Retrospective Analysis of R-COMP Therapy in Patients with Diffuse Large B-Cell Lymphoma (DLBCL): Assessing the Impact of Sample Selection Bias

Background: Retrospective studies are often criticized for their susceptibility to case selection bias compared to prospective studies, which include all patients consecutively and are thus less prone to such limitations. However, the larger sample sizes typical of retrospective studies can sometimes offset this drawback. On behalf of the Fondazione Italiana Linfomi (FIL), a substantial retrospective study involving 946 patients was conducted to examine the use of non-pegylated liposomal anthracycline (Myocet). This was followed by a prospective study, the Prospective Elderly Project, which enrolled 308 patients treated with the same liposomal anthracycline regimen. Methods: The objective of this analysis was to determine whether the patient cohort from the retrospective study significantly differed from the cohort in the prospective study. Statistical hypothesis testing was applied to assess whether the samples from both studies originated from the same underlying population. The Anderson–Darling test, a non-parametric statistical method, was utilized to evaluate and compare the overall survival distributions between the two patient cohorts. Results: The statistical tests produced conflicting results, suggesting a potential selection bias in the retrospective study or the possibility that the two groups were from the same population. These discrepancies may have arisen due to the choice of statistical methods or the quality of the data analyzed. Conclusions: This study highlights the challenges of comparing retrospective and prospective cohorts and underscores the importance of selecting appropriate statistical methodologies. The findings provide valuable insights and lay the groundwork for developing innovative approaches to improve such comparisons in future research.