Efficacy and Tolerance of Antipsychotics Used for the Treatment of Patients Newly Diagnosed with Schizophrenia: A Systematic Review and Meta-Analysis

This systematic review compared the efficacy and tolerance of oral antipsychotics (APDs) used in the treatment of schizophrenia following the PRISMA-P© statement (n = 21). The primary outcomes of interest were clinical response measured with symptoms’ improvement, tolerance to side effects and discontinuation reasons. There was better individual patients’ response to aripiprazole vs. ziprasidone and quetiapine ((CDSS p = 0.04), BPRS p = 0.02, YMRS p = 0.001) and ziprasidone vs. quetiapine (CGI p = 0.02, CDSS p = 0.02). Aripiprazole was more tolerated than risperidone, ziprasidone and quetiapine (p < 0.05). Quetiapine was more tolerated than aripiprazole, ziprasidone and risperidone (p < 0.05). Ziprasidone was more tolerated than quetiapine haloperidol and olanzapine (p < 0.05). Risperidone was more tolerated than olanzapine (p = 0.03) and haloperidol was more tolerated than olanzapine and quetiapine (p < 0.05). Olanzapine caused less discontinuation than quetiapine; quetiapine caused less discontinuation than ziprasidone, aripiprazole and haloperidol; ziprasidone caused less discontinuation than quetiapine, aripiprazole and haloperidol; aripiprazole caused less discontinuation than quetiapine, ziprasidone and olanzapine and olanzapine caused less discontinuation than ziprasidone and haloperidol (p < 0.05). It was concluded that individual patient clinical response, tolerance to side effects and life-threatening side effects remain the most reliable basis for selecting and continuing the use of APD.


Introduction
Schizophrenia affects 24 million (1 in 300, 0.32%) people worldwide and men are more affected and generally at a younger age than women [1].Acute, uncontrolled schizophrenia can lead to disability, impairment and severe distress.An individual diagnosed with schizophrenia has a higher likelihood (2-3 fold) of dying prematurely compared to the general population, due to infectious diseases, metabolic disorders and cardiovascular issues [1].The symptoms of schizophrenia are classed into positive symptoms (hallucination, delusion and disordered thinking) and negative symptoms (social withdrawal, appear emotionless and flat, disorganised speech, lack of drive and self-neglect) [2].
In England, the annual cost of schizophrenia to society and the public sector is GBP 11.8 billion and GBP 7.2 billion, respectively [3].The increased mortality risk in people with schizophrenia is complex.The incidence of psychosis in schizophrenia triggers a cascade of socioeconomic and lifestyle factors, including unhealthy lifestyle behaviour such as excessive alcohol consumption, smoking and poor diet that in return can result in adverse physical health outcomes [4,5].Suicide is the most common cause of death in this population; however, individuals with schizophrenia have been found to experience increased mortality from all causes [6].Semahegn et al. [7] found that 49% of patients diagnosed with

Review Aim
The aim of this review was to compare the evidence of effectiveness and tolerability of antipsychotic drugs (APDs) to treat psychosis in APD-naïve (first episode of psychosis) patients after both short-term (≤12 weeks) and long-term (>12 weeks) use.The following aspects were investigated, based upon evidence published in the current literature: 1.
Comparing the effectiveness of APDs used for the treatment of adults diagnosed with schizophrenia.

2.
Comparing the tolerability of APDs used for the adult patients diagnosed with schizophrenia.

3.
Comparing the discontinuation rate and reason of APDs used for the adult patients diagnosed with schizophrenia.

Methods and Design
This systematic review was reported according to the PRISMA statement and used a PRISMA flow chart to depict the process of the search strategy for the included published papers, following the Cochrane Handbook for Systematic reviewer's methodological guidelines [19].This project was prospectively registered and submitted on the PROS-PERO database (registration number: CRD42022311060, appendix 4).Data synthesis was performed using Review Manager (RevMan©) software V.5.4.1.A statistical analysis was conducted whenever similar data including continuous and dichotomous outcomes were available, and, in cases when it was not possible, a narrative analysis was conducted.
For both continuous and dichotomous outcomes, the mean difference and relative risk (RR), respectively, with a 95% confidence interval (CI) were reported.Results were considered significant when p < 0.05 and when the upper limit of 95% CI was less than 1 and the lower limit did not cross the line of no effect.A total effect estimate for dichotomous data was calculated as RR with 95% CI.In dichotomous data, the pooled effect size with 95% CI was obtained using a random effect (RE) model and fixed effect (FE) model.The RR with 95% CI was calculated using the Cochran-Mantel-Haenszel (M-H2) test.
An overall effect estimate for a continuous outcome was reported as the mean difference (MD).The effect sizes for continuous outcomes were centred on zero MD, values greater than zero favoured the intervention group and those less than zero were favouring the comparisons.For studies reporting a continuous outcome, effect size was calculated using the inverse variance (IV) method.I 2 describes the percentage of total variation across trials that is due to heterogeneity rather than chance or sampling errors [20].When the I 2 value was 30% or under for subgroups, the FE model was used for reporting, and when it was >30%, the RE model was used for more precision of reporting on the results.
A subgroup analysis was performed where there was more than one variable or indicators to demonstrate outcomes and when sufficient data were available (more than one study).The subgroup analysis determined the difference in pooled effect sizes between subgroups [21].Subdividing data into subgroups was performed to explore diverse outcomes, or to address particular inquiries concerning specific categories of interventions.As the systematic review was time-limited, publications from January 2000 to December 2021 were considered suitable for inclusion.The first APD was in use since 1950, the newest was marketed in 2020 and the newest FDA was registered in 2021.The systematic review ended in January 2022 and only randomised clinical trials (RCTs), which included oral routes of administration (tablets), were included.

Search Strategy
The PICO framework (population, intervention, control comparison intervention treatment/placebo/standard of care, outcomes) was used where the following applied: Population (P): 15 years of age or over, APD-naïve patients, diagnosed with confirmed schizophrenia or presenting psychosis with short-term (0-12 weeks) or long-term (>12 weeks) treatment.
Intervention (I): Treatment with APD.Comparators (C): Head to head, one APD compared to another APD.Outcomes (O): The outcome will determine the clinical effectiveness of the APDs in managing the symptoms, remission, preventing relapse and emerging of a side effect and promoting adherence to therapy.
Settings (S): Community, primary or secondary care.Three databases were reviewed: PubMed©, CINHAL© and ScienceDirect™.The medical subject headings (MeSH) search terms used in the PubMed database were (Randomised OR randomized) AND (schizophrenia) AND (naïve OR first-episode) AND (efficacy OR effectiveness OR effect) AND (adherence OR non-adherence OR compliance OR discontinuation OR withdrawal OR remission) AND (antipsychotic OR psychotropic OR psychotic medication OR psychiatric medication) AND (complication OR hospitalization OR relapse OR side effect OR tolerability OR adverse effect OR symptom OR risk OR clinical).

Selection Criteria
Adults diagnosed with schizophrenia aged 15 years old or over.-APD vs. APD/s, independent of whether they were first-generation (FGA) or secondgeneration (SGA) agents or the dose of administration.-APD-naïve patients or lifetime APD short-term use history.
A total of 14,417 articles were identified; the search was then restricted to the period of 2020-2021, and 'oral antipsychotics vs. oral antipsychotics' was added, which reduced the total publications to 166.The titles and abstracts were then screened to identify studies that meet the inclusion criteria, where 145 were removed as shown in Figure 1, leaving 21 publications for inclusion in the systematic review.The authors individually screened all studies and in case of disagreement, it was resolved with consensus (S1 and S2).

Data Analysis
The analysis model (FE vs. RE) was selected based on heterogeneity, when the I 2 value was above 30% for any of the subgroups; reporting was based on the RE model for precision of findings [22].The forest plots were only possible when there were two or more studies that shared similar outcomes.The statistical unit for any of the domains of study was based on numbers of patients.Due to lack of numbers of studies in each domain of the analysis (less than 10 studies), a funnel plot was not produced to assess publication bias [23].In addition, the chi-squared (X 2 , or Chi 2 ) test was used for heterogeneity in the forest plots using the formulae of I 2 = 100 × X 2 f/X 2 for quantifying the inconsistency of the discontinuation rate in one APD vs. another.
The included studies data reporting was inconsistent, as some studies presented the results in multiple means, e.g., means, standard means, mean difference, standard division, mean standard division, range, points or score, frequencies or percentages; accordingly, it was not possible to collate or further analyse.When a statistical analysis was not possible, reported results were tabulated and narratively explained.

Data Analysis
The analysis model (FE vs. RE) was selected based on heterogeneity, when the I 2 value was above 30% for any of the subgroups; reporting was based on the RE model for precision of findings [22].The forest plots were only possible when there were two or more studies that shared similar outcomes.The statistical unit for any of the domains of study was based on numbers of patients.Due to lack of numbers of studies in each domain of the analysis (less than 10 studies), a funnel plot was not produced to assess publication bias [23].In addition, the chi-squared (X 2 , or Chi 2 ) test was used for heterogeneity in the forest plots using the formulae of I 2 = 100 × X 2 f/X 2 for quantifying the inconsistency of the discontinuation rate in one APD vs. another.
The included studies data reporting was inconsistent, as some studies presented the results in multiple means, e.g., means, standard means, mean difference, standard division, mean standard division, range, points or score, frequencies or percentages; accordingly, it was not possible to collate or further analyse.When a statistical analysis was not possible, reported results were tabulated and narratively explained.Comparing PANSS, BPRS, CGI and CDSS, no significant differences were reported (p > 0.05) [24][25][26].For olanzapine and haloperidol, PANSS (total p = 0.02 and p = 0.019, negative p = 004 and general p = 0.003) and MADR (p = 0.02) were significantly different between the two APDs [25,27].For olanzapine and quetiapine, only PANSS negative scores were significantly different (p = 0.017) in the study by McEvoy et al. [28] but not (p > 0.05) in the study by San et al. [25].Comparing olanzapine and risperidone, it was concluded that there were no significant differences between the two APDs (p > 0.05) [25,26,28].Only PANSS positive symptoms' scores were significantly different (p = 0.031); other reported scores were not significantly different between quetiapine and risperidone (p > 0.05) [25,26,28].

Efficacy of Antipsychotics after Long-Term Treatment
The mean score at the endpoint analysis for aripiprazole and ziprasidone showed that the heterogeneity between the included studies was considerable in subgroup 88.2.5 (I 2 = 81%); the RE model was also used to ensure high precision of the reporting on this outcome.The mean change from baseline to the endpoint analysis showed that the heterogeneity on the FE model between the included studies was low (I 2 = 0-3%) for the subgroups.There was no significant difference in any of the subgroups' analyses except for the CDSS mean score at the endpoint (p = 0.04) and BPRS mean change between the baseline and endpoint (p < 0.001).Ziprasidone showed a lower mean score at the endpoint on CGI, BPRS, SANS and CDSS and a higher mean change from the baseline to endpoint on BPRS, SANS and CDSS.Aripiprazole showed a lower mean score at the endpoint on SAPS and YMRS and a higher mean change from the baseline to endpoint on SAPS and YMRS (Table 2).
The mean score at the endpoint analysis for aripiprazole and quetiapine showed that the heterogeneity between the included studies was moderate in subgroup 88.1.5(I 2 = 52%).Accordingly, the RE model was used to ensure high precision of the reporting on this outcome.Additionally, the mean change from the baseline to endpoint analysis showed that the heterogeneity on the FE model between the included studies was low (I 2 = 0-6%) for all the subgroups.There were no significant differences in any of the subgroups' analyses except for BPRS mean change between the baseline and endpoint (p = 0.002) and the YMRS mean score at the endpoint (p = 0.01).Regarding symptoms' improvement, aripiprazole showed a lower mean score at the endpoint on CGI, BPRS, SAPS and YMRS and a higher mean change between the baseline and endpoint on all scales except CDSS.Quetiapine showed a lower mean score at the endpoint and a higher mean change between the baseline and endpoint on CDSS only (Table 3).

Aripiprazole Aripiprazole
* Lower scale score is favourable; first value (columns 3 and 4): all studies average mean improvement endpoint, second value (columns 3 and 4): all studies average mean change from baseline, column 5: p-values (all studies average mean improvement at endpoint and mean change from baseline to endpoint) and column 6: drug with possible higher impact on the patient response (first line for mean score at endpoint, second line is for mean change between baseline and endpoint).The mean score at the endpoint analysis for ziprasidone and quetiapine showed that the heterogeneity between the included studies was low in all subgroups.Additionally, the mean change from the baseline to endpoint analysis showed that the heterogeneity on the FE model between the included studies was low (I 2 = 0%) for all the subgroups.There was no significant difference in any of the subgroups' analyses except for the CGI (p = 0.02) and CDSS (p = 0.020) mean score at the endpoint.Regarding symptoms' improvement, ziprasidone showed a lower mean score at the endpoint on CGI, BPRS, SANS, SAPS and YMRS and a higher mean change between the baseline and endpoint on CGI, SANS and SAPS.Quetiapine showed a lower mean score at the endpoint on CDSS only and a higher mean change from the baseline to endpoint on BPRS, SAPS, CDSS and YMRS (Table 4).For olanzapine and haloperidol, significant differences were reported for remission (p = 0.036) and MADRS (p = 0.045) and PANSS total scores (p < 0.001) [25,36]; however, there were no significant differences between olanzapine and risperidone reported (p > 0.05) [25,28,37].For olanzapine and quetiapine, the 'disorganized' mean change at 3 years [37], total PANSS [38] and positive PANSS at 52 weeks [28] were the only significantly different parameters (p < 0.05, p < 0.001, p = 0.013, respectively).For olanzapine and ziprasidone, CGI total mean change and SAPS mean change at 3 years [36] and PANSS total reduction at 12 months [36] were the only significant parameters between the two drugs (p < 0.05, p < 0.05, p < 0.001, respectively).Regarding ziprasidone and haloperidol, at 3 years, CGI total mean change, CDSS mean change, positive mean change [36] and PANSS total reduction [38] at 12 months showed significant differences between the two drugs (p < 0.05, p < 0.05, p < 0.05, p < 0.001, respectively).For quetiapine and haloperidol, at 3 years, CDSS mean change [35] and total PANSS [36] were statistically significantly different (p < 0.05, p < 0.001, respectively).The chlorpromazine group took significantly longer to achieve remission compared to the clozapine group [32,38].Kahn et al. [37] showed that the PANSS score significantly decreased for amisulpride and olanzapine compared to all other APDs (p < 0.001) between the baseline and the 12-month reviews, but there were no significant differences between all other APDs (p > 0.05).

Side Effects 7.2.1. Tolerance of Antipsychotics after Short-Term Treatment
Cardiovascular side effects were not intentionally excluded, as the data identified in the included studies were insufficient to focus on cardiovascular outcomes in particular.The analysis included all side effects that were reported in the studies included in this review.Included studies reported on all or some of the following: concentration difficulties, increased fatigability, sleepiness, memory impairment, depression, restlessness, increased duration of sleep, rigidity, akinesia, tremors, increased salivation, constipation, vertigo, amenorrhea, galactorrhoea, diminished sexual desire, orgasmic dysfunction, erectile dysfunction, ejaculatory dysfunction and weight gain.
The heterogeneity between Gómez-Revuelta et al. [30] and Robinson et al.'s [31] studies was moderate in subgroup 42.1.1(I 2 = 60%), and the RE model analysis was performed.The diminished sexual desire subgroup was the only subgroup with a significant difference between the two APDs, favouring aripiprazole (4.7%) over risperidone (12.5%) (p = 0.01) (Table 5).For akathisia after short-term treatment, the heterogeneity of the included studies was moderate (I 2 = 32%) in the FE model; RE was created and the finding was reported based on the RE (Figure 2).There was no significant difference between aripiprazole and risperidone (p = 0.440); however, akathisia events were higher with aripiprazole (18.7%) compared to risperidone (15.2%).For akathisia after short-term treatment, the heterogeneity of the included studies was moderate (I 2 = 32%) in the FE model; RE was created and the finding was reported based on the RE (Figure 2).There was no significant difference between aripiprazole and risperidone (p = 0.440); however, akathisia events were higher with aripiprazole (18.7%) compared to risperidone (15.2%).Robinson et al. [32] suggested that parkinsonian symptoms' prevalence was lower with aripiprazole (14.8%) than risperidone (15.5%), but not significantly different (p = 0.750).For cumulative EPS, the heterogeneity of the included studies was low (I 2 = 0%) and the FE was used to report on this outcome (Figure 3).There was no significant difference (p = 0.25) but EPS events were higher with aripiprazole (13%) than risperidone (9.5%).The reported RR (95% CI) was 1.22 (0.75, 3.06), indicating that the type of the drug had a high impact on causing treatment-emerged EPS in the study population.Wang et al. [26] and Gómez-Revuelta et al. [30] reported that all side effects for aripiprazole and risperidone were significantly different (p < 0.05) except for leukopaenia and ECG abnormalities at week 6.Amr et al. [29] and San et al. [25] reported on side effects for quetiapine and haloperidol, and only weight gain, blood glucose levels, insomnia and dizziness were not significantly different (p > 0.05).San et al. [25], Crespo-Facorro et al. [39] and Wang et al. [26] reported on side effects for quetiapine and ziprasidone, where weight gain, increased duration of sleep and somnolence were significantly different (p = 0.003 for all).
Robinson et al. [32] suggested that parkinsonian symptoms' prevalence was lower with aripiprazole (14.8%) than risperidone (15.5%), but not significantly different (p = 0.750).For cumulative EPS, the heterogeneity of the included studies was low (I 2 = 0%) and the FE was used to report on this outcome (Figure 3).There was no significant difference (p = 0.25) but EPS events were higher with aripiprazole (13%) than risperidone (9.5%).The reported RR (95% CI) was 1.22 (0.75, 3.06), indicating that the type of the drug had a high impact on causing treatment-emerged EPS in the study population.For akathisia after short-term treatment, the heterogeneity of the included studies was moderate (I 2 = 32%) in the FE model; RE was created and the finding was reported based on the RE (Figure 2).There was no significant difference between aripiprazole and risperidone (p = 0.440); however, akathisia events were higher with aripiprazole (18.7%) compared to risperidone (15.2%).Robinson et al. [32] suggested that parkinsonian symptoms' prevalence was lower with aripiprazole (14.8%) than risperidone (15.5%), but not significantly different (p = 0.750).For cumulative EPS, the heterogeneity of the included studies was low (I 2 = 0%) and the FE was used to report on this outcome (Figure 3).There was no significant difference (p = 0.25) but EPS events were higher with aripiprazole (13%) than risperidone (9.5%).The reported RR (95% CI) was 1.22 (0.75, 3.06), indicating that the type of the drug had a high impact on causing treatment-emerged EPS in the study population.Wang et al. [26] and Gómez-Revuelta et al. [30] reported that all side effects for aripiprazole and risperidone were significantly different (p < 0.05) except for leukopaenia and ECG abnormalities at week 6.Amr et al. [29] and San et al. [25] reported on side effects for quetiapine and haloperidol, and only weight gain, blood glucose levels, insomnia and dizziness were not significantly different (p > 0.05).San et al. [25], Crespo-Facorro et al. [39] and Wang et al. [26] reported on side effects for quetiapine and ziprasidone, where weight gain, increased duration of sleep and somnolence were significantly different (p = 0.003 for all).
Crespo-Facorro et al. [40] and Wang et al. [26] found that weight gain, somnolence, increased duration of sleep, treatment-emergent akathisia and EPS were significantly different between quetiapine and aripiprazole (p < 0.05).San et al. [25] and Wang et al. [26]  Wang et al. [26] and Gómez-Revuelta et al. [30] reported that all side effects for aripiprazole and risperidone were significantly different (p < 0.05) except for leukopaenia and ECG abnormalities at week 6.Amr et al. [29] and San et al. [25] reported on side effects for quetiapine and haloperidol, and only weight gain, blood glucose levels, insomnia and dizziness were not significantly different (p > 0.05).San et al. [25], Crespo-Facorro et al. [39] and Wang et al. [26] reported on side effects for quetiapine and ziprasidone, where weight gain, increased duration of sleep and somnolence were significantly different (p = 0.003 for all).

1.
Tolerance of antipsychotics after long-term treatment The FE model showed low heterogeneity between studies comparing aripiprazole and ziprasidone in all subgroups' analyses (I 2 between 0 and 30%).Table 6 illustrates that there was a significant difference between the two drugs in six subgroups: sleepiness (p = 0.03), increased duration of sleep (p = 0.003), rigidity (p = 0.02), erectile dysfunction (p = 0.005), ejaculatory dysfunction (p = 0.02) and weight gain (p = 0.01).Based on the reported 95% CI, the type of APDs had no consistent impact on causing side effects in the study population.
The heterogeneity of the included studies was low (I 2 = 1%) and the FE model was used to report this outcome.There was no significant difference between the two drugs (p = 0.28), where akathisia events were higher with ziprasidone (31%) than aripiprazole (25.5%).The reported RR (95% CI) was 0.83 (0.60, 1.16), indicating that the type of the drug had a low impact on causing treatment-emerged akathisia in the study population (Figure 4).The heterogeneity of the included studies was low (I 2 = 1%) and the FE model was used to report this outcome.There was no significant difference between the two drugs (p = 0.28), where akathisia events were higher with ziprasidone (31%) than aripiprazole (25.5%).The reported RR (95% CI) was 0.83 (0.60, 1.16), indicating that the type of the drug had a low impact on causing treatment-emerged akathisia in the study population (Figure 4).Crespo-Facorro et al. [34] concluded that parkinsonian symptoms were higher with ziprasidone (19.6%) than aripiprazole (17.7%) (p > 0.05).The heterogeneity of the included studies was low (I 2 = 0%).The EPS total events were not significantly different (p = 0.85); however, events were higher with aripiprazole (21.2%) than ziprasidone (20.2%).The reported RR (95% CI) was 1.05 (0.64, 1.74), indicating that the type of the drug had no impact on EPS development (Figure 5).Crespo-Facorro et al. [34] concluded that parkinsonian symptoms were higher with ziprasidone (19.6%) than aripiprazole (17.7%) (p > 0.05).The heterogeneity of the included studies was low (I 2 = 0%).The EPS total events were not significantly different (p = 0.85); however, events were higher with aripiprazole (21.2%) than ziprasidone (20.2%).The reported RR (95% CI) was 1.05 (0.64, 1.74), indicating that the type of the drug had no impact on EPS development (Figure 5).
Crespo-Facorro et al. [34] concluded that parkinsonian symptoms were higher with ziprasidone (19.6%) than aripiprazole (17.7%) (p > 0.05).The heterogeneity of the included studies was low (I 2 = 0%).The EPS total events were not significantly different (p = 0.85); however, events were higher with aripiprazole (21.2%) than ziprasidone (20.2%).The reported RR (95% CI) was 1.05 (0.64, 1.74), indicating that the type of the drug had no impact on EPS development (Figure 5).The FE model showed low heterogeneity between all studies in the subgroups' analyses (I 2 = 0%) but it was moderate in subgroup 15.1.1 (I 2 = 32%) and substantial in subgroup 15.1.17(I 2 = 69%); the RE analysis was used for reporting on this outcome.The side effect profile was significantly different in five subgroups: sleepiness (p < 0.001), increased duration of sleep (p = 0.001), tremors (p = 0.04), erectile dysfunction (p = 0.002) and galactorrhoea (p = 0.03).Based on the reported RR, the type of APDs had a low impact on causing side effects in the study population (Table 7).The FE model showed low heterogeneity between all studies in the subgroups' analyses (I 2 = 0%) but it was moderate in subgroup 15.1.1 (I 2 = 32%) and substantial in subgroup 15.1.17(I 2 = 69%); the RE analysis was used for reporting on this outcome.The side effect profile was significantly different in five subgroups: sleepiness (p < 0.001), increased duration of sleep (p = 0.001), tremors (p = 0.04), erectile dysfunction (p = 0.002) and galactorrhoea (p = 0.03).Based on the reported RR, the type of APDs had a low impact on causing side effects in the study population (Table 7).
Heterogeneity between all studies comparing quetiapine vs. ziprasidone was low for most of the subgroups' analyses (I 2 = 0-18%) but it was moderate in subgroups 16.1.10(I 2 = 53%), 16.1.17(I 2 = 36%) and 16.1.18(I 2 = 44%); the RE analysis was used for reporting on this outcome.There were significant differences in four subgroups: rigidity (p < 0.03), vertigo (p = 0.05), amenorrhoea (p = 0.006) and weight gain (p = 0.003).Based on the reported RR, the type of APDs had a low impact on causing side effects in the study population (Table 8).Heterogeneity between all studies comparing quetiapine vs. ziprasidone was low for most of the subgroups' analyses (I 2 = 0-18%) but it was moderate in subgroups 16.1.10(I 2 = 53%), 16.1.17(I 2 = 36%) and 16.1.18(I 2 = 44%); the RE analysis was used for reporting on this outcome.There were significant differences in four subgroups: rigidity (p < 0.03), vertigo (p = 0.05), amenorrhoea (p = 0.006) and weight gain (p = 0.003).Based on the reported RR, the type of APDs had a low impact on causing side effects in the study population (Table 8).
The heterogeneity of the three included studies was low (I 2 = 0%); the FE model was used to report on akathisia (Figure 7).There was a significant difference between the two drugs (p = 0.005); akathisia events were higher with ziprasidone (30.6%) than quetiapine (16.3%).The reported RR (95% CI) was >1 (1.87 [1.21, 2.91]), indicating that the type of the drug had a high impact on causing treatment-emerged akathisia in the study population.The heterogeneity of the three included studies was low (I 2 = 0%); the FE model was used to report on akathisia (Figure 7).There was a significant difference between the two drugs (p = 0.005); akathisia events were higher with ziprasidone (30.6%) than quetiapine (16.3%).The reported RR (95% CI) was >1 (1.87 [1.21, 2.91]), indicating that the type of the drug had a high impact on causing treatment-emerged akathisia in the study population.At week 52, San et al. [25] did not observe significant differences (p > 0.05) between quetiapine and olanzapine in UKU psychiatric side effects (2.3 ± 1.4 vs. 1.7 ± 1.8, respectively), neurological side effects (0.3 ± 0.8 vs. 0 ± 0, respectively) or the glucose level that showed to be similar for both drugs (4.6 ± 0.5 vs. 4.6 ± 0.7, respectively).Weight gain was slightly higher with olanzapine (+9 Kg) than quetiapine (+6 Kg) but remained not significantly different (p > 0.05).The FE model showed moderate heterogeneity between all studies comparing olanzapine vs. quetiapine and it was low in most subgroups (I 2 = 0%), but it was substantial in subgroup 19.1.4(I 2 = 65%); the RE analysis was used for reporting on this outcome.Weight gain was significantly different (p < 0.001).Based on the reported RR, the type of APDs had a low impact on causing side effects in the study population (Table 9).At week 52, San et al. [25] did not observe significant differences (p > 0.05) between quetiapine and olanzapine in UKU psychiatric side effects (2.3 ± 1.4 vs. 1.7 ± 1.8, respectively), neurological side effects (0.3 ± 0.8 vs. 0 ± 0, respectively) or the glucose level that showed to be similar for both drugs (4.6 ± 0.5 vs. 4.6 ± 0.7, respectively).Weight gain was slightly higher with olanzapine (+9 Kg) than quetiapine (+6 Kg) but remained not significantly different (p > 0.05).The FE model showed moderate heterogeneity between all studies comparing olanzapine vs. quetiapine and it was low in most subgroups (I 2 = 0%), but it was substantial in subgroup 19.1.4(I 2 = 65%); the RE analysis was used for reporting on this outcome.Weight gain was significantly different (p < 0.001).Based on the reported RR, the type of APDs had a low impact on causing side effects in the study population (Table 9).Gómez-Revuelta et al. [36] reported a higher rate of EPS with olanzapine (30%) than quetiapine (20%) (p > 0.05); however, Kahn et al. [37] found a lower prevalence with olanzapine (7%) than quetiapine (8%) (p > 0.05).The BAS scores in two studies [28,36] did not involve any significant differences (p > 0.05); however, exact scores were not reported.The heterogeneity of the included studies was moderate (I 2 = 40%); the RE analysis was used for reporting on this outcome.There was no significant difference between the two drugs (p = 0.49); however, akathisia events were higher with quetiapine (16.7%) than olanzapine (14.3%).The reported RR (95% CI) was 0.81 (0.44, 1.48), indicating that the type of the drug had a low impact on causing treatment-emerged akathisia in the study population (Figure 8).Gómez-Revuelta et al. [36] reported a higher rate of EPS with olanzapine (30%) than quetiapine (20%) (p > 0.05); however, Kahn et al. [37] found a lower prevalence with olanzapine (7%) than quetiapine (8%) (p > 0.05).The BAS scores in two studies [28,36] did not involve any significant differences (p > 0.05); however, exact scores were not reported.The heterogeneity of the included studies was moderate (I 2 = 40%); the RE analysis was used for reporting on this outcome.There was no significant difference between the two drugs (p = 0.49); however, akathisia events were higher with quetiapine (16.7%) than olanzapine (14.3%).The reported RR (95% CI) was 0.81 (0.44, 1.48), indicating that the type of the drug had a low impact on causing treatment-emerged akathisia in the study population (Figure 8).[28,36,37].
Heterogeneity between all studies comparing risperidone vs. quetiapine was low in all subgroups except in subgroups 20.1.2(I 2 = 54%) and 20.1.3(I 2 = 43%); the RE analysis was used for reporting on this outcome.Increased sleep duration was significantly different (p = 0.02).Most events were higher in risperidone except sleepiness, increased duration of sleep and constipation, which were higher in quetiapine; this may suggest that quetiapine showed better tolerability in the study sample.Based on the reported RR, the type of APDs had a low impact on causing side effects in the study population (Table 10).[25] reported slightly higher UKU psychiatric and neurological side effect scores for risperidone (2.5 ± 3.2 and 1.1 ± 1.8) than quetiapine (2.3 ± 1.4 and 0 ± 0).Both weight gain and glucose levels were higher for risperidone (+7 kg, 4.7 ± 0.3) than for quetiapine (+6 kg, 4.6 ± 0.7).However, all parameters were not significantly different  [28,36,37].
Heterogeneity between all studies comparing risperidone vs. quetiapine was low in all subgroups except in subgroups 20.1.2(I 2 = 54%) and 20.1.3(I 2 = 43%); the RE analysis was used for reporting on this outcome.Increased sleep duration was significantly different (p = 0.02).Most events were higher in risperidone except sleepiness, increased duration of sleep and constipation, which were higher in quetiapine; this may suggest that quetiapine showed better tolerability in the study sample.Based on the reported RR, the type of APDs had a low impact on causing side effects in the study population (Table 10).
At week 52, San et al. [25] reported slightly higher UKU psychiatric and neurological side effect scores for risperidone (2.5 ± 3.2 and 1.1 ± 1.8) than quetiapine (2.3 ± 1.4 and 0 ± 0).Both weight gain and glucose levels were higher for risperidone (+7 kg, 4.7 ± 0.3) than for quetiapine (+6 kg, 4.6 ± 0.7).However, all parameters were not significantly different (p > 0.05).The reported EPS percentage in Gómez-Revuelta et al.'s study [36] was higher for risperidone (40%) than quetiapine (20%), but the difference was not significant (p = 0.043).The heterogeneity of the included studies was moderate (I 2 = 40%); the RE analysis was used for reporting on this outcome.There was no significant difference (p = 0.49); however, akathisia events were higher with quetiapine (16.7%) than risperidone (14.3%).The reported RR (95% CI) was 0.81 (0.44, 1.48), indicating that the type of the drug had a low impact on causing treatment-emerged akathisia in the study population (Figure 9).San et al. [25] concluded that the UKU psychiatric and neurological side effects were slightly higher in risperidone groups (2.5 ± 3.2 and 1.1 ± 1.8) than olanzapine groups (1.7 ± 1.8 and 0.3 ± 0.8).The results of weight gain were slightly greater in olanzapine groups (+9 kg, 4.6 ± 0.5) than risperidone groups (+7 kg, 4.7 ± 0.3) at week 52; however, there were no significant differences in any of the three outcomes (p > 0.05).The FE model showed low heterogeneity between all studies comparing olanzapine vs. risperidone, but it was moderate in subgroup 17.1.4(I 2 = 59%); the RE analysis was used for reporting on this outcome, where weight gain was found to have a significant difference (p = 0.03).Based on the reported RR, the type of APDs had a low impact on causing side effects in the study population (Table 11).Only an abnormal cholesterol level, abnormal prolactin level, abnormal serum aspartate aminotransferase (AST) level and abnormal serum alanine aminotransferase (ALT) level were statistically significantly different (p < 0.05) between olanzapine and haloperidol [25,33,35,36,39].The heterogeneity of the included studies was low (I 2 = 0%).There was a significant difference (p < 0.001) with weight gain events, which were higher with olanzapine (63.4%) than haloperidol (36.6%).The reported RR (95% CI) was 1.59 (1.33, 1.91), indicating that the type of the drug had a high impact on causing weight gain in the study population (Figure 10).San et al. [25] concluded that the UKU psychiatric and neurological side effects were slightly higher in risperidone groups (2.5 ± 3.2 and 1.1 ± 1.8) than olanzapine groups (1.7 ± 1.8 and 0.3 ± 0.8).The results of weight gain were slightly greater in olanzapine groups (+9 kg, 4.6 ± 0.5) than risperidone groups (+7 kg, 4.7 ± 0.3) at week 52; however, there were no significant differences in any of the three outcomes (p > 0.05).The FE model showed low heterogeneity between all studies comparing olanzapine vs. risperidone, but it was moderate in subgroup 17.1.4(I 2 = 59%); the RE analysis was used for reporting on this outcome, where weight gain was found to have a significant difference (p = 0.03).Based on the reported RR, the type of APDs had a low impact on causing side effects in the study population (Table 11).Only an abnormal cholesterol level, abnormal prolactin level, abnormal serum aspartate aminotransferase (AST) level and abnormal serum alanine aminotransferase (ALT) level were statistically significantly different (p < 0.05) between olanzapine and haloperidol [25,33,35,36,39].The heterogeneity of the included studies was low (I 2 = 0%).There was a significant difference (p < 0.001) with weight gain events, which were higher with olanzapine (63.4%) than haloperidol (36.6%).The reported RR (95% CI) was 1.59 (1.33, 1.91), indicating that the type of the drug had a high impact on causing weight gain in the study population (Figure 10).
Only an abnormal cholesterol level, abnormal prolactin level, abnormal serum aspartate aminotransferase (AST) level and abnormal serum alanine aminotransferase (ALT) level were statistically significantly different (p < 0.05) between olanzapine and haloperidol [25,33,35,36,39].The heterogeneity of the included studies was low (I 2 = 0%).There was a significant difference (p < 0.001) with weight gain events, which were higher with olanzapine (63.4%) than haloperidol (36.6%).The reported RR (95% CI) was 1.59 (1.33, 1.91), indicating that the type of the drug had a high impact on causing weight gain in the study population (Figure 10).The olanzapine group had less EPS events compared to haloperidol [33,38] (p < 0.001) and the severity of akathisia (BAS total score) was not significantly different (p > 0.05) [36], The olanzapine group had less EPS events compared to haloperidol [33,38] (p < 0.001) and the severity of akathisia (BAS total score) was not significantly different (p > 0.05) [36], but actual scores were not reported.Heterogeneity of the included studies was low (I 2 = 0%).The difference was significant (p < 0.001), but akathisia events were higher with olanzapine (26.7%) than haloperidol (8.6%).The reported RR (95% CI) was 3.19 (1.70, 6.00), indicating that the type of the drug had a high impact on causing akathisia in the study population (Figure 11).but actual scores were not reported.Heterogeneity of the included studies was low (I 2 = 0%).The difference was significant (p < 0.001), but akathisia events were higher with olanzapine (26.7%) than haloperidol (8.6%).The reported RR (95% CI) was 3.19 (1.70, 6.00), indicating that the type of the drug had a high impact on causing akathisia in the study population (Figure 11).The prevalence of side effects for the two APDs was not significantly different [25,36] in all measured outcomes (p > 0.05) between haloperidol and quetiapine.Figure 12 shows that the heterogeneity of the included studies was low (I 2 = 0%) and that the difference was significant (p = 0.02), with the akathisia events being higher with haloperidol (26.7%) than quetiapine (14.6%).The reported RR (95% CI) was 1.81 (1.09, 3.01), indicating that the type of the drug had a high impact on causing treatment-emerged akathisia in the study population.Only three studies [25,36,38] compared haloperidol and ziprasidone, and reported that there was no significant difference (p > 0.05) in all measured outcomes.Figure 13 shows that the heterogeneity of the included studies was low (I 2 = 0%) and that the difference was significant between the two drugs (p = 0.03), with weight gain events being higher with haloperidol (50%) than ziprasidone (34.4%).The reported RR (95% CI) was 1.45 (1.03, 2.04), indicating that the type of the drug had a high impact on causing weight gain in the study population.The prevalence of side effects for the two APDs was not significantly different [25,36] in all measured outcomes (p > 0.05) between haloperidol and quetiapine.Figure 12 shows that the heterogeneity of the included studies was low (I 2 = 0%) and that the difference was significant (p = 0.02), with the akathisia events being higher with haloperidol (26.7%) than quetiapine (14.6%).The reported RR (95% CI) was 1.81 (1.09, 3.01), indicating that the type of the drug had a high impact on causing treatment-emerged akathisia in the study population.
Pharmacy 2023, 11, x FOR PEER REVIEW 16 of 32 but actual scores were not reported.Heterogeneity of the included studies was low (I 2 = 0%).The difference was significant (p < 0.001), but akathisia events were higher with olanzapine (26.7%) than haloperidol (8.6%).The reported RR (95% CI) was 3.19 (1.70, 6.00), indicating that the type of the drug had a high impact on causing akathisia in the study population (Figure 11).The prevalence of side effects for the two APDs was not significantly different [25,36] in all measured outcomes (p > 0.05) between haloperidol and quetiapine.Figure 12 shows that the heterogeneity of the included studies was low (I 2 = 0%) and that the difference was significant (p = 0.02), with the akathisia events being higher with haloperidol (26.7%) than quetiapine (14.6%).The reported RR (95% CI) was 1.81 (1.09, 3.01), indicating that the type of the drug had a high impact on causing treatment-emerged akathisia in the study population.Only three studies [25,36,38] compared haloperidol and ziprasidone, and reported that there was no significant difference (p > 0.05) in all measured outcomes.Figure 13 shows that the heterogeneity of the included studies was low (I 2 = 0%) and that the difference was significant between the two drugs (p = 0.03), with weight gain events being higher with haloperidol (50%) than ziprasidone (34.4%).The reported RR (95% CI) was 1.45 (1.03, 2.04), indicating that the type of the drug had a high impact on causing weight gain in the study population.Only three studies [25,36,38] compared haloperidol and ziprasidone, and reported that there was no significant difference (p > 0.05) in all measured outcomes.Figure 13 shows that the heterogeneity of the included studies was low (I 2 = 0%) and that the difference was significant between the two drugs (p = 0.03), with weight gain events being higher with haloperidol (50%) than ziprasidone (34.4%).The reported RR (95% CI) was 1.45 (1.03, 2.04), indicating that the type of the drug had a high impact on causing weight gain in the study population.
Only three studies [25,36,38] compared haloperidol and ziprasidone, and reported that there was no significant difference (p > 0.05) in all measured outcomes.Figure 13 shows that the heterogeneity of the included studies was low (I 2 = 0%) and that the difference was significant between the two drugs (p = 0.03), with weight gain events being higher with haloperidol (50%) than ziprasidone (34.4%).The reported RR (95% CI) was 1.45 (1.03, 2.04), indicating that the type of the drug had a high impact on causing weight gain in the study population.Three studies [25,36,37] reported no statistically significant differences between olanzapine and ziprasidone (p = 0.05) in all measured outcomes.Figure 14 shows that the heterogeneity of the included studies was low (I 2 = 0%) and that the difference was Figure 13.Weight gain for haloperidol vs. ziprasidone long-term treatment-FE [36,37].
Three studies [25,36,37] reported no statistically significant differences between olanzapine and ziprasidone (p = 0.05) in all measured outcomes.Figure 14 shows that the heterogeneity of the included studies was low (I 2 = 0%) and that the difference was significant (p < 0.001) in weight gain with higher events with olanzapine (81%) compared to ziprasidone (34.4%).The reported RR (95% CI) was 2.26 (1.69, 3.03), indicating that the type of the drug had a high impact on causing weight gain in the study population.
Pharmacy 2023, 11, x FOR PEER REVIEW 17 of 32 significant (p < 0.001) in weight gain with higher events with olanzapine (81%) compared to ziprasidone (34.4%).The reported RR (95% CI) was 2.26 (1.69, 3.03), indicating that the type of the drug had a high impact on causing weight gain in the study population.Additionally, one study [36] reported that aripiprazole administration was more likely to be associated with akinesia (p = 0.004)).In two studies [27,38], the authors found that the clozapine group experienced fewer side effects compared to the chlorpromazine group at 52 weeks but it was not significant (p = 0.05).Gómez-Revuelta et al. [36] reported EPS was higher with risperidone (40%) than aripiprazole (23.8%) but it was not significant (p = 0.456).Kahn et al. [37] found a higher prevalence of EPS with haloperidol than any SGA groups (34% vs. 6-17%; p < 0.001), and higher proportions of patients on haloperidol or ziprasidone experienced akathisia than with other APDs (26-28% vs. 10-16%; p < 0.01).

Discontinuation of APDs
San et al. [25] showed olanzapine had a longer period of use (260.2days) than quetiapine (187.1), which was similar in Gómez-Revuelta et al.'s study [36] (855 days vs. 60 days) (p < 0.05).The FE model showed low heterogeneity between studies comparing quetiapine vs. olanzapine and it was low for most of the analyses (I 2 = 0-22%), but it was considerable in subgroup 56.1.3(I 2 = 84%); the RE analysis was used for reporting on this outcome.One subgroup had statistically significant differences between the two APDs (lack of efficacy, p < 0.001).Based on the reported RR, the type of APDs had a low impact on causing treatment discontinuation the study population, which might be due to the patients' factors (Table 12).Time to discontinuation was longer with risperidone than with quetiapine (p < 0.05) (786 days vs. 60 days [36] and 206.2 days vs. 187.1 days [25]).Heterogeneity between studies comparing quetiapine vs. risperidone was low in all subgroups (I 2 = 0%) and considerable in subgroup 55.1.1 (I 2 = 86%); the RE analysis was used for reporting on this outcome.Discontinuation of APDs due to side effects was statistically significantly different between the two APDs (p = 0.02); the RR showed that the type of APDs had a low impact on causing treatment discontinuation in the study population (Table 13).San et al. [25] reported that time to discontinuation for ziprasidone was shorter than quetiapine (142.7 vs. 187.1),which was not significantly different between the two APDs (p > 0.05).There was heterogeneity between all studies in all subgroups (I 2 = 0-2%).Discontinuation due to lack of efficacy and side effects involved significant differences (p < 0.001).Based on the reported RR, the type of APDs had a high impact on causing treatment discontinuation in the study population (Table 14).In one study, haloperidol had a shorter time to discontinuation (125 days) compared to quetiapine (187.1 days) [25], but in another study [35], it was shorter for quetiapine (60 days) than haloperidol (295 days) and it was significant (p < 0.05).Heterogeneity between studies comparing quetiapine and haloperidol in four subgroups was low (I 2 = 0%) but considerable in subgroup 58.1.1 (I 2 = 82%); the RE analysis was used for reporting on this outcome.Discontinuation due to side effects involved significant differences between the two APDs (p = 0.01).Based on the reported RR, the type of APDs had a high impact on causing treatment discontinuation in the study population (Table 15).The mean time to any cause of discontinuation in the quetiapine-treated group [36] was only 60 days vs. 452 days for aripiprazole (p < 0.05); however, the gap between the two drugs was smaller in another study [39] (77.24 days vs. 106.71days, respectively, p < 0.05).Heterogeneity between all studies comparing quetiapine and aripiprazole was low in all subgroups (I 2 = 0-9%).Discontinuation due to lack of efficacy (p < 0.001) and lack of compliance (p = 0.04) was significantly different between the two APDs.Based on the reported RR, the type of APDs had a high impact on causing treatment discontinuation in the study population (Table 16).In one study [36], a longer time (in days) to discontinuation was reported in the aripiprazole group (452 days) than in the ziprasidone group (251 days), but it was slightly shorter in aripiprazole-treated groups (106.71)than in the ziprasidone group (129.88) in another study [39], which was not significantly different (p > 0.05).There was heterogeneity between all studies comparing aripiprazole and ziprasidone in all subgroups (I 2 = 0%).Discontinuation due to lack of compliance and side effects (p < 0.001) was statistically significantly different between the two APDs.Based on the reported RR, the type of APDs had a high impact on causing treatment discontinuation in the study population (Table 17).The mean discontinuation time (in days) seen in the olanzapine group (855 days) and risperidone group (786 days) was not significantly different (p > 0.05) in one study [35], but it was significant (p < 0.05) in another study [25] where risperidone-treated patients had a lower (206) number of days than in the olanzapine group (260).Heterogeneity was moderate between all studies comparing risperidone and olanzapine in subgroup 61.1.2(I 2 = 41%); the RE analysis was used for reporting on this outcome with no significant differences between the two drugs in lack of efficacy (p = 0.65) and side effect (p = 0.25) subgroups.Based on the reported RR, the type of APDs had a high impact on causing treatment discontinuation in the study population (Table 18).The time to discontinuation was reported in two studies [25,36] for risperidone and ziprasidone (786 days vs. 251 days and 206 vs. 142.7,respectively), which was significantly different (p < 0.05).Heterogeneity between all studies comparing risperidone and ziprasidone in all subgroups was moderate (I 2 = 54, 44, 35%); the RE analysis was used for reporting on this outcome.Only lack of efficacy (p = 0.90), side effects (p = 0.42) and drop out (p = 0.59) were significantly different between the two drugs.Based on the reported RR, the type of APDs had a high impact on causing treatment discontinuation in the study population (Table 19).Patients on haloperidol (295 days) had less time (in days) to discontinuation [36] compared to risperidone (786 days), which was similar to one other study [25] (125 days vs. 206 days), which was significant in both studies (p < 0.05).Heterogeneity between all studies comparing risperidone and haloperidol for all subgroups (I 2 = 0%) was low.Only lack of efficacy (p = 0.30), side effects (p = 0.81) and drop out (p = 0.46) were significantly different.Based on the reported RR, the type of APDs had a high impact on causing treatment discontinuation in the study population (Table 20).Only San et al. [25] showed the mean time to discontinuation in the ziprasidone group was less (142.7 days) than in the olanzapine group (260.2 days), which was significantly different between the two APDs (p < 0.05).Heterogeneity between all studies comparing olanzapine and ziprasidone was substantial in the 74.1.4subgroup (I 2 = 70%); the RE analysis was used for reporting on this outcome.Discontinuation due to lack of efficacy (p = 0.01), side effects (p < 0.001) and lack of compliance (p = 0.05) was significantly different.Based on the reported RR, the type of APDs had a high impact on causing treatment discontinuation in the study population (Table 21).San et al. [25] concluded that the olanzapine group mean time to discontinuation was longer (260 days) than with haloperidol (125 days) (p > 0.05).Heterogeneity between all studies comparing olanzapine and haloperidol was considerable in subgroup 75.1.4(I 2 = 68%); the RE analysis was used for reporting on this outcome.Lack of efficacy (p < 0.001) and side effects (p = 0.001) were significantly different.Based on the reported RR, the type of APDs had a high impact on causing treatment discontinuation in the study population (Table 22).The mean time to discontinuation was longer for ziprasidone (142.7 days) than haloperidol (125 days) in one study [26] but it was not in another [35] (ziprasidone: 251 days vs. haloperidol: 295 days).There was heterogeneity between all studies comparing haloperidol and ziprasidone in all subgroups (I 2 = 0-33%).Discontinuation due to lack of compliance (p = 0.01) was the only significantly different reason.Based on the reported RR, the type of APDs had a high impact on causing treatment discontinuation in the study population (Table 23).Girgis et al. [38] reported that 26% of the clozapine group and 10% of the chlorpromazine group remained on their assigned treatments for the duration of the study (p = 0.01).The median amount of time until first discontinuation was longer in the clozapine group (39 months) over the chlorpromazine group (23 months); the difference was significantly longer (p = 0.01).Reasons for discontinuation in the chlorpromazine group were withdrawn consent: 7.5%, side effects: 1.25%, lack of efficacy: 1.25%, death: 2.5%, did not attend follow up: 10% and imprisonment: 1.25%.Reasons for discontinuation in the clozapine group were withdrawn consent: 7.5%, side effects: 1.25%, death: 2.5%, lost to follow up: 8.75% and escape from the hospital: 1.25%.

Risk of Bias
Included were 21 randomised controlled studies to compare the APD outcomes with one another.Among the total of 21 studies, 11 studies remained at a low risk of bias.In addition, Stauffer et al. [42] showed participants were predominantly male; however, it was unclear whether this had affected the outcome of the study, which was considered under 'other' bias.Perkins et al.'s study [43] could not be included in the narrative or statistical analysis as it did not report data for the different APDs but rather the class in general.In addition, it was not possible to statistically analyse Lieberman et al. [32] and Girgis et al.'s [38] data for clozapine and chlorpromazine due to the lack of population size (Figure 15).

Review Limitations
This study has some limitations, which were mitigated to the minimum impact possible on the reported results.There was disparity in the way numerical data were reported; as such, in some occasions, a statistical analysis could not be performed.The number of measured outcomes and the scales they were measured on also varied between studies.The systematic review was time-limited and not a live review, as it is part of a degree not for a provider; as such, only studies up to 2022 were included in the analysis; updated systematic reviews are planned for the near future.In some subgroups, only two studies were available, which may have impacted the generalisation of the finding to a wider population.This systematic review did not consider confounders such as age, gender or ethnicity, which are known to possibly impact the severity of side effects and response to treatment and should be considered in future systematic reviews.Moreover, this study only included patients who were either APD-naïve or with short-term history of APD use (0-16 weeks); therefore, a future systematic review may consider investigating discontinuation reasons in patients with chronic psychosis.

Discussion
This systematic review gives a comprehensive view about the efficacy and tolerability of APDs in direct paired comparison.Twenty-one RCTs were included, and APDs were paired and compared.The fixed model analysis was used when the heterogeneity in all

Review Limitations
This study has some limitations, which were mitigated to the minimum impact possible on the reported results.There was disparity in the way numerical data were reported; as such, in some occasions, a statistical analysis could not be performed.The number of measured outcomes and the scales they were measured on also varied between studies.The systematic review was time-limited and not a live review, as it is part of a degree not for a provider; as such, only studies up to 2022 were included in the analysis; updated systematic reviews are planned for the near future.In some subgroups, only two studies were available, which may have impacted the generalisation of the finding to a wider population.This systematic review did not consider confounders such as age, gender or ethnicity, which are known to possibly impact the severity of side effects and response to treatment and should be considered in future systematic reviews.Moreover, this study only included patients who were either APD-naïve or with short-term history of APD use (0-16 weeks); therefore, a future systematic review may consider investigating discontinuation reasons in patients with chronic psychosis.

Discussion
This systematic review gives a comprehensive view about the efficacy and tolerability of APDs in direct paired comparison.Twenty-one RCTs were included, and APDs were paired and compared.The fixed model analysis was used when the heterogeneity in all subgroups was <30%, and the random model analysis was used when the heterogeneity was >30%.A minimum of two studies were required to conduct a statistical analysis.A statistical analysis was not possible for all short-term uses of APDs (0-12 weeks) based on the data available from the selected studies; accordingly, data from the included studies were narratively reported in Section 2 of this paper.Only three pairs of APDs made it possible to statistically compare using the mean difference to measure symptoms' improvement on one or more assessment tools (PANSS, CGI, BPRS, SANS, SAPS, CDSS and YMRS) after long-term use (>12 weeks.APDs were paired and compared based on their side effects; the use of medications from other classes such as hypnotics, anticholinergics or any others to manage those side effects, which allow the continuation of therapy (data reported in Section 3, part 3) and discontinuation rate and reasons (data reported in Section 3, part 4).A statistical analysis was not possible for some measured outcomes (side effects or concomitant medication used to manage them, discontinuation rate or reasons) due to the data available from the selected studies.Eight systematic reviews were identified to compare this systematic review's findings with their findings.Eight drugs were the most compared in those systematic reviews and were compared to the findings from this systematic review.Any data from this systematic review will be stated as a numerical summary or final outcome (p-value or total score) and labelled as 'this review' to avoid repeating the result sections stated previously in the results.Additionally, where comparison with the literature was made for one drug, e.g., olanzapine vs. risperidone, it was not repeated again under risperidone to prevent repetition.See S4.
Huhn et al. [45] [48], olanzapine (0.74 [0.67 to 0.81]) produced significantly more weight gain than all other drugs.Olanzapine (1.00 [0.73 to 1.33]) did not cause significantly more EPS than the placebo.Kishimoto et al. [50] reported sedation and/or somnolence with olanzapine was significantly less than clozapine (p < 0.001) and quetiapine, and olanzapine was significantly associated more with sedation than risperidone (p = 0.010).Parkinsonism symptoms showed to be significantly associated more with risperidone than olanzapine (p < 0.001).Zhang et al. [51] concluded that both olanzapine and risperidone increased weight significantly more than haloperidol (p < 0.01) in the short-term treatment analysis.It was also shown in both short-term treatment (p < 0.05) and long-term treatment (p < 0.001) that akathisia was less likely with olanzapine and risperidone than haloperidol.This study also reported that EPS was not as frequent with both olanzapine (p < 0.001) and risperidone (p < 0.001) compared with haloperidol.The authors reported that in the short-term analysis, patients on haloperidol required less anticholinergics (p < 0.001), benzodiazepines (p = 0.02) and beta-blockers (p < 0.01) compared to olanzapine.Zhu et al. [49] found that olanzapine was associated with a lower rate of parkinsonian symptoms than haloperidol (0.10 [0.03 to 0.29]) and risperidone (0.24 [0.07 to 0.78]).
Summary of findings: Based on the above findings, while limited to the population of the included studies, olanzapine improved symptoms in more patients (second to clozapine), was better tolerated except for weight gain and allowed more patient treatment continuation and for a longer period than other FGAs and other APDs except for clozapine, which is comparable to that reported in the reviewed systematic reviews.
Kishimoto et al. [50] reported that quetiapine involved significantly higher all-cause discontinuation compared with ziprasidone (p = 0.031).Zhang et al. [51] reported discontinuation due to patient decision and non-adherence, and only quetiapine showed a significantly lower rate than haloperidol (p < 0.05).Huhn et al. [45] reported that quetiapine, 0.85 (0.82 to 0.89), was higher than the placebo.
Summary of findings: Based on the above findings, while limited to the population of the included studies, quetiapine improved symptoms in more patients than those treated with FGAs but in less patients than those treated with clozapine, olanzapine, aripiprazole and ziprasidone.It was better tolerated except for weight gain and allowed more patient treatment continuation and for a longer period than other FGAs and other APDs except for clozapine, olanzapine, aripiprazole and ziprasidone, which is comparable to that reported in the reviewed systematic reviews.
Summary of findings: Based on the above findings, while limited to the population of the included studies, aripiprazole and ziprasidone were similar in improving symptoms in more patients than those treated with FGAs but to a lesser extent than that achieved with clozapine and olanzapine.They were better tolerated and allowed more patient treatment continuation and for a longer period than other FGAs and other APDs except for clozapine and olanzapine, which is comparable to that reported in the reviewed systematic reviews.).In the systematic review by Lieberman et al. [33], the authors found that only the SANS total score was statistically significantly different between clozapine and chlorpromazine after short-term use (p = 0.01), which is similar to Hartling et al.'s [47] systematic review where clozapine and chlorpromazine were statistically significantly different on the BPRS scale (p = 0.001).
Summary of findings: Based on the above findings, while limited to the population of the included studies, clozapine improved symptoms in more patients, was better tolerated except for weight gain and allowed more patient treatment continuation and for a longer period than other APDs, which is comparable to that reported in the reviewed systematic reviews.9.5.Amisulpride Huhn et al. [45] showed the overall reduction in symptoms was significantly more for amisulpride (−0.Huhn et al. [45] showed amisulpride (14.10 [7.71 to 20.45]) caused significantly more QTc prolongation than the placebo.Amisulpride (1.56 [0.91 to 2.23]) and aripiprazole (1.46 [1.11 to 1.83]) were higher than the placebo but less sedating than other APDs.Leucht et al. [48] reported that amisulpride (1.60 [0.88 to 2.65]) did not cause significantly more EPS than the placebo.Kahn et al. [37] also found that the total discontinuation rate in their amisulpride group was similar to that in ziprasidone (37%) and olanzapine (28%) groups, but lower than haloperidol (61%) and quetiapine (48%) groups.In Huhn et al.'s study [45], amisulpride involved significantly lower all-cause discontinuation than

Pharmacy 2023 ,
11,  x FOR PEER REVIEW 5 of 32 publications for inclusion in the systematic review.The authors individually screened all studies and in case of disagreement, it was resolved with consensus (S1 and S2).

1 .
Efficacy of Antipsychotics after Short-Term Treatment

Table 5 .
[31,35] of plotted RR for aripiprazole vs. risperidone[31,35].Columns 3 and 4: Prevalence of events; column 5: RR (95% CI)-when the estimated effect size or point estimate is >1, it indicates high impact, =1 indicates no impact, <1 indicates low impact; and column 6: Drug type impact on the patient experience of side effects. * Columns 3 and 4: Prevalence of events; column 5: RR (95% CI)-when the estimated effect size or point estimate is >1, it indicates high impact, =1 indicates no impact, <1 indicates low impact; and column 6: Drug type impact on the patient experience of side effects. * Columns 3 and 4: Prevalence of events; column 5: RR (95% CI)-when the estimated effect size or point estimate is >1, it indicates high impact, =1 indicates no impact, <1 indicates low impact; and column 6: Drug type impact on the patient experience of side effects. *
At week 52, San et al.
* Columns 3 and 4: Prevalence of events; column 5: RR (95% CI)-when the estimated effect size or point estimate is >1, it indicates high impact, =1 indicates no impact, <1 indicates low impact; and column 6: Drug type impact on the patient experience of side effects.