Early intervention services (EIS) have been shown to improve outcomes in first-episode psychosis (FEP).Reference Correll, Galling, Pawar, Krivko, Bonetto and Ruggeri1 Key components of EIS generally include case management, low-dose second-generation antipsychotics and family intervention.Reference Malla, Norman, McLean, Scholten and Townsend2 However, most of the evidence for superior effectiveness of EIS comes from high-income countries (HICs).
Past reports of better outcomes in psychotic disorders in low- and middle-income countries (LMICs) compared with HICsReference Harrison, Hopper, Craig, Laska, Siegel and Wanderling3–Reference Jablensky, Sartorius, Ernberg, Anker, Korten and Cooper6 have been criticisedReference Cohen, Patel, Thara and Gureje7 for not taking into account mortality using standard mortality ratios, the underreporting of suicides in LMICs, the lack of representative samples and inadequately measured outcomes. Despite these criticisms, this observation may hold true for IndiaReference Hopper and Wanderling4 for patients who actually receive adequate treatment.
Whether the differences in outcomes reported in the past between India and HICs would persist if similar EIS principles and treatment components were applied in both settings has never been examined. Hypotheses focused on cultureReference Kulhara and Chakrabarti8 have been proposed to explain any differences in outcome. However, what might facilitate better outcomes within a culture remains unexplored. The impact of overall social support on outcome has been well documented.Reference Gayer-Anderson and Morgan9 A greater role of families, as a specific and deeper form of social support, in India might be one important facilitator, given the context of a family structure in which the individual remains integrated with and supported by the family amidst a general lack of state support.Reference Nunley10
We have previously confirmed differences in clinical and functional outcomes in a pilot studyReference Iyer, Mangala, Thara and Malla11 using relatively small FEP samples in Montreal (Canada) and Chennai (India). On the basis of these findings, we conducted a larger study across the two sites, comparing multiple outcomes that were established a priori and included clinical, functional and subjective domains. Here we report only on clinical outcomes (symptoms). We hypothesised that (a) following application of key components of EIS, clinical outcomes would be better for patients in Chennai compared with those in Montreal; and (b) that such differences would be explained by differences in previously known predictors of outcomes, especially family support.
Method
Research and treatment context
This longitudinal, 2-year prospective outcomes study was conducted between 2012 and 2018 in Montreal (population 3.2 million), Canada, and Chennai (population 6 million), India. Within Montreal's McGill University network, the Prevention and Early Intervention Program for Psychoses (PEPP) is comprised of a larger service situated (since 2003) at the Douglas Mental Health University Institute and a smaller service at the McGill University Health Centre.Reference Iyer, Jordan, MacDonald, Joober and Malla12 The former serves a defined catchment area (total population 300 000) in south-west Montreal, with a significant proportion from minority ethnic groups (30%), and the latter mostly serves central Montreal (no defined catchment area), with a sizable student population. The two services are attached to a psychiatric and a general hospital respectively, with designated in-patient beds, and operate largely as out-patient community programmes.
Chennai, a large metropolis in South India, has a predominantly Tamil-speaking population with literacy rates of 90%. It hosts the Schizophrenia Research Foundation (SCARF), which is a non-governmental organisation and a World Health Organization research centre. Not being a catchment area-based service, SCARF accepts patients from all sources and all parts of Chennai. Its capacity for EIS was built collaboratively between 2006 and 2008 and adapted to sensitivities of the Indian cultural context and resource constraints, prior to the pilot project (2008–2010). It does not have an emergency psychiatric service but has access to beds when needed. We acknowledge that neither Chennai nor Montreal can be considered as representing the entire world of urban India or urban Canada respectively.
Both PEPP and SCARF followed EIS protocols for treatment of FEP, i.e. second-generation antipsychotic medications in lowest effective doses, case management, family psychoeducation and individual family intervention, cognitive-behavioural therapy when indicated and an overall recovery orientation.Reference Iyer, Mangala, Thara and Malla11 Services, including medication, were available to patients free of charge at both sites.
All study procedures complied with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008; and were approved by the Institutional Review Board at SCARF, Chennai, and the Research Ethics Board at McGill University, Montreal.
Participants
Inclusion criteria: an FEP defined as meeting DSM-IV-TR13 criteria for a primary DSM-IV-TR13 diagnosis of either schizophrenia-spectrum psychotic disorder or affective psychosis irrespective of the time since onset of psychosis, and not having received antipsychotic medication for more than 30 days since the onset of psychosis; age 16–35 years; and ability to communicate fluently in Tamil or English in Chennai, and French or English in Montreal.
Exclusion criteria: IQ below 70; psychosis secondary to a central nervous system disorder (e.g. epilepsy) or a medical condition; and a primary diagnosis of substance dependence. Patients who met the inclusion criteria but had a comorbid diagnosis of substance misuse were included.
Admission criteria for the two EIS were the same as the study entry criteria, except for the Montreal network also admitting 14- to 16-year-olds.
Sample size calculation: for clinical outcomes, using pilot dataReference Iyer, Mangala, Thara and Malla11 on symptom assessments, we assumed a difference of 15% in reduction of positive and negative symptoms between the two samples. We found that a total sample of 200 would allow us to detect a significant time × site interaction with 90% power at a type I error of 0.05. Conservatively factoring in attrition rates of 15% at SCARF and 24% at PEPP, we had estimated that a sample of 150 would be required at each site to observe significant differences in outcome. Supplementary Fig. 1, available at https://doi.org/10.1192/bjp.2020.126, provides details of patient recruitment. All consecutive patients (248 in Montreal; 244 in Chennai) entering the treatment programmes were assessed for inclusion in the study. Those meeting the inclusion criteria were approached for participation in the study. The final sample comprised 333 individuals with previously untreated FEP (n = 168 in Chennai; n = 165 in Montreal). Informed consent was obtained from all participants. Participants were offered a 24-month treatment programme.
Assessments
Sociodemographic and clinical information
Diagnosis was established through the Structured Clinical Interview for DSM-IV (SCID).Reference First, Spitzer, Gibbon and Williams14 Sociodemographic information, age at onset of psychosis and duration of untreated psychosis (DUP) were obtained by administering the Circumstances of Onset and Relapse Schedule (CORS).Reference Malla, Norman, Schmitz, Manchanda, BÉChard-Evans and Takhar15 DUP was defined as the number of weeks between the onset of psychosis and the initiation of antipsychotic medication. The suitability of the CORS in cross-cultural research and interrater reliability between raters at the two sites were established on 10 randomly selected cases (intraclass correlation coefficient ICC = 0.89–0.97) in the pilot study.Reference Iyer, Mangala, Thara and Malla11
Symptom assessments
The Scale for Assessment of Positive Symptoms (SAPS)Reference Andreasen16 and the Scale for Assessment of Negative Symptoms (SANS)Reference Andreasen17 were used to rate symptoms at baseline and at months 2, 3, 6, 12, 18 and 24. Positive and negative symptom severity were based on total SAPS and SANS scores respectively. The latter excluded scores on the items of ‘inappropriate affect’ and ‘poverty of content of speech’ and the items for the ‘attention’ subscale, as these have been shown not to be part of the negative symptoms domain.Reference Malla, Norman, Williamson, Cortese and Diaz18
Medication
All participants were prescribed a second-generation antipsychotic medication, with the specific choice based on clinical consideration and availability. Prescription of other psychotropic medications was based on clinical need. Adherence to medication was recorded monthly on the basis of information obtained from participants and/or their family members. Participants were regarded as adherent in a given month if they took the antipsychotic medication 76–100% of the time in that month. This method has been validated and used in other studies.Reference Cassidy, Rabinovitch, Schmitz, Joober and Malla19,Reference Jordan, Lutgens, Joober, Lepage, Iyer and Malla20 Percentage of adherence was calculated by dividing the number of months that the participant was adherent by the number of months for which they were prescribed antipsychotic medication, multiplied by 100.
Remission
The Remission in Schizophrenia Working Group's consensus criteriaReference Andreasen, Carpenter, Kane, Lasser, Marder and Weinberger21 were used to define remission. Participants were in positive or negative remission if they scored 2 or less on each global subscale of ‘hallucinations, delusions, bizarre behaviour and formal thought disorder’ on the SAPS or ‘flat affect, alogia, avolition and apathy; and anhedonia’ on the SANS respectively. Remission status was based on the symptom severity criterion every month. The total duration of positive, negative and total remission during the 2-year period was calculated for each participant on the basis of repeated symptom ratings at baseline and months 2, 3, 6, 12, 18 and 24. The cumulative duration of symptom remission has been reported to be a strong predictor of functional (occupational and social) outcomes.Reference Jordan, Veru, Lepage, Joober, Malla and Iyer22 For missing data, ratings at each time point were carried forward for the next period for up to a maximum of 6 months. Clinical outcome was assessed using three variables: change in symptom severity over 24 months, duration of remission and remission status at month 24. Early remission was defined as meeting remission criteria by month 3.
Family support
Family support and the quality of family relationships were based on ratings on the following two questions, derived from the Wisconsin Quality of Life Index-Provider VersionReference Becker, Shaw and Reib23: ‘During the past four weeks, this person has received _____ support’, assessed on a Likert-type scale, with 1 being infrequent, 2 moderate and 3 good; and ‘How would you describe the quality of this person's relationship with his/her family in the last 4 weeks?’, assessed on a Likert-type scale from 0 ‘none’ to 5 ‘excellent’. This was evaluated at months 3, 12 and 24. This measure was chosen for its simplicity and its limited potential for cultural bias. This assessment was made about the participant by the service provider (typically the case manager) on the basis of observation and enquiry. The score on the support received was multiplied by the score on the quality of relationship to obtain a weighted score (range 0–15) as a measure of overall family support. Higher scores indicate higher levels of family support.
Training and interrater reliability across sites
Staff at both sites were rigorously trained during the pilot phase and the current study. Interrater reliability (IRR) was established using four videotaped interviews of participants (two from each site), which were rated on the SAPS and SANS by all raters. The Cronbach's alpha for global subscale scores on the SAPS ranged from 0.93 to 0.99, and from 0.86 to 0.97 for the SANS. This compares with pilot phase IRRs of 0.82–0.87 on the PANSS.Reference Iyer, Mangala, Thara and Malla11
Data analyses
Data analyses were carried out using SPSS 22.0 for Windows.24 The two samples were compared using two inferential statistics: independent samples t-tests for continuous variables and χ2 for categorical variables. Repeated-measures analysis of variance (R-MANOVA) was used to examine the effect of time and site on symptom severity, controlling for baseline severity of each symptom domain. Any emerging site differences were adjusted for the influence of family support at 3 months. Sensitivity analyses were repeated to account for missing data on family support.
Linear and logistic regressions were conducted to determine predictors of duration and status of remission respectively. To address the hypotheses of the present study, the variables were selected on the basis of differences between the sites: gender (male/female); presence of a comorbid substance use or dependence disorder (yes/no); diagnosis (schizophrenia spectrum disorder/affective psychosis); age at onset of psychosis; baseline SAPS/SANS scores; ‘early remission status’ (yes/no); level of family support; and site (Chennai versus Montreal). We also controlled for baseline remission status. Effect sizes were computed using Cohen's d and partial eta squared (ƞp2) and interpreted according to the following guidelines: small: Cohen's d = 0.2 and ƞp2 = 0.01; medium: Cohen's d = 0.5 and ƞp2 = 0.06; large: Cohen's d = 0.8 and ƞp2 = 0.14.Reference Cohen25,Reference Gray and Kinnear26 All multivariate analyses were carried out on participants who completed the study (supplementary Fig. 1).
Results
Baseline sociodemographic and clinical characteristics
Participants at both sites were in their mid-twenties, although significantly younger and more often male and single in Montreal than in Chennai (Table 1). Most participants in Montreal and almost all in Chennai lived with their families. For occupational status, a substantial proportion of the women in the Chennai sample described their full-time role as homemakers.
Sociodemographic and clinical characteristics at entry for treatmenta
SCID, Structured Clinical Interview for DSM-IV; DUP, duration of untreated psychosis; SAPS, Scale for Assessment of Positive Symptoms; SANS, Scale for Assessment of Negative Symptoms.
a. The sample sizes vary because of missing data.
b. Analysis on log of the mean.
P < 0.05 is significant.
Most participants had a primary diagnosis of schizophrenia-spectrum disorder, albeit significantly more in Chennai. A significantly higher proportion had a concurrent substance use disorder (mostly cannabis) in Montreal. Age at onset was significantly lower in Montreal, although there was no difference in DUP between the two sites. At baseline, participants in Montreal showed a significantly higher level of positive symptoms but no differences were observed on severity of negative symptoms.
A comparison of individuals included in the study and those who declined to participate (n = 35) in Montreal revealed no significant differences on demographic and clinical characteristics, except for a lower median DUP among non-participants (4 weeks) compared with participants (11.57 weeks). In Chennai, only 6 individuals declined participation. In Montreal 31 (18.8%) participants did not complete the study, compared with 2 (1.2%) in Chennai. A comparison of completers with non-completers in Montreal did not reveal any differences on any demographic or clinical characteristics except that the proportion with family contact was higher among the completers than non-completers (50 v. 29.3%, χ2 = 4.10, d.f. = 1, P = 0.043).
Medication
On entry to the programme, almost all participants in Chennai (166/168) and most participants in Montreal (141/165) were prescribed and agreed to take antipsychotic medication. The daily dose of antipsychotic (chlorpromazine (CPZ) equivalents)Reference Woods27,Reference Leucht, Samara, Heres and Davis28 at the start of treatment was higher in Montreal (mean 202.59 mg, s.d. = 146.44) compared with Chennai (mean 168.11 mg, s.d. = 78.04) (t = 2.671, d.f. = 186.401, P < 0.01). There were no significant differences in the use of other psychotropic medications. Olanzapine and aripiprazole were the most common antipsychotic medication used in Chennai (n = 111; 63.4%) and Montreal (n = 65, 38%) respectively.
Outcomes
Mortality
Four participants (2.38%), all women (age 27–34), died at the Chennai site, all within 3 months of entry to the programme, compared with none in Montreal over the entire 2-year period. One died from recurrence of remitted thyroid cancer and the other three from suicide.
Family support
Family support (supplementary Table S1) was higher in Chennai than in Montreal at month 3 (mean 9.39, s.d. = 4.25 v. mean 8.31, s.d. = 3.91), month 12 (mean 10.64, s.d. = 3.72 v. mean 8.49, s.d. = 4.07) and month 24 (mean 10.6, s.d. = 3.41 v. mean 8.79, s.d. = 3.46). There was an effect of time (F = 7.007, d.f. = 1.915, P < 0.001, ƞp2 = 0.03) and site (F = 14.05, d.f. = 1, P < 0.001, ƞp2 = 0.06) but no time × site interaction (F = 2.56, d.f. = 1.915, P = 0.08, ƞp2 = 0.01). Family support increased over time at both sites.
Decrease in positive and negative symptom severity
Results of R-MANOVA showed that for both positive and negative symptoms, there were significant main and interaction effects of time and site (Table 2). Participants at both sites improved significantly over the 24 months. However, the improvement was greater in Montreal for positive symptoms (ƞp2 = 0.19) and in Chennai for negative symptoms (ƞp2 = 0.08). This analysis controlled for baseline symptoms. The time × site interactions reported in Table 2 were adjusted for the effect of family support at month 3, the latter being the main variable of interest that might account for site differences. Following this adjustment, the interaction remained significant for negative symptoms (F = 3.95 (d.f. = 1.49), P < 0.03, ƞp2 = 0.20) and for positive symptoms (F = 54.33 (d.f. = 1.51), P < 0.001, ƞp2 = 0.21). These results suggest that the greater reduction of negative and positive symptoms in Chennai and Montreal respectively is independent of family support. In the case of negative symptoms, it may be enhanced by such support.
Repeated measures analysis of variance for change in symptom severity over time
ƞp2, partial eta square (effect size); SAPS: Scale for Assessment of Positive Symptoms; SANS: Scale for Assessment of Negative Symptoms.
P < 0.05 is significant.
Duration of remission
Over the 24-month period, the average duration of positive symptom remission was similar in Montreal (mean 15.53 months, s.d. = 7.46) and Chennai (mean 15.65 months, s.d. = 7.93) (t = −0.13, d.f. = 331, P = 0.89, Cohen's d = −0.01). On the other hand, compared with Montreal, participants in Chennai were in negative symptom remission for a significantly longer period (mean 16.10 months, s.d. = 7.56 v. mean 9.78 months, s.d. = 8.13) (t = −7.35, d.f. = 331, P < 0.001, Cohen's d = −0.80).
Is the greater duration of negative symptom remission in Chennai independent of other confounders?
Multiple linear regression analysis was conducted only to establish whether any of the differences between the two samples (Table 1) contributed to the difference in duration of negative symptoms, independent of site. The results of the linear regression (n = 236) indicated that attaining early negative symptom remission and site (Chennai) significantly predicted a longer duration of negative symptom remission and accounted for 45% of variance in outcome (Table 3).
Regression analyses for negative remission
To further explore the relationship between family support and duration of negative symptom remission, bivariate correlations with family support at each site revealed modest correlation at months 3 and 12 in Chennai (r = 0.31) and only at month 12 in Montreal (r = 0.26).
Remission status at 24 months
The percentage of participants in positive and negative symptom remission increased over the 2-year period at both sites. The percentage of participants in positive symptom remission at month 24 was comparable between Montreal and Chennai (82.6 and 83.3% respectively; χ2 = 0.33, d.f. = 1, P = 0.86), whereas only 60.3% of participants in Montreal were in negative symptom remission compared with 81.5% in Chennai (χ2 = 16.12, d.f. = 1, P < 0.001).
Are site differences in negative symptom remission status at month 24 explained by differences in other characteristics between the two samples?
As there was no difference in rate of positive symptom remission between the two sites, binary logistic regression analyses were used to examine this question for negative symptom remission status only. This analysis (n = 216) revealed that only early remission (higher rate in Chennai) of negative symptoms was an independent predictor of remission status at 24 months. Neither site nor family support accounted for this difference (Table 3).
Early remission and impact on differences in outcome
‘Early remission’ (by month 3) appears to be strongly and consistently associated with better outcome for duration of negative symptom remission as well as remission status at 24 months, independent of site. In Chennai 141 (84%) participants were in negative symptom remission at 3 months compared with 65 (39%) in Montreal (χ2 = 69.98, P < 0.001). Therefore, we conducted a logistic regression with early negative symptom remission as the dependent variable and characteristics generally associated with better outcome on negative symptoms as independent variables (Table 4). Our results show that early negative symptom remission was independently associated with shorter DUP and site (Chennai) and only marginally by not having comorbid substance misuse. Given that there was no significant difference in DUP across the two sites, it can be concluded that the difference in the rate of early remission of negative symptoms is associated with site alone.
Logistic regression analysis to examine predictors of early negative remission
Discussion
We had hypothesised that clinical outcomes would be better in Chennai than in Montreal. This was confirmed for negative but not for positive symptoms. Positive symptoms respond well to antipsychotic medication, especially in an EIS environment, as indicated by previous reports of high rates of remission (80–85%).Reference Malla, Norman, Schmitz, Manchanda, BÉChard-Evans and Takhar15 This appears to be the case in both settings. Indeed, the higher magnitude of improvement in positive symptoms in Montreal could very likely be the result of regression to the mean.
Better negative symptom outcomes reported in EIS compared with regular care are likely related to the integration of several psychosocial interventions with medication.Reference Lutgens, Gariepy and Malla29 Our results show that in addition to site (Chennai), the better outcomes on negative symptoms are associated with greater family support at 3 months in the case of reduction in symptoms over time and by higher rates of early negative symptom remission (within 3 months) in the case of duration of remission. On the other hand, the status of being in remission at 24 months is associated mostly with the higher rate of early remission in Chennai. Achieving early remission of negative symptoms seems to be a key ingredient for better outcome on negative symptoms. Our examination of the predictors of early remission suggests that this is driven largely by site and DUP, the latter being the same across the two sites. These results further confirm our first hypothesis of better outcome in the EIS in Chennai compared with the one in Montreal, but only for negative symptoms.
We also demonstrated that family support early on (at 3 months), which was higher in Chennai, further enhances better negative symptom outcome independent of site. Family support, however, increases with time at both sites and may contribute to better outcomes in Montreal as well, if increased sufficiently and early enough.
Our results suggest that there may indeed be something other than, or in addition to, family support that influences both early remission as well as its sustenance in Chennai. Our family support measure may not have captured nuances of family involvement. There may be other unexplored contextual aspects of family involvement or some other patient or service characteristics that explain inter-site differences in negative symptom outcome. Other inter-site differences may pertain to the trust of patients and families in treatment providers and attributions of relative responsibility for care to patients, families and the state.Reference Pope, Jordan, Venkataraman, Malla and Iyer30,Reference Pope, Malla and Iyer31
Our model explained a considerable proportion of variance (45%) in duration of negative symptom remission and suggests that characteristics that differentiate the EIS in Chennai from that in Montreal make up the ‘site’ variable and could be influencing the outcome. However, there are likely to be other unexplored variables that contribute to this difference. A substantial amount of unexplained variance may be attributable to individual variations across participants related to genetic or other neurobiological or environmental factors not accounted for in this model (e.g. social deprivation, history of trauma).Reference Suvisaari, Mantere, Keinänen, Mäntylä, Rikandi and Lindgren32,Reference Díaz-Caneja, Pina-Camacho, Rodríguez-Quiroga, Fraguas, Parellada and Arango33
Better outcome for negative symptoms in Chennai is unlikely to be related to differences in baseline symptom levels, as these were controlled for in our analyses. Variation in the rating of negative symptoms across the two sites is also unlikely to explain this difference, given the high level of interrater reliability on measurement of negative symptoms across the two sites. Although a comorbid substance misuse diagnosis did not contribute significantly to any of the regression models, we did not measure continuous use of cannabis over time. Very low rates of drug use in Chennai have also been reported in previous studies.Reference Thara34,Reference Chand, Thirthalli and Murthy35
Strengths
Our study has several strengths. The samples of patients with essentially untreated FEP at both sites were very well characterised, met the same inclusion and exclusion criteria and received treatment in an EIS with components of treatment that were similar within respective cultural and economic contexts. Considerable time and resources had been used in building capacity at the Chennai site during the pilot phase. Rigorous training of staff and regular evaluation of interrater reliability were conducted across the two sites.
Limitations
Differences in service structures may have resulted in participants in Montreal being more representative of a catchment area sample compared with those in Chennai. In Chennai, a large metropolitan area in a LMIC, participants and families may have come from a much larger and wider-spread population base. Participant characteristics varied across the two sites and may reflect true differences in the epidemiology of psychotic disorders in India and Canada (e.g. difference in gender distribution, age at onset and substance misuse) or sampling bias (severity of symptoms). However, these confounders did not explain the site differences in our multivariate analyses. Our controlling for all baseline differences between the two samples is likely to have minimised but not eliminated bias, which must be acknowledged in interpreting differences.
Implications
Our findings suggest that an EIS with essential elements of integrated care, including case management, contextualised to local circumstances of a LMIC urban environment can produce clinical outcomes that are even better than those in an urban setting in an HIC. The differences in outcome in FEP across the two very different economic and cultural environments are not attributable to differences in some known predictors of outcome that may apply across disparate environments. Clinicians may need to pay particular attention to achieving early remission of symptoms by addressing factors associated with it, while also being vigilant about risk of suicide, especially for women in India. Increasing family support and reducing treatment delay may lead to both early remission and improvements in longer-term outcome, across geographies. Improving family support early in the treatment of young people with FEP may be necessary in Western settings such as Montreal, where family involvement may not occur as naturally as in LMIC settings such as Chennai. There may indeed be a dose–response relationship between family support and outcome, and family support may need to be sustained over the entire period of treatment.
Supplementary material
Supplementary material is available online at https://doi.org/10.1192/bjp.2020.126.
Data availability
The data that support the findings of this study are available from the corresponding author, A.M., upon reasonable request.
Acknowledgements
We thank Megan Pope, Heleen Loohuis, Nicole Pawliuk, Kevin MacDonald, Emily Schorr, Helen Martin, and service providers, patients and families at PEPP, SCARF and PEPP-MUHC.
Author contributions
Study design and protocol: A.M., S.N.I., R.T., R.P., R.J., N.S.; implementation and data acquisition: S.N.I., G.M., A.M., R.P., R.J., H.C.M.; data analysis: A.T.; statistical input and interpretation: N.S., A.M., S.N.I., R.T., R.P., R.J.; first draft of the manuscript: A.M., A.T., S.N.I.; all authors contributed to several drafts and have approved the final manuscript.
Funding
This research was funded by an R-01 grant from the National Institutes of Health (NIH 5R01MH093303-05). The funding source had no further role in study design, data collection and analysis, writing of the report, and in the decision to submit the report for publication.
Declaration of interest
A.M. served as a research consultant to and gave lectures at conferences supported by Lundbeck and Otsuka and was on an advisory board meeting for those two companies, for which he received honoraria. S.I. reports grant and salary awards from the Canadian Institutes for Health Research and Fonds de recherche du Québec–Santé, outside the submitted work. H.C.M. reports grants from Acadia, Amgen and SyneuRx international; grants, honoraria and non-financial support from Janssen, Lundbeck and Otsuka; honoraria and non-financial support from Pfizer; honoraria from Sunovion, Purdue, HLS therapeutics, Mylan and Shire within the past 3 years, unrelated to the submitted work. R.J. served as speaker and member of advisory board committees for Pfizer, Janssen, BMS, Sunovian, Myelin, Otsuka, Lundbeck, Shire and Perdue; he also received grants from Janssen, BMS, Otsuka, Lundbeck, Astra Zeneca and HLS, within the past 3 years, unrelated to the submitted work.
ICMJE forms are in the supplementary material, available online at https://doi.org/10.1192/bjp.2020.126.
eLetters
No eLetters have been published for this article.