BACKGROUND: Self-expandable metallic stent (SEMS) placement is widely carried out for distal malignant biliary obstruction, and survival analysis is used to evaluate the cumulative incidences of SEMS dysfunction (e.g. the Kaplan-Meier [KM] method and the log-rank test). However, these statistical methods might be inappropriate in the presence of 'competing risks' (here, death without SEMS dysfunction), which affects the probability of experiencing the event of interest (SEMS dysfunction); that is, SEMS dysfunction can no longer be observed after death. A competing risk analysis has rarely been done in studies on SEMS. PATIENTS AND METHODS: We introduced the concept of a competing risk analysis and illustrated its impact on the evaluation of SEMS outcomes using hypothetical and actual data. Our illustrative study included 476 consecutive patients who underwent SEMS placement for unresectable distal malignant biliary obstruction. RESULTS: A significant difference between cumulative incidences of SEMS dysfunction in male and female patients via theKM method (P = 0.044 by the log-rank test) disappeared after applying a competing risk analysis (P = 0.115 by Gray's test). In contrast, although cumulative incidences of SEMS dysfunction via the KM method were similar with and without chemotherapy (P = 0.647 by the log-rank test), cumulative incidence of SEMS dysfunction in the non-chemotherapy group was shown to be significantly lower (P = 0.031 by Gray's test) in a competing risk analysis. CONCLUSION: Death as a competing risk event needs to be appropriately considered in estimating a cumulative incidence of SEMS dysfunction, otherwise analytical results may be biased.
BACKGROUND: Self-expandable metallic stent (SEMS) placement is widely carried out for distal malignant biliary obstruction, and survival analysis is used to evaluate the cumulative incidences of SEMS dysfunction (e.g. the Kaplan-Meier [KM] method and the log-rank test). However, these statistical methods might be inappropriate in the presence of 'competing risks' (here, death without SEMS dysfunction), which affects the probability of experiencing the event of interest (SEMS dysfunction); that is, SEMS dysfunction can no longer be observed after death. A competing risk analysis has rarely been done in studies on SEMS. PATIENTS AND METHODS: We introduced the concept of a competing risk analysis and illustrated its impact on the evaluation of SEMS outcomes using hypothetical and actual data. Our illustrative study included 476 consecutive patients who underwent SEMS placement for unresectable distal malignant biliary obstruction. RESULTS: A significant difference between cumulative incidences of SEMS dysfunction in male and female patients via theKM method (P = 0.044 by the log-rank test) disappeared after applying a competing risk analysis (P = 0.115 by Gray's test). In contrast, although cumulative incidences of SEMS dysfunction via the KM method were similar with and without chemotherapy (P = 0.647 by the log-rank test), cumulative incidence of SEMS dysfunction in the non-chemotherapy group was shown to be significantly lower (P = 0.031 by Gray's test) in a competing risk analysis. CONCLUSION:Death as a competing risk event needs to be appropriately considered in estimating a cumulative incidence of SEMS dysfunction, otherwise analytical results may be biased.