Efficacy and tolerability of aripiprazole versus D2 antagonists in the early course of schizophrenia: a systematic review and meta-analysis

Early intervention is essential for favorable long-term outcomes in schizophrenia. However, there is limited guidance in the scientific literature on how best to choose between dopamine D2 receptor (D2R) partial agonists and D2R antagonists in early stages of schizophrenia. The aim of this meta-analysis was to directly compare D2R partial agonists with D2R antagonists for efficacy and tolerability, using randomized controlled trials (RCTs) that involved participants diagnosed with first-episode psychosis, schizophrenia, or related psychotic disorders with a duration of illness ≤5 years. Fourteen RCTs, involving 2494 patients, were included in the meta-analysis. Aripiprazole was the only identified D2R partial agonist, and was not significantly different from pooled D2R antagonists for overall symptom reduction or all-cause discontinuation. However, aripiprazole was more favorable than pooled D2R antagonists for depressive symptoms, prolactin levels, and triglyceride levels. Specifically, aripiprazole was more favorable than paliperidone for triglyceride levels and more favorable than risperidone and olanzapine, but less favorable than ziprasidone, for weight gain. In addition, aripiprazole was less favorable for akathisia compared with second-generation D2R antagonists, in particular olanzapine and quetiapine, and less favorable for discontinuation due to inefficacy than risperidone. Lastly, aripiprazole was more favorable than haloperidol for various efficacy and tolerability outcomes. In conclusion, aripiprazole’s efficacy did not differ substantially from D2R antagonists in the early course of schizophrenia, whereas differential tolerability profiles were noted. More double-blind RCTs are required comparing the efficacy and tolerability of aripiprazole as well as other D2R partial agonists with D2R antagonists in early stages of schizophrenia.


INTRODUCTION
Schizophrenia is a debilitating psychiatric disorder that affects 0.87% of the general population over a lifetime 1 . The onset of schizophrenia is typically between the ages of 14 and 35 2 , and the disease is associated with a reduction in lifespan by approximately 16.3-18.7 years compared with the general population 3 . Recent research shows that antipsychotic treatment is associated with a lower mortality risk compared with no treatment during follow-up periods 4 . This illustrates the importance of utilizing antipsychotics in patients with schizophrenia, but how these medications should be used for optimal outcomes requires further research.
First-episode psychosis (FEP) refers to the first time an individual experiences psychotic symptoms, which gives significant distress, confusion, and fear to many individuals 5 . There is well-established evidence that the duration of untreated psychosis is negatively associated with long-term outcomes, indicating that intervening early and effectively is important in FEP 5 . The literature suggests that an illness duration of less than 5 years is considered an acceptable period of time to define the early stage of schizophrenia 6 . Thus, a person who has FEP, or is in an early stage of schizophrenia, should be provided with evidence-based pharmacotherapy in a timely manner.
Current treatment guidelines recommend second-generation antipsychotics, including risperidone, quetiapine, olanzapine, and aripiprazole, as first-line treatment for schizophrenia 7 . Aripiprazole's pharmacology is distinct from most antipsychotics in that it is a dopamine D 2 receptor (D 2 R) partial agonist, rather than a full D 2 receptor antagonist 8 . Owing to this difference in pharmacology, D 2 R partial agonists (including brexpiprazole and cariprazine) have also been referred to as third-generation antipsychotics 8 . To date, there is limited evidence as to how D 2 R partial agonists differ from D 2 R antagonists in the early course of schizophrenia. The most recent network meta-analysis of the efficacy and tolerability of antipsychotics in FEP included 19 randomized controlled studies (RCTs), of which only one RCT involved a D 2 partial agonist, aripiprazole 9 .
The small number of RCTs in the literature that compared D 2 partial agonists with D 2 antagonists in FEP led us to conduct a systematic review that does not restrict the target population to FEP but also includes all individuals in the early course of their disease, using the evidence-based definition (i.e., a duration of illness less than 5 years) 6 . The objective of our study was to conduct a metaanalysis comparing the efficacy and tolerability of D 2 R partial agonists with D 2 R antagonists in the early course of schizophrenia.
When only blinded RCTs were considered (4 double-blind, 2 single-blind), aripiprazole remained more favorable than D 2 R antagonists for depressive symptoms (p = 0.03), prolactin levels (p = 0.008), and triglyceride levels (p = 0.002). In addition, aripiprazole was more favorable than D 2 R antagonists for total cholesterol (p < 0.001) and glucose levels (p = 0.001), and less favorable than D 2 R antagonists for discontinuation due to inefficacy (p = 0.02). Including only the double-blind RCTs led to aripiprazole being more favorable than D 2 R antagonists for prolactin (p = 0.008), triglyceride (p = 0.03), total cholesterol (p < 0.001), and glucose levels (p = 0.001), and less favorable for discontinuation due to inefficacy (p = 0.02). Restricting the analysis to open-label RCTs (N = 7 trials) led to aripiprazole being less favorable than secondgeneration D 2 R antagonists for akathisia (p = 0.02) and anticholinergic use (p < 0.001), whereas no significant difference emerged for blinded RCTs.

Aripiprazole versus individual D 2 R antagonists
Comparisons of aripiprazole with each of the D 2 R antagonists that led to significant results and were based on at least two comparisons are summarized in Table 3. Aripiprazole was associated with larger reductions in overall and negative symptoms than haloperidol and in depressive symptoms than haloperidol and risperidone. Aripiprazole was more favorable for discontinuation due to any cause and adverse events than haloperidol, but less favorable for discontinuation due to inefficacy than risperidone. Aripiprazole was more favorable for metabolic adverse effects than risperidone (weight gain), paliperidone (triglyceride levels), and olanzapine (weight gain), but less favorable than ziprasidone (weight gain). Aripiprazole was less favorable for extrapyramidal side effects than quetiapine (akathisia) and olanzapine (akathisia and use of anticholinergics).
Studies comparing aripiprazole with risperidone were sufficient in number to be analyzed in short-term and long-term trials. For overall symptom reduction, aripiprazole did not significantly differ from risperidone in short-term trials (N = 7 trials, n = 732, SMD = 0.13, 95% CI = −0.13 to 0.39, p = 0.34), but demonstrated significantly greater efficacy than risperidone in long-term trials (N = 3 trials, n = 198; SMD = −0.74, 95% CI = −1.25 to −0.24, p = 0.004), and a significant difference was found between the two durations (χ 2 = 6.44, p = 0.01). For positive symptom reduction, aripiprazole was significantly less efficacious than risperidone in short-term trials (N = 4 trials, n = 444; SMD = 0.25, 95% CI = 0.06 to 0.44, p = 0.009), but was not significantly different in long-term trials (N = 2 trials, n = 158; SMD = −0.40, 95% CI = −0.82 to 0.02, p = 0.06), and no significant subgroup difference was found between short-term and long-term trials. For negative symptom reduction, no significant difference between aripiprazole and risperidone was found in neither short-term nor long-term trials, and no subgroup difference between the two durations was noted. For all-cause discontinuation and discontinuation due to adverse events, no significant difference between aripiprazole and risperidone was found in neither short-term nor long-term trials, and no subgroup difference between the two durations was noted. For discontinuation due to inefficacy, aripiprazole was significantly less favorable than risperidone in short-term trials (N = 2 trials, n = 410; RR = 1.77, 95% CI = 1.13 to 2.76, p = 0.01), but was not significantly different in long-term trials (N = 2, n = 482; RR = 1.66, 95% CI = 0.85 to 3.24, p = 0.14), and no subgroup difference between the two durations was noted.

Heterogeneity
Heterogeneity (I 2 > 50%) was present for overall symptom reduction, positive symptom reduction, discontinuation due to adverse events and inefficacy, use of anticholinergics, incidence of akathisia, weight gain, total cholesterol levels, triglyceride levels, fasting glucose levels, prolactin levels, and incidence of sedation ( Table 2). The short-term study (i.e., 8 weeks) that compared aripiprazole with risperidone and olanzapine 22 and the long-term study (i.e., 3 years) that compared aripiprazole with risperidone, quetiapine, olanzapine, ziprasidone, and haloperidol 23 were each source of heterogeneity for positive symptom reduction, and removal of each study did not change the significance of the result. The long-term study (i.e., 3 years) that compared aripiprazole with quetiapine and ziprasidone 29 was the source of heterogeneity for triglyceride levels, and removal of the study did not change the significance of the result. The long-term study (i.e., 1 year) that compared aripiprazole with paliperidone and ziprasidone 11 was the source of heterogeneity for glucose levels, and removal of the study led to aripiprazole being significantly D.D. Kim et al.  more favorable than D 2 R antagonists (SMD = −0.21, 95% CI = −0.34 to −0.07). The long-term study (i.e., 1 year) that compared aripiprazole with haloperidol 10 was the source of heterogeneity for incidence of akathisia, and removal of the study led to aripiprazole being significantly less favorable than D 2 R antagonists (RR = 1.39, 95% CI = 1.08 to 1.80). The long-term study (i.e., 3 years) that compared aripiprazole with risperidone, quetiapine, olanzapine, ziprasidone, and haloperidol 23 and the long-term study (i.e., 1 year) that compared aripiprazole with risperidone and olanzapine 33 were each source of heterogeneity for discontinuation due to inefficacy, and removal of the former study led to aripiprazole being less favorable than D 2 R antagonists (RR = 2.25, 95% CI = 1.24 to 4.09). The short-term study (i.e., 12 weeks) that compared aripiprazole with quetiapine 19 and the long-term study (i.e., 3 years) that compared aripiprazole with risperidone, quetiapine, olanzapine, ziprasidone, and haloperidol 23 were each source of heterogeneity for incidence of sedation, and removal of each study did not change the significance of the result. No single study was the source of heterogeneity for overall symptom reduction, discontinuation due to adverse events, use of anticholinergics, weight gain, total cholesterol levels, and prolactin levels.

Meta-regression
In our meta-regression analysis, associations of covariates with overall symptom reduction (n = 14 studies) and all-cause discontinuation (n = 11 studies) were examined (Supplementary Table 1). Aripiprazole's effect on overall symptom reduction relative to D 2 R antagonists was positively associated with trial duration (estimate = −0.01, p = 0.020) and baseline symptom severity (estimate = −0.02, p = 0.041). None of the covariates demonstrated significant associations with all-cause discontinuation.

Publication bias
We assessed publication bias for overall symptom reduction (n = 14 studies) and all-cause discontinuation (n = 11 studies) (Supplementary Fig. 3). Egger's test did not indicate substantial asymmetry in the funnel plots for both outcomes (p = 0.546 for overall symptom reduction and p = 0.917 for all-cause discontinuation).

DISCUSSION
Our meta-analysis directly compared aripiprazole with D 2 R antagonists for efficacy and tolerability in the early course of schizophrenia. Aripiprazole was the only identified D 2 R partial agonist and was compared with various D 2 R antagonists, including risperidone, paliperidone, quetiapine, olanzapine, ziprasidone, perospirone, and haloperidol. Our results indicate that aripiprazole's efficacy was comparable to D 2 R antagonists (as a group), however; it was more favorable in terms of depressive symptoms, triglyceride levels, and prolactin levels. When stratified according to trial duration, aripiprazole was additionally more favorable for total cholesterol levels (short-term), glucose levels (short-term), and sedation (long-term), but was less favorable for akathisia (short-term). These results are largely consistent with existing evidence that aripiprazole has antidepressant effects and a lower tendency to cause hyperprolactinemia, metabolic dysregulation, and sedation [34][35][36][37][38] .
Several mechanisms of action that are unique to aripiprazole may explain such results. Aripiprazole's D 2 partial agonistic properties have prolactin-sparing effects 34,35 . Its partial agonistic activity at postsynaptic 5-HT 1A and 5-HT 2C receptors, coupled with desensitization of presynaptic 5-HT 1A receptors, would lead to an increased serotonergic tone, which may play a role in improving depressive and anxiety disorders 38,39 . Aripiprazole's relatively  lower affinity for H 1 receptors compared with other secondgeneration antipsychotics may explain its lower propensity to induce sedation 35,36 . Aripiprazole's lack of anticholinergic effects, relatively modest antagonism of H 1 receptors, and partial agonistic activity at D 2 and 5-HT 1A receptors would explain its favorable metabolic profiles compared with other second-generation antipsychotics, apart from ziprasidone 37 . These results are consistent with existing evidence that both aripiprazole and ziprasidone, have favorable metabolic profiles relative to many other antipsychotics 37 .
It should be noted that aripiprazole (in short-term trials) was more frequently associated with akathisia compared with D 2 antagonists, especially quetiapine and olanzapine. This is consistent with a recent network meta-analysis involving patients with FEP, where aripiprazole was less favorable than quetiapine and olanzapine for akathisia 9 . Nevertheless, aripiprazole was not associated with higher discontinuation due to adverse events than D 2 R antagonists, including quetiapine and olanzapine, indicating that the severity of akathisia may have been tolerable. Aripiprazole-induced akathisia may be attributed to its proserotonergic effects as well as its functional selectivity for D 2 receptors. In the case of the latter, aripiprazole may be acting as a full antagonist in certain brain regions (e.g., striatum) where there are higher levels of D 2 receptor expression 35,40 .
Differential results were found compared with a recent network meta-analysis involving patients with multi-episode, chronic schizophrenia 41 . We found that compared with haloperidol, aripiprazole was more favorable in terms of overall, negative, and depressive symptoms and discontinuation due to any cause and adverse events. This is consistent with a recent network metaanalysis involving patients with FEP that found haloperidol to be less efficacious than second-generation antipsychotics 9 . However, aripiprazole's efficacy was not significantly different from haloperidol's in patients with multi-episode, chronic schizophrenia 41 . Also, we found that aripiprazole was less favorable than risperidone in terms of discontinuation due to inefficacy, which is consistent with the finding in patients with multi-episode, chronic schizophrenia where risperidone was more efficacious than aripiprazole 41 .
Such differential results between FEP and multi-episode, chronic schizophrenia may be explained by dopamine supersensitivity, which is a key predictor of poor long-term outcomes in schizophrenia 42 . Individuals with multi-episode, chronic schizophrenia are likely to be exposed to chronic D 2 receptor blockade, increasing the propensity to develop dopamine supersensitivity 42 . Since dopamine supersensitivity is associated with upregulation of D 2 receptors, D 2 R antagonists with high affinities for the D 2 receptor may be more effective than antipsychotics with lower affinities to treat the related psychosis, with the exception of clozapine that has a lower affinity but has the potential to reverse dopamine supersensitivity 43 . Although aripiprazole has a very high affinity for the D 2 receptor (0.34 nM), its action as a partial agonist would predict worsening of psychosis in an environment of dopamine supersensitivity. This is consistent with reports showing that switching patients with chronic schizophrenia to aripiprazole can be less effective or even lead to psychotic worsening 44 .
However, it should be noted that aripiprazole has been shown to prevent and reverse dopamine supersensitivity and thus may be more effective over the long term 45 . In support of this, our sensitivity analysis found that aripiprazole, although less favorable than risperidone for discontinuation due to inefficacy over a short term, was more efficacious than risperidone over a longer term. Although more studies are required to replicate this finding, the data from animal studies and our preliminary analysis may provide a rationale for using aripiprazole prior to D 2 R antagonists in the treatment of FEP.
Our meta-analysis has several limitations to be considered. First, the number of studies included in our meta-analysis was relatively small. Second, many of the included studies were open-label RCTs, which could have increased the risk of performance and detection bias. Third, aripiprazole was the only identified dopamine D 2 R partial agonist. Due to differences in pharmacological activity even among dopamine D 2 R partial agonists (aripiprazole, brexpiprazole, and cariprazine), there needs to be caution when extrapolating the results of our meta-analysis to D 2 R partial agonists other than aripiprazole. Fourth, heterogeneity was present in various outcomes in our meta-analysis. This was expected as D 2 R antagonists have unique receptor profiles and thus may have varying efficacy and tolerability profiles. However, we identified sources of heterogeneity for the majority of the outcomes and addressed this limitation via subgroup, sensitivity, and meta-regression analyses. Lastly, although multiple outcomes were examined in our review, multiple comparisons were not adjusted for. This may have increased the risk of type 1 error. However, given the relatively small number of studies for each outcome, such a statistical adjustment may be more appropriate when more studies become available in the future.
In conclusion, aripiprazole's efficacy did not differ substantially from D 2 R antagonists in the early course of schizophrenia, whereas it demonstrated greater antidepressant effects than D 2 R antagonists. Differential tolerability profiles were noted between aripiprazole and D 2 R antagonists, where aripiprazole appeared to have general advantages regarding prolactin and triglyceride levels over D 2 R antagonists, but may induce more akathisia. Evidence is still limited to draw strong conclusions. More doubleblind RCTs are required to better understand the relative effects of aripiprazole and other D 2 R partial agonists in the early course of schizophrenia.

Search strategy
The systematic review and meta-analysis were conducted according to the guidelines of Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2009 46 . Pre-registration with the International Prospective Register of Systematic Reviews was not undertaken for our review, but we ensured that no duplicate review is ongoing to date. We searched for RCTs comparing D 2 R partial agonists with D 2 R antagonists in the early course of schizophrenia that were published up to, and including 15 July 2020, using the MEDLINE, EMBASE, and PsycINFO, and Clinical-Trials.gov databases. The following search terms and associated MeSH terms were used: (1) ['aripiprazole' OR 'brexpiprazole' OR 'cariprazine'] AND (2) [('first episode' OR 'first-episode' OR '$naive' OR 'early*' OR 'recent*' OR 'prodrom*' OR 'youth*' OR 'adolescen*' OR 'child*') AND ('schizophreni*' OR 'schizophrenia spectrum' OR 'psychosis')] AND (3) [('random*') OR (('random*') AND ('comparative' OR 'comparison*' OR 'compare*' OR 'versus' OR 'vs*' OR 'open*')) OR ('$blind*')]. We also manually searched the reference lists of all relevant retrieved articles for potential studies eligible for inclusion in our analysis. Two authors (D.D.K. and L.L.) independently screened for relevant articles and any discrepancy was resolved following a discussion between the two authors or with R.M.P. Corresponding authors of the included studies were contacted for any missing data.

Selection criteria
Studies were included in the quantitative analysis if they enrolled patients who met the diagnostic criteria for schizophrenia or schizophrenia-spectrum disorders and were in their first episode (as identified by authors) or were in the early course of their illness. We defined early stages of illness as follows: (1) FEP, 2) antipsychotic-naïve, and (3) ≤5 years of mean duration of illness 6 . We placed no restriction on language, blinding, publication year, age, sex, ethnicity, settings, or trial duration.

Data extraction
The following data were extracted: publication year, sample size, age, sex, duration of illness, trial duration, doses of antipsychotics used, baseline symptom severity, change/endpoint scores for overall, positive, negative, and depressive symptoms, incidence of akathisia, sedation, discontinuation due to any cause, adverse events, and inefficacy, use of medications to treat extrapyramidal symptoms (EPS) (i.e., anticholinergics), and changes in body mass index and levels of prolactin, total cholesterol, triglycerides, and fasting glucose.
Mean antipsychotic doses were converted to chlorpromazine equivalents according to methods provided elsewhere 47,48 . Change scores measured using the Positive and Negative Syndrome Scale (PANSS) were the primary efficacy outcome 49 . Other validated scales were considered if the PANSS was not available. If change scores were not available, authors were contacted for the data. Endpoint scores were used if authors did not respond to our request. D.D.K. and L.L. reviewed included studies and supplementary materials, extracted relevant data, and assessed risk of bias using the Cochrane Risk of Bias tool 50 .

Data analysis
The meta-analysis was performed using Cochrane Review Manager (version 5.4). Using random-effects models, aripiprazole was compared with D 2 R antagonists (as a group) and also with individual D 2 R antagonists. Standardized mean differences (SMDs) and risk ratios (RRs) were calculated for continuous and dichotomous variables, respectively, with 95% confidence intervals (CIs). SMDs less than 0 and RRs less than 1 indicated that aripiprazole was favored compared with D 2 R antagonists. An inverse variance method was used according to the Cochrane guideline when dichotomous and continuous variables needed to be combined. When necessary, online tools were used to calculate an effect size from F or t statistics or to combine multiple means and standard deviations 51 . Study heterogeneity was quantified using the I 2 statistic, where I 2 > 50% was considered substantial heterogeneity. If a study provided data for multiple timepoints, our main analysis included the study's planned duration. A secondary analysis was performed using the short-term (i.e., <6 months) and long-term (i.e., ≥6 months) data separately.
For outcomes that involved at least 10 studies, meta-regression analysis was performed and publication bias was assessed using R 52 . For the meta-regression analysis, we examined the effects of the following covariates: sample size, study year, age, trial duration, aripiprazole dose, baseline symptom severity, proportion of open-label studies, male participants, studies that included FEP patients, and studies that utilized risperidone or olanzapine. The rationale for including the proportion of risperidone or olanzapine as a covariate was based on a recent meta-analysis that has shown that risperidone and olanzapine tend to be more efficacious than other antipsychotics in multi-episode, chronic schizophrenia 41 . Baseline symptom scores measured on scales other than the PANSS were standardized using methods provided elsewhere 53 . Publication bias was assessed using funnel plots, Egger's regression test, and trim-and-fill procedure 54 .

DATA AVAILABILITY
The manuscript reports meta-analytic data based on original studies. Extracted data are available upon request.