Factors predicting outcome after pulmonary endarterectomy

Objective Few studies have reported predictive factors of outcome after pulmonary endarterectomy (PEA) in chronic thromboembolic pulmonary hypertension. The purpose of this study was to determine factors influencing mortality and predictors of hemodynamic improvement after PEA. Methods A total of 383 consecutive patients who underwent PEA between January 2005 and December 2009 were retrospectively reviewed. Among them, 150 were fully reevaluated 7.5±1 months after PEA by NYHA class, 6–minute walk distance (6MWD), percentage of predicted carbon monoxide transfer factor (TLCO) and right heart catheterisation. Results Mortality rates at 1 month, 1 year and 3 years were 2.8%, 6.9% and 7.5%, respectively. Preoperative pulmonary vascular resistance (PVR) independently predicted 1-month, 1- and 3-year mortality and age predicted mortality at 1 year and 3 years. Significant improvement in NYHA class and 6MWD were observed and PVR decreased from 773±353 to 307±221 dyn.sec.cm-5 (p<0.001). In 96 patients (64%), PVR decreased by at least 50% and/or was reduced to lower than 250 dyn.sec.cm-5. Preoperative cardiac output (CO) and TLCO predicted hemodynamic improvement. Conclusion PEA is associated with an excellent long-term survival and a marked improvement in clinical status and hemodynamics. Some preoperative factors including PVR, CO and TLCO can predict postoperative outcomes.


Introduction
Chronic thromboembolic pulmonary hypertension (CTEPH) is a rare complication of venous thromboembolism and is associated with an important morbidity and mortality. CTEPH results from obstruction of the pulmonary arterial bed by organized thrombus after acute or recurrent pulmonary emboli. Despite the advent of medical therapies [1] and the emergence of pulmonary angioplasty for CTEPH [2], the best-established treatment remains pulmonary endarterectomy (PEA), which is usually performed in expert surgical centers [3]. Eligibility criteria for surgery are determined by a multidisciplinary panel of physicians, surgeons and radiologists and are based on the amount of surgically accessible lesions assessed by imaging, the presence of comorbidities and the degree of hemodynamic impairment in symptomatic patients [3]. When successful, PEA markedly improves pulmonary hemodynamics, symptoms and functional status. However, several studies have demonstrated that some patients had persistent pulmonary hypertension (PH) after the procedure [4][5][6][7] Persistent PH after surgery represents the most important cause of postoperative morbidity and mortality but there is no consensus on its definition. Some authors used mean pulmonary artery pressure (mPAP) thresholds of 25-30 mmHg [5,7] whereas others used pulmonary vascular resistance (PVR) thresholds of 500-550 dyn.sec.cm -5 [4,6,8], although mPAP!25mmHg is the accepted definition of PH (3).
In a cohort of consecutive CTEPH patients who underwent PEA, we aimed to identify: (1) factors affecting short and long-term survival; (2) predictive factors of hemodynamic and/or functional status improvement.

Patients and methods
This retrospective study complied with the Declaration of Helsinki. Although French law does not require ethics committee approval or informed consent for retrospective data collection, the data was anonymised and complied with the requirements of the Commission Nationale Informatique et Liberté, the organisation dedicated to privacy, information technology, and civil rights in France. The committee approved the methods used to collect and analyse data on May 24, 2003 (approval number 842063).
wedge pressure (PAWP) 15mmHg measured by right heart catheterization) [3]. CTEPH was confirmed as the cause of pulmonary hypertension in the presence of mismatched perfusion defects on radionuclide ventilation/perfusion lung scan and typical lesions of CTEPH on computed tomography pulmonary angiogram and/or conventional pulmonary angiography. Patients were required to have received at least 3 months of adequate anticoagulation therapy before PEA.
Patients were not included in the analysis if the preoperative PVR was below 250 dyn.sec. cm -5 , or in the absence of either a preoperative hemodynamic assessment or postoperative follow-up.

Data collection and classification
Data were collected from the French PH registry or from routinely performed assessments in clinical practice. Predefined clinical, functional, hemodynamic and therapeutic data were collected using a standardized case report form. The period of observation was from the date of PEA until time of death or date of last known contact.

Outcomes evaluation
The primary outcome was short-term (within 1 month after surgery, i.e.; the early post-operative period) and long-term mortality (at 1 year and 3 years) after PEA.
Secondary outcomes were: 1. normalization of hemodynamics, defined by postoperative PVR <250 dyn.sec.cm -5 at first assessment after PEA; 2. clinically meaningful improvement in hemodynamics defined, as it has been suggested in previous publications [9,18], by a postoperative decrease in PVR by at least 50% and to an absolute value of less than 500 dyn.sec.cm -5 at first assessment after PEA. The threshold of 500 dyn.sec.cm -5 was chosen because it is strongly correlated with mortality in several studies [4,6,8]. Patients with preoperative PVR below 500dyn.sec.cm -5 were not included in this analysis; 3. improvement in NYHA by at least one functional class at first evaluation after PEA.

Statistical analysis
Continuous variables were expressed as mean (± standard deviation [17]) or median depending on normality of the distribution; 95% confidence intervals (95% CI) were computed for proportions according to the binomial distribution. For the comparison of pre-and postoperative data samples a paired Student's t-test was used for continuous variables and a Wilcoxon test for the improvement of NYHA classification. Univariate analysis was conducted in four steps: (1) an association between each predefined variable and endpoints was tested using univariable logistical regression and only variables with a p-value <0.2 were considered in further analysis; (2) correlations between variables were systematically searched within each and between sets of variables (in case of significant correlation, the most statistically significant variable was considered); (3) for continuous variables that were statistically significant, the discriminant power was determined by calculating the area under the curve (AUC) on receiver operating characteristic (ROC) curve analysis; and (4) interactions between variables (within and between sets of variables) were also estimated and considered for a p-values <0.1. Multivariate analysis was performed using a multivariable logistic regression. In the model, only variables with a p-value <0.2, which were not correlated and had no significant interaction (p-value >0.1), were included. Age and FEV1/FVC were also systematically included in the model. In this multivariate model, predictors were considered as statistically significant for pvalue of less than 0.05. Survival curves were derived by Kaplan-Meier method. Statistical analyses were performed using SPSS software (version 19.0; SPSS, Inc.; Chicago, Illinois).

Study population
Between January 2005 and December 2009, 383 patients underwent PEA. Among them, 193 were followed up after surgery at the French reference for PH. Twenty-one patients were excluded from the study for the following reasons: loss of follow-up (n = 6), final diagnosis of sarcoma (n = 2), final diagnosis of fibrosing mediastinitis (n = 1), no preoperative hemodynamic assessment (n = 2) and PEA performed in patients with PVR <250 dyn.sec.cm -5 (n = 10). Therefore, the primary objective (mortality) was analysed in 172 patients. As full postoperative hemodynamics were not assessed in 22 of these patients, the secondary endpoints were evaluated in 150 patients (Fig 1). The mean age of the patients was 60±14 years and 51% were female. A history of acute venous thromboembolism was reported in 114 patients (66%). At the time of surgery, 26 patients (15%) had medical conditions increasing the risk of CTEPH and including antiphospholipid syndrome (n = 6), splenectomy (n = 6), myeloproliferative disorder (n = 12), ventriculoatrial shunt (n = 4), implanted central venous access port (n = 3) and cardiac pacemaker (n = 4). Pre-PEA, 19 patients (11%) were receiving at least one PAH targeted treatment. The time between diagnosis of CTEPH and PEA was significantly longer in these patients than in patients who did not receive PAH targeted therapies before surgery (8.1 ±5 versus 1.2±1.9 months; p<0.001). At the time of pre-surgery evaluation, patients receiving bridging therapy had higher mPAP (55±12 versus 46±10 mmHg; p<0.001) but the level of PVR did not differ between the two groups (858±338 versus 754±341 dyn.sec.cm -5 ; p = 0.14). The preoperative patient characteristics are summarized in Table 1.

Primary outcome
Postoperative mortality rates at 1 month, 1 year and 3 years were 2.9%, 6.9% and 7.5%, respectively. Early mortality was mainly related to right ventricular failure. The causes of death are detailed in S1 Appendix. In multivariate analysis, preoperative PVR was a predictive factor of mortality at 1 month (odd ratio (OR, per 100dyn.sec.cm -5 ) 1.329, 95% confidence interval (CI), 1.035 to 1.706, p = 0.026), 1 year (OR, per 100dyn.sec.cm -5 1.252, 95% CI, 1.024 to 1.529, p = 0.028) and 3 years (OR, per 100dyn.sec.cm -5 1.245, 95% CI, 1.013 to 1.529, p = 0.037). Age also predicted 1-and 3-year mortality. There was a trend towards a deleterious effect of the use of PAH targeted therapies before PEA (duration of use before surgery, per month) on 1-month mortality (p = 0.053). The results of the univariate and multivariate analyses are shown in Table 2. No other baseline factors were predictors of mortality in univariate and multivariate analysis. Survival according to the optimal threshold of PVR determined by ROC analysis is represented in Fig 2 (AUC = 0.72, standard error 0.06, p = 0.008, sensitivity 77%, specificity 60%).

Normalization of pulmonary hemodynamics
Normalization of hemodynamics, defined by postoperative PVR <250 dyn.sec.cm -5 , was observed in 71 out of the 150 patients (47%). Multivariate analysis revealed that preoperative CO was the main predictive factor of normalization of hemodynamics ( Table 4).

Improvement of pulmonary hemodynamics
One hundred and nine patients (73%) had preoperative PVR >500 dyn.sec.cm -5 . Seventy of these 109 patients (64%) improved their PVR by at least 50% to reach a value of less than 500 dyn.sec.cm -5 . When mPAP was included in the multivariate model, TL CO (adjusted for FEV1/ FVC ratio) was a significant predictor of hemodynamic improvement (OR, 1.06; 95% CI, 1.01 to 1.10; p = 0.009). When PVR was included in the multivariate model, TL CO (adjusted for

Improvement of NYHA functional class
Postoperatively, 73% of patients (110/150 patients) improved their NYHA functional class by at least one class. In multivariate analysis, baseline TL CO , functional class and FEV1/FVC predicted improvement of NYHA functional class ( Table 6). A worse NYHA functional class at baseline predicted a greater improvement in NYHA functional class after surgery.

Discussion
The present study assessed the factors influencing short-and long-term mortality, the predictors of functional improvement, and the factors predicting improvement or normalization of pulmonary hemodynamics in patients who underwent PEA for CTEPH.

Short-and long-term mortality
In our cohort, the mortality rates at 1 month, 1 year and 3 years were 2.9%, 6.9% and 7.5%, respectively, which are similar to other CTEPH centres around the world [4,8,[12][13][14][15]. Our study supports the importance of preoperative PVR in predicting early post-operative, 1-and 3-year mortality and of age in predicting 1-and 3-year mortality. The main causes of deaths observed during the early postoperative period were related to right ventricular failure. In contrast, long-term mortality was associated with other causes than CTEPH. Several studies have shown that high preoperative PVR was associated with an increase in early postoperative mortality [6,9,11,15]. Tscholl et al. were first to demonstrate that patients with a high preoperative PVR (greater than 1136 dyn.sec.cm -5 ) had a markedly higher mortality risk [11]. Dartevelle et al. also reported a prognostic influence of PVR, with in-hospital mortality increasing when preoperative PVR was higher than 900 dyn.sec.cm -5 [9]. Lastly, in a large series of CTEPH patients treated by PEA at the University of California-San Diego, Madani et al. showed that preoperative PVR levels correlated with perioperative mortality, with patients experiencing a higher mortality when PVR exceeded 1000 dyn.sec.cm -5 [6]. Interestingly, in the present study, preoperative PVR was also identified as an independent risk factor of late death (at 1 and 3 years) in multivariate analysis, while in other studies, PVR measured immediately after PEA [6,7,10] or at 3-to 12-month post-PEA was found to be a predictor of long-term mortality [8,12]. Our study confirms that the most important risk factor for mortality remains the preoperative hemodynamic impairment assessed by the level of PVR. We found a threshold of 800 dyn.sec.cm -5 for identification of patients at high or low risk of death with sensitivity and specificity of 77% and 60%, respectively. In addition, patient selection for surgery is a complex process that is based on the concordance between the degree of obstruction and the level of PVR. Therefore, our cut-off value of PVR must be interpreted cautiously and should not be considered as the only criteria for surgical treatment decision. Lastly, age was also a predictor of 1 and 3-year mortality but did not predict in-hospital mortality. These results confirm that an advanced age is not systematically a criterion for inoperability [19,20].
We observed a trend towards a deleterious effect of the use of PAH targeted therapy before PEA with mortality increasing by 50% per month of use. In our study, PAH-targeted medical Their analysis revealed that the use of PAH targeted treatment before surgery resulted in delayed referral for PEA, was not associated with improvement in postoperative pulmonary hemodynamics and had no impact on postoperative outcome [23]. More recently, an international prospective registry including 27 centres in Europe and Canada assessed the long-term outcomes of a cohort of operated and not-operated patients. The authors found that 29% of operated patients had received bridging therapy with PAH-targeted drugs before PEA. These patients had higher preoperative PVR in comparison with those who did not receive bridging therapy but postoperative PVR and the rate of PEA related complications did not differ between the 2 groups. In addition, one of the independent predictors of mortality, discovered in the multivariate analysis, was bridging therapy with PH drugs [14]. Thus, these observations reinforce the idea that PAH targeted therapy should not delay a patient's referral to expert centres for PEA and if indicated, should be started after a multidisciplinary decision.

TL CO as a predictive factor of postoperative improvement
In the present study, we demonstrated for the first time that preoperative TL CO predicted improvements in hemodynamics and NYHA functional class after PEA. Despite these results, we did not find that TL CO was associated with postoperative mortality. Previous studies have shown that TL CO was a predictive factor of short-and long-term mortality. Condliffe et al. assessed factors influencing mortality in a cohort of 236 patients who underwent PEA and reported that preoperative TL CO was an independent predictor of perioperative mortality [24]. More recently, Suda et al. investigated the relationship between TL CO /alveolar ventilation (TL CO /V A ) and long-term outcomes after PEA and found that decreased TL CO /V A was associated with poor long-term outcomes. In addition, the authors showed the absence of correlation between TL CO /V A and proximal obstruction suggesting that TL CO /V A might be related to distal or microvascular disease [25]. In another study, Zoia et al. have reported that patients who underwent successful PEA improved gradually their TL CO after surgery due to the reversal of microvascular disease [26]. Thus, TL CO might reflect the degree/severity of small vessel vasculopathy and could be a simple tool to predict postoperative improvement.
The relationship between the decrease in TL CO and microvascular disease remains poorly understood [27]. Bernstein et al. showed that TL CO was slightly reduced in CTEPH patients. They demonstrated that this decrease in TL CO was predominantly caused by a reduction in pulmonary membrane diffusion capacity (Dm) and to a lesser extent by a low pulmonary capillary blood volume (Vc) [28]. Steenhuis et al. noted that Dm was also strongly associated with PVR. They hypothesized that the reduction in Dm was due to microvascular disease and, more specifically, to an alveolocapillary membrane thickening [29]. Interestingly, Dorfmüller et al. described histological findings from 17 CTEPH patients who were transplanted either after failure of PEA or because of inoperable disease. They reported an involvement of small pulmonary arteries but also venous remodelling and focal capillary haemangiomatosis [30].
Whether TL CO may be a predictor of mortality and a predictive factor of improvement of hemodynamics and NYHA functional class after PEA requires further confirmation in larger prospective trials.

Limitations
We acknowledge some limitations such as the retrospective and monocentric nature of the present study. In addition, some patients were not evaluated for secondary endpoints because of lack of hemodynamic data. The low mortality rate after PEA limited the power of the multivariate analysis. Lastly, we did not analyse the impact of the degree and location of pulmonary vascular obstruction determined by CT scan and pulmonary angiography on the post-PEA outcome because no objective radiological score has been yet established in CTEPH. Moreover, the surgical classification schema to separate patients between proximal (main and lobar) vs distal (segmental and subsegmental) chronic thromboembolic disease resection was not available for some patients. It is possible that this distinction is an important factor to predict longterm outcome after surgery. Finally, we identified different factors predicting mortality, improvement in hemodynamics, normalization of hemodynamics and improvement in functional status. These findings may be due to a lack of statistical power.
In conclusion, PEA is associated with an excellent long-term survival and a marked improvement in clinical status and hemodynamics. Preoperative PVR is a major determinant of short-and long-term mortality whereas TL CO may predict postoperative improvement of hemodynamics and NYHA functional class. These findings need to be confirmed by prospective studies including large cohorts of CTEPH patients.