A total of 466 patients were screened and 447 patients were enrolled in the double-blind phase of the study. After a 3-week run-in period, 400 patients (80 children and 320 adolescents) were included and randomly assigned to one of the four treatments [33]. A total of 352 (88%) patients completed the 12-week double-blind phase of whom 339 (96%) including 68 children and 271 adolescents, entered the open-label extension. Patients in the open-label extension had previously received: agomelatine, either at 10 or 25 mg (ago/ago, n = 170), placebo then agomelatine 10–25 mg (pcb/ago, n = 85), or fluoxetine then agomelatine 10–25 mg (fluox/ago, n = 84). Among the ago/ago patients, 39% remained on ago 10 mg during the extension phase, 45% started on ago 10 mg and increased to ago 25 mg, and 14.5% started on ago 10 mg, increased to ago 25 mg, and then decreased once more to ago 10 mg for the duration of the extension. Of the 339 patients included in the extension period, 187 (55.2%) completed the open-label extension: 93 (54.7%, ago/ago); 48 (56.5%, pcb/ago); and 46 (54.8%, fluox/ago).

The most frequent reason for premature study withdrawal during the extension period was recovery (20.4%), with a numerically higher frequency in patients who received agomelatine 10–25 mg in both the double-blind and extension periods (23.5%) than in the other two groups (16.5% in the pcb/ago group and 17.9% in the fluo/ago group). Withdrawals for non-medical reasons were reported in 14.7% who received agomelatine in both periods, 14.1% of pcb/ago patients, and 20.2% of fluo/ago patients. Other reasons for withdrawal were: adverse events (14, 4.1% patients overall), lack of efficacy (8, 2.4%), protocol violation (5, 1.5%), and loss to follow up (2, 0.6%). Of the 271 adolescents, 145 (53.5%) completed the open-label extension. The disposition of adolescents and their reasons for withdrawal were similar to the overall population.

Patient demographics and characteristics at the start of the extension period (W12) are shown in Table 1 for the overall population. Mean age (± SD) was 13.6 ± 2.7 years and 61.9% were female. BMI ranged from 13.5 to 35.2 kg/m2; 11 patients (3.2%) were considered underweight, 18.3% overweight and 6.2% obese. Current MDE was diagnosed as moderate in 62.5%, and severe without psychotic features in 37.5% of patients; MDE presented with melancholic features in 19.2%. At inclusion (W0), the mean duration of the current MDE for patients entering the extension period was 133.6 ± 134.1 days with a median of 90.0 days (range from 29 to 961 days). A total of 94 patients (27.7%) had a history of previous MDE, and 57 patients (16.8%) had a family history of mood disorders. Nearly two-thirds of patients (59.9%) reported at least one medical history besides MDD. At each visit during the extension period (except W14), most patients (≥ 94%) participated in the Manualized Psychosocial Counselling session.

Mean treatment duration in the extension period was 15.5 ± 7.5 months (median of 21.0 months) and mean tablet adherence was 97.4 ± 9.8%.

In the children and adolescent population combined, the mean CDRS-R raw total score gradually decreased over the W12-W104 period in all groups, indicating a continuous improvement of patients receiving agomelatine all along the extension period, whatever the treatment previously received during the double-blind period (Fig. 2). Mean W12 values for the three groups and overall were 42.1 ± 12.4 (ago/ago), 46.2 ± 15.1 (pcb/ago), 42.7 ± 11.7 (fluox/ago), and 43.3 ± 13.0 (overall), respectively, reflecting the difference of efficacy between treatments during the double-blind period. Mean changes from W12 to last post-W12 value in the three groups were − 16.3 ± 12.2, − 18.9 ± 16.1, and − 16.1 ± 15.5, respectively, leading to last post-W12 values of 25.8 ± 10.5 (ago/ago), 27.3 ± 12.1(pcb/ago), 26.6 ± 12.2 (fluox/ago), and 26.4 ± 11.3 overall. These differences were due to the heterogeneity at W12 between treatment groups. In all groups, agomelatine treatment was associated with a continuous decrease in depressive symptoms, but the most marked improvement in mean CDRS-R score was observed during the first 24 weeks of open-label treatment (Fig. 3).

CDRS-R raw total score: mean value (± SD) at each visit during the W12–W104 open-label extension period, according to the treatment previously received during the double-blind period

CDRS-R raw total score: mean value (± SD) at each visit during W000–W104 in patients receiving agomelatine for the duration of the study and continuing in the extension period

Similar results were observed in the adolescent population with changes from W12 to last post-W12 value of − 15.2 ± 12.1 (ago/ago, n = 134), − 18.4 ± 16.1 (pcb/ago, n = 69), and − 15.8 ± 14.9 (fluo/ago, n = 68).

Among patients receiving agomelatine in both periods, mean decreases from W0 in CDRS-R raw total score were: − 29.5 ± 14.0 at W24 (N = 160), − 41.6 ± 12.6 at W104 (N = 93), and − 38.8 ± 13.2 at last post-W0 value (N = 170).

In the total population, the rate of patients considered in remission gradually increased during the extension period from 13.6% at W12 (N = 339) to 83.5% at W104 (N = 187), whatever the treatment previously received during the double-blind period (Fig. 4). Values for adolescents were 14.0% (N = 271) and 80.1% (N = 146), respectively. When considering the last post-W12 value, 74.6% of patients in the total population (N = 338) and 72.2% of adolescents (N = 270) were considered in remission. For patients treated exclusively with agomelatine, 129 (75.9%) were in remission at last post-W12 visit, and 81 (87.1%) were in remission at W104.

Proportion of total population (children and adolescents) in remission (CDRS-R raw total score ≤ 28) during the extension period

In the overall population, there was a continuous decrease in mean CGI-S and CGI-I scores during the extension period indicating an improvement under agomelatine, whatever the treatment previously received during the double-blind period. Mean scores improved from 3.5 ± 1.1 at W12 to 1.7 ± 1.0 at W104 for the CGI-S score, and from 2.5 ± 1.0 at W12 to 1.5 ± 0.8 at W104 for the CGI-I score. Mean scores at the last post-baseline visit were 1.9 ± 1.1 for the CGI-S score and 1.6 ± 0.9 for the CGI-I score. The proportion of responders (defined as CGI-I score ≤ 2) increased from 49.6% at W12 to 87.8% at W104. The rate of responders at the last post-baseline visit was 84.9%.

Similar improvements in CGI scores during the extension period were observed in the adolescent population, with mean CGI-S score decreasing from 3.5 ± 1.1 at W12 to 1.7 ± 1.0 at W104 and mean CGI-I score decreasing from 2.5 ± 1.0 at W12 to 1.5 ± 0.9 at W104. The proportion of adolescent responders increased from 51.7% at W12 to 85.6% at W104.

Although this study was not designed as a relapse prevention study, it was nevertheless of interest to measure occurring relapses.

Among the 69 patients initially randomized to either agomelatine 10 or 25 mg and presenting at least a significant clinical response at W12 (defined as: either a CDRS-R score < 40 and a CGI-I score of 1 or 2 or a decrease of 50% or more on the CDRS-R score), eight patients (11.6%) relapsed during the W12-W40 period: six during the first 6 weeks of treatment and two beyond 6 weeks.

A total of 212 patients (62.5%) presented 620 treatment-emergent adverse events (TEAE) under agomelatine during the W12-W104 period: 61.8% of patients in the agomelatine both periods group, 64.7% in the pcb/ago group, and 61.9% in the fluox/ago group. Of these, 85 TEAE in 49 patients (14.5%) were considered related to treatment during the open-label extension: 15.3% agomelatine both periods, 16.5% pcb/ago, and 10.7% fluox/ago group. The most frequent treatment-related TEAEs (in more than three patients overall) were headache (2.4% of patients), dizziness (2.1%), dry mouth and thirst (1.8% each), somnolence and increased alanine aminotransferase (ALT) (1.2% each), and increased aspartate aminotransferase (AST) and nausea (0.9% each) (Table 2). The most frequent treatment-related TEAEs were the same in adolescents as in the total population.

A total of 19 patients (5.6%) presented with 32 severe TEAEs during the W12–W104 period.

Only four severe TEAEs in three adolescents (1.1%), two treated with ago/ago and one treated with pcb/ago, were considered related to treatment: headache, thirst, and dry mouth and increased appetite, which occurred in only one patient each (0.4%).

Thirteen patients (3.8%) experienced 18 TEAEs leading to agomelatine withdrawal during the extension period, of whom 11 were adolescents who experienced 16 TEAEs. Among the TEAEs leading to agomelatine withdrawal, 4 TEAEs in three patients (0.9%) were considered related to treatment: an increase of ALT and AST reported in the same patient (ago/ago group), one case of headache (pcb/ago), and one case of hypotension (pcb/ago group).

During the extension period, 12 patients (11 adolescents) presented emergent suicidal ideations on treatment according to the C-SSRS-C: three patients in the ago/ago group, four patients in the pcb/ago group and five patients in the fluox/ago group. Emergent suicidal ideation was rated as serious in one adolescent in the fluox/ago group. Two patients (one in each of the pcb/ago and fluox/ago groups, both adolescents) presented three emergent suicidal behaviors: both made emergent actual suicide attempt on treatment; in addition, the patient in the pcb/ago group also undertook emergent preparatory actions toward imminent suicidal behavior.

In the total population, the most common emergent liver potentially clinically significant abnormal (PCSA) values on treatment were high direct bilirubin (11.2%, 19 patients including 16 adolescents) and high indirect bilirubin (4.1%, seven patients including six adolescents).

A total of three patients (two adolescents and one child) had high emergent PCSA values (> 3 ULN) of ALT or AST on treatment. None of these events led to study withdrawal.

As expected in a pediatric population, patients gained an average of 4.2 ± 5.3 kg between W12 and W104. Among the adolescents there was a mean weight gain of 3.5 ± 5.5 kg between W12 and W104. Mean BMI slightly increased from W12 up to last post-baseline value under treatment by 0.48 ± 1.46 kg/m2. Analysis by class showed that most patients treated with agomelatine for both study periods remained in the same BMI class. Fourteen patients moved to a higher BMI class, including three patients moving from underweight to normal class. Eighteen patients moved to a lower BMI class, including 13 patients moving from overweight to normal class and five from obese to overweight.

In this study, Tanner Staging was used to document the development and sequence of secondary sex characteristics during puberty. At enrollment and W12, W52 and W104, Tanner stage of sexual development, based on pubic hair and genitalia appearance, was evaluated by physical examination conducted by a trained clinician. Among patients taking agomelatine for the duration of the study there was no evidence of any alterations to normal puberty development. At each visit, patient age and pubertal status were consistent with normal development and stable for the duration of the study.