Study design

We performed a cluster-randomized controlled trial in primary care, including patients who had been using ADM for at least nine months who wished to discontinue. Allocation in a 1:1 ratio was performed at the level of the GP practice to avoid contamination and to simplify research procedures for GPs and mental health nurses. Generation of the numbers for the randomization schedule was performed by a statistician (RA), based on block randomization with blocks of two, stratified for region (north, central and east Netherlands) and type of practice (single versus group). Allocation was performed by the researcher (CW) in collaboration with the statistician who consecutively allocated practices to the randomization scheme. Patients, GPs, and the research team were aware of treatment allocation. The mental health nurses were kept blind to allocation to ensure that SPD was offered in the same way to all participants, and both GPs and patients were asked not to reveal this information. The study was approved by the medical-ethical committee Arnhem-Nijmegen (2016–2527) and all patients gave written informed consent before entering the intervention phase of the study.

Study population and procedure

For a detailed description of the study procedures see the published protocol [39]. The trial was registered in a publicly accessible registry in December 2017 (ClinicalTrials.gov; NCT03361514), inadvertently 10 months after the start of recruitment due to the research team being insufficiently attentive to timely registration. Future trials will be registered prospectively. There were no changes to the design, procedures or outcomes after trial commencement, except for having a 12-month assessment instead of a 15-month assessment. Data collection took place between February 2017 and July 2019. GPs from three regions in the Netherlands (north, middle and east) were systematically retrieved and invited via e-mail. In case of non-response, we tried to reach them by phone at least twice. Participating GPs selected from their electronic database adults receiving antidepressants for the past 9 months or longer. They excluded patients currently treated by a psychiatrist; with substance use disorder; non-psychiatric indication for ADM (e.g. neuropathic pain); previous mindfulness training within the past 5 years; or cognitive or language impairments. GPs were allowed to specify additional reasons for exclusion. Next, GPs invited eligible patients to contact the researchers. Due to the cluster-randomized design, patients were aware of treatment allocation before deciding to participate in the study. Patients were asked to complete online informed consent and questions about baseline characteristics, after which they received a decision aid [30] on whether or not to discontinue ADM [30] to discuss with their GP. If they decided to discontinue, they were provided with a personal tapering schedule of 6 months based on a discontinuation protocol and invited for the baseline interview of the intervention phase of the study. Further assessments were after 6, 9 and 12 months. Qualified research assistants interviewed patients at each assessment with the SCID (Structural Clinical Interview for DSM-IV Axis I Disorders) [40]. Self-report measures were completed online using LimeSurvey.

Interventions

Primary outcome

Our primary outcome was the proportion of patients who had fully discontinued their ADM within 6 months (yes/no). Use of medication was measured with a diary and questionnaires at each assessment. We checked all diaries to see if participants had fully discontinued at some point between baseline and 6 months.

Secondary outcomes

Sample size calculation

We expected a discontinuation rate of 15% for patients in the SPD condition and 40% for patients in the SPD + MBCT condition. To detect this difference with a power of 80%, alpha of 0.05 (two-sided), intraclass correlation of 0.05 due to cluster randomization, which was estimated to range between 0.00 and 0.05 in primary care in a previous study [51], with an average of 5 participants per cluster and a minimum of 24 clusters, 59 patients per group would be needed. We used G-power to calculate the uncorrected sample size, after which we adopted the design factor to adjust the required sample size for cluster sampling, using the following formula: 1 + 0.05 * (n-1) [52]. Then, to account for an expected drop-out rate of 15%, we aimed to include 138 patients in the study. For further details we refer to the published protocol [39].

Statistical analysis

The clinical outcome data were analysed and reported on both intention-to-treat (ITT) and per-protocol (PP) analysis. The ITT sample included all patients who were willing to attempt tapering. The PP sample included patients who attended at least 4 sessions of MBCT, if allocated to MBCT. We did not include the SPD intervention as criterium for the per-protocol analysis as this intervention was optional. Multilevel analyses were used to account for the hierarchical structure of the data (i.e. multiple assessments nested within patients, patients nested within practices). Multilevel logistic regression models were used to test differences in discontinuation rates between the conditions. Relapse percentages of depression or anxiety disorders were compared with a Chi² test. Multilevel linear regression analyses were used for all continuous outcomes. To test the difference in change over time between the two groups, we used a model with group, time and the group x time interaction. The interaction term tests the difference in change over time between the groups and is reported in S1 Table. We explored whether any of the baseline characteristics were associated with the willingness to taper, regardless of group allocation, using Chi² analyses for categorical variables and independent t-tests for continuous variables. All analyses were performed in SPSS Statistics version 25.0.

Participant flow

Fig 1 illustrates the flow of GPs and participants. About a quarter of invited GPs participated in the study. Although specific data on reasons for exclusion of patients were only available for 12/36 practices, most common reasons for exclusion were currently treated by psychiatrist (38%), pain (15%), already stopped ADM (14%), current depression (7%), somatic comorbidity (5%) and current anxiety disorder (5%). Of the 276 patients who responded to the invitation letter, 157 (57%) decided not to discontinue ADM, and the remaining 119 (43%) were included (SPD+MBCT n = 73; SPD n = 46). Table 1 shows the baseline characteristics of patients in both randomization groups, separately for the ITT sample and those who were not willing to taper. Looking at the entire group of 276 responders, there were no baseline characteristics that seemed associated with the willingness to taper that is, those who proceeded to the intervention phase of the study versus those who dropped out after shared decision making, except for a trend suggesting that individuals who experienced adverse effects when using ADM may have been more willing to taper (54%) than those who did not experience adverse effects (42%).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Flow chart of general practices and patients with long-term use of antidepressant medication allocated to supported protocolized discontinuation (SPD) or SPD plus mindfulness-based cognitive therapy (MBCT).

Data on the numbers of ADM users, numbers excluded and numbers invited are weighed estimates based on specific data of 12/36 GPs (33%), combined with data on practice size, which were available for all GPs. ADM = antidepressant medication; SPD = supported protocolized discontinuation; MBCT = mindfulness-based cognitive therapy.

https://doi.org/10.1371/journal.pone.0290965.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Baseline demographic and clinical characteristics of long-term users of antidepressant medication (ADM) in primary care responding to the invitation to participate in a study on discontinuing ADM, cluster-randomized on the level of their general practice to mindfulness-based cognitive therapy (MBCT) plus supported protocolized discontinuation (SPD) or SPD alone.

https://doi.org/10.1371/journal.pone.0290965.t001

Primary outcome

The primary outcome was available for 117/119 participants (98.3%). In the ITT analysis, 31/71 (44%) of the participants discontinued ADM in the SPD+MBCT group versus 15/46 (33%) in the SPD group: OR 1.60 (95% CI 0.73 to 3.49), corresponding to a number needed to treat of 9. Only 20/73 patients allocated to MBCT (27%) completed the intervention. In the PP analysis, 10/20 (50%) of the participants discontinued ADM in the SPD+MBCT group versus 15/46 (33%) in the SPD group: OR 2.06 (95% CI 0.69 to 6.14), corresponding to a number needed to treat of 6.

Of the 31 patients who discontinued in the SPD+MBCT group, 13 (42%) restarted during the 12-month study period, 11 (35%) did not, and for seven (23%) it was unknown. Of the 15 patients who discontinued in the SPD group, nine (60%) restarted, five (33%) did not, and for one (7%) it was unknown.

Secondary outcomes

Based on the ITT analysis, relapse into depression or anxiety occurred in 15/73 participants (21%) in the SPD+MBCT group, of whom 6 had anxiety disorder and 9 depressive disorder, and in 5/46 participants (11%) in the SPD group, of whom 1 had anxiety and 4 depressive disorder. Although the confidence interval includes 1.0, the point estimate suggests that patients in the MBCT group are twice more likely to relapse than those in SPD (OR 2.12, 95% CI 0.71 to 6.30).

S1 Table shows that none of the secondary outcomes showed a different trajectory over time between the groups.

Harms

Aside from expected symptoms related to ADM withdrawal (S2 and S3 Tables), no serious adverse events were reported by participants. One participant reported suicidal thoughts, but not tendencies, at 6-month follow-up. Another participant reported that the assessments evoked stress, after which she decided to withdraw from the study. None of these events required medical help.

Discontinuation of antidepressants after MBCT

Discontinuation rates at 6-month follow-up were slightly higher in the SPD + MBCT group (44%) than in the SPD group (33%). This difference was smaller than we had hypothesized, which can in part be explained by the SPD group showing a higher discontinuation rate (33%) than we had anticipated (15%). Perhaps this was caused by our selective sample; those who did proceed to the intervention phase of the study were probably motivated to give discontinuation a try. Although the discontinuation rate in the MBCT group (44%) was in line with our a priori expectation (40%), it was at the low end of estimates from a systematic review, in which 40% to 95% of participants were able to discontinue when supported by psychological or psychiatric treatment [28]. In studies focusing specifically on MBCT in recurrent depression, previously reported discontinuation rates varied between 53% [38] and 70–75% [35, 36]. Relatively low discontinuation rates in the current study may be related to the fact that only a minority (27%) adhered to the MBCT protocol in the current study, which is much lower than for MBCT participants in routine clinical practice, estimated between 81–96% [53–55]. Possible reasons for these low participation rates are discussed below. Secondary outcomes did not differ between the groups.

Participation in the taper support interventions

The current study adopted a comprehensive approach to implementing taper support in primary care. This included an inclusive recruitment strategy in a relatively large number of GP practices, offering a decision aid to discuss advantages and disadvantages of tapering, materials to devise a gradual and patient-tailored tapering schedule, plus additional support from a mental health nurse (i.e. SPD) and a patient-tailored MBCT intervention. Thus, in theory, our approach seemed adequate to reach and support professionals and patients who may have experienced a lack of support in regular healthcare. Nevertheless, despite the perceived need for such support, actual participation in patient-tailored interventions was low, both at the level of the GP practices (only a quarter of the invited practices participated), and at the level of patients of whom only about one tenth responded to the study invitation. This is in line with previous studies that aimed to support tapering in long-term users [4, 56]. Reasons for non-participation for GPs were not systematically collected, but ‘lack of time’ was often mentioned. Reasons for non-participation of non-responding patients were unknown.

Of those who did respond to the invitation, less than half participated in the intervention study. Patients’ reasons for non-discontinuation included fear of recurring symptoms, pain, somatic comorbidity, social circumstances, or clinicians’ advice against tapering. Only a minority of the patients actually reported using the decision aid with their GP, so we do not know to what extent this document could be helpful in making a truly shared and balanced decision about whether or not to taper ADM. Perhaps some of the reasons for not tapering that patients mentioned in the research interviews (e.g. fear of worsening, pain, social circumstances) already shaped their decision before even using the decision aid. Another possible explanation is that the decision aid was insufficiently personalized, despite being co-created with diverse stakeholders to ensure a wide scope of topics. We may need to investigate patient expectations and values and GP’s considerations on an individual level. For example, patients’ beliefs about the causes of their symptoms were not included in the decision aid. If many patients adopted a biomedical framework and believed that their depressive symptoms for example were due to a biochemical imbalance that needs to be restored with medication, the relatively low engagement in tapering is not surprising.

The low participation rate in the MBCT intervention might have been due to our primary focus on withdrawal from ADM, so the resulting sample may have been less “mindfulness-minded” than a self-selecting sample. Perhaps patients who responded to our study invitation were interested in tapering their medication rather than in attending eight weekly 2,5 hour MBCT classes, despite being informed about the time investment of this intervention. Work and travel issues were the most common barriers for participation in MBCT.

Limitations

Our recruitment strategy was designed to keep the possible selection bias of mindfulness studies at a minimum, thereby increasing the generalizability of the findings. The other side of the coin, however, is that many of those allocated to MBCT did not choose to participate in the intervention after all. Similarly, SPD was used by less than half of the participants. Thus, a major limitation of the current study is that adherence to the interventions was low. Consequently, we are unable to say whether MBCT added to SPD has the potential to significantly increase successful tapering of ADM. The study was underpowered to detect a small difference in discontinuation rates. Another limitation is that we failed to systematically assess reasons for non-participation or non-discontinuation at all stages of the process when implementing our taper support strategies. For example, we do not know why participants did not use the SPD consultations as intended. Furthermore, the partnership between patient and GP was not explicitly targeted, and responsibility for making an appointment to discuss tapering was given to the patients only. In addition, the current study selected patients who all expressed a desire to discontinue their medication. Therefore, we cannot generalize our findings to all patients who are using antidepressant medication. For the aim of this study, however, it seemed most appropriate ‐ both on a practical and ethical level ‐ to select those who are motivated to taper their medication. Finally, a methodological limitation is that we used block sizes of two in our randomization schedule. Although we chose this smaller block size to prevent unequal distribution within cells that were expected to have lower numbers of GP practices due to the stratification variables (region: north, central and east Netherlands; and type of practice: single versus group), this small block size does come with a larger risk of prediction.

Future directions

Our findings of low engagement raise the question where and how we can find the patients who are ready and willing to taper. Participants did not actively ask for support but were invited in the context of the study. An alternative approach could be to target our interventions at those who actively request taper support. However, from previous studies we know that patients believe that it is the doctor’s responsibility to initiate the conversation about tapering ADM [20]. Without active recommendation by the GP and in absence of consistent guidelines or knowledge about when to start tapering, patients might leave the situation as it is. Therefore, a shared decision making process is essential. This does not have to be limited to GPs, and qualitative studies suggest that there might also be a role for other health care professionals (mental health assistants, psychiatrists or psychologists) or significant others (family or friends) [20]. To promote engagement, fostering the partnership between patients, health care teams and even researchers might be an important step forward, as suggested by a study investigating a coproduction model for opioid tapering [57]. A similar approach will probably be useful for antidepressant tapering too, with a focus on acceptable, feasible and tailored forms of support.

In addition, new forms of support are being developed, such as ADvisor [58], a digital intervention designed as part of the REDUCE programme [59] to support ADM discontinuation, which integrates different kinds of resources that people can use flexibly (including psychoeducation, relapse prevention techniques grounded in MBCT, goal setting exercises based on acceptance and commitment therapy, techniques for managing distress, tools for activity planning and information for family and friends).

Work or travel issues were most often mentioned as barriers to participation. Therefore it would be interesting to include different formats of MBCT, such as stand-alone, blended or therapist-assisted online mindfulness interventions. At the same time, this might challenge the possibly important role of peer support that emerged from the post-MBCT interviews with a subsample of the patients. Video-conferencing programs would overcome the issue of travel time, and may be more appealing for patients with physical problems. In stand-alone interventions, audiovisual materials such as interviews with ‘peers’ who tell about their experiences might help to connect with others’ experiences. Similar adaptations can also be considered for SPD or other forms of monitoring in primary care.

Furthermore, we might consider partial rather than full discontinuation as a valid outcome as well, given the evidence that most withdrawal problems might occur at the lowest doses [24] and there is preliminary evidence for relatively low depressive relapse rates after partial discontinuation [60]. In fact, mindfulness training may even help to be more flexible, self-compassionate and open-minded towards using different treatments to manage depressive symptoms [32].