A retrospective two-center cohort analysis was performed at the University Hospitals Leuven (UZL, Belgium) and Motol University Hospital in Prague (FNM, Czech Republic). Adult patients (≥18 years old) who underwent LuTx between 01/2010 and 12/2022 (Leuven) and 01/2018–03/2023 (Prague) were included. The final date of follow-up was October 31, 2023. The different inclusion periods reflected center-specific availability of complete and consistently retrievable data for the variables required for this analysis, particularly donor PFR, PGD grading, and post-transplant outcomes. Retransplant patients and combined organ transplant patients were excluded (Figure 1). Patients with missing key data were also excluded. All recipients provided written informed consent to approve use of medical data for research. The study was approved by the research ethics committee UZ/KU Leuven (S68554) and FN Motol (EK-530/21) and adhered to the principles of the Declaration of Helsinki and is compliant with the International Society for Heart and Lung Transplantation (ISHLT) Ethics statement.

Donor and recipient demographics as well as surgical and postoperative parameters were selected based on prior studies [15–20].

Donor and preservation variables included: donor type (donation after brain death [DBD] vs. DCD), age, sex, body mass index (BMI), duration of ventilation, duration of ex-vivo lung perfusion (EVLP), total ischemic time of first/second lung and the donor PFR. EVLP was used according to center practice for marginal/uncertain lungs, including impaired oxygenation, DCD/ECD features, radiographic findings, or other clinical concerns. Although donor smoking and heavy alcohol use were associated with an increased risk of PGD [21, 22], they were not included as variables in our analysis due to the large number of missing data.

Recipient variables were: age, sex, BMI, indication for transplantation, type of transplant, surgical approach, time on post-operative ECMO (extracorporeal membrane oxygenation, days), PGD grade at 0, 24, 48 and 72 h, length of intensive care unit (ICU) stay (days), length of hospital stay (days), grade and incidence of acute cellular rejection (ACR) within the first post-transplant year, time to CLAD or death (due 9/2022 for UZL and 9/2023 for FNM) and patient survival. A 2-year survival endpoint was selected to reflect early to intermediate-term outcomes influenced by donor and perioperative factors, while reducing confounding from long-term comorbidities. The immunosuppression protocol followed the profile previously described by Provoost et al. (Chapter 2.4) [23].

Primary endpoints were PGD grade 3 at 24 and at 72 h and 2-year patient survival. Secondary endpoints were the incidence of ACR within the first year after transplantation and CLAD-free survival. PGD was diagnosed based on the ISHLT consensus and was assessed by pulmonary edema on X-ray and the ratio of arterial partial pressure of oxygen to the fraction of inspired oxygen (PaO₂/FiO₂) [24]. Data on blood gases were acquired by extraction of electronic patient files. Early PGD was defined as PGD grade 3 at 24 h, late PGD as grade 3 at 72 h. Histological diagnosis of ACR was made using tissue samples obtained from transbronchial biopsy (either by forceps or cryobiopsy), through flexible bronchoscopy. The biopsies underwent evaluation based on ISHLT guidelines [25]. (Supplementary Material 1).

Donor PFR was defined as a ratio of arterial oxygen tension (PaO2) to fractional inspired oxygen (FiO2). No a priori categorization of donor PFR (e.g., <300 vs. ≥ 300) was used to define comparison groups; donor PFR was modeled as a continuous predictor in all primary analyses. For the Leuven cohort, the PFR was defined as the last value reported by Eurotransplant at the time of organ allocation. For the Prague cohort, it was the last value measured prior to procurement. All donor arterial blood gases were obtained after 10 min of ventilation at an FiO₂ of 100% and a positive end-expiratory pressure of 5 cm H₂O, with lung recruitment maneuvers performed beforehand when indicated. Donor PFR in DCD donors was defined using the same center-specific principle as in DBD donors. All donor PFR values were reviewed for plausibility and retained in the analysis if they were documented under the predefined ventilatory conditions and if the lungs were ultimately accepted and transplanted.

Statistical analyses were performed using SAS9.4 (Windows) by an experienced biostatistician (ABe). Continuous variables were reported as median (interquartile range), categorical variables as number (percentage). Hospital and ICU duration were summarized using Cumulative Incidence Functions (CIF) to account for the competing risk of in-hospital/ICU death.

Spearman correlations (ρ) and Kruskal–Wallis tests were used to evaluate relations between covariates and PFR.

Associations between P/F ratio and binary outcomes were assessed using logistic regression analyses (PGD>=3, ECMO, Any ACR>0 or 1). Survival outcomes (all-cause death and CLAD-free survival) were analyzed using Cox regression. Competing risk outcomes (ICU and hospital duration) were analyzed by means of a Fine&Gray model. ACR results over time were analyzed using a generalized estimating equations (GEE) multinomial model, while the number of ACR assessment >0/1 was assessed using a multinomial logistic model.

For all of the above models and outcomes, the linearity assumption was assessed for all continuous predictors by checking for improvement in the model when fitting the variables using non-linear functions (log-transformed, quadratic and restricted cubic spline) using Akaike’s Information Criterion (AIC).

The association between PFR and outcome was assessed using univariate models. Multivariable model was used that included all known predictors, to assess whether the association between PFR and outcome was independent from known predictors. As a sensitivity analysis, to assess the stability of the obtained association, a backward model reduction was performed on the known predictors (using p < 0.157 which corresponds to a decrease in AIC) before adding PFR to the model.

Data inspection revealed that the availability of ACR assessments was associated with outcome, i.e., missingness was not at random. Therefore, ACR data are analyzed similar to a pattern-mixture model whereby the number of assessments was added as a factor to the model [26].

All tests were 2-sided and assessed at a significance level of 5%. Due to the retrospective and exploratory nature of the study, no formal adjustment for multiple comparisons was applied; secondary analyses were interpreted descriptively and considered hypothesis-generating.

A total of 846 LuTx procedures were performed in Leuven between January 2010 and December 2022. In Prague, 245 LuTx between January 2018 and March 2023. After exclusion of 88 ineligible procedures, 1,003 eligible double or single LuTx remained. In 19 cases, important data were missing, resulting in 984 patients included in the analysis (Figure 1A). Median follow-up was 4 (1.6–7.4) years.

Recipient and donor demographics are summarized in Table 1. Recipient age was 57 (49–62) years of which 531 (54%) were male and 453 (46%) female. The most common indication for transplantation was chronic obstructive pulmonary disease, accounting for 492 (50%) cases, followed by pulmonary fibrosis (n = 299 [30%]), and cystic fibrosis (n = 106 [11%]). Recipient BMI was 23 (20–27) kg/m2.

Donors were aged 51 (39–61) years, with a BMI of 24 (22–27) kg/m². The donor population comprised 541 (55%) male and 443 (45%) female donors. DCD procurement was performed in 181 (18%) cases. In the present cohort, only three (1.6% of DCD) DCD donors were classified as Maastricht category IV, while all remaining DCD donors were category III [27]. Median time on ventilator was 3 (2–6) days. Median donor PFR was 435 (380–492). The distribution of PFR in 20-mmHg intervals is illustrated in Figure 1B. Supplementary Figure S1 shows the median and range (maximum and minimum) of PFR by year across the study period. No donor was supported with VA-ECMO at the time of donor PFR assessment.

Median total ischemic time for the first and second implanted lung was 226 (182–275) min and 382 (325–447) min, respectively. EVLP was performed in 26 (3%) cases and lasted 250 (164–316) min. Among donors managed with EVLP in the present cohort, donor PFR values ranged from 311 to 561 mmHg. Therefore, EVLP was not triggered exclusively by a PFR threshold below 300 mmHg, but was used as part of a broader clinical evaluation of donor lung suitability. Double LuTx was performed in 973 (99%) and single LuTx in 11 (1%) cases.

One hundred thirty-nine (14%) LuTx recipients developed PGD 3 at 24 h and 115 (12%) had PGD 3 at 72 h. Two hundred seventy-eight (29%) patients developed PGD 3 within the first 72 h. Transplant outcomes are summarized in Table 2. ICU stay was 8 (5–16) days. Post-operative ECMO was required in 79 (8%) patients and lasted 4 (2–6) days. At least one episode of ACR in A grade was observed in 293/843 (35%) cases. The total length of in-hospital stay was 28 (22–41) days. One, 2-, and 5- year patient survival was 89%, 84%, and 75%, and incidence of CLAD or death was 10%, 19%, and 36%, respectively. Because of differences in inclusion period and perioperative practice between centers, center-specific descriptive data are provided in Supplementary Table S1.

Relations between potential confounders and PFR were explored (Table 3). A negative correlation was observed between PFR and increasing age of both recipient and donor (ρ = −0.0786, p = 0.0136 and ρ = −0.0871, p = 0.0062), as well as increasing BMI of the recipient and donor (ρ = −0.0638, p = 0.0453, and ρ = −0.2634, p<0.0001) and longer time on ventilator (ρ = −0.0852, p = 0.0075). Lower PFR was observed when recipient was male (p = 0.0469).

Several statistical approaches were employed to reveal potential association of last donor PFR and severity and frequency of ACR grade A in LuTx recipients. No independent association was identified when comparing PFR with ACR outcome, with at least 1 ACR grade A > 0, at least 1 ACR grade A > 0 for patients alive at 6 months, with at least 1 ACR grade A > 1 or number of ACR grade A > 0 or >1 (Table 5).

Neither univariable (HR 0.999, 95% CI [0.997–1.000], p = 0.1065) nor multivariable (HR 1.000, 95% CI [0.998–1.001], p = 0.5404) analysis revealed a statistically significant association between PFR and overall survival at 2 years post-LuTx. Several correlations of overall survival with other variables were observed (Table 6). Especially, donor and recipient age were significant in both univariable (p = 0.0003 and p < 0.0001) and multivariable analysis (p = 0.0341 and p < 0.0001). Similarly, CLAD-free survival did not demonstrate a significant correlation with PFR (Supplementary Table S4).