Ethical practices

An independent review board (United States [US]) or ethics committee (ex-US sites) at each study site approved the study protocol and its amendments. The study was conducted in accordance with ethical principles that originated in the Declaration of Helsinki, current guidelines on Good Clinical Practices, and applicable regulatory and country-specific requirements. All individuals voluntarily provided written informed consent before participating in the study.

Study design

This phase 2, multicenter (28 sites in the US, 11 in Russia, 5 in the United Kingdom, 6 in Ukraine, 2 in Germany, 1 in Moldova), randomized, double-blind, placebo-controlled study (clinicaltrials.gov: NCT03559192) was conducted between July 2018 and May 2020. The study consisted of a 5-week screening phase and an 11-week double-blind treatment phase, the latter consisting of 3 periods: (1) a double-blind placebo lead-in period of variable duration (up to 3 weeks); (2) a 6-week double-blind treatment period; and (3) a withdrawal period during which (only) participants who completed double-blind treatment received placebo for the remaining time of the treatment phase. Investigators and participants were informed that the variable duration of treatment with placebo during the lead-in period could last from 0 to 3 weeks, and that participants could be randomly assigned to treatment with either aticaprant or continued placebo during the 6-week double-blind period.

Study population

Study participants were between 18 and 64 years of age with a diagnosis of MDD per DSM−5 criteria [27], without psychosis; the current episode was to be 18 months or shorter in duration. Eligible participants had been treated with a selective serotonin reuptake inhibitor (SSRI) or serotonin-norepinephrine reuptake inhibitor (SNRI), at an adequate dosage for at least 6 continuous weeks, but no more than 12 months for the current depressive episode of moderate-to-severe severity, and had inadequate response documented at screening (i.e., based on MADRS [26] total score ≥25).

Individuals who had failed (≤25% improvement based on the Massachusetts General Hospital Antidepressant Treatment Response Questionnaire [MGH-ATRQ] [28]) three or more antidepressants, including the current SSRI/SNRI, during the current episode of depression, despite adequate dose and duration (≥6 weeks), were not eligible for study participation. The study excluded individuals with potentially confounding psychiatric and general medical comorbidities. All inclusion and exclusion criteria are presented in the Supplement (Table S1).

Randomization and double-blind study drug

Eligible participants were randomly assigned (1:1), based on a computer-generated randomization schedule, to receive double-blind treatment with either 10 mg aticaprant (JNJ−67953964) or to continue matching placebo, each once daily (in fasting condition, before breakfast) for 6 weeks during the double-blind treatment period. Randomization was balanced by using randomly permuted blocks of 4, stratified by placebo lead-in response status and SHAPS score (≥20 versus <20). Treatment codes were assigned by a centralized interactive web response system. Participants continued taking the same SSRI/SNRI antidepressant/dose that they had received prior to study entry, with no changes permitted during the study. The use of quetiapine (≤100 mg) was allowed, primarily as a sedative, when it had been used in a stable dose for ≥8 weeks prior to screening and was continued unchanged during the study. Participants, investigators/site personnel, those assessing outcomes, and those analyzing the data were blinded to treatment group assignment.

Adherence was tracked by the site, with study drug dispensing and return recorded in the electronic case report form. Adherence to study drug was also monitored using a smart phone app that recorded study drug ingestion.

Efficacy assessments

Severity of depression was assessed by site-based, trained, certified, and blinded raters using the MADRS [26]. Severity of depressive illness was also assessed by investigators using the Clinical Global Impression–Severity (CGI-S; rated on a 7-point scale from 1 [normal – not at all ill] to 7 [among the most extremely ill patients]) [29].

Participants rated their hedonic capacity using the SHAPS, a reliable and validated 14-item instrument developed for use in MDD (score range 14–56; rating according Franken et al. [30]), with higher score indicating greater severity of anhedonia) [30, 31].

Investigators assessed anxiety using the Structured Interview Guide for the Hamilton Anxiety scale (SIGH-A; comprised of 14 items, each scored from 0 [not present] to 4 [maximum degree]) [32, 33] and the 6-item Hamilton Anxiety Subscale (HAM-A6; comprised of 5 psychic anxiety items [anxious mood, psychic tension, fears, intellectual disturbances, and anxious behavior] and 1 somatic item [muscular tension], each scored from 0 to 4) [34].

To assess the effect of study drug on aspects of cognitive and executive function, participants were asked to complete the Massachusetts General Hospital Cognitive and Physical Function Questionnaire (CPFQ) [35]. The CPFQ includes 7 questions about attention, energy, memory, mental acuity, and motivation, scored on a 6-point Likert scale, with higher values indicating worse function.

Samples were collected for exploratory plasma biomarkers and salivary cortisol; results will be reported elsewhere.

Safety assessments

Adverse events and other standard safety assessments (i.e., hematology, serum chemistry, urinalysis, physical and neurological examination, vital signs, electrocardiogram [ECG]) were monitored throughout the study. Investigators classified adverse events as mild (i.e., easily tolerated, caused minimal discomfort and did not interfere with everyday activities), moderate (sufficient discomfort to cause interference with normal activity), or severe (i.e., caused extreme distress, significant impairment of functioning or incapacitation, and prevented normal everyday activities).

At each study visit, investigators administered the Columbia Suicide Severity Rating Scale (C-SSRS) questionnaire [36] to solicit the occurrence, severity, and frequency of suicide-related ideation and behaviors.

Statistical methods

All randomized participants who received ≥1 dose of study drug during the double-blind treatment period were included in the safety analysis dataset, of whom those having ≥1 post-baseline MADRS assessment during the double-blind treatment period were included in a ‘full’ intent-to-treat (fITT) analysis dataset. Efficacy data were also analyzed for an “enriched” intent-to-treat (eITT) analysis dataset, which included all participants in the fITT analysis dataset who were non-responders at the end of the placebo lead-in period (i.e., <30% improvement in MADRS total score from screening or entry baseline). Investigators were blind to the definition of the improvement threshold. The 30% improvement threshold for detection of possible placebo response was selected based partly on a previous trial of ALKS−5461 (a combination of buprenorphine, a partial mu-opioid receptor agonist and KOR antagonist, and samidorphan, a potent mu-opioid receptor antagonist) [37], which used 50% improvement after 4 weeks of placebo (the first stage of their Sequential Parallel Comparison Design), and partly on our own adjustment for the shorter duration of the placebo lead-in, which was informed by a comparison of the prevalence of 30% and 50% improvement in participants assigned to placebo in previous Johnson & Johnson-sponsored placebo-controlled studies of other candidate antidepressants in MDD. The goal was to balance the response definition, while adjusting for the duration of the lead-in phase, and allowing for a typical attrition rate. The eITT analysis dataset was prespecified as the primary efficacy analysis dataset.

Statistical analyses were conducted using SAS, version 9.4. Analyses of efficacy endpoints were performed at a significance level of 0.20 (primary endpoint and other endpoints related to MADRS, 1-sided; secondary endpoints and other endpoints not pertaining to MADRS, 2-sided). Adjustment for multiple comparisons was not performed.

Efficacy endpoints and statistical analyses

The primary efficacy endpoint – change from treatment baseline to treatment week 6 in MADRS total score – was analyzed using a mixed-effects model using repeated measures (MMRM). The model included baseline MADRS score as a covariate; treatment (aticaprant, placebo), country, time, and time-by-treatment interaction as fixed effects; and a random patient effect.

In a pre-specified subgroup analysis, the impact of baseline anhedonia level (above versus below the baseline median SHAPS score) on the primary endpoint was summarized descriptively.

The overall differences between treatment groups based on the proportion of responders (defined by ≥30% and by ≥50% improvement from treatment baseline MADRS total score) and the proportion of participants in remission (MADRS ≤ 10) at the end of the 6-week double-blind treatment period were analyzed using Chi-square tests.

MADRS items most closely reflecting anhedonia symptoms (i.e., apparent sadness, reported sadness, concentration difficulties, inability to feel, and lassitude, referred to as the MADRS 5-item anhedonia factor subscale [38,39,40]; total score range 0–30) were examined post hoc according to the MMRM model described above for the primary efficacy endpoint analysis, but using the baseline MADRS 5-item anhedonia factor subscale score as covariate. Change in the MADRS 5-item anhedonia factor subscale score was also analyzed by baseline anhedonia level.

The same MMRM model, with respective baseline score as covariate, was also used to compare the treatment groups based on SHAPS score, SIGH-A total score, and HAM-A6 subscale score.

Analysis of safety endpoints

Treatment-emergent adverse events and other measures of safety were summarized descriptively for each treatment group.

Sample size determination

The sample size planned for this study was calculated assuming a treatment effect size of 0.45 at treatment week 6 for mean change from baseline in MADRS total score between aticaprant and placebo. The assumed effect size and an estimated standard deviation (SD) of 11 were derived from clinical trials of ALKS−5461 as adjunctive treatment in patients with MDD who had inadequate response to one or two courses of antidepressants [37, 41]. Based on an overall 1-sided significance level of 0.2 and SD of 11, randomization of about 90 individuals – 96 when adjusted for an anticipated 5% drop-out rate during the treatment period – was required to achieve 90% power. After adjusting for an estimated placebo response rate of 25% and drop-out rate of 10% during the placebo lead-in period, 142 participants were to enter the placebo lead-in period.

In accordance with the protocol, a blinded sample size re-estimation was conducted due to a higher-than-anticipated lead-in placebo response of 26.6% (Table S2), resulting in the inclusion of 181 participants.

The choice of alpha and beta (1-power) for this phase 2 study was intended to increase sensitivity for detecting a therapeutic signal while also maintaining a modest sample size. Thus, for the purpose of sample size estimation the power was set to a high value (power = 90%; beta = 0.10) but the type 1 error rate was specified at 1-sided alpha=0.20, as proposed by Lindborg and co-authors [42].

Consistent with the study design, the results for the analyses based on MADRS are characterized by 1-sided upper 80% CI, and those based on SHAPS by 2-sided 80% CI.