FormalPara Key Points

This study investigated the risk-profile of different second-generation antipsychotics (SGAs) for the treatment of schizophrenia through a meta-analysis by assessing variations between the start of treatment and the end of follow-up.

Olanzapine and risperidone reported the greatest weight gain and olanzapine the largest BMI increase. Paliperidone showed the highest increase in total cholesterol, but is the only drug reporting an increase in the HDL cholesterol. Quetiapine XR showed the highest decrease in fasting glucose. Lurasidone showed the lowest increase in body weight and a reduction in BMI and was also the only treatment reporting a decrease in total cholesterol and triglycerides. The highest increase in systolic and diastolic blood pressure was reported by quetiapine XR.

The evidence on the metabolic risk profile of SGAs may support clinicians in the selection of the appropriate treatment for each patient and the development of economic evaluation studies.

1 Introduction

Schizophrenia is a severe long-term mental health condition that involves cognitive, mood symptoms, behavioral and emotional dysfunctions. The symptoms of schizophrenia are usually classified into positive symptoms—any change in behavior or thoughts, such as hallucinations or delusions—and negative symptoms—where people appear to withdraw from the world around them, take no interest in everyday social interactions, and often appear emotionless and flat. Late adolescence and early adulthood are peak periods for the onset of this disease, that is generally characterized by repeated relapses as well as a worsening of psychopathology and social functioning.

Approximately 1.1% of the adult population is affected and the origin seems to derive from both genetic and environmental factors. According to the Diagnostic and Statistical Manual of Mental Disorders, 5th Edition (DSM-5) criteria [1], it is characterized by at least two of the following six symptoms, each present for a significant portion of time during a 1-month period: delusions, hallucinations, disorganized speech (e.g. frequent derailment or inconsistency), grossly organized behavior or catatonic and negative symptoms (e.g. decreased expression of emotions and abulia).

The treatment of schizophrenia includes antipsychotic (or neuroleptic) drugs. The efficacy of neuroleptics has been extensively investigated and the results show, not only a reduction in the risk of relapse, but also a lower risk of hospitalization for the subjects treated. This translates positively into the quality of life of these patients [2].

Antipsychotic drugs have been available from the mid-1950s; the older types are called typical or first-generation antipsychotics (e.g. chlorpromazine, haloperidol). In the 1990s, new antipsychotic drugs, called second-generation or “atypical” antipsychotics (SGAs) were developed. The first of these SGAs was clozapine, which was followed by risperidone, olanzapine, ziprasidone, quetiapine, amisulpride, sertindole, lurasidone, paliperidone, iloperidone, asenapine, aripiprazole and, more recently, brexpiprazole, cariprazine and zotepine (not in the USA). Some of these SGAs (e.g. paliperidone, aripiprazole, olanzapine, and risperidone) are also available in long-acting injectable (LAI) formulations. The main guidelines recommend SGAs as first choice in both the first episode and in exacerbations. The recommendations on the use of SGAs are supported by a lower incidence of adverse events [3] and, as a consequence, by low discontinuation of therapy [4]. However, SGAs can cause weight gain and considerable changes in the metabolism, which can increase the risk of diabetes and increase circulating cholesterol levels.

Since many SGAs are available, understanding how the many substances compare with each other is important. Few studies focused on the comparison of antipsychotics with placebo in terms of response [5] or considered the real-world effectiveness in preventing relapses [6]. These studies showed that patients improved with antipsychotics compared with placebo, and that clozapine and long-acting injectable antipsychotic medications were the treatments with the highest rate of prevention of schizophrenia relapse. A more recent study reported no consistent superiority of any SGA across efficacy outcomes [7] and most of the literature showed that the main differences between the diverse compounds arise from the tolerability profiles [5, 8,9,10,11,12], especially in terms of metabolic side effects [13].

In the literature there are some systematic reviews comparing side effects, including the metabolic profile of specific oral SGAs in the treatment of schizophrenia. Although most of these studies have been performed in RCT (considered as the gold standard for proving causability), meta-analyses, including observational studies, have been performed as well. The meta-analyses including randomized clinical trials compared the different antipsychotics with placebo [14, 15] or different antipsychotics head-to-head [7, 16] or performed both comparisons [17]. Effect sizes were in general reported as risk ratios for dichotomous outcomes (e.g. sedation) and as standardized mean differences or mean differences for continuous outcomes (e.g. weight gain). Meta-analyses on observational studies carried out comparisons between the various SGA treatments or with placebo in terms of risk of weight gain or risk of developing type 2 diabetes mellitus [18, 19].

The aim of the present paper was to investigate the metabolic and cardiovascular risk profile of the main oral SGAs used in the treatment of adult patients with schizophrenia on the grounds of a systematic review and meta-analysis. In light of the great importance given to the collection and analysis of real-world data for the evaluation of outcomes of new health technologies [20], randomized controlled trials and observational studies have both been considered. Contrary to other published reviews, we assessed the mean variation of the main metabolic parameters between the start of treatment and the end of follow-up for each SGA, reporting detailed results for the different follow-up horizons.

2 Methods

The systematic review of the literature was conducted in November 2020 based on the PRISMA criteria (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) [21], starting from a search of the four fundamental elements (population, intervention, comparator, outcomes).

From preliminary research, no studies were found that considered a population to be only European, consequently, no restrictions were imposed on the choice of the base-case population for the analyses.

The drugs taken into consideration were lurasidone, aripiprazole, olanzapine, paliperidone extended release (XR), quetiapine XR and risperidone, which are the products with the highest market share in the major European countries for the treatment of patients with schizophrenia (IQVIA—data on file [22]). The choice also included lurasidone (recently launched onto the market) as a stabilization drug.

Given that no reliable advantage of any SGA emerged across efficacy outcomes [7], the systematic review of the literature has been focused on metabolic and cardiovascular adverse events. In particular, for each drug, variations from final and baseline values have been retrieved for the following parameters (metabolic profile): body weight, body mass index (BMI), total and high-density lipoprotein (HDL) cholesterol, triglycerides, fasting glucose, systolic and diastolic blood pressure.

The scientific databases used for the systematic review of the literature were Pubmed and Web of Science. Studies were considered if published in English and related to an adult population (aged ≥ 18 years). The research period has been restricted to the last 10 years. No restrictions were applied to the type of study.

The literature search has been performed according to the following:

“schizophrenia” AND (“lurasidone” OR “quetiapine XR” OR “quetiapine extended release” OR “extended release quetiapine” OR “risperidone” OR “olanzapine” OR “aripiprazole” OR “paliperidone”) AND (“fasting glucose” OR “fasting plasma glucose” OR “FPG” OR “weight” OR “BMI” OR “HDL” OR “total cholesterol” OR “triglyceride*” OR “blood pressure” OR “hypertension” OR “cardiovascular risk” OR “diabetes”).

Abstract and full-text selection was conducted independently by two expert reviewers (CR, AB). Data were extracted using a customized template developed in Microsoft Excel based on the PICOS statement. Information recorded included study features, participants and treatments characteristics and metabolic profiles.

Data referring to the different treatments were retrieved from all comparative and non-comparative studies identified. Outcomes variations from the different studies, calculated as the difference between the value at the last follow-up and the baseline value, were pooled through a random effect meta-analysis (mean differences) [23] considering the available follow-up. The analyses were performed using Stata® software (StataCorp, version 14) through the “metan” command, which requires two input parameters, effect estimate and standard error. In case the standard deviation was reported for the effect estimate, it was transformed into standard error according to formulas presented in Burns et al. [24].

A test on the summary effect measure is given, as well as a test for heterogeneity, quantified using the I2 metric [25]: the higher the values (from 0% to 100%) the larger the heterogeneity across studies. For the meta-analyses, a broader inclusion criterion has been applied so no restrictions on heterogeneity level have been considered. Results are displayed in forest plots according to different ranges of follow-up duration: ≤ 6 months, 6 < months ≤ 12, 12 < months ≤ 24 and 24 < months ≤ 36; this will allow further uses of the meta-analysis results in the context of economic evaluations from short to medium time horizons.

An appraisal of the studies included in the analyses has been performed in order to assess their methodological quality and to determine the extent to which the studies addressed the possibility of bias in their design, conduct and analysis. All papers selected for inclusion in the systematic review have been critically evaluated by two appraisers (CR, AB) according to the JBI Critical appraisal tools for randomized controlled trials (RCTs) and cohort studies [26].

The level of evidence (LOE) of the studies was assessed according to a classification provided by the Agency for Healthcare Research and Quality (AHRQ) [27], which considers three categories: high (current evidence derived from RCTs without important limitations), moderate (current evidence derived from RCTs with important limitations or very strong evidence from observational studies or case series), low (current evidence from observational studies, case series or just opinions). In our case, RCTs with lack of double-blinding, failure to adhere to intention-to-treat analysis or methodological flaws (treatment groups dissimilar at the baseline) were considered together with prospective observational trials and pre-post studies as moderate LOE. Retrospective studies and case series were considered low LOE.

Scenario analyses have been performed by considering only RCTs and by removing the low-quality studies according to the LOE to evaluate the robustness of the results.

3 Results

Figure 1 shows the search process according to PRISMA flow-chart. Starting with 3076 identified papers, the analysis focused on 79 that contained useful data for performing the meta-analyses. These were prospective studies (34%, n = 27), retrospective studies (5%, n =4) and RCTs (61%, n =48), with a total of 37,467 participants (median 69 participants/arm, range 7–5204). The mean age of the population was 36 ± 7.3 years and 62% were male. The number of studies with each individual SGA were: 49 olanzapine, 27 risperidone, 20 aripiprazole, 19 lurasidone, 13 quetiapine and 6 paliperidone.

Fig. 1
figure 1

Preferred reporting items for systematic reviews and meta-analyses (PRISMA) flow chart

Full size image

Table 1 presents the characteristics of the studies included in the quantitative synthesis, which presents only study arms whose drugs are considered in the present study, regardless of whether they were compared with other treatments or placebo, while Table 2 presents the metabolic parameters extracted. Parameter variations for total and HDL cholesterol, triglycerides and fasting glucose were expressed in mg/dL. To convert millimoles per liter to milligrams per deciliter, we multiplied total and HDL cholesterol values by 38.6, triglycerides values by 88.6 and fasting glucose values by 18.

Table 1 Studies characteristics
Full size table
Table 2 Parameter variations according to meta − analysis results (values are expressed as mean change and 95% CI)
Full size table

Supplementary Table 1 reports metabolic parameters derived by the considered studies while the Supplementary material shows a detailed analysis of results according to the forests plots for the different treatments, follow-up periods and parameters considered. A summary of the main findings considering the complete follow-up horizon of studies is presented in Table 2. The appraisal of the studies according to the risk of bias and LOE is reported in Supplementary Table 2.

From the meta-analyses, lurasidone was shown to be the treatment with a lower increase in body weight (0.43 kg) and with a decrease in BMI (− 0.10 kg/m2); it was also the only treatment reporting a decrease in total cholesterol (− 8.01 mg/dL) and triglycerides (− 5.33 mg/dL) and the highest decrease in HDL cholesterol (− 2.05 mg/dL).

Olanzapine and risperidone reported the largest weight gain of 4.52 and 4.19 kg, respectively, with significant differences compared with the other treatments. Olanzapine also reported the greatest variation in BMI (1.59 kg/m2) compared with the other SGAs and significant effects on the variation of triglycerides (33.10 mg/dL) and fasting glucose (6.24 mg/dL). Paliperidone showed the highest increase in total cholesterol (14.69 mg/dL) but reported a positive increase in the HDL cholesterol (0.57 mg/dL). Aripiprazole was another treatment showing a large increase in triglycerides (18.63 mg/dL).

The assessment of the variations in diastolic blood pressure was not possible for paliperidone due to lack of data. The highest increase in systolic and diastolic blood pressure was reported by quetiapine XR—2.60 and 2.77 mm Hg, respectively. Quetiapine XR was also the only drug reporting a decrease in fasting glucose (− 0.59 mg/dL).

The parameters reporting the higher heterogeneity (I2 > 50%) were body weight (aripiprazole, olanzapine, risperidone), BMI (aripiprazole, olanzapine), HDL cholesterol (olanzapine, risperidone), total cholesterol (olanzapine, risperidone), triglycerides (olanzapine, paliperidone, risperidone), fasting glucose (aripiprazole, olanzapine, risperidone), systolic blood pressure (olanzapine) and diastolic blood pressure (olanzapine, risperidone).

The scenario analysis performed considering only data from RCTs (see Supplementary Table 3) confirmed in general the results of the base-case analysis, with the exception of aripiprazole, which showed an increase in cholesterol HDL (0.59 vs − 0.62 mg/dL) and risperidone, which reported a decrease in triglycerides (− 3.69 vs 9.39 mg/dL) and in systolic blood pressure (− 2.33 vs 1.07 mm Hg). The scenario analysis conducted excluding low-quality studies (see Supplementary Table 4) showed only small variations in a limited set of parameters compared with the base case.

4 Discussion and Conclusion

Schizophrenia is a serious mental illness that affects how a person thinks, feels, and behaves. If left untreated, the symptoms of schizophrenia can be persistent and disabling. Despite its low prevalence (about 1% of the population) it has great health, social and economic burdens not only for patients but also for families, caregivers, and society. Comorbidities related to metabolic disorders and cardiovascular diseases, such as diabetes, hypertension, metabolic syndrome, and obesity are excessively prevalent among patients with schizophrenia. Compared with the general population, schizophrenia patients have nearly twice the risk of diabetes and metabolic syndrome [106] and an increased risk of mortality for cardiovascular disease, with patients’ life expectancy reduced by about 15 years [107]. Although some modifiable cardiovascular disease risk factors, such as sedentary lifestyle, may be associated with schizophrenia, several antipsychotics have been associated with an increased risk of weight gain and other metabolic abnormalities.

The literature reports some meta-analyses [7, 14,15,16,17,18,19] based on RCTs or observational studies which compared antipsychotics with each other and possibly with placebo in terms of relative risks or differences for the considered parameters. In contrast to these studies, the present work considered both RCTs and observational studies in order to provide results that may be also be extended to clinical practice contexts. Moreover, for each SGA we assessed the mean variation of the metabolic parameters between the start of treatment and the end of follow-up, thus providing immediate and clinically tangible results.

The analyses showed that metabolic effects are not statistically different across medicines although presenting great variations. For weight and BMI gain, respectively, olanzapine and risperidone and olanzapine alone reported significant differences compared with the other SGAs. In particular, olanzapine and risperidone reported a weight gain of 4.52 and 4.19 kg, respectively, while olanzapine reported an increase in BMI of 1.59 kg/m2. From the meta-analyses, lurasidone was shown to be the treatment with the lowest increase in body weight (0.43 kg) and with a decrease in BMI (− 0.10 kg/m2). These results are in line with a recent published study that provided a systematic review and meta-analysis of randomized trials lasting at least 6 months comparing SGAs head-to-head in schizophrenia and related disorders [7]. The paper reported that weight gain was greater with olanzapine than with all other non-clozapine SGAs and risperidone was significantly worse than several other SGAs. Olanzapine and clozapine have also been reported as the drugs causing greater weight gain compared with most other agents in another recent narrative review [108]. Huhn and colleagues [15] showed that placebo was preferred to olanzapine and risperidone when considering weight increase (mean difference, olanzapine: 2.78 kg, 95% CI 2.44–3.13; risperidone 1.44 kg, 95% CI 1.05–1.83).

The results on total cholesterol and fasting glucose are in line with those reported by Rummel-Kluge and colleagues [16] who showed that olanzapine produced a greater cholesterol increase than aripiprazole and risperidone, while cholesterol increase with quetiapine was greater than with risperidone. From our meta-analyses lurasidone showed a decrease in total cholesterol (− 8.01 mg/dL) and triglycerides (− 5.33 mg/dL) and a moderate variation in fasting glucose (1.78 mg/dL). Concerning fasting glucose, olanzapine produced the highest increase compared with the other drugs. Our data are in accordance with those derived from the meta-analysis of RCTs and observational studies [14, 18, 19, 109].

The present study has some limitations. First, changes in patients’ metabolic profiles have been derived from studies that reported, for each drug, different mean dosages per patient, highlighting that the dose is personalized according to patients’ characteristics. Second, the study focused on the analysis of metabolic side effects, without considering the impact of different side effects on patients’ quality of life. However, this was out of the scope of the analysis and, furthermore, there are difficulties in assessing the quality of life of patients with schizophrenia because of their cognitive impairments and lack of insight into their disease [110]. Third, the study focused on the analysis of metabolic effects due to the different treatments and did not consider the management of other adverse events.

Despite these limitations, this paper provides evidence on differences in the metabolic effects of SGAs, in a context where recent indications showed no consistent differences in their relative effectiveness.

These findings have important implications not only for clinical practice but also for health economics studies. On the one hand, because currently available antipsychotics vary more with regard to adverse effects than with efficacy, the selection of the appropriate treatment should do no harm to the patient, being mindful that untreated disease can commonly have greater adverse effects than medications. On the other hand, this analysis summarized the evidence on the metabolic impact of SGAs that could be the benchmarks for drugs launched into the market for the same indication, thus integrating the treatment cost with the cost for the management of the metabolic effects. Our findings could be used to perform cost-effectiveness or cost-utility analyses comparing new options with existing treatments and the budget impact of new treatments. A budget impact analysis could also be carried out to estimate the economic impact of a change of prescription mix for current treatment options.