A systematic review of oral steroids in cystic fibrosis: statistical challenges

Tags: Poster
Preston C, Ashby D, Cheng K, Smyth RL

Introduction/Objective: To consider statistical issues arising from systematic reviews of chronic diseases.

Methods: The Cochrane Cystic Fibrosis Group has identified 350 randomised controlled trials (RCTs) and is now undertaking reviews of specific interventions. In order to do this it is usual to summarise results in a common form and then average them. For trials with binary endpoints the methodology for a meta-analysis is well established, but for trials with longitudinal data this is more complex. Cystic Fibrosis (CF) is a chronic disease, where lung function is often the primary outcome measure and while longitudinal data is collected it is frequently not utilised in the analysis. Three trials which examine the use of oral steroids in the treatment of CF patients have been identified to determine whether there is a role for oral steroids in the treatment of CF patients with mild to moderate pulmonary disease. Methods for combining aggregate data from published reports presented in different ways will be addressed.

Results: The design problems encountered in these studies are similar to those found in other areas, and although all three trials reported being double-blind, placebo-controlled only two performed an intention to treat analysis. The duration, dosage and method of analysis also varied between the studies.

Discussion: The problems of the different study designs and presentations of aggregate data encountered in this review will be discussed together with possible approaches to solutions including reading estimates from graphs at particular time points.