The REAL (REctal Anastomotic Leak) score for prediction of anastomotic leak after rectal cancer surgery

Abstract

We would like to thank Dr. Reddavid and colleagues [1] for their letter that allows us to clarify some of the choices we had to make, while performing the literature review and data-analysis that led us to the conception of the REAL score [2]. Dr. Reddavid is certainly right when she affirms that the individual participant data meta-analysis did not take into consideration the influence of the experience of surgeons on the outcomes in rectal cancer surgery. It is true that half of the centres included have a low cumulative volume (< 20 cases per year), but they account for only 1035 cases out of 9735 (10.6%). Although some difference in results could be expected in determining the accuracy of the threshold value, the subgroup analysis shows just a slight increase in specificity (Table 1). Nevertheless, we appreciated the comment and replaced the data online with the new data (www. real-score .org). The second point concerns the potential influence of any independent factors such as malnutrition, anastomotic technique (end-to-end vs side-to-end, j-pouch vs straight colorectal anastomosis) and intraoperative contamination. We are aware of the paper by Asteria et al. [3]. Unfortunately, in retrospective databases, similar data are often missing. We preferred the inclusion of a high number of individuals, rather than a selection of few individuals with a complete database. In fact, this makes it possible to verify on a large scale the influence of even a small number of variables. To pursue this, we have used a multiple imputation algorithm strategy. With multiple imputation algorithms, it is possible to produce different complete datasets from incomplete data by imputing missing data several times with appropriate functions, depending on variable type. Each complete data set is analyzed using a complete data method and the results are then combined to obtain the inferential syntheses. Nevertheless, in some cases, this was evidently not reasonable and we had to skip several variables. Dr. Reddavid also raises the issue of the role of laparoscopy and risk of anastomotic leak. In fact, we could have analysed which was the initial approach, laparoscopic or open, and the eventual conversion to open surgery. Unfortunately, the number of missing data about conversions, which play a major role in the outcomes, led us to ignore this variable. Moreover, our aim was to assign a risk factor to each variable we were able to study objectively and this way to build a score assessing the risk of an anastomotic leak which would help surgeons to decide whether to protect a colorectal anastomosis with a stoma. Further studies are surely mandatory in this field, and we are happy to announce that the Italian Society for Endoscopic Surgery (SICE) is just about to start a prospective evaluation of the REAL score, including the collection of all the variables cited above as missing, to allow a better preoperative evaluation of the risk of anastomotic leak.