On the efficacy and prescription of serotonin reuptake inhibitors

Abstract

This thesis addresses several controversial issues regarding commonly used antidepressants, including selective serotonin reuptake inhibitors (SSRIs), serotonin and noradrenaline reuptake inhibitors (SNRIs), and tricyclic antidepressants (TCAs). To this end, patient-level data from 28 placebo-controlled trials on SSRIs and 13 corresponding trials on the SNRI duloxetine, were requested and obtained. Data from Swedish nation- or region-wide registers were also utilised. Questions addressed include i) whether the antidepressant effects of SSRIs and SNRIs are too small to be clinically meaningful, ii) if the antidepressant effects of SSRIs and SNRIs are present also in those not severely depressed, iii) if the effects of an SNRI are present only in those reporting side effects, iv) how an SNRI influences suicidality, v) whether the placement of a commonly used cut-off for response to treatment might inflate the apparent efficacy of SSRIs and SNRIs, vi) if there are differences between SSRIs and an SNRI in terms of reducing specific symptoms of depression, vii) if a TCA, amitriptyline, is more effective than an SSRI in reducing core symptoms of depression, and viii) to what extent SSRI doses eliciting maximum antidepressant effects are prescribed to depressed individuals in Sweden. Using a population of 8262 patients treated with an SSRI (citalopram, fluoxetine, paroxetine, or sertraline), we explored whether meaningful antidepressant effects are at hand also in those not severely depressed. While the usual outcome measure used in these trials, the sum score of the Hamilton Depression Rating Scale 17-item version (HDRS-17-sum), to some extent revealed increasing drug-placebo differences in those more severely ill, this was not the case for the core depressive symptoms included in the unidimensional HDRS-6 subscale, or for the individual HDRS item depressed mood. Meaningful antidepressant effects were thus evident also for those not severely depressed, and were similar in magnitude to those in the severe group. In a population of 3575 duloxetine-treated patients, the above observations were replicated, as were several other previous findings on the SSRIs. Thus, we could refute that the effects of duloxetine are too small to be relevant, and instead concluded that these have been underestimated. An early improvement on certain depressive symptoms was at hand already after one week of treatment. The effect of duloxetine was not restricted to those with severe depression, and was not secondary to side effects breaking the blind. No evidence for a suicide-provoking effect of duloxetine was found. No significant differences in terms of item-specific effects adjusted for overall efficacy were found between duloxetine and the SSRIs. Using both the SSRI and duloxetine populations, we found no evidence that differences between active treatment and placebo in terms of response to treatment, meaning a ≥50% reduction of baseline HDRS scores, be inflated due to choosing a cut-off close to the median. Drug-placebo separation was instead largely independent of chosen cut-off, except for extreme values. In line with previous studies on mean symptom ratings, the separation with respect to response rates was larger when measured using cut-offs on HDRS-6-sum or the depressed mood item, and was highly dependent on the SSRI dose. When comparing the effects of a TCA, amitriptyline, to those of an SSRI, sertraline, using a population of 1241 individuals, a significant superiority of amitriptyline to sertraline was noted when using HDRS-17-sum, but not for most core symptoms of depression, where amitriptyline was only numerically superior. Instead, the difference observed with respect to HDRS-17-sum in previous meta-analyses seems to be largely explained by amitriptyline outperforming the SSRIs on items related to sleep, or those expected to be negatively impacted by side effects of SSRIs. Finally, using register-based data on 50365 Swedish individuals, we found that a majority of those collecting an SSRI never receive a dose eliciting maximum antidepressant efficacy, as judged from fixed-dose trials. The three most common SSRIs - citalopram, escitalopram, and fluoxetine - were all found to be underdosed. No support was found for higher doses being less tolerable in individuals aged below 65 years, but in those aged 65 and above. To conclude, we found no evidence for several of the arguments underlying the widespread questioning of commonly used antidepressants. By exploring effects on individual symptoms of depression, we could instead, with high consistency and in two different populations, show that the effects of SSRIs and SNRIs have been underrated due to the use of an insensitive outcome measure, HDRS-17-sum. This was also found to be true for dichotomous cut-offs measuring response to treatment. The clinical profiles of SSRIs and duloxetine were highly similar to each other, but not to that of amitriptyline. We could also conclude that most patients prescribed an antidepressant are never treated with a dose that has been found to be maximally effective in clinical trials.