This NMA of 69 randomized, controlled trials comprising 10,285 participants and 25 different therapies for TRD found that six of the 25 included treatments were associated with significantly higher response rates than placebo: ECT, minocycline, TBS, rTMS, ketamine, and aripiprazole. The strongest effects were found for neuromodulatory treatments followed by antipsychotics and agents that target the NMDAR. Notably, this study is the first to incorporate neuromodulatory treatments alongside both established antidepressants and novel, rapid-acting antidepressants [6]. Previous analyses either did not incorporate recent evidence from ketamine and SPs [4] or excluded neuromodulatory treatments [5, 7, 8] and other pharmacological agents [6, 12]. Previous studies also used a broader definition of TRD that allowed for lack of response to just one rather than two antidepressant trials; the latter is the accepted definition of treatment-resistance adopted by the EMA and FDA [1, 2, 7, 8, 12].

In addition to the six treatments associated with significantly higher response rates than placebo, ECT, TBS, rTMS, and ketamine were also effective when considering both total endpoint depression score and remission. Aripiprazole and the olanzapine/fluoxetine combination treatment were effective when considering remission only. The study also found that antipsychotics (quetiapine, olanzapine, aripiprazole, and brexpiprazole) as well as fluoxetine and olanzapine/fluoxetine combination treatment were significantly less well tolerated than placebo. While these results are based on a substantial amount of evidence for most treatments (e.g., aripiprazole, n = 1933; rTMS, n = 1810; ketamine, n = 1109; TBS, n = 683; olanzapine/fluoxetine combination, n = 477), for other treatments fewer studies meeting relevant inclusion criteria were available (e.g., minocycline, n = 41; ECT, n = 208).

Interestingly, this analysis found that ECT was most likely to be the most efficacious treatment, though it should be noted that few randomized, controlled trials exist for ECT. Previous NMAs drawing from head-to-head studies found mixed results regarding the efficacy of ECT for TRD [6, 8, 12]. Our analysis relied on three head-to-head studies that compared ECT to rTMS or tDCS. Consistent with our findings, a recent head-to-head study suggested that ECT may be more effective than ketamine in participants with MDD, though treatment resistance was not explicitly defined [13]. However, the largest comparative effectiveness trial to date, which included mostly outpatients with TRD, found that ketamine was noninferior to ECT [14]. Nevertheless, that study also found a considerable preference for ketamine (31 participants receiving ECT dropped out of the study versus four receiving ketamine), and typical responders to ECT may have been underrepresented (eg, inpatients or those with severe or catatonic depression). Broadly, however, the evidence is limited; there has been no sham-controlled study of ECT for either TRD or non-TRD since 1985.

Crucially, convincing antidepressant effects were also observed for other neuromodulatory treatments such as rTMS and TBS, consistent with previous evidence [4, 6, 12]. Robust effects were noted across all efficacy outcomes, and these were sustained in sensitivity analyses; good tolerability was also observed. Both rTMS and TBS were significantly more effective than aripiprazole but showed only a numerical advantage compared to ketamine. However, similar to previous studies, ECT was found to be superior to rTMS [15]. It should be noted that TBS was noninferior to rTMS in the present analysis, which may justify use of this more convenient modality.

Echoing previous analyses, this study found promising outcomes for ketamine and, to a lesser degree, other NMDAR-targeting medications like minocycline [5, 6, 16]. This finding is corroborated by recent evidence demonstrating the superior efficacy of ketamine compared to the traditional augmentative agent quetiapine [17]. Interestingly, the only differences observed between different forms of ketamine were found when including the TRANSFORM studies, specifically between intranasal (S)-ketamine, with or without an additional oral antidepressant, and IV-administered ketamine. This finding aligns with the results of Bahji and colleagues, who reported a tendency for IV racemic ketamine to be more efficacious than intranasal (S)-ketamine [18]. However, racemic IV ketamine studies were generally of shorter duration, highlighting the need for research on the long-term effects of this treatment.

Interestingly, other NMDAR antagonists such as nitrous oxide and lanicemine exhibited no significant antidepressant effects in our analysis. In this context, it should be noted that ketamine may exert its antidepressant effects via several different mechanisms of action [19]. For example, anti-inflammatory processes appear to play a role in the antidepressant effects of both ketamine and minocycline [19, 20]. Nevertheless, a recent systematic review [21] that also included a large randomized controlled trial with 173 patients who had not responded to only one previous antidepressant [22] found inconclusive results. It is also important to note that this analysis included only a single study on minocycline, which also stood out as an outlier in terms of severity of depressive symptoms (mean HAM-D score = 34.5). Broadly, however, the existing evidence suggests that individuals with more severe depression tend to exhibit a better response to antidepressants but a poorer response to placebo [23], which may have inflated the effect of minocycline.

Antipsychotic augmentation with quetiapine or brexpiprazole, despite being FDA-approved, was generally not effective for treating TRD in our analysis. Similar to a prior analysis [16], aripiprazole was an exception and displayed modest antidepressant effects. Consequently, the significant group-level effect of antipsychotics was largely attributable to the evidence supporting the efficacy of aripiprazole. This particularity might be attributable to aripiprazole’s distinct receptor profile compared to other atypical antipsychotics; specifically, it acts as a partial agonist at the 5-HT1A, D2, and D3 receptors and as a 5-HT2A receptor antagonist. In support of this hypothesis, brexpiprazole—considered the successor to aripiprazole and also displaying activation at the 5-HT1A, D2, D3, and 5-HT2A receptors—had the second largest effect size among antipsychotics after aripiprazole. Brexpiprazole’s lack of efficacy in the present analysis may be explained by our use of response rate—a more conservative measure—whereas the original studies used change in depressive symptom score as their primary outcome. Notably, all six aripiprazole and all five brexpiprazole trials included in our analysis were sponsored by pharmaceutical companies, whereas the evidence for all other effective treatments was at least partially drawn from trials not sponsored by industry. Regarding quetiapine, it should be noted that FDA approval was based on studies that adopted the less stringent definition of treatment resistance as lack of response to a single antidepressant and which were thus not included here [24]. Moreover, when analyzing antidepressant classes, both neuromodulatory treatments and NMDAR-targeting agents were more efficacious than antipsychotic augmentation. Antipsychotics were also the least tolerated drug class in our analysis. Specifically, antipsychotic treatments have been associated with a range of side effects, including metabolic dysfunction, weight gain, tardive dyskinesia, and possibly increased mortality compared to other antidepressants [25, 26]. These results require the cautious use of these compounds as augmentative strategies for TRD.

Interestingly, no significant effect was seen for lithium in treating TRD, in contrast to previous analyses [27, 28]. However, this discrepancy may have been due to the fact that prior analyses defined TRD as lack of response to one antidepressant rather than two as well as the inclusion of individuals with bipolar depression. Another point of interest is that this NMA is the first to also include SPs, specifically psilocybin and ayahuasca. Despite moderate ORs, these agents displayed no significant antidepressant effects. Subsequent studies are needed to ascertain whether these observed trends can be substantiated.

Despite these valuable findings, the analysis is also associated with several limitations. First, most of the included studies were conducted over a relatively brief period that rarely exceeded 6-8 weeks. As a salient example of the implications of this limitation, fluoxetine studies found that 31-41% of patients who showed no improvement after 6 weeks nevertheless achieved remission by week 12 [29]. Second, time points at which outcomes were assessed varied across studies, with some outcomes measured within days (e.g., ketamine) and others over months (e.g., DBS). While most studies had a duration of several weeks, this discrepancy still limits the comparability of treatments based on a uniform timeline. Third, our analysis could not determine the longevity of the observed antidepressant effects. Addressing these questions requires more time-consuming and ethically complex research comparing treatments over several months to years. Fourth, our study provides no evidence regarding individual or group-level factors that might influence the effect of a given treatment, such as comorbidities or combination with another antidepressant treatment. Relatedly, the included trials were conducted over a timespan of several decades and, thus, methodological quality and the characteristics of included patients might vary. However, the sensitivity analyses largely demonstrated the robustness of our results across various factors such as severity of depressive symptoms, sex, and age. The lack of significant effects for some treatments (e.g., ECT, minocycline) but not others (e.g., rTMS, ketamine) in specific sensitivity analyses underscores the scarcity of the evidence for these treatments. Finally, alternative treatment approaches (e.g., psychotherapy, light therapy, sleep deprivation, aerobic exercise) were excluded from our analysis. The challenge of employing a placebo for these treatments complicates their comparison with the included therapies. Nevertheless, there is evidence for the efficacy of psychotherapy plus antidepressant treatment compared to antidepressant treatment alone for TRD [30].

In conclusion, the findings of this NMA underscore the efficacy of several available treatments for TRD, including ECT, rTMS, TBS, ketamine, aripiprazole, minocycline, and the combination of olanzapine/fluoxetine; the first four treatments demonstrated efficacy across all outcomes. The analysis also found no effects for mood stabilizers and antipsychotics other than aripiprazole and a low tolerance for all included antipsychotics. These results provide valuable information for researchers and clinicians, and the novel evidence presented herein may help guide the selection of antidepressant treatments for individuals with TRD.