Introduction
Weight gain has long been recognized as a side-effect of antipsychotic (AP) drugs (Allison et al. Reference Allison, Mentore, Heo, Chandler, Cappelleri, Infante and Weiden1999; Lieberman et al. Reference Lieberman, Stroup, McEvoy, Swartz, Rosenheck, Perkins, Keefe, Davis, Davis, Lebowitz, Severe and Hsiao2005). This AP side-effect has recently become a major concern in the treatment of psychosis because weight gain is a potential contributor to increased co-morbidity (Wirshing, Reference Wirshing2001), including glucose intolerance, diabetes mellitus (Hedenmalm et al. Reference Hedenmalm, Hagg, Stahl, Mortimer and Spigset2002; Scheen & De Hert, Reference Scheen and De Hert2007), metabolic syndrome (McEvoy et al. Reference McEvoy, Meyer, Goff, Nasrallah, Davis, Sullivan, Meltzer, Hsiao, Stroup and Lieberman2005), sleep apnoea (Wirshing et al. Reference Wirshing, Pierre and Wirshing2002) and cardiovascular diseases (Robinson et al. Reference Robinson, Woerner, Delman and Kane2005). All of these conditions can lead to increased mortality (Kurzthaler & Fleishhacker, Reference Kurzthaler and Fleischhacker2001; Wirshing, Reference Wirshing2001). Furthermore, fear of weight gain is one of the main factors contributing to poor compliance found in AP treatment (Perkins, Reference Perkins1999, Reference Perkins2002), so it may adversely affect the clinical outcome (Allison et al. Reference Allison, Mentore, Heo, Chandler, Cappelleri, Infante and Weiden1999; Allison & Casey, Reference Allison and Casey2001).
The mechanisms by which AP drugs induce weight gain have yet to be fully elucidated. AP antagonism to serotonin or histamine receptors (5-HT2c, H1), which modulate appetite and body weight, has been implicated. Estimates of mean weight gain associated with AP agents have varied greatly and possible confounders could be the extent of previous AP treatment or a history of schizophrenia and other risk factors for weight gain in psychiatric patients (Ganguli et al. Reference Ganguli, Brar and Ayrton2001). There is much interest in determining possible additional risk factors for weight gain associated with AP therapy. Schizophrenia and other psychotic disorders are known to be associated with a higher risk of obesity and metabolic disorders; in particular, the body mass index (BMI) of patients with schizophrenia exceeds the general population estimates (Allison et al. Reference Allison, Mentore, Heo, Chandler, Cappelleri, Infante and Weiden1999; Caballero, Reference Caballero2003; De Hert et al. Reference De Hert, van Winkel, Van Eyck, Hanssens, Wampers, Scheen and Peuskens2006). Patients with schizophrenia or other psychotic disorders have well-recognized lifestyle risk factors for overweight and obesity, such as lack of exercise, sedentary habits and a poor diet (Brown et al. Reference Brown, Birtwistle, Roe and Thompson1999; Kelly et al. Reference Kelly, Conley and Carpenter2005). A study by Ryan et al. (Reference Ryan, Flanagan, Kinsella, Keeling and Thakore2004) observed that, compared with normal controls, higher levels of visceral fat stores were present even in first-episode schizophrenia drug-naive patients.
It emerges from such studies that there is considerable variability among patients in terms of tolerance towards AP side-effects such as weight gain. The amount of weight gain differs considerably from patient to patient and the clinically prominent individual differences in drug effects on body weight result from a combination of genetic and environmental influences (Basile et al. Reference Basile, Masellis, McIntyre, Meltzer, Lieberman and Kennedy2001; American Diabetes Association, 2004). Moreover, the time course of AP weight gain varies greatly between studies considering chronic patients and research on first-episode psychosis (FEP) patients; among the first group weight gain reaches a plateau after the first year of treatment whereas among FEP it seems to go on increasing after the first year of treatment (Alvarez-Jiménez et al. Reference Alvarez-Jiménez, González-Blanch, Crespo-Facorro, Hetrick, Rodríguez-Sánchez, Pérez-Iglesias and Vázquez-Barquero2008; Perez-Iglesias et al. Reference Perez-Iglesias, Crespo-Facorro, Martinez-Garcia, Ramirez-Bonilla, Alvarez-Jimenez, Pelayo-Teran, Garcia-Unzueta, Amado and Vazquez-Barquero2008a). Thus, only studies involving drug-naive patients could provide evidence of a risk of weight gain among patients starting AP therapy; and only drug trials conducted on patients starting AP therapy for the first time are able to detect any causal relationship between AP exposure and weight gain and the time course of weight gain.
We have conducted a review of the literature regarding the amount and time course of weight gain induced by APs in drug-naive patients and carried out a meta-analysis of mean weight gains. To our knowledge, this is the first such meta-analysis on this topic.
Method
Studies were identified by systematic searches from 1997 to July 2008 in Medline, PsycINFO, EMBASE and The Cochrane Library database to identify primary studies describing weight gain in AP-naive patients. Search terms included: antipsychotic agents, drug naive, first episode psychosis, weight gain, and BMI. The titles, abstracts and full text of identified papers were screened for eligibility by two reviewers (B.F.G. and D.G.). The search was supplemented by additional relevant papers identified by a manual search through reference lists from articles retrieved and review articles.
The following criteria were applied for selection: participants in the study were AP-naive at baseline; participants in the study had been treated with APs since baseline; APs were not combined with other medication to reduce the potential weight gain due to AP therapy; participants were aged >15 years; the study presented weight and/or BMI changes during AP treatment as outcome; and the results were published in English as a full report.
Statistical analysis
We analysed the mean weight gain by the standard deviation (s.d.) for all the studies considered. For studies that analysed weight gain in patients treated with various APs, we calculated a mean weight change of the whole sample. For those that did not report any s.d. value, for every analysis we calculated it from the s.d. weighted mean in studies that did. As a control group, for each study we created a sample out of an equal number of subjects. For the control group we considered a mean weight gain of 0.00 kg and the s.d. as the weighted mean of s.d. in the studies considered in each analysis. We performed a first analysis of all studies selected. Then we repeated the analysis considering only studies with a follow-up at ⩽12 weeks, studies without co-therapy, out-patients and patients with physical co-morbidity within the sample and independent studies to reduce heterogeneity among studies. Finally, we analysed studies grouped by differing follow-up duration, to analyse the time course of weight gain in AP-naive patients.
Data were entered into the Cochrane Collaboration review manager software (RevMan version 4.2) and analysed by RevMan analysis 1.01 (RevMan Reference RevMan2003). Heterogeneity between studies was assessed by the χ2 test. Individual and pooled weighted mean differences and associated 95% confidence intervals (CIs) were calculated. A fixed-effect model was used in all analyses. We used this model despite the moderate heterogeneity across studies, given that we had no a priori reason to hypothesize data coming from different populations and because the main aim of the present analysis was to ascertain the best estimate of a single effect size rather than the range of effect sizes across populations (Munafò & Flint, Reference Munafò and Flint2004). We assessed the publication bias visually with funnel plots and statistically with the methods of Egger et al. (Reference Egger, Smith, Schneider and Minder1997).
Results
The literature search and selection produced 844 studies, 243 of which were retrieved after screening the title and abstracts (Fig. 1). Eighteen of these papers met our criteria. The main reasons for exclusion at this stage were: articles did not report data on body weight gain or did not involve drug-naive patients (199); studies were carried out on a sample with only a percentage of drug-naive patients or they did not report participants' previous drug history (17); studies where olanzapine was combined with the drug to reduce the potential weight gain (fluoxetine, reboxetine, famotidine, betahistine or metformin) (6); case reports (3). One additional paper was identified by checking references to selected papers. In all, 19 papers were included, reporting on 11 studies.
Fig. 1. Selection of studies.
One paper (Wu et al. Reference Wu, Zhao, Zhai, Guo and Guo2007) was excluded because it used the same sample as the previous Wu et al. study (Reference Wu, Zhao, Liu, Zhai, Guo, Guo and Tang2006). Zhang et al. (Reference Zhang, Yao, Zhang, Chen, Sun, Yao, Hou and Zhang2003) reported data on nearly the same sample as the previous one by Reynolds et al. (Reference Reynolds, Zhang and Zhang2002), but the latter paper provided data only on BMI whereas the former also reported data in kg; thus we considered the Reynolds et al. paper (Reference Reynolds, Zhang and Zhang2002) for BMI data and the Zhang et al. paper (Reference Zhang, Yao, Zhang, Chen, Sun, Yao, Hou and Zhang2003) for kg data. Three studies (Saddichha et al. Reference Saddichha, Manjunatha, Ameen and Akhtar2007a, Reference Saddichha, Shaul and Sayeedb, Reference Saddichha, Manjunatha, Ameen and Akhtar2008b) were excluded because of preliminary and/or incomplete data in the study by Saddichha et al. (Reference Saddichha, Ameen and Akhtar2008a), two studies (Perez-Iglesias et al. Reference Perez-Iglesias, Crespo-Facorro, Amado, Garcia-Unzueta, Ramirez-Bonilla, Gonzalez-Blanch, Martinez-Garcia and Vazquez-Barquero2007, Reference Perez-Iglesias, Vazquez-Barquero, Amado, Berja, Garcia-Unzueta, Pelayo-Terán, Carrasco-Marín, Mata and Crespo-Facorro2008b) because of preliminary and/or incomplete data in the study by Perez-Iglesias et al. (Reference Perez-Iglesias, Crespo-Facorro, Martinez-Garcia, Ramirez-Bonilla, Alvarez-Jimenez, Pelayo-Teran, Garcia-Unzueta, Amado and Vazquez-Barquero2008a), and one (Reynolds et al. Reference Reynolds, Zhang and Zhang2003) because it referred to a subgroup of patients from the study by Reynolds (Reference Reynolds, Zhang and Zhang2002).
Of the 12 papers selected (Table 1), five reported AP-naive patients' weight gain (in kg) and BMI (Zhang et al. Reference Zhang, Yao, Liu, Fang and Reynolds2004; Arranz et al. Reference Arranz, San, Duenas, Centeno, Ramirez, Salavert and Del Moral2007; Perez-Iglesias et al. Reference Perez-Iglesias, Crespo-Facorro, Martinez-Garcia, Ramirez-Bonilla, Alvarez-Jimenez, Pelayo-Teran, Garcia-Unzueta, Amado and Vazquez-Barquero2008a; Saddichha et al. Reference Saddichha, Ameen and Akhtar2008a; Tarricone et al. Reference Tarricone, Serretti, Ferrari Gozzi, Mandelli, Grieco, Mellini, Bigini, Berti and Berardi2008), four studies reported only BMI values (Reynolds et al. Reference Reynolds, Zhang and Zhang2002; Ryan et al. Reference Ryan, Flanagan, Kinsella, Keeling and Thakore2004; Templeman et al. Reference Templeman, Reynolds, Arranz and San2005; Wu et al. Reference Wu, Zhao, Liu, Zhai, Guo, Guo and Tang2006) and three only the weight gain in kg (Yap et al. Reference Yap, Mahendran, Lim, Liow, Lee, Phang and Tiong2001; Luty et al. Reference Luty, Kelly and McCreadie2002; Zhang et al. Reference Zhang, Yao, Zhang, Chen, Sun, Yao, Hou and Zhang2003).
Table 1. Description of studies on antipsychotic-naive patients
BMI, Body mass index; NOS, not otherwise specified; AP, antipsychotic; FGA, first-generation AP; SGA, second-generation AP; SS, statistically significant difference; AD, antidepressant; BDZ, benzodiazepine; EPS, extrapyramidal symptoms.
a BMI data from the Luty et al. study (Reference Luty, Kelly and McCreadie2002) were not considered because they are given for the whole sample, including patients who discontinued AP treatment during the follow-up.
b These two studies had almost the same sample; we used for the BMI data of Reynolds et al. (Reference Reynolds, Zhang and Zhang2002) and the kg data of Zhang et al. (Reference Zhang, Yao, Zhang, Chen, Sun, Yao, Hou and Zhang2003).
As shown in Table 1, six studies reported BMI data at 4–8 weeks' follow-up (Reynolds et al. Reference Reynolds, Zhang and Zhang2002; Templeman et al. Reference Templeman, Reynolds, Arranz and San2005; Wu et al. Reference Wu, Zhao, Liu, Zhai, Guo, Guo and Tang2006; Arranz et al. Reference Arranz, San, Duenas, Centeno, Ramirez, Salavert and Del Moral2007; Saddichha et al. Reference Saddichha, Ameen and Akhtar2008a; Tarricone et al. Reference Tarricone, Serretti, Ferrari Gozzi, Mandelli, Grieco, Mellini, Bigini, Berti and Berardi2008) and four studies reported kg data at 4–8 weeks' follow-up (Yap et al. Reference Yap, Mahendran, Lim, Liow, Lee, Phang and Tiong2001; Arranz et al. Reference Arranz, San, Duenas, Centeno, Ramirez, Salavert and Del Moral2007; Saddichha et al. Reference Saddichha, Ameen and Akhtar2008a; Tarricone et al. Reference Tarricone, Serretti, Ferrari Gozzi, Mandelli, Grieco, Mellini, Bigini, Berti and Berardi2008). Four studies reported BMI data at 10–12 weeks' follow-up (Reynolds et al. Reference Reynolds, Zhang and Zhang2002; Zhang et al. Reference Zhang, Yao, Liu, Fang and Reynolds2004; Templeman et al. Reference Templeman, Reynolds, Arranz and San2005; Perez-Iglesias et al. Reference Perez-Iglesias, Crespo-Facorro, Martinez-Garcia, Ramirez-Bonilla, Alvarez-Jimenez, Pelayo-Teran, Garcia-Unzueta, Amado and Vazquez-Barquero2008a) and three studies reported kg data at 10–12 weeks' follow-up (Zhang et al. Reference Zhang, Yao, Zhang, Chen, Sun, Yao, Hou and Zhang2003, Reference Zhang, Yao, Liu, Fang and Reynolds2004; Perez-Iglesias et al. Reference Perez-Iglesias, Crespo-Facorro, Martinez-Garcia, Ramirez-Bonilla, Alvarez-Jimenez, Pelayo-Teran, Garcia-Unzueta, Amado and Vazquez-Barquero2008a). Finally, three studies reported BMI data at 24–48 weeks' follow-up (Ryan et al. Reference Ryan, Flanagan, Kinsella, Keeling and Thakore2004; Templeman et al. Reference Templeman, Reynolds, Arranz and San2005; Perez-Iglesias et al. Reference Perez-Iglesias, Crespo-Facorro, Martinez-Garcia, Ramirez-Bonilla, Alvarez-Jimenez, Pelayo-Teran, Garcia-Unzueta, Amado and Vazquez-Barquero2008a), one study reported kg data at 48 weeks' follow-up (Perez-Iglesias et al. Reference Perez-Iglesias, Crespo-Facorro, Martinez-Garcia, Ramirez-Bonilla, Alvarez-Jimenez, Pelayo-Teran, Garcia-Unzueta, Amado and Vazquez-Barquero2008a) and one kg data at 2.5 years' follow-up (Luty et al. Reference Luty, Kelly and McCreadie2002). Among the studies found, nine reported mean AP doses that are within the optimal therapeutic range as presented in Table 1.
Meta-analysis of BMI main gains
As the first step we calculated the mean BMI gain from analysis of the nine studies selected, in each study considering data from the last observation available. Only two studies lasted more than 12 weeks (Templeman et al. Reference Templeman, Reynolds, Arranz and San2005; Perez-Iglesias et al. Reference Perez-Iglesias, Crespo-Facorro, Martinez-Garcia, Ramirez-Bonilla, Alvarez-Jimenez, Pelayo-Teran, Garcia-Unzueta, Amado and Vazquez-Barquero2008a). We found a mean BMI increase of 1.97 (95% CI 1.81–2.12, p<0.00001); homogeneity between studies proved to be low (test for heterogeneity: χ2=241.26, df=8, p<0.00001) (Fig. 2). In the following steps we excluded studies with possible factors affecting heterogeneity. First, we excluded studies that only provided follow-up >12 weeks (Ryan et al. Reference Ryan, Flanagan, Kinsella, Keeling and Thakore2004). We considered data from the studies by Templeman et al. (Reference Templeman, Reynolds, Arranz and San2005) and Perez-Iglesias et al. (Reference Perez-Iglesias, Crespo-Facorro, Martinez-Garcia, Ramirez-Bonilla, Alvarez-Jimenez, Pelayo-Teran, Garcia-Unzueta, Amado and Vazquez-Barquero2008a) at 12 weeks. We found eight studies where analysis resulted in a mean BMI gain of 1.51 (95% CI 1.36–1.67, p<0.00001; test for heterogeneity: χ2=42.75, df=7, p<0.00001). Then we excluded studies with co-therapies having a potential side-effect on body weight (see Table 1) (Perez-Iglesias et al. Reference Perez-Iglesias, Crespo-Facorro, Martinez-Garcia, Ramirez-Bonilla, Alvarez-Jimenez, Pelayo-Teran, Garcia-Unzueta, Amado and Vazquez-Barquero2008a; Tarricone et al. Reference Tarricone, Serretti, Ferrari Gozzi, Mandelli, Grieco, Mellini, Bigini, Berti and Berardi2008); and we found six studies where analysis resulted in a mean BMI gain of 1.43 (95% CI 1.32–1.53, p<0.00001; test for heterogeneity: χ2=108.65, df=5, p<0.00001). At this point, we excluded studies carried out on non-hospitalized patients (Templeman et al. Reference Templeman, Reynolds, Arranz and San2005) to reduce the possible effect of unbalanced meals on body weight, and we found five studies on hospitalized patients, resulting in a mean BMI gain of 1.18 (95% CI 1.05–1.30, p<0.00001; test for heterogeneity: χ2=26.83, df=4, p<0.0001). Among the five studies on hospitalized patients we then excluded the study by Arranz et al. (Reference Arranz, San, Duenas, Centeno, Ramirez, Salavert and Del Moral2007), which did not specify whether patients with physical co-morbidity were excluded, and we found a mean BMI gain of 1.14 (95% CI 1.01–1.27, p<0.00001; test for heterogeneity: χ2=23.13, df=3, p<0.0001). Finally, we excluded three sponsored studies (Reynolds et al. Reference Reynolds, Zhang and Zhang2002; Wu et al. Reference Wu, Zhao, Liu, Zhai, Guo, Guo and Tang2006; Saddichha et al. Reference Saddichha, Ameen and Akhtar2008a), resulting in a mean BMI gain of 1.08 (95% CI 0.95–1.22, p<0.00001; test for heterogeneity: χ2=17.30, df=2, p=0.0002).
Fig. 2. Mean body mass index gain in antipsychotic-naive patients.
We also analysed the mean BMI time course considering study data at different times of follow-up. Three studies that reported data at different times of follow-up were considered for all such follow-up evaluations (Reynolds et al. Reference Reynolds, Zhang and Zhang2002; Templeman et al. Reference Templeman, Reynolds, Arranz and San2005; Perez-Iglesias et al. Reference Perez-Iglesias, Crespo-Facorro, Martinez-Garcia, Ramirez-Bonilla, Alvarez-Jimenez, Pelayo-Teran, Garcia-Unzueta, Amado and Vazquez-Barquero2008a). In view of the low number of studies in the sample, we decided to analyse the mean weighted BMI gains from studies at three different times of follow-up (Fig. 3) and found a continuous rise in the weighted mean BMI; for studies lasting 4–8 weeks (Reynolds et al. Reference Reynolds, Zhang and Zhang2002; Templeman et al. Reference Templeman, Reynolds, Arranz and San2005; Wu et al. Reference Wu, Zhao, Liu, Zhai, Guo, Guo and Tang2006; Arranz et al. Reference Arranz, San, Duenas, Centeno, Ramirez, Salavert and Del Moral2007; Saddichha et al. Reference Saddichha, Ameen and Akhtar2008a; Tarricone et al. Reference Tarricone, Serretti, Ferrari Gozzi, Mandelli, Grieco, Mellini, Bigini, Berti and Berardi2008) we found a mean BMI gain of 1.15 (95% CI 1.06–1.24, p<0.00001; test for heterogeneity: χ2=86.14, df=5, p<0.00001); for studies lasting 10–12 weeks (Reynolds et al. Reference Reynolds, Zhang and Zhang2002; Zhang et al. Reference Zhang, Yao, Liu, Fang and Reynolds2004; Templeman et al. Reference Templeman, Reynolds, Arranz and San2005; Perez-Iglesias et al. Reference Perez-Iglesias, Crespo-Facorro, Martinez-Garcia, Ramirez-Bonilla, Alvarez-Jimenez, Pelayo-Teran, Garcia-Unzueta, Amado and Vazquez-Barquero2008a) we found a mean BMI gain of 1.80 (95% CI 1.62–1.97, p<0.00001; test for heterogeneity: χ2=26.01, df=3, p<0.00001); and for studies lasting 24–48 weeks (Ryan et al. Reference Ryan, Flanagan, Kinsella, Keeling and Thakore2004; Templeman et al. Reference Templeman, Reynolds, Arranz and San2005; Perez-Iglesias et al. Reference Perez-Iglesias, Crespo-Facorro, Martinez-Garcia, Ramirez-Bonilla, Alvarez-Jimenez, Pelayo-Teran, Garcia-Unzueta, Amado and Vazquez-Barquero2008a) we found a mean BMI gain of 3.87 (95% CI 3.48–4.26, p<0.00001), and here homogeneity between studies proved to be high (test for heterogeneity: χ2=2.88, df=2, p=0.24).
Fig. 3. Body mass index (BMI) mean change at three different follow-up times.
Meta-analysis of weight gain in kg
Overall, from analysis of the eight studies selected, considering data at the last observation available for every study, the mean weight gain was 4.85 kg (95% CI 4.23–5.47, p<0.00001); homogeneity among the studies proved to be low (test for heterogeneity: χ2=76.91, df=7, p<0.00001). We analysed studies reporting weight gain in kg following the same steps reported above for BMI studies and found that the heterogeneity decreased, consistently with the BMI analyses, and that weight gain in kg was directly correlated with the duration of AP treatment (data available on request).
Publication bias
The funnel plot (not shown, available on request) carried out on all BMI studies included in Fig. 1 did not suggest any evidence of publication bias. This finding is confirmed by Egger's regression test analysis (β=0.31, p=0.411). The funnel plot on the kg studies included in Fig. 3 was asymmetric but the Egger analysis did not find any publication bias (β=−0.41, p=0.32).
Discussion
Our findings clearly show a rapid, highly significant and continuously growing weight gain in patients treated for the first time with AP drugs. To our knowledge, this is the first systematic review and meta-analysis of weight gain in AP-naive patients.
We found that weight gain due to AP therapy in AP-naive patients is highly significant from the first months of treatment. The exclusion of studies with adjunctive risk factors, such as co-therapies with a body-weight effect or out-patient status possibly leading to an unbalanced diet, consolidates the causal role of AP drugs on our weight gain findings at around 3.8 kg and 1.2 points BMI within the first 12 weeks of treatment.
Moreover, in analysing the time course of kg and BMI mean increases we found that AP-naive patients' body weight gain increases continuously during AP treatment. We found that heterogeneity among studies with a longer follow-up was low, although the mean kg and BMI gains were still highly significant.
Our results show that even in AP-naive patients weight gain is prominent from the first weeks of treatment, as Allison et al. (Reference Allison, Mentore, Heo, Chandler, Cappelleri, Infante and Weiden1999) found in long-term AP-treated patients, where weight gain occurred after a mean of 10 weeks of treatment. Moreover, our results add evidence for the causal role of AP in weight gain, as the previous meta-analysis on this topic (Allison et al. Reference Allison, Mentore, Heo, Chandler, Cappelleri, Infante and Weiden1999) was carried out on long-term AP-treated patients, without considering the role of possible adjunctive risk factors and not reporting the BMI change but only the weight gain in kg. Some studies have noted that FEP patients treated with APs have a higher risk of gaining weight than patients suffering from chronic psychosis (Wetterling & Mussigbrodt, Reference Wetterling and Mussigbrodt1999; Kelly et al. Reference Kelly, Conley and Carpenter2005; Alvarez-Jiménez et al. Reference Alvarez-Jiménez, González-Blanch, Crespo-Facorro, Hetrick, Rodríguez-Sánchez, Pérez-Iglesias and Vázquez-Barquero2008). As a result, it has been posited that patients previously unexposed to AP medication are particularly vulnerable to this AP side-effect (Wetterling & Mussigbrodt, Reference Wetterling and Mussigbrodt1999; Basson et al. Reference Basson, Kinon, Taylor, Szymanski, Gilmore and Tollefson2001; Kinon et al. Reference Kinon, Basson, Gilmore and Tollefson2001; Meyer, Reference Meyer2002). The results of our meta-analysis, rigorously including all AP-naive samples, now provide conclusive evidence that weight gain associated with AP therapy in AP-naive patients is prominent and occurs rapidly within the first few weeks. Thus, clinicians should be aware of the high probability of causing weight gain in AP-naive patients and should strictly monitor such patients and also give them dietary counselling at the earliest opportunity.
Our findings also show that weight gain is a continuous process in drug-naive patients. This result is consistent with findings from the few studies that carried out more than one observation at the follow-up (Reynolds et al. Reference Reynolds, Zhang and Zhang2002; Templeman et al. Reference Templeman, Reynolds, Arranz and San2005; Perez-Iglesias et al. Reference Perez-Iglesias, Crespo-Facorro, Martinez-Garcia, Ramirez-Bonilla, Alvarez-Jimenez, Pelayo-Teran, Garcia-Unzueta, Amado and Vazquez-Barquero2008a). Our meta-analysis does not prove that weight gain reaches a plateau, as suggested for chronic long-treated patients (Henderson et al. Reference Henderson, Cagliero, Gray, Nasrallah, Hayden, Schoenfeld and Goff2000). The Perez-Iglesias et al. (Reference Perez-Iglesias, Crespo-Facorro, Martinez-Garcia, Ramirez-Bonilla, Alvarez-Jimenez, Pelayo-Teran, Garcia-Unzueta, Amado and Vazquez-Barquero2008a) randomized control trial on AP-naive patients extending over 1 year of continuous AP treatment showed that weight gain was faster during the first 12 weeks of treatment and then plateaued. However, there are too few studies on AP-naive patients lasting more than 12 weeks to confirm the possibility that weight gain reaches a plateau when a patient is receiving long-term treatment, and the timing of any such plateau is still controversial and requires further investigation (Wirshing et al. Reference Wirshing, Wirshing, Kysar, Berisford, Goldstein, Pashdag, Mintz and Marder1999; Taylor & McAskill, Reference Taylor and McAskill2000; McGavin & Goa, Reference McGavin and Goa2002; McIntyre et al. Reference McIntyre, Trakas, Lin, Balshaw, Hwang, Robinson and Eggleston2003; Lee et al. Reference Lee, Leung and Wong2004; Wirshing, Reference Wirshing2004; Gentile, Reference Gentile2006; Brecher et al. Reference Brecher, Leong, Stening, Osterling-Koskinen and Jones2007; Henderson, Reference Henderson2007).
The seminal Allison et al. (Reference Allison, Mentore, Heo, Chandler, Cappelleri, Infante and Weiden1999) review reported that, in long-term AP-treated patients, weight gain was highly variable among different APs; in their review the drugs most associated with this side-effect were clozapine and olanzapine (4–4.5 kg over 10 weeks). Unfortunately, the studies that have examined weight gain in AP-naive patients receiving AP drugs are still too few to distinguish clearly differences in weight gain-inducing potential among the different APs. Moreover, risperidone and olanzapine are the AP drugs most analysed in these studies, whereas the results on other APs are insufficient to draw any definitive conclusions. Although we were unable to perform a direct comparison among the different APs, the studies reviewed indicated that weight gain occurs with all medication.
Limitations
This is the first meta-analysis of weight gain in AP-naive patients. Although AP treatment should be initiated as soon as psychotic symptoms are recognized (Kane et al. Reference Kane, Leucht, Carpenter and Docherty2003; Kelly et al. Reference Kelly, Conley and Carpenter2005), therapy for first-episode schizophrenia has been poorly studied by controlled clinical trials and many studies focus on the efficacy rather than the safety of APs. The large majority of data on weight gain are generated from short-term studies on long-term treated patients and this situation potentially represents a significant bias in identifying the true effects of APs on patients' body weight. In addition, a large proportion of patients treated with APs receive multiple psychotropic medications for concomitant symptoms, and these medications (e.g. mood stabilizers and antidepressants) are frequently associated with weight gain (Kinon et al. Reference Kinon, Basson, Gilmore and Tollefson2001; Meyer, Reference Meyer2002). Our meta-analysis reports data on weight gain due to AP treatment in a sample of truly AP-naive in-patients, without adjunctive drug therapy. However, our review has several limitations. First, because of the small number of studies found, we could only check a few possible risk factors for weight gain. Several clinical characteristics of the patients included in these studies were not reported and this limited the possibility of looking further at the effect of weight gain in AP-naive patients. For example, most studies did not report the percentage of patients who were underweight, but only said that patients were on average in the normal BMI range at baseline, and did not report data about previous therapies with drugs that can potentially influence body weight. Again, we calculated standard deviations for studies that did not report them, taking a weighted mean of the standard deviation from the studies that did. We created a control group using the mean of the number of patients from all studies considered. For the control group we considered a mean weight gain of 0.00 kg and 0.0 BMI, similar to the weight gain that was found in the meta-analysis by Allison et al. (Reference Allison, Mentore, Heo, Chandler, Cappelleri, Infante and Weiden1999) after placebo (−0.5 kg); the s.d. was considered as the weighted mean s.d. of the studies considered in that analysis. This method may have inflated the results but only in the direction of significance, not in the direction of effect size. Moreover, we retrieved studies published in the past 11 years to avoid duplicating the results of Allison et al.'s meta-analysis (Reference Allison, Mentore, Heo, Chandler, Cappelleri, Infante and Weiden1999); this may have led us to miss some studies. Finally, our results on weight gain in kg need to be interpreted with caution because there was evidence of funnel plot asymmetry; however, the Egger analysis did not seem to find any publication bias.
Conclusions
Overall, the studies reviewed clearly show a causal relationship between AP treatment and weight gain because: (1) the studies were carried out on AP-naive patients; (2) the weight gain set in during the first 12 weeks of treatment; and (3) the weight gain was notable even when patients with co-therapy and with unbalanced meals were excluded. Our finding that weight gain sets in early and increases continually in AP-naive patients during AP treatment should encourage clinicians to set up an early weight-monitoring programme and to continue this as long as AP treatment lasts.
Weight gain is a significant long-term health issue because it is associated with insulin resistance and the resultant metabolic effects such as elevated triglycerides, diabetes and hypertension, all of which may increase the risk of cardiovascular disease (Haupt, Reference Haupt2006) and trigger well-established long-term medical consequences. Overweight is also a problem of great concern, particularly in young people, because of the substantial negative impact it has on patients' quality of life, giving rise to poor self-esteem and lack of acceptance in public places, and such distress can lead to low adherence to treatment (Haupt, Reference Haupt2006). Furthermore, unlike extrapyramidal motor side-effects, weight gain and other metabolic side-effects have no well-established treatments (Robinson et al. Reference Robinson, Woerner, Delman and Kane2005). Thus it seems that young people at their first episode of psychosis are at high risk of morbidity and mortality with AP therapy (Addington et al. Reference Addington, Mansley and Addington2003). Innovative approaches are called for, aimed at achieving long-term compensatory weight loss (Addington et al. Reference Addington, Mansley and Addington2003). As indicated by Alvarez-Jimenez et al. (Reference Alvarez-Jimenez, Gonzalez-Blanch, Vazquez-Barquero, Perez-Iglesias, Martinez-Garcia, Perez-Pardal, Ramirez-Bonilla and Crespo-Facorro2006), early behavioural intervention (diet, education, and exercise) seems to be effective in weight control and needs to be initiated before ascertainment of weight gain.
Declaration of Interest
None.
