Publication bias in neurosurgery: lessons from series of unruptured aneurysms.

Abstract

Bias favouring publication of research with "positive" results over studies with "negative" results is widely suspected. The present investigation addressed this problem in the field of neurosurgery through a review of recent literature concerning outcome of surgery for unruptured intracranial aneurysms. A Medline search was performed seeking case series of surgical treatment for unruptured intracranial aneurysms that analyzed 50 or more patients. Ten type I studies (retrospective studies from a single institution; 1457 patients) met these entry criteria. In general, type I studies reported excellent surgical outcome, with mean combined mortality and morbidity of 7.8% (95% confidence interval (CI), 6.4% to 9.2%). We found 4 multicenter or community-based studies (type II studies; 5401 patients). Mean combined mortality and morbidity in the type II studies was 20.3% (95% CI, 19.2% to 21.4%), much higher than in type I studies. Relative risk was 2.6 (95% CI, 2.2 to 3.1) for patients in type II studies compared with those in type I studies. Two meta-analyses (type III studies) described combined mortality and morbidity of 5.0% and 12.7%, essentially corresponding to results of type I studies. Publication bias is present in the neurosurgical literature. Studies with an excellent surgical outcome are more likely to be published than those with an average outcome. Conclusions of literature reviews or meta-analyses based on published work therefore may be misleading. The solution to the problem would be community-based prospective registration of all patients who underwent surgery, providing a sampling frame free from publication bias.