Redo coloanal anastomosis for anastomotic leakage after low anterior resection for rectal cancer: an analysis of 59 cases

We thank the authors for their comments on our article in Colorectal Disease.(1) The authors address the methodological issues caused by the relatively small sample size and the subsequent low number of events presented in our study. We acknowledge that the confidence interval for the association between anastomotic leakage of the redo anastomosis and subsequent failure is wide, and that this is probably due to sparse data bias. This is illustrated by the 2x2 table for anastomotic leakage and success rate (Table 1). The authors referred to Greenland et al., who suggested penalization as an optimal statistical method to address this problem.(2) Simple Firth penalization, by adding 0.5 to each cell, resulted in an odds ratio of 0.026 for the univariable analysis for anastomotic leakage compared to 0.020 before penalization, which indeed indicates the bias as addressed by the authors.(3) Meanwhile, we have only included two variables in our multivariate analysis, with ten events per variable, which leads to a limited number of different combinations between the dependent and independent variables, therefore limiting the risk of bias for the multivariable analysis. We agree that the outcomes of our regression analysis should be interpreted with great caution. Nevertheless, the association between new anastomotic leakage and failure of the anastomosis remains plausible.. This article is protected by copyright. All rights reserved