McIntyre et al¹, in their excellent paper appearing in this issue of the journal, successfully tackle a critical issue in the field of depression: how should we define treatment-resistant depression (TRD) and how can we best manage it? They point out that a consensus definition of TRD with demonstrated predictive utility in terms of clinical decision-making and health outcomes does not currently exist, and that the definition adopted by the US Food and Drug Administration (FDA) and the European Medicines Agency (EMA) – i.e., failure to respond to two or more antidepressant regimens despite adequate dose and duration and adherence to treatment – remains at the moment the reference one.

Although it makes perfect sense that regulatory agencies rely on a definition which can be easily adopted in clinical trials, as it captures a large population with unmet needs, should researchers and clinicians routinely use the same approach? The Massachusetts General Hospital Staging Model (MGH-S) provides an example of a definition of TRD which integrates the number of failed trials with the intensity/optimization of each trial, without assuming a hierarchy of antidepressant classes². An observational study of patients receiving antidepressant therapy (N=78,477)³ applied this model to claims data from the MarketScan Research Databases over a 24-month time period. Annual costs for patients with mild TRD (MGH-S score: 3.5-4) were 1,530$ higher than those for non-TRD patients, and annual costs for patients with complex TRD (MGH-S score ≥6.5) were 4,425$ higher than those for non-TRD patients (all p<0.001). A 1-point increase in the MGH-S score was associated with a 590$ increase in annual costs (p<0.001). There is, therefore, a clinical utility in adopting a staging method to evaluate the cost-effectiveness of new treatments for TRD.

The MGH-S model has been recently updated to reflect some of the new treatments for TRD, including ketamine/esketamine, transcranial magnetic stimulation (TMS) and vagus nerve stimulation. The new version of the model⁴ provides a score for the characteristics of depression (including severity of the episode, presence or not of psychotic features, presence of suicidal ideation, and presence of anxious distress) (maximum score = 10), and a score for treatment history, considering the number of medication trials, the number of augmentation treatment trials, and the use of the above-mentioned new treatments as well as of electroconvulsive therapy (ECT) (maximum score = 25). We look forward to a wider adoption of this model in research and advanced clinical settings.

As pointed out by McIntyre et al, intravenous ketamine and intranasal esketamine (co-administered with an antidepressant) have an established efficacy in the management of TRD, while some second-generation antipsychotics are proven effective as adjunctive treatments to antidepressants in partial responders, but only the olanzapine-fluoxetine combination has established efficacy in FDA-defined TRD. However, despite the current FDA indication, the results of a pooled analysis⁵ suggest that adjunctive aripiprazole can be an effective intervention for patients whose symptoms worsen during antidepressant monotherapy, challenging the view that its benefits are limited to partial responders to antidepressants. The same may be true for other second-generation antipsychotics, supporting the need for further investigations.

On the other hand, the authors highlight that ECT is regarded as an effective acute and maintenance intervention in TRD, with preliminary evidence suggesting its superiority over acute intravenous ketamine. The adoption of ECT in clinical practice, however, remains somewhat limited by the complexity of its administration and the possible adverse events.

The authors also argue that manual-based psychotherapies are not established as efficacious on their own in TRD, although offering significant symptomatic relief when added to conventional antidepressants. Nevertheless, the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) trial compared the effectiveness of cognitive therapy and pharmacotherapy as second-step strategies for outpatients with major depressive disorder (MDD) who had received inadequate benefit from an initial trial of citalopram⁶. Among participants who were assigned to second-step treatment, those who received cognitive therapy had similar response and remission rates to those assigned to medication strategies, suggesting that there may be a role for cognitive therapy in TRD.

The recent approval for the treatment of MDD of the combination of dextromethorphan (an uncompetitive N-methyl-D-aspartate receptor antagonist and sigma-1 receptor agonist) and bupropion (a norepinephrine-dopamine reuptake inhibitor), as well as the recent FDA filing of a new drug application for the GABAergic modulator zuranolone, raise the possibility that the scenario concerning medications available for MDD will change significantly in the near future. A large number of other novel compounds developed with non-monoamine molecular targets are currently in phase 2 or 3, again questioning whether clinicians in the next few years will continue to routinely use monoamine-based therapies in the initial algorithm for the treatment of depression. The current construct of TRD, focusing on the lack of response to what have been considered the first- or second-line monoamine-based treatments for MDD, may consequently become obsolete.

Regardless of the methodology used to assess TRD patients, it is absolutely critical to carefully select subjects for randomization in clinical trials. “Professional patients” or duplicate subjects are a common problem in TRD trials, and may threaten the integrity of these studies. A number of digital platforms have been developed to identify duplicate subjects and allow investigators to exclude them from trials, as well as new methodologies to better document the treatment history of patients, including the measurement of blood levels of the ongoing therapies.

To avoid the issues of diagnostic misclassification and severity of illness grade inflation, it is essential to ensure that patients enrolled for TRD trials really fulfill the needed requirements. Patients recruited in these trials may present with a heterogeneous group of symptoms representing several syndromes or subtypes, subsumed under the same diagnosis in the DSM-5 classification system. The SAFER interview⁷ has been developed to delineate a more symptom-specific and ecologically valid approach to the identification of the appropriate patients for MDD clinical trials through an independent assessment. It has been reported⁸ that, overall, 15.3% of MDD patients who had been deemed eligible at research sites were not eligible after the structured interview, with the most common reason being that patients did not meet the study requirements for level of treatment resistance. In MDD trials utilizing the SAFER interview as a tool to confirm eligibility, placebo response rates ranged between 13.0% and 27.3%, below the 30% to 40% average in antidepressant clinical trials, suggesting a benefit of the quality assurance provided by this interview. This reminds us of the importance to make sure that the right patients get into our TRD trials.

In conclusion, there are currently several challenges in the definition and management of TRD, and the most significant feature of McIntyre et al's paper is probably its ability to reflect this evolving scenario.

McIntyre 等人¹在本期杂志上发表的优秀论文中，成功解决了抑郁症领域的一个关键问题：我们应该如何定义难治性抑郁症 (TRD) 以及如何最好地管理它？他们指出，目前不存在在临床决策和健康结果方面具有预测效用的 TRD 共识定义，美国食品和药物管理局 (FDA) 和欧洲药品管理局 (EMA) 采用的定义）——即，尽管有足够的剂量、持续时间和坚持治疗，但对两种或多种抗抑郁药治疗方案没有反应——目前仍然是参考方案。

尽管监管机构依赖一个可以在临床试验中轻松采用的定义是完全有道理的，因为它涵盖了大量需求未得到满足的人群，但研究人员和临床医生是否应该经常使用相同的方法？马萨诸塞州总医院分期模型 (MGH-S) 提供了 TRD 定义的示例，该模型将失败试验的数量与每次试验的强度/优化相结合，而不假设抗抑郁药物类别的层次结构²。一项针对接受抗抑郁治疗的患者 (N=78,477) ^{3 的}观察性研究将该模型应用于 MarketScan 研究数据库 24 个月期间的索赔数据。轻度 TRD 患者（MGH-S 评分：3.5-4）的年度费用比非 TRD 患者高出 1,530 美元，复杂 TRD 患者（MGH-S 评分≥6.5）的年度费用比非 TRD 患者高出 4,425 美元。非 TRD 患者的数据（所有 p<0.001）。MGH-S 评分每增加 1 分，年度费用就会增加 590 美元（p<0.001）。因此，采用分期方法来评估 TRD 新疗法的成本效益具有临床实用性。

MGH-S 模型最近进行了更新，以反映 TRD 的一些新疗法，包括氯胺酮/艾氯胺酮、经颅磁刺激 (TMS) 和迷走神经刺激。新版本的模型⁴提供了抑郁症特征的评分（包括发作的严重程度、是否存在精神病特征、是否有自杀意念以及是否存在焦虑困扰）（最高分 = 10），以及治疗史，考虑药物试验的数量、强化治疗试验的数量以及上述新疗法以及电休克治疗 (ECT) 的使用（最高分 = 25）。我们期待在研究和高级临床环境中更广泛地采用该模型。

正如 McIntyre 等人所指出的，静脉注射氯胺酮和鼻内艾氯胺酮（与抗抑郁药联合给药）在治疗 TRD 方面具有明确的疗效，而一些第二代抗精神病药已被证明可以作为部分缓解者中抗抑郁药的辅助治疗有效，但只有奥氮平-氟西汀组合已确定对 FDA 定义的 TRD 有效。然而，尽管目前有 FDA 的指示，但汇总分析的结果⁵表明，辅助阿立哌唑可以成为抗抑郁药物单一治疗期间症状恶化的患者的有效干预措施，这对“其益处仅限于抗抑郁药物部分反应者”的观点提出了质疑。对于其他第二代抗精神病药物来说可能也是如此，这支持了进一步研究的必要性。

另一方面，作者强调，ECT 被认为是 TRD 的有效急性和维持干预措施，初步证据表明其优于急性静脉注射氯胺酮。然而，ECT 在临床实践中的采用仍然受到其管理的复杂性和可能的​​不良事件的限制。

作者还认为，尽管手动心理疗法与传统抗抑郁药物相结合可以显着缓解症状，但其本身对 TRD 的治疗效果并不理想。尽管如此，缓解抑郁症的序贯治疗替代方案 (STAR*D) 试验比较了认知疗法和药物疗法作为重度抑郁症 (MDD) 门诊患者的第二步策略的有效性，这些患者从西酞普兰初始试验中获益不足⁶。在分配到第二步治疗的参与者中，接受认知治疗的参与者与分配到药物策略的参与者有相似的反应和缓解率，这表明认知治疗可能在 TRD 中发挥作用。

最近批准右美沙芬（一种非竞争性 N-甲基-D-天冬氨酸受体拮抗剂和 sigma-1 受体激动剂）和安非他酮（一种去甲肾上腺素-多巴胺再摄取抑制剂）的组合用于治疗 MDD，以及最近向 FDA 提交的文件GABA 能调节剂 zuranolone 的新药申请，增加了有关 MDD 药物的情况在不久的将来将发生重大变化的可能性。用非单胺分子靶点开发的大量其他新型化合物目前正处于 2 期或 3 期阶段，再次质疑临床医生在未来几年是否会在抑郁症治疗的初始算法中继续常规使用基于单胺的疗法。当前的 TRD 结构侧重于对 MDD 的一线或二线单胺治疗缺乏反应，因此可能会过时。

无论使用何种方法评估 TRD 患者，在临床试验中仔细选择随机受试者绝对至关重要。“专业患者”或重复受试者是 TRD 试验中的常见问题，并可能威胁到这些研究的完整性。已经开发了许多数字平台来识别重复的受试者，并允许研究人员将他们排除在试验之外，以及新的方法来更好地记录患者的治疗历史，包括测量正在进行的治疗的血液水平。

为了避免诊断错误分类和疾病严重程度分级膨胀的问题，必须确保参加 TRD 试验的患者真正满足所需的要求。这些试验中招募的患者可能会出现代表多种综合征或亚型的一组异质症状，这些症状在 DSM-5 分类系统中属于同一诊断。SAFER 访谈^{7 的}开发旨在描述一种更具症状特异性且生态学上有效的方法，通过独立评估来识别适合 MDD 临床试验的患者。据报道⁸总体而言，在研究中心被认为符合资格的 MDD 患者中，有 15.3% 在结构化访谈后不符合资格，最常见的原因是患者不符合研究对治疗抵抗水平的要求。在使用 SAFER 访谈作为确认资格的工具的 MDD 试验中，安慰剂反应率在 13.0% 至 27.3% 之间，低于抗抑郁药物临床试验中 30% 至 40% 的平均水平，这表明该访谈提供的质量保证是有益的。这提醒我们确保合适的患者参加我们的 TRD 试验的重要性。

总之，目前 TRD 的定义和管理存在一些挑战，McIntyre 等人的论文最显着的特点可能是它能够反映这种不断发展的情况。