This was a randomised, double-blind, placebo-controlled, parallel-group, multi-centre, fixed-dose study in adult patients (aged ≥ 18 years) with bipolar I disorder conducted from August 2012 to December 2015 at 21 centres in China. The study consisted of four phases: screening phase, open-label (OL) phase, randomised double-blind (RD) phase and a follow-up visit (Additional file 1: Fig. S1). Patients meeting the eligibility criteria entered the OL phase to receive lamotrigine monotherapy or adjunctive therapy escalated to a target dose of lamotrigine 200 mg/day monotherapy for up to 16 weeks.

Beginning at Week 7 of the OL phase, patients who met the response criteria (Clinical Global Impressions of Severity [CGI-S] score ≤ 3 maintained for ≥ 4 continuous weeks) and who were maintained for ≥ 1 week on lamotrigine 200 mg/day monotherapy, while demonstrating compliance with the study treatment, were enrolled in the RD phase and randomised (1:1) to lamotrigine 200 mg/day or placebo. Patients who completed 36 weeks of randomised treatment, reached the study endpoint (intervention for relapse and/or recurrence of a mood episode) or withdrew from the study early (including the OL phase and RD phase) were followed up for 14 days after the last dose.

Patients were randomised (1:1) sequentially (block of four) through GlaxoSmithKline’s interactive voice response system, following a computer-generated randomisation schedule. Both investigators and patients were blinded to treatment allocation.

The study was conducted in accordance with the International Conference on Harmonization Good Clinical Practices, applicable country-specific requirements and the ethical principles outlined in the Declaration of Helsinki (2008). The study protocol was approved by an Independent Ethics Committee or Institutional Review Board at each study centre. Written informed consent was obtained from each participant prior to any study-specific procedures.

When lamotrigine was used as adjunctive therapy with valproate during the OL phase, the prespecified starting dose of lamotrigine and subsequent increases in dose were reduced to half; however, when used with carbamazepine, the starting dose of lamotrigine and subsequent dose adjustment rates were doubled. Concomitant antiepileptic drugs and psychotropic medications (except fluoxetine) were permitted for the treatment of mood episodes during the OL phase but were discontinued at least 2 weeks (at least 3 weeks for lithium) prior to entering the RD phase. During the OL and RD phases, the short-term (two to three times/week, duration < 2 weeks) use of chloral hydrate, lorazepam, clonazepam, estazolam or oxazepam was permitted as required for control of agitation, irritability, insomnia and hostile behaviour. No other psychotropic, antidepressant or antimanic drug was permitted unless the patient had reached the study endpoint, at which point study treatment was stopped and one or more of the following psychotropic medications were prescribed: antidepressants, antipsychotics (with or without anticholinergic medications), benzodiazepines (exceeding the above dosages) and anticonvulsants/mood stabilisers. Concomitant medications for co-morbid conditions (including diabetes and hypertension) were permitted at the discretion of the investigator. During the RD phase, lamotrigine (200 mg) or matching placebo tablets were administered orally, once daily in the evening.

Patients of either gender (aged ≥ 18 years) were recruited from inpatient or outpatient clinics. Patients entering the OL phase were diagnosed (using medical records and clinical interviews) with bipolar I disorder and had either a current or most recent depressed (296.5×), hypomanic (296.40), manic (296.4×), or mixed (296.6×) episode (as defined by Diagnostic and Statistical Manual of Mental Disorders-IV [DSM-IV] criteria) within the last 60 days. Patients diagnosed with 296.5 × must have had at least one well-documented manic, hypomanic or mixed episode, as defined by DSM-IV criteria, within 3 years of enrolment. Patients diagnosed with 296.40, 296.4 × or 296.6 × must have had at least one well-documented additional manic, hypomanic or mixed episode and one depressed episode, as defined by DSM-IV criteria, within 3 years of enrolment. Patients were excluded from the OL phase if they met DSM-IV criteria for rapid cycling in the 12-month period prior to enrolment; had significant DSM-IV Axis II diagnosis; had current or previous diagnosis of an Axis I disorder with the exception of bipolar disorder, had signs or symptoms of psychosis, were at suicidal risk; had a history of substance abuse or dependence; had received fluoxetine within 4 weeks prior to the OL phase, used oral contraceptives or other hormonal preparations containing oestrogen within 2 weeks prior to OL phase entry; or had a history or current diagnosis of epilepsy, were morbidly obese (body mass index > 35 kg/m2) or were pregnant or lactating women. Factors leading to withdrawal before entry to the RD phase included signs or symptoms of psychosis, the need for treatment of a manic or mixed episode during the OL phase (with new courses of lithium/psychotropic drugs or other drugs with a half-life greater than 14 days), becoming actively suicidal and/or having a score ≥ 3 on item 3 of the Hamilton Depression Rating Scale (HAMD), or testing positive for an illicit drug on laboratory analysis administered before randomisation or alcohol abuse/addiction.

Patients were withdrawn from the study if they required intervention for a relapse and/or recurrence of a mood episode (primary endpoint during the RD phase); had a medically relevant adverse event (AE) or intercurrent illness or were pregnant; demonstrated significant non-compliance with the protocol or investigational treatment; had an inability to tolerate the drug; developed a rash or hypersensitivity reaction; had a prolonged QT interval (QTc > 500 ms or uncorrected QT > 600 ms or, in cases with bundle branch block, QTc > 530 ms based on average QTc value of triplicate electrocardiograms [ECGs]); or discontinued treatment.

Demographic and baseline characteristics were assessed by study investigators at screening or baseline. During the RD phase, patients were assessed at weekly intervals for the first month, biweekly intervals for the second month and then at monthly intervals for up to 36 weeks. The HAMD, Young Mania Rating Scale (YMRS), Clinical Global Impressions of Improvement (CGI-I), CGI-S and Global Assessment Scale (GAS) were used as indices for intensity and duration of mood symptoms.

The primary endpoint was intervention for relapse and/or recurrence of a mood episode, defined as first prescription of any additional pharmacotherapy or electroconvulsive therapy necessary for the treatment of a relapse and/or recurrence of a depressive, manic, hypomanic or mixed episode. The primary efficacy outcome measure was the time from entry into the RD phase to intervention for relapse and/or recurrence of a mood episode (TIME). Secondary efficacy outcome measures included: time to intervention for manic, hypomanic or mixed episode (TIMan); time to intervention for depressive episode (TIDep); overall survival in the study (TIME-SIS); changes from baseline to Week 36 in CGI-S (Guy 1976), CGI-I (Guy 1976), HAMD (Hamilton 1960), YMRS (Young et al. 1978), GAS (Endicott et al. 1976); and change in weight from baseline during the RD phase. Safety assessments included: monitoring of AEs, including treatment-emergent adverse events (TEAEs); clinical laboratory tests; vital signs; ECGs; physical examinations; and suicidality, as determined by the Columbia Suicide Severity Rating Scale (C-SSRS). Patient-reported quality of life changes were assessed at each visit using the Short Form 36 (SF-36) questionnaire (Garratt et al. 1993). The eight domains of the SF-36 were physical functioning, role physical, role emotional, bodily pain, vitality, mental health, social functioning and general health. The SF-36 was administered by well-trained personnel who passed the consistency evaluation with a certification record.

Based on previous studies (Calabrese et al. 2003; Bowden et al. 2003), the median time to intervention was 97 days in the lamotrigine group and 58 days in the control group. To detect a treatment difference of the same magnitude in the current study, a target sample size of 178 patients (89 patients in each group) was estimated to provide 90% power at the significance level of 0.05. The minimum regulatory requirement specified by the State Food and Drug Administration of China was 100 pairs of patients. Assuming a 20% dropout rate, at least 250 patients were required for the RD phase (125 patients in each group). Considering 60% of patients who entered the OL phase would be randomised to RD phase, the total number of patients to be included in the OL phase was determined to be 416.

The full analysis population (FAP) for the OL and RD phases included all patients who received at least one dose of the study medication and provided at least one post-baseline efficacy/health outcome assessment during the OL and RD phases, respectively. The safety population included all patients who received at least one dose of the study medication in the corresponding phase (OL or RD phase). No interim analyses were planned or performed for this study.

All analyses were conducted using SAS software (Version 9, SAS Inc., Cary, NC, USA). A superiority test was performed to compare efficacy between two groups. The primary efficacy outcome of TIME was assessed using the Cox proportional hazards regression model, including site and CGI-S baseline score as covariates. Data for patients who discontinued from the study were censored. Kaplan–Meier plots of TIME are presented by treatment group during the RD phase, together with hazard ratios (HR), 95% (two-sided) confidence intervals (CI) and p values. The secondary outcome measures TIME-SIS, TIDep and TIMan were analysed in a similar way, with data censored both for patients who discontinued and for patients who received intervention for a mood episode of the opposite polarity to the relevant endpoint; p values were also provided based on the log-rank tests to describe any significant distribution difference between treatment groups for these endpoints. The between-group comparisons for HAMD, YMRS, CGI-S, CGI-I, GAS scales and body weight (secondary outcome measures) as well as for quality of life scores (SF-36) were performed using analysis of covariance (ANCOVA) or rank sum test.

Post hoc analyses (Cox proportional hazards regression model) were conducted on the FAP to compare the efficacy (TIME) of lamotrigine and placebo in a subgroup of patients with moderate/severe mood symptoms at baseline. One post hoc analysis defined symptoms by CGI-S score (CGI ≥ 4; moderate to severe symptoms), the second defined symptoms by HAMD (≥ 18)/YMRS (≥ 10) scores. A further analysis explored the effect of type of index (most recent) episode (depressive or manic/hypomanic) and first recorded episode (depressive or mixed/manic) on TIME.

Baseline characteristics, including past psychiatric conditions and OL phase lamotrigine exposures were summarised descriptively for each RD treatment population.