Challenges of research on treatment‐resistant depression: a clinician's perspective

I think we should be grateful to McIntyre et al ¹ for their extraordinarily thorough and balanced review of treatment‐resistant depression (TRD). They note that this condition poses a plethora of clinical research challenges. Here I offer a few suggestions that might make research more cost‐efficient and clinically generalizable.

First, we should develop tools to systematically identify treatable causes of depression. Many medical conditions (e.g., neurological, infectious, oncological, endocrine) as well as medications may cause depression that appears to be treatment resistant. Without a thorough medical review, depression‐causing disorders and medications will escape notice.

To my knowledge, there is no evidence‐based or widely agreed upon set of laboratory tests with demonstrated “yield” rates to detect treatable medical, iatrogenic and even post‐surgical causes of depression in any patient population. This issue is analogous to searching for treatable causes of cognitive impairment before diagnosing Alzheimer's disease. We need to launch a multi‐care‐system clinical research effort to gather evidence in apparent TRD patients and key subgroups (e.g., elderly, medically underserved) by which to prioritize laboratory, neuropsychological and other assessments to effectively identify treatable causes of depression, before embarking on an often‐lengthy series of TRD treatments.

Second, TRD studies should recruit both early and delayed treatment failures, to reduce cost and enhance generalizability. Regulatory authorities require unsatisfactory acute‐phase treatment response to at least two well‐delivered antidepressant regimens. However, both acute and later (continuation or maintenance phase) failures are commonly seen in practice ² , ³ . Late failures are not rare. For example, relapse rates in active medication arms in randomized, placebo‐controlled continuation and maintenance phase studies of recurrent depression range from 3 to 45% ³ , with the majority around 20‐25%. Most of these relapses typically occur within 3‐4 months after the successful acute‐phase treatment in continuation trials. These are cases of TRD, because they have failed on the same initially successful acute treatment – just a bit later. Indeed, the greater the number of initially failed acute‐phase treatment trials, the greater the relapse rate and the sooner relapse occurs on the active treatment ² .

Whether acute and delayed failures are biologically distinct is unknown. Certainly, they present the same clinical challenge: the failure of a specific medication to which most clinicians will not return. Their treatment options are the same: augmentation, combination, switch, or dose adjustments. Many clinicians believe that patients who respond acutely but cannot sustain the benefit (the late or delayed failures) are more likely to benefit from the next‐step treatment as compared to those who have no benefit at all acutely. To address this concern, one could stratify patients based on acute vs. delayed failure.

Another revision of TRD trial design to increase feasibility, generalizability and cost‐efficiency might be to include patients whose depression has been insufficiently responsive to either two sequential monotherapies or one monotherapy followed by an augmentation trial. In both treatment sequences, these depressions have not responded to two different agents in two distinct attempts, each of which is expected to have an antidepressant effect. This revision would include only augmentation agents approved either by the US Food and Drug Administration or the European Medicines Agency (e.g., aripiprazole or quetiapine), or possibly lithium, given the positive randomized controlled trials (RCTs). While this revision also raises the issue of heterogeneity, patients with TRD are biologically heterogeneous regardless of the types and numbers of prior treatment failures. Again, stratification could address this concern if needed.

TRD trials to evaluate a next treatment step, whether monotherapy or augmentation, would simply require that eligible participants’ depressions be severe and persistent enough to call for a new treatment and that participants consent. For those switching (as in studies of new monotherapies), the control could be a low dose of the experimental agent or placebo. If the aim is to evaluate a new adjunctive agent, those choosing to switch would be ineligible. This approach avoids the issue of deciding how much of a prior benefit is needed to enter either type of study. In short, wouldn't a sample of TRD patients with either acute or delayed failures from representative treatment sequences (monotherapy‐monotherapy or monotherapy‐augmentation) be more ecologically valid (be a truer representation of the most common types of TRD), more generalizable and less costly, yet with well‐protected internal validity?

On the other hand, ensuring both an adequate dose and duration for each of the two failed treatment trials is essential to establishing valid cases of TRD. In typical practice, doses are not titrated consistently, and trial durations vary from 2 to 6 weeks. These shortcomings are often addressed with standardized tools, such as the Antidepressant Treatment History Form (ATHF) ⁴ or the Maudsley staging method ⁵ . However, these approaches use 4‐6‐week thresholds to define an adequate duration. While earlier responses are common, especially in non‐TRD patients, even 6 weeks is likely too short, as suggested by the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) trial ⁶ , in which half of the remissions and one third of the responses occurred after 6 weeks in the first step. We found similar results from STAR*D in the second medication switch step, with half the responses occurring after 6 weeks ⁷ .

More recent RCTs conducted for regulatory approval also indicate that a 6‐week threshold is too short, because response rates in the placebo or control arms were much higher than the “expected” 15% based on the third step of STAR*D². For example, in a pivotal trial with esketamine ⁸ , there was a 52% response rate in the control condition (initiating a new antidepressant along with placebo), as compared to 64% for esketamine, following at least two failed prior trials, using the 6‐week threshold. Similarly, higher‐than‐expected response rates were reported in the placebo cell (35%) in a study of cariprazine as an adjunct treatment for TRD ⁹ . Notably, this study also required only the minimum effective dose to qualify for a prior “failed” trial, and just one failed trial was sufficient for study entry. Both factors likely contributed to the higher‐than‐expected placebo response rates in this “TRD” group.

The 15% remission rate in the third treatment step of STAR*D occurred after two prior steps each of which could take 12+ weeks, during which doses were driven to individually titrated maximum, using increases informed by symptom and side‐effect measures at each visit. Accepting patients with a minimum or “adequate” dose based on a staging method or historical assessment tool likely leads to acceptance of some underdosed apparent TRD patients into treatment trials.

In summary, the quality, generalizability, cost‐efficiency and validity of TRD trials could be improved by admitting a wider range of patients with acute and delayed failures from typical treatment steps, but trial durations and dosages should be elevated, when possible, to ensure that each case of TRD is valid.