Competing risk: a potential risk factor for misleading results of length of stent patency

We read with great interest the article by Kim et al. [1] about a retrospective study evaluating the effects of chemotherapy on the patency of self-expandable metal stents (SEMS) for malignant outlet obstruction caused by gastric cancer. Kim et al.’s analysis showed that greater time-to-progression (TTP) and first-line chemotherapy were associated with significantly longer SEMS patency. Better local tumor control in patients with a better response to chemotherapy theoretically provides longer SEMS patency. However, these “reasonable” results regarding SEMS patency and its association with chemotherapy should be interpreted with caution due to the presence of competing risks.

“Competing risks” are defined as causes other than those included in the analysis that may affect or compromise the probability of occurrence of an event of interest [2]. Kim et al. [1] calculated the cumulative incidence functions (CIFs) of SEMS dysfunction as 1 – Kaplan – Meier estimate and compared these using the log rank test; however, this conventional method of calculation overestimates CIFs in the presence of competing risks. Thus, a competing risk analysis should also be considered to compare appropriately the length of patency of SEMS in two groups with a high rate of competing risk events.

We propose that the deaths prior to SEMS dysfunction in Kim et al.’s study should be treated as a competing risk for SEMS dysfunction. Although Kim et al. did not provide detailed survival data, the prognosis was probably much better in patients who had had a long TTP or had undergone first-line chemotherapy. As death prior to SEMS dysfunction is a competing risk for SEMS dysfunction, it was more likely to occur in patients who had a short TTP or had undergone salvage chemotherapy, and CIFs of SEMS dysfunction were likely to be overestimated in the groups with poor prognosis. As a result, a longer survival time might have favorably affected the CIF of SEMS dysfunction as calculated by 1 – Kaplan – Meier estimate in groups with a long TTP and first-line chemotherapy, leading to apparently better SEMS patency. We believe CIFs of SEMS dysfunction should be estimated and compared using competing risk analysis [2] [3] instead of conventional methods, especially when comparing SEMS patency times of two groups with different prognoses, as carried out by van Hooft JE et al. [4] in their study of duodenal SEMS. Furthermore, Kim et al. [1] separately computed CIFs of SEMS migration and restenosis; however, these two causes of SEMS dysfunction are potential competing risk events for each other. The counterpart of the Cox proportional hazards model in a competing risk framework was also proposed by Fine and Gray [5]. Thus, we believe the validity of Kim et al.’s results would have been enhanced by utilizing a competing risk regression model.

Finally, we wish to comment on the prognostic factors included in the multivariate analysis. Initially, only baseline characteristics should be included in the Cox proportional hazards model; particular attention is required, e. g., the landmark method [6], when factors pertaining to the follow-up period are included in the analysis, such as response to chemotherapy or TTP, as in Kim et al.’s study.

• References

• 1 Kim CG, Park SR, Choi IJ et al. Effect of chemotherapy on the outcome of self-expandable metallic stents in gastric cancer patients with malignant outlet obstruction. Endoscopy 2012; 44: 807-812
• 2 Tai BC, Machin D, White I et al. Competing risks analysis of patients with osteosarcoma: a comparison of four different approaches. Stat Med 2001; 20: 661-684
• 3 Gray RJ. A class of K-sample tests for comparing the cumulative incidence of a competing risk. Ann Stat 1988; 16: 1140-1154
• 4 van Hooft JE, van Montfoort ML, Jeurnink SM et al. Safety and efficacy of a new non-foreshortening Nitinol stent in malignant gastric outlet obstruction (DUONITI study): a prospective, multicenter study. Endoscopy 2011; 43: 671-675
• 5 Fine JP, Gray RJ. A proportional hazards model for the subdistribution of a competing risk. J Am Stat Assoc 1999; 94: 496-509
• 6 Anderson JR, Cain KC, Gelber RD. Analysis of survival by tumor response. J Clin Oncol 1983; 1: 710-719