Selection and characteristics of included studies

After removal of duplicates, 5450 records were identified and screened. Out of 118 full-text articles, 17 articles reporting on 10 unique RCTs were included (figure 1 and online supplemental appendix 3). Table 1 summarises the characteristics of included studies. Due to considerable methodological differences, the three RCTs examining pharmacotherapy for partial remission could not be meta-analysed and were included only in the qualitative review (table 1 and online supplemental table 2). One study compared cognitive therapy (CT) augmentation with lithium augmentation while antidepressant medication (ADM) was continued.36 Another RCT examined oestrogen augmentation to ADM in women with perimenopausal depression experiencing partial remission after ADM.37 Lastly, one trial investigated the efficacy of iloperidone augmentation of ADM.38

Figure 1

Preferred Reporting Items for Systematic Reviews and Meta-analyses flow diagram of study selection and inclusion. MDE, major depressive episode.

Table 1

Overview and characteristics of included studies

This resulted in a final set of 7 included RCTs in the quantitative analysis, including 1024 patients who received psychotherapy (n=517) or a control intervention (n=507). All RCTs examined a cognitive behavioural therapy (CBT)-derived psychotherapy strategy: CT (k=1),28 guided self-help CBT (k=2),39 40 rumination-focused CBT (k=2),41 42 group-based mindfulness-based cognitive therapy (MBCT; k=1)43 and MBCT-based guided self-help (k=1).44 Three studies (n=290)28 40 41 examined psychotherapy as add-on to ADMc and four studies (n=734) examined psychotherapy as an add-on to TAU compared with TAU only.39 42–44 Treatment length ranged from 6 to 20 weeks. The weighted mean age of participants was 46.56 years with all studies only including patients aged 18 years or older, and the majority were female (71.23%). Publication dates ranged from 1999 to 2020. All studies determined the presence of partial remission of MDD on a symptom-level cut-off with a clinical interview. Time to relapse was only examined in one study.44 Study duration (ie, treatment and follow-up) ranged from 1.5 to 72 months, with a median follow-up of 13.5 months.

Treatment efficacy: depressive symptom severity

Overall, psychotherapy was associated with superior outcomes on depression severity immediately post-treatment (Hedges’ g=0.50; 95% CI 0.23 to 0.76; k=7), compared with control conditions (figure 2 and table 2). Heterogeneity was moderate (I2=51%; 95% CI 0% to 79%). A wide PI was observed, indicating that future studies might find no or small effects (95% PI −0.06 to 1.05). Separating this effect for different informants, psychotherapy was not associated with better outcomes compared with control conditions for clinician-rated symptom severity (Hedges’ g=0.37; 95% CI −0.04 to 0.77; I2=38%; 95% CI 0% to 79%; k=4), although such associations were found for self-reported severity (Hedges’ g=0.48; 95% CI 0.23 to 0.73; I2=46%; 95% CI 0% to 77%; k=7). At 1–12 months follow-up, the pooled effect size was small but non-significant (Hedges’ g=0.36; 95% CI −0.30 to 1.02; k=3), with moderate heterogeneity (I2=67%; 95% CI 0% to 91%). Similarly, depression severity pooled effect sizes were small and non-significant after follow-up of 12 months or more (Hedges’ g=0.02; 95% CI −0.09 to 0.12; I2=0%; 95% CI 0% to 90%; k=3).

Table 2

Effects for treatment on depressive symptomatology severity, remission and relapse rates

Figure 2

Forest plot of random effects analysis on depressive symptom severity at post-treatment, comparing treatment with control conditions.

A sensitivity analysis detected no outliers, and the funnel plot did not indicate small-study effects bias (online supplemental figure 1). The trim and fill procedure by Duval and Tweedie slightly attenuated the main effect size at post-treatment (Hedges’ g=0.44; 95% CI 0.14 to 0.73; I2=61.0%; 95% CI 15% to 82%; k=7), indicating that the effect might be slightly overestimated. Subgroup analyses revealed that treatment was associated with greater efficacy in studies using a TAU-only comparison group (Hedges’ g=0.59; 95% CI 0.32 to 0.86; I2=21%; 95% CI 0% to 88%; k=4), as opposed to studies with ADMc control conditions (Hedges’ g=0.24; 95% CI −0.31 to 0.79; I2=14%; 95% CI 0% to 91%; k=3) (psubgroup=0.023). Additional subgroup analyses, meta-regression and Egger’s test could not be performed because of a lack of studies.

Treatment efficacy: remission and relapse rates

Patients receiving psychotherapy were about two-and-a-half times more likely to achieve full remission post-treatment, compared with control conditions (OR 2.57; 95% CI 1.71 to 3.87; I2=0%; 95% CI 0% to 79%; k=5). At follow-up 6 months or longer, patients receiving treatment were almost twice as likely to be in full remission (OR 1.75; 95% CI 1.21 to 2.53; I2=0%; 95% CI 0% to 90%; k=3), indicating that treatment was superior to control conditions in attaining full remission. These results are presented in table 2. Pooled remission rates immediately post-treatment were estimated at 44.0% for the treatment group (95% CI 27.1% to 62.5%) and 20.6% for the control conditions (95% CI 10.4% to 36.7%). Remission rates were higher at follow-up 6 months or longer in both treatment (53.7%; 95% CI 34.4% to 72.0%) and control conditions (37.0%; 95% CI 15.6% to 65.1%).

Immediately post-treatment, psychotherapy was not significantly associated with differences in the odds of relapse, compared with those in the control conditions (OR 0.17; 95% CI 0.01 to 4.83, I2=63%; 95% CI 0% to 89%; k=3). Similarly, as can be seen in table 2, after follow-up 6 months or more, no significant associations were found (OR 0.46; 95% CI 0.21 to 1.03, I2=0%; 95% CI 0% to 90%; k=3). Pooled relapse rates immediately post-treatment were 5.7% for the treatment group (95% CI 0.4% to 48.3%) and 29.0% for the control conditions (95% CI 7.2% to 68.4%). These were somewhat higher at follow-up 6 months or longer, with 15.6% (95% CI 6.6% to 32.8%) experiencing a relapse after treatment compared with 30.6% (95% CI 14.5% to 53.3%) in the control conditions.

Quality assessments

Overall, the risk of bias in the included studies was considerable. Methodological quality was rated as low with a high risk of bias (figure 3 and online supplemental figure 2). Two studies41 44 were rated as having ‘some concerns’ and five28 39 40 42 43 were judged at high risk of bias (see justifications in online supplemental appendix 4). The quality of the evidence (GRADE) was downgraded and judged at very low—low for all pooled outcomes (see online supplemental table 3).

Figure 3

Quality of the evidence: weighted risk-of-bias plot of included trials.

Conclusions and clinical implications

Partial remission of MDD is a debilitating clinical state related to relapse, chronicity, morbidity and dysfunction. As about one-third of patients with MDD achieves only partial remission,3–7 and the best treatment option for partial remission is unclear, objectively synthesising the available evidence becomes imperative. This systematic review and meta-analysis is the first to assess treatment efficacy, specifically focused on partial remission. We identified seven RCTs examining effects of CBT-derived psychotherapy, while, surprisingly, only three pharmacotherapy RCTs were found. This systematic review gives a first indication suggesting that psychotherapy for patients with partial remission of MDD is associated with superior effects on short-term depressive symptom severity and remission rates, compared with control conditions. That is, patients receiving treatment that targets partial remission were almost two-and-a-half times more likely to achieve full remission. While remission rates remained superior at follow-up, the effects on symptomatology wane in the long term after treatment cessation. Moreover, there seems to be a minimal effect of psychotherapy on relapse rates compared with control groups, in this at-risk group.

It is disappointing that treatment gains on depressive symptomatology, while evident for remission status, were not maintained at follow-up. This may likely be explained by the limited studies available at follow-up (k=3). Yet, short-term effects are still of clinical relevance, as partial remission is a robust predictor of chronicity.3 5 7 15–19 Moreover, previous meta-analyses showed that psychotherapy for acute MDD, including CBT, has enduring and prophylactic effects even after treatment termination, comparable to continuing pharmacotherapy after remission.45 46 In contrast, preventive effects of ADM disappear when they are discontinued or tapered.47 Another explanation for limited long-term outcomes in our meta-analysis may be that individuals in partial remission of MDD represent a distinct population in which different mechanisms contribute to the maintenance of symptomatology and relapse. Note that these short-term effects are consistent with those found for subthreshold depression, where psychotherapy had a small-to-moderate effect on symptomatology immediately post-treatment.48 Notably, treatment effects were significantly larger in studies using TAU-only control conditions, compared with studies in which all patients were taking ADM. This is unsurprising, as ADM is an established treatment option for MDD and would thus comprise an active control condition, leading to fewer group differences.8 10 Furthermore, treatment was associated with superior remission rates at both post-treatment (44%) and follow-up (54%).

In contrast to studies focusing on relapse prevention in remitted MDD, we found no significant effect of psychotherapy on relapse rates. While the overall relapse rate was lower than expected,3 5 16 18 there was a pronounced trend at follow-up indicating superior associations with treatment compared with control conditions. It should be noted that the effects may be underestimated due to a lack of statistical power, as well as the time it takes to observe effects on relapse rates. These findings seem to support the distinction between relapse rates as an indication of long-term durability of effects, compared with remission. Longer follow-up times may thus be a more representative exposure to estimate efficacy of relapse prevention.49 Contrary to our findings, such preventive effects of psychotherapy have repeatedly been found among patients with remitted MDD.9

Because pharmacological RCTs targeting partial remission are scarce, we could not provide a robust estimate of their effectiveness. Pharmacotherapeutic trials examined different augmentation strategies but yielded conflicting results. Nevertheless, the current study overall provides initial evidence to suggest that offering psychotherapy specifically to patients in partial remission is associated with superior outcomes. These results should still be interpreted with caution, given the limited number of studies, as well as differences in methodological design and treatments. Therefore, importantly, no firm conclusions can be made for some outcomes and treatment modalities, and results are insufficiently robust to definitively guide clinical practice. Nevertheless, findings inform shared decision-making as they suggest that CBT may be at least for the short term an effective treatment for patients in partial remission. Future research should address these issues to optimise depression outcomes in partially remitted patients and reduce their future risk profile.

Our systematic review and meta-analysis has several strengths. To our knowledge, it is the first systematic study to provide an overview of treatment options specifically targeting partial remission in MDD and their efficacy. Additionally, we used a rigorous study protocol with strict inclusion criteria, followed a comprehensive search strategy and, most importantly, stated a clear definition for partial remission of MDD to increase sample homogeneity, in line with previous research.7 27 28

However, several limitations should be noted. Our study was limited by the number of RCTs available, and the relatively small samples in some trials, especially at follow-up. Especially pharmacotherapy RCTs specifically targeting partial remission and trials examining relapse rates were scarce. This may have led to underpowered designs and results should be interpreted with caution. Another source of uncertainty is the considerable heterogeneity and its wide CIs, possibly due to differences in the pooled interventions (eg, treatment length or self-help vs operator-led interventions) or follow-up length. We were unable to adequately examine these differences using extracted study and intervention characteristics by applying subgroup and meta-regression analyses, given the number of studies required to ensure adequate power. Furthermore, the level of evidence was rated as very low or low, due to an overall poor methodological quality of primary studies. While the risk of bias was considerable, our results may still provide valuable input for clinical guidelines. Even more so, it helps setting the agenda for future research by underscoring the importance to conduct high-quality trials with longer follow-ups to examine the efficacy of treatments that target partial remission. Lastly, this meta-analysis was inherently limited by the availability and usage of aggregate data, and we could not take individual patient profiles, including prognostic or moderating factors, into account. Future research should pool RCT data using an individual patient data meta-analysis approach, offering more power and precision. This might allow for the differentiation of specific patient profiles. Future research on the efficaciousness of treatment for this at-risk population should incorporate such a personalised medicine approach, to improve clinical practice.

Given the chronic trajectory of depression in patients in partial remission,3 17 20 we call for well-designed and adequately powered studies to examine innovative, sequential or combined treatment options to get patients well and facilitate endurance of effects. Treatment aimed at mechanisms underlying poor response may help keep patients well. For example, cognitive impairments are commonly seen in partial remission,12 and predict poorer treatment response.50 Specifically targeting such underlying mechanisms may help improve outcomes.