Crawford, Jang, Lau, McGill and Khandaker2020). Most of the analysed studies show that among the individual parameters, waist circumference relates the most to changes in MetS prevalence, finding MetS prevalence higher in those with the highest abdominal perimeter. Additionally, we found that the prevalence of altered waist circumference is 14% in naïve patients with FEP.
Limitations of the current work itself should be noted: heterogeneity across studies which may be due to disparity in MetS criteria. Although we tried our best to account for potential heterogeneity resulting from the different MetS criteria (conducting sensitivity analysis according to studies that used IDF or ATP-IIIA and conducting separated meta-analysis with studies that reported prevalence with ATPIII-A and with IDF) we were still unable to account for variations in all criteria (e.g. JIS-2009 and WHO criteria). We were not able to exclude patients/controls that were prescribed other psychiatric/physical health medications other than antipsychotics known to impact metabolic function and we could not account for the level of depressive symptom or comorbid depression in our meta-regressions, a factor known for being associated with obesity-related outcomes (Lasserre et al., Reference Lasserre, Glaus, Vandeleur, Marques-Vidal, Vaucher, Bastardot and Preisig2014), thus we cannot exclude that depression is influencing our prevalence estimates. It is also known that people with psychotic disorders are less likely to present to physical health services compared with the general population. As such, there is a risk of under-reporting and thus under-estimating the prevalence of MetS in this cohort.
In terms of limitations related to the included studies in meta-analysis, two studies (Owiredu et al., Reference Owiredu, Osei, Amidu, Appiah-Poku and Osei2012; Saloojee et al., Reference Saloojee, Burns and Motala2018) were the only ones conducted on Afro-descendant ethnicity and the overall prevalence increased when those were removed in sensitivity analyses. However, one of them (Saloojee et al., Reference Saloojee, Burns and Motala2018) was the only study that reported cannabis consumption, which has been associated with low odds of MetS in both general population (Vidot et al., Reference Vidot, Prado, Hlaing, Florez, Arheart and Messiah2016) and patients with FEP (Stiles, Alcover, Stiles, Oluwoye, & McDonell, Reference Stiles, Alcover, Stiles, Oluwoye and McDonell2020), low odds of overweightness (Vazquez-Bourgon et al., Reference Vazquez-Bourgon, Setien-Suero, Pilar-Cuellar, Romero-Jimenez, Ortiz-Garcia de la Foz, Castro and Crespo-Facorro2019) and low odds of non-alcoholic fatty liver (Vazquez-Bourgon et al., Reference Vazquez-Bourgon, Ortiz-Garcia de la Foz, Suarez-Pereira, Iruzubieta, Arias-Loste, Setien-Suero and Crespo Facorro2019) in patients with FEP. We were not able to see the influence of cannabis on prevalence accurately. Unfortunately, we do not have enough data to accurately see the influence of age on prevalence as most studies reported the mean and not age range, and the few reported age ranges are not mutually excluding. Besides, the control group studies remain relatively low.
To conclude, our findings of increased rates of MetS in patients with antipsychotic-naïve FEP suggest that we are underestimating cardiovascular risk in this cohort, especially in those of non-Caucasian origin. The role of cannabis in the modulation of MetS requires additional research. Early predictors of cardiovascular risk for schizophrenia should be determined considering different patient phenotypes according to precision medicine. Future research should focus on the predictors of cardiovascular risk including common molecular and environmental factors, as our findings support that altered metabolic parameters in FEPs are not exclusively due to antipsychotic treatments.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291721002853.
Financial support
This research received no specific grant from any funding agency, commercial or not-for-profit sectors.
Conflict of interest
No potential competing interest is reported by the authors.
