Truong et al. [1] suggested that our findings were confounded by a selection bias. The cohorts of the study were balanced using propensity score matching. Multiple comorbidities were selected to be balanced; before running the analysis, the two cohorts were matched by selectively removing individuals to render the differences for the selected comorbidities statistically nonsignificant. The selection of comorbidities to balance was based on the typical comorbidities of obstructive sleep apnea (OSA) patients reported in the literature [2, 3]. However, it is impossible to prove empirically that a full set of cofounders has been included in the propensity score matching model [4], especially in a retrospective real-world data study. We considered that it was also essential to avoid adding too many cofounders to the propensity score matching model, which could have led to “overfitting” the data and a decrease in the population representativeness of the sample [5]. Only patients with data going back at least three months before the procedure were included in the electronic health records to minimize bias due to data incompleteness.

The study’s index event was the initiation of continuous positive airway pressure (CPAP)/surgery after OSA diagnosis, and the 5-year follow-up also included the beginning of treatment.