Objectives:

To inform future discussions and decisions about how to define treatment-resistant depression (TRD) and specify the important outcomes measured in research studies, and to clarify how trials or observational studies might best be designed and conducted to inform clinical practice and health policy.

Data sources:

To provide a comprehensive understanding of how experts and investigators have defined and studied TRD, we first performed a narrative review of relevant literature. We considered consensus statements, practice guidelines, government materials, and other literature published from 1/1/1995 through 8/18/2017, except for systematic reviews (limited to start 1/1/2005). Next, we performed a systematic review of published studies of TRD interventions (1/1/2005 through 8/18/2017) indexed in MEDLINE®, EMBASE, PsycINFO, and Cochrane Library.

Review methods:

Trained personnel dually reviewed all titles and abstracts for eligibility. Studies marked for possible inclusion by either reviewer and those with inadequate abstracts underwent dual full-text review. Disagreements were resolved by consensus discussion. One member of the research team abstracted data; a senior investigator reviewed abstractions for accuracy and completeness.

Results:

Our narrative review indicated that no consensus definition existed for TRD. We identified four basic definitions for TRD (3 for major depressive disorder [MDD]; 1 for bipolar disorder). Based on frequency of reporting in the literature, the most common TRD definition for MDD required a minimum of two prior treatment failures and confirmation of prior adequate dose and duration. The most common TRD definition for bipolar disorder required one prior treatment failure. For all TRD definitions, no clear consensus emerged on defining adequacy of either dose or duration. Little agreement exists about the best approach to diagnose TRD or the preferred outcome measure, although the Hamilton Depression Rating Scale was the most used. We found general agreement about minimizing bias by using randomization; studies have not focused on minimizing placebo effects. Evidence about the risk factors (e.g., age, sex, number of prior failed treatments, and length of current depressive episode) associated with TRD and data to assess potential prognostic factors were limited.

Only 17 percent of intervention studies enrolled study populations that met frequently specified criteria for TRD. Most studies (88%) were randomized controlled trials; all studies applied some exclusion criteria to limit potential confounders. Depressive outcomes and clinical global impressions were commonly measured; functional impairment and quality-of-life tools were rarely used.

Conclusions:

No agreed-upon definition of TRD exists; although experts may converge on two as the best number of prior treatment failures, they do not agree on definitions for adequacy of either dose or duration or outcomes measures. Critical to advancing TRD research are two key steps: (1) developing a consensus definition of TRD that addresses how best to specify the number of prior treatment failures and the adequacy of dose and duration; and (2) identifying a core package of outcome measures that can be applied in a standardized manner. Our recommendations about stronger approaches to designing and conducting TRD research will foster better evidence to translate into clearer guidelines for treating patients with this serious condition.