Survival Analysis and Cox Proportional Hazards Model Reporting in Pediatric Leukemia Studies—a Systematic Review

Abstract

Survival (overall, event free, etc.) is the most-used outcome in clinical oncology studies. This study analyzed methodological reporting of survival analysis in pediatric leukemia studies, focusing on Cox proportional hazards (PH). We performed a systematic review of studies published between 2012 and 2021 in the five highest-ranking oncology and five highest-ranking hematology journals. The included studies had to focus on pediatric leukemia and utilize survival analyses. We extracted data on how the survival analysis methodology was reported and focused on Cox proportional hazards modeling and whether the PH assumption was checked. We screened 561 studies and included 103 in the analysis. The most-used crude survival analysis method was Kaplan–Meier, as 96 (94%) of the 103 studies applied it. Adjusted survival analysis was performed in 80 (78%) of the included studies, and the Cox PH model was used in 77 (96%) of these studies. The PH assumption was mentioned in 18 (23%) of the 77 studies that used the Cox PH model. Only nine studies (12%) stated how the PH assumption was assessed. We noted 10 (13%) studies with possible violations of the PH assumption. Overall, we found a need for improvement in the reporting of survival analysis and especially PH assumption in pediatric leukemia studies. The Cox PH model was the most-used adjusted survival analysis method but checking of the background assumption was not reported in most of the studies.