Introduction
The World Health Organization describes depression as a leading cause of disability, with an estimated 280 million individuals affected worldwide. 1 Although effective treatments for depression exist, more than 25% of patients with major depressive disorder (MDD) do not respond to 2 or more treatments.Reference Zhdanava, Pilon and Ghelerter 2 Further, the onset of benefits of antidepressants can be slow,Reference Hillhouse and Porter 3 and guidelines now suggest that it takes up to 12 weeks of treatment to ensure an optimal treatment response.Reference Rush 4 , Reference Lam, McIntosh and Wang 5
MDD treatments have evolved significantly over the past 60 years, 6 with therapies becoming increasingly targeted or selective. Before the mid-1950s, the only effective medical treatment for severe depressive episodes was electroconvulsive therapy.Reference Hyun-Hee Kim 7 Iproniazid, a medication used to treat tuberculosis, was the first drug identified with antidepressive effects; within a few years, its mechanism of action was linked to the inhibition of monoamine oxidase.Reference Hillhouse and Porter 3 Monoamine oxidase inhibitors (MAOIs) do not act on specific receptors but increase the levels of serotonin, norepinephrine, and dopamine in the brain by preventing their enzymatic oxidation.Reference Hillhouse and Porter 3 , Reference Pereira and Hiroaki-Sato 8 Shortly thereafter, the therapeutic effects of imipramine, the first drug to be classified as a tricyclic antidepressant (TCA), were identified in the course of research to develop safer and more effective antipsychotic drugs than chlorpromazine.Reference Hillhouse and Porter 3 TCAs inhibit presynaptic norepinephrine and, to a lesser extent, serotonin reuptake transporters.Reference Hillhouse and Porter 3
Following the goal to develop interventions with fewer side effects than these serendipitously discovered medications, selective serotonin reuptake inhibitors (SSRIs) became the most commercially successful class of antidepressant drugs following the introduction of fluoxetine in 1987.Reference Pereira and Hiroaki-Sato 8 , Reference López-Muñoz and Alamo 9 Though reduced, side effects associated with SSRIs were still problematic for some patients, and others did not achieve meaningful symptomatic improvement, leading to the development of antidepressants such as bupropion, venlafaxine, reboxetine, and mirtazapine.Reference Pereira and Hiroaki-Sato 8 , Reference Fava, Rush and Thase 10 Although these medications offered additional options for patients, none were able to supplant SSRIs as the standard first choice for first-line therapy. Moreover, as was the case with the TCAs and MAOIs, there was often substantial latency between beginning therapy and the onset of meaningful clinical benefits.Reference Hillhouse and Porter 3 , Reference Pereira and Hiroaki-Sato 8 Moreover, despite having certain advantages in tolerability and safety indices, the so-called second generation of antidepressants was not more effective than the TCAs or MAOIs.Reference Pereira and Hiroaki-Sato 8 Some suggested that the unmet needs in the pharmacotherapy of depression, such as the long latency to response and an apparent plateau in effectiveness across classes of antidepressants, were attributable to the fact that all these medications targeted monoaminergic mechanisms.Reference Pereira and Hiroaki-Sato 8 It was further posited that novel targets for pharmacotherapy would need to be identified in order for the next generation of antidepressants to emerge.Reference Pereira and Hiroaki-Sato 8
By the late 1990s, interest in the glutamatergic system and its importance in the neurobiology of depression had grown. It was recognized that a single intravenous dose of the anesthetic drug ketamine, which blocks the effects of glutamate on the N-methyl-D-aspartate receptor, could have rapid and large antidepressant effects.Reference Pereira and Hiroaki-Sato 8 , Reference Berman, Cappiello and Anand 11 The discovery that the antidepressant effects of ketamine last for a number of days had a transformative effect on depression treatment research, including the commercial development of one of its stereoisomers, esketamine, for intranasal administration.Reference Pereira and Hiroaki-Sato 8 , Reference Berman, Cappiello and Anand 11 , Reference Ionescu, Fu and Qiu 12 However, ketamine and esketamine are classified as controlled substances that can cause dissociation and cardiovascular side effects that warrant up to 2 hours’ of monitoring, which limits their potential for widescale clinical use.Reference Pereira and Hiroaki-Sato 8 , Reference Sanacora and Schatzberg 13 , Reference Alan and Schatzberg 14 Nevertheless, the recognition of one novel target for pharmacotherapy that yielded a potentially large and rapid effect for depressed patients helped to restore therapeutic optimism that potentially better options for our patients were on the horizon.Reference Pereira and Hiroaki-Sato 8 , Reference Berman, Cappiello and Anand 11 , Reference Ionescu, Fu and Qiu 12
In the mid-1990s, as it became apparent that the drugs available were not “one size fits all,” a strategy for managing treatment-resistant MDD using an algorithmic approach began to emerge.Reference Thase and Rush 15 This approach, coupled with a systematized monitoring of symptoms and side effects known as measurement-based care, served as the platform for a large-scale study: Sequenced Treatment Alternatives to Relieve Depression (STAR*D).Reference Rush, Trivedi and Wisniewski 16 , Reference Fava, Rush and Trivedi 17 STAR*D comprised a 4-level treatment algorithm in which a patient with depression moved from 1 treatment level to the next, starting with citalopram at level 1 and escalating through levels 2 to 4, which included various switching and combination categories if full remission was not achieved.Reference Fava, Rush and Trivedi 17
In clinical practice, combination treatments are used by many patients to combat treatment resistance and comorbidity.Reference Cuijpers, de Wit, Weitz, Andersson and Huibers 18 Results from a meta-analysis indicate that combined treatment results in small-to-moderate improvements in depression compared with psychotherapy or pharmacotherapy alone or with psychotherapy plus a placebo pill.Reference Cuijpers, de Wit, Weitz, Andersson and Huibers 18 However, many treatments with different mechanisms of action have been found to have significant adjunctive antidepressant effects.Reference Cuijpers, de Wit, Weitz, Andersson and Huibers 18 , Reference Cuijpers, Noma, Karyotaki, Vinkers, Cipriani and Furukawa 19 There is little guidance available on the use of one adjunctive therapy over another; new studies are needed to operationalize our understanding of the combination effect.Reference Cuijpers, Noma, Karyotaki, Vinkers, Cipriani and Furukawa 19 , Reference Guidi and Fava 20
Although continuation and maintenance treatment is generally recommended after a successful response to acute treatment, it is unclear how long maintenance therapy should continue to prevent subsequent recurrent depressive episodes. A measurement-based care approach could enable the monitoring of potential relapse-preventative or disease-modifying effects that have eluded the current treatment armamentarium.Reference Rush, Trivedi and Wisniewski 16 , Reference Scott and Lewis 21
The development of pharmacotherapy for MDD has evolved from chance findings to a more targeted neurobiological approach.Reference Pereira and Hiroaki-Sato 8 A range of specific and targeted therapies are now available; however, there is no objective guidance on how to choose from the many available medications.Reference Zanardi, Prestifilippo, Fabbri, Colombo, Maron and Serretti 22 Moreover, despite our understanding of the pathophysiology of MDD evolving from single brain region or monoamine deficits to more network-based models with corresponding subtyping,Reference Li, Friston, Mody, Wang, Lu and Hu 23 treatments are generally not targeted to individual phenotypes.
The purpose of this review is to examine why the current approach to MDD often results in treatment failure, the impact of placebo response in clinical trials for MDD, and why more targeted pharmacotherapy for MDD, such as through precision psychiatry, may be beneficial for short-term optimization toward an early treatment response, and in the long-term to reduce the number of trials and ineffective courses of therapy to achieve remission. Further, we will discuss the role of precision psychiatry and how it can be used to inform phenotypes for more targeted treatment and provide suggested approaches for future clinical studies. See Table 1 for a summary of the key points discussed.
Table 1. Key points
Reasons for treatment failure in MDD
As highlighted above, over one-quarter of patients do not respond to 2 or more treatments and are categorized as having treatment-resistant depression.Reference Zhdanava, Pilon and Ghelerter 2 Some patients continue to be significantly burdened by depression despite usual treatment efforts and are classified as having difficult-to-treat depression.Reference McAllister-Williams, Arango and Blier 24 It is important to note that while the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) provides specific criteria for the diagnosis of MDD, 25 no such criteria exist for difficult-to-treat depression. While several factors, such as symptom onset time and severity, early treatment response, psychiatric comorbidities, frontal electroencephalography theta activity, neuroimaging, and peripheral markers, have been identified as predictors of antidepressant response,Reference Jambor, Juhasz and Eszlari 26 the rate of nonresponse to antidepressants is still high.Reference Fava and Davidson 27 As such, it is important to first understand the possible reasons for treatment failure.Reference Cuthbert and Insel 28 , Reference Vahid-Ansari, Zhang, Zahrai and Albert 29
Medication nonadherence is an endemic problem that commonly contributes to the apparent “failure” or cessation of the effect of a course of antidepressant therapy after an initial response.Reference Marasine and Sankhi 30 The rates of nonadherence at 4-month follow-up among older adults ranged from 29% to 40% in the USA.Reference Kales, Kavanagh and Chiang 31 , Reference Kales, Nease and Sirey 32 In another study examining nonadherence rates in primary care and psychiatric populations across different countries, about 50% of patients were found to discontinue antidepressant medications prematurely.Reference Sansone and Sansone 33 A systematic review of 21 studies indicated that patient factors (e.g., forgetfulness, comorbidities, and misconceptions about the disease), medication factors (e.g., polypharmacy, side effects, and pill burden), healthcare system-related factors (e.g., physician-patient interactions) and sociocultural factors contributed to the antidepressant nonadherence in patients with MDD.Reference Marasine and Sankhi 30 For all of these reasons, assessment of the history of all patients with difficult-to-treat depression should begin with a careful consideration of adherence.
Another potential reason for treatment failure is the “blunt instrument” nature of antidepressants: even relatively selective antidepressants act on many receptors in the brain, often with unwanted effects in the periphery.Reference Cuthbert and Insel 28 , Reference Blackburn 34 For example, SSRIs are presumed to work by improving the function of serotoninergic neurotransmission in the brainReference Vahid-Ansari, Zhang, Zahrai and Albert 29; however, serotonin is linked to the regulation of not only emotion, mood, stress, appetite, and sleep but also the control of vascular resistance and blood pressure, heart function, mammary gland development, and digestion.Reference Vahid-Ansari, Zhang, Zahrai and Albert 29 , Reference Berger, Gray and Roth 35 SSRIs lead to remission in 30% of patientsReference Vahid-Ansari, Zhang, Zahrai and Albert 29 , Reference Rush, Warden and Wisniewski 36 but are associated with a wide range of side effects, such as memory impairment, somnolence, decreased concentration, fatigue, weight gain, headache, sexual dysfunction, and dizziness.Reference Anagha, Shihabudheen and Uvais 37 , Reference Ferguson 38 As individual serotonin neurons are highly branched, sending input to multiple forebrain structures (Figure 1), the global targeting of serotonin by SSRIs likely activates antagonistic pathways that may contribute to the side effects.Reference Vahid-Ansari, Zhang, Zahrai and Albert 29 These unwelcome effects may impact the tolerability and acceptability of SSRIs and may increase the likelihood of medication nonadherence.Reference Sansone and Sansone 33 Individual serotonin neurons are highly branched and send input to multiple forebrain structures, the midbrain, and the hindbrain (cerebellum). Hence, they target the entire central nervous systemReference Vahid-Ansari, Zhang, Zahrai and Albert 39; serotonin also targets other tissues and cells.Reference Berger, Gray and Roth 40 , Reference Watanabe, Rose, Kanayama, Shirakawa and Aso 41
Figure 1. Serotonin neurons target multiple brain structures and other organs, tissues, and cells.
Treatment failure may also be related to the inter-individual variability in treatment response, which has been shown to be heritable and so, in part, is affected by genetic variation.Reference Pain, Hodgson and Trubetskoy 42 In a sample of 2799 patients treated with antidepressants, 42% of individual differences in antidepressant response were explained by genetic variants,Reference Tansey, Guipponi and Hu 43 which are likely acting together to express a range of behavioral and somatic traits.Reference Andreassen, Hindley, Frei and Smeland 44 A previous study in a Chinese population identified single nucleotide polymorphisms that resulted in poorer treatment responses to fluoxetine and venlafaxine.Reference Bi, Ren and Guo 45 However, several genome-wide association studies have not been able to identify genetic associations to robustly predict antidepressant response to date,Reference Pain, Hodgson and Trubetskoy 42 , Reference Biernacka, Sangkuhl and Jenkins 46 -Reference Tansey, Guipponi and Perroud 48 with extant studies either reporting trivial variance explained by genetics or potentially overestimating, due to sample size, the genetic contributions to antidepressant response through the use of genome-wide complex traits or similar analyses.Reference Tansey, Guipponi and Hu 43 , Reference Li, Tian and Hinds 49 , 50
Misdiagnosis can also lead to treatment failure and may occur for a variety of reasons, including comorbid disorders and the heterogeneous nature of depressive disorders.Reference Shen, Zhang, Xu, Zhu, Chen and Fang 51 The problem can be compounded by an incomplete understanding of the patient’s condition, resulting in an incomplete or superficial clinical assessment,Reference Ayano, Demelash and yohannes 52 leading to a failure to differentiate symptoms of unipolar (i.e., recurrent episodes of MDD) and bipolar depression (BD),Reference Shen, Zhang, Xu, Zhu, Chen and Fang 51 or identifying and addressing mixed depressive states in a person who has never suffered a discrete hypomanic or manic episode.Reference Koukopoulos, Sani and Ghaemi 53 , Reference Pacchiarotti, Mazzarini and Kotzalidis 54 For example, the DSM-5 definition of mixed depression combines manic and depressive symptoms only where the symptoms do not overlap, thereby excluding psychomotor agitation, irritability, and distractibility, which are common symptoms experienced during mixed states.Reference Koukopoulos, Sani and Ghaemi 53 This can lead to an improper diagnosis and treatment, which may affect the patient’s outcome.
Case illustration for a patient with mixed features in bipolar depression
A patient with bipolar I depression presented with mixed features for several months with uncontrollable panic, emotional instability, and symptoms of inattention. The patient also had comorbid anxiety and attention deficit hyperactivity disorder (ADHD) and was being treated with lithium (1500 mg/day) monotherapy. Historically, the patient found only modest benefit from combination treatments with quetiapine, gabapentin, SSRI, and stimulants for 12 months and experienced inadequate response. However, when lithium was increased by 300 mg to 1800 mg, the patient experienced further resolution of depressive symptoms, including agitation.
This case suggests that when patients with bipolar I depression with comorbid anxiety and ADHD experience breakthrough mixed feature symptoms, optimizing mood stabilization through dose-finding and adjustment of current medications before treating anxiety and/or ADHD could be a useful first step and mitigate unhelpful polypharmacy. A careful history and stepped decision-making can impact the treatment outcomes for patients who are difficult to treat.
The impact of placebo response in clinical trials for MDD; adapting trial design
Treatments for depression have both specific and non-specific effects.Reference Brown and Peciña 55 In clinical trials, the impact of the non-specific elements of treatment is estimated for the sample by the placebo response rate. However, at the level of the individual, it is usually not possible to separate the specific and non-specific effects of treatment.Reference Khan and Brown 56 Placebo response rates have been documented in hundreds of clinical trials of MDD,Reference Parker, Ricciardi and Hadzi-Pavlovic 57 , Reference Furukawa, Cipriani and Atkinson 58 with some evidence that the non-specific component of treatment response has increased over the past 30 years.Reference Khan, Fahl Mar, Faucett, Khan Schilling and Brown 59 There is concern that the problem of increasing placebo response has been particularly problematic for investigators studying pharmacotherapy of MDD in children and adolescents.Reference Feeney, Hock, Fava, Hernández Ortiz, Iovieno and Papakostas 60 Factors associated with variation between studies in placebo response rates may include study intervals, the diagnosis criteria used or rater biases in judging depression, baseline severity, trial length, and number of study sites.Reference Parker, Ricciardi and Hadzi-Pavlovic 57 , Reference Furukawa, Cipriani and Atkinson 58 , Reference Feeney, Hock, Fava, Hernández Ortiz, Iovieno and Papakostas 60 , Reference Stone, Yaseen, Miller, Richardville, Kalaria and Kirsch 61
High placebo response rates in clinical trials contribute to trial failures and delay the development of new antidepressants.Reference Rutherford and Roose 62 –Reference Cheung, Thiyagarajah and Abraha 64 Limiting the number of trial sites, enrolling patients with higher baseline severity at study entry, and implementing protections against expectancy have helped to curb the growth in placebo response rates.Reference Brown and Peciña 55 , Reference Stone, Yaseen, Miller, Richardville, Kalaria and Kirsch 61 Across the past few decades, patients with MDD were likely to benefit from an antidepressant drug by 15% beyond a placebo effect.Reference Stone, Yaseen, Miller, Richardville, Kalaria and Kirsch 61 Therefore, to adequately address the negative impact of a high placebo response on signal detection, the field needs a better understanding of the developmental, behavioral, social, and biological underpinnings of the placebo response and to effectively model prevailing mechanisms that drive study dropout when placebo is used in clinical trials.Reference Gomeni, Lavergne and Merlo-Pich 65
Neuroimaging using positron emission tomography (PET) has shown the changes in the brain due to placebo treatment in a 2-week single-blind, randomized lead-in of 2 identical oral placebos, followed by 10 weeks of open-label treatment.Reference Peciña, Bohnert and Sikora 66 The oral placebos were described to participants either as being a fast-acting antidepressant agent (active) or disclosed to be an inactive placebo (inactive). When compared with the inactive placebo, clinical responses to the “active” placebo treatment were associated with increased placebo-induced μ-opioid neurotransmission in the subgenual anterior cingulate cortex, nucleus accumbens, midline thalamus, and amygdala.Reference Peciña, Bohnert and Sikora 66 These results indicate that the variability in patient expectancy likely plays a role in placebo response, and design manipulations that inhibit placebo responses could help separate drug-specific treatment effects in clinical trials.Reference Brown and Peciña 55
Furthermore, post hoc examination of clinical trial databases revealed that early improvement or lack of response in the first 2 weeks of blinded therapy is a powerful predictor of subsequent response or nonresponse after 6 weeks of therapy.Reference Szegedi, Jansen, van Willigenburg, van der Meulen, Stassen and Thase 67 , Reference Kudlow, McIntyre and Lam 68 Strategies such as the sequential parallel comparison design (SPCD) attempt to capitalize on these observations to reduce the placebo response and to increase the efficiency of signal detection in clinical trials.Reference Mathew, Gueorguieva, Brandt, Fava and Sanacora 69 –Reference Fava, Memisoglu and Thase 71 SPCD is a two-stage study design in which a much higher proportion of patients are randomized to receive a double-blind placebo in the first stage.Reference Liu, Kim and Han 70 , Reference Fava, Memisoglu and Thase 71 At the end of stage 1, patients from the placebo group are classified as placebo responders or nonresponders; the latter are then re-randomized in a blinded fashion to active drug or placebo in stage 2.Reference Liu, Kim and Han 70 , Reference Fava, Memisoglu and Thase 71 However, regulatory agencies such as the Food and Drug Administration (FDA) have not determined if studies using the SPCD method are more likely to succeed than studies using more conventional designs.
Potential benefits of a more targeted approach
A systematic review and meta-analysis of 522 double-blind studies found that of the 21 antidepressant drugs studied, all were more effective than placebo in adults with MDD.Reference Cipriani, Furukawa and Salanti 72 This suggests that we already have effective antidepressants if treatment is based on neurobiology, neuronal networks of depression, and precision pharmacology, with its focus on diagnosis-based science, not symptoms.Reference Cuthbert and Insel 28 , Reference Blackburn 34 However, clinical guidelines are often limited, as they give general information about drug classes and guidance for treatment selection but do not provide further details for the individual compounds.Reference Zanardi, Prestifilippo, Fabbri, Colombo, Maron and Serretti 22 , Reference Serretti 73 , 74 There is poor guidance in prescribing guidelines about the possible strategies to personalize antidepressant prescriptions.Reference Zanardi, Prestifilippo, Fabbri, Colombo, Maron and Serretti 22 , Reference Serretti 73 , Reference Fabbri and Serretti 75 Thus, the choice of an effective antidepressant treatment from over 40 available compounds is still a challenge, as prescription is often based on the personal experience of the clinician.Reference Zanardi, Prestifilippo, Fabbri, Colombo, Maron and Serretti 22 , Reference Serretti 73
The identification of robust clinical criteria and biomarkers (e.g., neuroimaging biotypes, genetic variants) for guiding both a mechanistic understanding of the disease and treatment choice is important in depression.Reference Williams and Hack 76 Due, in part, to the practical challenges of deep phenotyping with serum and neuroimaging tools with unknown or variable degrees of reliability and validity,Reference Uher, Tansey and Dew 77 , Reference Godlewska 78 consideration should also be given to factors such as past response to antidepressant medication, family pharmacological history, pharmacogenomics to optimize tolerability, and possible drug interactions, which can change medication plasma levels and pharmacodynamics.Reference Zanardi, Prestifilippo, Fabbri, Colombo, Maron and Serretti 22 , Reference Cuthbert and Insel 28 , Reference Serretti 73 This is especially important given the known associations between depression and health comorbidities such as inflammation and cardiometabolic disease risk.Reference Deif and Salama 79 Considering these factors may lead to more targeted therapy as the right treatments can be matched to the right patients, thereby increasing the benefit–risk ratio.Reference Cuthbert and Insel 28 Personalized treatments could also improve remission rates and reduce the risk of relapse, leading to recovery and better functioning in patients with depression,Reference Fabbri and Serretti 75 and reduce the need for a trial-and-error approach associated with drug adverse effects that can erode patient trust and hope.
It is also worth noting that personalization of therapy to improve outcomes is not limited to pharmacotherapies. Personalization achieved through the optimization of stimulation targets and parameters of transcranial magnetic stimulation has demonstrated improved efficiency as compared with standard neuromodulation protocols.Reference Fang, Godlewska, Cho, Savitz, Selvaraj and Zhang 80
Precision psychiatry findings for a more circuit-driven approach
Given the heterogeneity of depressionReference Goldberg 81 and the relatively modest efficacy of existing antidepressants, a move away from simply symptom-based diagnosis is urgently needed. One way this could be achieved is through precision psychiatry, which is an approach to psychiatric treatment that is based on understanding the neurobiological mechanisms that cause symptoms so that treatment can be tailored precisely to those mechanisms.Reference Blackburn 34 Concisely, precision psychiatry may be viewed as the right treatment for the right patient,Reference Trivedi 82 with the understanding that timing of treatment may also play a key role.
A shift away from the classification structure of DSM-4 to biologically based diagnosis was initiated in 2009 by the United States National Institute of Mental Health as part of a long-term strategic initiative with their Research Domain Criteria Project (RDoC).Reference Cuthbert and Insel 28 , Reference Blackburn 34 , Reference Insel 83 While DSM-5 does incorporate some neuroscience not included in previous versions,Reference Williams 84 the RDoC aims to develop new ways of classifying mental disorders based on dimensions of observable behavior and neurobiological measures.Reference Cuthbert and Morris 85 , Reference Williams, Carpenter, Carretta, Papanastasiou and Vaidyanathan 86
The dimensions of the RDoC are organized into six superordinate domains of functioning: negative valence, positive valence, cognition, social processes, arousal/regulatory systems, and sensorimotor systems.Reference Cuthbert and Morris 85 Each domain contains several constructs characterized by data from behavior or cognitive function, evidence for a neural circuit, and relevance to psychopathology.Reference Cuthbert and Morris 85 RDoC considers mental disorders from a translational point of view in two steps: in the first step, it determines the primary behavioral functions of the brain and specifies neurobiological systems responsible for these functions; in the second step, psychopathology in terms of dysfunction of different kinds in particular systems is considered from an integrative, multi-systems point of view,Reference Cuthbert and Insel 28 thereby enabling deep phenotyping. Importantly, the RDoC was developed not only to generate the initial constructs framework but also to evolve with scientific progress. Although RDoC is largely theory-based, there is ongoing investigation and validation of the proposed constructs in a data-driven way, as well as in optimizing tools to assess RDoC constructs.Reference Gordon 87 , Reference Cuthbert 88 While principles of the RDoC have extended into clinical studies, regulatory bodies including the FDA and European Medicines Agency, continue to base study population inclusion criteria on DSM-5 or the International Statistical Classification of Diseases and Related Health Problems diagnostic coding for MDD, 89 , 90 presenting possible challenges for this approach.
Technical advances and improved knowledge have provided new insights into the brain circuits that underlie cognitive and emotional functioning.Reference Williams 84 For example, a possible neural circuit taxonomy has been proposed to address the gap between advances in brain imaging and clinical practice for mental disorders,Reference Williams 84 in place of a symptom-led taxonomy. Certain dysfunctions in large-scale circuits that control emotional and cognitive functions describe distinct biotypes of depression and anxiety, which may commonly co-occur in individuals.Reference Williams 84 For example, six neural circuits have been proposed in dysfunctions expressed in depression and anxiety: default mode, salience, negative affect, positive affect (reward), attention, and cognitive control.Reference Williams 84 Another framework that has been proposed is based on brain lesions networking mapping, where it was shown that functional connectivity between lesion locations and the left dorsolateral prefrontal cortex was strongly associated with depression.Reference Padmanabhan, Cooke and Joutsa 91 Consequently, this neural circuit is thought to hold promise for precision targeted therapy in individuals with depression.
Other examples of neural circuit-based biotypes may inform pharmacotherapy, the most used treatment for MDD.Reference Williams 84 In a clinical trial, anterior insula hyperactivation during resting metabolism was identified (via PET scanning) as a differential biomarker of remission for escitalopram.Reference McGrath, Kelley and Holtzheimer 92 Similarly, amygdala reactivity to emotional faces was used to identify individuals who are unlikely to respond to particular types of antidepressants in the randomized international Study to Predict Optimized Treatment for Depression (iSPOT-D) clinical trial that combined antidepressant therapy with pre-/post-neuroimaging scans.Reference Williams, Korgaonkar and Song 93
Neuroimaging may be used to achieve a more precise diagnosis based on characterizing the underlying neural circuit function, thereby providing the clinician with additional data to inform treatment choices, such as selecting an appropriate pharmacotherapy and limiting side effects.Reference Fang, Godlewska, Cho, Savitz, Selvaraj and Zhang 80 , Reference Williams and Hack 94 , Reference Shalbaf, Brenner and Pang 95 In another example, using iSPOT-D data, remission on standard first-line antidepressants depended on pre-treatment connectivity between the posterior cingulate cortex and the anterior cingulate cortex.Reference Goldstein-Piekarski, Staveland, Ball, Yesavage, Korgaonkar and Williams 96 Similarly, the comparative effectiveness of existing therapeutics can be explored based on neural circuit changes in response to different compounds. For example, through analyzing data collected from the iSPOT-D trial, it was demonstrated that sertraline responders had higher functional connectivity at baseline between the dorsolateral prefrontal cortex/supramarginal gyrus and supramarginal gyrus/middle temporal gyrus when compared with nonresponders.Reference Tozzi, Goldstein-Piekarski, Korgaonkar and Williams 97 The opposite was observed for the venlafaxine-extended release group, where responders had lower functional connectivity in these regions.Reference Tozzi, Goldstein-Piekarski, Korgaonkar and Williams 97 Following treatment with sertraline, reduction of connectivity in the precentral and superior temporal gyri was associated with symptom improvement; for the venlafaxine-extended release group, symptom improvement correlated with enhancement of connectivity between the orbitofrontal cortex and subcortical regions.Reference Tozzi, Goldstein-Piekarski, Korgaonkar and Williams 97
Resting-state electroencephalography (rsEEG) has also been used to predict the outcome of sertraline versus placebo in a neuroimaging-coupled, placebo-controlled antidepressant study.Reference Wu, Zhang and Jiang 98 EEG may be a more accessible tool for use in clinical practice, even with its relatively reduced spatial and temporal resolution compared with magnetic resonance imaging (MRI).Reference Wu, Zhang and Jiang 98 , Reference Al-Janabi 99 Symptom improvement predicted using the sertraline rsEEG signature was associated with prefrontal neural connectivity and was found to be consistent across different study sites and EEG equipment.Reference Wu, Zhang and Jiang 98
Examples of drug development techniques that may support precision psychiatry
Many innovative strategies have been used in the development of pharmacological agents some of those used in MDD, which may prove useful for precision psychiatry approaches, are discussed in Table 2.
Table 2. Examples of pharmacological development strategies being implemented to meet current unmet needs for patients with depression which may support precision psychiatry
Abbreviation: fMRI, functional magnetic resonance imaging; KNCQ, Voltage-gated potassium channels; KOR, kappa opioid receptor; MDD, major depressive disorder; MoA, mechanism of action; MT1, melatonin receptor type 1A; MT2, melatonin receptor type 1B; NMDA, N-methyl-D-aspartate; NR2B, N-methyl D-aspartate receptor subtype 2B; SNRI, serotonin-norepinephrine reuptake inhibitors; SOC, standard-of-care; SSRI, selective serotonin reuptake inhibitor.
Notably, the proof-of-mechanism strategy employed for aticaprant, a kappa opioid receptor antagonist 100 , Reference Krystal, Pizzagalli and Smoski 101 may be one of the most useful approaches for the development of psychiatric agents, specifically in the realm of precision psychiatry. Aticaprant was developed based on the Fast-Fail Trials initiative developed by the National Institute of Mental Health.Reference Krystal, Pizzagalli and Smoski 101
As many pharmacological agents will fail to be approved for their possible indications, the concept of “Fast-Fail” was developed with the goal of eliminating these agents at earlier, less costly stages of clinical development.Reference Krystal, Pizzagalli and Mathew 102 To be developed under “Fast-Fail,” potential agents must meet four requirements: (1) Compelling preclinical research establishing that engaging the target would likely have a therapeutic effect on the brain; (2) Engagement of the target by a compound can be measured in a robust method; (3) The compound specifically engages with the target and preclinical safety data supports human trials; (4) A brain biomarker with a therapeutic potential to serve as the proof-of-mechanism outcome measure for the study.Reference Krystal, Pizzagalli and Mathew 102 Although aticaprant was identified for Phase III clinical trial by the “Fast-Fail” trial approach, 100 there are inherent drawbacks to applying this strategy. Firstly, in psychiatric disorders, there is a limited availability of biomarkers suitable for study outcomes, and not all targets of interest have a robust means of measuring target engagement. It may be possible for the investigational drug to impact other targets and cause clinical changes but not engage the prespecified target. Based on the “Fast-Fail” criteria, this would result in a negative study result, highlighting the need for an established, sufficiently sensitive primary outcome.Reference Krystal, Pizzagalli and Mathew 102
Mechanistically driven approaches have been used in other areas of pharmacological development, such as for the development of valbenazine, a reversible vesicular monoamine transporter-2 (VMAT-2) inhibitor used for the treatment of tardive dyskinesia. 103 , Reference Grigoriadis, Smith, Hoare, Madan and Bozigian 104 Tetrabenazine is an approved treatment for chorea associated with Huntington’s disease and has demonstrated improvements in hyperkinetic movement disorders. 105 , Reference Miguel, Mendonca and Barbosa 106 Valbenazine and tetrabenazine have a common isomer, which was found to be the most potent inhibitor of VMAT-2, supporting the development of this mechanism-based therapeutic.Reference Grigoriadis, Smith, Hoare, Madan and Bozigian 104 These developments highlight the potential benefits of mechanistically driven clinical research, which may support precision psychiatry approaches.
Precision psychiatry to inform phenotypes
There is a need to develop combinatorial diagnostic approaches and tools that can be applied in precision psychiatry to inform phenotypic profiles of patients in clinical settings.Reference Lin, Lin and Lane 107 For example, multi-omics and neuroimaging data can be used as biomarkers to achieve a more precise diagnosis that will assist clinicians in offering the right treatment.Reference Williams and Hack 94 , Reference Lin, Lin and Lane 107
The use of artificial intelligence methods is still in its infancy in terms of forecasting drug treatments in psychiatry. In time, probabilistic symptom targeting, as well as deep learning algorithms, may be used to predict treatment response, prognosis, diagnosis, and detection of potential biomarkers.Reference Lin, Lin and Lane 107 , Reference Athreya, Brückl and Binder 108 For example, a machine learning algorithm using a multidomain data integration model consisting of peripheral blood and cognitive markers was used to predict the diagnosis of bipolar disorder.Reference Fernandes, Karmakar and Tamouza 109 Compared with control, a sensitivity of 80% and specificity of 71% was observed for bipolar disorder, suggesting that these blood and cognitive biomarkers could be used by clinicians for diagnosis depending on the clinical situation.Reference Fernandes, Karmakar and Tamouza 109 Similarly, a probabilistic graphical model followed by unsupervised machine learning was used to identify specific depressive symptoms and thresholds of improvement that predicted antidepressant response by 4 weeks and the achievement of remission, response, or nonresponse by 8 weeks in 947 patients with depression.Reference Athreya, Brückl and Binder 108 Specific thresholds of change in 4 depressive symptoms, namely depressed mood, feelings of guilt and delusion, work and activities, and psychic anxiety, at 4 weeks predicted the subsequent outcome at 8 weeks to SSRI therapy with an average accuracy of 77%.Reference Athreya, Brückl and Binder 108 In another study, a multisite trial of sertraline versus placebo for adults with MDD was performed using a combination of machine learning with a Personalized Advantage Index (PAI).Reference Webb, Trivedi and Cohen 110 The study determined whether individualized treatment recommendations can be generated based on endophenotype profiles coupled with clinical and demographic characteristics.Reference Webb, Trivedi and Cohen 110 The study found that a subset of patients with MDD optimally suited to sertraline could be identified based on pre-treatment characteristics, which included higher baseline severity of depressive symptoms, older patients, higher neuroticism, less impairment in cognitive control, and being employed.Reference Webb, Trivedi and Cohen 110 Further work is needed, including prospective tests in which the PAI model is built and tested in 2 different samples, but the results of this study demonstrate the potential to use algorithms to predict treatment outcomes. Ultimately, comparative effectiveness trials of relatively comparable treatments or treatment approaches will be a cornerstone for precision psychiatry.
Notably, improved accessibility and increased sharing of health-related data between institutions and sectors for research and clinical uses may further advance the use of artificial intelligence.Reference Villanueva, Cook-Deegan and Koenig 111 Facilitating analysis of the electronic health record with the use of artificial intelligence allows for more personal care by identifying at-risk patients for early intervention or for generating an actionable insight for these patients.Reference Khalifa, Albadawy and Iqbal 112
Neuroimaging techniques such as PET and MRI have been used to study the impact of genetic variants on drug target engagement.Reference Silberbauer, Rischka and Vraka 113 A placebo-controlled, crossover study of healthy volunteers and patients with MDD used these neuroimaging techniques to evaluate serotonin transporter occupancy after infusion with citalopram (an SSRI) to assess the impact of ABCB1 gene variants on drug target engagement in the brain.Reference Silberbauer, Rischka and Vraka 113 Six ABCB1 single nucleotide polymorphisms were tested, and lower serotonin transporter occupancy was found in ABCB1 rs2235015 minor allele carriers compared with major allele homozygotes, as well as in men compared with women.Reference Silberbauer, Rischka and Vraka 113 These results highlight the potential of imaging genetics for precision pharmacotherapy in psychiatry.
Use of pharmacogenomics to target treatment
The first large-scale study to utilize pharmacogenetic (PGx)-guided selection in MDD yielded mixed results.Reference Pérez, Salavert and Espadaler 114 This was a prospective, double-blind, randomized controlled trial conducted in Spain to assess whether PGx-guided treatment is more effective than unguided treatment in improving drug response and tolerability.Reference Pérez, Salavert and Espadaler 114 Although no difference in sustained response (primary endpoint) was observed between patients receiving PGx-guided treatment and patients receiving treatment as usual during the study period, the PGx-guided treatment group had a higher responder rate at Week 12. This effect was stronger in patients with 1–3 previously failed psychiatric treatments, with a 2.4-fold increase in the odds of response for these patients. Additionally, PGx-guided treatment resulted in an improved likelihood of achieving better medication tolerability compared with treatment as usual. The results suggest that the use of PGx information to guide treatment adjustments may be justified if traditional first-line treatment fails.Reference Pérez, Salavert and Espadaler 114
Another multicenter, prospective, double-blind, randomized controlled trial in the USA used pharmacogenetic testing to guide medication management recommendations for depression and anxiety based on gene-drug and drug-drug interactions for over 40 medications used in the treatment of depression and anxiety.Reference Bradley, Shiekh and Mehra 115 Response and remission rates at Weeks 8 and 12 were significantly higher for patients receiving PGx-guided treatment compared with patients treated with the usual standard of care. There was no statistical difference in adverse drug events between the two groups.Reference Bradley, Shiekh and Mehra 115 The randomized controlled Precision Medicine in Mental Health Care; PRIME Care trial of 1944 patients with MDD compared treatment guided by pharmacogenomic testing versus usual care.Reference Oslin, Lynch and Shih 116 The PRIME Care study demonstrated that pharmacogenomic testing for drug-gene interactions reduced the prescription of drugs with predicted drug-gene interactions compared with the usual care. However, while remission rates were modestly higher at Weeks 8 and 12 in the pharmacogenomic testing group compared with patients receiving usual care, no advantage was observed at Week 24.Reference Oslin, Lynch and Shih 116
A pharmacogenomic and survival analysis was used to determine suitable antidepressants for the Chinese population.Reference Bi, Ren and Guo 45 A total of 610 patient samples were treated with a selection of SSRIs, serotonin norepinephrine reuptake inhibitors (SNRIs), noradrenergic, and specific serotonergic antidepressants (NaSSA) or TCAs.Reference Bi, Ren and Guo 45 The study indicated that treatment with SSRIs and SNRIs was more efficacious than with TCAs and NaSSAs in the Chinese population. The study also showed that certain genetic variants were significantly susceptible to a worse response to fluoxetine; these genes were present on the neurotrophin pathway in patients with depression comorbid with anxiety.Reference Bi, Ren and Guo 45
Further, a phase 2b trial in participants with treatment-resistant depression utilized the novel genomic biomarker Denovo Genomic Marker 4 (DGM4) to predict the antidepressant response of a novel agent, liafensine. Results of this biomarker-guided study indicated significant improvements in treatment-resistant depression following treatment with liafensine, leading to a Fast Track designation by the FDA. 117 – 119
The clinical decision of whether to use pharmacogenomic testing should be guided by a risk–benefit analysis. While the cost of implementing pharmacogenomic testing is likely high, there are potential benefits to the individual patient in providing precise care.Reference Oslin, Lynch and Shih 116
Challenges to the application of precision psychiatry
While some benefits of precision psychiatry have been outlined above, there are several challenges that may limit clinical translation and utility at this time. The majority of clinical trials recruit patients with mild and moderate severities, and generalization in real-life practice cannot be made to patients with severe disease.Reference Deif and Salama 79
Disagreement exists around the validity of grouping depression into specific subtypes based on symptoms and the presence of specific endophenotypes.Reference Bayes and Parker 120 There are variations in the methods of data collection and technical complexity required to process and analyze multi-omics data from large datasets and/or artificial intelligence.Reference Deif and Salama 79
The cost-effectiveness of some of the techniques used in precision psychiatry is still not well known, nor is the cost of appropriate training of healthcare staff in these different techniques.Reference Deif and Salama 79 Ethical concerns, such as protecting the privacy and security of data and patient stratification (risk of discrimination against patients in less privileged groups), also exist.Reference Deif and Salama 79 Additional studies related to the cost-effectiveness of precision psychiatry are warranted to ensure improved treatment approaches are accessible to all patients.
Other challenges faced in the precision psychiatry field are the lack of validated biomarkers that can serve as viable targets for precise therapeutics,Reference Goldstein-Piekarski, Staveland, Ball, Yesavage, Korgaonkar and Williams 96 as well as a lack of comparative effectiveness studies. Guidance on personalized treatments, including the type and length of treatments, and more studies using extended follow-up of individuals treated for depression are needed. The results of a meta-analysis indicated that maintenance therapy should be continued for at least 6 months after remission.Reference Kato, Hori and Inoue 121 This meta-analysis also suggested that continuing antidepressants for another year led to lower relapse rates in patients with MDD, and flexible dose adjustment based on symptoms could help prevent relapse.Reference Kato, Hori and Inoue 121
Precision psychiatry for children, adolescents, and specific populations of adults
In the USA, 17% of adolescents had at least one major depressive episode in 2020. 122 The risk of experiencing MDD is highest during adolescence and is associated with adverse consequences persisting well into adulthood.Reference Chahal, Gotlib and Guyer 123 During adolescence, the neurocircuitry involved in depression is still developing, and it is crucial that mental health problems are identified and treatment is initiated during this period.Reference Chahal, Gotlib and Guyer 123 , Reference Sequeira, Battaglia, Perrotta, Merikangas and Strauss 124
Due to the substantial side effects associated with psychiatric medications, clinicians often initiate pharmacotherapy at low doses in children and adolescents and slowly titrate the dose. This may increase the risk of under-treatment and may lead to a medication change due to the lack of treatment response.Reference Wehry, Ramsey, Dulemba, Mossman and Strawn 125
An increase in the data available for pediatricians to augment existing treatment guidelinesReference Cheung, Zuckerbrot, Jensen, Laraque and Stein 126 would be of great benefit. Much of the work in pharmacogenomic testing has been conducted in adults, though there has recently been an increase in studies applying these tests as a predictor of treatment response and medication tolerability in pediatric patients.Reference Wehry, Ramsey, Dulemba, Mossman and Strawn 125 The rates of placebo response are higher in children and adolescents than in adults, exacerbating the difficulties in establishing efficacious treatments.Reference Rutherford and Roose 62 , Reference Meister, Abbas and Antel 127
Digital phenotyping refers to the “moment-by-moment quantification of the individual-level human phenotype in situ using data from smartphones and other personal digital devices.”Reference Torous, Kiang, Lorme and Onnela 128 This field may be of particular relevance to child and adolescent psychiatry.Reference Sequeira, Battaglia, Perrotta, Merikangas and Strauss 124 An ongoing study, Texas Resilience in Adolescent Development, follows participants 10–24 years of age at risk for depression. The aim of the study is to uncover the sociodemographic, lifestyle, clinical, psychological, and neurobiological factors that contribute to mood disorder onset, recurrence, progression, and differential treatment response.Reference Trivedi, Chin Fatt and Jha 129
Other populations of adults that could benefit from precision psychiatry are older adults and those with BD, as there are clinical challenges in differentiating BD from non-bipolar depression, which often leads to delays in diagnosis and accurate treatment. A more targeted treatment could also be used for pregnant people with depression during the perinatal or postpartum period. Perinatal depression (PND) is heterogeneous as there are likely multiple contributing etiologies and neurohormonal responses.Reference Kimmel, Bauer and Meltzer-Brody 130 , Reference Patterson, Balan, Morrow and Meltzer-Brody 131 Neurosteroid targets that attend to the neurohormonal context of depression in the postpartum period have recently been approved by the FDA.Reference Patterson, Balan, Morrow and Meltzer-Brody 131 – 133 Determining the biological features responsible for PND could shed light on how precision psychiatry may be used to tailor treatment options.Reference Kimmel, Bauer and Meltzer-Brody 130
Suggested approaches for future clinical studies
Overall, a more targeted approach to treatment using precision psychiatry may offer future benefits to patients. Continued use of the National Institute of Mental Health Fast-Fail Trials may serve as a starting point. As this initiative is based on preclinical research establishing that the engaging target could have a therapeutic effect on the brain and using brain biomarkers to serve as the proof of mechanism outcome measure for the study,Reference Krystal, Pizzagalli and Mathew 102 and therefore may offer an improved strategy for the development of psychiatric agents.
Additionally, increased use of human induced pluripotent stem cells (iPSCs) in psychiatric research may also help drive precision psychiatry efforts. IPSCs offer a reproducible method for modeling human diseases,Reference Nicholson, Ting and Chan 134 and in some conditions, iPSCs models have demonstrated aspects of the intended disease compared with controls.Reference Moretti, Bellin and Welling 135 , Reference Park, Arora and Huo 136 Future research may benefit from further use of iPSCs, including the use of brain organoids,Reference Nakazawa, Hashimoto, Takuma and Hashimoto 137 and in disease processes where murine and human physiology vary.Reference Park, Arora and Huo 136
Conclusions
An unmet need exists in MDD to develop diagnostic tools and more targeted therapy using precision psychiatry-based approaches so that the right treatments can be matched to the right patients. Identification of clinical biomarkers may allow for a more precise approach to treatment, in which specific disease mechanisms (which may sometimes be shared across multiple disorders) are targeted. PGx testing, neuroimaging, and machine learning approaches have been used with some success in trial settings, and some neural circuit biotypes associated with MDD have been identified. The core challenge remains as targeted receptor physiology is only part of the complex dysfunction due to the wide distribution of many receptors. Further advances in precision psychiatry pharmacotherapy may hinge on the spatial identification of selective subclasses of receptors. Still, clinical biomarkers that can effectively predict response to treatment based on individual phenotypes are needed. Personalized treatments could improve remission rates and reduce the risk of relapse, leading to overall better functioning in patients with depression.
Acknowledgments
Boehringer Ingelheim Pharmaceuticals, Inc. was given the opportunity to review the manuscript for medical and scientific accuracy as well as intellectual property considerations.
Author contribution
The authors of this manuscript meet the criteria for authorship as recommended by the International Committee of Medical Journal Editors. All authors drafted the work, or reviewed it critically for important intellectual content, gave final approval of the version to be published, and agreed to be accountable for all aspects of the work.
Financial support
The authors received no direct compensation related to the development of the manuscript. Editorial support in the form of preparing draft sections based on input from authors, collation and incorporation of author feedback to develop subsequent drafts, assembling tables and figures, copyediting, and referencing was provided by Blessing Anonye, PhD, of Avalere Health Global Limited, and was funded by Boehringer Ingelheim Pharmaceuticals, Inc.
Disclosures
MKS: Dr. Singh has received research support from the National Institutes of Health, the Patient Centered Outcomes Research Institute, and the Brain and Behavior Research Foundation. She is on a data safety monitoring board for a study funded by the National Institute of Mental Health. She has, in the past 3 years, consulted for AbbVie, Alkermes, Alto Neuroscience, Boehringer Ingelheim, Johnson and Johnson, Karuna Therapeutics, Inc., and Neumora. She receives honoraria from the American Academy of Child and Adolescent Psychiatry and royalties from American Psychiatric Association Publishing.
MT: Dr. Thase has received grant/research support from Acadia, Alkermes, AbbVie, AssureRx Health (now Myriad Neuroscience), Axsome, Intra-Cellular Therapies, Janssen Pharmaceuticals, Inc., National Institute of Mental Health, Otsuka, Patient Centered Outcomes Research Institute, and Takeda. He is a consultant for Axsome, Clexio Biosciences Ltd, GH Research, Janssen Pharmaceuticals, Inc., H. Lundbeck A/S, Otsuka Pharmaceutical Co., Ltd., Pfizer Inc., Sage Pharmaceuticals, Seelos Therapeutics, Sunovion Pharmaceuticals Inc., Takeda Pharmaceutical Company Ltd, and receives royalties from American Psychiatric Association Publishing, Inc., Guilford Publications, Herald House, Wolters Kluwer, and W.W. Norton & Company, Inc.
Open access
