Key Considerations in the Pharmacological Management of Treatment-Resistant Depression

Abstract

Introduction:

Treatment-resistant depression (TRD) is a complex, multifactorial, and biologically heterogeneous disorder with debilitating outcomes. Understanding individual reasons why patients do not respond to treatment is necessary for improving clinical recommendations regarding medication regimens, augmentation strategies, and alternative treatments.

Areas covered in this review:

This manuscript reviews evidence-based treatment strategies for the clinical management of TRD. Current developments in the field and potential future recommendations for personalized treatment of TRD are also discussed.

Expert opinion:

Treatment guidelines for TRD are limited by the heterogeneous nature of the disorder. Furthermore, current strategies reflect this heterogeneity by emphasizing disease characteristics as well as drug trial response or failure. Developing robust biomarkers that could one day be integrated into clinical practice has the potential to advance specific treatment targets and ultimately improve treatment and remission outcomes.

1. Introduction

Major depressive disorder (MDD) is a chronic illness that is estimated to affect as many as one in five adults in the United States during their lifetime [1]. The World Health Organization (WHO) estimates that, worldwide, over 300 million people suffer from depression and that MDD is the leading cause of ill health and disability [2]. MDD, which has been associated with increased risk of death at any age [3], is most commonly treated with monoaminergic antidepressant medications, for which over 29 million prescriptions are written annually in the United States [4]. Despite the widespread use of antidepressants, many patients do not respond to treatment.

Treatment-resistant depression (TRD) is a complex, multifactorial, and biologically heterogeneous disorder defined by lack of clinical response to two successive antidepressant treatment trials of adequate dose and duration during the current major depressive episode. The prevalence of TRD is significant, encompassing an estimated 15–30% of MDD cases [5]. TRD is associated with increased risk of suicide, cognitive impairments, medical comorbidities, functional impairment, and economic costs [6]. In particular, the incidence of suicidal ideation and behavior associated with TRD is substantially higher than for individuals with MDD in general, and the life-time risk of suicide attempt is close to 30% [7]. Although few effective treatments are available specifically for TRD, current developments in the field suggest two areas of particular interest; the first is increasing biological insights into the pathophysiology of TRD, and the second is the development of mechanistically novel, rapid-acting therapeutics that act on the glutamatergic system.

This manuscript reviews evidence-based treatment strategies for the clinical management of TRD, discusses current clinical and research insights, and explores potential future recommendations for personalized treatment of TRD. Evidence was drawn from a literature search conducted using the following search terms: treatment-resistant depression, depression, chronic depression, antidepressant, psychopharmacology, outcome, and (bio)marker in addition to combinations of these terms. The search was conducted in PubMed, Scopus, and Google Scholar between August 2020 and December 2020, with no restrictions on time period. References from published treatment guidelines were also reviewed in detail. Citations were chosen based on quality of data and perceived relevance to the topics under discussion.

2. Clinical strategies for TRD

2.1. Selecting and optimizing initial antidepressants

In its categorization of depressive disorders, the Diagnostic and Statistical Manual of Mental Disorders (DSM-5) details several specifiers and clinical dimensions of clinical utility in the initial selection of antidepressants [8]. For instance, because sleep disturbance is common in depressive disorders and a key consideration in selecting treatment options, antidepressants such as mirtazapine, trazodone, and agomelatine are favored for patients with significant insomnia. For depressive disorders with psychotic features, superior outcomes have been reported for combination therapies that include an antidepressant and an antipsychotic versus either treatment alone or placebo [9] (see Section 2.4, below). In depression associated with cognitive dysfunction, vortioxetine is recommended because of its potential to improve processing speed, verbal learning, and recall domains [10]. Bupropion is often recommended when hypersomnolence and fatigue as well as weight gain or sexual dysfunction are concurrent [11]. Finally, for patients with comorbid painful conditions—particularly neuropathic pain and fibromyalgia—duloxetine and other serotonin norepinephrine reuptake inhibitors (SNRIs) are advised [12]. Extreme sensory processing patterns have also been linked to depression and hopelessness and may have a characteristic pattern in MDD patients [13], though evidence for a link between these patterns and particular treatment strategies have yet to be established.

Attempts to categorize subtypes of depression in a manner that could help predict both clinical course and response to antidepressant treatment date back to the categorization of atypical depression as a separate diagnosis; specifically, medication trials had shown that this group exhibited greater response to monoamine oxidase inhibitors (MAOIs) than to tricyclic antidepressants (TCAs) [14, 15]. Recent research on biological differences between atypical and melancholic depression suggests that hypothalamic-pituitary-adrenal (HPA) axis or autonomic nervous system differential reactivity may be at play [16]. It should be noted that a significant (estimated between 56–75%) proportion of MDD patients endorse an anxious distress specifier, reflecting the high prevalence of anxious comorbidity in depressed patients [1]. Comorbid substance use is another significant factor given its biological impact on key reward circuitry in the brain. In this context, treatment approaches for comorbid alcohol use disorder that target both depression and addiction with anti-craving compounds, such as combination sertraline and naltrexone, have been found to be superior to monotherapies [17, 18]. Taken together, the evidence reviewed above reinforces the notion that, given the heterogeneity of TRD, clinicians must consider the complexity of potential comorbidities as well as environmental factors when selecting an antidepressant. Although a comprehensive discussion on the nature of these interactions and a review of treatment strategies for comorbidities is beyond the scope of this discussion, we refer the interested reader to several recent guidelines on this topic [8, 19].

2.2. Clinical considerations: switching medications

Although it is common clinical practice to switch from one antidepressant to another after a failed trial, results are often underwhelming. Thus, a comprehensive assessment of a patient’s treatment history is crucial for ensuring that treatment trials were of sufficient dose and duration. In addition, an inquiry into the nature of treatment discontinuation may provide guidance for further management. A shared-decision making approach is particularly important, as proactive discussions of treatment goals, options, and anticipated medication side effects can improve patient compliance and outcomes over the course of treatment trials.

In the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) study, 727 patients with MDD who failed to respond to an initial trial of citalopram were randomized into treatment groups with bupropion, sertraline, or venlafaxine for 14 weeks [20]. As assessed by the Quick Inventory of Depressive Symptomatology (16-item Self-Report) (QIDS-SR-16), remission rates were approximately 25% for each group, suggesting that any of these medications can be considered after an unsuccessful trial of a selective serotonin reuptake inhibitor (SSRI). A further analysis of 235 patients who had failed to respond to two consecutive treatment trials found that remission rates were 8% in those subsequently treated with mirtazapine and 12.4% in those subsequently treated with nortriptyline [21].

This variability in antidepressant treatment response has been attributed to the biological heterogeneity of depression. Consequently, both mechanism of action and pharmacokinetics are relevant in determining course of treatment when a patient fails to respond to an initial antidepressant trial. Although novel genomic testing tools are available, differences in antidepressant response rates can be approximated by considering “class” switches to account for this heterogeneity, i.e. switching patients from an SSRI to a TCA [22] or from a TCA to an MAOI [23]. In particular, poor or ultra-rapid CYP2D6 or CYP2C19 metabolizers should consider avoiding TCAs [24], rapid CYP2C19 metabolizers are advised to avoid citalopram, escitalopram, and sertraline, and rapid CYP2D6 metabolizers should also consider avoiding paroxetine and fluvoxamine [25]. In addition, pooled data from four clinical trials comprising 1496 participants with SSRI-resistant depression found superior remission rates with a “class” switch to a non-SSRI antidepressant (bupropion, mirtazapine, or venlafaxine) [26]. Similar remission rates were reported in the VAST-D study—a large, multi-site, randomized, single-blind study of 1,522 patients who did not respond to at least one antidepressant treatment trial—when participants switched to bupropion, a norepinephrine-dopamine reuptake inhibitor [27].

Finally, the National Institute for Health and Care Excellence (NICE) has recommended vortioxetine as a treatment option for TRD; this newer antidepressant is classified as a serotonergic modulator and has a mechanism of action similar to vilazodone [28]. Due to its unique mechanism of action, vortioxetine has generated considerable interest as an additional treatment option [29]. However, evidence for its effectiveness compared to other antidepressants is presently limited, and it remains unclear whether there is any direct advantage to this agent [30].

2.3. Clinical considerations: combining antidepressants

Another common strategy for treating TRD is the addition of a second antidepressant. Mirtazapine and bupropion are often considered for this purpose because of their unique mechanisms of action; for instance, mirtazapine—which is a noradrenergic and specific serotonergic antidepressant—is often used to promote appetite and induce sleep, although known side effects include weight gain and hypersomnolence. Bupropion, which is a norepinephrine-dopamine reuptake inhibitor, has no effect on serotonin, histamine, acetylcholine, or adrenaline receptors and does not cause sedation, hypotension, sexual dysfunction, weight gain, or anticholinergic side effects [31].

It should be noted that although it is common to add a second antidepressant to a patient’s treatment regimen, clinical evidence of its benefit over SSRI monotherapy remains mixed [6]. One double-blind study found that adding mirtazapine to an antidepressant significantly lowered Hamilton Depression Rating Scale (HAM-D) scores [32]. However, a more recent and larger study found no significant differences in remission rates when adding mirtazapine to ongoing treatment with either an SSRI or SNRI [33]. With regard to bupropion, results from the VAST-D trial found remission rates of 27% for participants whose treatment regimens were augmented with bupropion [27], and a systematic meta-analysis of bupropion’s effectiveness across 51 studies found that augmentation therapy with bupropion benefited patients with TRD [11].

2.4. Clinical considerations: augmentation strategies

2.4.1. Lithium

Lithium, a mood stabilizer used to treat bipolar disorder, has a unique mechanism of action that remains to be fully elucidated. Evidence suggests that it is an effective adjunct treatment for TRD [34]. The STAR*D study showed only modest remission rates for TRD patients who received augmented treatment with lithium (15.9%) [35]. However, a meta-analysis of nine randomized controlled trials found that lithium augmentation was more effective than antidepressant treatment alone, and that there were no significant differences in adverse events between the two groups [36]. Another meta-analysis comparing 11 augmentation agents across 48 randomized controlled trials found significant effects for lithium augmentation [37]. Notably, one of the key advantages of lithium augmentation is that it lowers suicide risk; one study found an 88.5% decreased risk for suicide or suicide attempts in TRD patients who received concomitant lithium treatment [38].

2.4.2. Atypical antipsychotics

Extensive evidence supports the use of atypical antipsychotics as adjunctive therapy for MDD; aripiprazole, quetiapine, and brexpiprazole have all received FDA approval for this indication [39], and the combination of olanzapine and fluoxetine (marketed as Symbiax by Eli Lilly) has also been FDA-approved for the treatment of TRD. Although risperidone has not been approved by the FDA for the treatment of TRD, it has also been used clinically as an augmentation strategy, with some studies demonstrating efficacy [40, 41]. In one meta-analysis that compared 11 augmentation agents across 48 randomized controlled trials, the most robust evidence emerged for quetiapine and aripiprazole as augmentation strategies [37]. Furthermore, results from the VAST-D trial found remission rates of 29% for participants whose treatment regimens were augmented with aripiprazole [27]. It should be noted that although ziprasidone has not received FDA approval for this indication, it has also been considered as an augmentation strategy for TRD in light of its ability to inhibit monoamine reuptake and its relatively favorable side effect profile [42]. One open-label randomized controlled trial of ziprasidone used adjunctively with escitalopram found that this agent significantly improved outcomes in TRD, but data are both limited and mixed; as a result, ziprasidone is not presently recommended as an augmentation strategy [43].

In depressive episodes marked by psychotic features, augmentation with atypical antipsychotic medications may be considered earlier in the disease course. However, clinical diagnosis may need to be further scrutinized and treatment side effects such as metabolic effects, weight gain, sedation, and akathisia should be closely monitored by clinicians to balance the risk and benefits to patients.

2.4.3. Triiodothyronine (T3)

The use of triiodothyronine (T3) as a potential augmentation strategy stems from previous studies indicating an association between thyroid hormone levels and severity of depressive symptoms as well as from studies that used T3 and levothyroxine as adjunctive treatments for depression [44]. Because of individual differences in thyroid hormone metabolism, T3 is the preferred treatment due to its faster response time [45]; however, there is concern regarding its long-term use, although it is well tolerated at low doses. A meta-analysis comparing 11 augmentation agents across 48 randomized controlled trials found significant effects for T3 augmentation [37]. Another meta-analysis of eight studies found that patients who received T3 augmentation were twice as likely to exhibit an antidepressant response than those who received standard treatment [46]. However, the STAR*D study found a remission rate of 24.7% for TRD patients who received augmented treatment with T3, which was not significantly better than other options [35]. Interestingly, results from another meta-analysis suggested that T3 augmentation might be more successful with TCAs than with SSRIs [47]. In addition, some of the reviewed studies found lower baseline thyroid function in the group that responded to T3, suggesting that antidepressant response to this agent may be due to sub-optimal thyroid function, but these findings were not consistent [47].

2.4.4. Psychostimulants

Psychostimulants have been suggested as an augmentation strategy in TRD, but sufficient evidence is presently lacking to justify their use in light of the potential risks of addiction and cardiac side effects. Nevertheless, stimulants are frequently prescribed off-label to target specific depressive characteristics such as poor energy and concentration [48]. In this context, modafinil and methylphenidate are often selected as initial stimulant augmentations given their relatively superior safety profiles. A meta-analysis of six randomized controlled trials found that depressive symptoms were significantly improved with modafinil augmentation [49]. Given that arousal has been identified as a dimension of the Research Domain Criteria (RDoC) [50], electroencephalogram (EEG)-guided biomarkers could be used to identify subgroups of depressed patients most likely to benefit from augmentation with a stimulant. Further studies of psychostimulants as an augmentation strategy are warranted, particularly using a targeted approach.

2.4.5. Buprenorphine

There has been considerable interest in the opioid system as a target for novel therapeutic agents given its involvement in mood regulation and incentive salience [51]. Furthermore, numerous studies have demonstrated that the opioid partial agonist buprenorphine has antidepressant effects [52]. However, a recent meta-analysis of randomized, placebo-controlled trials evaluating its use as an augmentation strategy in TRD found that buprenorphine did not reduce the severity of depressive symptoms compared to placebo [53]. Building on this work, two multi-center, placebo-controlled, Phase 3 trials investigating combination treatment with buprenorphine / samidorphan (BUP/SAM), an investigational opioid modulator known as ALK 5461, found that adjunctive use of this agent was superior to placebo in reducing Montgomery-Åsberg Depression Rating Scale (MADRS) scores in TRD patients at a dose combination of 2mg/2mg in the second study (FORWARD-5 (n=407)) as well as in a pooled analysis of both studies [54]. Although ALK 5461 was well-tolerated, the FDA has not approved the drug for MDD, and more clinical data are needed to clarify its therapeutic efficacy.

2.4.6. SAMe (S-Adenosyl Methionine)

S-Adenosyl Methionine (SAMe) is an amino acid metabolite with enzymatic function in the synthesis of monoamine neurotransmitters. Studies have found decreased SAMe concentrations in neurological and psychiatric disorders such as MDD, leading to many clinical trials evaluating its potential as a treatment or supplement for depression [55]. Although some evidence suggests that adjunctive use of SAMe improves depressive symptoms, these improvements were not significant compared to conventional treatments [56].

2.4.7. Levo-methylfolate

Another agent of interest as an augmentation strategy is levo-methylfolate, given its role in monoamine synthesis and findings of lower levels of folate derivatives in patients with inadequate antidepressant response [57]. However, studies of its efficacy have found mixed results, and a recent meta-analysis based its conclusion that levo-methylfolate could be used adjunctively with SSRI therapy on limited evidence [58]. The heterogeneity in response to levo-methylfolate may be due to underlying biology; some have suggested that levo-methylfolate treatment might best be targeted based on serum folate levels, increased body-mass index (BMI), and methylenetetrahydrofolate reductase (MTHFR) polymorphisms [59, 60].

2.5. Novel rapid-acting antidepressants: glutamate and GABA-based agents

Evidence of glutamatergic system impairments in MDD has emerged from clinical studies demonstrating that sub-anesthetic (0.5 mg/kg) IV administration of the glutamatergic modulator ketamine is associated with rapid antidepressant response [61–63]. In contrast to conventional monoaminergic antidepressants that are associated with a 30–35% response rate and a lag time of weeks to months until onset of action, ketamine has a 65–70% response rate, with antidepressant effects observed within 24 hours following single and repeated administrations [64]. Ketamine has also been found to reduce suicidal thoughts within one day of administration, with anti-suicide ideation effects lasting up to one week in MDD patients [65].

Despite its rapid and robust antidepressant effects, clinical concerns remain associated with ketamine use, including tolerability of its dissociative and vasomotor effects. Safety concerns associated with its long-term use also remain. The Canadian Network for Mood and Anxiety Treatments (CANMAT) Task Force recently provided guidance on ketamine use in clinical practice, recommending it as a tertiary treatment despite noting that the antidepressant effects associated with single-dose IV racemic ketamine infusion for TRD were strong [66]. The CANMAT guidelines also noted the need for case-based assessments for maintenance ketamine infusions that take potential risks and benefits into account.

While not fully elucidated, ketamine’s mechanism of action is hypothesized to involve downstream actions on the monoaminergic system, the opioidergic system, the glutamatergic system, and the gamma aminobutyric acid (GABA) system as well as signal transduction cascades such as mechanistic target of rapamycin (mTOR), cellular proliferation, and neuroplasticity cascades [67]. Neurochemical and functional imaging studies have also corroborated the glutamatergic and GABA-ergic dysfunction previously identified in studies of individuals with depression and, furthermore, demonstrated that ketamine treatment partly reverses these neurochemical disturbances [68]. These findings have advanced our conceptualization of the pathophysiology of depression and paved the way for developing other novel antidepressants.

In particular, these discoveries have led to increased interest in both the glutamatergic and GABA pathways as potential targets for the development of novel antidepressants [69]. This is underscored by the FDA and European Medicines Agency (EMA) approval of esketamine, the S[+] enantiomer of ketamine formulated for intranasal use. After positive Phase 3 trials, esketamine was approved for adults with TRD (NCT01998958; NCT02133001) and for adjunctive use in adults with major depression with acute suicidal ideation or behavior (NCT03039192; NCT03097133). Based on favorable Phase 3 study results (NCT02942004; NCT02942017), in March 2019 the FDA also approved the neuroactive steroid brexanolone, a positive allosteric modulator of GABA_A receptors, as the first drug to target postpartum depression. Based on encouraging results from Phase 2 studies (NCT03000530), in March 2020 the FDA then granted breakthrough status to SAGE-217, a next-generation allopregnalone analogue formulated for once-daily oral dosing. Despite recent setbacks in distinguishing the efficacy of SAGE-217 from placebo in a Phase 3 trial for MDD (NCT03672175), additional studies are currently underway to explore the efficacy of this agent for both MDD (NCT04442490) and postpartum depression (NCT04442503). The ultimate usefulness of brexanolone and SAGE-217 in individuals with TRD is presently unknown but warrants investigation.

2.5.1. Insights from biomarkers of response to ketamine

The search for biomarkers to identify mood disorder sub-populations most likely to benefit from ketamine treatment remains a priority. Previous studies have identified potential clinical pretreatment predictors of response to ketamine in TRD. In particular, clinical data including BMI [70], history of suicide [70], and family history of alcohol use disorder have been implicated as prognostic factors [71]. Other potential factors include adiponectin levels [72], central glutamine to glutamate ratios [73], abnormalities in delta sleep ratio in polysomnography [74], neuroimaging markers of the anterior cingulate cortex activity [75], genetic variations such as the Val66Met BDNF allele [76], and processing speed in cognitive function testing [77]. However, it should be noted that although several of these biomarkers provide insights into ketamine’s mechanism of action, none are yet ready for clinical use.

Non-invasive neuroimaging techniques have also revealed several potential biomarkers of ketamine’s antidepressant efficacy, although most of these have focused on acute rather than longer-term antidepressant effects and, again, none are presently ready for clinical use [78]. One study found that increased pretreatment rostral anterior cingulate cortex (ACC) reactivity to fearful faces predicted heightened antidepressant response to ketamine [75]. Magnetoencephalography (MEG) response during spatial working memory tasks has also been correlated with antidepressant response to ketamine [75]. In addition, greater intra-infusion dissociation, as measured by the Clinician Administered Dissociative States Scale (CADSS), was found to be correlated with increased antidepressant effects at 230 minutes and one week post-infusion [79]; however, the evidence to date for this link is mixed, and the literature does not presently support the conclusion that dissociation is necessary for antidepressant response to ketamine [80]. Studies to understand the significance of ketamine’s dissociative side effects at the molecular, biomarker, and psychological levels remains ongoing.

Abnormalities in resting-state connectivity patterns have previously been described in functional neuroimaging studies of patients with MDD, particularly in brain networks involved with cognitive control of attention and emotion and the Default Mode Network (DMN), which is deactivated during cognitive tasks [81]. The DMN includes regions such as the medial prefrontal cortex and posterior cingulate cortex and has been associated with introspection [82]. Previous studies also described hyperconnectivity between the subgenual ACC and the MPC in MDD [83]. Increased pre-treatment neural activity in the rostral anterior cingulate cortex (rACC) has also been associated with greater antidepressant response across different interventions [84]. In a randomized, placebo-controlled, double-blind, crossover ketamine challenge in healthy volunteers, ketamine was associated with decreased functional connectivity between the rACC and medial prefrontal cortex [85], suggesting that antidepressant response may be associated with reduced hyperconnectivity between these two regions in patients with MDD. Ketamine administration at therapeutic and subanesthetic levels has also been shown to substantially increase gamma oscillations and gamma power measures [86]. In this context, delayed gamma power measurements may be potential biomarkers of both ketamine response and synaptic potentiation that may help differentiate between responders and non-responders. The findings suggest that there may be an optimal level of gamma power that is essential for mediating ketamine’s antidepressant effects, although the underlying mechanism linking acute gamma oscillations and antidepressant response remains unknown [86].

Finally, it is important to note that, despite these intriguing preliminary findings, these biomarkers are not presently ready for clinical use. Further research into prognostic biomarkers for ketamine response will ultimately further our understanding of its effects and help develop the next generation of antidepressants. Although a more in-depth discussion of experimental therapeutics and their treatment targets is beyond the scope of this review, we refer the interested reader to several previous articles on this topic (see [87–90]).

3. Inflammation as a potential target in depression

Peripheral inflammation can cause inflammatory cells to cross the blood-brain barrier (BBB), causing brain inflammation and dysfunction. An ever-growing literature has linked some degree of neuroinflammatory pathogenesis to most psychiatric disorders [91]; this has, in turn, led to increased interest in plasma inflammatory biomarkers in depression [92]. In a recent analysis of mRNA expression of inflammatory markers, researchers identified increased inflammasome activation and glucocorticoid resistance in participants with TRD and drug-free depression [93]. Specifically, six mRNAs (P2RX7, IL-1-beta, IL-6, TNF-alpha, CXCL12, and GR) were able to distinguish TRD patients from treatment-responsive ones. In addition, levels of pro-inflammatory cytokines such as interleukin-6 (IL-6) and tumor necrosis factor alpha (TNF-α) have been reported to be higher in individuals with MDD than controls [94]. Interestingly, the SSRI fluoxetine was found to reduce both central and peripheral levels of the pro-inflammatory cytokine interleukin-1 beta (IL-β) in animal models [95].

Ketamine’s anti-inflammatory properties suggest that inflammatory markers may be a potential predictor of antidepressant response worth examining [72]. Previous studies noted the association between increased BMI and chronic inflammation [96] and, indeed, one recent review found that high BMI was one of the most consistent predictors of antidepressant response to ketamine [97]. Furthermore, research into specific adipokines (which are known to exert anti-inflammatory and insulin-sensitizing effects) found that low pretreatment plasma adiponectin levels correlated with rapid antidepressant response to ketamine in individuals with TRD [72]. Omega-3 polyunsaturated fatty acids and their metabolites have also been studied for their anti-inflammatory properties, but studies exploring the antidepressant effects associated with their adjunctive use have returned mixed results [98]; nevertheless, the findings are promising enough to warrant a more tailored investigative approach.

Another area of interest is traumatic childhood experiences. Studies have suggested that the association between the severity of such experiences and poor antidepressant treatment response may be due to the effects of trauma on the HPA axis, as evidenced by increased cortisol levels and glucocorticoid resistance [99]. This suggests that this subgroup with glucocorticoid resistance could be targeted for specified treatment. In this context, the second-generation tetracycline antibiotic minocycline is being investigated as a potential treatment for CNS inflammatory disorders; this agent, which has known immunomodulatory effects and has been shown to normalize glucocorticoid levels through its actions on the HPA [100], is able to cross the BBB and suppress inflammatory pathways directly at the microglia [101]. In addition to normalizing glucocorticoid levels and acting on the HPA axis, minocycline has also been implicated in the kynurenine pathway [102]. A large randomized controlled trial is in progress to validate the antidepressant efficacy of adjunctive minocycline treatment [103].

Finally, cyclooxygenase-2 (COX-2) inhibitors have been explored as adjunctive therapies given their role in blocking prostaglandin production [104]. A large, recent, double-blind, placebo-controlled, randomized clinical trial found that, compared to placebo, low-dose aspirin did not affect depression outcomes in a geriatric population [105]. TNF antagonism with the chimeric monoclonal antibody infliximab was found to be efficacious when used specifically in TRD patients with elevated levels of C-reactive protein (CRP) [106]. Furthermore, a recent study found that, in unmedicated individuals with MDD, elevated levels of 18kDa translocator protein (TSPO), a biomarker of neuroinflammation, predicted response to celecoxib, a non-steroidal anti-inflammatory drug (NSAID) and selective COX-2 inhibitor [107]. However, given the preliminary nature of much of this evidence, the role of inflammatory biomarkers in selecting anti-inflammatory treatments warrants further investigation.

4. Non-Pharmacological Treatments

Cognitive behavioral therapy (CBT) and other psychotherapies, alone or in combination with medication, have demonstrated efficacy in decreasing depressive symptoms for patients with TRD [108, 109]. Neurostimulation and neuromodulation strategies such as electroconvulsive therapy (ECT), repetitive transcranial magnetic stimulation (rTMS), deep brain stimulation (DBS), and vagus nerve stimulation (VNS) are also available for individuals with TRD. Although a comprehensive discussion of non-pharmacological interventions for TRD is beyond the scope of this article, we refer the interested reader to several recent reviews of neuromodulation strategies [110, 111].

5. Limitations and treatment guidelines for TRD

This review of psychopharmacological approaches to TRD is not a structured guideline of recommendations. Rather, we have broadly discussed some of the treatment options available to clinicians as well as the supporting evidence for these strategies and potential future developments in the field.

As noted above, one key consideration in the treatment of TRD is the use of augmentation strategies. While a comprehensive and systemic discussion of augmentation strategies is beyond the scope of this review, a number of guidelines exist that describe augmentation strategies in TRD and advise clinicians regarding treatment selection, dose consideration, monitoring, and discontinuation. Interestingly, a recent review assessed 10 guidelines and compared these publications on the basis of quality domains outlined in the AGREE II recommendations [19, 112]. Among the augmentation strategies reviewed, lithium and atypical antipsychotics were the most consistently recommended by all 10 guidelines, while none of the other 15 augmentation strategies investigated were universally recommended. Although there is wide variability in the content and quality of these guidelines—factors that may be limited by their publication date or limited clinical trial evidence—there is also a subjective quality inherent in any balancing of factors such as efficacy and tolerability. The review calls for greater standardization between guidelines and recommends evidence grading systems such as those in the World Federation of Societies of Biological Psychiatry (WFSBP) guidelines on how to grade treatment evidence for guideline development [113].

Another approach is sequential treatment optimization for MDD. This guidance is based on disease trajectories with four stages of treatment: Stage 1: initial monotherapy and dose escalation; Stage 2: augmentation and switching strategies; Stage 3: third-line interventions; and Stage 4: experimental interventions [114]. Treatment strategies for TRD would fall under Stage 2. This disease staging reflects the characterization of TRD based on disease course, but in cases where acuity and severity supersede chronicity, rapid-acting interventions such as ketamine or ECT should be considered. Future treatment guidelines would ultimately benefit from developing clinically applicable tools to stratify subgroups for outcome prediction.

6. Conclusion

The heterogeneous nature of depression complicates treatment options. Indeed, TRD is defined by a disease course marked by failure to respond to multiple treatment trials, reflecting the complexity of this multifactorial disorder. Clinicians should seek to employ informed management strategies such as medication class switches and augmentation treatments to account for clinical characteristics and treatment response. Understanding the biological nature of response to treatments such as ketamine may better inform the pathophysiology of TRD and further help clinicians make informed decisions about treatment options. Towards this end, the development of robust biomarkers for clinical practice has the potential to supplant the trial-and-error disease trajectory of TRD, with biologically targeted treatments leading to early intervention and ultimately improved remission outcomes.

7. Expert Opinion

7.1. Treatment strategies

The treatment interventions outlined for TRD in this review include antidepressant class switching and augmentation strategies. Antidepressant class switches and combinations should be informed by the clinical dimensions of any individual patient’s presentation. Although pharmacogenetic tools are emerging to help guide clinicians, they are mostly limited to pharmacokinetic information and have limited utility. Augmentation strategies with the most evidence-based support include atypical antipsychotics, lithium, T3 and, more recently, ketamine and esketamine. Atypical antipsychotics can be considered for initial augmentation but cautiously selected considering their side effect profile. Lithium seems to have similar efficacy for augmentation to atypical antipsychotics, but evidence is mostly limited to comparison studies with TCAs. While lithium is effective for suicide prevention, for acutely suicidal patients, treatment with ECT, ketamine, and esketamine is advised given their efficacy and rapid onset of action.

A significant weakness in the available data is the lack of quality head-to-head studies to prioritize treatment selection. Comparisons of meta-analytic data have been attempted and some have highlighted the significant effect size for esketamine compared to atypical antipsychotics [115], but these analyses are in part limited by variability in how TRD is defined. While there is no universally accepted definition for TRD, it is commonly recognized as a failure to respond to two or more antidepressants in succession. Nevertheless, differences persist regarding how the criteria underlying TRD are defined in both clinical practice and research, reflecting controversies not only about its categorization but also the inherent challenges associated with defining a heterogeneous disorder. For example, some authors have suggested that part of the challenge associated with defining TRD may be due to a significant subset of TRD patients having a bipolar diathesis [116]. While different groups have sought to stage and better define TRD using a variety of clinical factors, the clinical utility of such attempts remains to be determined. Future studies may help establish a hierarchical recommendation of interventions; for now, however, a shared decision-making approach is preferred for determining which individual factors should be prioritized in treatment.

7.2. Future directions

7.2.1. Reconceptualizing the diagnosis of heterogeneous TRD

Advances in the treatment of TRD patients are limited by our understanding of the neurobiological underpinnings of any individual patient’s course of illness. Clinical risk factors for TRD can help predict disease course; these include early onset of first episode, severe and frequent recurrence of depressive episodes, comorbidity with anxiety disorder, current suicidal risk, and presence of melancholic features. In addition, factors that can help guide treatment course are particularly useful. Specifically, clinicians should reconsider the diagnosis after failure to respond to antidepressant treatment, as patients with a bipolar diathesis may benefit from mood stabilizers or medications approved for bipolar depression.

Determining why some patients benefit from certain antidepressants while others do not is critical for advancing our biological understanding of depression treatments and more accurately identifying targeted treatments. Biomarkers have specific diagnostic and prognostic value and are crucial for advancing personalized treatment for patients. The development of the RDoC by the NIMH highlights efforts to identify therapeutically relevant biomarkers and neural circuits associated with psychiatric disorders [50]. The RDoC model’s approach aims to improve risk prediction, early detection, and individualized response and ultimately steer clinical practice towards precision psychiatry.

7.2.2. Biomarkers and precision medicine

While no biomarkers are currently in routine clinical use, there is continued interest in developing biomarkers that can ultimately help clinicians decide the most appropriate next treatment steps for individual patients [117]. In this context, the heterogeneity of depression is underscored by studies such as a genome-wide association study (GWAS) of MDD patients that found that up to 42% of the variance in antidepressant response was associated with common genetic variations [118]. Pharmacogenetic-guided treatments have been reported to improve response rates, and such improvements in antidepressant response as a result of genetic testing were more consistent in the TRD group [119]. However, at present, most genetic panels are limited to pharmacokinetic insights that inform side effects rather than provide treatment response prediction. Ultimately, reconceptualizing depression as heterogeneous with regard to phenotype and etiology may improve our approach to discovering biomarkers [120].

Recent scientific advances that explore precise brain microcircuits have prompted a multi-dimensional systems biology approach that incorporates diverse datasets in an effort to elucidate pathophysiology and circuit dynamics; such efforts to parse depression into biologically-defined subtypes are critical to developing novel antidepressant drugs [121]. Converging findings from functional neuroimaging in patients and optogenetic animal model behavioral studies are already helping the field identify key neuroanatomical substrates and circuit-based understanding of the neurophysiology of specific depressive behaviors and symptoms [122].

In addition, the combination of advances in data science and artificial intelligence, powered by large consortiums such as EMBARC [123] and ENIGMA [124] that combine neuroimaging with genomic data has been encouraging. The utility of applying such methods is exemplified by the development of an EMBARC-trained machine learning algorithm capable of predicting robust antidepressant response to sertraline [125]. Another research group recently leveraged data from ENIGMA that integrated clinical imaging markers with genetic and postmortem patient transcriptional data to identify somatostatin interneurons and astrocytes as replicable cell-level correlates of depression and negative affect [126].

Finally, the development of biomarkers also has the potential to transform how clinical trials designed for psychiatric illnesses are conducted, similar to advances in cancer research. For instance, the National Cancer Institute’s MATCH trial (NCT02465060) recruits patients with different tumor types but performs extensive genotyping and molecular stratification to place participants into targeted treatment arms, a strategy known as an “umbrella” design trial. In another type of study—often referred to as a “basket” trial—patients are recruited to a targeted treatment. Such recruitment is based on the molecular signature for a particular mutation that can manifest as different cancers depending on the organ; the use of imatinib to treat BCR-Abl transposition (NCT00154388) is one such example. In psychiatry, the effectiveness of ketamine for treating both TRD and bipolar depression suggests that there is a unique subtype of depression that may be mediated through the glutamatergic pathway rather than the monoaminergic one. In this context, if a robust biomarker for glutamatergic dysfunction is identified in a basket trial, it could power the study of future glutamatergic drug targets. Thus, both basket and umbrella clinical trial designs have the potential to leverage biomarkers to better study targeted treatments across complex syndromes in psychiatry. These are encouraging advances in the search to identify biomarker-guided phenotypes for the targeted treatment of TRD, with the ultimate goal of improving individual outcomes.

Article Highlights.

• Treatment options for treatment-resistant depression (TRD) include switching antidepressant classes and augmentation strategies.

• Shared decision-making approaches are recommended to tailor treatment choices in light of particular clinical dimensions and to improve compliance.

• Substantial evidence suggests that augmentation therapy with FDA-approved atypical antipsychotics may be efficacious, but caution is warranted given the potential for adverse effects.

• Augmentation with lithium, particularly for tricyclic antidepressants (TCAs), has been shown to be efficacious and helps reduce suicide risk.

• Ketamine and esketamine are rapid-acting antidepressants with significant efficacy in the treatment of TRD. Ketamine and electroconvulsive therapy (ECT) are the preferred treatments for acutely suicidal patients.

• The heterogeneous nature of TRD remains a significant barrier to treatment. Although not ready for clinical integration, the development of biomarkers to ultimately predict treatment response holds great promise for improving outcomes.