We included 150 studies, from which 97 studies with 333 arms (n = 36 326 participants) provided usable data for meta-analysis. The PRISMA diagram of the search was provided (supplementary appendix 4). For the a priori defined analyses of specific patient-subgroups, data were only available for the chronic patients with acute exacerbations and patients with predominant negative symptoms (three studies on amisulpride, one of them also on olanzapine). The median study duration was 6 weeks, ranging from 4 to 26 weeks. Participants were already overweight at baseline with a median baseline weight of 79.4 kg, interquartile range IQR [73.8 kg, 84.8 kg] and BMI of 26.7 kg/m2 IQR [25.6 kg/m2, 28.5 kg/m2]. Detailed description of included studies was provided (supplementary appendix 6). For overall risk of bias of included studies, 66 studies were rated as low, 25 moderate, and 6 high (supplementary appendix 7).
The dose-response curves of the primary outcome are presented in Figure 1.
Antipsychotics in patients with acute exacerbations of chronic symptoms and with predominant negative symptoms. Figure 1 shows the dose-response curves for individual antipsychotics. The dose-response curve represents the mean differences of weight gain (in kg) comparing a given dose of the drug to non-exposure (i.e., 0 mg/d or placebo). The dotted lines are 95% confidence intervals. We used knot locations at the 25th, 50th, and 75th percentiles to anchor the curves, except for asenapine and sertindole, knot locations at the 10th, 50th, and 90th percentiles were used. Y-axis represents mean differences of weight gain for the dose-response curve. X-axis represents doses. Marks along the X-axis indicate available dose data. Different formulations were pooled. Chouinard 1993 and Marder 1994 are the Canadian and the American part of the RIS-INT-3 study, only combined data were available. n. studies = number of studies; n. arms = number of arms; N = number of participants.
Subgroup of Chronic Patients With Acute Exacerbations Amisulpride
A single 4-week dose-finding study (n = 241) compared amisulpride 400 mg/d, 800 mg/d and 1200 mg/d with 100mg/d. Amisulpride produced negligible weight gain (maximum MD = 0.14 kg) and the dose response curve was in essence flat (P-value = 0.52, Figure 1 Amisulpride).
Aripiprazole. Ten placebo-controlled studies (n = 2694) were included, eight studies examining aripiprazole oral,[38–44] one aripiprazole maintena, and one aripiprazole lauroxil. Doses between 2 mg/d and 30 mg/d were examined. Study durations ranged from 4 to 12 weeks (median 4 weeks). The dose-response curve suggested a fairly linear relationship between dose and weight gain (P-value < .01). It increased up to 10mg/d and afterwards a slower increase, but even at 30 mg/d, the MD of weight gain was small (0.97 kg, Figure 1 Aripiprazole).
Asenapine. Five studies (n = 1775) of six week duration examined asenapine doses between 5 mg/d and 20 mg/d with placebo, in which four were on asenapine oral[47–50] and one on asenapine maleate transdermal patch (HP3070). The dose-response curve plateaued at approximately 10 mg/d and 1.5 kg MD of weight gain (P-value < .01, Figure 1 Asenapine).
Brexpiprazole. Four studies (n = 2069) examined brexpiprazole doses between 0.25 mg/d and 5 mg/d with placebo.[52–55] Study durations were 6 weeks. The hyperbolic curve plateaued around 2 mg/d at 1.06 kg MD of weight gain (P-value < .01, Figure 1 Brexpiprazole).
Cariprazine. Four studies (n = 1874) of six week duration examined doses between 1.5 mg/d and 9 mg/d with placebo.[42,56–58] The dose-response curve plateaued around 4 mg/d with a slight increase of weight (MD = 0.62 kg, P-value < .01, Figure 1 Cariprazine).
Clozapine. A single small study of 16 weeks duration in treatment resistant patients (n = 43) was included. It compared doses of 100 mg/d, 300 mg/d and 600 mg/d. The dose-reponse curve appeared to be linear, however, due to the small sample size and limited data, the confidence interval was extremely wide and no statistically significant dose-response relationship was detected (P-value = .25). The maximum MD of weight gain was 3.75 kg (Figure 1 Clozapine).
Haloperidol. Twelve placebo-controlled studies (n = 2044) compared haloperidol doses between 4 mg/d and 20 mg/d,[32,34,39,44,47,60–66] from which eleven studies used single haloperidol doses as an active comparator in the evaluation of a second-generation antipsychotic, only one study was a dose-finding study for haloperidol. Study durations ranged from 4 to 8 weeks (median 6 weeks). The dose-response curve plateaued at 8 mg/d (P-value < .01), and the MD of weight gain at the plateau was mild (0.73 kg, Figure 1 Haloperidol).
Iloperidone. Four placebo-controlled dose-finding studies (n = 1905) examined iloperidone doses between 4 mg/d and 24 mg/d.[66,67] Study durations were 4 and 6 weeks. The dose-response curve had a relatively narrow confidence interval and reached a plateau at approximately 12 mg/d (P-value < .01) and a MD of weight gain of 2.26 kg (Figure 1 Iloperidone).
Lumateperone. Three dose-finding studies (n = 1093) lasting between 4 and 6 weeks compared lumateperone doses between 20 mg/d and 120 mg/d with placebo.[68–70] There was obvious weight gain in the placebo groups, means from 0.83 kg to 1.82 kg, and the differences in weight gain between lumateperone arms and placebo were small. The maximum MD of weight gain was small (0.65 kg) and no overall dose-response relationship was detected (P-value = .27, Figure 1 Lumateperone).
Lurasidone. Nine studies (n = 3124) of 6 weeks duration examined lurasidone doses between 20 mg/d and 160 mg/d.[65,71–78] The dose-response curve reached a plateau at 60 mg/d (maximum MD = 0.51 kg, P-value < .01, Figure 1 Lurasidone).
Olanzapine. Sixteen studies (n = 3575) examined olanzapine doses between 1 mg/d and 40 mg/d, 15 of which examined olanzapine oral[48,61,76,79–90] and one olanzapine LAI. In 11 out of 16 studies olanzapine was used as an active comparator for the evaluation of another second-generation antipsychotic. Study durations ranged between 4 and 8 weeks (median 6 weeks). The dose-response curve did not plateau at the highest examined dose (P-value < .01) (40mg/d with a corresponding MD = 3.62kg). Nevertheless, the slope of the curve was smaller beyond 10 mg/d and only one study examined olanzapine dose of 40 mg/d (n = 195) explaining the wide confidence intervals beyond 20 mg/d (Figure 1 Olanzapine).
Paliperidone. Ten studies (n = 3577) examined paliperidone doses between 1.5 mg/d and 15 mg/d, in which six were paliperidone oral studies[82–84,88,92,93] and four paliperidone LAI.[94–97] Study durations ranged between 6 and 13 weeks (median 6 weeks). The dose-response curve did not approach a clear plateau (P-value < .01) at the highest examined dose (MD = 1.95 kg at 15 mg/d) (Figure 1 Paliperidone).
Quetiapine. Quetiapine IR and ER were pooled in the primary analysis. In seven studies (n = 2336) of 6 to 8 weeks duration, quetiapine oral doses between 75 mg/d and 1200 mg/d[60,77,98–102] were examined, where doses above 1000 mg/d came from two studies in treatment-resistant patients.[99,102] The dose-response curve was approximately bell-shaped (P-value < .01), peaking at MD of 1.48 kg at around 600 mg/d (Figure 1 Quetiapine).
Risperidone. Seventeen studies (n = 5244) compared risperidone doses between 2 mg/d and 16 mg/d. Thirteen studies examined risperidone oral formulations,[43,50,56,62,63,66,69,70,103–108] 4 risperidone LAIs, including one study on risperidone RBP-7000, a sustained-release subcutaneous injection, two risperidone consta, an intramuscular injection[30,110] and one risperidone ISM, a new intramuscular injection. Study durations ranged between 4 and 12 weeks (median 6 weeks). The dose-response curve plateaued at approximately 5 mg/d and the maximum MD of weight gain was 1.82 kg (P-value < .01, Figure 1 Risperidone).
Sertindole. Three studies (n = 712) compared sertindole doses between 12 mg/d and 24 mg/d with placebo.[32–34] Study durations were 6 and 8 weeks. The dose-response curve was approximately bell-shaped with a peak at 17 mg/d where the MD of weight gain was 3.49 kg (P-value < .01, Figure 1 Sertindole).
Ziprasidone. Nine studies were eligible for inclusion, investigating a wide range of doses (4 mg, 10 mg, 40 mg, 60 mg, 120 mg, 200 mg, and 320 mg/day).[67,112–119] However, only two placebo-controlled studies (n = 599, 4 and 6 weeks) provided data for meta-analysis between 80 mg/d and 160 mg/d. The shape of the curve was estimated only by these two dose arms. The dose-response curve was bell-shaped with a peak of 1.24 kg MD in weight gain at around 80 mg/day (P-value = .02, Figure 1 Ziprasidone).
Zotepine. Only one study with data was eligible from our search, which compared 300 mg/d zotepine (n = 53, mean dose = 240.57 mg/d) with placebo (n = 53) and chlorpromazine. The MD of weight gain at 240.57 mg/d was 3.8 kg, but as there was only one dose arm a dose-response curve could not be estimated.
Subgroup of Patients With Predominant Negative Symptoms
Only three, comparably long studies (one 12 weeks and two 26 weeks, n = 482) were available in this subgroup.[121–123] All of them examined low doses of amisulpride which are sufficient for this indication (50–150 mg/d). There was a monotonic increasing dose-response relationship with no sign of plateau. The dose of 150mg/d has the largest MD of 2.67 kg (P-value < .01). Notably, all placebo arms in these studies were associated with weight loss, means ranging from 0.2 to 1.98 kg (Figure 1 Amisulpride for negative symptoms).
One of the 26-week studies (n = 173) also compared olanzapine oral 5 mg/d and 20 mg/d with placebo. The dose-response curve plateaued at approximately 9 mg/d with considerable weight gain (MD = 5.8kg, P-value < .01, Figure 1 Olanzapine for negative symptoms) and wide confidence intervals.
Our secondary outcome was the number of patients with weight gain, usually reported as the number of patients with at least 7% increase from baseline (68 studies), other criteria used in studies were at least 5% increase (3 studies), 10 kg increase (1 study) or weight gain as an adverse event (16 studies). For the rest that didn't report this binary data, we used an imputation method to analyze. Dose-response curves observed in the secondary outcome were comparable to the primary outcome (supplementary appendix 9).
In the sensitivity analyses, results were largely unchanged (supplementary appendix 10).
One notable difference was that the dose-response curve of haloperidol became bell-shaped when the only true dose-finding study was analyzed. The maximum mean difference of weight gain was 2.23 kg around 6 mg/d, which was a considerable weight increase in a short term (8 weeks).
For the primary outcome considerable heterogeneity (VPC ≥ 50%) across studies was observed for amisulpride, asenapine, cariprazine, olanzapine, and quetiapine, and lower levels of heterogeneity for the other antipsychotics (supplementary appendix 11).
Heterogeneity assessments of secondary outcome was also provided in supplementary appendix 11.
Small-study Effect/Publication Bias
For the primary outcome, we investigated small-study effects for haloperidol, olanzapine, paliparidone, and risperidone, since there were at least ten studies available for these antipsychotics. We conducted pairwise meta-analyses comparing these antipsychotics with placebo. Visual assessments of the contour-enhanced funnel plots and the statistically results from the Egger's regression tests suggested small-study effects for olanzapine (P-value < .01) in a way that small studies reported more weight gain compared to large trials. Small-study effects were not detected in haloperidol (P-value = .44), paliperidone (P-value = .63) and risperidone (P-value = .72) studies. Detailed analyses are provided (supplementary appendix 12).
Schizophr Bull. 2022;48(3):643-654. © 2022 Oxford University Press