Rationale: Pulmonary hypertension from pulmonary arterial hypertension or parenchymal lung disease is associated with an increased risk for primary graft dysfunction after lung transplantation.
Objective: We evaluated the clinical determinants of severe primary graft dysfunction in pulmonary hypertension and developed and validated a prognostic model.
Methods: We conducted a retrospective cohort study of patients in the multicenter Lung Transplant Outcomes Group with pulmonary hypertension at transplant listing. Severe primary graft dysfunction was defined as PaO2/FiO2 ≤200 with allograft infiltrates at 48 or 72 hours after transplantation. Donor, recipient, and operative characteristics were evaluated in a multivariable explanatory model. A prognostic model derived using donor and recipient characteristics was then validated in a separate cohort.
Results: In the explanatory model of 826 patients with pulmonary hypertension, donor tobacco smoke exposure, higher recipient body mass index, female sex, listing mean pulmonary artery pressure, right atrial pressure and creatinine at transplant, cardiopulmonary bypass use, transfusion volume, and reperfusion fraction of inspired oxygen were associated with primary graft dysfunction. Donor obesity was associated with a lower risk for primary graft dysfunction. Using a 20% threshold for elevated risk, the prognostic model had good negative predictive value in both derivation and validation cohorts (89.1% [95% confidence interval, 85.3–92.8] and 83.3% [95% confidence interval, 78.5–88.2], respectively), but low positive predictive value.
Conclusions: Several recipient, donor, and operative characteristics were associated with severe primary graft dysfunction in patients with pulmonary hypertension, including several risk factors not identified in the overall transplant population. A prognostic model with donor and recipient clinical risk factors alone had low positive predictive value, but high negative predictive value, to rule out high risk for primary graft dysfunction.
Lung transplantation is a therapeutic option for patients with pulmonary hypertension (PH) from parenchymal lung disease or pulmonary arterial hypertension (PAH) refractory to medical therapy. Unfortunately, severe pretransplant PH and PAH are associated with a two- to threefold increased risk for primary graft dysfunction (PGD), a form of acute lung injury that occurs within 72 hours of transplantation (1–3). The most severe form of PGD, grade 3 PGD, is associated with a longer duration of mechanical ventilation, increased intensive care unit length of stay, and higher 30-day mortality (1, 4, 5). Grade 3 PGD also increases the risk for bronchiolitis obliterans syndrome (5, 6).
Despite the link between PH and PGD, little is known about the clinical risk factors associated with the development of PGD within this specific patient population. Better understanding of these clinical risk factors may affect perioperative management and provide therapeutic targets in the future. We aimed to identify donor, recipient, and operative characteristics associated with the development of PGD among patients with PH and to develop a prognostic model for clinical use.
Some of the results of this study were presented in abstract and poster form at the 2015 International Society for Heart and Lung Transplantation conference, April 15–18, Nice, France (7) and American Thoracic Society International Conference, May 15–20, Denver, Colorado (8).
We included subjects enrolled in the ongoing, multicenter prospective Lung Transplant Outcomes Group cohort who underwent lung transplantation between March 1, 2002, and October 31, 2012 (2, 4, 9). The study sample consisted of transplanted subjects with a mean pulmonary artery pressure (mPAP) ≥25 mm Hg, measured by right heart catheterization at the time of listing for lung transplantation. A small number of subjects (11%) did not have preoperative right heart catheterization, and therefore perioperative hemodynamic measurements were substituted. The institutional review boards at each site approved our study, and written informed consent was obtained from each subject.
PGD grade was assessed prospectively, using the International Society for Heart and Lung Transplantation criteria, defined by the PaO2/FiO2 ratio and the presence of infiltrates within the allograft or allografts (1). Two physicians blinded to the clinical information independently interpreted each center’s chest radiographs with adjudication of conflicts by a third physician (kappa for consensus on subject-level grade 3 PGD classification = 0.95). The primary outcome was grade 3 PGD at 48 or 72 hours after reperfusion (herein referred to as PGD), which has been validated and used in previous observational studies (2, 4, 9).
Potential risk factors for PGD were selected a priori on the basis of previous studies or biological plausibility (2, 3, 6, 9–12). Body mass index (BMI) was calculated from measured height and weight and was assessed for inclusion as a linear variable and as a categorical variable. Race/ethnicity was grouped into three categories: Caucasian, African American, or other (including Hispanic and Asian Pacific Islander). Mechanism of donor death was categorized into head trauma, anoxia, stroke, and other, including blunt trauma and suicide. Pretransplant recipient diagnosis was categorized into five groups: chronic obstructive pulmonary disease (COPD), interstitial lung disease (ILD), cystic fibrosis, PAH, and other, including sarcoidosis and bronchiolitis obliterans syndrome.
Continuous variables were summarized with mean and standard deviation or median and interquartile range, where appropriate. Categorical variables were summarized by frequency and percentages. Differences in candidate prognostic variables between those with and without PGD were assessed with unpaired t, Mann-Whitney, or chi-squared tests. As with many prognostic studies based on clinical risk factors generated through regular clinical care, some subjects had missing values for one or more variables. Through the method of chained equations, we created 20 imputed datasets using multinomial, ordinal, and linear regression models for missing data (13–15). After the imputation process, out-of-range values were truncated to fall within the appropriate clinical range.
We evaluated all clinically meaningful donor, recipient, and operative candidate predictors based on our examination of their distribution and by preliminary assessments of their association with PGD. Donor, recipient, and operative risk factors with a P value ≤0.20 on bivariate logistic regression were considered for inclusion into the multivariable explanatory model for PGD. We defined a priori the potential interaction between donor and recipient BMI. Collinearity was assessed using Pearson and Spearman correlation coefficients for continuous measures and by cross-classification for categorical factors. To generate a parsimonious multivariable model, covariates that were not confounders based on a less than 20% change in odds ratio (OR) were eliminated by purposeful backward selection. Because subsequent models were nested, the global fit of each model was assessed using the likelihood ratio test. To account for relatedness among subjects at each center, a second explanatory model was created using conditional logistic regression stratified by transplant center. Confidence intervals for point estimates considered the additional variance arising from the imputation of missing values. We used marginal standardization to estimate average risk for PGD from the final explanatory logistic regression model for selected categorical variables and for the continuous variable mPAP.
For the testing and validation of a prognostic model (better suited for clinical use at the bedside), participants were randomly assigned to the derivation or validation cohort in a 1:1 ratio. Model generation in the derivation cohort proceeded using donor and recipient risk factors and the same methods as outlined earlier. The estimates of the predictors identified in the derivation cohort were used to calculate the predicted risk for PGD in the validation cohort. Performance characteristics of the model were assessed. We also evaluated the discriminative ability of the model by using the c-statistic. Overall point estimate and confidence intervals for the performance characteristics were generated using Rubin’s method (16). Reporting of the prognostic model follows the items listed in transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD) (see Table E3 in the online supplement) (17).
A P value <0.05 was used to indicate statistical significance. Statistical analyses were performed using STATA software version 12.0 through 14.1 (StataCorp LP, College Station, TX).
During the study period, 1,678 subjects underwent lung transplantation in the Lung Transplant Outcomes Group (Figure 1). Of those, 1,624 (97%) had available hemodynamic data. Compared with those in our overall cohort, the 3% with missing hemodynamics were younger (40.4 ± 15.0 vs. 54.5 ± 12.7), were more likely to undergo transplantation for cystic fibrosis (42.6% vs. 12.6%), and had a higher incidence of PGD (31.5% vs. 16.1%). A total of 826 subjects fulfilled criteria for pulmonary hypertension with mPAP ≥25 mm Hg on right heart catheterization and made up the study sample. The degree of PH was less severe in subjects transplanted for cystic fibrosis, COPD, or ILD compared with subjects with sarcoidosis or PAH (Figure 2).
A total of 157 subjects with PH (19.0%) fulfilled criteria for PGD (Table 1). Lower donor BMI and donor smoking exposure were associated with the development of PGD. Recipient BMI, mPAP, and pulmonary vascular resistance were higher in those with PGD. The majority of our cohort underwent bilateral lung transplantation without a significant difference in transplant type based on PGD status (111 [71.2%] with PGD vs. 482 [72.2%] without PGD; P = 0.52). A higher proportion of subjects with PGD received cardiopulmonary bypass (CPB) during the transplant procedure (106 [68.9%] vs. 263 [39.8%]; P < 0.001). The PGD group received a greater volume of packed red blood cell (pRBC) transfusion (P = 0.005).
|PGD (N = 157)||No PGD (N = 669)||P Value|
|Age, yr||34.6 (14.1) (n = 149)||34.9 (13.7) (n = 643)||0.78|
|Sex, female n (%)||60 (38.5) (n = 156)||232 (34.8) (n = 665)||0.39|
|Race, n (%)||n = 155||n = 668||0.73|
|Caucasian||102 (65.8)||426 (64.4)|
|African American||29 (18.7)||142 (21.5)|
|Other||24 (15.5)||94 (14.2)|
|BMI category, kg/m2, n (%)||n = 149||n = 643||0.02|
|<18.5||7 (4.7)||15 (2.3)|
|18.5-24.9||80 (53.7)||284 (44.2)|
|25-29.9||44 (29.5)||213 (33.1)|
|≥30||18 (12.1)||131 (20.4)|
|Mechanism of death, n (%)||n = 156||n = 661||0.74|
|Head trauma||58 (37.2)||243 (36.8)|
|Anoxia||13 (8.3)||75 (11.4)|
|Stroke||62 (39.7)||252 (38.1)|
|Other||23 (14.7)||91 (13.8)|
|Tobacco smoke exposure, n (%)||64 (45.5) (n = 141)||188 (30.4) (n = 618)||0.001|
|Age, yr||52.6 (12.7)||53.5 (12.9)||0.45|
|Sex, female n (%)||74 (47.1)||266 (39.8)||0.09|
|Race, n (%)||n = 668||0.11|
|Caucasian||118 (75.2)||550 (82.3)|
|African American||28 (17.8)||80 (12.0)|
|Other||11 (7.0)||38 (5.7)|
|BMI category, kg/m2, n (%)||n = 156||n = 658||0.18|
|<18.5||9 (5.8)||57 (8.7)|
|18.5-24.9||52 (33.3)||248 (37.7)|
|25-29.9||55 (35.3)||229 (34.8)|
|≥30||40 (25.6)||124 (18.8)|
|Pulmonary diagnosis, n (%)||<0.001|
|COPD||42 (26.8)||266 (39.8)|
|Interstitial lung disease||61 (38.9)||215 (32.1)|
|Cystic fibrosis||9 (5.7)||91 (13.6)|
|PAH||18 (11.5)||36 (5.4)|
|Other†||27 (17.2)||61 (9.1)|
|Creatinine, mg/dl||0.9 [0.7–1.1]||0.9 [0.7–1.0]||0.10|
|Right atrial pressure, mm Hg||14 [10–19]||12 [8–16]||0.002|
|Mean PA pressure, mm Hg||34 [29–43]||30 [27–35]||<0.001|
|PCWP, mm Hg||12 [9–16] (n = 142)||13 [10–16] (n = 597)||0.20|
|Cardiac output, L/min||5.1 [4.2–5.9] (n = 127)||5.2 [4.4–6.1] (n = 548)||0.22|
|Cardiac index, L/min/m2||2.8 [2.4–3.2] (n = 126)||2.6 [2.2–3.1] (n = 540)||0.06|
|PVR, Wood units||3.9 [2.6–6.2] (n = 124)||3.3 [2.5–4.6] (n = 539)||0.005|
|Transplant type, n (%)||0.52|
|Single||39 (25.0)||171 (25.6)|
|Bilateral||111 (71.2)||482 (72.2)|
|Heart/lung||6 (3.9)||15 (2.3)|
|Ischemic time, min||321 [270–388]||320 [261–389]||0.65|
|CPB use, n (%)||106 (68.9)||263 (39.8)||<0.001|
|pRBC volume >1L, n (%)||55 (35.0)||156 (23.3)||0.005|
|Reperfusion FiO2, %||88 [44–98] (n = 92)||50 [25–96] (n = 424)||<0.001|
|Reperfusion FiO2 category, n (%)||n = 92||n = 424||<0.001|
|21–40%||21 (22.8)||183 (43.2)|
|>40%||71 (77.2)||241 (56.8)|
Using multivariable logistic regression in the 826 recipients with PH, donor tobacco smoke exposure, recipient female sex, higher recipient BMI, listing mPAP, right atrial pressure (RAP), and creatinine at time of transplant were associated with PGD (Table 2). Perioperative CPB use, pRBC transfusion volume, and reperfusion FiO2 also increased PGD risk (Table 2). Donor obesity was associated with a lower risk for PGD (OR, 0.52; 95% confidence interval [CI], 0.28–0.95; P = 0.03). Standardized predicted risks of PGD are displayed in Figures 3 and 4. There is a positive correlation between mPAP and the standardized risk for PGD (Figure 3).
|Variable||OR||95% CI||P value|
|Donor tobacco smoke exposure||2.07||1.37–3.13||0.001|
|Donor body mass index, kg/m2|
|Recipient female sex||1.52||1.02–2.28||0.04|
|Recipient body mass index, per 1 kg/m2 increase||1.05||1.01–1.09||0.02|
|Right atrial pressure, per 1 mm Hg increase||1.03||1.01–1.06||0.01|
|Mean PA pressure, per 10 mm Hg increase||1.17||1.00–1.37||0.05|
|Creatinine, per 1 mg/dL increase||1.57||1.21–2.20||0.009|
|Cardiopulmonary bypass use||2.52||1.63–3.89||<0.001|
|Packed red blood cell volume|
|Reperfusion FiO2, per 10 mm Hg increase||1.14||1.05–1.24||0.001|
Because pretransplant diagnosis was collinear with mPAP (r = 0.44; P < 0.001) and CPB (Chi-square P < 0.001), it could not be included in the main model with these other variables. Therefore, we created another model and substituted pretransplant diagnosis for mPAP and CPB. In this model, when compared with COPD, the diagnosis of ILD (OR, 2.03; 95% CI, 1.25–3.40; P = 0.004), PAH (OR, 3.14; 95% CI, 1.53–6.43; P = 0.002), and sarcoidosis, bronchiolitis obliterans syndrome, and other diagnosis (OR, 2.93; 95% CI, 1.57–5.47; P = 0.001) had an increased risk of PGD.
In the conditional logistic regression model conditioned on transplant center, donor tobacco smoke exposure, higher recipient BMI, RAP, and creatinine were significantly associated with PGD. Although the point estimates for female sex, mPAP, pRBC transfusion volume, and reperfusion FiO2 were similar to the explanatory model not conditioned on transplant center (Table E1), the P values for these variables increased in the conditional model grouped by center, likely because of the smaller sample size (one center dropped out of the analysis because there were no PGD cases).
Because the risk factors in the conditional logistic regression model conditioned on transplant center were similar to the overall explanatory model, and because we were interested in a generalizable rather than a localized (transplant-center-specific) model, we elected to randomly generate a derivation and validation cohort for the prognostic model (instead of grouping by center). A predicted probability cut-off of 20% was used, as it optimized the sensitivity and specificity of the model and because it approximated the overall PGD incidence. Selecting a threshold that approximates the incidence of the outcome increases the net benefit of a prognostic model (18). Table E2 shows the donor, recipient, and operative characteristics of the derivation and validation cohorts. The derivation and validation cohorts were similar, other than there being somewhat more obese patients in the validation cohort. In the derivation cohort, donor tobacco smoke exposure, female sex, recipient obesity, and higher mPAP were independently associated with PGD (Table 3). Donor obesity had a borderline protective effect on PGD (OR, 0.42; 95% CI, 0.17–1.03; P = 0.058). In the derivation cohort, the model had a sensitivity to predict PGD of 57.1% (95% CI, 44.1–70.2%), specificity of 74.5% (95% CI, 69.8–79.3%), positive predictive value of 31.9% (95% CI, 23.6–40.3%), and negative predictive value (NPV) of 89.1% (95% CI, 85.3–92.8%) (Table 4). In the validation cohort, the model had a sensitivity of 46.5% (95% CI, 33.8–59.2%), specificity of 69.3% (95% CI, 64.0–74.6%), positive predictive value of 28.2% (95% CI, 19.9–36.5%), and NPV of 83.3% (95% CI, 78.5–88.2%) (Table 4). Receiver operating characteristic curves (ROC) for both the derivation and validation cohorts are displayed in Figure 5. The closed red circle on the validation ROC curve corresponds to the sensitivity and false-positive rate (calculated as one minus the specificity) for the threshold for PGD of 20% that was used to generate the prognostic model.
|Variable||OR||95% CI||P value|
|Donor tobacco smoke exposure||2.19||1.23–3.91||0.008|
|Donor body mass index, kg/m2|
|Recipient female sex||1.92||1.09–3.39||0.02|
|Recipient body mass index, kg/m2|
|Mean PA pressure, per 10 mm Hg increase||1.48||1.19–1.83||<0.001|
|Derivation (N = 413)||Validation (N = 413)|
|Receiver operator characteristic||72.4%||61.3%|
|Sensitivity||57.1% (44.1–70.2%)||46.5% (33.8–59.2%)|
|Specificity||74.5% (69.8–79.3%)||69.3% (64.0–74.6%)|
|Positive predictive value||31.9% (23.6–40.3%)||28.2% (19.9–36.5%)|
|Negative predictive value||89.1% (85.3–92.8%)||83.3% (78.5–88.2%)|
|Prevalence of PGD in the sample||17.4%||20.6%|
In a large, prospective cohort study, we have identified several donor, recipient, and operative characteristics associated with the development of grade 3 PGD in subjects with PH, a population known to be at higher risk of PGD (9, 11). This is one of the first studies, to our knowledge, that specifically evaluated the risk factors for grade 3 PGD in this high-risk population. Although some risk factors, including mPAP, diagnosis of PAH, recipient BMI, donor smoking, pRBC, reperfusion FiO2, and CPB use, were similar to those for PGD in the overall lung transplant population, it is unclear whether the mechanism by which these factors modulate the risk of PGD is similar in those with and without PH. Furthermore, we have detected several risk factors for PGD not previously identified, including RAP and creatinine.
Higher pretransplant RAP and creatinine at the time of transplant were significantly associated with PGD and may reflect decompensated left ventricular (LV) and/or right ventricular (RV) systolic or diastolic function with poor tissue perfusion. The fact that they were not collinear (r = 0.04) and were both significantly associated with PGD in a multivariable model suggests each might reflect a different component of decompensated heart failure, including hypervolemia and impaired cardiac output.
Both transplant diagnosis and mPAP were significantly associated with PGD; it is difficult to conclude whether it is the underlying lung disease diagnosis itself or the level of PA pressure elevation that is causal, as these variables were collinear. Pathologic changes in the pulmonary vasculature vary on the basis of the underlying lung disease, and therefore may explain the variability of PGD based on diagnosis (19–22). For instance, in COPD, vasculature remodelling appears to be mediated by tobacco smoke exposure and hypoxic vasoconstriction and results in the development of neointimal lesions with proliferating smooth muscle-like cells in pulmonary arteries (20–22). These lesions differ from the plexiform lesions in PAH characterized by phenotypically altered and proliferating endothelial-like cells. The mechanism by which ILD leads to PH is unknown and may even vary by ILD type, but may involve oxidant–antioxidant imbalance, the endothelin system, or autoimmunity (19). The effect of these differences in mechanisms and pathologic changes on subsequent PGD after transplant is unknown, but may involve different local effects on pulmonary vasculature compliance and resistance and RV and LV function. Alternatively, mPAP elevation itself, regardless of underlying etiology, may increase the risk for PGD. This is supported by the exposure–response relationship we observed between mPAP and the standardized risk for PGD (Figure 3). Elevated mPAP may increase the risk for PGD through its effect on RV and LV morphology and function. Preserved RV function, as measured by speckle-tracking echocardiography, was associated with the development of PGD, possibly because of increased shear stress on pulmonary vasculature in the transplanted allograft (23). RV pressure and volume overload is associated with LV atrophy and impaired relaxation (24, 25). We have previously shown that impaired LV relaxation is associated with the development of PGD (26). Lastly, although the mechanism by which PH and underlying lung disease increase the risk for PGD may involve local effects on pulmonary arterial compliance and RV and LV morphology, they may also do so by provoking systemic effects including circulation of pro-inflammatory cytokines and activation of innate immunity. Interleukin 6 is associated with mortality in PAH and with PGD (27, 28). Circulating pentraxin 3 is integral in angiogenesis and innate immunity and is elevated in both PAH and PGD (29–31). Further studies are necessary to understand the mechanism or mechanisms by which PH increases the risk for PGD.
Higher recipient BMI was associated with PGD risk similar to previous studies (2, 12). In our study, BMI expressed as a continuous variable was associated with PGD, whereas BMI categorized according to World Health Organization cutpoints was not. This difference might reflect that very high (or low) BMI values are driving the association, and this truncation of the extreme measures dampens the association. Alternatively, BMI categories according to World Health Organization cutpoints for BMI categories may not accurately reflect obesity as it relates to PGD risk. We have previously shown that BMI ≥30 kg/m2 was a poor predictor of total body fat-defined obesity (32). Alternate approaches to modeling BMI, especially with large numbers of patients across the range of BMI values, might lead to improved understanding of the relationship between recipient BMI and PGD. In addition, alternative measures of adiposity, including computed tomography scan, dual-energy X-ray absorptiometry, and biomarkers, may improve PGD risk assessment (12, 32). Although recipient obesity was associated with an increased risk for PGD in our study, donor obesity was associated with a decreased risk for PGD. The association of donor obesity and recipient PGD was not modified by recipient BMI (test for interaction, P = 0.46). Previous studies have documented a lower risk for other forms of acute lung injury, including acute respiratory distress syndrome, among obese patients (33). One theory suggests obesity induces low-grade inflammation that triggers anti-inflammatory, antioxidant, and other mechanisms that protect the lung against subsequent insults (33–35). Donor obesity might activate anti-inflammatory mechanisms that blunt the subsequent injuries induced by ischemia-reperfusion injury at the time of transplant. Further studies are necessary to better understand this relationship.
Donor tobacco smoke exposure was associated with a twofold increase in the odds of developing PGD. Previous studies have shown that donor smoking also increases the risk for PGD and mortality in the overall transplant population (2, 6). The exact mechanism by which tobacco exposure increases the risk for PGD in the overall population remains unknown; however, it may include increased oxidative stress and epithelial injury (36, 37). Both PH and donor smoking exposure have been associated with increased lipid peroxidation products, suggesting the potential for overlapping mechanisms involving oxidative stress (37, 38). Despite the link between donor tobacco exposure and increased PGD and mortality in the overall lung transplant population, the increased risk associated with accepting a lung from such donor is less than the risk of death while remaining on the transplant list, given the limited donor pool (39). It is unknown whether the same conclusion is true for those with PH.
Several operative variables were associated with PGD among those with PH, including CPB use, larger pRBC transfusion volume, and higher reperfusion FiO2. Although these risk factors overlap with those in the overall transplant population (2, 6), these risk factors may warrant distinct consideration in those with PH. Larger pRBC transfusion volumes may be especially problematic in those with RV or LV dysfunction resulting from PH. Higher FiO2 requirements among those with PH undergoing sequential bilateral lung transplantation may confound reperfusion FiO2. Hyperoxia may also increase LV filling pressures (40) and exacerbate LV diastolic dysfunction among those with PH. Future studies evaluating these perioperative risk factors in PH may translate into changes in perioperative management.
The ability to detect those at high risk for PGD would better inform patients of their posttransplant mortality, identify subjects for inclusion into research studies, and assist in allocating already scarce organs. Important considerations for developing a prognostic model have been recently discussed (41). Assessment of our prognostic model’s performance (Table 4) in both the derivation and validation cohorts yielded moderate sensitivity and specificity with high NPV. Although the NPV of our model is high, the prevalence of PGD at most centers is only 20–30%, which may limit the clinical utility of the model. However, it may help identify a cohort with lower PGD risk for future clinical trial design. The relatively low positive predictive value of our model highlights the difficulty of identifying high-risk recipients and suggests clinical risk factors alone are inadequate to identify high-risk subjects with PH. Further evaluation is warranted to determine whether inclusion of biomarkers into the model would improve clinical utility.
Our study has several limitations. Although we have previously published data on risk factors for PGD using the Lung Transplant Outcomes Group database (2), this study focuses on risk factors in patients with PH at especially high risk for PGD. Given the observational nature of our study, we were unable to fully exclude other disease processes with similar radiographic appearances as PGD including diffuse pneumonia or significant pulmonary contusion. However, this definition has been used in other large center trials and has previously demonstrated good construct validity (1, 2, 4, 6). Despite extensive and standardized variable collection, unmeasured confounding is possible. Specifically, detailed information regarding perioperative management, including fluid management, use of pulmonary vasodilators, and emergent versus planned CPB use, were unavailable and represent potential confounders. Preoperative echocardiographic data were not available, but should be incorporated into future trials evaluating the mechanistic link among PH, RV and LV dysfunction, and PGD. There remains a potential for selection bias, as a small number of subjects (54 of 1,678) were excluded because of missing hemodynamics, although this only represents 3% of the overall cohort.
In conclusion, we identified several risk factors associated with the development of grade 3 PGD among those with PH that should be the focus of future mechanistic studies. We demonstrated that a prognostic model for grade 3 PGD in those with PH based on clinical risk factors alone had high NPV, which identifies low-risk transplant recipients.
|1 .||Christie JD, Carby M, Bag R, Corris P, Hertz M, Weill D. Report of the ishlt working group on primary lung graft dysfunction part ii: Definition. A consensus statement of the international society for heart and lung transplantation. J Heart Lung Transplant 2005;24:1454–1459.|
|2 .||Diamond JM, Lee JC, Kawut SM, Shah RJ, Localio AR, Bellamy SL, Lederer DJ, Cantu E, Kohl BA, Lama VN, et al. Clinical risk factors for primary graft dysfunction after lung transplantation. Am J Respir Crit Care Med 2013;187:527–534.|
|3 .||Kuntz CL, Hadjiliadis D, Ahya VN, Kotloff RM, Pochettino A, Lewis J, Christie JD. Risk factors for early primary graft dysfunction after lung transplantation: A registry study. Clin Transplant 2009;23:819–830.|
|4 .||Christie JD, Bellamy S, Ware LB, Lederer D, Hadjiliadis D, Lee J, Robinson N, Localio AR, Wille K, Lama V, et al. Construct validity of the definition of primary graft dysfunction after lung transplantation. J Heart Lung Transplant 2010;29:1231–1239.|
|5 .||Daud SA, Yusen RD, Meyers BF, Chakinala MM, Walter MJ, Aloush AA, Patterson GA, Trulock EP, Hachem RR. Impact of immediate primary lung allograft dysfunction on bronchiolitis obliterans syndrome. Am J Respir Crit Care Med 2007;175:507–513.|
|6 .||Whitson BA, Nath DS, Johnson AC, Walker AR, Prekker ME, Radosevich DM, Herrington CS, Dahlberg PS. Risk factors for primary graft dysfunction after lung transplantation. J Thorac Cardiovasc Surg 2006;131:73–80.|
|7 .||Porteous M, Lederer DJ, Palmer SM, Cantu E, Shah RJ, Bellamny S, Lama VN, Bhorade SM, Crespo MM, Wille KM, et al. Clinical prediction model for pgd among patients with pulmonary hypertension. J Heart and Lung Transplant 2015;34:S254–S255.|
|8 .||Porteous M, Lederer DJ, Palmer SM, Cantu E, Shah RJ, Bellamny S, Lama VN, Bhorade SM, Crespo MM, McDyer JF, et al. Explanatory model for primary graft dysfunction among patients with pulmonary hypertension. Am J Respir Crit Care Med 2015;191:A1428.|
|9 .||Shah RJ, Diamond JM, Cantu E, Flesch J, Lee JC, Lederer DJ, Lama VN, Orens J, Weinacker A, Wilkes DS, et al. Objective estimates improve risk stratification for primary graft dysfunction after lung transplantation. Am J Transplant 2015;15:2188–2196.|
|10 .||Barr ML, Kawut SM, Whelan TP, Girgis R, Bottcher H, Sonett J, Vigneswaran W, Follette DM, Corris PA. Report of the ishlt working group on primary lung graft dysfunction part iv: Recipient-related risk factors and markers. J Heart Lung Transplant 2005;24:1468–1482.|
|11 .||Fang A, Studer S, Kawut SM, Ahya VN, Lee J, Wille K, Lama V, Ware L, Orens J, Weinacker A, et al. Elevated pulmonary artery pressure is a risk factor for primary graft dysfunction following lung transplantation for idiopathic pulmonary fibrosis. Chest 2011;139:782–787.|
|12 .||Lederer DJ, Kawut SM, Wickersham N, Winterbottom C, Bhorade S, Palmer SM, Lee J, Diamond JM, Wille KM, Weinacker A, et al. Obesity and primary graft dysfunction after lung transplantation: The lung transplant outcomes group obesity study. Am J Respir Crit Care Med 2011;184:1055–1061.|
|13 .||Azur MJ, Stuart EA, Frangakis C, Leaf PJ. Multiple imputation by chained equations: What is it and how does it work? Int J Methods Psychiatr Res 2011;20:40–49.|
|14 .||van Buuren S. Multiple imputation of discrete and continuous data by fully conditional specification. Stat Methods Med Res 2007;16:219–242.|
|15 .||White IR, Royston P, Wood AM. Multiple imputation using chained equations: Issues and guidance for practice. Stat Med 2011;30:377–399.|
|16 .||Schafer JL. Multiple imputation: A primer. Stat Methods Med Res 1999;8:3–15.|
|17 .||Moons KG, Altman DG, Reitsma JB, Ioannidis JP, Macaskill P, Steyerberg EW, Vickers AJ, Ransohoff DF, Collins GS. Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (tripod): Explanation and elaboration. Ann Intern Med 2015;162:W1–W73.|
|18 .||Steyerberg E. Clinical prediction models: A practical approach to development, validation, and updating. New York: Springer; 2009.|
|19 .||Behr J, Ryu JH. Pulmonary hypertension in interstitial lung disease. Eur Respir J 2008;31:1357–1367.|
|20 .||Chaouat A, Naeije R, Weitzenblum E. Pulmonary hypertension in copd. Eur Respir J 2008;32:1371–1385.|
|21 .||Christensen CC, Ryg MS, Edvardsen A, Skjonsberg OH. Relationship between exercise desaturation and pulmonary haemodynamics in copd patients. Eur Respir J 2004;24:580–586.|
|22 .||Sakao S, Voelkel NF, Tatsumi K. The vascular bed in copd: Pulmonary hypertension and pulmonary vascular alterations. Eur Respir Rev 2014;23:350–355.|
|23 .||Tudorache I, Sommer W, Kuhn C, Wiesner O, Hadem J, Fuhner T, Ius F, Avsar M, Schwerk N, Bothig D, et al. Lung transplantation for severe pulmonary hypertension–awake extracorporeal membrane oxygenation for postoperative left ventricular remodelling. Transplantation 2015;99:451–458.|
|24 .||Hardziyenka M, Campian ME, Reesink HJ, Surie S, Bouma BJ, Groenink M, Klemens CA, Beekman L, Remme CA, Bresser P, et al. Right ventricular failure following chronic pressure overload is associated with reduction in left ventricular mass: Evidence for atrophic remodeling. J Am Coll Cardiol 2011;57:921–928.|
|25 .||Hardziyenka M, Campian ME, Verkerk AO, Surie S, van Ginneken AC, Hakim S, Linnenbank AC, de Bruin-Bon HA, Beekman L, van der Plas MN, et al. Electrophysiologic remodeling of the left ventricle in pressure overload-induced right ventricular failure. J Am Coll Cardiol 2012;59:2193–2202.|
|26 .||Porteous MK, Ky B, Kirkpatrick JN, Shinohara R, Diamond JM, Shah RJ, Lee JC, Christie JD, Kawut SM. Diastolic dysfunction increases the risk of primary graft dysfunction after lung transplant. Am J Respir Crit Care Med 2016;193:1392–1400.|
|27 .||Heresi GA, Aytekin M, Hammel JP, Wang S, Chatterjee S, Dweik RA. Plasma interleukin-6 adds prognostic information in pulmonary arterial hypertension. Eur Respir J 2014;43:912–914.|
|28 .||Krishnadasan B, Naidu BV, Byrne K, Fraga C, Verrier ED, Mulligan MS. The role of proinflammatory cytokines in lung ischemia-reperfusion injury. J Thorac Cardiovasc Surg 2003;125:261–272.|
|29 .||Tamura Y, Ono T, Kuwana M, Inoue K, Takei M, Yamamoto T, Kawakami T, Fujita J, Kataoka M, Kimura K, et al. Human pentraxin 3 (ptx3) as a novel biomarker for the diagnosis of pulmonary arterial hypertension. PLoS One 2012;7:e45834.|
|30 .||Diamond JM, Lederer DJ, Kawut SM, Lee J, Ahya VN, Bellamy S, Palmer SM, Lama VN, Bhorade S, Crespo M, et al. Elevated plasma long pentraxin-3 levels and primary graft dysfunction after lung transplantation for idiopathic pulmonary fibrosis. Am J Transplant 2011;11:2517–2522.|
|31 .||Diamond JM, Meyer NJ, Feng R, Rushefski M, Lederer DJ, Kawut SM, Lee JC, Cantu E, Shah RJ, Lama VN, et al. Variation in ptx3 is associated with primary graft dysfunction after lung transplantation. Am J Respir Crit Care Med 2012;186:546–552.|
|32 .||Singer JP, Peterson ER, Snyder ME, Katz PP, Golden JA, D’Ovidio F, Bacchetta M, Sonett JR, Kukreja J, Shah L, et al. Body composition and mortality after adult lung transplantation in the united states. Am J Respir Crit Care Med 2014;190:1012–1021.|
|33 .||Memtsoudis SG, Bombardieri AM, Ma Y, Walz JM, Chiu YL, Mazumdar M. Mortality of patients with respiratory insufficiency and adult respiratory distress syndrome after surgery: The obesity paradox. J Intensive Care Med 2012;27:306–311.|
|34 .||Morris DL, Singer K, Lumeng CN. Adipose tissue macrophages: Phenotypic plasticity and diversity in lean and obese states. Curr Opin Clin Nutr Metab Care 2011;14:341–346.|
|35 .||Bustamante AFRJ. The obesity ards paradox: “A pre-conditioning cloud. J Pulm Respir Med 2012;2:e122.|
|36 .||Ware LB, Lee JW, Wickersham N, Nguyen J, Matthay MA, Calfee CS, California Transplant Donor N. Donor smoking is associated with pulmonary edema, inflammation and epithelial dysfunction in ex vivo human donor lungs. Am J Transplant 2014;14:2295–2302.|
|37 .||Diamond JM, Porteous MK, Jackson Roberts L II, Wickersham N, Rushefski M, Kawut SM, Shah RJ, Cantu E III, Lederer DJ, Chatterjee S, et al. The relationship between plasma lipid peroxidation products and primary graft dysfunction after lung transplantation is modified by donor smoking and reperfusion hyperoxia. J Heart Lung Transplant 2016;35:500–507.|
|38 .||Cracowski JL, Cracowski C, Bessard G, Pepin JL, Bessard J, Schwebel C, Stanke-Labesque F, Pison C. Increased lipid peroxidation in patients with pulmonary hypertension. Am J Respir Crit Care Med 2001;164:1038–1042.|
|39 .||Bonser RS, Taylor R, Collett D, Thomas HL, Dark JH, Neuberger J. Cardiothoracic Advisory Group to NHSB, Transplant, the Association of Lung Transplant P. Effect of donor smoking on survival after lung transplantation: A cohort study of a prospective registry. Lancet 2012;380:747–755.|
|40 .||Mak S, Azevedo ER, Liu PP, Newton GE. Effect of hyperoxia on left ventricular function and filling pressures in patients with and without congestive heart failure. Chest 2001;120:467–473.|
|41 .||Labarere J, Renaud B, Fine MJ. How to derive and validate clinical prediction models for use in intensive care medicine. Intensive Care Med 2014;40:513–527.|
This study was supported by the National Institutes of Health (Grants T32 HL007891, K24 HL103844, R01 HL087115, R01 HL081619, R01 HL096845, K23: K23 HL121406) and the Actelion Entelligence Grant.
Author Contributions: M.K.P. designed the study, performed data collection and data analysis and interpretation, wrote the first draft, and approved the final draft; S.L.B. and A.R.L. assisted with data analysis and data interpretation, reviewed and revised the manuscript, and approved the final draft; J.C.L., D.J.L., S.M.P., E.C., R.J.S., V.N.L., S.M.B., M.M.C., J.F.M., K.M.W., J.B.O., P.D.S., A.B.W., S.A., D.S.W., and L.B.W. assisted with data interpretation, reviewed and revised the manuscript, and approved the final draft; and J.M.D., S.M.K., and J.D.C. assisted with study design, contributed to data collection and interpretation, reviewed and revised the manuscript, and approved the final draft.
This article has a data supplement, which is accessible from this issue's table of contents online at www.atsjournals.org.