Table 1.

Reference table for included articles.

Reference [year]	Antidepressant(s)	Design and population	Time to relapse and measures	Relapse outcomes and other findings	Strengths and limitations
Doogan and Caillard24 [1992]	Sertraline (50–200 mg/day flexible dosing) mean dose was 69.3–89.2 mg/day	RCT, n = 185 MDD patients; randomized to sertraline or placebo for 44-week maintenance phase	HAM-D and CGI severity; CGI severity score	13% of sertraline patients relapsed compared with 45.7% placebo (p < 0.001)	Mean time to relapse was not specified
Kamijima et al.25 [2006]	Sertraline (50–100 mg/day flexible dosing)	RCT, n = 117; randomized to either sertraline or placebo for 16-week maintenance phase	Time to relapse rate curve was significantly higher in sertraline compared with placebo (p = 0.026) HAM-D, CGI-I, MADRS	8.5% of sertraline patients relapsed; significantly lower relapse rate compared with placebo (19.5%) (p < 0.016)	The time to relapse was not specified
Aberg-Wistedt et al.26 [2000]	Sertraline (50–150 mg/day) versus paroxetine (20–40 mg/day)	RCT, n = 353 MDD patients randomized to 24-week treatment	HAM-D, CGI-I, CGI-S, MADRS	Slightly higher rate of relapse in patients with paroxetine (9% versus 2% with sertraline); continuation phase significantly increased remission rates for both SSRIs	The mean time to relapse was not reported
Anghelescu et al.12 [2006]	Paroxetine (20–40 mg/day) versus WS®5570 3 × 300 mg/day	RCT multisite study 133 adult MDD outpatients for 160-week maintenance	WS®5570 (89.5%) and sertraline (92.9%) of patients continued to be treatment responders at end of maintenance phase HAM-D	WS®5570 was equally effective as paroxetine in preventing relapse after recovery from an episode of moderate-to-severe depression	High attrition rate for both the WS®5570 and paroxetine groups; median time of relapse was not reported
Emslie et al.27 [2004]	Fluoxetine versus placebo	RCT Depressed children–adolescents randomized to 36-week fluoxetine or placebo	Time to relapse was shorter for patients with placebo than fluoxetine (71.2 ± 9.5 days versus 180.7 ± 17.0 days, respectively p = −0.046) measured by CDRS	Fewer relapse rates in fluoxetine patients (34%) versus placebo (60%)	More studies needed to determine predictors of relapse and recurrence
Kornstein et al.28 [2006]	Escitalopram (10 or 20 mg/day)	RCT, n = 73 MDD patients	252 days, 27% relapse rate, measured by MADRS, HAM-D	Escitalopram is effective in reducing recurrence when used as maintenance therapy; discontinuing maintenance with escitalopram and switching to placebo led to depression recurrence even in people with well-controlled depression
Fava et al.29 [1994]	Mainly TCAs (i.e. desipramine, amitriptyline, imipramine, mianserin) ± CBT	RCT n = 40 depressed adults with residual depressive symptoms were randomized to 20-week pharmacotherapy or pharmacotherapy + CBT 2-year follow up	No significant difference in relapse rates between both groups at 2 year follow up (15% with CBT versus 35% pharmacotherapy only)	Although the relapse rates were insignificantly different at 2 years, CBT group improved residual symptoms more than pharmacotherapy only (p < 0.0001)	Limitation was the heterogeneous antidepressants used; concomitant use of benzodiazepines
Fava et al.30 [1996]	Mainly TCAs (i.e. desipramine, amitriptyline, imipramine, mianserin) ± CBT	RCT n = 40 depressed adults with residual depressive symptoms were randomized to 20-week pharmacotherapy or pharmacotherapy + CBT 4-year follow up	At 4-year follow up, CBT had lower relapse rate (35% versus 70%) and greater improvement in residual symptoms (p < 0.0001)		Strength was the long-term follow up < 4 years; limitation was the heterogeneous antidepressants used; concomitant use of benzodiazepines
Gulec et al.22 [2011]	Fluoxetine, mirtazapine, mianserin, amitriptyline, tianeptine	Open-label trial, n = 60 MDD patients for 52-week maintenance phase	A return of index major depressive episode following onset of full or partial remission; score of 18 or more on HAMD-17 for at least 2 consecutive weeks in partial or full remission.	Relapse rate at week 52 was 8.33% for full remitters and 25% for partial remitters (threefold); partially remitted patients had higher relapse rate as compared with fully remitted patients having depression	Focus was the relapse rate on partially versus full remitters; many concomitant medications (stimulants, benzodiazepines and antipyschotics) were coprescribed; maintenance phase drug treatments were unspecified
Dotoli et al.13 [2006]	Fluvoxamine SSRI	Maintenance phase treatment study, n = 101 remitted patients (80 unipolar, 21 bipolar) during 6-month follow up.	Relapsed patients had a longer mean duration of index depressive episodes than nonrelapsed patients (23.3 ± 24.2 versus 17.2 ± 17.3 weeks) measured by SASS	8.9% relapsed within first 2 months of continuation	Median time to relapse was not compared between fluvoxamine, no placebo control
Peselow et al.18 [2015]	SSRIs versus SSRIs and SSRI versus (fluoxetine, escitalopram, sertraline, paroxetine) ± CBT	Naturalistic long-term study, n = 387 patients who had remitted or responded with SSRIs	Escitalopram had highest prophylactic efficacy or prevention of depressive episodes (36%), fluoxetine (33.3%), sertraline (21.3%), paroxetine (12.85), but ns between SSRIs MADRS	41% SSRI + CBT were episode free compared with 18% of patients in the SSRI-only group who maintained remission (p < 0.05)	Small size of CBT group
Pundiak et al.3 [2008]	Fluoxetine, paroxetine, sertraline	Naturalistic long-term study, n = 60 MDD patients, entered into 1-month continuation phase followed by a 5-year maintenance phase with SSRI monotherapy and assessed every 3 months for mood symptoms	Median time of relapse with SSRI was 38 months, exceeding the 10-month time of patients who had discontinued	Continuation of SSRIs had better probability of maintaining remission during the first year than with discontinuation [Survival probability 0.70 versus 0.40]) significant difference for over 30 months	Major strength was the long term follow up. Community, pragmatic setting Limitation was the lack of separating analyses based on the SSRI (fluoxetine versus paroxetine versus sertraline)
Kaymaz et al.19 [2008]	SSRIs (fluoxetine, sertraline, citalopram, paroxetine) and TCAs (amitriptyline, imipramine, nortriptyline, maprotiline, dothiepin)	Meta-analysis of RCTs (30 trials with 4890 participating patients)	Reduction of relapse risk was significant for SSRIs: (OR = 0.24, Cl 0.20–0.29), TCAs: (OR = 0.29, Cl 0.23–0.38) at 1-year follow up CGI-I, CGI-S, MADRS, HAM-D, GAS, MINI, Raskin depression scale, RDC, SCID-P	Antidepressants are effective in reducing relapse risk in maintenance phase; however, no significant difference seen between SSRIs and TCAs	Direct comparison between TCA and SSRI; limitation was that neither the time to relapse nor the mean relapse rate for both groups was reported
Bauer et al.31 [2009]	Venlafaxine versus sertraline, citalopram, escitalopram and placebo	Meta-analysis of RCT trials up to 2007 based on venlafaxine trials with MDD patients	HAM-D, BDI, MADRS	Long-term treatment with venlafaxine was effective at reducing relapse after major depressive episode (OR = 0.37, Cl 0.27–0.51) compared with placebo; venlafaxine was more effective than SSRIs and as effective as TCA in treating major depressive episode via remission and response rates.	Wide CIs were due to the heterogeneity of the analyzed clinical trials
Garnock-Jones et al.20 [2010]	Escitalopram (10 or 20 mg/day)	Meta-analysis of three RCTs, n = 311 MDD patients randomized to maintenance phase of escitalopram or placebo for 24–52 weeks.	Time of relapse was significantly longer in escitalopram than placebo (p < 0.05) MADRS and HAM-D	Escitalopram effectively prevented relapse in adults who had previously responded to escitalopram	One of the three studies analyzed did not report the number of relapsed patients (calculated here ~14 patients or 9% relapse rate)
Yang et al.23 [2010]	Fluoxetine ⩾ 40 mg/day	Post hoc analysis, n = 262 MDD patients who responded to fluoxetine by week 12 (131 continued fluoxetine, 131 switched to placebo)	Based on time to relapse as a dependent variable, the hazard ratio = 1.55 or 2.4; symptom questionnaire (SQ) and 90-item self-report Hopkins symptom checklist (SCL-90)	Fluoxetine relapse rate at 6 months was 35.2% and 61.8% for placebo; at 1 year, 45.9% for fluoxetine and 72% for placebo	Time to relapse was a dependent variable in a prediction model but the mean time to relapse was not described

RCT, randomized controlled trial; MDD, major depressive disorder; HAM-D, Hamilton Depression Rating Scale; CGI, Clinical Global Impression Scale; CGI-S, Clinical Global Impression Severity Scale; SSRIs, selective serotonin reuptake inhibitors; SASS, Social Adaptation Self-Evaluation Scale; GAS, Global Assessment Scale; MINI, Mini International Neuropsychiatric Interview; MADRS, Montgomery-Asberg Depression Rating Scale; WS^®5570, Hypericum extract; CDRS, Children’s Depression Rating Scale; CBT, cognitive behavior therapy; TCAs, tricyclic antidepressants; ns, not significant; OR, odds ratio; CI, confidence interval; RDC, Research Diagnostic Criteria; SCID-P, Structured Clinical Interview D for DSM-IV - Patient Version; BDI, Beck Depression Inventory.