LAUSR

Editor We congratulate Saunders and colleagues1 on their large, single-centre study demonstrating that major postoperative complications, but not anastomotic leaks, impact on long-term survival. However, we would like to comment on several issues regarding their analysis which might be of interest to the readership. First, this is an extremely heterogeneous population, including a range of procedures from pharyngolaryngooesophagectomy for proximal oesophageal cancers to distal gastrectomy for distal gastric cancers. Owing to the varied morbidity and mortality profile between these different surgical approaches, type and location of cancers, and severity and timing of management of postoperative anastomotic leak, it is possible that the impact of complications on long-term outcome may differ. However, none of these is explored in the current paper. As such, a subgroup analysis assessing long-term outcomes depending on the operation performed (oesophagectomy versus gastrectomy) or the severity and location of the anastomotic leak should be performed. Second, there is selection bias in procedure choice (i.e. Ivor-Lewis, McKeown) not adjusted for within the multivariate analysis in Table S3 (supporting information). For example, assessment of the impact of postoperative complications will be better understood by adjusting for co-variates such as tumour location and histology, extent of lymph node dissection and yield, and type of anastomosis. A metaanalysis of more than 74 000 patients undergoing oesophagectomy identified that the majority of factors associated with anastomotic leaks were mainly patient-related2. Hence, it remains unclear if these factors were adjusted in the current paper. For these reasons, results from this study must be interpreted with caution. Moving forward, improved patient selection, prehabilitation programmes,