Discussion

The DEPRE’5 study, a pragmatic multicentre five-arm randomised controlled trial, assessed the efficacy of different treatment strategies for patients with MDD after an unsuccessful SSRI treatment. SSRI dose optimisation resulted in the lowest response rates, suggesting other alternatives should be considered. No significant differences in primary outcomes were found when comparing each strategy individually with the SSRI optimisation arm. However, the lower-than-planned recruitment rate could explain the lack of greater benefits of some of the explored treatments. Switching to venlafaxine, combining SSRIs with psychotherapy and, to a lesser extent, lithium augmentation offered better outcomes in terms of response rate, change in depressive symptoms and/or global improvement compared to dose optimisation. All conditions were safe, but patients receiving psychotherapy experienced fewer adverse events.

Although antidepressant dose optimisation is debated, ^{Reference Jakubovski, Varigonda, Freemantle, Taylor and Bloch8,Reference Furukawa, Cipriani, Cowen, Leucht, Egger and Salanti9} real-world prescription patterns show that it is still a common approach for managing depressive episodes. ^{Reference Gauthier, Guérin, Zhdanava, Jacobson, Nomikos and Merikle10} Surprisingly, SSRI monotherapy remains popular even after a second treatment failure, with long delays before exploring alternative strategies such as augmentation or combination. ^{Reference Jain, Higa, Keyloun, Park, Bonafede and Tung19,Reference Gillain, Degraeve, Dreesen, De Bruecker, Buntinx and Beke20} While higher SSRI doses may boost efficacy, ^{Reference Jakubovski, Varigonda, Freemantle, Taylor and Bloch8} we found that using a second-line approach – either expanding the mechanism of action or adding psychotherapeutic interventions – doubled the likelihood of response in the case of inadequate SSRI response.

Switching to venlafaxine showed potential as one of the more beneficial second-line strategies in our sample. While somewhat controversial, previous literature supports the increased effectiveness of dual-acting antidepressants in patients with severe depression or inadequate SSRI response. ^{Reference Papakostas, Fava and Thase21,Reference Bradley and Lenox-Smith22} In this trial, most patients on venlafaxine received at least 150 mg/day, a dose modulating both monoaminergic pathways without compromising tolerability. ^{Reference Furukawa, Cipriani, Cowen, Leucht, Egger and Salanti9}

Lithium augmentation showed a trend towards greater symptom reduction at the end-point. However, the small sample size may have limited the ability to demonstrate more robust benefits. Although mean plasma levels were within the desired range, the SSRI + Li group showed lower adherence, which may have also affected the outcomes. Lithium users experienced more tremors and gastrointestinal discomfort but with no differences in drop-out rates. Meta-analyses support the efficacy of lithium augmentation in cases of inadequate antidepressant response, even when combined with SSRIs. ^{Reference Bauer, Adi, Ricken, Sverus and Pilhatsch23,Reference Strawbridge, Carter, Marwood, Bandelow, Tsapekos and Nikolova24} Preliminary analyses suggest it may be more cost-effective than other augmentation strategies, such as atypical antipsychotics, with similar acceptability and tolerability. ^{Reference Edwards, Hamilton, Nherera and Trevor25} The low usage rate may be because of clinicians’ inexperience, the need to monitor lithium levels or concerns about its narrow safety range. Despite less promotion and research, lithium remains a long-standing, consistent strategy in MDD. Further research should clarify whether lithium augmentation’s benefits are specific to certain groups of patients. ^{Reference Bauer, Adi, Ricken, Sverus and Pilhatsch23,Reference Álvarez, Pérez-Solá, Pérez-Blanco, Queraltó, Torrubia and Noguera26}

Our results support the utility of PST, a structured and goal-oriented cognitive–behavioural intervention, chosen for its effectiveness, applicability and limited duration, comparable to the other treatment arms. ^{Reference Bell and D’zurilla27} Beyond the STAR*D study, ^{Reference Thase, Friedman, Biggs, Wisniewski, Trivedi and Luther28} limited research has compared the effectiveness of combining psychotherapy with antidepressants in patients with previous treatment failures. Initial evidence suggests that combined treatment may be more effective than single treatments. ^{Reference Malhi, Bell, Bassett, Boyce, Bryant and Hopwood7,Reference Cuijpers, Noma, Karyotaki, Vinkers, Cipriani and Furukawa29} Although SSRIs relieve some symptoms faster than psychotherapy, the latter, when combined with antidepressants, can be more effective in achieving behavioural activation and positive experiences, crucial aspects for a full recovery. Psychotherapy, alone or combined, is often better accepted by patients with mild to moderate depression. This matches the strong adherence seen in the SSRI + PST group and the lower adverse event rate.

The low remission rates are consistent, albeit slightly lower than those observed in comparable pragmatic studies. ^{Reference Mohamed, Johnson, Chen, Hicks, Davis and Yoon30,Reference Lindhiem, Bennett, Trentacosta and McLear31} In contrast to the STAR*D study, which relied on an equipoise-stratified randomisation scheme, our participants were randomised using a classic randomisation approach regardless of the patient and clinician preferences, and raters were blinded. These factors influence expectations, treatment adherence and perceived efficacy. ^{Reference Dunlop, Kelley, Aponte-Rivera, Mletzko-Crowe, Kinkead and Ritchie32} In clinical practice, choosing a particular second-line strategy relies on the degree of the previous antidepressant response (i.e. optimisation, augmentation or combination in the case of a partial response versus switching in the case of null response) or the type of predominant symptoms. ^{Reference Bauer, Severus, Möller and Young33–Reference Korte, Prins, Krajnc, Hendriksen, Oosting and Westphal35} Unfortunately, we could not assess differences between groups based on the previous dose or the degree of response to the initial SSRI. The presence of specific stressful life events or maladaptive coping strategies could make psychotherapy the most appropriate option. Further research should identify clinical and neurobiological features that could moderate treatment responses to these alternatives, leading to more effective therapeutic decisions.

The duration of the depressive episode is one of the factors most clearly associated with the risk of persistence and treatment failure. ^{Reference Bartova, Dold, Kautzky, Fabbri, Spies and Serretti36} In our study, almost 30% of the participants had an episode lasting over a year, over half had experienced previous episodes, 48% had at least one first-degree relative with a psychiatric history and nearly 10% had a lifetime history of suicidal behaviour. These factors could have influenced our patients’ likelihood of remission within 6 weeks. ^{Reference Bartova, Dold, Kautzky, Fabbri, Spies and Serretti36}

Judging by the symptom progression over the follow-up period, some participants may have achieved remission with the assigned treatment beyond that end-point. The VAST-D study, which tested the efficacy of different strategies in a sample of US veterans with MDD unresponsive to a first-line antidepressant, ^{Reference Mohamed, Johnson, Chen, Hicks, Davis and Yoon30} showed even lower remission rates at week 6 compared to our study, with improvements in the subsequent weeks. Eight to 12 weeks may be necessary to achieve the maximum symptomatic reduction with a new line of antidepressant treatment, especially after initial treatment failure. ^{Reference Bauer, Severus, Möller and Young33} In any case, our results support that the failure of one treatment reduces the odds of remission ^{Reference Rush, Trivedi, Wisniewski, Nierenberg, Stewart and Warden3} and underscores the need to use other strategies than dose optimisation.

Several limitations have to be acknowledged. First, the low recruitment rate resulted in a lower-than-expected sample size and compromised our statistical power. Recruitment was challenging because participating sites often encountered participants who had been treated with multiple therapeutic lines at the initial assessment. Site was not included as a covariate, although the centres’ homogeneity, standardised protocols and centralised randomisation likely mitigated site-related variability. In addition, the study’s lack of a double-blind design and the use of the LOCF approach for imputing missing data may have influenced our findings. Since depressive symptoms tend to improve over time, a more extended follow-up period would have been needed. The degree of prior SSRI response, as well as clinical and patient’s expectations are rarely assessed in similar trials but may influence real-world treatment choices and outcomes. For some, optimising the dose of an ineffective antidepressant can be discouraging; for others, concerns about trying a new therapeutic option may make adjusting the dose of a known antidepressant preferable. Baseline doses of SSRIs were mostly appropriate but not maximal, supporting flexible adjustments in the SSRI-Opt group based on prior response, tolerability and perceived efficacy. The neutral way in which participants were informed about the common use and potential benefits of each arm, along with randomised treatment allocation and blinded assessment, may have helped mitigate these effects. Finally, other treatment alternatives (i.e. atypical antipsychotic augmentation, SSRI with α2-antagonist combination or other psychotherapeutic approaches) were not explored. ^{Reference Strawbridge, Carter, Marwood, Bandelow, Tsapekos and Nikolova24,Reference Davies, Ijaz, Williams, Kessler, Lewis and Wiles37}

In conclusion, patients with MDD who do not respond to SSRIs may benefit from exploring alternative second-line strategies beyond simply increasing doses. With limited statistical power, switching to venlafaxine and adding psychotherapy yielded more consistent results compared to dose optimisation.