# Table 5 Primary and secondary outcomes at week 9 for step 2 randomisation

Continue sertraline Combine with mirtazapine Switch to mirtazapine Combine vs continue Switch vs continue Combine vs switch
Least squares mean (95% CI) Least squares mean (95% CI) Least squares mean (95% CI) Adjusteda difference
(95% CI)
P value
(95% CI)
P value
(95% CI)
P value
PHQ-9 9.26 (8.79 to 9.72) 8.27 (7.80 to 8.74) 8.25 (7.79 to 8.71) −0.99 (−1.55 to − 0.43)
P = 0.0006
P = 0.0012 b
−1.01 (− 1.56 to − 0.46)
P = 0.0004
P = 0.0012 b
0.02 (− 0.54 to 0.58)
P = 0.94
P = 0.94 b
BDI-II 18.68 (17.87 to 19.49) 16.59 (15.77 to 17.40) 17.06 (16.25 to 17.86) −2.10 (−3.12 to − 1.07)
P < 0.0001
− 1.62 (− 2.63 to − 0.61)
P = 0.0017
−0.47 (− 1.49 to 0.55)
P = 0.36
FIBSER 5.34 (5.05 to 5.63) 5.43 (5.14 to 5.72) 5.59 (5.30 to 5.88) 0.09 (−0.32 to 0.50)
P = 0.66
0.25 (−0.16 to 0.66)
P = 0.22
− 0.16 (− 0.57 to 0.25)
P = 0.44
Raw numbers (%) Raw numbers (%) Raw numbers (%) Adjustedc OR
(95% CI)
P value
(95% CI)
P value
(95% CI)
P value
Proportion of response 233/537 (43.4%) 273/527 (51.8%) 275/550 (50.0%) 1.41 (1.11 to 1.85)
P = 0.0055
1.39 (1.08 to 1.78)
P = 0.0109
1.03 (0.81 to 1.33)
P = 0.79
Proportion of remission 132/537 (24.6%) 188/527 (35.7%) 174/550 (31.6%) 1.80 (1.36 to 2.38)
P < 0.0001
1.51 (1.14 to 1.99)
P = 0.0037
1.19 (0.92 to 1.55)
P = 0.19
Proportion of continuation of the allocated treatment up to week 9 452/551 (82.0%) 400/537 (74.5%) 427/558 (76.5%) 0.61 (0.45 to 0.83)
P = 0.0014
0.68 (0.50 to 0.92)
P = 0.0131
0.90 (0.67 to 1.19)
P = 0.46
Mean (SD) Mean (SD) Mean (SD)
Sertraline prescribed at week 9 (mg/day) 71.7 (30.0), n = 520 71.1 (29.1), n = 502 6.5 (19.0), n = 524
Mirtazapine prescribed at week 9 (mg/day) 0.6 (3.9), n = 520 15.1 (11.8), n = 501 18.4 (13.0), n = 524
1. aThe linear mixed-effects repeated-measures model included fixed effects of PHQ-9 at week 3, treatment, visit (as categorical), treatment-by-visit interaction and minimisation variables for step 2 randomisation (step 1 treatment, 50% or greater reduction on PHQ-9 by week 3, moderate or greater impairment on FIBSER at week 3) and random effects for individuals and sites
2. bHochberg method was used for adjustment of multiplicity for the primary outcome (italicised P values for PHQ-9) but not for secondary outcomes, as postulated in the statistical analysis plan [33]
3. cWe used the logistic regression model adjusted for sites, step 1 treatment, 50% or greater reduction on PHQ-9 by week 3 and moderate or greater impairment on FIBSER at week 3
4. BDI-II Beck Depression Inventory 2nd edition, FIBSER Frequency, Intensity and Burden of Side Effects Rating, PHQ-9 Patient Health Questionnaire-9