LAUSR

To the Editor: We appreciate the authors’ insights into the clinical outcomes of open versus minimally invasive surgery for esophageal cancer, with results obtained from a large Japanese inpatient database. By propensity score (PS) matching, Sakamoto and colleagues showed that patients who underwent minimally invasive esophagectomy (MIE) experienced better in-hospital survival and a lower incidence of anastomotic leakage and reoperation. They also showed that, compared to open esophagectomy, MIE was associated with a shorter hospital length-of-stay but a significantly longer mechanical ventilation period, and longer anesthesia time. Although they openly discuss the limitations of their findings, there are 2 critical methodological issues that require further discussion. First, the authors used PS matching to account for clustering patients within hospitals. The PS matching method has been used for balancing background characteristics and controlling for selection biases in retrospective studies. However, PS matching to analyze clustered or hierarchically structured data remains controversial. Ignoring or misspecification of a potential cluster effect can introduce severe bias caused by the exemption of cluster-level covariates. To handle the clustering effects, it may be more methodologically reasonable for the authors to consider within-cluster matching using a global PS or create a 2-level hierarchical model with the unique hospital number as a random effect to account for clustering of patients by hospital. Regarding the impact