Summation
-
The available evidence suggests that electroconvulsive therapy is one of the most efficacious treatment strategies, in terms of augmentation of clozapine in patients with schizophrenia.
-
In terms of augmentation with antipsychotics, available evidence suggests aripiprazole be superior to other antipsychotic agents.
-
Evidence for augmentation with other agents is limited.
Considerations
-
There are limited numbers of studies that have evaluated the efficacy of various treatment strategies in patients with schizophrenia responding poorly to clozapine.
-
Most of the studies were considered underpowered and limited by small sample sizes.
-
There is a need to expand the data by designing better quality studies.
Introduction
About 30% of patients with schizophrenia fulfil the criteria of treatment-resistant schizophrenia (TRS) (i.e., fail to respond to two adequate trials of antipsychotics) (Souza et al., Reference Souza, Kayo, Neto, Elkis and Buckley2010; Porcelli et al., Reference Porcelli, Balzarro and Serretti2012). For patients with TRS, clozapine is considered the mainstay of treatment (Šagud, Reference Šagud2015; Leung et al., Reference Leung, Gadelrab, Ntephe, McGuire and Demjaha2019). Evaluation of data of patients with TRS suggests that 30–70% of patients fail to respond to clozapine (Taylor et al., Reference Taylor, Shapland, Laverick, Bond and Munro2000; De Berardis et al., Reference De Berardis, Fornaro, Anastasia, Vellante, Valchera, Cavuto, Perna, Di Nicola, Serafini, Carano, Pompili, Orsolini, Tomasetti, Di Emidio, Martinotti, Di Giannantonio and Kim2019; Roerig, Reference Roerig2019; Kudva & Gupta, Reference Kudva and Gupta2016). This sub-group of patients are termed ‘clozapine non-responders’ or ‘clozapine-resistant’ (Souza et al., Reference Souza, Kayo, Neto, Elkis and Buckley2010; Barber et al., Reference Barber, Olotu, Corsi and Cipriani2017) or having ‘ultra-resistant’ schizophrenia (Porcelli et al., Reference Porcelli, Balzarro and Serretti2012; Elkis and Meltzer, Reference Elkis and Meltzer2007). These patients are challenging to manage and often lead to frustration among the clinicians. Although the available treatment guidelines suggest multiple treatment options for this sub-group of patients, the evidence for various suggested options is meagre.
The treatment strategies which have been evaluated for this group of patients include augmentation of clozapine with a second antipsychotic medication, mood stabiliser, antidepressant, other groups of medications such as memantine and glycine, electroconvulsive therapy (ECT), repetitive transcranial magnetic stimulation (rTMS), transcranial direct current stimulation (tDCS), Occupational therapy, and cognitive behaviour therapy (CBT) (Grover & Sahoo, Reference Grover and Sahoo2019). A few reviews have also evaluated the data for the use of these treatments in clozapine non-responders (Barber et al., Reference Barber, Olotu, Corsi and Cipriani2017; Buchanan et al., Reference Buchanan, Kreyenbuhl, Kelly, Noel, Boggs, Fischer, Himelhoch, Fang, Peterson, Aquino and Keller2010; Wang et al., Reference Wang, Zheng, Li, Wang, Cai, Yang, Ungvari, Xiang and Correll2018; Barbui et al., Reference Barbui, Signoretti, Mulè, Boso and Cipriani2009; Wagner et al., Reference Wagner, Löhrs, Siskind, Honer, Falkai and Hasan2019). There are specific studies on intervention modalities like antiepileptic drugs and second-generation antipsychotics (Zheng et al., Reference Zheng, Xiang, Yang, Xiang and de Leon2017; Bartoli et al., Reference Bartoli, Crocamo, Di Brita, Esposito, Tabacchi, Verrengia, Clerici and Carrà2019). The most recent meta-analysis on the topic was published in 2019 (Wagner et al., Reference Wagner, Löhrs, Siskind, Honer, Falkai and Hasan2019). In a recent literature review, authors reported the identification of 21 reviews and meta-analyses, which have evaluated the evidence for clozapine augmentation or combination strategies. This review showed that the maximum number of trials included in any meta-analysis was 46 in numbers (Siskind et al., Reference Siskind, Lee, Ravindran, Zhang, Ma, Motamarri and Kisely2018). One of the meta-reviews reviewed the data from 21 reviews and meta-analyses by using the Scottish Intercollegiate Guidelines Network (SIGN) criteria to grade the available evidence to the four different levels and concluded that none of the treatment strategies met the grade A criteria (Wagner et al., Reference Wagner, Löhrs, Siskind, Honer, Falkai and Hasan2019; Harbour & Miller, Reference Harbour and Miller2001). However, the combination of clozapine with first- or second-generation antipsychotics, augmentation of clozapine with ECT for persistent positive symptoms, and use of clozapine with certain antidepressants (fluoxetine, duloxetine, citalopram) for persistent negative symptoms were reported to have grade B level evidence (Wagner et al., Reference Wagner, Löhrs, Siskind, Honer, Falkai and Hasan2019; Harbour & Miller, Reference Harbour and Miller2001). The evidence for augmentation of clozapine with lithium and anticonvulsants, glutamatergic agents, rTMS, tDCS, or CBT met grades C–D criteria only. Many of the previous meta-analyses and reviews have not addressed the risk of bias adequately, and all of these have not included studies evaluating the efficacy of non-pharmacological measures like CBT and occupational therapy in the meta-analysis. It is important that evidence-based approach for different augmentation strategies, the quality of evidence for each specific strategy needs to be analysed thoroughly, and the risk of bias needs to be documented separately. In this background, this systematic review and meta-analysis aimed to assess the evidence for augmentation of clozapine with various strategies and to assess the risk of bias in randomised controlled trials (RCTs) in patients not showing adequate response to clozapine.
Methods
Search strategy
For the identification of relevant literature, electronic searches were done using Google Scholar, PubMed and Science Direct, and Scopus databases. The keywords included clozapine, augment*, resistan*, non-respon*, schizophrenia, psychosis, clozapine resistant, and clozapine refractory in various permutations and combinations. Initially, the word ‘clozapine’ was combined with each of these terms (e.g., augment*), and then the terms were substituted one after the other, with clozapine remaining as a search term. In the later searches, permutations and combinations of more than two words were used. The searches were carried out in April 2022. Additional literature was identified from the cross-references and through hand searches.
Selection of studies
Placebo-controlled RCTs published in English-language peer-reviewed journals that evaluated the efficacy/effectiveness of interventions in patients with clozapine resistance were included. The studies which did not present separate data for intervention were excluded. Similarly, studies for which effect sizes could not be computed with the available data were also excluded.
Data extraction
Abstracts of the identified studies were sifted through and the full texts of the papers were obtained. The extracted data were coded independently by two investigators (SS and SS). Information pertaining to study characteristics, demographic characteristics, and clinical details was extracted. Discrepancies between the investigators were resolved after a mutual discussion, involving the third author (SG) and a re-evaluation of the published paper.
Risk of bias
The Jadad scale (Jadad et al., Reference Jadad, Moore, Carroll, Jenkinson, Reynolds, Gavaghan and McQuay1996), which has good validity and reliability (Clark et al., Reference Clark, Wells, Huët, McAlister, Salmi, Fergusson and Laupacis1999), is a five-point scale for assessing the quality of RCTs and was used to quantify the risk of bias. The rating is done based on reporting of randomisation, blinding, and reporting of withdrawals and drop-outs.
Statistical analysis
Effect sizes and 95% confidence intervals (CI) for the efficacy of the intervention were determined. If any study reported more than one outcome measure, then the primary outcome measure or the measure considered to be most clinically relevant was used to calculate effect size. If any study did not explicitly mention a primary efficacy variable, then the proportion of patient’s improved or the improvement in the positive psychotic symptoms was taken for computation of effect sizes. Hedges g test, which gives a robust measure for both categorical and continuous data, was used to calculate effect sizes. The logit method was used for determining the effect size and confidence intervals for the dichotomous variables.
The obtained effect sizes were entered into OpenMetaAnalyst software for the computation of standardised mean differences. Random effects models were used due to the heterogeneity of studies. The standardised mean differences and effect sizes were computed for each of the following categories of interventions: antipsychotics, antidepressants, mood stabilisers and anticonvulsants, memantine, amino acids and alternate treatments, ECT, rTMS, and other interventions. For assessing the heterogeneity, I2 test was used. Forest plots were constructed to represent the studies graphically.
Results
Initial search of the literature showed 6271 studies. Abstracts of all these studies were screened for eligibility and 60 abstracts were considered to be relevant. We excluded the articles which reported findings of the meta-analysis, open-label studies, retrospective studies, review articles, treatment guidelines, on-going clinical trials/ study protocols, non-placebo-controlled comparative studies, articles published in languages other than English, case series, case reports, and those articles for which full texts were not available. An attempt was made to assess the full text of all these studies. Unfortunately, we could not access the full texts of six of the articles involving valproate augmentation. Of the remaining 54 studies, 9 could not be included, as the data were insufficient to calculate the effect sizes. After excluding all these studies, we were left with 45 placebo-controlled RCTs included in the meta-analysis.
Antipsychotics
The antipsychotics that have been commonly evaluated as the augmenting agent to clozapine include risperidone, aripiprazole, haloperidol, sulpiride, and ziprasidone, with the maximum number of studies on risperidone (Table 1 and Supplementary Figure 1). The sample sizes across different studies ranged from 6 to 207, and the duration of trials ranged from 8 weeks to 24 weeks. The most commonly used primary outcome measure in these trials was Brief Psychiatric Rating Scale (BPRS) or Positive and Negative Syndrome Scale (PANSS). When the meta-analysis of these trials was carried out, it was seen that overall antipsychotics were not more effective than placebo (standardized mean difference = 0.103, 95% CI −0.288 to 0.493, p < 0.001, I2 = 85%). This was mainly due to negative findings from one of the studies on risperidone. The pooled effect size for the five risperidone trials was −0.274 (95% CI −1.261 to 0.714, I2 = 91%). Similarly, the pooled effect size for the three aripiprazole trials was 0.574 (95% CI from 0.112 to 1.037, I2 = 67%).
Trials of antipsychotic augmentation of clozapine
BPRS: Brief Psychiatric Rating Scale; PANSS: Positive and Negative Syndrome Scale; Ris: Risperidone; Ami: Amisulpride; Sul: Sulpiride; Pim: Pimozide; Ser: Sertindole; Arip: Aripiprazole; H: Haloperidol; Zip: Ziprasidone; M: Males; F: Females.
Mood stabilisers
In terms of mood stabilisers, studies have evaluated lamotrigine, topiramate, and valproate as the augmenting agents in the descending order of frequency (Table 2 and Supplement Figure 2). Outcome measures used in these studies were PANSS and clinician-rated improvement status. The sample sizes in these studies have ranged from 4 to 80, and the duration of trials varied from 10 to 24 weeks. Overall, when compared to placebo, the mood stabilisers and anticonvulsants were found to be more effective (standardized mean difference = 0.265, 95% CI 0.016–0.515, p = 0.194, I2 = 28%). The pooled effect size for lamotrigine was 0.145 (95% CI −0.154 to 0.445, I2 = 0%), and that for topiramate was 0.392 (95% CI −0.045 to 0.830, I2 = 56%).
Trials of mood stabiliser and anticonvulsant augmentation of clozapine
PANSS: Positive and Negative Syndrome Scale; PANSS-P: Positive symptom subscale of PANSS; T: Topiramate; Lamo: Lamotrigine.
Antidepressants
In terms of antidepressants, only three RCTs have evaluated antidepressants for augmentation of clozapine (Supplement Table 1, Supplement Figure 3). While one study with mirtazapine and another with duloxetine showed significant improvement in BPRS and PANSS scores, respectively, another study evaluated fluoxetine and did not find a significant change in BPRS positive symptoms score. Overall, antidepressants were found to be conclusively better than placebo (standardized mean difference = 2.062, 95% CI 0.187–3.938, p < 0.001, I2 = 92.1%). However, it must be remembered that the analysis was limited to only three studies. Among the antidepressant, largest effect size was seen for mirtazapine.
Other agents
Few studies have evaluated memantine, amino acids, and alternate treatments for the augmentation of clozapine (Table 3 and Supplementary Figure 4). Memantine and glycine were evaluated in two trials, while D-cycloserine, D-serine, sarcosine, ampakine CX516, and ginkgo biloba were evaluated by one study each. The sample sizes for these studies ranged from 16 to 52, and the duration of the trials ranged from 4 weeks to 13 weeks. PANSS, BPRS, and Scale for Assessment of Negative Symptoms (SANS) were used to evaluate the outcome. These interventions were not seen to be conclusively effective (standardized mean difference = 0.378, 95% CI −0.027 to 0.784, p = 0.006, I2 = 57%). The two studies of memantine had a pooled effect size of 0.948 (95% CI −1.689 to 3.584, I2 = 94%), while the three studies of glycine had an effect size of 0.118 (95% CI −0.399 to 0.635, I2 = 0%).
Trials of memantine, amino acids, and alternate treatments for augmentation of clozapine
BPRS: Brief Psychiatric Rating Scale; PANSS: Positive and Negative Syndrome Scale; PANSS-P: Positive symptom subscale of PANSS; SANS: Scale for the assessment of negative symptoms; Mem: Memantine; Sar: Sarcosine; Ser: D-Serine; Gly: Glycine, G: Gingko Biloba; Mod: Modafinil; Mino: Minocycline; M: Males; F: Females.
Somatic and psychosocial treatments
Some RCTs have evaluated ECT, rTMS, occupational therapy, and CBT (Table 4 and Supplementary Figure 5). The duration of these interventions has ranged from 10 days (for rTMS) to 6 months (for occupational therapy). Two studies used BPRS, two used PANSS, and the occupational therapy trial used the Scale for Interactive Observation in Occupational Therapy as the outcome measure. Among these RCTs, trials evaluating ECT, one out of two studies of rTMS, and trials of occupation therapy and CBT showed that these active interventions were more effective than the control conditions. Overall, these interventions were seen to be effective (standardized mean difference = 0.781, 95% CI 0.401–1.161, p = 0.581, I2 = 0%).
Trials of other interventions for augmentation of clozapine
OT: occupational therapy; CBT: cognitive behaviour therapy; BPRS: Brief Psychiatric Rating Scale; PANSS: Positive and Negative Syndrome Scale; EOITO: Scale for Interactive Observation in Occupational Therapy; BF: Befriending; M: Males; F: Females; TAU: treatment as usual.
The various specific interventions are presented in Supplementary Figure 6 which shows the sub-group analysis which suggests that sulpiride, ziprasidone, mirtazapine, D-cycloserine, ampakine X516, ECT, occupation therapy, and CBT are effective (the confidence intervals did not span zero) augmentation strategies in patients not responding to clozapine.
Exploratory analyses were conducted to see whether there was an effect of the study duration, sample size, or outcome measure on the efficacy. The sample size (tau = 0.212, p = 0.041) had some relationship with the effect size (i.e. larger studies tended to have smaller effect size, implying the possibility of file drawer effect). The type of outcome measure (coded into BPRS, PANSS, and others) did not relate to the effect sizes.
Funnel plot of all included studies
The Egger test of asymmetry of all the studies suggested no significant evidence of publication bias (B = 0.569, p = 0.209) (Supplementary Figure 7). The visual inspection of the studies does suggest the possibility of publication bias.
Risk of bias
When the risk of bias was evaluated by using Jadad score, the scores ranged from 1 to 5 with a median of 4 (mean of 3.93, mode 5) (Supplementary Table 2). Only six studies had Jadad score of less than 3 and none had a score of 0. The maximum number of studies had used intention to treat analysis, while 15 studies (n = 15) had not relied on such statistical procedures or had not mentioned using statistical methods to estimate intention to treat (Muscatello et al., Reference Muscatello, Bruno, Pandolfo, Micò, Bellinghieri, Scimeca, Cacciola, Campolo, Settineri and Zoccali2011a; Mossaheb et al., Reference Mossaheb, Sacher, Wiesegger, Klein, Spindelegger, Asenbaum, Dudczak and Kasper2006; Zoccali et al., Reference Zoccali, Muscatello, Bruno, Cambria, Micò, Spina and Meduri2007; Vayısoğlu et al., Reference Vayısoğlu, Anıl Yağcıoğlu, Yağcıoğlu, Karahan, Karcı, Gürel and Yazıcı2013; Muscatello et al., Reference Muscatello, Bruno, Pandolfo, Micò, Scimeca, Di Nardo, Santoro, Spina and Zoccali2011b; Behdani et al., Reference Behdani, Hebrani, Rezaei Ardani and Rafee2011; Tiihonen et al., Reference Tiihonen, Halonen, Wahlbeck, Repo-Tiihonen, Hyvärinen, Eronen, Putkonen, Takala, Mehtonen, Puck, Oksanen, Koskelainen, Joffe, Aer, Hallikainen, Ryynänen and Tupala2005; Buchanan et al., Reference Buchanan, Kirkpatrick, Bryant, Ball and Breier1996; Zoccali et al., Reference Zoccali, Muscatello, Cedro, Neri, La Torre, Spina, Di Rosa and Meduri2004; Veerman et al., Reference Veerman, Schulte, Smith and de Haan2016; Potkin et al., Reference Potkin, Jin, Bunney, Costa and Gulasekaram1999; Evins et al., Reference Evins, Fitzgerald, Wine, Rosselli and Goff2000; Diaz et al., Reference Diaz, Bhaskara, Dursun and Deakin2005; Goff et al., Reference Goff, Leahy, Berman, Posever, Herz, Leon, Johnson and Lynch2001; Dean et al., Reference Dean, Mancuso, Bush, Copolov, Do, Cuénod, Conus, Rossell, Castle and Berk2015).When the association of the year of publication and risk of bias (measured with Jadad scale) were evaluated, a statistically significant correlation (Kendall tau correlation of 0.312, P = 0.008) was seen.
Discussion
Clozapine resistance poses a major clinical challenge to clinicians. Clinicians are always on the lookout for choosing an agent that can improve the clinical outcome of the patient. This systematic review and meta-analysis suggest that there are limited numbers of RCTs that have evaluated various agents among patients with clozapine resistance. It is evident from the data that most of these studies are of small sample size, with the largest study sample being of 487. Further, the limitation of the existing literature, which must be considered while interpreting the results of this systematic review and the meta-analysis, includes inconsistency in defining clozapine resistance. The definitions have varied significantly across different studies. Hence, findings of the present meta-analysis must be construed in this background.
One of the recent meta-analyses on the same topic included 46 RCTs, with searches in both Chinese and non-Chinese databases. In contrast, the present meta-analysis was limited to studies published in English only. However, when one compares the studies included in the current meta-analysis with this meta-analysis, there is only an overlap of 26 studies. We did not include 15 studies published in the Chinese language and we did not include another study, as it did not provide information separately for clozapine (Berk, Reference Berk2014). Accordingly, the present meta-analysis provides broader details on the studies published in the English language.
When one evaluates the effect size of interventions, it is evident that the effect size for various antipsychotic medications was 0.103; however, when the effect size was evaluated for specific antipsychotics for which more than one trial was available, the effect size for risperidone was −0.274 (negative effect size primarily due to one study) (Yağcioğlu et al., Reference Yağcioğlu, Kivircik Akdede, Turgut, Tümüklü, Yazici, Alptekin, Ertuğrul, Jayathilake, Göğüş, Tunca and Meltzer2005) and that for aripiprazole was 0.574. The effect size for mood stabilisers was 0.265, with that for lamotrigine being 0.145, and that for topiramate being 0.392 (with either being not significantly better than placebo). The effect size for glycine was 0.118 and the effect size for ECT was 0.743. In general, it is suggested that effect size of 0.2 is considered to be small, 0.5 is considered to be medium, and 0.8 is considered to be large. Accordingly, it can be said that among the various interventions, the effect size is highest for mirtazapine, followed by D-cycloserine. However, it is important to note that these were based on single small sample size study. Accordingly, the findings must be interpreted in this background. The previous meta-analysis suggests that lamotrigine may be of some benefit among patients with clozapine resistance for positive symptoms and general psychopathology symptoms, but not for negative symptoms (Zheng et al., Reference Zheng, Xiang, Yang, Xiang and de Leon2017; Tiihonen et al., Reference Tiihonen, Wahlbeck and Kiviniemi2009). Present meta-analysis provides further credence to these findings. The high effect size for ECT suggests that it should be preferred in patients not responding to clozapine if there are no contraindications for ECT. However, it is important to note that this is also based on a single study, but the sample size of this study was larger than that for mirtazapine and D-cycloserine. If we attempt to evaluate the efficacy of various agents as per the SIGN criteria, findings of the present study support the conclusion of the recent review of all the available systematic reviews and meta-analysis, which suggest that ECT has evidence of grade B (Wagner et al., Reference Wagner, Löhrs, Siskind, Honer, Falkai and Hasan2019). However, our meta-analysis contradicts the assertion that the level of evidence for anticonvulsants, that is, lamotrigine, is that of grade C or D (Wagner et al., Reference Wagner, Löhrs, Siskind, Honer, Falkai and Hasan2019).
Present meta-analysis shows that the addition of a second antipsychotic may not be useful in clozapine non-responders, though certain antidepressants may help. Some of the previous meta-analyses, (Barber et al., Reference Barber, Olotu, Corsi and Cipriani2017) which have also limited themselves to RCTs, suggest that in contrast to placebo, addition of second antipsychotic medication in patient with inadequate response to clozapine does not lead to significant improvement in positive symptoms, as suggested by small effect sizes in the present study, which were non-significant too. Further, the evidence from the available meta-analyses also offers a high degree of heterogeneity with minimal or no publication bias. Present meta-analysis and systematic review support the existing literature.
Several novel antipsychotics and other approaches are likely to be used in clozapine-resistant schizophrenia in the future. rTMS has been evaluated, and possibly other neuromodulation techniques may find considered for evaluation of clozapine-resistant cases. The adjunctive therapies may include approaches to reduce distress due to psychotic symptoms. In fact, the largest study for clozapine augmentation was on CBT. Assessment of outcomes has generally been done using PANSS and BPRS scales. Uniform application of these instruments and reporting their findings is likely to help maintain comparability across studies.
Findings of the present meta-analysis are limited by the existing literature. We could not include some RCTs, especially those involving valproate, due to non-availability and some other studies due to lack of data to calculate effect sizes. We have used the Jadad criteria for the assessment of the risk of bias, while other methods like Cochrane recommendations have become more popular now. The Egger’s test may not have picked up publication bias due to low sample sizes. Additionally, we did not evaluate the inter-rater agreement during the search process.
To conclude, based on the findings of the present systematic review and meta-analysis, it can be said that ECT or antipsychotics maybe preferred over other treatment strategies, and if this is not feasible, then lamotrigine must be considered much before the use of other agents.
Prisma flow diagram.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/neu.2022.30
Acknowledgements
None.
Author contributions
SG, SS, and SS contributed to the conception and design of, or acquisition of data or analysis and interpretation of data, drafting the article and revising it critically and provided final approval of the version to be published.
Financial support
None.
Conflict of interest
None.
