We screened all patients who underwent lung transplantation from January 2016 to June 2021 and focused on double lung transplant recipients mechanically ventilated through endotracheal tubes at ICU admission. Criteria for exclusion were: 1) post-operative venoarterial/venovenous extracorporeal membrane oxygenation (ECMO), 2) single-lung transplantations, 3) re-transplantations, 4) extubation delay for reasons other than weaning unreadiness and 5) re-intubation for reasons other than post-extubation respiratory failure.

Lung protective mechanical ventilation is applied to all patients after transplant to prevent ventilator-induced lung injury [9–11], a factor that may contribute to PGD [12], which is particularly relevant to undersized allografts [13–15]. Protective ventilation involves tidal volumes of 6–8 mL/kg recipient-predicted body weight (PBW), plateau pressures of less than 30 cmH₂O, and positive end-expiratory pressure (PEEP) adjusted to F_iO₂ without exceeding 12–14 cmH₂O [16]. Once the level of sedation has been reduced and patients are deemed ready for ventilator weaning, they are placed on assisted ventilation, providing continuous positive airway pressure (CPAP) with low pressure support. The next step is extubation if pressure support ventilation is maintained for several hours without signs of respiratory distress. When extubated, non-invasive respiratory support may be applied in response to clinical signs of respiratory muscle fatigue and blood gas analysis showing impaired gas exchange, including nasal high-flow cannula (NHFC) and non-invasive mask ventilation (NIV) [17]. Tracheostomies are performed in cases of re-intubation or when patients cannot sustain pressure support ventilation, preventing extubation.

Patients’ baseline demographics, clinical characteristics (e.g., pulmonary hemodynamics before transplant), comorbidities, procedural features in transplant (e.g., the predicted total lung capacity (pTLC) ratio [18]), and donor parameters were collected from the hospital’s electronic and Eurotransplant database. Research team members examined the radiologists’ reports for terms congruent with PGD, including pulmonary edema, reperfusion injury, and primary graft dysfunction, determining the P/F ratio based on P_aO₂ and FiO₂. The 2016 ISHLT definition was used to diagnose PGD [2], focusing on PGD grades at T+72 h after transplantation among patients still receiving invasive mechanical ventilation.

We assessed the trajectories and duration of invasive ventilation after transplant (including extubations, re-intubations, and tracheostomies), and we determined the time between ICU admission and weaning readiness, defined as the first transition from pressure-controlled to pressure-support ventilation. Ventilatory variables with corresponding arterial blood gas analyses (ABGs) were recorded in each patient during pressure-controlled ventilation up to 36 h after entering the ICU or until extubation (if performed within 36 h of admission), with the median of these values used for the analysis. We also analyzed non-invasive respiratory support (NHFC and NIV) following (first) extubation, studying the number of patients treated with these techniques and the duration of treatment. Finally, we assessed the number and percentage of transplant recipients dying in the ICU.

Pressure-controlled ventilation (Evita XL, Dräger Medical GmbH, Lübeck, Germany) was used via endotracheal tubes on all patients admitted to the ICU after transplantation. Variables collected included inspired oxygen fraction (F_iO₂), respiratory rate, tidal volume (VT), minute ventilation, peak inspiratory airway pressure (P_peak), and PEEP, with the following indexes calculated: P/F ratio (quotient of partial pressure of oxygen to inspired oxygen fraction), dynamic driving pressure (∆P_aw; defined as P_peak–PEEP), dynamic lung-thorax compliance (LTC_dyn; defined as VT/∆P_aw), ventilatory ratio (VR; a measure of ventilatory efficiency correlating with the pulmonary dead space fraction) [19], and MP [6] utilizing a simplified formula for pressure-controlled ventilation [20].

During artificial ventilation, MP is the pressure-volume work per minute provided by the ventilator, as determined by respiratory rate, tidal volume, and applied airway pressure. To account for lung dimensions interacting with MP, we normalized power to the recipients’ predicted body weight (PBW-MP) [21] and LTC_dyn (LTC_dyn-MP) [7, 8]. PBW correlates with the total lung capacity of a healthy individual. Dynamic compliance with its temporal changes is a surrogate of ventilated lung volume, accounting for the force required to overcome the respiratory system’s resistance and elastance, which is critical during ventilator weaning [7, 8].

On a physical level, normalizing MP to lung volume surrogates determines the intensity of mechanical stress exerted on the respiratory system based on applied airway pressure and respiratory rate, referred to as mechanical “power density”. Power density, previously termed “specific mechanical power” [22], is equivalent to tidal “strain energy density” per minute, which refers to the stored energy per unit of volume resulting from a body’s linear elastic deformation [23]. Consequently, power density describes the time rate of energy transfer per unit of volume displaced and is inherently equivalent to pressure application per time. A detailed description of the calculated ventilatory indexes is provided in the online supplement (see Supplementary material).

Each period of invasive mechanical ventilation (by endotracheal tube or tracheal cannula) was summarized to determine the total “duration of invasive ventilation” after transplant considering four clinical scenarios: 1) patients successfully extubated at the first attempt, 2) patients successfully extubated at the second or third attempt (refers to individuals experiencing extubation failure with re-intubation), 3) patients tracheotomized after one or more failed extubations, and 4) patients tracheotomized without previous extubation (referred to as primary tracheostomy) (see Supplementary Figure S1).

Although “prolonged mechanical ventilation” after lung transplantation has no universal definition [4, 5], the present study refers to PMV as invasive ventilation lasting longer than 72 h based on previous reports demonstrating that most patients were extubated within this period [4, 24].

“Weaning readiness” refers to the first transition from pressure-controlled to pressure-support ventilation, typically indicating the point where spontaneous breathing with subsequent extubation seems feasible. “Non-invasive respiratory support” refers to applying NHFC and/or NIV within 24 hours and for at least six hours daily following (first) extubation.

The study’s primary outcome was the correlation between the duration of invasive ventilation and MP. Moreover, the ability of MP and other ventilatory indexes to predict PMV and the evaluation of patients according to the level of respiratory support required after surgery (including those with post-extubation non-invasive respiratory support and primary tracheostomy) were secondary outcomes.

Descriptive and frequency statistics summarized patients’ baseline characteristics. A Chi-square or Fisher’s exact test was used when comparing categorical variables. Depending on the continuous variables’ homogeneity of variance, determined by the Kolmogorov-Smirnov normality test, differences were analyzed through Student’s t-test or Mann-Whitney U-test. Similarly, we performed bivariate comparisons of the ventilatory variables and indexes in recipients with and without post-transplant PMV and between patients requiring varying levels of respiratory support following surgery. We conducted a (factorial) repeated-measures analysis of variance (ANOVA) to detect within-subject time effects and between-group differences in ventilatory variables’ trajectories after transplant.

We correlated the duration of invasive ventilation after lung transplant with MP and other ventilatory indexes using Spearmans’ correlation coefficient (ρ). The parameters’ ability to predict PMV was assessed through receiver operating characteristic (ROC) curve analysis, with diagnostic accuracy expressed as the area under the ROC curve (AUROC) [25]. In the absence of an external test cohort, we performed 2-time repeated, 5-fold cross-validation to evaluate the internal validity of these indexes [26]. We performed a multivariable binary logistic regression analysis to determine whether MP was independently associated with PMV. Recipient, transplant, and donor characteristics deemed clinically relevant a priori and those with a p-value of less than 0.2 in the univariable analysis have been utilized as input variables for multivariable model development. We performed a forward selection of variables, with Hosmer & Lemeshow test and Nagelkerke R² employed to evaluate the model’s goodness-of-fit. Probabilities were expressed as odds ratios (OR) with 95% confidence intervals (95%CI).

We conducted sensitivity analyses by either redefining the threshold for prolonged ventilation (to more than 96 h and 7 days, respectively) or restricting the analysis to patients following a particular clinical scenario (as described above) [27].

Since there were no comparable studies on prolonged ventilated lung transplant recipients to determine sample size, we recruited patients to the maximum extent possible. We performed two-tailed tests; statistical significance was indicated by p < 0.05. The analyses were conducted with MedCalc^® statistical software v20.305 (MedCalc Software Ltd, Ostend, Belgium; 2023).

Baseline characteristics between PMV and non-PMV subjects differed regarding pulmonary hemodynamics before transplant, provision of intraoperative ECMO, and number of red cell transfusions received during transplant. PMV was more prevalent in females (58.5% versus 32.3%, p < 0.01), accounting for reduced recipients’ and donors’ heights and predicted body weights within the PMV group (Table 1).

PMV (>72 h) occurred in 82 patients (35%), with invasive ventilation totaling a median of 189 h [IQR 108–666 h] versus 32 h [21–45 h] in the non-PMV group (Table 2). In addition, patients were ventilated for more than 96 h and more than 7 days in 28% and 18%, respectively.

For PMV, the time between admission to the ICU and weaning readiness (as defined above) and the time to extubation was longer. A higher proportion of these patients also underwent primary tracheostomy, which resulted in fewer extubations. Among PMV subjects extubated, the percentage receiving non-invasive respiratory support was higher (61.4% versus 38.7%, p < 0.01). They also required more re-intubations (within a median of 48 h [32–90 h] following extubation) and tracheostomies (performed after a median of 11 days [7–14 days]), resulting in longer lengths of stay in the ICU (Table 2).

Overall, 2079 ventilatory variables with respective ABGs were collected over a median of 17.7 h [10.0–31.0 h] following ICU admission. Lung protective ventilation using low tidal volumes per recipients’ and donors’ PBW was provided to the PMV and non-PMV group (6.3 mL/kg [5.6–6.8 mL/kg] vs. 6.2 mL/kg [5.6–6.7 mL/kg], p = 0.409; and 5.7 mL/kg [5.1–6.7 mL/kg] vs. 5.8 mL/kg [5.2–6.4 mL/kg], p = 0.836).

PMV patients showed lower median P/F ratios (239 mmHg [184–308 mmHg] vs. 296 mmHg [254–345 mmHg], p < 0.01), higher ventilatory ratios (1.30 [1.17–1.50] vs. 1.18 [1.04–1.32], p < 0.01) and lower dynamic compliance (27 mL/cmH₂O [22–31 mL/cmH₂O] vs. 32 mL/cmH₂O [28–38 mL/cmH₂O], p < 0.01). Since minute ventilation was not different between groups, these patients also displayed higher MP (18.1 J/min [13.8–21.2 J/min] vs. 14.7 J/min [12.1–17.3 J/min], p < 0.01) owing to higher peak pressures. Consequently, power density was higher for PMV (e.g., LTC_dyn-MP 7049 cmH₂O²/min [5,255–8,299 cmH₂O²/min] vs. 4,432 cmH₂O²/min [3,781–5,889 cmH₂O²/min], p < 0.01) (Table 3).

Figure 2 illustrates the ventilatory variables of interest at different time points, stratified by group membership. According to the ANOVA, MP and power density declined consistently within both groups and similar to bivariate comparisons, there were significant differences in trajectories of variables between the PMV and non-PMV group following transplant (MP: F[1, 235] = 27.3, p < 0.01; LTC_dyn-MP: F[1, 235] = 66.6, p < 0.01) (see Supplementary Tables S1, S2).

MP demonstrated a poor correlation with the duration of invasive ventilation (Spearmans’ coefficient 0.252 [95% CI 0.129–0.368]), which is consistently stronger with power density (ρ = 0.452 [0.345–0.548] for LTC_dyn-MP) (Figure 3, see Supplementary Table S3). There was a poor diagnostic performance in predicting PMV for MP (AUROC 0.66 [95% CI 0.60–0.72], DOR 5.6, MCC 0.29), significantly increasing with power density (AUROC 0.78 [0.72–0.83], DOR 15.0, MCC 0.44 for LTC_dyn-MP) (Figure 4; Table 4, see Supplementary Table S4; Supplementary Figure S2). The sensitivity analysis based on the 96-h and 7-day thresholds for defining prolonged ventilation or analyzing only 186 cases successfully extubated at the first attempt provided similar results (see Supplementary Tables S5–S8).

Patients were stratified by whether they had received post-extubation NHFC/NIV and compared to patients with primary tracheostomy. The respiratory mechanics of these three groups differed significantly, with tracheotomized patients demonstrating less favorable mechanics and outcomes (e.g., lower dynamic compliance, higher power density, and a longer duration of invasive ventilation). Similarly, extubated patients requiring non-invasive respiratory support showed higher power densities than those receiving pure oxygen insufflation. Re-intubations and tracheostomies were also significantly higher (see Supplementary Table S9).

In univariable analysis, MP and power density (LTCdyn-MP) were independently related to PMV. After adjusting for recipient, transplant, and donor characteristics, only power density (OR 1.54 per cmH₂O²/min * 10⁻³ [95% CI 1.30–1.83]; p < 0.01) remained in the multivariable model, along with female gender (OR 2.45 [95% CI 1.29–4.66]; p < 0.01), recipients’ mean pulmonary artery pressure before transplant (OR 1.05 mmHg⁻¹ [95% CI 1.02–1.09]; p < 0.01), and P/F ratio after transplant (OR 0.57 mmHg⁻¹ * 10⁻² [95% CI 0.36–0.88]; p = 0.012) (see Supplementary Tables S10, S11).

There were 18 out of 237 patients with PGD grade 3 (7.6%), 28 patients with PGD grade 2 (11.8%), and 11 patients with PGD grade 1 (4.6%). Moreover, we identified a subgroup of 22 (9.3%) mechanically ventilated patients without PGD having a similar duration of mechanical ventilation, P/F ratios, MP, and power density levels. The patients with PGD grade 3 compared with grades 1/2 had longer durations of invasive ventilation and fewer extubations, despite no significant differences in baseline characteristics, donor parameters, or ventilatory indexes (see Supplementary Table S12).

PMV was more prevalent in females; they also experienced more re-intubations and tracheostomies, and respiratory mechanics significantly differed from male recipients. Females demonstrated lower tidal volumes and higher airway pressures (along with lower dynamic compliance), but MP was higher in males owing to substantially higher tidal volumes (despite lower airway pressure). After normalizing MP to dynamic compliance, females exhibited significantly higher power density levels (see Supplementary Table S13).

This study has limitations. First, nearly half of the patients screened were excluded, mainly due to single lung transplantation and requiring ECMO after surgery, resulting in a highly selected group of patients and limiting the generalizability of the results. However, given that both conditions impact either the respiratory systems’ mechanical properties or pulmonary gas exchange, this step was necessary to determine whether there is a relationship between early graft function, MP, and PMV. Second, although we performed cross-validation, it is uncertain whether the results apply to other centers due to the study`s retrospective nature, monocentric design, and lack of an external validation group. Third, ventilatory variables were recorded during pressure-controlled ventilation, but we cannot rule out that some patients already had spontaneous breathing activity, affecting measurements of some variables (e.g., tidal volume) and consequently distorting values for dynamic compliance and MP. Fourth, since this was a retrospective analysis restricting the available ventilatory variables, we used a simplified equation for MP calculation [20], and correlations and predictions of PMV might have been more accurate with a more comprehensive formula [45]. Fifth, we did not assess primary graft dysfunction systematically at the designated time points T0, T+24, T+48, and T+72 [2]. However, evidence suggests that PGD scores are ineffective at predicting PMV [4], which might be related to difficulties in adequately grading PGD and the fact that scoring does not provide a physiological assessment of the allograft [46].

In double lung transplant recipients without postoperative extracorporeal support, PMV was observed in 35% of cases and related to impairments in respiratory mechanics and pulmonary gas exchange. A significant correlation was found between MP and the duration of post-transplant invasive ventilation, with power density showing stronger correlations and predicting PMV more accurately. These indexes may assist clinicians in identifying patients at risk for PMV, which is associated with inferior outcomes such as increased re-intubation and tracheostomy rates. Clinical strategies may change in the event of persistently high power density after transplant, such as providing early tracheostomy or prophylactic non-invasive respiratory support following extubation. Findings need to be confirmed in future studies and whether differences in MP may ultimately impact long-term clinical outcomes after lung transplantation.