Eveleigh 2018.

Study characteristics
Methods	Design: cluster‐randomised controlled trial Prerandomisation phase: no Duration post randomisation: 52 weeks Unit of clustering: general practice Aim: to assess the effectiveness of a tailored recommendation to withdraw antidepressant treatment
Participants	Country: Netherlands Setting: outpatients from 45 general practices Type of AD: SSRI, SNRI, TCA, or other antidepressant except MAOI Duration of antidepressant treatment prerandomisation: ≥ 9 months; median duration, years (range): 8.0 (1 to 48) discontinuation, 9.5 (1 to 56) usual care Duration of antidepressant treatment post stabilisation: not described Total numbers of randomised participants: 146 (70 discontinuation, 76 usual care) Primary diagnosis ‐ discontinuation: any lifetime psychiatric diagnosis 76%, depression 56%, panic disorder or agoraphobia 19%, generalised anxiety disorder 31%, social phobia 23% ‐ usual care: any lifetime psychiatric diagnosis 63%, depression 46%, panic disorder or agoraphobia 17%, generalised anxiety disorder 17%, social phobia 26% Median duration of antidepressant use at inclusion, years (range): 8.0 (1 to 48) discontinuation; 9.5 (1 to 56) usual care Gender distribution (male): discontinuation 71%, usual care 68% Mean age, years (SD): discontinuation 56 (12.9), usual care 56 (14.3) Severity of depressive symptoms at randomisation: not described Inclusion criteria ‐ long‐term antidepressant use (≥ 9 months); all antidepressants were included, except MAOI ‐ written informed consent Definition of remission: no indication for long‐term antidepressant treatment in line with Dutch guidelines Exclusion criteria ‐ current treatment in a psychiatric inpatient or outpatient clinic ‐ appropriate use of long‐term antidepressants according to the Dutch guidelines for depressive and anxiety disorders (i.e. a history of recurrent depression (≥ 3 episodes) and/or a recurrent psychiatric disorder with ≥ 2 relapses after antidepressant discontinuation) ‐ history of psychosis, bipolar disorder, or obsessive‐compulsive disorder ‐ current diagnosis of substance use disorder, excluding tobacco, because of the necessity for specialised treatment ‐ non‐psychiatric indication for long‐term antidepressant usage (e.g. neuropathic pain) ‐ hearing impairment and/or insufficient understanding of the Dutch language Age was not an exclusion criterion
Interventions	Intervention 1: a patient‐specific letter was sent to the GP with the recommendation to discontinue the antidepressant without placebo. Information on antidepressant tapering and the withdrawal syndrome was provided. The GP invited the patient to discuss the recommendations. No treatment restrictions were imposed in case of a relapse or onset of a new psychiatric disorder after discontinuation. A return slip was included to ascertain the patient’s intention to comply with the recommendation Tapering scheme: a gradual tapering programme was recommended Intervention 2: GPs were unaware which patients participated in this study and continued usual care: no restrictions on GPs to deliver care or to refer to specialised mental health care, including continuation or discontinuation of psychotropic drugs Comparison: discontinuation with psychosocial interventions vs continuation of antidepressant (or usual care)
Outcomes	Primary outcome ‐ proportion of participants who successfully discontinued their long‐term antidepressant use after 1 year defined as no antidepressant use during preceding 6 months and absence of a depressive or anxiety disorder during 1‐year follow‐up, as assessed by the CIDI Secondary outcomes ‐ severity of global distress and global psychopathology and depressive symptoms, assessed by the Brief Symptom Inventory (BSI‐53) sum score and the Centre for Epidemiological Studies Depression Scale (CESD) at 3, 6, 9, and 12 months' follow‐up ‐ somatic comorbidity measured by TiC‐P ‐ quality of life at 3, 6, 9, and 12 months' follow‐up (reported in Eveleigh 2014) ‐ costs at baseline, 3, 6, 9, and 12 months' follow‐up (reported in Eveleigh 2014)
Notes	Funding: study supported by Netherlands Organization for Health Research and Development (ZonMW) grant COI: study authors declare that no competing interests exist Other: low participation rate: 15% of participants deemed eligible by the GP; consented to participate
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Comments: randomisation was done by picking an envelope Quote: "random assignment was ensured by picking a sealed envelope with intervention or control group"
Allocation concealment (selection bias)	Low risk	Comment: sealed envelopes were used to conceal allocation Quote: "random assignment was ensured by picking a sealed envelope with intervention or control group"
Blinding of participants and personnel (performance bias) All outcomes	High risk	Comment: no blinding or incomplete blinding; outcome is likely to be influenced by lack of blinding
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: blinding of outcome assessment ensured; unlikely that blinding could have been broken Quote: "interviewers who conducted the baseline and follow‐up interviews as well as the psychiatrist and general practitioner who judged the indication of maintenance treatment will remain blinded throughout the trial"
Incomplete outcome data (attrition bias) All outcomes	High risk	Comment: dropout was higher in discontinuation group (n = 20/70; 28%) compared to continuation group (n = 10/76; 13%). Intention to comply was much lower in discontinuation group (34/70) compared to continuation group (74/76)
Selective reporting (reporting bias)	High risk	Comment: the study protocol is available. Most predefined primary and secondary outcomes were reported in a predefined way. Withdrawal symptoms were measured but were not reported. No adverse events was not an outcome. No CESD; BSI‐53 results reported only at 3 months' follow‐up ‐ not at end of study
Other bias	Low risk	Bias for cluster‐RCT Recruitment bias: low risk Comment: participants were recruited before clusters had been randomised Quote: “to prevent contamination between intervention and control groups, a cluster randomisation was performed with the general practice as the unit of clustering. Random assignment was executed after patient recruitment was concluded per practice; a practice was either an intervention practice or a control practice” Baseline imbalance: low risk Comment: there was baseline comparability of clusters from the data presented Quote: "patient characteristics were well balanced at randomisation; any differences were not statistically significant" Incorrect analysis: low risk Comment: study authors used intra‐class correlation (ICC) adjustment for clustering in the analysis Quote: "because our trial is cluster randomised, calculations to determine the minimum number of general practices is stricter than in a non‐clustered trial. To account for this, we used an ICC of 0.05" Loss of clusters: unclear risk Comment: insufficient information to permit judgement