Response rates to sequential trials of antipsychotic medications according to algorithms or treatment guidelines in psychotic disorders. A systematic review and meta-analysis

Background: There is a relative lack of research evaluating the outcomes when treatment guidelines or algorithms for psychotic disorders are followed. This systematic review and meta-analysis determined the response rates to antipsychotic medications at different stages of these algorithms and whether these response rates differ in first episode cohorts. Methods: Data sources: A systematic search strategy was conducted across four databases PubMed, EMBASE, PsycINFO (Ovid) and CINAHL. Studies that had sequential trials of different antipsychotic medications were included. A meta-analysis of proportions was performed using random effects models and sub-group analysis in first episode psychosis studies. Results: Of the 4078 unique articles screened, fourteen articles, from nine unique studies, were eligible and included 2522 participants. The proportion who experienced a response to any antipsychotic in the first stage of an algorithm was 0.53 (95 % C.I.:0.38,0.


Background
Antipsychotic medications are the mainstay of treatment for psychotic disorders and are at least moderately effective in reducing the overall severity of symptoms for individuals with a diagnosis of schizophrenia (Leucht et al., 2009).Pharmacological guidelines or algorithms for the use of antipsychotic medications in the treatment of psychotic disorders have been developed with this evidence base and share a number of recommendations, such as the use of secondgeneration antipsychotic medication as first line and the principle of starting at a low dose and slowly titrating upwards according to response and tolerability (Keating et al., 2017).A further consistent recommendation within these algorithms is that clozapine should be considered as a third-line agent for individuals with a diagnosis of schizophrenia if two previous antipsychotic medications have not been effective.
Clozapine is the most effective antipsychotic for the reduction of positive symptoms (Siskind et al., 2016) and reducing hospitalisations (Land et al., 2017).Furthermore, a meta-analysis of 24 studies, of which all but one were naturalistic observational studies, demonstrated that continuous treatment with clozapine was associated with a greatly reduced crude mortality rate (Vermeulen et al., 2019).In addition, a recent registry study showed that while nearly one third of individuals were non-adherent with their antipsychotic medication, only 5 % did not take clozapine as prescribed (Lieslehto et al., 2022), indicating that despite the associated side-effects, it is an acceptable treatment for individuals.Due to the risks of agranulocytosis and the subsequent regulatory requirements that were introduced, clozapine is not used as a firstline treatment (Thien et al., 2018), with the exception of in China (Ruan et al., 2023) and also for the treatment of psychosis in individuals with Parkinsons disease (Friedman, 2024).
Despite the superior outcomes associated with clozapine, only a small proportion of individuals with a diagnosis of treatment-resistant schizophrenia actually have a trial of clozapine, which can be preceded by years of delay from the time from which they were eligible (Thien and O'Donoghue, 2019).Thus, there is a disconnect between what is recommended in clinical guidelines and what is occurring in routine clinical practice.Despite the resources employed to develop clinical guidelines for mental health disorders, there is a relative lack of research evaluating their implementation and outcomes (Nguyen et al., 2020).
Therefore, this systematic review and meta-analysis aimed to determine the proportion of individuals with a psychotic disorder who achieved a response to: (i) the 1st trial of an antipsychotic medication; (ii) the 2nd trial of an antipsychotic medication; (iii) clozapine used in the 3rd stage and (iv) the 3rd trial of an antipsychotic medication, other than clozapine.As individuals with a first episode of psychosis are typically naïve to antipsychotic medication, sub-group analysis will be performed separately in the studies that included individuals with a first episode psychosis and those with an enduring psychotic disorder.

Protocol registration
The protocol for this systematic review was registered with PROS-PERO (application reference: 384433).

Eligibility criteria
Inclusion criteria: This systematic review included studies that met the following inclusion criteria: (i) including a clinical population of people with a psychotic disorder, including but not limited to: schizophrenia spectrum disorders, affective psychotic disorders and other psychotic disorders, such as brief psychotic disorder, substance induced psychotic disorder or psychotic disorder not otherwise specified; (ii) treatment algorithm/ guideline must include at least two sequential trials of antipsychotic medications and a trial is considered at least four weeks; (iii) the study must include information pertaining to the proportion of participants who achieve a response at each stage of the treatment algorithm; (iv) published in English language and (v) no age restrictions will be applied to the participants and there was no restriction based on year of publication.Studies were excluded If the intervention group was provided with additional interventions, such as psychosocial or family-based interventions and the effect of these additional interventions could not be differentiated from the pharmacological effects.In addition, algorithms or guidelines that are exclusively directed towards rapid tranquilization were not included.Finally, case reports and case series were not included.

Information sources
Search terms were entered into four databases: PubMed, EMBASE, PsycINFO (Ovid) and CINAHL.In addition, the reference list of relevant articles, including any reviews, were screened to identify any further articles.Articles were screened at title and abstract, then at full text level by two authors (FP and HP).The grey literature was also examined, with abstracts from large, international conferences with a focus on psychotic disorders and the treatment of psychotic disorders were reviewed, such as the Schizophrenia International Research Society (SIRS) conference (Schizophrenia Research & Schizophrenia Bulletin) and IEPA International Early Psychosis Association conference (Early Intervention for Psychosis).The authors of potentially eligible studies were contacted, for example, some studies may have provided the overall results of the use of an algorithm and in these cases, the response rates after each stage of the algorithm were requested from the authors.

Data items
For eligible articles, the following data was extracted: setting, participantssex, mean age, demographic details, study design type, clinical practice guideline evaluated, study sample size in each group, comparison group, diagnostic groups included, co-morbidities of participants, clinical characteristicse.g.symptom severity, functioning, outcome(s) evaluated, instruments used to determine outcomes, followup rate, primary outcome, specifically response rates.One author (BOD) extracted data from each eligible study.

Study risk of bias assessments
The Cochrane risk of bias tool for randomised trials was used to assess quality and risk of bias in randomised trials (Sterne et al., 2019) and the Newcastle-Ottawa scale was used to assess quality and risk of bias in non-interventional studies (Wells et al., 2009).

Primary outcome & measure of effect
The primary outcome was the proportion of individuals who experienced a response to antipsychotic medication in the first, second and third stage of an algorithm.The proportion was determined from the numerator (number of individuals who responded) divided by the denominator and is presented as a number between 0 and 1, with 95 % confidence intervals.The denominator included all individuals who started a trial stage of a particular antipsychotic medication, even those who may have discontinued the medication due to intolerability or for other reasons, i.e. the denominator will represent the 'intention to treat' approach.This method provided a more conservative estimation of the proportion who experienced a response.

Synthesis methods
The meta-analysis of proportions of means was conducted in Stata v18.The Freeman-Tukey double-arcsine transformation of proportions was used for each study, as there were a number of studies with proportions close to 0 or 1 and in this circumstances, this particular variance-stabilizing transformation is recommended.Random effects models were used to account for the risk of heterogeneity between studies.

Determining eligibility of studies
We included any study that had at least two stages of different antipsychotic medications and provided response rates for each stage.Some studies may have had multiple arms and evaluated an intervention that was not relevant to this review, such as an augmentation strategy or a psychosocial intervention.In this circumstance, we only included the data pertaining to the antipsychotic medication.In the CATIE study, the second and third stage of the algorithm each consisted of two different trials.Therefore, the response rate to each of the antipsychotic medications from each separate trial were combined, to provide the overall response rate to each medication at each stage.In the Optimise trial, some participants remained on amisulpride and entered the third stage with having had only one trial of an antipsychotic medication.The results pertaining to the response to clozapine in those who had amisulpride alone in stage 1 and 2 and those who had amisulpride in stage 1 and olanzapine in stage 2 were provided by the investigators of the OPTiMiSE study, which facilitated the study being included in this review.The study by Yoshimura et al. (2017) did not provide the response rates for clozapine when it was used in the 3rd stage, as some participants had been prescribed clozapine as a 2nd, 3rd and 4th line medication in the study and the results were not provided separately for clozapine in these different stages and therefore this study was not included in the analysis pertaining to clozapine as a 3rd line medication.

Ethics approval and consent
Ethical approval was not required for this systematic review and as individual level data was not utilised, individual consent was not required.

Study selection
A total of 4078 unique articles were identified and of these, 4038 were excluded from the screening of the title and abstracts.A total of 40 articles were reviewed in full and of these 14 articles from 9 unique studies were eligible to be included.The reasons for excluding articles are presented in Fig. 1.

Study characteristics
The characteristics of the included studies are presented in Table 1.The fourteen eligible articles pertained to nine unique studies.Three of the studies were observational cohort studies (Agid et al., 2011b;Drosos et al., 2020;Yoshimura et al., 2019a) and six studies had at least one component of randomisation (Hirano et al., 2013;Kahn et al., 2018a;Kane et al., 1988;Shalev et al., 1993;Stroup et al., 2003;Suzuki et al., 2007).The findings from the CATIE study relevant to this systematic review were published across five articles (Lieberman et al., 2005;McEvoy et al., 2006;Stroup et al., 2009;Stroup et al., 2007;Stroup et al., 2006).Two other studies randomised participants to one of six groups, which consisted of the different possible sequences of three different antipsychotic medications across three stages (Shalev et al., 1993;Suzuki et al., 2007).One study had the design in which the prescribing psychiatrist was randomised to employ the algorithm or treatment as usual (Hirano et al., 2013).Four of the studies included participants with a first episode of psychosis (FEP) and these studies included a total of 907 participants.Each of these studies applied different criteria to the definition of FEP in relation to the duration of previous exposure to antipsychotic medication, ranging from four weeks (Yoshimura et al., 2019a), six weeks lifetime exposure (Kahn et al., 2018a) and twelve  weeks (Drosos et al., 2020).The previous exposure to antipsychotic medication was not explicitly stated in one study, although it is stated that it participants underwent their first trial of an antipsychotic within the study, so it can be assumed that there no prior exposure was permitted (Agid et al., 2011b).Four studies included individuals with an established and enduring diagnosis of schizophrenia (Hirano et al., 2013;Lieberman et al., 2005;Shalev et al., 1993;Suzuki et al., 2007) and one study included individuals with a diagnosis of treatment-resistant schizophrenia (Kane et al., 1988).The definition of response in each individual study is presented in Table 1.A description of the interventions provided according to the treatment algorithm for each study is provided in Table 2.

Results of individual studies
The results of each individual study are presented in Supplementary Table 1 and this includes findings on the response rates to each specific antipsychotic medication at the different stages of each study and the mean (±sd) dose used to achieve a response if available.

Response rates in the 1st stage of an algorithm
There were data pertaining to 2522 participants in relation to whether they experienced a response to their first trial of antipsychotic medication.The overall proportion who experienced a response for the 1st stage of an algorithm was 0.53 (k = 8 95 % C.I.: 0.38, 0.68, I 2 = 97.5 %).In the FEP studies, the proportion that achieved a response was 0.63 (k = 4 95 % C.I. 0.45, 0.79, I 2 = 96.2%) compared to 0.42 (k = 4 95 % C. I. 0.22, 0.62, I 2 = 93.9%) in the non-FEP studies and this difference between groups was not significant (p = 0.12).The forest plot of the meta-analysis for the 1st stage of an algorithm is presented in Fig. 2.

Response rates in the 2nd stage of an algorithm
In the 2nd stage of any algorithm, there were data pertaining to 1031 participants and the overall proportion who achieved a response was 0.26 (k = 11, 95 % C.I.: 0.15, 0.39, I 2 = 93.5 %).In the FEP studies, the proportion that achieved a response was 0.34 (k = 4, 95 % C.I. 0.16, 0.55, I 2 = 85.1 %) compared to 0.22 (k = 7, 95 % C.I. 0.08, 0.39, I 2 = 95.3 %) in the non-FEP studies and this difference between groups was not significant (p = 0.35).The forest plot of the meta-analysis for the 2nd stage of an algorithm is presented in Fig. 3.

Response rates to clozapine in the 3rd stage of an algorithm
There were data pertaining to 205 participants who were prescribed clozapine as a 3rd line treatment and the proportion that achieved a response was 0.43 (k = 4, 95 % C.I.: 0.19.0.69, I 2 = 90.4%).In the FEP studies, the proportion that achieved a response to clozapine in the 3rd stage was 0.45 (k = 2, 95 % C.I. 0, 0.97, I 2 = 93.2%) compared to 0.41 (k = 2, 95 % C.I. 0.19, 0.65, I 2 = 85.2 %) in the non-FEP studies and this difference between groups was not significant (p = 0.90).The forest plot of the meta-analysis for the response rates to clozapine in the 3rd stage of an algorithm is presented in Fig. 4.

Response rates to other antipsychotic medications (not clozapine) in the 3rd stage of an algorithm
In the 3rd stage of any algorithm, there were data pertaining to 382 participants who were prescribed an antipsychotic medication other than clozapine.The overall proportion who achieved a response was 0.26 (k = 7, 95 % C.I.: 0.05, 0.54, I 2 = 92.9%).In the FEP studies, the proportion that achieved a response was 0.15 (k = 2, 95 % C.I. 0.01, 0.37, I 2 = 0 %) compared to 0.26 (k = 5, 95 % C.I. 0.05, 0.54, I 2 = 92.9%) in the non-FEP studies and this difference between groups was not significant (p = 0.35).The forest plot of the meta-analysis for the response rates of other antipsychotic medications in the 3rd stage of an algorithm is presented in Fig. 5.  If, after 2 trials, the response criteria was not met, then a trial of clozapine was offered.

Suzuki et al., 2007, Japan
The algorithm was based on the availability of antipsychotic medications in Japan and this included olanzapine, quetiapine and risperidone.
Participants were randomly assigned to one of the six treatment arms that consisted of all of the possible combinations of the three medications and flexible dosing allowed.Each medication trial represented an individual stage and there was a maximum of 3 stages.After 4 weeks, if there was non-response (BPRS score at 4 weeks ≥90 % at baseline) then participant progressed to the next stage.If there was a partial response (BPRS Score ≤ 90 % but ≥70 %) then same antipsychotic medication continued for a further 4 weeks and if no change then, they progressed to the next stage.CATIE trial Lieberman et al., 2005Stroup et al., 2006Stroup et al., 2009 Protocol for clinical trial.The findings pertaining to the 3 stages of the CATIE trial were published across 5 articles.After stage 1, participants could decide to participate in the Stroup et al., 2006  Randomised controlled trial of olanzapine, perphenazine, quetiapine, risperidone or ziprasidone for up to 18 months.Participants randomised to perphanzaine who did not respond, entered the arm described in Stroup et al., 2007 and those initially randomised to olanzapine, quetiapine, risperidone or ziprasidone who did not respond entered the Stroup et al., 2006 arm.Participants randomised to receive either olanzapine, quetiapine, risperidone or ziprasidone but not allocated to the medications that was not used in stage 1.
Open label trial of aripiprazole, clozapine, fluphenazine decanoate, olanzapine, perphenazine, quetiapine, risperidone or ziprasidone.Participants could also select to have two of the above antipsychotic medications.Stroup et al., 2007 Participants randomised to olanzapine, quetiapine or risperidone.McEvoy et al., 2006 Participants who had been randomised & discontinued either olanzapine, quetiapine, risperidone or ziprasidone in stage 1 were randomised to clozapine or another antipsychotic not used in Stage 1. Shalev et al., 1993, Israel Algorithm consisted of the possible treatment combinations of 3 different first-generation antipsychotics (haloperidol (HPD), levomepromazine (LPZ) & perphenazine (PPZ)).These medications were chosen as they were chemically distinct and varied in potency.
Participants were randomly assigned to six treatment schedules based on the possible combinations of the 3 medications.Prescribing psychiatrists were blinded to sequence in which medications were to be given.After 4 weeks, if there was a non0response (<30 % improvement on BPRS) then the participant progressed to the next stage of the sequence that they had been randomised to.

Summary of findings
The main findings of this meta-analysis are that when algorithms are applied to the treatment of psychotic disorders, there is an initial high response in the first stage, but then the response rate reduces substantially for the second and third stage, unless clozapine is initiated.These results were also consistent in the first episode cohorts, with a high response rate in the first stage which reduces incrementally with each stage, unless clozapine is initiated, to which nearly half of individuals achieved a response.

Clinical implications
There are a number of important clinical implications from this meta-analysis.The first is that there is an initial high response rate to antipsychotic medication when an algorithm is used and this is highest in the first episode cohorts.Following the initial antipsychotic medication, the response rate falls substantially unless clozapine is initiated.
Our findings around response to clozapine are in keeping with a previous meta-analysis which found a clozapine response rate of 40 % (Siskind et al., 2017).Almost one-quarter of people with first-episode psychosis or schizophrenia will develop treatment-resistant schizophrenia (Siskind et al., 2022), and there have been recent advances in  attempting to identify treatment resistance as early as possible in these cohorts, using demographic, clinical and biomarkers (Smart et al., 2022;Yang et al., 2022).While developing tools to identify treatment resistance early is undoubtably a worthwhile pursuit, it can be identified clinically within three months by the non-response to two antipsychotic medications (Kane et al., 2019).As already highlighted, there tends to be long delays before clozapine is commenced and it is likely that a multifaceted approach would be required to reduce these delays.This would include further education and promotion for prescribing clinicians, having an efficient system for haematological monitoring and the resources and guidelines for the safe commencement of clozapine in the community, as nearly half of the individuals eligible for clozapine stated that the need to be admitted to hospital deters them from having a trial of clozapine (Gee et al., 2017).Recent studies have demonstrated that it can be safe, effective, and cost saving to initiate clozapine in the community for selected individuals (Butler et al., 2022).A recent qualitative study interviewed the psychiatrists providing care for individuals who were eligible to have a trial of clozapine but were still yet to have a trial (Jakobsen et al., 2023).The expected non-adherence with blood tests and also with the medication were the two commonest reasons for not prescribing clozapine.However, this perception is in contrast to the evidence that adherence to clozapine is higher than other antipsychotic medications (Lieslehto et al., 2022), even in individuals who were non-adherent with their antipsychotic medication prior to initiating clozapine (Brodeur et al., 2022).One of the main findings from the CATIE study was the very high rate of discontinuation of all antipsychotic medications (Davis et al., 2011).Therefore, a previous history of non-adherence should not preclude an individual from a trial of clozapine and even if this is a concern, clozapine levels can now be measured more frequently using point of care testing (Kamhi-Nesher et al., 2022), which can then give an indication of adherence.
There have been recent calls for the haematological monitoring requirements for clozapine to be changed and to be confined to the first year of use, as after that, the risk of agranulocytosis is similar to other commonly prescribed antipsychotic medications (Oloyede et al., 2023).While this could reduce a barrier to the timely use of clozapine, other important and potentially fatal adverse effects, such as gastrointestinal Fig. 3. Meta-analysis of proportion of response rates of all antipsychotic medications in the 2nd stage.hypomobility, myocarditis and increased risk of pneumonia, need to be continually monitored for and managed (de Leon et al., 2023;Every-Palmer and Ellis, 2017).
Early intervention for psychosis services (EIS) should be resourced appropriately to facilitate the commencement of clozapine, as the typical tenure of care of EIS covers the period in which treatment resistance should be identified (Stokes et al., 2020;Thien et al., 2018).An underlying philosophy of the pharmacological management of first episode has been to 'start low and go slow' (International Early Psychosis Association Writing Group, 2005) based on the findings that individuals with a first episode of psychosis respond to lower doses of potent antipsychotic medications (Mouaffak et al., 2021).However, an important balance needs to be maintained, as titration needs to occur slowly enough that only the minimum dose required to achieve remission is prescribed, yet at the same time ensure that trials of different antipsychotic medications occur in a timely enough manner so that identification of treatment resistance is not delayed.As a longer delay to the commencement of clozapine is associated with poorer outcomes compared to those who start clozapine earlier (Yoshimura et al., 2017).
In terms of the clinical applicability of these findings, it needs to be acknowledged that the reality of clinical practice is different to rigid and strict protocols, as it is common for antipsychotic medications to be changed for a variety of reasons, such as side-effects, patient preference and the ability to provide specific medications in the long-acting injection form.Therefore, while algorithms can provide a structure and guidance to the treatment of psychotic disorders, it cannot be expected that they are always followed precisely, and treatment needs to be tailored to the individual in relation to their preferences and the tolerability of the recommended treatment.In addition to this, it needs to be considered that the data derived from randomised clinical trials that inform these algorithms and guidelines are biased due to the participants in clinical trials are more likely to be adherent, have less comorbidities and have less risks and this could explain why findings from RCTs are more likely to over-estimate the effectiveness of antipsychotic medications (Efthimiou et al., 2024).
The findings from this meta-analysis provide percentages for the likelihood of a response at different stages of treatment of a psychotic disorder.This information can assist an individual affected by a psychotic disorder and their caregivers in making an informed decision as to whether to proceed with a trial of clozapine or a different antipsychotic medication, knowing the likely response rates.

Further research
There are several ways in which further research could progress this area.First, while there has been a moderate amount of research determining the effectiveness of the use of algorithms and treatment guidelines for psychotic disorders, which facilitated this meta-analysis, there is a dearth of knowledge on how to implement their use in routine clinic practice.Specialist consultation services have been established in Australia and the UK to assist prescribing clinicians in identifying, diagnosing and treating individuals with treatment resistant schizophrenia in the community (Beck et al., 2014;Edwards et al., 2002).The other aspect of algorithms that could be developed is expanding the recommendations beyond positive symptoms.None of the algorithms included in this study included treatment recommendations for negative or depressive symptoms or cognitive deficits in people affected by psychotic disorders, all of which cause significant impairment and reduce quality of life (Helldin et al., 2020;Herniman et al., 2019;Lyne et al., 2018).A further area in which there is need to develop pharmacological algorithms is for the first episode non-schizophrenia spectrum psychotic disorders.Three of the four FEP studies included in this systematic review only included participants with a diagnosis of a schizophrenia spectrum disorder and this leaves uncertainty about the effectiveness of these algorithms in the other disorders represented within first episode cohorts, such as affective psychotic disorders, drug induced psychosis, brief psychotic disorder or psychotic disorder not otherwise specified.Additionally, there have been indications that long acting injectable antipsychotic medications can lead to superior outcomes in some domains for people with early psychosis, such as relapse rates and hospitalisation (Lian et al., 2022), however current algorithms do not include recommendations in relation to the use of administering antipsychotic medications via this method.

Limitations
The findings of this systematic review and meta-analysis need to be considered within the limitations of the study.First, data was only extracted pertaining to the response rate at each stage of an algorithm or guideline and our analysis did not utilise a control group.This means that the response rates observed cannot be directly attributed to the use of the medication or algorithm, as control groups were lacking.It is possible that a number of the studies that were classified as 'enduring psychotic disorders' included a combination of participants with a first episode psychotic disorders and those with an enduring disorder.Furthermore, different definitions of response were used in different studies and this could explain some of the variation in the results observed, for example, the CATIE study took time to discontinuation as the primary outcome and therefore a response was defined as remaining on the antipsychotic medication, as opposed to any improvement in symptoms or functioning.In addition, the number of participants in the meta-analysis pertaining to the response rates to clozapine was low and were likely to have insufficient power.Unfortunately, it is unlikely that there will be many more clinical trials involving clozapine as an intervention, as it now available as a generic medication and its effectiveness has been established.Therefore, further interventional studies in relation to clozapine should look at reducing barriers to its initiation.A further limitation is that we grouped all antipsychotic medications together, as the purpose was to determine the overall response rates at different stages, but it is possible that different antipsychotic medications were used more frequently in different stages and could account for some of the variation of responses observed.

Conclusions
The use of algorithms to promote evidence-based care can result in superior outcomes in psychotic disorders, particularly for those with a first episode psychosis.They could also assist in ensuring people with
arm or the McEvoy 2006 arm.

Fig. 2 .
Fig. 2. Meta-analysis of proportion of response rates of all antipsychotic medications in the 1st stage.

Fig. 4 .
Fig. 4. Meta-analysis of proportion of response rates to clozapine in the 3rd stage of an algorithm.

Fig. 5 .
Fig. 5. Meta-analysis of proportion of response rates to other antipsychotics (not clozapine) in the 3rd stage of an algorithm.

Table 1
Characteristics of the eligible and included studies in the systematic review.

Table 2
Details of the interventions in each of the different stages of each treatment algorithm