Antipsychotic-Induced Weight Gain

Dose-Response Meta-Analysis of Randomized Controlled Trials

Hui Wu; Spyridon Siafis; Tasnim Hamza; Johannes Schneider-Thoma; John M Davis; Georgia Salanti; Stefan Leucht


Schizophr Bull. 2022;48(3):643-654. 

In This Article


We used dose-response meta-analysis to identify the possible relationships between weight gain and doses of 17 antipsychotics. For most drugs, the dose-response curves showed an initial dose-related increase in weight which plateaued at higher doses, while for aripiprazole, olanzapine, and paliperidone, the curves did not reach a plateau. Notable bell-shaped curves were found for quetiapine and ziprasidone. The magnitudes of weight gain were generally consistent with the results of previous network meta-analyses, which excluded subtherapeutic doses and pooled not only fixed- but also flexible doses of the same compound.[7,124]

The pharmacological mechanisms of antipsychotic-induced weight gain are not entirely clear. Antipsychotics differ in their receptor-binding profiles, yet they all target D2/3 receptors.[125] D2/3 antagonism could interfere with reward signaling and lead to weight gain.[126] Above certain doses, dopamine receptors can be fully bound and the increasing dose-response curves could plateau, as it can be observed for antipsychotics that act primarly as dopamine antagonists, such as haloperidol. In addition to the dopaminergic antagonism, multiple and synergistic pathways have been suggested, such as antagonism of serotonin (5-HT2C), histamine (H1), and muscarinic receptors.[127,128] Therefore, it is conceivable that dose-response curves of antipsychotics involving these receptors, such as clozapine and olanzapine, don't reach a plateau, as observed in our results (limited data were available for clozapine). Some antipsychotics may also have anti-hyperphagic mechanisms. For example, antipsychotics that act as 5-HT1A partial agonists could possibly induce less weight gain or even weight loss, when higher doses are used. Such hypothesis could be one of the underlying explanations for the dose-response curves that we found for lurasidone, quetiapine, and ziprasidone.[129] Finally, higher doses might be associated with more side-effects such as EPS and higher drop-out rates leading to less exposure to antipsychotic medications and potentially less weight gain.

Other rationales could also contribute to elucidating the dose-response curves. Characteristics of datasets, such as higher doses not examined, not enough distinguishable doses, weight gain imputed using last-observation-carried forward (LOCF) data could compromise the curves. Different subgroups of participants[130,131] could differ in their predisposition to gain weight. For example, we found sparse data in patients with predominant negative symptoms, who gained 2kg more than those did in the general group for both amisulpride and olanzapine.[121–123] This subgroup of patients might have lower body weight before initiation of antipsychotics, which could make them more vulnerable to antipsychotic-induced weight gain.[132,133] Furthermore, recovery of the negative symptoms may compensate the weight loss due to poor self-care and persistent disability.[134] In these comparably longer studies (one for 12 weeks, two 26 weeks), more weight gain happened after prolonged exposure to antipsychotics.[135]

In addition, a binary outcome (number of paticipants with significant weight gain) was investigated as a secondary outcome to test the robustness of the curves. Although dichotomous outcomes being more straighforward to clinical interpretation, ceiling effects when high cut-offs are applied make small changes less distinguishable. In our case, differences between continuous and dichotomous data were more obvious in antipsychotics with mild weight gain, such as aripiprazole and risperidone LAIs.

The one-stage approach allowed us to analyze a larger set of fixed-dose studies than a two-stage approach that excludes studies with less than three dose level groups,[26] which is more suitable for the complex weight gain question. However, our analysis has certain limitations. Fixed-dose studies with less than two dose levels of the same compound (or placebo) could not be analyzed using this approach. These non-dose-finding studies aimed to investigate head-to-head comparisons and could introduce heterogeneity. Nevertheless, a more comprehensive synthesis of evidence would require dose-response network meta-analysis, though their methods are still under development and not widely applied.[136,137] We were able to collate 150 eligible studies, from which 1/3 studies didn't provide usable data, otherwise, the precision of the curves could have been optimized. We also performed several sensitivity analyses to test the robustness of the results, however, not all heterogeneity was resolved and heterogeneity measures (i.e., VPC plots) should be interpreted with caution when only a few studies are available. Judgements of the curves rely on visual inspection, and due to the limited data, over-interpretation should be avoided. Antipsychotic-induced weight gain occurs quickly, often within the first weeks,[138] and more rapid than in the long-term.[139] However, almost all antipsychotics cause weight gain in the long-term,[135] and those with more weight gain in the short-term seem to cause more weight gain also in the long-term.[140] This work focused on short-term treatment from the point of early management/prevention of antipsychotic-induced weight gain.[141] We excluded maintenance studies to avoid methodological and clinical heterogeneity. In those studies, patients were all pre-treated which would limit the additional weight gain in the randomized phase. We found much fewer long-term data on this issue, moreover, some of which were not dose-finding studies. Only sparse data were available for individual antipsychotics. However, patients usually need to stay on the treatments for relapse prevention, which calls for urgency to investigate dose-response relationship for weight gain in longer durations.

Antipsychotic-induced weight gain is so far not clearly and completely understood.[142] Pillinger et al. found that increased baseline weight, male sex, and nonwhite race are more vulnerable to antipsychotic-induced metabolic dysregulation.[124] Moderators (e.g. demographics, baseline BMI, etc.) and mediators (e.g. adverse effects, comedications, etc.) were not sufficiently investigated in our analysis. In this work, participants were already overweight at baseline, 69% were male and almost half were nonwhite race. We only had sparse data on two a priori defined patient-subgroups (chronic patients with acute exacerbations and patients with predominant negative symptoms). Studies rarely reported data of subgroups and meta-regressions analyzing study-level data of participant-level predictors have limited statistical power and are prone to ecological fallacy. This question could be better elucidated by an individual-participant-data meta-analysis.

Individual variability should be considered, by means such as using plasma concentration of antipsychotics[143] and genetic polymorphisms,[128] unfortunately, those were not commonly reported in trials and not adopted by us, yet they would be important to better understand pharmacologic modulation. Limitations of generalizability of clinical trial data to the real-life practice should not be omitted.[144,145]

Antipsychotic-induced weight gain is a complex question, and investigating its dose-response relationships helps clinicians to optimize treatment plans for patients. Second-generation antipsychotics differ not only in their propensities but also in the shape of the dose-response curves for weight gain. Early on monitoring of weight gain should always be kept in mind.