Article type
Year
Abstract
Background:
Some literature reviews have demonstrated that trial design characteristics - such as sponsorship, type of comparison and intervention dose - may influence randomized controlled trials (RCTs) results. In psychiatry, there are examples of studies investigating the role of these variables as possible source of bias in RCTs on antipsychotics, but evidence on antidepressants is scant.
Objectives:
This study aimed to examine whether the outcome of RCTs favored fluoxetine, where it was the investigational agent, and favored the comparator when fluoxetine was the reference ('wish' bias) using a Cochrane Review comparing fluoxetine with any other antidepressant. Moreover, the hypothetical role of drug dose as mediator of treatment effect was investigated applying the Baron and Kenny model.
Methods:
We extracted data from all RCTs included in a Cochrane review on fluoxetine. We extracted data on study characteristics, drug dose, and efficacy. We recorded whether fluoxetine was the experimental or reference drug according to the authors’ stated aims of the study.
Results:
We included a total of 173 comparisons (more than 24,000 participants). Meta-analysis of efficacy of fluoxetine revealed a statistically significant advantage for fluoxetine in trials where fluoxetine was the investigational agent (standardised mean difference (SMD), fixed-effect -0.07, 95% confidence interval (CI) -0.13 to -0.02). Where fluoxetine was the reference agent, no significant differences were identified. By means of a meta-regression analysis, which adjusted for potential confounders, we found a statistically significant association between trials where fluoxetine was the investigational agent and overall treatment estimate, favoring fluoxetine. However, the Baron and Kenny model failed to suggest a mediating role of drug dose in influencing treatment effect.
Conclusions:
The evidence that the outcome of fluoxetine trials varied according to whether this drug was used as a new compound or a reference one suggests the presence of 'wish' bias. However, the direct relationship between fluoxetine and outcome does not seem to be mediated by drug dose.
Some literature reviews have demonstrated that trial design characteristics - such as sponsorship, type of comparison and intervention dose - may influence randomized controlled trials (RCTs) results. In psychiatry, there are examples of studies investigating the role of these variables as possible source of bias in RCTs on antipsychotics, but evidence on antidepressants is scant.
Objectives:
This study aimed to examine whether the outcome of RCTs favored fluoxetine, where it was the investigational agent, and favored the comparator when fluoxetine was the reference ('wish' bias) using a Cochrane Review comparing fluoxetine with any other antidepressant. Moreover, the hypothetical role of drug dose as mediator of treatment effect was investigated applying the Baron and Kenny model.
Methods:
We extracted data from all RCTs included in a Cochrane review on fluoxetine. We extracted data on study characteristics, drug dose, and efficacy. We recorded whether fluoxetine was the experimental or reference drug according to the authors’ stated aims of the study.
Results:
We included a total of 173 comparisons (more than 24,000 participants). Meta-analysis of efficacy of fluoxetine revealed a statistically significant advantage for fluoxetine in trials where fluoxetine was the investigational agent (standardised mean difference (SMD), fixed-effect -0.07, 95% confidence interval (CI) -0.13 to -0.02). Where fluoxetine was the reference agent, no significant differences were identified. By means of a meta-regression analysis, which adjusted for potential confounders, we found a statistically significant association between trials where fluoxetine was the investigational agent and overall treatment estimate, favoring fluoxetine. However, the Baron and Kenny model failed to suggest a mediating role of drug dose in influencing treatment effect.
Conclusions:
The evidence that the outcome of fluoxetine trials varied according to whether this drug was used as a new compound or a reference one suggests the presence of 'wish' bias. However, the direct relationship between fluoxetine and outcome does not seem to be mediated by drug dose.