Effect of VV-ECMO as a BTT on postoperative 30-day outcomes and 1-year mortality - primary matched subgroups analysis
After matching, BTT and non-BTT patients were well-balanced with respect to the age, preoperative laboratory characteristics and main diagnosis (all standardized mean differences [d] <0.1; Table 3), but there was still imbalance (d >0.1) in the proportion of female patients, BMI and hemoglobin levels (lower in BTT patients) (Table 3). Intraoperatively, CPB was used less often, while the use of ECMO was considerably higher in BTT patients than in non-BTT patients after matching (Table 3). Regarding 30-day outcomes, need for postoperative ECMO (73.0% vs. 8.6%), delayed chest closure (11.9% vs. 6.3%) and incidence of AKI requiring RRT (63.6% vs. 29.7%) were higher in BTT vs. non-BTT patients (Table 3). However, chest drainage within 24 hours, incidence of surgical re-exploration, tracheostomy, chest infection, and 30-day mortality appeared similar in the two matched subgroups, while 1-year mortality was lower in BTT patients (Table 3). With further adjustment for the unbalanced covariates (gender, BMI and hemoglobin level), preoperative VV-ECMO support was associated with around 20-fold higher odds of postoperative ECMO (frequentist and Bayesian estimates; Table 4). It was also associated with around 4-fold higher odds of AKI requiring RRT: the Bayesian estimate (95%CrI 1.31-14.2) appeared robust (a rather high E-value indicated a rather low susceptibility to unmeasured confounding) and was more precise than the frequentist estimate (95%CI 0.43-39.2), leaving some uncertainty about this effect (Table 4). There was also a tendency of higher odds of tracheostomy (OR 2.3), but both frequentist and Bayesian estimates were imprecise (Table 4). VV-ECMO as a BTT did not appear associated with other 30-day outcomes including mortality or with 1-year mortality (Table 4).