Determining knowledge transfer gaps in the life cycle of evidence for chemotherapy in non-small cell lung cancer (NSCLC) through cumulative meta-analysis

6555 Background: The uptake of research findings into practice is often delayed (Antman et al. JAMA 1992). We investigated the relative role of evidence in determining clinical recommendations and practice for NSCLC. Novel chemotherapy agents (NCA) approved for use in Canada between 1992 and 2002 were eligible. Methods: We conducted a systematic review of evidence for vinorelbine (V), paclitaxel (P) and gemcitabine (G) in combination with a platinum agent for the treatment of advanced NSCLC. Primary endpoint of efficacy was median survival. For each included randomized controlled trial (RCT), the publications dates of abstracts and journal articles were considered. At each time point when new data was available, a new meta-analysis using the Follman methodology was performed. Bibliometric analyses were performed to identify key milestones for each drug (e.g. Health Canada Notice of Compliance, provincial drug funding, and clinical recommendations). Results: 3,399 references were obtained for NCA in advanced NSCLC. Eligibility review identified 20 references for V representing 6 RCTs (1994–2002), 16 for P representing 4 RCTs (1997–2000), and 10 for G representing 7 RCTs (1998–2003). All drugs trended towards median survival benefit throughout the time of analysis. However, over time the estimated effect for V became weaker, remained stable for P and became stronger for G. The 1997 ASCO guideline recommended V and P as standard therapy; G was only cited as a promising investigational agent. By 2003, the ASCO guideline recommended G as a standard. Only V received Ontario provincial funding in 1997, while G and P received funding in 2002 and 2003 respectively. Conclusions: This study demonstrates the relatively small pool of RCT evidence for NCAs. These three commonly used NCAs demonstrate different patterns of evolution of evidence. For advanced NSCLC, the time gap between evidence and clinical recommendations is short. Caution should be used for generating recommendations using early results, when the evidence base is not stable, which may either over or underestimate true effectiveness. No significant financial relationships to disclose.