Citation bias favoring positive clinical trials of thrombolytics for acute ischemic stroke: a cross-sectional analysis

BACKGROUND Citation bias occurs when positive trials involving a medical intervention receive more citations than neutral or negative trials of similar quality. Several large clinical trials have studied the use of thrombolytic agents for the treatment of acute ischemic stroke with differing results, thereby presenting an opportunity to assess these trials for evidence of citation bias. We compared citation rates among positive, neutral, and negative trials of alteplase (tPA) and other thrombolytic agents for stroke. METHODS We used a 2014 Cochrane Review of thrombolytic therapy for the treatment of acute stroke to identify non-pilot, English-language stroke trials published in MEDLINE-indexed journals comparing thrombolytic therapy with control. We classified trials as positive if there was a statistically significant primary outcome difference favoring the intervention, neutral if there was no difference in primary outcome, or negative for a significant primary outcome difference favoring the control group. Trials were also considered negative if safety concerns supported stopping the trial early. Using Scopus, we collected citation counts through 2015 and compared citation rates according to trial outcomes. RESULTS Eight tPA trials met inclusion criteria: two were positive, four were neutral, and two were negative. The two positive trials received 9080 total citations, the four neutral trials received 4847 citations, and the two negative trials received 1096 citations. The mean annual per-trial citation rates were 333 citations per year for positive trials, 96 citations per year for neutral trials, and 35 citations per year for negative trials. Trials involving other thrombolytic agents were not cited as often, though as with tPA, positive trials were cited more frequently than neutral or negative trials. CONCLUSIONS Positive trials of tPA for ischemic stroke are cited approximately three times as often as neutral trials, and nearly 10 times as often as negative trials, indicating the presence of substantial citation bias.