Major depressive disorder (MDD) is a debilitating mental disorder characterised by emotional, cognitive and physical symptoms that impact an individual's ability to function in multiple domains.Reference Chand and Arif1 The global impact of MDD cannot be understated, with it being one of the leading causes of disability.Reference Friedrich2 As of 2023, the World Health Organization estimates that approximately 5–6% of adults experience MDD around the world.Reference World Health3 This figure is expected to grow, as the incidence of MDD appears to be increasing steadily over time.Reference Akil, Gordon, Hen, Javitch, Mayberg and McEwen4 Making matters worse, traditional pharmacotherapies that act on monoamine neurotransmitter systems such as selective-serotonin-reuptake-inhibitors (SSRIs) and selective noradrenaline reuptake inhibitors (SNRIs)Reference Yoshimura, Hori, Ikenouchi-Sugita, Umene-Nakano, Ueda and Nakamura5 are often ineffective, of minimal benefit or poorly tolerated. Up to 50%Reference Akil, Gordon, Hen, Javitch, Mayberg and McEwen4 of people with MDD do not achieve a response with these medications and experience treatment-resistant depression (TRD). This suggests that other mechanisms beyond those directly involving monoamine neurotransmission are likely contributing to the development of depressive symptoms in a high proportion of people. Clearly a deeper understanding of the biology of MDD is urgently needed to facilitate the development of alternative treatments and address the unmet needs of individuals with TRD across populations and settings.
Converging evidence supports the role of dysregulation in the innate immune system in the aetiology of MDD, and elevated inflammation is thought to contribute to treatment resistance.Reference Orsolini, Pompili, Tempia Valenta, Salvi and Volpe6 C-reactive protein (CRP), an acute-phase reactant protein commonly used as a marker of peripheral inflammation, has been reported to be higher in patients with TRD compared with non-TRD patients.Reference Orsolini, Pompili, Tempia Valenta, Salvi and Volpe6,Reference Qiao, Geng, Qian, Zhu and Zhao7 Furthermore, higher levels of interleukin-6 (IL-6), a cytokine which induces CRP synthesis in the liver,Reference Bermudez, Rifai, Burning, Manson and Ridker8 have been detected in individuals with MDD whose symptoms were refractory to SSRIs and SNRIs when compared with those who responded to these medications.Reference Yoshimura, Hori, Ikenouchi-Sugita, Umene-Nakano, Ueda and Nakamura5 While standard antidepressant medications have some anti-inflammatory activity,Reference Patel, Keating and Dale9 they alone may not be adequate to address underlying inflammation in a subset of people with MDD. In theory, adjunctive medications to target inflammation may improve antidepressant treatment efficacy. However, clinical trials of repurposed anti-inflammatory agents in MDD and TRD have reported mixed results.Reference Husain, Strawbridge, Stokes and Young10 A recent systematic review and meta-analyses concluded that no clear recommendations could be made regarding the use of anti-inflammatories in MDD given the current state of research in this area.Reference Simon, Arteaga-Henríquez, Fouad Algendy, Siepmann and Illigens11 Inflammation has, however, been linked to individual features of depression, such as anhedonia, appetite dysregulation, sleep disturbances, low energy and suicidality.Reference Miola, Dal Porto, Tadmor, Croatto, Scocco and Manchia12 It is possible that anti-inflammatory medications may improve these symptoms individually, though research in this area is limited.
Furthermore, despite nearly 40% of the global population residing in lower-middle income countries (LMICs),13 most research on MDD, its treatment and pathophysiology, has involved populations in high income countries (HICs). The high burden of mental illness in LMICs, together with limited access to evidence-based mental health interventions,Reference Rathod, Pinninti, Irfan, Gorczynski, Rathod and Gega14 highlights the need for studies to identify effective and accessible psychiatric treatment protocols in these settings. The goal of this study is to determine the prevalence of inflammation in people with TRD from studies in a LMIC setting, and to explore associations between specific depressive symptoms and peripheral inflammation in this population.
Method
Study design
This study is a cross-sectional secondary analysis of data pooled from two multicentre randomised controlled trials (RCTs) in Pakistan, evaluating the safety and efficacy of adjunctive anti-inflammatory medications in adults with TRD. Treatment resistance in both trials was defined as non-response to treatment with two antidepressant medications for a minimum of 4 weeks at their minimum effective dosage during the current major depressive episode. Both trials followed British National Formulary and Maudsley Prescribing Guidelines for determining minimum effective medication dosages. CRP was measured at baseline and after 12 weeks of treatment in both studies.
The first trial was a multicentre, two arm, placebo-controlled RCT investigating simvastatin (20 mg daily) added to standard treatment (i.e. psychotropic medications and out-patient psychiatric care) versus placebo added to standard treatment (ClinicalTrials.gov Identifier: NCT03435744). Participants included 150 out-patients with MDD and TRD, aged 18–75 years treated at psychiatric clinics across five urban centres in Pakistan, with a score of 14 or higher on the 24-item Hamilton Depression Rating Scale (HAMD-24). Exclusion criteria included: diagnosis of psychotic disorder or bipolar disorder; current use of statins; unstable medical or neurological problems; autoimmune or inflammatory disorders; alcohol or substance use disorder; active suicidal ideation; and pregnancy or breastfeeding. This study was approved by The National Bioethics Committee of the Pakistan Health Research Council. Detailed methodology has previously been reported.Reference Husain, Chaudhry, Khoso, Kiran, Khan and Ahmad15
The second was a 12-week, multicentre, placebo-controlled RCT of minocycline (200 mg daily) added to standard treatment for TRD versus placebo added to standard treatment (ClinicalTrials.gov identifier: NCT02263872). The study included 41 out-patients aged 18–65 years treated at four out-patient psychiatric clinics in Karachi, Pakistan. Exclusion criteria included: relevant medical illnesses or concomitant penicillin therapy; concomitant anticoagulant therapy; presence of a seizure disorder; current use of valproic acid; diagnosis of substance use disorder; pregnant or breastfeeding; and the presence of a primary psychotic disorder. This study was approved by the ethics committee of the Karachi Medical and Dental College and Dow University of Health Sciences, Pakistan.Reference Husain, Chaudhry, Husain, Khoso, Rahman and Hamirani16
Ethics approval and informed consent was not required for the secondary analysis. Both original trials obtained ethics approval and written informed consent from research participants.Reference Husain, Chaudhry, Khoso, Kiran, Khan and Ahmad15,Reference Husain, Chaudhry, Husain, Khoso, Rahman and Hamirani16
Demographic and clinical variables
CRP was selected as the sole biomarker of inflammation in this study because of the availability of data across both clinical trials, as well as its known associations with MDD.Reference Dessoki, Khattab, Moris, Abdelhakim, Lotfy and Salah17 Inflammation was defined as blood CRP levels of greater than or equal to 3 mg/L. This cut-off was chosen to reflect guidelines from the United States Centers for Diseases Control and Prevention, and the American Heart Association, which considers CRP levels over 3 mg/L to be indicative of elevated inflammation and high risk for the development of cardiovascular disease.Reference Ridker18 This cut-off has previously been used to represent low-grade inflammation in other depression studies.Reference Chamberlain, Cavanagh, de Boer, Mondelli, Jones and Drevets19,Reference Jones, Mahmood, Hodsoll, Chaudhry, Khoso and Husain21
Demographic and clinical variables were included based on their availability in the entire sample across both trials. Demographic variables included age, gender, marriage status (‘single’, ‘married’ and ‘widowed or divorced’), years of education and socioeconomic status (SES) (‘lower’ and ‘middle/upper’). Lower SES was defined as a monthly income of 17 000 Pakistani rupees or lower. Clinical variables included weight in kilograms and body mass index (BMI), use of anticholinergic medication, typical and atypical antipsychotic or benzodiazepine medications, number of hospital admissions and clinical assessment scores. Clinical assessments included the Hamilton Depression Rating Scale (HAMD),Reference Hamilton22 General Anxiety Disorder-7 (GAD-7),Reference Spitzer, Kroenke, Williams and Löwe23 Clinical Global Impression scale (CGI)Reference Busner and Targum24 and Patient Health Questionnaire-9 (PHQ-9).Reference Kroenke, Spitzer and Williams25 Individual depressive symptoms were measured as individual item scores on the first 17 items of the HAMD.
Statistical analysis
The sample was divided based on inflammatory status into ‘high CRP’ (CRP ≥ 3 mg/L) and ‘non-inflamed’ (CRP < 3 mg/L) groups. Between-group comparisons were performed using Mann–Whitney U Test for continuous and ordinal variables and with Chi-squared or Fisher exact test for categorical variables.
Following a similar statistical strategy applied previously,Reference Jones, Mahmood, Hodsoll, Chaudhry, Khoso and Husain21 a series of logistic regression models were built to assess inflammatory status as a binary dependent variable (CRP ≥ or <3 mg/L), with one symptom, clinical or demographic variable included in each model as an independent variable. Age, gender, socioeconomic status and BMI were selected a priori as covariates due to their known influence on both depression and inflammation. CRP was also examined as a continuous variable using linear regression, with the same independent variables and covariates. The distribution of CRP was positively skewed and thus was normalised with log10 transformation. Statistical significance was defined as P-values of <0.05. Results from this exploratory analysis were not corrected for multiple comparisons.
Finally, we explored whether anti-inflammatory treatment improved symptoms associated with inflammation in logistic regression for participants with high CRP. Regression analysis was used for simvastatin treatment, with HAMD item score at 12 weeks as the dependent variable, and treatment group (placebo or simvastatin) and baseline HAMD item score as independent variables, with the covariates included. Response to treatment with minocycline was assessed using descriptive statistics due to the small number of participants from the original trial (n = 16) with CRP ≥ 3 mg/L. All statistical analyses were performed with statistical computing software R.26
Results
Sample and baseline characteristics
The total sample across both trials included 191 adult out-patients with MDD and TRD. Baseline CRP data were available for 171 participants. Of these, three were omitted because of missing weight data, leaving a total of 168 participants in the final sample.
Baseline characteristics and comparisons between high CRP and non-inflamed groups are summarised in Table 1. The final sample for the primary analysis included 92 females (55%) and 75 males (45%) with a mean age of 37.1 (9.6) years. Most participants were married (80%), with an average of 6.3 (5.4) years of education. A total of 60% of participants in the sample fell into the ‘lower’ socioeconomic class, and all participants were prescribed an antidepressant. In all, 87% (n = 146) had elevated plasma CRP at baseline.
Table 1 Baseline demographic and clinical variables
CRP, C-reactive protein; BMI, body mass index (weight in kilograms divided by height in metres squared); HAMD, Hamilton Depression Rating Scale; PHQ-9, Patient Health Questionnaire-9; GAD-7, General Anxiety Disorder-7; CGI, Clinical Global Impressions.
Statistical tests: a. Mann–Whitney U Test.
b. Chi-Squared Test.
c. Fisher's Exact Text.
Units: CRP measured in mg/L.
Demographic and clinical variables that were distributed unequally between high CRP and non-inflamed groups included age (P = 0.014) and number of hospitalisations (P = 0.021), with values being higher for individuals with high CRP.
Associations between clinical and demographic variables and high CRP in regression analysis
Results of logistic regression are summarised in Table 2 and Fig. 1. Symptoms that were associated with a higher likelihood of inflammation were: insomnia early in the night (HAMD item#4) (odds ratio 2.33, 95% CI 1.16–5.25, P = 0.028), early morning insomnia (HAMD item#6), (odds ratio 2.65, 95% CI 1.29–6.38, P = 0.015) and psychic anxiety (HAMD item#10) (odds ratio 3.79, 95% CI 1.39–21.7, P = 0.044). Gastrointestinal symptoms and/or low appetite (HAMD item#12) (odds ratio 0.38, 95% CI 0.14–0.86, P = 0.032) and general somatic symptoms (HAMD item#13) (odds ratio 0.34, 95% CI 0.14–0.74, P = 0.011) were negatively associated with inflammation. The variable ‘number of hospitalisations’ was heavily skewed because of a high number of 0 values (n = 128, 76%) and was thus omitted from logistic regression analysis. No other clinical or demographic variables were associated with CRP status in logistic regression. Logistic model plots and ROC curves are provided in Supplementary Figures 1 and 2, respectively.
Fig. 1 Estimated odds ratios for inflammation (C-reactive protein ≥ 3 miligrams per litre (mg/L).
Table 2 Associations between peripheral inflammation and depressive symptoms, clinical and demographic variables in logistic regression
HAMD, Hamilton Depression Rating Scale; PHQ-9, Patient Health Questionnaire-9; GAD-7, General Anxiety Disorder-7; CGI, Clinical Global Impressions.
Results of linear regression are difficult to interpret and may lack validity due to high variability in plasma CRP in the sample and heteroscedasticity in the residuals and are thus not presented here. A summary of results from linear regression is available in Supplementary Table 1 available at https://doi.org/10.1192/bjo.2024.735.
High variability in CRP may reflect the presence of active infection in some participants at the time of measurement, typically considered plasma CRP > 10 mg/L.Reference Nehring, Goyal and Patel27 While infection cannot be confirmed, we repeated the logistic and linear regressions while omitting people with baseline plasma CRP > 10 mg/L (n = 33) to account for this possibility. In logistic regression, the same symptoms remained significantly associated with CRP status, with results displayed in Supplementary Table 2. Symptom categories positively associated with CRP in linear regression were early waking insomnia (HAMD item#4) (B = 0.49, CI 0.15–0.84, P = 0.005), work and activities (HAMD item#7) (B = 0.36, CI 0.11–0.61, P = 0.005), psychomotor retardation (HAMD item#8) (B = 0.30, CI 0.04–0.56, P = 0.026), agitation (HAMD item#9) (B = 0.0.32, CI 0.04–0.59, P = 0.023) and psychic anxiety (HAMD item#10) (B = 0.0.47, CI 0.21–0.73, P ≤ 0.001). Symptoms negatively associated with CRP were general somatic symptoms (HAMD item#13) (B = 0.−0.15, CI −0.88 to −0.13, P = 0.008) and weight loss (HAMD item#16) (B = −0.53, CI −0.84 to −0.22, P = 0.001). CGI scores were positively associated with CRP (B = 0.38, CI 0.08–0.68, P = 0.012). The results of linear regression are summarised in Supplementary Table 3.
Associations between adjunctive anti-inflammatory treatments and change in symptom scores for symptoms associated with high CRP
Of the 146 participants in the sample with high plasma CRP, week 12 data for the above-mentioned HAMD items were missing from 13 participants, leaving a total of n = 133 participants in the treatment response analysis. Of these, 117 received adjunctive simvastatin or placebo, and 16 received adjunctive minocycline or placebo. The results of regression analysis are presented in Table 3. There were three levels to the symptom score for the variables insomnia early in the night (HAMD item#4) and early morning insomnia (HAMD item#6) (i.e. these symptoms could be scored either 0, 1 or 2), thus these variables were assessed using ordinal regression. In our pooled data-set, psychic anxiety (HAMD item #10) had two levels (anxiety scores were either 0 or 1 for every participant in the data-set) and was assessed using logistic regression. Response to treatment with adjunctive simvastatin did not differ significantly from placebo added to standard treatment for any depression symptoms analysed. With a small sample size of 16 participants in the minocycline/placebo group, we were underpowered to perform regression analysis for adjunctive minocycline treatment. Therefore, the number and proportion of people in each treatment group whose symptom scores improved, did not change or worsened are presented in Table 4. Fisher's exact test was used to compare differences in the proportion of individuals whose symptoms did and did not improve (those whose symptoms worsened or did not differ from baseline) between minocycline and placebo groups. Overall, a higher proportion of people who received adjunctive minocycline experienced improvements in all three symptoms relative to those who received placebo plus standard treatment. This was only significant for anxiety (P = 0.003), as all participants who received minocycline reported improvements, compared with just 22% in the placebo group. A heat map visualising the degree of change in symptoms scores post-treatment with minocycline compared to placebo shows greater reductions in symptom scores with minocycline (Fig. 2).
Table 3 Response to simvastatin treatment for inflammation-associated symptoms in regression analysis
HAMD, Hamilton Depression Rating Scale.
Statistical tests:
a. Ordinal regression.
b. Logistic regression.
Table 4 Response to treatment with minocycline
HAMD, Hamilton Depression Rating Scale.
Statistical tests: Fisher's exact test.
Fig. 2 Change in depression symptom scores pre to post-treatment with minocycline versus placebo for symptoms associated with high plasma CRP (≥3 mg/L)Reference Zazula, Husain, Mohebbi, Walker, Chaudhry and Khoso33. Heat map of change in HAMD item score for insomnia (early) (HAMD#4); insomnia (morning) (HAMD#6); and psychic anxiety (HAMD#10) with treatment with either minocycline or placebo. Numbers indicate the actual change in score from baseline to 12 weeks. Blue indicates an increase in symptom score, while red signifies reduction in score. HAMD, Hamilton Depression Rating Scale.
Discussion
The prevalence of inflammation in this LMIC sample with MDD was much higher (87%, n = 146) than the approximately 25% prevalence that has been reported from studies investigating associations between low-grade inflammation and depression in higher income countries.Reference Osimo, Baxter, Lewis, Jones and Khandaker28 We previously found similarly high rates of low-grade inflammation in adult people with bipolar depression in Pakistan.Reference Jones, Mahmood, Hodsoll, Chaudhry, Khoso and Husain21 Furthermore, there were a relatively high number of individuals in the sample with CRP values over 10 mg/L (n = 33). LMIC populations have heightened exposure to pro-inflammatory stressors like air pollution,Reference Arias-Pérez, Taborda, Gómez, Narvaez, Porras and Hernandez29 pathogenic microbesReference Goddard, Ban, Barr, Brown, Cannon and Colford30 and lower SES. This may have contributed to the high prevalence of inflammation seen in this sample of people with TRD, and may reflect higher levels of inflammation in Pakistan in general. Though it remains uncertain whether high rates of inflammation are unique to clinically depressed populations in this region, psychosocial and environmental stressors are known to increase the risk of MDD and predict poorer prognosis.Reference Gilman, Trinh, Smoller, Fava, Murphy and Breslau31 It is possible that inflammatory processes may underlie this enhanced risk. Future research should aim to clarify the relationship between elevated inflammatory markers and depression by investigating rates of inflammation in depressed and non-clinical populations in Pakistan and other LMICs.
Inflammation has been shown to induce ‘sickness behaviours’ that mimic symptoms of MDD such as anhedonia, weight and appetite disturbances, fatigue and psychomotor slowing.Reference Orsolini, Pompili, Tempia Valenta, Salvi and Volpe6 Inflammation has also been associated with cognitive dysfunction, anxietyReference Orsolini, Pompili, Tempia Valenta, Salvi and Volpe6 and suicidal ideation.Reference Miola, Dal Porto, Tadmor, Croatto, Scocco and Manchia12 Our analysis found early night and early morning insomnia and anxiety remained associated with elevated CRP when controlling for age, BMI, SES and gender. Associations between insomnia symptoms and inflammation are consistent with previous studies. Insomnia has been repeatedly associated with elevated CRP through large longitudinal and experimental studies.Reference Orsolini, Pompili, Tempia Valenta, Salvi and Volpe6 Furthermore, both sleep disturbance and inflammation have been shown to be influence the risk of depression.Reference Irwin, Olmstead and Carroll32 The association between elevated CRP and anxiety symptoms in depression is also well established.Reference Orsolini, Pompili, Tempia Valenta, Salvi and Volpe6 It should be noted, however, that this association only just met cut-offs for significance in our analysis, with a wide 95% CI indicating low precision (95% CI 1.39–21.7, P = 0.044). These results should therefore be interpreted cautiously, and further research should be conducted to confirm this association in LMIC settings.
Contrary to what has been reported previously,Reference Orsolini, Pompili, Tempia Valenta, Salvi and Volpe6 non-inflamed people in this sample were at greater risk of gastrointestinal and general somatic symptoms. Our group has previously reported insomnia in the middle of the night and suicidality to be negatively associated with low-grade inflammation in Pakistani adults with bipolar depression,Reference Jones, Mahmood, Hodsoll, Chaudhry, Khoso and Husain21 which also contrasts with findings from higher income countries. These results indicate that depressive disorders may present differently in higher and lower income countries, and may reflect the unique contributing factors and vulnerabilities faced by the populations in each setting. Future research should aim to investigate the basis for these distinct associations in Pakistan and other LMICs.
When people with CRP levels of >10 mg/L were omitted, our linear regression analysis found that in addition to anxiety and insomnia, other symptoms positively associated with CRP included psychomotor retardation, agitation, and difficulties with work and other activities. These findings are largely consistent with the existing literatureReference Orsolini, Pompili, Tempia Valenta, Salvi and Volpe6 as well as sickness behaviour theory.Reference Irwin, Olmstead and Carroll32 Agitation has been associated with neuroinflammation in mental disorders,Reference Larsen, Stunes, Vaaler and Reitan34 and increased agitation with higher baseline CRP may be a consequence of heightened anxiety and disturbed sleep. Symptoms negatively associated with CRP in linear regression were general somatic symptoms and weight loss. CRP has previously been associated with increased appetite in depressed populations,Reference Chae, Baumert, Nübel, Brasanac, Gold and Hapke35 which may explain the negative association with weight loss observed here. It should be noted, however, that the original trials used different versions of the SPINREACT CRP-turbilatex agglutination test with differing sensitivities (the assay used in the simvastatin study had a lower detection limit of 1 mg/mL,36 while the lower limit in the minocycline study was 2 mg/L37). This limits any interpretation of linear regression results. These findings require replication in larger and more diverse LMIC patient samples, ideally using a higher sensitivity CRP assay.
Treatment with simvastatin did not improve insomnia or anxiety symptoms in Pakistani adults with TRD and high CRP. This finding is consistent with results from the original trial, which concluded simvastatin did not provide any therapeutic benefit over that of placebo overall in Pakistani adults with TRD.Reference Husain, Chaudhry, Khoso, Kiran, Khan and Ahmad15 Moreover, a recent systematic review found no improvement in sleep-related symptoms with statin treatment in any study, and little to no improvement in anxiety.Reference De Giorgi, Rizzo Pesci, Quinton, De Crescenzo, Cowen and Harmer38 It is possible that the strength of the anti-inflammatory effects of simvastatin at the 20 mg dosage used in the parent RCT may not be adequate for the treatment of symptoms associated with inflammation.
In a smaller sample of 16 people, greater reductions in symptom scores were observed with adjunctive minocycline relative to placebo, particularly for anxiety symptoms, and to a lesser extent, early morning waking and early night insomnia. Despite the small sample size, our results are in line with other studies investigating minocycline augmentation treatment for MDD and TRD. In a pooled data analysis of 112 individuals with TRD which included people from the minocycline pilot trial included in the current analysis, Zazula et alReference Zazula, Husain, Mohebbi, Walker, Chaudhry and Khoso33 reported that minocycline significantly improved anxiety symptom severity on the HAMD and the GAD-7, in addition to functional status and depressive symptoms.Reference Liu, Yue, Hu, Cheng, Wang and Wang39 The use of minocycline to treat anxiety symptoms is also supported by a preclinical study which demonstrated chronic administration of minocycline reduced stress-induced anxiety behaviours.Reference Liu, Yue, Hu, Cheng, Wang and Wang39 While larger and more diverse samples are needed, these results indicate augmentation with minocycline may be a viable strategy for treating comorbid depression and anxiety symptoms in people in Pakistan with elevated CRP, and possibly other LMIC settings.
This study was limited by a relatively small sample size, and by the availability of biomarker and other data measured in the original studies. MDD is a complex, heterogeneous condition, with many factors (e.g. genetics, adverse life experiences) contributing to symptom profiles that we were unable to assess in this analysis. Additionally, we did not correct for multiple comparisons. While correction for multiple comparisons was not necessary in this context given the exploratory nature of the analysis,Reference Bender and Lange40 our results would not withstand Bonferroni correction. Finally, our investigation of symptom response to minocycline was limited by a very small sample size, and so findings are difficult to extrapolate. Furthermore, as our analysis was exploratory, we did not compare the efficacy of each medication treatment directly. Therefore, at present, clinical inferences cannot be drawn from these results. Despite these limitations, this study provides preliminary evidence to support minocycline as an adjunctive treatment for a subset of inflammation-associated symptoms. Larger, more comprehensive trials investigating treatment response to minocycline for individual symptoms of depression may be warranted to confirm this finding.
Overall, this study's findings reveal that deeper insights about associations between inflammation and depression can be gained through investigating symptoms individually. Furthermore, these results support the notion that individuals with specific clinical phenotypes and evidence of peripheral inflammation may benefit from a more targeted and individualised approach with repurposed anti-inflammatory agents. Given the high prevalence of inflammation in this LMIC sample, targeting inflammation and increasing stress resilience may improve treatment outcomes for individuals with MDD in these settings.
Supplementary material
Supplementary material is available online at https://doi.org/10.1192/bjo.2024.735
Data availability
Data will be made available upon request to the corresponding author.
Acknowledgements
This work was funded in part by an Academic Scholars Award from the Department of Psychiatry, University of Toronto and a Canadian Institutes of Health Research Tier 2 Canada Research Chair awarded to M.I.H. We thank Dr. Sheng Chen (Centre for Addiction and Mental Health, Biostatistics Core) for providing statistical consultation.
Author contributions
E.F. formulated the research question, extracted the data, carried out the data analysis and drafted the manuscript. B.D.M.J. contributed to the data analysis plan. J.H. provided consultation on statistical methods and assisted with data analysis and interpretation. M.I.H. formulated the research question, conceived the study design and helped draft the manuscript. N.H., A.B.K., A.H.Y. and I.B.C. contributed to data analysis and drafting the manuscript. All authors have reviewed and approved the final manuscript.
Funding
This research received no specific grant from any funding agency, or commercial or not-for-profit sectors.
Declaration of interest
M.I.H. receives research support from the Brain and Behavior Research Foundation, Canadian Institutes of Health Research, Centre for Addiction and Mental Health Foundation, the Physician Services Incorporated Foundation and the University of Toronto. He has provided consultancy to Mindset Pharma, PsychEd Therapeutics and Wake Network. The other authors have no conflicts to disclose. M.I.H. and A.H.Y. are members of the editorial board for British Journal of Psychiatry Open, though they were not involved in the review or decision-making process of this manuscript.
Open access
eLetters
No eLetters have been published for this article.