Aim
The primary aim was to test whether the brief intervention delivered in-person or by telephone is superior to usual care in terms of post-stroke depression treatment response (percent reduction in depressive symptoms) or remission in the short-term (immediately following treatment) and at 1 year post treatment.
Design
Living well with stroke 2 (LWWS 2) was a randomized controlled efficacy study comparing two modes of delivery of a brief psychosocial behavioural intervention with usual care in both ischemic and haemorrhagic stroke survivors. It was designed as a three arm study with a pre-planned combining of the two intervention arms for primary analysis if they were found to be statistically equivalent. The Institutional Review Board of the University of Washington (Human Subjects Division) approved this study for protection of human subjects.
Participants
One hundred people who were within 4 months of an ischemic or haemorrhagic stroke (verified by Computerized Tomography or Magnetic Resonance Imaging) consented to participate in this RCT. Clinical depression was identified by the screening score of the Geriatric Depression Scale [12] and verified at entry to the trial by the Diagnostic Interview and Structured Hamilton, previously validated as consistent with criteria from the 4th Edition of the Diagnostic and Statistical Manual of Mental Disorders [10].
Protocol
This study followed the same recruitment and treatment protocols as described in living well with stroke 1 (LWWS 1) [3, 11]. The primary difference from LWW1 was the mode of delivery—offering treatment by telephone as well as in-person, and a reduction in the length of treatment (6 weeks vs 8 weeks). The treatment content remained the same and is described briefly.
Screening for eligibility
Patients were recruited from six university and community hospitals in the Seattle, WA area. Consecutive patients newly admitted with stroke were introduced to the study by intermediaries. Study research nurses obtained written consent from those who agreed to be screened for eligibility and administered the 30 item Geriatric Depression Scale [12]. Those who scored ≥ 11 were invited to join the full study. Clinical depression was verified as noted above. All screening and full study data were collected after the participant’s discharge from the hospital. Recruitment occurred from May 2010 to December 2014 and follow-up from August 2010 to December 2015. The trial ended in December 2015 when the last follow up was completed. The study flow chart is shown in Fig. 1.
A letter identifying each person’s participation in the study, but not their treatment allocation, and recommended antidepressant therapy was sent to the participant’s primary provider. In addition, all participants received written materials from the American Stroke Association about stroke recovery, including information on depression. All participants received ongoing medical care from their own provider, including antidepressant adjustment as determined by their provider.
Brief psychosocial–behavioural intervention
Those randomized to intervention (either telephone or in-person) had one in-person orientation session with the psychosocial nurse practitioner therapist, either in their home or at our study offices. They received the participant manuals, discussed goals and expectations of each session, and learned how to fill out the homework sections.
Participant manuals were revised from those used in LWWS1 to reflect fewer sessions, based on a pilot with people who had completed LWWS1 control group follow up. The content of the intervention was adapted from the Seattle Protocols [13] in which cognitive behavioural therapy was conducted with adults with mild dementia. It was designed to challenge negative cognitions, reduce distortions and enable more adaptive ways of viewing specific situations or events. A behavioural intervention, used with adults with more moderate or severe dementia, increased the level of positive activities and decreased negative ones. Problem-solving was taught to provide effective strategies for behaviour change [13]. LWWS was revised from the Seattle Protocols for people with PSD, and aimed to reduce depression through pharmacotherapy of serotonin pathways, boosted by challenging negative cognitions and behaviours, enhancing engagement in and perception of positive events, and immediate provision of social support. LWWS2 content was unchanged in that we taught participants about the relationship of depression and stroke, that depressive symptoms are observable and potentially modifiable behaviours and that these behaviours and associated negative cognitions can be changed through observation and interaction. Participants were helped to identify activities that they found pleasant to do and to build these into their daily activities. Problem solving approaches were used to tailor this treatment to the circumstances of each participant and the challenges each faced in stroke recovery. This content was outlined in our prior publications and the manuals are available from the corresponding author [3, 11].
Following the in-person orientation session, each of the subsequent six sessions occurred either in-person, usually at the participant’s home, or by telephone. Topics were as follows: (1) introduction to behavioural therapy for depression after stroke, pleasant events; (2) scheduling pleasant events: problems and planning; (3) managing depression behaviours: problem-solving techniques; (4) changing negative thoughts and behaviours; (5) problem-solving in depth; (6) review of skills, generalization & strategies for maintenance of skills. At the end of each session, the participant provided an evaluation of the session and a rating of mood on a 0 (very happy) to 9 (very depressed) scale. Session length ranged from 10 to 80 min, with the telephone sessions somewhat shorter than the in-person ones (average 26 min versus 38 min).
As in LWSS 1, a family member or informal caregiver could opt to participate and provide data with the participant’s agreement. Fourteen family caregivers of the 72 participants in the intervention groups chose to be part of the study. One session included content directed particularly at the caregivers.
Sessions were recorded with the participants’ consent and treatment integrity was evaluated by listening to 10% of the recordings, in conjunction with the session checklist. No problems were detected over the course of the study (Additional file 1).
Participants in the intervention arms saw their primary care or stroke provider for stroke follow up care and were provided antidepressants as prescribed by their providers. Since antidepressant treatment is an informal standard of care in the community where this study place, we chose to have antidepressants prescribed by the participant’s medical care provider, rather than by the therapist in our study. The letter sent to providers by the principal investigators and the study psychiatrist stated: “The drugs that are recommended in the literature for post-stroke depression include the serotonin selective reuptake inhibitors (SSRI) sertraline as the primary antidepressant medication and citalopram or paroxetine as alternative medications for those who do not respond to sertraline. These SSRIs have been used extensively in at-risk cardiovascular populations without adverse cardiac effects.”
Usual care control
Participants randomized to usual care reported on their progress at follow up visits in their homes from the research nurses at 8, 21 weeks, and 12 months following entry to the study. Seven family caregivers joined the 28 participants in this arm. As with our prior work we did not provide a “time and attention” control since the studies from which LWWS was adapted and other problem-solving interventions have not shown active control to be equivalent to active intervention [14]. Antidepressants were prescribed by the participant’s usual care provider in the same manner as they were for participants in the intervention arms.
Study timetable and assessments
Baseline data consisted of demographics, medical history, stroke characteristics, National Institutes of Health Stroke Scale score [15], Geriatric Depression Scale [12], Diagnostic Interview and Structured Hamilton Rating Scale for Depression [10], Barthel Index [16], Stroke Impact Scale [17], and percent perceived recovery (overall stroke impact) [17]. All were assessed at entry before randomization. Measures of depressive symptoms, stroke impact, and perceived recovery were assessed at 8 weeks (just following the intervention period), at 21 weeks and at 12 months following entry. Percent reduction in HDRS and remission data from the 8 week and 12-month time points were the four primary endpoints used to evaluate the primary aim.
Randomization and masking
The algorithm used for randomization was a modification of the minimization method described on page 84 of Pocock [18]. The algorithm was based on an imbalance score which measured, for a given set of random assignments, how far out of balance the study would be within strata for each factor and then summed over factors [19]. When a new subject was available for randomization, we computed what the imbalance score would be if this subject were assigned to usual care, or to telephone intervention, or to in-person intervention. Then a randomization was done to allocate two intervention participants to each control with each new assignment having a higher probability of less imbalance. The schema did not require equal numbers in each arm.
This procedure balanced the three groups with respect to important predictors of outcomes, including stroke severity (National Institutes of Health Stroke Scale—NIHSS), severity of baseline depression (Hamilton Rating Scale for Depression—HRSD), age, gender, and type of stroke (ischemic or haemorrhagic) (see Table 2). The study statistician generated the algorithm, which was securely stored and accessible only by the statistician and research nurse supervisor. Research nurses entered relevant assessment data and the research nurse supervisor informed the participant of the study arm to which they were randomized, thus masking outcome assessors to the participant’s randomization status. Participants were asked not to reveal their study arm to the outcome assessors. We did not detect any breaches in masking.
Sample size and statistical analysis
Our original recruitment plan called for 75 in each of the three arms, providing a power of 92% to detect an odds ratio of 2.7 between either intervention and control or between the two intervention groups. This was the odds ratio found in our initial study at the primary endpoint. We planned to combine the two intervention groups if there was no significant difference between them to increase the power to detect a difference between combined intervention and control. A lower rate of sadness and higher than previous rate of refusal led to a necessary reduction in the sample size. We reprogrammed our randomization algorithm to allocate remaining recruits on a ratio of 2 in-person, 2 telephone, 1 usual care. Assuming a total final sample of 165, this reduced power to 74% for an odds ratio of 2.7 and 80% for an odds ratio of 2.9. Continued low enrollment reduced our final sample to 100, which an interim analysis had shown as having similar effect size with our original estimate.
Analysis of outcomes was conducted per protocol and without imputation. Percent reduction in HRSD was the primary symptom response endpoint, as well as remission, defined here as HRSD score ≤ 10 (no longer meeting depression criteria) [20, 21]. We used analysis of variance for continuous variables (percent reduction in HRSD) and logistic regression for binary variables (remission or not).