FormalPara Key Points

This systematic review indicated that once-monthly paliperidone palmitate (PP1M) may be an effective and tolerable option for managing patients with schizophrenia who were previously on oral antipsychotics (OAPs).

Our findings suggest PP1M may reduce the severity of disease and symptoms as well as improve functioning states in patients with schizophrenia with OAP history.

1 Introduction

Schizophrenia is a chronic, serious mental disorder characterised by symptoms such as delusions, hallucinations, disorganised behaviour, emotional withdrawal, and progressive cognitive deficits [1]. It affects approximately 24 million people or 0.32% of the population worldwide [2]. Although schizophrenia is not as common as many other mental disorders, the severity of the disability and the unclear aetiology makes it the 11th leading causes of disability worldwide [3]. Patients with schizophrenia have a 15- to 25-year shorter life expectancy than those without the disorder [4].

Schizophrenia was typically treated with oral antipsychotics (OAPs). The use of second-generation antipsychotics (SGAs) (e.g., paliperidone and risperidone) was more prevalent than first-generation antipsychotics (FGAs) (e.g., chlorpromazine and haloperidol). However, the outcomes associated with OAPs were often suboptimal. For various reasons, approximately one-third of patients with schizophrenia did not adhere to their OAP treatment regimen [5]. To improve adherence, long-acting injectable antipsychotics (LAI) were proposed as an effective tool [6], although they remain underutilised in clinical practice [7, 8]. Several studies suggested LAIs may offer an advantage in increasing patient adherence and reducing relapse and rehospitalisation when compared with OAPs [9,10,11]. Guidelines from the American Psychiatric Association and other organisations have recommended LAIs as a treatment alternative to OAPs for patients with poor or uncertain adherence as well as for patients transitioning from inpatient to outpatient status [12,13,14].

Several studies have demonstrated once-monthly paliperidone palmitate (PP1M), an SGA LAI, is associated with improvements in psychotic symptoms, disease severity, and functional outcomes in patients with newly diagnosed or acute schizophrenia [15, 16]. Although some studies have reported benefits of PP1M—namely improved functioning and symptom severity—for patients who have not responded to prior administration of OAPs [17, 18], there has been a lack of a systematic analysis and synthesis of the data. Thus, we performed a systematic review and meta-analysis to assess the efficacy and safety of PP1M in patients switching from OAP use with the aim of providing evidence that may be used to guide clinical practice.

2 Methods

This study followed the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA 2020) extension statement [19]. A protocol was registered in the PROSPERO database (CRD42022363726).

2.1 Data Sources and Literature Search

An electronic database search of PubMed, EMBASE, and the Cochrane Library was conducted on 19 July 2022 using the following keywords: ‘oral antipsychotics’, ‘Schizophrenia’, ‘LAI’, ‘Paliperidone Palmitate’, and ‘1 month’. We applied no restrictions on language, date, or publication status. The reference lists of the included studies were searched to identify further eligible studies. The search strategy is provided in Online Resource 1.

2.2 Study Selection

Studies were included if they met the following criteria:

Participants patients with a diagnosis of schizophrenia—as defined by the original study—who had a history of receiving OAPs. For instance, schizophrenia was defined by the International Statistical Classification of Diseases and Related Health Problems-10th revision (ICD-10) [20] or Diagnostic and Statistical Manual of Mental Disorders-Fourth/Fifth Edition (DSM-IV/DSM-5) [21, 22]. There were no restrictions on the age of the patient.

Intervention patients with a prior history of OAP use and who were currently switching to PP1M.

Comparator there were no limitations on comparators as we only pooled data from single arms of included studies.

Outcomes primary outcomes included patient relapse rate, hospitalisation rate, and the change from baseline in the Positive and Negative Syndrome Scale (PANSS) total score. The secondary outcomes included the changed number of inpatient visits, changed length of hospital stays, change from baseline in the Clinical Global Impressions-Severity (CGI-S) score and the Personal and Social Performance (PSP) total score, response rate defined as at least 20%, 30% or 40% reduction in the PANSS score, proportion of treatment discontinuation, total Treatment Emergent Adverse Events (TEAEs), and metabolic-related TEAEs.

Study design randomised controlled trials (RCTs), single-arm interventional trials, and observational studies.

To limit the effect of LAIs, we excluded patients with a history of previous treatment with a LAI that was reported at baseline. For intervention, current treatment with oral antipsychotics was considered ineligible. Case reports, case series, and reviews were excluded. We included studies that reported findings in English only. Two reviewers (XL and MJ) independently screened articles for eligibility. Any disagreements were resolved through discussion with a third reviewer (CY).

2.3 Data Extraction

Two reviewers (XL and CY) extracted the following data from each study into a pre-defined data extraction form: trial characteristics (i.e., country, study design, centre, NCT number, and sample size), participant characteristics (i.e., age, body mass index, sex, prior oral antipsychotics, stages, and ethnicity), baseline data (i.e., number of inpatient visits, PANSS score, CGI-S score, and PSP score), and outcome. Any disagreements were resolved by discussion with a third reviewer (MJ).

2.4 Quality Assessment

The quality of included studies was assessed independently by two reviewers (XL and CY). The Revised Cochrane risk-of-bias tool for randomised trials (RoB2) was employed to assess the risk of bias in RCTs [23]. The 9-point Newcastle-Ottawa Scale (NOS) instrument that contains 8 items was employed for non-randomised studies and cohort studies [24]. For before and after trials, we used the 12-item National Institutes of Health (NIH) quality assessment tool for before-after (Pre-Post) study without control group [25]. Any disagreements were resolved by discussion with a third reviewer (MJ).

2.5 Data Synthesis and Analysis

The R software (version 4.1.0) was used for meta-analyses (two tails with an α = 0.05 unless indicated). For dichotomous outcomes, we extracted the reported rates and calculated data as the proportion and 95% confidence interval (CI) for each study, and then pooled them by meta-analysis. When no event was observed, we added a fixed value (typically 0.5) to the event number of the intervention group following methods described in the Cochrane handbook [26]. For continuous outcomes, we extracted the means and standard deviations (SDs) of the changed score for each study and calculated a pooled changed mean and 95% CI. When the changed SD was not observed, we calculated it from the reported endpoint and baseline value [27]. Also, we used baseline SD when the changed or endpoint SD was not available. For those studies reporting no available data, we imputed SD from another study by considering whether the studies used the same measurement scale, had the same degree of measurement error, had the same time interval between the baseline and post-intervention measurement as well as a similar population [27]. When data were insufficient for meta-analysis, we described the outcome narratively. We divided the follow-up time into short- (≤ 13 weeks), medium (14–26 weeks) and long term (≥ 27 weeks) according to a previously published Cochrane review [28].

A fixed-effects model was used for meta-analyses. Where substantial heterogeneity was identified—defined as I2 ≥ 50% coupled with p < 0.1 from Q test [28, 29]—we explored the sources of heterogeneity from the following aspects: baseline PANSS score, previous concomitant medications, and number of previous hospitalisations. We synthesised data using a random-effects model with restricted maximum likelihood (REML)-based estimation of between-study variance tau2 when heterogeneity was significant, and the source was not identified. A funnel plot used for assessing publication bias was not performed due to the insufficient number of studies for each outcome (i.e., less than 10 studies).

Subgroup analyses were carried out on primary outcomes according to the OAP treatment history, ethnicity (i.e., Asian vs non-Asian), and the stage of schizophrenia at which switching was initiated, where possible. Sensitivity analysis was conducted on meta-analyses including data transformation or imputation to confirm the robustness of the conclusion.

3 Results

3.1 Results of Study Selection

The database search returned 2198 results and provided 1467 unique citations after duplicates were removed. A further 795 articles were excluded at the title-abstract screening stage. Full papers were sought for the remaining 672 articles. Of these, 628 full papers were retrieved and assessed for eligibility according to the inclusion and exclusion criteria. The remaining 44 articles were not successfully retrieved due to unauthorised access. Among the 628 full papers, 602 studies were excluded: 99 studies were excluded due to ineligible patients; 419 studies were excluded due to ineligible intervention; 20 studies were excluded due to ineligible outcomes; 48 studies were excluded due to ineligible study design, and 16 studies were excluded because they were duplicates. Finally, 15 studies with 26 references [17, 18, 20,21,22, 30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50] met full eligibility criteria and were included in the review. No additional eligible studies were identified by manual searching. The study selection process is shown in Fig. 1.

Fig. 1
figure 1

The PRISMA flow chart of study selection. PRISMA preferred reporting items for systematic reviews and meta-analyses, PP1M once-monthly paliperidone palmitate

Full size image

3.2 Characteristics of Included Studies and Participants

The included 15 studies comprised four RCTs (26.7%) [21, 31, 38, 44], 2 cohort studies (13.3%) [33, 50], and 9 pre-post studies (60.0%) [20, 22, 30, 32, 34, 45, 46, 48, 49]. Of these, 3 studies were from the USA [20, 34, 49], 2 from Italy [21, 33], 2 from China [30, 48], 1 from Croatia [32], 1 from Romania [50], and 6 were from multiple countries [22, 31, 38, 44,45,46]. The sample size ranged from 10 to 1169 (Table 1). Across the included studies, a total of 4740 patients were included. The mean age ranged from 28.7 to 52.0 years. Except for 2 studies [21, 50] that did not report sex information, the remaining 13 studies reported that 67.9% of included patients were male. The follow-up period ranged from 1 week to 18 months. The most reported OAPs previously used were risperidone and paliperidone. Others including amisulpride, olanzapine, haloperidol, fluphenazine, ziprasidone, quetiapine, aripiprazole, sertindole, chlorpromazine, penfluridol, perphenazine, and clozapine. Detailed patient characteristics are shown in Table 2.

Table 1 Characteristics of included studies
Full size table
Table 2 Characteristics of included patients
Full size table

3.3 Relapse Rate and Hospitalisation Rate

One study with 157 patients reported 19 of 157 patients (12%) had relapsed by the long-term follow-up (95% CI 0.07–0.18) [38]. The long-term pooled hospitalisation rate of two studies was 18% (95% CI 0.15–0.20, I2 = 34%) with a total of 731 patients [20, 22].

3.4 Change in the PANSS Total Score from Baseline

Five studies (N = 1390 patients) reported a change in the PANSS total score at the short-term follow-up that ranged from − 30.87 to − 15.40. The pooled mean change was − 21.69 (95% CI − 30.02 to − 13.36, I2 = 99%, Fig. 2a) [30, 31, 33, 38, 46]. Two studies (N = 1041 patients) reported a change in the total score at the medium-term follow-up ranging from − 18.30 to − 11.70, and the pooled mean change was − 14.98 (95% CI − 21.45 to − 8.51, I2 = 97%, Fig. 2b) [38, 45]. Two studies (N = 533 patients) reported a change in the total score at the long-term follow-up that ranged from − 25.70 to − 11.30. The pooled change was − 17.88 (95% CI − 31.94 to − 3.82, I2 = 91%, Fig. 2c) [22, 33]. Results at Weeks 1, 5, 9, 12 and 13 are provided in Online Resource 2 Figure S1.

Fig. 2
figure 2

The single arm meta-analysis plot of change from baseline in the PANSS total score at short-, medium-, and long-term periods. PANSS Positive and Negative Syndrome Scale

Full size image

3.5 Change in the CGI-S Score from Baseline

Three studies (N = 926 patients) reported a reduction in the CGI-S score at the short-term follow-up that ranged from 0.99 to 1.84 from baseline [30, 31, 46]. The pooled change was − 1.42 (95% CI − 1.89 to − 0.96, I2 = 95%, Fig. 3a). Two studies (N = 618 patients) reported a reduction in the score at the medium-term follow-up that ranged from − 0.74 to − 0.60. The pooled change was − 0.61 (95% CI − 0.69 to − 0.53, I2 = 0%, Fig. 3b) [21, 45]. Another two studies (N = 633 patients) reported a reduction in the score at the long-term follow-up that ranged from − 2.60 to − 0.80. The pooled change was − 1.70 (95% CI − 3.46 to − 0.06, I2 = 100%, Fig. 3c) [22, 32]. Results at Weeks 1, 5, and 9 are provided in Online Resource 2 Figure S2.

Fig. 3
figure 3

The single arm meta-analysis plot of change from baseline in the CGI-S score at the short-, medium- and long-term periods. CGI-S Clinical Global Impression-Severity

Full size image

3.6 Change in the PSP Score from Baseline

Improvement in PSP score at the short-term follow-up was reported in five studies (N = 1413 patients) and ranged from 5.8 to 19.34 [30, 31, 33, 41, 46]. The pooled change was 13.99 (95% CI 6.94–21.05, I2 = 98%, Fig. 4a). Three studies (N = 1129 patients) reported an improvement of 8.00–12.41 score at the medium-term follow-up [21, 41, 45]. The pooled mean change was 9.27 (95% CI 7.31–11.23, I2 = 73%, Fig. 4b). Another two studies (N = 519 patients) reported an increase in the score at the long-term follow-up that ranged from 10.50 to 13.00 [33, 41]. The pooled result after meta-analysis was 10.69 (95% CI 9.05–12.33, I2 = 0%, Fig. 4c). Results at Week 5, 12, and 13 are provided in Online Resource 2 Fig. S3.

Fig. 4
figure 4

The single arm meta-analysis plot of change from baseline in the PSP score at the short-, medium-, and long-term periods. PSP Personal and Social Performance

Full size image

3.7 Response Rate—at Least 20%, 30% or 40% Reduction in the PANSS Score from Baseline

Response rate was defined as at least a 30% and 40% reduction in the PANSS score (Online Resource 2 Figure S4). The pooled response rate at the short-term follow-up was 57% (95% CI 0.32–0.82, I2 = 99%, 3 studies, N = 1341 patients) [22, 30, 46] when defined as at least a 30% reduction and 53% (95% CI 0.46–0.60, one study, N = 210 patients) [46] when defined as at least a 40% reduction in the baseline PANSS score. The medium-term response rate was 51% (95% CI 0.47–0.56, one study, N = 589 patients) [45] and 30% (95% CI 0.27–0.34) [45], respectively. Only one study (N = 521 patients) reported a response rate-defined by at least a 30% reduction in the score of 74% at the long-term follow-up (95% CI 0.70–0.78) [22]. We did not identify studies that reported a response rate defined as at least a 20% reduction in the PANSS score. Results at Weeks 1, 5, 9, 13 are provided in Online Resource 2 Figure S5.

3.8 Changes in Length of Stay Hospitalisation and Number of Inpatient Visits from Baseline

The length of stay hospitalisation and number of inpatient visits at the medium-term follow-up were reported in two studies (N = 846 patients) [34, 49]. The pooled result showed a reduction in hospital stay of 5.45 days (95% CI − 14.77 to 3.87, I2 = 100%). The pooled change in the number of inpatient visits was − 0.79 (95% CI − 1.27 to − 0.31, I2 = 100%). Four studies (N = 558 patients) reported the long-term follow-up results of these two outcomes [20, 32, 48, 49]. The pooled reduction of length of stay was 12.98 days (95% CI − 18.15 to − 5.47, I2 = 97%), and the pooled change in number of inpatients visits was − 0.79 (95% CI − 1.27 to − 0.31, I2 = 98%). See Online Resource 2 Figure S6 for details.

3.9 Discontinuation Rate

Short term, the overall pooled weighted discontinuation rate was 31% (95% CI 0.21–0.41, I2 = 86%, three studies, N = 926 patients) [30, 31, 46] due to all-cause, 4% (95% CI 0.02–0.06, I2 = 54%, three studies, N = 932 patients) [30, 31, 39] due TEAEs, and 7% (95% CI 0.02–0.11, I2 = 84%, three studies, N = 926 patients) [30, 31, 46] due to lack of efficacy. Medium term, meta-analyses showed a 24% discontinuation rate (95% CI 0.00–0.54, I2 = 96%, two studies, N = 173 patients) [21, 44] due to all-cause, 7% (95% CI 0.05–0.09, I2 = 0%, two studies, N = 730 patients) [17, 44] due to TEAEs, and 2% (95% CI 0–0.06, one study, N = 137 patients) due to lack of efficacy [44]. Long term, the discontinuation rate was 14% (95% CI 0.02–0.43) due to non-adherence [33]. See Online Resource 2 Figure S7 for details.

3.10 Adverse Events

The short-term pooled total TEAE rate was 54% (95% CI 0.29–0.79, I2 = 98%, three studies, N = 932 patients) [30, 31, 46], the medium-term was 54% (95% CI 0.43–0.64, I2 = 91%, three studies, N = 1182 patients) [38, 44, 45], and the long-term was 69% (95% CI 0.56–0.82, I2 = 89%, two studies, N = 674 patients) [38, 40]. Results of at least one serious TEAE rate and possibly drug-related TEAE rate are shown in Online Resource 2 Figure S8. The range of long-term abnormal weight gain was 9.8% (one study, N = 521 patients) [22]. The short-term increase in weight ranged from 1.5 to 7.6% (three studies, N = 985 patients) [30, 31, 46] and the long-term increase ranged from 1.5 to 7.5% (three studies, N = 1110 patients [22, 38, 44]. One study (N = 593 patients) reported a 3.2% medium-term increase in weight [45]. In addition, the short-term rate of glucose-related TEAEs was 0.5% (one study, N = 212 patients) [46] and the long-term rate was 0.6% (one study, N = 521 patients) [22]. Results of other metabolic-related TEAEs are shown in Online Resource 2 Table S1. The most reported TEAEs included injection-site pain, weight increase, headache, insomnia, constipation, upper respiratory tract infection, dizziness, nasopharyngitis, somnolence, and extrapyramidal symptom TEAEs (i.e., akathisia and tremor).

3.11 Subgroup and Sensitivity Analysis

The pre-planned subgroup analyses were conducted to explore the source of heterogeneity. However, due to insufficient data, only analyses stratified by OAP history were conducted. There was a significant subgroup difference (p = 0.02) in short-term changes in the PANSS score between patients with a history of olanzapine (mean change − 25.50, 95% CI − 30.94 to − 20.06) and risperidone (mean change − 17.61, 95% CI − 21.42 to − 13.81) use [30, 31, 33]. This indicated that a history of OAP use may be one of the sources of heterogeneity in the short-term change in PANSS score results (Online Resource 3 Table S2). The sensitivity analyses were only available for the change in the PANSS total score. After removing data with imputed SDs, the result did not substantially differ from the result reported with imputed SDs in the review (Online Resource 3 Table S3).

3.12 Quality Assessment of Included Studies

The overall quality assessment result for each study is shown in Table 1. For assessment details pertaining to RoB2, NOS, and NIH tool, see Online Resource 4.

4 Discussion

4.1 Summary of Findings

This review summarises the data supporting the efficacy and safety of switching from OAPs to PP1M in the treatment of patients with schizophrenia as reported in 15 studies with a total of 4740 patients. Our findings indicated the relapse rate was 12% among moderately ill patients over 26 weeks. The long-term rehospitalisation rate was 18%, although the severity and stage of illness was not extractable to provide greater precision. In addition, the PANSS total scores, CGI-S scores, and PSP total scores were improved at short-, medium- and long-term follow-ups. According to subgroup analysis results, there was a greater reduction in the PANSS score among patients who previously used olanzapine than among former risperidone users (− 25.5 vs − 17.61). Compared with previous research, our findings also suggested that a safety profile associated with switching to a PP1M was similar to that observed with OAP use. The sensitivity analyses regarding data imputation did not alter our conclusions about the primary outcomes, although careful interpretation of these results is needed due mainly to limited evidence.

4.2 Change in the PANSS, CGI-S, and PSP Scores from Baseline

The reduction on PANSS and CGI-S scores indicated PP1M clinically reduces the symptoms and severity of schizophrenia in patients with a history of prior OAP use. The short-term (PANSS: − 21.69; CGI-S: − 1.42) and long-term (PANSS: − 17.88; CGI-S: − 1.7) changes in the score met their minimum clinically important difference (MCID) (PANSS MCID: 15 [51]; CGI-S MCID: 1 [52]). The medium-term change was − 14.98 in the PANSS score, which just met the MCID cut-off point, and − 0.61 in the CGI-S score, which was lower than its MCID. However, it might be inappropriate to discredit the benefits of switching to PP1M when the MCID is not met. A previous prospective single-arm study, (N = 546), for example, set a 10-point reduction as a meaningful change on the PANSS score, which indicates MCID should not be used as the only criterion [40]. In addition, the greater reduction in the PANSS score among patients with olanzapine compared with patients with risperidone should be interpreted with caution as the data were very limited [30, 31, 33]. In this review, we did not have a definition of previous OAP users. Although the study providing relevant data mentioned that patients were included when they had been on the same oral antipsychotic given for the treatment of schizophrenia in an adequate therapeutic dose, there was no limitations or restrictions on frequency, duration, and dose of OAPs. Also, patients were not randomly assigned to OAP groups. The difference on relevant clinical characteristics may confound interpretation of our findings. Thus, future studies would be expected to provide more comprehensive and detailed data on this aspect for a better understanding of the effects of switching to a PP1M regimen.

The improvement in the PSP score (approximately 10%) at the short-, medium- and long-term follow-ups demonstrated a clinically meaningful improvement over time in patients' daily function [53]. In addition, the reduced length of stay and inpatient visits at medium- and long-term follow-ups may indicate a clinical benefit of the treatment effectiveness associated with switching to a PP1M. These improvements can be seen as beneficial and important from the perspective of reducing the burden on family caregivers, society, and the health system [36, 54, 55].

Our results demonstrated that over 50% patients showed at least a 30% and 40% reduction from the baseline PANSS total score during various treatment periods. Although a reduction of 50% is recognised as a more meaningful measurement of response rate, few studies set this as a criterion. It should be noted that 50% was considered a cut-off threshold among patients taking OAPs [56]. Thus, along with the change in PANSS and CGI-S scores in our review, setting a lower cut-off point may be more appropriate to reflect clinically meaningful improvement associated with switching patients from OAPs to PP1Ms.

4.3 Comparisons with Previous Research

Compared with the efficacy and safety of PP1M reported in a previous long-term double-blind randomised controlled trial (RCT) with 747 patients [57], our review showed consistent results, such as a low relapse rate (12% vs 18%), reduced PANSS total score (− 17.88 vs − 11.6), and high incidence of total TEAEs (75% vs 76%). Likewise, the short-term change in the PANSS score observed in this review was consistent with a previously unpublished 13-week, double-blind RCT, although the sample size was not retrieved) (− 21.69 vs − 18.6) [58]. Therefore, our findings may support extending the extrapolation of PP1M to patients with history of prior OAP use.

Compared with oral SGAs, switching to PP1M results in a lower relapse rate (12%). This is consistent with results reported in a previous review six RCTs with 525 patients, where the 1-year relapse rate for LAIs (27%) was lower than that of patients who received oral medication (42%) [59]. Patients treated with oral SGAs reported a wide range (29.9–73.7%) of all-cause discontinuation rates at both short- and medium-term follow up in previous studies (a 4-week double-blind RCT with 141 patients, a 12-week single-blind RCT with 72 patients, and a systematic review of 9 included studies with 648 patients) [60,61,62]. Conversely, our results showed a 31% short-term and 24% medium-term discontinuation rate in patients who were switched to a PP1M. For the long-term, we found a much lower discontinuation rate due to non-adherence (14%) when switched to PP1M compared with approximately 30% discontinuation rate reported for oral SGAs in an 18-month RCT with 1493 patients [63]. The lower discontinuation rate associated with switching to a PP1M may be beneficial in terms of greater symptom improvement and a higher response rate [55]. Of note, the time to discontinuation of antipsychotics may influence the inclusion and sequence of administration of other drugs, thereby influencing the switching strategies [64]. Contrarily, a recently published RCT (EULAST trial) with 533 patients with early-phase schizophrenia reported that there was no significant difference in time to all-cause discontinuation between the combined oral and combined LAI treatment groups (71% vs 64%) [65]. The contrary results may be explained by different baseline characteristics. For example, in our review, only patients with previous OAP history and without LAI history were included. Whereas in the EULAST trial, only patients with previous history of investigated drugs were excluded. Different drug history may affect the results.

Despite the fact that a minority of patients prescribed LAIs may still encounter instances of missed or delayed injections for various reasons, such as the COVID-19 pandemic, that may potentially compromise the effectiveness of the treatment regimen, our findings suggested a lower discontinuation rate due to non-adherence in patients switching to PP1M compared to that reported for OAPs in previous research (14% vs 30%). This may provide evidence to support the ability of LAI, including PP1M, to circumvent covert non-adherence commonly associated with oral antipsychotic medications. By eliminating the possibility of undisclosed non-adherence, LAIs offer enhanced assurance regarding treatment adherence, which can positively impact therapeutic outcomes in individuals with schizophrenia [8]. Subsequent investigations are required to obtain direct evidence that substantiates the assumed improvement in adherence among patients who transition to PP1M compared to those receiving OAPs.

Compared to previous study results in a 15-month RCT with 444 patients, the pooled overall TEAE rate for patients switching to PP1M (69%) in this review was much lower than oral FGAs (91.4%) and slightly lower than oral SGAs (77.6%) [66]. At least one serious TEAE rate was similar to that reported for OAPs in a 24-month RCT with 764 patients (8% vs 12.7%) [67]. Consistent with a previous study of SGA-LAIs, the high incidence of injection-site pain and weight gain are findings of concern regarding switching to PP1M [68], indicating the need for careful monitoring in clinical practice. Other commonly reported adverse events such as headache and insomnia were often rated as mild or moderate in intensity [17].

It is worthy of note that no included studies provided a clear definition of discontinuation and adverse events or details on how they assessed these indicators. Different methodologies employed for assessing adverse events may have potential influence on reported results [69]. This indicated the importance of detailing assessment methods for adverse events in future research, in order to enhance the accuracy of safety awareness.

Previous reviews suggested LAIs, including paliperidone palmitate, may be an effective treatment for patients experiencing early psychosis in terms of relapse, hospitalisation reduction, and symptom improvement [70, 71]. Although we were not able to assess the effect of stages of schizophrenia due to insufficient data, two studies that enrolled recent-onset (≤ 5 years) or early-stage patients included in our review reported consistent results with previous reviews. Zhang 2015 reported PP1M significantly improved the PANSS total score from baseline to Month 18 and reduced the percentage of recent-onset patients requiring hospitalisation in the past 12 months [22]. The other study reported that PP1M was associated with significantly improved function status at both short- and long-term follow-ups [41].

4.4 Strengths and Limitations

Our study has some strengths and limitations. To our knowledge, this is the first comprehensive and most current meta-analysis evaluating the efficacy and safety of switching patients with schizophrenia from OAPs to PP1M. The results may guide clinical decisions and fill the evidence gap in the field of PP1M. Moreover, this study was carried out with good quality control as we conducted the review strictly according to the Cochrane and PRISMA standard.

However, there are several limitations. First, of note, although our findings demonstrated several advantages of switching to PP1M in the short-, medium- and long-term time periods, the results should be interpreted with caution. Different follow-up periods should be carefully compared and potential influencing factors taken into consideration. For instance, the smaller absolute change in the PANSS score at the medium-term follow-up should not be interpreted as an indication of a poorer symptom improvement after switching to PP1M at that time period in light of the clear benefits noted at other time points. Furthermore, the pooled results in meta-analysis might be influenced by multiple factors, which may limit the comparability of results between follow-up periods. For example, some research suggests that greater improvement might be due, in part, to a higher baseline PANSS score, shorter duration of illness, greater baseline severity, and a more recent onset phase [40, 72,73,74]. Furthermore, of the 15 included studies, six involved patients who had a baseline PANSS total score exceeding 70 [30, 31, 33, 38, 45, 48]. Additionally, two studies specifically recruited patients in an acute stage [46, 49], implying that these individuals were transitioning from OAPs to a PP1M injection during periods characterised by worsened illness and instability. Consequently, the initial measurements of symptomatology could have been elevated, rendering it plausible that the subsequent improvement observed upon switching to the LAI formulation could be partially ascribed to intensified clinical intervention during periods of illness, as well as a natural alleviation of symptoms over time [75]. Moreover, it is possible that patients diagnosed with schizophrenia had limited clinical monitoring and support while receiving OAPs. However, upon transitioning to PP1M, they would have received monthly visits from a mental health professional responsible for administering the injection. This increased level of clinical monitoring would have allowed for prompt response to reports of side effects and changes in mental state. Therefore, for individuals who experienced an escalation in clinical monitoring, on-going clinical review, and improved communication, any observed benefits subsequent to the switch might be partially attributed to these factors. On the other hand, the increase in clinical care is recognised as the unique benefit of LAIs compared to OAPs in managing schizophrenia [76].

Second, the study design and sample size may limit the generalisability of the results. Unfortunately, many studies did not report sufficient details to enable stratification of the results according to possible influencing factors [20, 34, 36, 38, 49, 50]. Thus, future studies focusing on confounders that might impact the effect of switching to PP1M would be beneficial. Additionally, the treatment effect might also be influenced by the definition of outcomes. Because we included the hospitalisation rate defined by the original studies, the variation in the threshold of hospitalisation may affect the pooled results.

Third, only one study reported the discontinuation rate due to non-adherence, which could compromise the treatment effect [33]. Fourth, regarding the limitations of methodology, because we performed a single arm meta-analysis, the results were pooled from single-arm studies or single arms of a comparative study where the blinding and sample size calculation methods were not reported for most studies. This may lead to an under- or over-estimation of the efficacy and safety of switching to PP1M. The nature of proportional data could lead to important heterogeneity, which was reflected by an I2 over 50% in most of our outcomes. In addition, our findings need to be interpreted carefully and not overly generalised due to the limited evidence included for each outcome and the various follow-up periods. Also, our findings should be carefully generated to patients in early stages of schizophrenia since there is a lack of studies in this area.

4.5 Clinical Implications

Overall, our results provide clinicians with evidence that switching patients with schizophrenia to PP1M may be an effective option for those previously taking OAPs. This approach can reduce relapses and rehospitalisation, and probably improve adherence in the long term [20, 22, 33, 38]. In addition, it can reduce symptoms, disease severity and improve social functioning over the short, medium, and long term [21, 22, 30,31,32,33, 38, 41, 45, 46]. However, the findings should be interpreted and applied with caution as the results were based on limited evidence. The influence of disease stage, ethnicity, and history of OAP use remains unclear. Thus, future high-quality comparative studies are needed to confirm the benefits of switching patients to PP1M. Possible influencing factors should be explored in future studies to generalise the findings more precisely.

5 Conclusions

Our findings indicate switching patients with schizophrenia from OAPs to PP1M is effective and tolerated and is associated with lower relapse and hospitalisation rates, PANSS, CGI-S and PSP scores as well as a comparable safety profile to oral OAPs. Compared with previously published research of OAPs, switching to PP1M may also improve adherence and have a similar safety profile to oral OAPs.