A meta-analysis that wasn't because of inappropriate trial design and reporting

Article type
Johansen HK, Gotzsche PC
Introduction/Objective: To compare the efficacy of amphotericin B and fluconazole in cancer patients with neutropenia.

Methods: Randomised trials comparing amphotericin B with fluconazole were collected. A modified Cochrane search strategy was used. Data were extracted independently by two observers. Mortality, invasive fungal infections (fungi in blood, oesophagus, lungs or deep tissues (microscopically confirmed)) and colonisation were the outcome measures.

Results: Eight full articles, one interim analysis and three congress abstracts were identified. Three of the studies, comprising 44% of the 2919 patients in the trials, had three arms: fluconazole, amphotericin and nystatin. However, the results were not given for amphotericin and nystatin separately but as a combined "polyene" group. This is surprising, since nystatin has never been shown to be effective against systemic fungal infections (in fact, two of the three authors admitted in the discussion section of their papers that nystatin is a poor drug!). I v. amphotericin was used in 5 studies, corresponding to 20% of the randomised patients. In the remaining 80%, oral amphotericin was used which is also surprising since oral amphotericin is poorly absorbed. Accordingly, we found in a previous meta-analysis of placebo controlled trials of antifungals that all amphotericin trials except one (which has been published only in Japanese) used i.v. amphotericin.

Discussion: Since the design and reporting of these studies were biased in favour of fluconazole we decided not to perform a meta-analysis on the data currently available to us. We suspect undue industry influence because Pfizer, the company that produces fluconazole, was thanked for their support in 7 of the 9 articles.