Disentangling pharmacological and expectation effects in antidepressant discontinuation among patients with fully remitted major depressive disorder: study protocol of a randomized, open-hidden discontinuation trial

The primary aim of our trial is to disentangle how pharmacological and expectation effects contribute to discontinuation symptom load in antidepressant discontinuation. We hypothesize that treatment (discontinuation vs. continuation) and treatment expectation (high vs. moderate) interact in modulating discontinuation symptom load among patients. If we find a significant interaction effect of treatment and treatment expectation, we expect that i) patients who remain on their antidepressant will show a higher discontinuation symptom load with moderate than with high expectation, ii) patients with moderate treatment expectation will show a higher discontinuation symptom load if the antidepressant is discontinued versus continued, and iii) patients who discontinue their antidepressant will show a higher discontinuation symptom load with high than with moderate treatment expectation. We assume that the relationship between treatment expectation and discontinuation symptom load will vary according to stress ratings, antidepressant-related side effects, prior discontinuation experience, neuroticism, anxiety, somatosensory amplification, and illness framework.

This prospective, randomized, parallel-group, partly blinded, open-hidden discontinuation trial with a 2 × 2-factorial design will systematically vary the factors Treatment and Expectation to investigate the interaction of pharmacological and expectation effects in modulating discontinuation symptom load. 196 patients with fully remitted MDD and an indication to discontinue antidepressant use will be randomly allocated to open discontinuation (OD), hidden discontinuation (HD), open continuation (OC), or hidden continuation (HC) of their antidepressant (Fig. 1). Our trial is part of a collaborative research center (CRC; TRR 289 Treatment Expectation: treatment-expectation.de/en/).

A block randomization with varied block sizes will be performed externally by the Medical Biometry department of the University Medical Center Hamburg-Eppendorf, Hamburg, Germany. Block size will be unknown to the study staff. Participants will be allocated 1:1:1:1 to OD, HD, OC, and HC groups. Randomization will be stratified by duration of antidepressant use, with 24 months as a marker for long-term use (< 24 months vs. > 24 months, 3:7), and antidepressant-associated risk of developing discontinuation symptoms (moderate [citalopram, escitalopram, sertraline, duloxetine] vs. higher or unknown risk [paroxetine, venlafaxine, mirtazapine], 1:1).

Digital randomization lists will contain allocation sequences and will be passed on to unblinded randomization officers who have no personal contact with participants to assign participants to interventions. In case of assignment to one of the double-blinded hidden trial arms, both participants and study staff will be blinded to group assignment. Participants in the hidden trial arms, their prescribing physicians, and study staff will be debriefed 13 weeks post-baseline at t9. Staff responsible for data analysis will be blinded to group assignment. Digital and paper key lists containing pseudonym, name, contact details, experimental group, medication, and current dosage will be prepared to ensure emergency unblinding (e.g., in case of hospitalization).

Participants will be recruited from the psychiatric outpatient clinics of the University Medical Center Hamburg-Eppendorf, Hamburg, Germany. The study will be advertised via support groups, psychiatric practices, general practitioners, and pharmacies. Digital and analogue methods such as media and newspaper articles, advertisement in public transport, social media, online forums, and leaflets will be used. Interested patients can contact the study team for further study information and a first screening telephone interview (S1). Potential participants will be invited for an on-site screening interview (S2), consisting of an in-depth clinical interview to assess eligibility. Indication for discontinuation will be assessed in accordance with German guideline recommendations [26] and in consultation with the prescribing physician. Detailed eligibility criteria are summarized in Table 1.

The target sample size of N = 168 (n = 42 in each experimental group) is based on medium to large effect sizes for expectation effects on depressive symptoms and treatment outcome [47, 50] and hypothesized small to medium interaction effects with pharmacological modulations. A medium-sized effect (f = 0.22) for the interaction of expectation and pharmacological modulation was assumed in an ANOVA (with main effects and interactions) with a between-within interaction using a power of 80% and two-sided significance level of 0.05. Including oversampling with an anticipated dropout rate of 15%, our recruitment target is N = 196 (with n = 49 per group). Sample size calculation was conducted using G*Power (version 3.1.9.2; [57]).

Assessments will be conducted according to pre-defined standard operating procedures (SOPs) at the Institute of Systems Neuroscience, University-Medical Center Hamburg-Eppendorf, Hamburg, Germany. Figure 2 shows participant flow through the trial. A study physician or psychologist will ask eligible patients to give written informed consent for study participation, blood analysis, optional saliva sampling, and optional rsfMRI assessment. A release from the medical confidentiality obligation of the prescribing physician will be obtained. Prescribing physicians will be asked to confirm absence of objection against discontinuation, MDD as primary diagnosis for initial antidepressant prescription, current type and dosage of antidepressant, duration of antidepressant use, medication switches, and frequency of visits. Participants will then be randomized, invited for baseline assessment t0, and informed about group allocation.

Assessments during the 13-week experimental phase, i.e., weekly from t0 to t5 and biweekly from t5 to t9, will be interview- and questionnaire-based. Each study visit begins with a clinical interview to assess depressive symptoms, adverse events (AEs), adherence, and clinical impression. Participants will then be asked to fill digital questionnaire batteries programmed with LimeSurvey [58]. A study physician or psychologist will examine safety-relevant data and, if necessary, discuss these data immediately with the participant. Otherwise, the study visit will end with handing out the study medication. Saliva samples will be taken between t0 and t1, blood samples at t1 and t9, and rsfMRI measurements at t0 or t1. At t9, an emotional interference paradigm will be administered.

Following the 13-week experimental phase, participants will enter a 39-week clinical observation phase. The clinical observation phase consists of phone or online calls and online-questionnaires at FU1-3, i.e., 26, 39, and 52 weeks post-baseline. At FU3, participants who continued antidepressant use will be offered an individualized discontinuation plan. All participants will receive a reimbursement of €150 for full participation in the study, or pro rata reimbursement in case of incomplete participation. There will be two payouts of €75, one following completion of the experimental phase and another following completion of the trial. In case of participation in rsfMRI measurement, participants will receive an additional €30.

Our primary outcome discontinuation symptom load over the course of the experimental phase will be assessed with the Discontinuation Emergent Signs and Symptoms Scale (DESS; [59]). The DESS is a self-report measure incorporating 43 symptoms of antidepressant discontinuation. Participants will rate intensity of each discontinuation symptom on a 4-point response-scale (0 ‘not present’, 1 ‘mild’, 2 ‘moderate’, 3 ‘severe’; [60]). Sum scores range between 0–129. Higher scores indicate more pronounced discontinuation symptoms. Discontinuation symptom load will be measured as area under the curve (AUC) based on assessments from t2 until t9, adjusted for baseline symptoms reported at t1. Table 2 provides an overview of assessments according to SPIRIT guidelines [61].

Secondary outcomes will include i) discontinuation symptom load over the clinical observation period, ii) recurrence over the experimental period, iii) recurrence over the course of the complete trial evaluated in a time-to-event analysis, differences in iv) stress and v) state anxiety from baseline to end of the experimental phase, and vi) attentional and emotional processing at the end of the experimental phase. Discontinuation symptom load over the clinical observation period will be based on DESS assessments from t9 until FU3 [59] to cover reported symptoms between week 13 until week 52 and measured as AUC. Both recurrence over the experimental period (t1-t9) and recurrence over the course of the trial (t1-FU3, evaluated in a time-to-event analysis) will be monitored via expert-rating and self-report measure. Recurrence will be defined as appearance of a new depressive episode after full, sustained remission of depressive symptoms. Study physicians and psychologists will use the Montgomery-Asberg Depression Rating Scale (MADRS) to assess depressive symptom severity by ten items with seven intensity ratings (0–6) each [62, 63]. Sum scores range between 0–60. Higher scores indicate more pronounced depressive symptoms. To standardize assessments, a German translation of the Structured Interview Guide for the Montgomery-Asberg Depression Rating Scale (SIGMA) will be used [64]. Participants will assess depressive symptoms with the Beck Depression Inventory (BDI-II; [65, 66]), a self-report measure that includes 21 items with 4 response options (0–3). Sum scores range between 0–63. Higher scores indicate more pronounced depressive symptoms. If recurrence is suspected, as indicated by BDI-II score > 19 or MADRS score > 21 over two consecutive study visits, corresponding sections of the expert-rated Structured Clinical Interview for DSM-5 – Clinician Version (SCID-5-CV; [56]) will be conducted to (dis-)confirm recurrence. In case of recurrence, beginning of the new depressive episode will be examined and recorded as calendar week. The Perceived Stress Scale (PSS-10; [67, 68]) will be used to assess stress. The PSS-10 includes ten items with five response options (0 ‘none of the time’ to 4 ‘very often’). Sum scores range between 0–40. Difference scores will be calculated by subtracting scores at t1 from scores at t9 and range between -40 to 40. Higher scores indicate increased stress. The State-Trait-Anxiety-Depression-Inventory (STADI; [69, 70]) will be used to assess state anxiety. Two anxiety-related state subscales include five statements with four response options (1–4) each. State anxiety scores base on sum scores of both state anxiety subscales and range between 10–40. Difference scores will be calculated by subtracting scores at t1 from scores at t9, resulting in a range between -30 to 30. Higher scores indicate increased state anxiety.

We will use a modified emotional Posner task to assess attentional and emotional processing at t9. The Posner task manipulates attentional resources and provokes emotional responses using facial stimuli, activating limbic, prefrontal, and visuo-spatial brain circuits [71]. In short, participants respond as fast as possible to a dot target by button pressing, while neutral, happy, sad, or fearful faces are presented as distractors. Targets are preceded by either spatially-directing cues leading to covert shifts in the attentional focus (i.e., low attentional resources to process distractors) or non-spatial cues, leaving the attentional focus on the faces. We will measure reaction times in milliseconds (ms) for each condition and calculate difference scores for reaction times to happy - neutral faces, sad - neutral faces, and fearful - neutral faces under high attention to faces.

Stress, antidepressant-related side effects, prior discontinuation experience, neuroticism, trait anxiety, somatosensory amplification, and illness framework will be explored as modulators. Stress will be rated with the PSS-10 (see secondary outcomes). Antidepressant-related side effects will be measured with the Generic Assessment of Side Effects Scale (GASE, [72]). The GASE is a self-report measure including 36 symptom descriptions with 4 severity ratings (0 ‘not present’, 1 ‘mild’, 2 ‘moderate’, 3 ‘severe’). For every reported symptom, participants will indicate severity and whether they perceive this side effect as antidepressant-related. Sum scores of antidepressant-related symptom ratings range between 0–108. Higher scores indicate more pronounced antidepressant-related side effects. Prior discontinuation experience will be assessed with a modified version of the Generic Rating Scale for Treatment Pre-Experiences (GEEE_PRE, [73]). Participants will indicate improvement (0 ‘no improvement’—10 ‘greatest improvement imaginable’) and worsening (0 ‘no worsening’—10 ‘greatest worsening imaginable’) of general condition attributed to the most recent discontinuation attempt on numeric rating scales (NRS). Difference scores will be calculated by subtracting worsening scores from improvement scores, resulting in a range between -10 to 10. Lower scores indicate a more negative discontinuation experience. Participants with no prior discontinuation experience will be assigned a value of 0. Neuroticism will be assessed with the Emotional Stability subscale of the Brief Big Five Inventory (BFI-10; [74]). The Emotional Stability subscale includes two items rated on a five-point scale (1 ‘disagree strongly’—5 ‘agree strongly’). Sum scores range between 2–10, with higher scores indicating higher neuroticism. The State-Trait-Anxiety-Depression-Inventory (STADI; 69, 70) will be used to assess trait anxiety. Two anxiety-related trait subscales include 5 statements with four response options (1–4) each. Trait anxiety scores base on sum scores of both trait anxiety subscales and range between 10–40. Higher scores indicate higher trait anxiety. The Somatosensory Amplification Scale (SSAS; [75, 76]) will be used to assess the amount of somatosensory amplification. Ten items will be rated on a five-point scale (1 ‘not at all true’—5 ‘extremely true’). Sum scores range from 10–50, with higher scores indicating higher symptom amplification. Finally, participants will be asked about their illness rationale. A single item will be used to assess whether participants perceive MDD as a more biologically or psychologically caused disorder (0 ‘biologically caused’—10 ‘psychologically caused’). A higher score indicates an inclination towards psychological causes underlying MDD.

During the experimental phase, treatment adherence to study medication will be assessed in two ways. First, participants will indicate the number of days the study medication was taken since the last study visit. Second, study staff will assess treatment adherence via blood analysis at t1 and t9. Blood samples will be collected using white tubes without gel and will be analyzed with Liquid Chromatography Mass Spectrometry (LC–MS/MS) in accordance with EU guidelines as stated in the In-vitro-Diagnostic Device Regulation (IVDR; [77]). Storage and analysis will be conducted at the Department of Legal Medicine, University Medical Center Hamburg Eppendorf, Hamburg, Germany. During the clinical observation phase, participants will indicate whether they remained on or off antidepressants. In case of non-adherence, possible reasons will be surveyed.

Optional saliva samples will be taken between t0 and t1 to assess cortisol awakening response and salivary alpha-amylase activity as objective stress markers. Salivary analyses will be conducted in the laboratory of the Institute of Medical Psychology and Behavioral Immunobiology, University Hospital Essen, Essen, Germany, as part of central scientific project Z02 within our CRC 289 Treatment Expectation. Brain imaging data on functional and structural connectivity will be collected on-site. Data will be contributed to central scientific project Z03 within our CRC 289 Treatment Expectation. Standardized MR protocols will be provided by Z03, who will use the data for pooled and meta-analytic analyses. Furthermore, we will include assessments of past experiences with antidepressant use and discontinuation, well-being, and substance use. As part of the CRC standard battery, we will measure treatment expectations, personality traits, state and trait anxiety and depression, behavioral approach and avoidance tendencies, psychopathology, subjective impairment, perceived warmth and competence of both study physician/psychologist and participant, suspicions about treatment, and extensive demographic and medical characteristics.

Safety endpoints comprise depressive symptoms (BDI-II, MADRS) including inspection of suicidality and recurrence (SCID-5 CV; see above), current treatment effects, AEs including burdensome life events, and clinical impression according to psychopathological findings. A modified version of the Generic Rating Scale for Treatment Effects (GEEE_ACT; 73) will be used to measure self-reported current treatment effects. Participants indicate worsening of their condition and treatment side effects (0 ‘no worsening’—10 ‘greatest worsening imaginable’, and 0 ‘no complaints’—10 ‘greatest complaints imaginable’, respectively) on an NRS with eleven response options (0–10) each. A score ≥ 8 on either item indicates severe distress. Occurrence of AEs will be assessed via interview, followed by an expert-rating of intensity and causal relation to study treatment. AEs will be graded with regard to i) intensity according to Common Terminology Criteria for Adverse Events (Grade 1 ‘mild’, Grade 2 ‘moderate’, Grade 3 ‘severe’, Grade 4 ‘life threatening or disabling’, Grade 5 ‘fatal resulting in death’; CTCAE; [78]), and ii) causality according to World Health Organization—The Uppsala Monitoring Centre system (1 ‘certain’, 2 ‘probable/likely’, 3 ‘possible’, 4 ‘unlikely’, 5 ‘conditional/unclassified’, 6 ‘unassessable/unclassifiable’; WHO-UMC; [79]).

Analyses will be conducted per endpoint, i.e., as soon as all data of the according endpoint are available. A statistical analysis plan (SAP) will be established in collaboration with the Medical Biometry department of the University Medical Center Hamburg-Eppendorf, Hamburg, Germany, and published on ClinicalTrials.gov prior to analyses.

Analyses of the primary endpoint and of secondary endpoint discontinuation symptom load over the clinical observation phase will be based on the intention-to-treat population. For all other secondary endpoints, the full analysis set will be used. No interim analyses are planned.

Confirmatory testing of the interaction of treatment and treatment expectation in modulating discontinuation symptom load will be performed using a two-way ANCOVA. The model will contain the nominally scaled between-subject factors treatment (discontinuation vs. continuation) and treatment expectation (high vs. moderate), their interaction, the interval-scaled dependent variable discontinuation symptom load (measured as AUC) and, as covariates, the two binary stratification variables long-term use (yes vs. no) and risk of developing discontinuation symptoms (high vs. moderate) as well as discontinuation symptoms at baseline. Missing data of the primary outcome will be imputed using linear interpolation or, in case of individual termination of study treatment or rescue medication, as last observation carried forward.

If analyses reveal a significant interaction effect of treatment and treatment expectation, exploratory post-hoc comparisons will be made between experimental groups: A post-hoc comparison between the two continuation groups is planned to test whether participants show higher discontinuation symptom load with moderate (HC) than with high (OC) treatment expectations. Hence, conclusions could be drawn about a nocebo-induced effect of treatment expectation on discontinuation symptom load. To examine the extent to which discontinuation symptom load may be influenced by pharmacological factors alone, a post-hoc comparison will be performed between the moderate treatment expectation groups who discontinued (HD) versus remained on antidepressants (HC). A post-hoc comparison between participants who discontinued antidepressants will examine whether high treatment expectation (OD) is associated with higher discontinuation symptom load than moderate treatment expectation (HD).

We assume that the relationship between treatment expectation and discontinuation symptom load will vary according to stress ratings, antidepressant-related side effects, prior discontinuation experience, neuroticism, anxiety, somatosensory amplification, and illness framework. In case of a significant interaction within the primary analysis, individual moderators will be included as 3^rd factor in the model and three-way interactions will be analyzed.

Secondary endpoints will be analyzed as follows: Interactions of pharmacological and expectation effects in modulating i) discontinuation symptom load over the clinical observation phase measured as AUC, ii) stress, and iii) state anxiety will be examined following the model of our primary analysis. A binary logistic model will be used to predict iv) recurrence during the 13-week experimental period. A Cox proportional hazards model will be used to examine v) recurrence over the course of the complete trial. Finally, we will examine the interaction of expectation and emotion in modulating vi) attentional and emotional processing in a linear mixed model. This model will include a bias score (emotional versus neutral faces, measured in ms) as dependent variable and the factors expectation concerning occurrence of discontinuation symptoms (high (OD) vs. moderate (HC, HD) vs. none (OC)) and emotion (happy vs. sad vs. fearful), their 3 × 3 interaction, the stratification variables as covariates and random intercept for the individual patients. Analysis of secondary and safety endpoints and modulating factors will be explorative. Further specifications on statistical analyses can be found in the SAP.

Concerning analysis of the blood samples, we will check adherence to study medication by individually comparing antidepressant blood serum levels at t1 and t9. We will assess whether antidepressant blood serum levels lie within therapeutic range at t1. Finally, we will exploratively examine the relationship between initial antidepressant blood serum levels and discontinuation symptom load reported by the two discontinuation groups using simple correlation analyses.

Safety endpoints will be assessed and documented at every measurement point. Safety assessments and consequent actions follow a pre-defined, step-wise procedure (Fig. 3). Criteria for individual withdrawal from study treatment will be reviewed during each study visit and, in case of monitoring, during each additional visit. Treatment will be terminated in case of participant’s withdrawal of his/her informed consent, pregnancy, medical or psychological objections by the study physician or psychologist, or insufficient compliance regarding requirements for study participation including non-adherence. In these cases, the study physician informs about further treatment options and participants will be invited to continue participation in all measurement points. If a participant falls ill during the discontinuation process, a physician may decide to extend the discontinuation process by the duration of the illness (max. 4 weeks). If so, the participant will continue to receive the currently administered dose of medication from the study team. Measurements will be postponed accordingly. No additional interim measurements will take place. The continuation of the entire study will be questioned by the principal investigator in case of medical or psychological concerns/reoccurring AEs with possible causal relation to the study treatment regimen, insufficient study activity (e.g. enrolment rate < 5 per year), or unforeseeable complications that do not justify study continuation.

This trial was approved by the ethics committee of the Hamburg Medical Chamber (reference number: PV7151, 16.12.2019). Research will be performed in accordance with the Declaration of Helsinki. All participants will provide written informed consent to participate. This trial has been registered at ClinicalTrials.gov (NCT05191277).