Ventilation parameters and early graft function in double lung transplantation

Background: Currently, the primary graft dysfunction (PGD) score is used to measure allograft function in the early post-lung transplant period. Although PGD grades at later time points (T48 hours and T72 hours) are useful to predict mid- and long-term outcomes, their predictive value is less relevant within the first 24 hours after transplantation. This study aimed to evaluate the capability of PGD grades to predict prolonged mechanical ventilation (MV) and compare it with a model derived from ventilation parameters measured on arrival at the intensive care unit (ICU).

Methods: A retrospective single-center analysis of 422 double lung transplantations (LTxs) was performed. PGD was assessed 2 hours after arrival at ICU, and grades were associated with length of MV (LMV). In addition, peak inspiratory pressure (P_IP), ratio of the arterial partial pressure of oxygen to fraction of inspired oxygen (P/F ratio), and dynamic compliance (cDyn) were collected, and a logistic regression model was created. The predictive capability for prolonged MV was calculated for both (the PGD score and the model). In a second step, the created model was externally validated using a prospective, international multicenter cohort including 102 patients from the lung transplant centers of Vienna, Toronto, and Budapest.

Results: In the retrospective cohort, a high percentage of extubated patients was reported at 24 hours (35.1%), 48 hours (68.0%), and 72 hours (80.3%) after transplantation. At T0 (time point defined as 2 hours after arrival at the ICU), patients with PGD grade 0 had a shorter LMV with a median of 26 hours (interquartile range [IQR]: 16-47 hours) than those with PGD grade 1 (median: 42 hours, IQR: 27-50 hours), PGD grade 2 (median: 37.5 hours, IQR: 15.5-78.5 hours), and PGD grade 3 (median: 46 hours, IQR: 27-86 hours). However, IQRs largely overlapped for all grades, and the value of PGD to predict prolonged MV was poor. A total of 3 ventilation parameters (P_IP, cDyn, and P/F ratio), determined at T0, were chosen on the basis of clinical reasoning. A logistic regression model including these parameters predicted prolonged MV (>72 hours) with an optimism-corrected area under the curve (AUC) of 0.727. In the prospective validation cohort, the model proved to be stable and achieved an AUC of 0.679.

Conclusions: The prediction model reported in this study combines 3 easily obtainable variables. It can be employed immediately after LTx to quantify the risk of prolonged MV, an important early outcome parameter.