Selective serotonin reuptake inhibitors (SSRIs) for stroke recovery

Stroke is a major cause of adult disability. Selective serotonin reuptake inhibitors (SSRIs) have been used for many years to manage depression and other mood disorders after stroke. The 2012 Cochrane Review of SSRIs for stroke recovery demonstrated positive effects on recovery, even in people who were not depressed at randomisation. A large trial of fluoxetine for stroke recovery (fluoxetine versus placebo under supervision) has recently been published, and it is now appropriate to update the evidence.

To determine if SSRIs are more effective than placebo or usual care at improving outcomes in people less than 12 months post-stroke, and to determine whether treatment with SSRIs is associated with adverse effects.

For this update, we searched the Cochrane Stroke Group Trials Register (last searched 16 July 2018), the Cochrane Controlled Trials Register (CENTRAL, Issue 7 of 12, July 2018), MEDLINE (1946 to July 2018), Embase (1974 to July 2018), CINAHL (1982 July 2018), PsycINFO (1985 to July 2018), AMED (1985 to July 2018), and PsycBITE March 2012 to July 2018). We also searched grey literature and clinical trials registers.

We included randomised controlled trials (RCTs) that recruited ischaemic or haemorrhagic stroke survivors at any time within the first year. The intervention was any SSRI, given at any dose, for any period, and for any indication. We excluded drugs with mixed pharmacological effects. The comparator was usual care or placebo. To be included, trials had to collect data on at least one of our primary (disability score or independence) or secondary outcomes (impairments, depression, anxiety, quality of life, fatigue, healthcare cost, death, adverse events and leaving the trial early).

We extracted data on demographics, type of stroke, time since stroke, our primary and secondary outcomes, and sources of bias. Two review authors independently extracted data from each trial. We used standardised mean differences (SMDs) to estimate treatment effects for continuous variables, and risk ratios (RRs) for dichotomous effects, with their 95% confidence intervals (CIs). We assessed risks of bias and applied GRADE criteria.

We identified a total of 63 eligible trials recruiting 9168 participants, most of which provided data only at end of treatment and not at follow-up. There was a wide age range. About half the trials required participants to have depression to enter the trial. The duration, drug, and dose varied between trials. Only three of the included trials were at low risk of bias across the key 'Risk of bias' domains. A meta-analysis of these three trials found little or no effect of SSRI on either disability score: SMD -0.01 (95% CI -0.09 to 0.06; P = 0.75; 2 studies, 2829 participants; moderate-quality evidence) or independence: RR 1.00 (95% CI 0.91 to 1.09; P = 0.99; 3 studies, 3249 participants; moderate-quality evidence). We downgraded both these outcomes for imprecision. SSRIs reduced the average depression score (SMD 0.11 lower, 0.19 lower to 0.04 lower; 2 trials, 2861 participants; moderate-quality evidence), but there was a higher observed number of gastrointestinal side effects among participants treated with SSRIs compared to placebo (RR 2.19, 95% CI 1.00 to 4.76; P = 0.05; 2 studies, 148 participants; moderate-quality evidence), with no evidence of heterogeneity (I² = 0%). For seizures there was no evidence of a substantial difference. When we included all trials in a sensitivity analysis, irrespective of risk of bias, SSRIs appeared to reduce disability scores but not dependence. One large trial (FOCUS) dominated the results. We identified several ongoing trials, including two large trials that together will recruit more than 3000 participants.

We found no reliable evidence that SSRIs should be used routinely to promote recovery after stroke. Meta-analysis of the trials at low risk of bias indicate that SSRIs do not improve recovery from stroke. We identified potential improvements in disability only in the analyses which included trials at high risk of bias. A further meta-analysis of large ongoing trials will be required to determine the generalisability of these findings.