Analysis of conflicting results of 2 non-randomized controlled trials

Tags: Poster
Peinemann F1, Kulig M2
1IQWiG, Cologne, Germany, 2G-BA, Berlin, Germany

Background: In the framework of a systematic review on patients with rhabdomyosarcoma with vs. without autologous stem cell transplantation, we identified a total of 4 non-randomized controlled studies. Two studies reported statistically significant, yet contradictory, estimates of the overall survival at 3 years. Klingebiel 2008 reported 22% vs. 55% and Hosoi 2007 reported 53% vs. 18%. The difference of treatment effect in 2 further non-randomized studies did not reach statistical significance. Carli 1999 reported 40% vs. 28% and Suita 2005 reported 73% vs. 52%. Objective: The aim of this analysis was to explore potential influence factors potentially explaining the contradictory results. Methods: We investigated risk of bias and compared study and patient characteristics. The distribution of prognostic factors, such as age of patients, histological subtype, bone metastasis, and tumor size were also analyzed, to search for a possible explanation of the differences between the studies. Results: Criteria for assignment of patients to the treatment groups were not reported and a considerable number of enrolled patients were not analyzed in both studies. Patients with embryonal rhabdomyosarcoma younger than 10 years of age were excluded from Hosoi 2007 but included in Klingebiel 2008. Controls were treated differently in Hosoi 2007 (standard chemotherapy) and Klingebiel 2008 (oral maintenance). Other heterogeneous disease and treatment conditions, such as bone marrow involvement and type of chemotherapy were observed. Differences regarding proportions of prognostic factors between the treatment groups, such as age and histologic subtype were noticeable but were based on small numbers of patients. Conclusion: High risk of bias, heterogeneity of disease and treatment conditions, and unbalanced distribution of prognostic factors might have contributed to a possible distortion of results.