Potential efficacy of dopaminergic antidepressants in treatment resistant anergic-anhedonic depression results of the chronic anergic-anhedonic depression open trial – CADOT

Introduction Among treatment-resistant depression (TRD), we identified anergic-anhedonic clinical presentations (TRAD) as putatively responsive to pro-dopaminergic strategies. Based on the literature, non-selective monoamine oxidase inhibitors (MAOI) and dopamine D2 receptor agonists (D2RAG) were sequentially introduced, frequently under the coverage of a mood stabilizer. This two-step therapeutic strategy will be referred to as the Dopaminergic Antidepressant Therapy Algorithm (DATA). We describe the short and long-term outcomes of TRAD managed according to DATA guidelines. Method Out of 52 outpatients with TRAD treated with DATA in a single expert center, 48 were included in the analysis [severity – QIDS (Quick Inventory of Depressive Symptomatology) = 16 ± 3; episode duration = 4.1 ± 2.7 years; Thase and Rush resistance stage = 2.9 ± 0.6; functioning – GAF (Global Assessment of Functioning) = 41 ± 8]. These were followed-up for a median (1st – 3rd quartile) of 4 (1–9) months before being prescribed the first dopaminergic treatment and remitters were followed up 21 (11–33) months after remission. Results At the end of DATA step 1, 25 patients were in remission (QIDS <6; 52% [38–66%]). After DATA step 2, 37 patients were in remission (77% [65–89%]) to whom 5 patients with a QIDS score = 6 could be added (88% [78–97%]). Many of these patients felt subjectively remitted (GAF = 74 ± 10). There was a significant benefit to combining MAOI with D2RAG which was maintained for at least 18 months in 30 patients (79% [62–95%]). Conclusion These results support TRAD sensitivity to pro-dopaminergic interventions. However, some clinical heterogeneities remain in our sample and suggest some improvement in the description of dopamine-sensitive form(s).


Introduction
Until recently, depression was mostly considered to be a single entity, thus legitimizing one-size-fits-all therapeutic guidelines (1)(2)(3).This view, based on an atheoretical approach of consensus criteria, has some limitations.These are epitomized by the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) trial showing the virtual absence of superiority of one therapeutic strategy over another, resulting in 33% of non-remitters despite four successive lines of treatment (4), a proportion which grows up to 54% when dropouts are considered, i.e., when STAR*D is analyzed as a whole from an intention-to-treat perspective (ITT) (5).
Yet the reification of 'depressive disorder' as equivalent to a disease or a syndrome underpinned by a common pathophysiology is more of a belief than a factual statement (6,7).Depression is more likely a broad defined condition encompassing multiple syndromes or diseases whose different causes make them more responsive to specific interventions.Considering heterogeneities in mood disorders aims to avoid the trial-and-error approach promoted by current guidelines and to focus on the most appropriate treatments, i.e., precision medicine (8,9).The analogy between dopamine withdrawal apathy and anergic-anhedonic depression led us to reinterpret the latter as a mesolimbic hypodopaminergic syndrome likely to respond to dopamine-enhancing strategies.
In the 2010s, movement disorders specialists described the emergence of apathy after surgery for Parkinson's disease.Five months after subthalamic nucleus deep brain stimulation, half of the patients developed a loss of motivation, a difficulty in initiating goal-directed cognitions and actions by themselves coming together with anhedonia, which could develop to emotional numbness and depression (10).Accumulating evidence suggests that this symptomatology is the result of mesolimbic hypodopaminergia: (i) apathy is unmasked by the rapid reduction of dopaminergic drugs allowed by the dramatic improvement of motor symptoms, (ii) it is reversed by dopamine D2 receptor agonists (D2RAG) (10,11), and (iii) dopaminergic denervation was found in the ventral striatum, amygdala, and medial and lateral orbito-frontal cortices of apathetic patients (12).Dopamine withdrawal apathy was a reminder of some clinical pictures of depression, not only in their clinical presentation but also their pharmacological reactivity.Soon after its discovery, iproniazid, the first non-selective irreversible monoamine oxidase inhibitor (MAOI), was reported to be especially effective on 'inert psychasthenic-anhedonic reactions' (13).Ten years later, the Pittsburgh group showed 'anergic depression' , characterized by a loss of initiative experienced as a volitional inhibition (anergia), anhedonia, and reversed vegetative symptoms (hypersomnia and hyperphagia), to be selectively responsive to tranylcypromine (TCP)another MAOI (14-16).Another decade later, the Columbia group redefined 'atypical depression' as emotional hyper-reactivity, leaden paralysis, and reversed neurovegetative symptoms, and showed its responsivity to phenelzine -again a MAOI (17,18).Anergic and atypical depressions were originally thought to be of different natures: a clinical picture of manic-depressive illness for the former (i.e., endogenous) vs. non-endogenous (i.e., neurotic) depression frequently developing on personality disorders for the latter ( 19).Yet they shared features like fatigue and reversed neurovegetative symptoms and responded to MAOI.In the following, this designation refers only to irreversible and non-selective drugs.By inhibiting both A and B isoforms, MAOI is one of the very few classes of pro-dopaminergic antidepressants (20).
Analogical reasoning with the mesolimbic hypodopaminergic syndrome led us to treat three patients with chronic and severe treatment-resistant depression (TRD) presenting prominent anergicanhedonic features with MAOI and D2 receptor agonists (D2RAG).The effect was dramatic, which led us to formalize the 'Dopaminergic Antidepressant Therapy Algorithm' (DATA) for treatment-resistant anergic-anhedonic depression (TRAD).
But in line with the seminal report of the Pittsburgh group, one patient was already diagnosed as bipolar (and treated with lithium) whereas the two 'apparently unipolar' patients switched to mania (14,19).Like the Pittsburgh group, this led us to speculate that TRAD belongs to the bipolar spectrum and we designed diagnostic criteria and therapeutic guidelines accordingly, i.e., DATA initially emphasized the preventive introduction of antimanic mood-stabilizers (see methods).The Chronic Anergic-anhedonic Depression Open Trial (CADOT) was designed to evaluate the effectiveness of DATA guidelines in the routine management of TRAD (efficacy and sideeffects).The following article reports the short and long-term outcomes and explores the predictors of remission.(19,24).Treatment-resistant anergic-anhedonic depression (TRAD) hereafter refers to the absence of response to two lines of antidepressants in anergic-anhedonic patients fulfilling the criteria listed in Table 1.

DATA: a two-step dopaminergic antidepressant therapy algorithm
The clinical characteristics of TRAD were supposed to be predictive of a good response to a dopaminergic strategy, i.e., DATA (Figure 1).Not included in the counting of steps is the preparatory phase.It was first recommended to (i) exclude contraindications for MAOI and/or D2RAG, (ii) remove medications that may impede dopamine transmission, e.g., foremost to stop D2-blockers, and (iii) initiate an antimanic mood-stabilizer.Lithium was initially strongly recommended.Over time, though lithium still was proposed as the first choice, lamotrigine was considered in the absence of previous manic or hypomanic episodes or bipolar relatives.
DATA was presented as a two-step dopaminergic approach aiming at full remission.The first step (DATA1) consisted of the introduction of a first dopaminergic class.The decision was discussed with the patient but a switch to MAOI was presented as the first choice relative to the add-on of D2RAG to the current antidepressant (preferentially a serotonin-norepinephrine reuptake inhibitor or SNRI).In 2013, iproniazid was the only MAOI available in France but was discontinued in 2015.By then, TCP was proposed as the first-choice because of its tolerance profile and efficiency in the Pittsburgh studies (15,16,19).Phenelzine was only considered as a first line MAOI in cases of prominent anxiety (25).Transdermal selegiline was mostly a second choice.Titration was adapted to tolerance but generally performed on a weekly basis up to the minimum effectiveness.The prescriber waited about four weeks to assess effectiveness, during which side effects were managed, before eventually increasing doses up to remission or the maximally tolerated dose.
D2RAG were considered in add-on, i.e., combined with an ongoing antidepressant (preferentially an SNRI).Pramipexole (PPX) was the first-choice D2RAG, initially with a three-dosesper-day scheme (as in Parkinson's disease).In 2018, concerns about D2RAG-induced heart failure and reports of efficacy when delivered once a day led to the recommendation of a single dose in the evenings (26).
In the absence of subjective remission after dose optimization, side effect management, and the sufficient duration (≥6 weeks) of DATA1, the patient was proposed to enter DATA2, i.e., the combination of MAOI with D2RAG.

Chronic anergic-anhedonic depression open trial -CADOT design
CADOT was designed as a monocentric, prospective observational mirror-image study.It started in 2013 after proof-ofconcept in a few patients and ended in June 2021.As CADOT aimed to evaluate DATA as a guideline for the treatment of chronic

Diagnostic criteria for anergic-anhedonic depression
Rank A: all of the following must be fulfilled Anergia: loss self-generated action and thought, volitional inhibition, poor motivation.
Anhedonia and/or emotional numbness.
Significant impact on personal or professional life (a decrease ≥20 pts. in GAF score and/or a GAF score ≤50) and significant duration ≥3 months.
Rank B: one or more of the following must be present Sadness secondary to self-criticism or guilt in relation with the poverty of action and emotion ¤ .
Atypical depression or at least one reversed neurovegetative symptom (one or more): Overeating and/or weight gain (≥10% of initial weight), Hypersomnia (≥10 h of sleep per day or increase ≥+2 h).±heavy sensation in the limbs and/or interpersonal rejection reactivity.
Mood reactivity and/or polymorphic and/or fluctuating clinical manifestations.
First degree relative suffering of bipolar I disorder or WKL-MDI.
Exclusion criteria: symptoms are not attributable to Moderate or absence of thought inhibition and dysexecutive syndrome † .
Current drug abuse or withdrawal ‡ .

Medical condition (including neurodegenerative diseases).
Translation from French criteria first posted on 10 Dec 2012, last updated on 13 Oct 2013 (French original version, 2013's English translation).GAF, Global Assessment of Functioning; WKL-MDI, Wernicke-Kleist-Leonhard's definition of manic-depressive illness.*"Mixed states are defined as the co-occurrence of both the manic and depressive pole among the different domains: affect, thought, and psychomotricity.[…] Incomplete states are an extension of the former concept, meaning that aside from being excited and inhibited, a single domain can also be completely unaffected" (22); ¤ The lack of motivation, the difficulty in initiating action, or the poverty of emotion leading to the feeling of failure to meet commitments or obligations, to self-criticism, to withdraw from social or family life, or to guilt for not feeling affectionate emotions towards loved ones.♱ Poverty of action is not primarily explained by difficulties in problem solving, i.e., orienting and maintaining attention, and planning and monitoring actions, as seen in dementia or severe obsessive-compulsive disorder (mental overload and procrastination).‡ Opioid withdrawal syndrome, especially during or after tapering of opioid substitution therapy, responds to reintroduction, whereas D2RAG may promote addictive behavior.1, patients were screened for contraindication to MAOI and/or D2RAG before being proposed DATA (27).Patients' characteristics and assessments were collected prospectively except the duration of the remission and tolerance data which were retrospectively extracted from patients' files.Treatment resistance was assessed using the Thase and Rush staging model (28).Symptom severity was evaluated by the Quick Inventory of Depressive Symptomatology clinician-rated version (QIDS), which is more sensitive to TRAD-specific 'atypical' features (hyperphagia and hypersomnia) (29).The QIDS was first assessed at the visit when the patient was prescribed DATA1 treatment, i.e., MAOI or D2AGO (excluding the preparatory phase).Interim and final evaluations were not performed at a specific time after initiation but when, after adapting the treatment at best, the physician felt that the patient was remitted or, if non-remitted, that no further improvement could be achieved.Finally, the Global Assessment of Functioning Scale (GAF) was only administrated at the beginning and at the end of the trial (with no interim evaluation) (30).
For all patients, the diagnosis was rediscussed with the physician in charge and management difficulties, e.g., deviation from the guidelines, were retrospectively examined.
CADOT received approval from the ethics committee of the University Hospitals of Strasbourg (approval reference no.CE-2023-31).All patients were contacted and consented to the use of their data.

Analysis plan 2.3.1. Primary outcome and analysis
The primary outcome was the proportion of remitters (Rm) defined by a QIDS score <6 (29).
The trial was analyzed from an ITT perspective.In accordance with the mirror image design, the only planned comparison was the proportion of Rm after DATA1 and after DATA2, i.e., the added value of the combination relative to a single drug (Odds ratio; χ 2 -test, α = 0.05).

Secondary outcomes and analyses
The proportion of responders (Rs) was computed to compare our results with the literature.Patients were considered as Rs if they were Rm or had their QIDS score reduced by at least 50%.
The absence of response could be related to resistance despite adequate trial, but also to early dropout, e.g., due to side effects.An adequate trial was defined as treatment maintained for at least 6 weeks at the minimum effective dose (MAOI ≥30 mg/d TCP-eq; D2RAG > 1 mg/d PPX-eq).
The time to remission was calculated from the first prescription to the first visit showing a remission that subsequently proved to be stable (≥6 weeks).In addition to the time of effectiveness, the time to Rm includes the progressive dose increases and the management of side effects which sometimes led to a change of molecule.
The time to trial completion was calculated from the first prescription to the management decision visit, i.e., pursuing the treatment, moving to DATA2, or quitting.For Rm, the time to the end of the trial equaled the time to remission.In non-remitters (nRm), the time to trial completion could be shorter if the patient quit because of side effects or longer because it took more time to conclude that the treatment had been best adapted.
Considering the paucity of missing data, no data imputation was performed.The Kaplan-Meier method was used to generate the rightcensored survival curves of Rm and Rs.Post-hoc exploratory Student, Wilcoxon, or χ 2 -test were performed on groups characteristics for continuous, non-Gaussian, and categorial data, respectively.Two-step Dopaminergic Antidepressant Therapy Algorithm (DATA) for treatment resistant anergic-anhedonic depression (TRAD).Leadin/preparation period: patients who accepted were first proposed to stop or reduce D2 receptor blockers and to introduce a mood stabilizer (preferably lithium).DATA1 (first step): patients were advised to switch to a non-selective (irreversible) monoamine oxidase inhibitor (MAOI -first choice).The add-on of dopamine D2 receptor agonist (D2RAG) to current antidepressant (classical) was presented as the second choice.DATA2 (second step): all non-remitters or patients who did not feel like before were proposed to combine MAOI and D2RAG.At each step, the prescriber was allowed to change molecules within the same class and adapt titration speed to optimize tolerance.Evaluations with QIDS were performed after sufficient stability to allow management decisions to maintain, stop, or combine treatments.Considering their exploratory purpose, no adjustment was made for multiple comparisons (α = 0.05, two-sided).Effect size was converted to standardized distance ('d' -see Supplementary material).

Conventions
Proportions are given in percentages, sometimes followed by 95% confidence intervals [in square brackets], computed using the normal approximation to the binomial distribution.Continuous values are summarized by their mean ± standard deviation except time to remission and time to trial completion which had left fat-tailed distributions.These were summarized by their median and their interquartile range in round brackets (Q1 -Q3).

Patients
In total, 56 of 61 outpatients diagnosed with TRAD were proposed with MAOI as the first line treatment.Four chose and responded to other therapeutic options: three bupropion-SNRI combinations and one individualized rTMS (7%) (35).Of the 52 subjects who entered DATA1, four (8%) received MAOI but could not be included in the analysis because of insufficient data: one Rm, one Rs, and two non-responders (nRs: i.e., implicitly non-remitters nRm).None received the combo, i.e., DATA2 (see Figure 2 for the flow-chart of inclusions).Flow chart of Chronic Anergic-anhedonic Depression Open Trial (CADOT).In total, 56 out of 61 TRAD outpatients presenting to Strasbourg's TRD reference center between 2013 and mid-2021 were proposed to be treated along DATA guidelines (none had contraindications to dopaminergic treatment).Of the four patients who wanted to try another option and remitted (Rm), three benefited from the combination of bupropion to an SSRI which is already a dopaminergic strategy.Because bupropion is not covered by the French health insurance system, it was not considered in this analysis.Only a few patients who could afford it had a bupropion trial before entering DATA.It should be noted that no patients responded to mood stabilizers alone.However, antidepressant washout periods longer than 8 weeks (mood stabilizers alone) were systematically offered to bipolar patients only until 2017.The lack of success led us to be less insistent thereafter.The baseline demographic and clinical characteristics of the 48 (93%) patients who were included are reported in Table 2, Section 2a.TRAD was severe (QIDS = 16 ± 3) and heavily disabling (GAF = 41 ± 8).The current episodes were resistant to treatments (≈stage III, i.e., having failed at least three trials with different classes including tricyclic antidepressants) and of long duration (3.3 ± 2.4 years).Thirty patients (63%) were chronically depressed (i.e., >2 years).

Study process
The flow chart of the study is reported in Figure 3.The long preparation/lead-in period of 19  weeks corresponds to the time it took to get patient approval of the DATA as a treatment scheme (meanwhile other treatments could be tried), to introduce a mood stabilizer, and to stop incompatible treatments, i.e., D2RAG and other antidepressants if MAOI was accepted as the first step.

Stoppers and switchers
Four nRs did not proceed to DATA2 because of side effects and switched treatment arms, i.e., from MAOI to D2RAG (n = 2) or the reverse (n = 2), of whom two remitted (Figure 3, T2).If we include them, 32 (67%) patients stopped after DATA1: 22 Rm, 27 Rs (+5), and 5 nRs (Figure 3, T3).The five Rs/nRm who decided to stop after DATA1 were satisfied with the result and maintained the treatment.All were at the edge of Rm (QIDS = 6) and had a good functional outcome (GAF = 73 ± 6).Regarding nRs, one patient could neither take MAOI nor D2RAG at an adequate dose because of side effects (two or more drugs of each class had been tried).Of the four nRs who achieved adequate doses and duration, only one continued the treatment (QIDS = 10); the three others stopped and did not want to test another molecule because of poor efficacy relative to side effects (QIDS = 17 ± 4).

DATA 2: combo
Only 16 patients proceeded to DATA2, of whom three were already Rm and three were Rs/nRm.All aimed at remission ad integrum.Accordingly, when entering DATA2, the average severity was mild to moderate (QIDS = 10 ± 4).Relative to the 32 patients who stopped after DATA1, patients who entered DATA2 tended to have more severe (p < 0.06) and more resistant depression at onset (p = 0.008, see Table 2, Section 2c).
As already mentioned, because of side effects, four patients switched from MAOI to D2RAG or vice-versa rather than combining the drugs.As indicated in Figure 3, the two patients who switched from D2RAG to MAOI remitted while the two patients who switched from MAOI to D2RAG were nRs (including the patient for whom neither MAOI nor D2RAG could be maintained at a sufficient dose and duration).These patients were only mentioned for their first trial in Table 2 Section 2b but included in the global outcome (ITT and per-protocol analyses; Figure 3-T5, Table 4).
Finally, regarding the exclusion criteria, five (10%) patients were later diagnosed or suspected of having a slowly developing form of synucleinopathies (four definite Parkinson's diseases, one probable prodromal form) (37).The single nRs after DATA2 subsequently developed cognitive and neurological disorders of probable mixed etiology (vascular and neurodegenerative), explaining the poor functional outcome despite mild symptom severity (Figure 3).Nine subjects had or developed significant personality traits with improvement, seven of whom fulfilled the 'atypical' criteria during the episode (see Supplementary material).

Intention-to-treat analysis 3.3.1. Proportion of Rm, Rs, and nRs
Of the 48 patients with TRAD who entered DATA, 37 (77%) were Rm with a total time to remission of 25 [14-55] weeks, i.e., about 6 months (adding the time of DATA1 and DATA2 for Rm after the combination, Figure 3-T5).Though not formerly Rm according to the classical threshold, the five Rs who stopped after DATA1 while at the edge of remission (QIDS = 6) were closer to the group of Rm than the group of nRs: they felt their needs were met and had no or little functional impairment (GAF = 73 ± 6, i.e., similar to Rm = 74 ± 11 but significantly different from nRs = 44 ± 8; p = 9 × 10 −4 ).
When patients started with MAOI, the combination with a D2RAG resulted in a significant increase in Rm and Rs (n = 17 → 29 and n = 25 → 32; p = 5 × 10 −6 and 2 × 10 −3 , respectively).Starting from D2RAG combined with a 'classical' antidepressant, the switch to a MAOI resulted in an increase of Rm (n = 6 → 8).This failed to reach significance likely due to the small sample size (n = 10) since the odds ratios are of the same magnitude (Table 3).

Duration of remission (survival analysis)
The 48 patients included in the analysis have been followed-up for a median of 3 [1.4-4.8] years since beginning DATA.The Kaplan-Meier estimator of the duration of remission considered 37 Rm and 5 Rs/nRm together.At 18 months, 79% [62-95%] of the sample remained in Rm (24 patients, i.e., 57% of the sample, remained in the analysis -Figure 4).Six out of nine relapses occurred when tapering antidepressant treatments (67%) and could be successfully treated by returning to therapeutic doses.

Predictors of Rm and Rs
If the resistance of depression predicted the outcomes of DATA1, this was no more the case for DATA2 and in the (global) ITT analysis.The only negative predictor of Rm or Rs in the ITT analysis was a concurrent diagnosis of DSM-atypical depression (Table 4, see Supplementary material for effect size and personality traits).

Discussion
The main findings of our study are the good remission and response rates in patients with TRAD by using dopaminergic treatments as the first line, and the significant added value of their combination and their maintenance in time.We will first compare these results to the literature before discussing their limits.

DATA1 comparison with the literature
In the absence of a control group, absolute values of Rs (and Rm) may be interpreted relative to two references: (i) the number of Rm with other attempts implemented during the nearly half-year lead-in period (23 ± 18 w), i.e., 7% [0-14%] (Figure 2), and (ii) the proportion of Rs when combining stages 3 and 4 of the STAR*D trial, i.e., 22% [18-26%] over an average of 15 weeks (Figure 5, shaded areas).These included a switch to other antidepressants (nortriptyline, mirtazapine, or TCP), augmentation (lithium or triiodothyronine) and combination strategies (venlafaxine + mirtazapine) (38-40).
In the following section, comparisons are performed on fixedeffect meta-analyses of Rs reported in the literature (ITT analysis, i.e., including dropouts, Figure 5).Only randomized controlled studies are considered when discussing MAOI but not when considering D2RAG.

Switching to a MAOI
In non-stratified TRD, our results are better than expected when compared with the seven studies using TCP: 68% [50-85%] vs. 43% [37-50%].This difference can neither be accounted for by differences in severity, resistance, doses, or treatment duration, all being in the same range (41).Hence, there might be an advantage to stratify TRD patients.In the following, TRAD will be compared with two other clinical subgroups: atypical depression and anergic depression.
In the seven studies on atypical depression, the weighted average of Rs to phenelzine was 61% [54-67%] (18, [43][44][45][46][47][48].Two thirds of these patients were chronically depressed, and many may have suffered from dysthymia or double depression regarding the moderate level of severity (QIDS ≈ 11) and resistance (22-48% of TRD).Hence, our results stand up well since the sample had much less favorable characteristics and only half of our patients had atypical features.In CADOT, 83% of the patients received TCP rather than phenelzine and received higher doses, i.e., 50 mg/d vs. 40 mg/d of TCP-eq (59 mg/d of phenelzine).This difference in doses has probably less to do with the nature of the drug than with the naturalistic setting allowing to take the time needed for dose optimization and to maximize tolerance.In CADOT, only 5% of the patients could not achieve the appropriate dose and duration vs. 19% in the randomized control trials on atypical depression.If the proportion of Rs is computed on the number of patients who received an adequate treatment rather than the number of patients at inclusion, i.e., per-protocol rather than ITT analysis, there is no more difference: 71% [65-78%] in atypical depression vs. 69% in our study.
All the four studies in anergic depression are from the Pittsburgh group in which the weighted average of Rs to TCP was 73% [62-84%] (14-16, 19).They mostly recruited inpatients who were more often bipolar (49% vs. 35%), but they had similar symptom severity and were less resistant (<51% vs. 100%).Again, the rigid timeframe of the Pittsburgh's randomized controlled trials accounted for the lower doses and shorter treatment duration before evaluation relative to CADOT (35 vs. 50 mg/d TCP-eq; 6 vs. 15 weeks).

PPX (D2RAG) in TRD
Of the ten reports on the use of PPX in TRD, three were randomized control trials, three included bipolar patients, and only one addressed its use in combination with a (non-MAOI) antidepressant as in DATA1 (42).None questioned the added value of D2RAG in specific TRD sub-types.Our results of PPX in DATA1 are below the range of those in the published studies, i.e., 50% [22-78%] vs. 62% [56-68%] despite the use of (much) higher doses, 3.6 mg/d vs. 1.4 mg/d (26).

Interim conclusion
The average proportion of Rs of 65% after DATA1 supports the added value of considering TRAD as most likely to benefit from a dopaminergic antidepressant strategy.Kaplan-Meier estimator of the remission 'survival' function.The five responders who were not formally remitters according to their QIDS = 6 but who were satisfied with MAOI are included in the pool of remitters (Rm).Vertical tick-marks state patients who were still Rm at their last visit (right-censored survival).The numbers at risk and their proportion relative to the initial sample are indicated below the time abscise.The envelop corresponds to the 95% confidence interval (Gaussian approximation to the binomial distribution).At 6, 12, and 18 months, 95% [88-100%], 86% [73-98%], and 78% [61-95%] of the patients did not relapse, respectively.This compares best with cohorts treated in expert centers (see Supplementary material).

The benefit of DATA2 combination
To the best of our knowledge, this is the first report of the combination of MAOI and D2RAG in the management of TRD (notwithstanding the anergic-anhedonic features) (49).When considering only the group of patients who entered DATA2, the proportion of Rm was unexpectedly high both in the whole sample (n = 16) and when considered in only the subgroup of nRm (n = 13), i.e., 94% [82-100%] and 92% [78-100%], respectively, being fully remitted (QIDS = 2, Table 4, Section 4b).In line with earlier reports on a similar clinical picture (13), it is of interest that all the five ECT-resistant patients who entered DATA2 remitted, suggesting that this might not be of unfavorable prognosis in the case of anergic-anhedonic depression.
However, the results of DATA2 cannot be interpreted on their own.The significant proportion of nRm after DATA1 who refused to enter DATA2 benefited less from the first line of treatment (QIDS = 16) than those who continued (QIDS = 10 -Figure 3).A complete absence of response to a first dopaminergic agent may have been interpreted by the patient and his/her psychiatrist as predictive of a low probability of response to the continuation of this strategy.Lastly, lithium discontinuation in two patients may have contributed to the improvement ('*' Figure 4C).Accordingly, DATA efficacy on TRAD can only be assessed from an ITT perspective.Though somewhat less striking, the increase of Rm from 52 to 77% and of Rs from 65 to 88% remains significant.

Efficiency of treatment maintenance
Even if all patients with a QIDS score ≤6 could not be maintained under the same treatment and the same doses up to 6 months, 95% [88-100%] of them did not relapse.This is much higher than the 6 months follow-up after steps 3 and 4 of the STAR*D trial in which only 38% [22-55%] were still in remission (4).Conversely, our maintenance rate is consistent with the 80% [71-90%] (n = 71/57) of TRD patients followed up in tertiary care (50) (detailed in the Supplementary material).Interestingly, these authors also reported a positive association between maintenance with MAOI and the absence of relapse.

TRAD in the context of French psychiatric care
It is uneasy to ascertain the generalizability of these results out of the context of French psychiatry.In our region, ECT was only available in an inpatient setting and the specificities of French health insurance coverage might have biased our population with TRAD.
In France, bupropion, the only norepinephrine-dopamine reuptake inhibitor available, is not reimbursed for depression.Yet its dopaminergic valence could make it particularly beneficial in cases of anergicanhedonic depression (51).As bupropion is hardly ever used for depression in France, it may be that TRAD are not only more frequent, but also less resistant to dopaminergic antidepressant strategies relative to other countries in which bupropion is a second line treatment, alone or in combination (52)(53)(54).In CADOT, only five patients accepted to support the cost of bupropion and tried it in combination with the SNRI duloxetine before trying DATA; three of them remitted with the association, so that only two entered DATA -of whom all were Rm.Furthermore, since structured psychotherapy is not reimbursed too, only two of our patients had an adequate trial of cognitivebehavioral therapy before trying DATA.This may have been particularly beneficial to patients with atypical features (48) and may account for the poor prognostic value of this characteristic in our ITT analysis (see Table 4, Section 4c, and the Supplementary material for personality traits).Moreover, the most effective psychotherapy for chronic depression, CBASP (Cognitive Behavioral Analysis System of Psychotherapy), is not available in France (55).

Limitations related to the prospective observational design
Regarding DATA1, it is impossible to conclude a superiority of MAOI or D2RAG as the first line treatment (as it was non-significant whether switchers were or were not included).Samples were not randomized because the decision was up to the patient, although the psychiatrist might have also played a role.For example, the difference in resistance between the MAOI and D2RAG groups (Table 4, Section 4a) likely reflects the prescriber's belief in the higher chance of success with the former.
Overall, as enthusiastic as they may seem, the results of DATA1-2 benefited from the high motivation of patients and their caregivers and the flexibility of the prescribing context.Considering that the treatment would have to be maintained after remission, prescribers were particularly careful to optimize tolerance from the very start.Depending on the side effects and the patient's preference, either the dose escalation was adapted or the molecule was switched to another from the same class.Only quaternary care facilities permit practitioners to devote the necessary time to do so, which may account for the low proportion of patients who could not have at least one adequate trial (n = 1, 2% [0-6%]).
Moreover, the results may have been favorably influenced by the flexible evaluation schedule and the long lead-in period (a median of 4 months).On the one hand, this allowed the establishment of the relationship of trust necessary for patients to accept riskier drugs and ascertained the resistance of the depression (when other lines of treatment were tried).On the other hand, it may have increased expectations and the placebo effect, not to mention that our belief in DATA for TRAD may sometimes have led us to convince the patient to pursue the trial where neutral prescribers would not have insisted.
Finally, the analysis of the whole TRAD cohort (see Supplementary material for an example) suggested some improvements to the description and the model of mesencephalic hypodopaminergia (Foucher et al., in preparation).The refinement of this dopamine-sensitive anergic-anhedonic syndrome could benefit from brain imaging biomarkers which would facilitate the implementation of multicenter randomized control trials.Without them, DATA-like guidelines are not recommended before traditional resources are exhausted and out of the hands of trained prescribers.

Conclusion
These results are not the first to suggest the existence of an anergic-anhedonic depressive syndrome distinct by its sensitivity to MAOI.By extending this sensitivity to D2RAG, they lend support to the hypothesis of mesencephalic hypodopaminergia as pathophysiological substratum.It is all the more timely to establish the existence of such an anergic-anhedonic syndrome and its relationship with 'positive valence systems' , as new pro-dopaminergic interventions are emerging, e.g., triple reuptake inhibitors (56) and deep brain stimulation of the medial forebrain bundle (57).

FIGURE 5
FIGURE 5Comparison of the proportion of responders (Rs) in CADOT with the literature.The first reference (gray vertical band) is our lead-in period of 24 weeks during which four patients remitted with other treatment than DATA (but three remitted with bupropion + SSRI, i.e., a dopaminergic strategy).The second is the proportion of Rs when combining stages 3 and 4 of the STAR*D trial (38-40).Regarding monoamine oxidase inhibitors (MAOI), the results in non-stratified treatment-resistant depression (TRD) were extracted from(41).MAOI in atypical depression: weighted average of Rs from an intention-to-treat perspective (see references in text).Proportion of TRD and bipolar depressions (BPD) are indicated on the right.MAOI in anergic depression: weighted average of Rs from an intention-to-treat perspective.All studies are from the Pittsburgh group (see references in text).The single dopamine D2 receptor agonist (D2RAG) on which we have sufficient data is pramipexol (PPX).The proportion of Rs with PPX in TRD is recomputed from 10 studies reviewed in reference (42).

TABLE 1
Diagnostic criteria of anergic-anhedonic depression used in the study.

TABLE 2
(22)lations' baseline characteristics.SNRI, serotonin and norepinephrine reuptake inhibitors; rTMS, repetitive transcranial magnetic stimulation; ECT, electro-convulsive therapy.Means are expressed as "mean (±standard deviation)"; medians are expressed as "median (1st-3rd quartile)."*Accordingto DSM-5 criteria (36); ¤Corresponds to the duration of the episode at the time of the first contact with Strasbourg's expert center on TRD plus the lead-in period, i.e., until the first DATA1 treatment.†Theconcepts of mixed or incomplete states and mood reactivity or fluctuation are defined according to Karl Leonhard(22).For sections 2b and 2c, the p-values of group-comparisons are in bold when the difference is significant (α = 0.05, bilateral, uncorrected for multiple testing), or in gray italic when only a trend (threshold = 0.1, bilateral, uncorrected for multiple testing).Effect sizes are provided in the Supplementary material.

TABLE 4
Predictors of Rm and Rs with DATA.