Predicting Survival in Pulmonary Arterial Hypertension
Insights From the Registry to Evaluate Early and Long-Term Pulmonary Arterial Hypertension Disease Management (REVEAL)
Background— Factors that determine survival in pulmonary arterial hypertension (PAH) drive clinical management. A quantitative survival prediction tool has not been established for research or clinical use.
Methods and Results— Data from 2716 patients with PAH enrolled consecutively in the US Registry to Evaluate Early and Long-Term PAH Disease Management (REVEAL) were analyzed to assess predictors of 1-year survival. We identified independent prognosticators of survival and derived a multivariable, weighted risk formula for clinical use. One-year survival from the date of enrollment was 91.0% (95% confidence interval [CI], 89.9 to 92.1). In a multivariable analysis with Cox proportional hazards, variables independently associated with increased mortality included pulmonary vascular resistance >32 Wood units (hazard ratio [HR], 4.1; 95% CI, 2.0 to 8.3), PAH associated with portal hypertension (HR, 3.6; 95% CI, 2.4 to 5.4), modified New York Heart Association/World Health Organization functional class IV (HR, 3.1; 95% CI, 2.2 to 4.4), men >60 years of age (HR, 2.2; 95% CI, 1.6 to 3.0), and family history of PAH (HR, 2.2; 95% CI, 1.2 to 4.0). Renal insufficiency, PAH associated with connective tissue disease, functional class III, mean right atrial pressure, resting systolic blood pressure and heart rate, 6-minute walk distance, brain natriuretic peptide, percent predicted carbon monoxide diffusing capacity, and pericardial effusion on echocardiogram all predicted mortality. Based on these multivariable analyses, a prognostic equation was derived and validated by bootstrapping technique.
Conclusions— We identified key predictors of survival based on the patient’s most recent evaluation and formulated a contemporary prognostic equation. Use of this tool may allow the individualization and optimization of therapeutic strategies. Serial follow-up and reassessment are warranted.
Clinical Trial Registration— URL: http://www.clinicaltrials.gov. Unique identifier: NCT00370214.
Received July 30, 2009; accepted May 3, 2010.
Pulmonary arterial hypertension (PAH) is a fatal disease that has no satisfactory predictive model of survival. The only existing predictive equation, derived from the National Institutes of Health (NIH) Registry of Primary Pulmonary Hypertension (1983 to 1987),1 may not be applicable to the broader World Health Organization (WHO) group I PAH population or accurately reflect survival in the current treatment era. Substantial advances, including safe and effective therapies2,3 and a revised classification system,4 necessitate a new prognostic equation.
Editorial see p 106
Clinical Perspective on p 172
Although 6-minute walk distance (6MWD) and other end points are considered potential surrogates for survival of patients with PAH, they have never been thoroughly tested for their predictive abilities. However, these factors are often used to make critical decisions about the utility and efficacy of present-day therapeutics.5 The Registry to Evaluate Early and Long-Term PAH Disease Management (REVEAL) is a multicenter, observational, US-based registry initiated in 2006 and designed to study longitudinal clinical course and disease management in ≈3000 patients with WHO group I PAH.6 A prespecified objective of the REVEAL Registry was to identify predictors of short- and long-term survival reflecting current treatment and clinical variables. Using these results, we assessed the prognostic value of multiple factors, enabling more accurate risk stratification, and we developed an algorithm for predicting survival in patients with PAH.
REVEAL Study Design
REVEAL is an observational prospective registry study that consecutively enrolled patients with a diagnosis of WHO group I PAH meeting prespecified hemodynamic criteria at 54 geographically diverse community and university PAH specialty care facilities in the United States.6 Both newly and previously diagnosed patients have been enrolled and will be followed up for at least 5 years unless discontinued from the study because of withdrawal of consent, death, or loss to follow-up. There are no protocol-mandated tests, treatments, or visit schedules. Study objectives and methods were prespecified in an Institutional Review Board–approved protocol, and all participants or their legal guardians gave written informed consent.
Data from right heart catheterization were categorized as meeting traditional or expanded hemodynamic criteria for WHO group I PAH (ie, pulmonary capillary wedge pressure ≤15 versus 16 to 18 mm Hg); only patients meeting traditional hemodynamic criteria are included in the analyses to develop a prognostic equation for survival in PAH. All WHO group I subgroups were analyzed except for pulmonary hypertension of the newborn.
Patient data are collected electronically at the time of enrollment and updated quarterly as available. Demographic data include age, sex, race, and ZIP code; median income in the patient’s ZIP code is used as a proxy for socioeconomic status.7 The most recent data collected at the time of enrollment include WHO group I PAH subgroup classification, modified New York Heart Association (NYHA)/WHO functional class, 6MWD with concurrent Borg dyspnea scale, pulmonary function testing, hemodynamic measurements, and acute vasodilator test results if available. A ≥10-mm Hg decrease in mean pulmonary artery pressure to <40 mm Hg without a decrease in cardiac output defined vasoreactivity.8 Comorbid conditions such as renal insufficiency are determined by each investigator. Blood test results are categorized as low, normal, and high; however, results for brain natriuretic peptide (BNP) levels are quantitative. Qualitative echocardiographic results (none, mild, moderate, moderate/severe, or severe) are recorded for right ventricular dysfunction and pericardial effusion (yes/no), whereas numeric results are captured for Tei index only from those who routinely perform this modality. Information on other quantitative echocardiographic parameters such as tricuspid annular plane systolic excursion were not captured in this data set.
Missing and out-of-range data are queried at the point of data entry; additional clarifications are sent as queries to sites. Onsite monitoring of source data is performed at 20% of participating sites annually.
Survival was estimated from time of enrollment with all-cause mortality as the end point. Because of the observational nature of REVEAL, survival analyses involved a large number of candidate predictor variables from a wide range of diagnostic tests. As a result of natural practice pattern variation, few patients had every test performed, and some tests were performed more recently than others.
There were 2 steps to the survival analysis. First, univariable Cox regression models were used to identify subgroups with better-than-average, worse-than-average, poor, and extremely poor 1-year survival. One-year survival of 90% to 95% was considered average on the basis of the 1-year survival estimate for the full patient cohort, and the other 4 categories were defined on the basis of 5-percentage-point increments (eg, 85% to 90% for worse than average, 80% to 85% for poor, etc).
The univariable analyses identified predicted cut points to transform continuous variables into subgroups. An indicator variable was created for every subgroup associated with better- or worse-than-average survival but not for continuous or categorical variables associated with average survival. To avoid excluding patients with missing tests, univariable analyses were also performed for the “missing” category. If the indicator variable for “missing” was associated only with an average survival, patients with missing data became part of the reference group. For time-sensitive hemodynamic data, additional models were run, restricting to tests that had been performed within the 1 or 2 years preceding study entry. To account for the possibility of interactions, we created sex-specific age and WHO group I PAH subgroup pathogenesis indicators. No other interactions were considered.
In the second step of the survival analysis, the full set of indicator variables identified in the univariable analyses were entered into a stepwise multivariable Cox regression model. Because of the way that the indicator variables were created (ie, patients with missing data were a subgroup or were part of the reference group), all patients had a complete set of covariates. An α level of 0.05 was used for model entry in the primary analysis. Proportional-hazards assumptions were confirmed with a Kolmogorov-type supremum test.9
The discriminatory ability of the model was assessed with the c index.10 The assessment was repeated for subgroups of maximally treated patients and for WHO group I PAH subgroups to ensure generalizability. Cross-validation was used to compute the c index to approximate an independent validation.11 For each of 1000 bootstrap samples, the stepwise model was refit and reassessed on the original data. Following the approach of Harrell et al,12 optimism associated with both the c index and calibration was assessed so that a bias-corrected shrinkage factor could be included in the final predictive equation. The optimism estimates for discrimination were applied to compute corrected c indexes, and separately, optimism estimates for comparing predicted values with observed (Kaplan-Meier) values were computed for each 5-percentile increment in the distribution of predicted values.
Multiple sensitivity analyses were conducted (1) using α levels of 0.2, 0.1, 0.01, and 0.001 for model entry, (2) censoring at the time of transplantation, and (3) modeling survival from time of diagnosis rather than time of enrollment. Details of the following 3 aspects of the model development are located in the Appendix in the online-only Data Supplement: a more complete flow of the model-building process, sensitivity analyses using models of time from diagnosis, and bootstrap cross-validation.
Characteristics at Enrollment
A total of 2716 consecutively enrolled patients met all analysis criteria. Mean age was 50 years; 79% were women; and 73% were white (Table 1⇓). Forty-seven percent had idiopathic PAH (IPAH); 86% were in modified NYHA/WHO functional class II to III, and 14% were newly diagnosed by right heart catheterization within 3 months preceding enrollment. Median times from echocardiograms, hemodynamic measurements, and percent predicted carbon monoxide diffusing capacity (Dlco) to enrollment were 2.8, 11.2, and 15.4 months, respectively.
PAH therapies included prostacyclin analogs in 1092 (41.6%), endothelin receptor antagonists in 1231 (46.9%), and phosphodiesterase-5 inhibitors in 1301 (49.6%) patients. A total of 1087 (40.0%) and 687 (26.2%) patients received combination PAH therapies or an intravenous prostacyclin analog, respectively. Calcium channel blockers were used for PAH treatment or as a concomitant therapy for 251 (9.2%) and 428 (15.8%) patients, respectively.
The mean duration of follow-up (after enrollment) among survivors was 521 days (range, 0 to 731 days); 5 patients (0.2%) had no follow-up, and 97.5% of survivors were followed up for ≥12 months. There were 340 deaths, and 33 patients underwent lung transplantation. The observed 1-year survival from the date of enrollment was 91.0% (95% confidence interval [CI], 89.9 to 92.1; Figure 1).
Predictors of Survival
Several demographic, functional, laboratory, and hemodynamic parameters were independently associated with survival in the multivariable model (Figure 2). Variables associated with a >2-fold increase in hazard ratio (HR) included PAH associated with portal hypertension (HR, 3.6; 95% CI, 2.4 to 5.4), a family history of PAH (HR, 2.2; 95% CI, 1.2 to 4.0), men >60 years of age (HR, 2.2; 95% CI, 1.6 to 3.0), modified NYHA/WHO functional class IV (HR, 3.1; 95% CI, 2.2 to 4.4), and pulmonary vascular resistance >32 Wood units (HR, 4.1; 95% CI, 2.0 to 8.3). Other variables associated with significantly increased risk of death included PAH associated with connective tissue disease (CTD), renal insufficiency, modified NYHA/WHO functional class III, resting systolic blood pressure (BP) <110 mm Hg, resting heart rate >92 bpm, 6MWD <165 m, BNP >180 pg/mL, presence of pericardial effusion, percent predicted Dlco ≤32%, and mean right atrial pressure >20 mm Hg within the year preceding enrollment. Scleroderma and nonscleroderma CTD categories had similar coefficients and were combined in the final model. Four variables were associated with increased 1-year survival: modified NYHA/WHO functional class I, 6MWD ≥440 m, BNP <50 pg/mL, and percent predicted Dlco ≥80%.
Time from diagnosis was not independently associated with survival. Without adjustment for other variables, patients newly diagnosed within 90 days of enrollment had a nonsignificant elevated risk (HR=1.20; P=0.24) compared with patients diagnosed >90 days before enrollment, and no difference was present after adjustment for the multivariable model (HR=0.93; P=0.63). Without adjustment for other variables, years from diagnosis was significantly associated with a decreased risk (HR=0.94 per year; P=0.002); however, this difference did not persist after adjustment (HR=1.01 per year; P=0.53).
Sensitivity Analysis: Censoring at Transplant
Results of censoring patients at the time of lung transplant were consistent with those in the primary analysis. All terms remained in the model using the prespecified α=0.05 criterion.
Based on the Cox proportional-hazard multivariable analysis, a preliminary prognostic equation was derived from the independent prognosticators of survival. Patients were divided into 20 equally sized groups (half-deciles) stratified by predicted survival. The Kaplan-Meier estimates for each group were compared with the preliminary predicted survival (Figure 3A), showing good apparent calibration. Optimism-corrected Kaplan-Meier estimates (Figure 3B) suggested a need for a small shrinkage adjustment, primarily because the optimism-corrected Kaplan-Meier estimate for the lowest half-decile demonstrated slightly better survival than predicted (58.8% versus 54.4%). After application of a shrinkage correction to develop a final prognostic equation, the calibration was nearly perfect (Figure 3C).
Predicted 1-year survival is computed as follows: S0(1)exp(Z′βγ) where S0(1) is the baseline survivor function (0.9698), Z′β is the linear component, and γ is the shrinkage coefficient (0.939). The core of the prognostic equation is Z′β, the linear component of the Cox model presented in Figure 2. Starting with a base value of 0, the linear component is increased or decreased according to the variable coefficients summarized in Table 2.
To further summarize the mortality risk stratification produced by the final prognostic equation, 5 risk groups were defined. The predicted risk, after shrinkage correction, was classified as low (>95% 1-year survival) for 1374 patients. Patients in the low risk category had a median of 1 of a possible 15 risk factors and a median of 1 of 4 possible protective factors. The average risk (90% to 95% 1-year survival), moderately high risk (85% to 90% 1-year survival), high risk (70% to 85% 1-year survival), and very high risk (<70% 1-year survival) strata had a median of 2, 3, 4, and 6 risk factors, respectively, and a median of 0 protective factors. Because of the nature of the formula, patients within the same strata with more risk factors than the median were more likely to have ≥1 protective factors. Patients in higher risk categories had proportionately lower observed 1-year survival (Figure 4). Table 3 describes a sample patient for each of the 5 strata.
In addition to calibration and shrinkage, discrimination was assessed. The c index, defined as the probability that a randomly chosen survivor has a lower risk estimate than a randomly chosen death, was calculated for estimates from the NIH survival equation and for the REVEAL multivariable model, before and after bootstrap correction for optimism (Table 4). The ability of the REVEAL multivariable model to discriminate between low- and high-risk patients was considerably greater than that of the NIH survival equation (0.772 versus 0.588), even after correcting for the optimism bias inherent in using the same data set for developing and assessing the model (corrected c index=0.744). The model was also able to accurately discriminate between higher- and lower-risk patients in these specific subgroups: (1) maximally treated patients (ie, on intravenous prostacyclins or combination PAH therapies), (2) newly diagnosed patients, (3) pulmonary capillary wedge pressure <12 mm Hg, and (4) IPAH/familial PAH or other forms of PAH. Estimates were consistently greater than those from the NIH survival equation for IPAH/familial PAH even when many of the tests associated with the indicator variables in the model were unavailable.
One of the major goals of the REVEAL Registry was to design a widely applicable, contemporary, clinically relevant model to predict outcome in patients with WHO group I PAH. Through analysis of multiple prognostic factors in 2716 consecutively enrolled patients with PAH, we developed a prognostic equation that predicts 1-year survival. Multiple, incremental clinical measures in the equation make it a more valuable predictor of survival compared with each measure assessed individually.
Assessment of prognosis guides individual therapeutic decisions. Using information commonly obtained in patients with WHO group I PAH, one can calculate the risk and estimate 1-year survival from time of assessment (regardless of when the patient was diagnosed). The prognostic equation was developed to be applicable at any point in the course of the disease based on the patient’s most recent evaluation. The ability of the equation to discriminate between lower- and higher-risk patients was demonstrated in the entire cohort and in several clinical subsets (listed in Table 4). Patients with more protective factors than risk factors (n=396) exhibited 1-year survival of 98.7%, and patients in the lowest risk decile had a predicted 1-year survival of 99.0%, similar to the age-adjusted estimated survival rate for the general US population in 2005 (99.2%).13 Specific PAH therapies were not included as candidate predictors of survival in this study for 2 reasons. First, we believe that these determinations are best left to head-to-head randomized controlled trials. Second, and more important, prognosis is more related to a change of the specific therapy in a modifiable risk factor (ie, 6MWD, BNP, or hemodynamic parameter) than to an individual class of therapy per se. Thus, the relative importance of a particular drug in changing prognosis is diluted by the change in a patient’s functional capacity.
This study assimilates previously noted and newly confirmed PAH prognostic findings into a cohesive predictive formula that weighs each one and resolves the relevance of each factor. Importantly, the risk assessment is derived from a multivariable model and thus weighs each risk factor within the context of the other variables. Therefore, some subgroups such as CTD-scleroderma have lower HRs than might have been expected because of other associated risk factors that also contribute to the prognostic equation. This illustrates the importance of a quantitative model rather than a qualitative assessment of the risk factors for each patient. The multivariable model provides considerably better risk stratification than any single variable used alone. Functional class, 6MWD, and BNP variables each had better prognostic value than the NIH equation but had less discriminatory capacity than the full prognostic equations.
Our analyses confirmed increased mortality risk in patients with PAH associated with portal hypertension14 or scleroderma15 (and found increased risk among patients with PAH associated with any CTD). Increased mortality risk was also confirmed in patients with renal insufficiency16 or any pericardial effusion on echocardiogram.17 Previous investigators have demonstrated low systolic BP at peak exercise to be associated with poor outcome18; we found that resting systolic BP <110 mm Hg and resting heart rate >92 bpm were associated with worse survival. Prior studies have demonstrated the utility of 6MWD in predicting outcome,19 but a recent meta-analysis has raised doubts about this association.20 We found that 6MWD thresholds of ≥440 m are associated with longer survival and <165 m with increased mortality. We have extended the utility of low percent predicted Dlco as a prognosticator, demonstrating that it is also an important discriminator at high levels even after the exclusion of patients with anemia and despite differing methodologies used at multiple sites.
The NIH registry identified mean right atrial pressure, cardiac index, and mean pulmonary artery pressure as important predictors of survival.1 Our analyses confirmed the importance of hemodynamic parameters obtained by right heart catheterization. However, in contrast to the NIH registry and consistent with contemporary hypotheses, mean pulmonary artery pressure was not an important predictor of survival. When adjusted for all other risk factors making up the final multivariable model, only an elevated mean right atrial pressure within the year preceding study enrollment and a markedly increased pulmonary vascular resistance were independent risk predictors.
Despite its reported importance in predicting outcome, particularly in IPAH,8,21,22 acute vasoreactivity did not result in an overall survival advantage at 1 year when weighed against other evaluable factors in the multivariable analysis. The 155 vasoreactive patients were more often in the lowest predicted risk category (59% versus 51% overall), with the greatest differences seen among 83 vasoreactive patients with IPAH, suggesting that the advantages of being acutely vasoreactive are captured by other variables. Conceivably, the relatively low importance of acute vasoreactivity may also be due to the inclusion in our analysis of all patients with WHO group I PAH. This may have minimized the overall survival effect of this particular factor because the degree of vasoreactivity important in predicting outcome is seen in only a small proportion of patients with IPAH.22 We do not believe, however, that this finding should question the usefulness of vasoreactive testing or suggest that testing and treating vasoreactive patients should be abandoned.
Certain findings in our analysis constitute new associations. We noted a survival disadvantage for those with a family history of PAH. Prior studies reported no significant difference in survival when patients with IPAH and familial PAH with and without BMPR2 mutations were compared, despite earlier onset and more severe disease in BMPR2 mutation–positive cases.23 Because BMPR2 mutations are detected in only ≈70% of familial cases, it is possible that family history identifies increased mortality risk better than BMRP2 mutation detection because alternative, yet unidentified, mutations also may be more closely linked to survival.
Congenital heart disease (CHD), interestingly, was not associated with a survival advantage, regardless of the type of defect or repair status. We were also unable to detect a difference between patients with repaired and unrepaired CHD. The reason for this is unclear but may simply reflect the lesser importance of demographic factors when compared against factors that depict the clinical status of the patient or the reduced power to detect differences in small subgroups with relatively few events. Previous reports of better survival in patients with PAH associated with CHD (APAH-CHD) versus patients with IPAH/familial PAH or PAH associated with CTD were predominantly natural history data derived from unoperated Eisenmenger patients and do not include patients with PAH associated with CHD that has been repaired or patients with small, clinically insignificant congenital systemic to pulmonary shunts, whether or not they are on PAH therapies. Current PAH treatment regimens (including surgical repair now performed more often in patients with increased pulmonary vascular resistance than in the past) suggest that either patients with PAH associated with CHD are being treated less aggressively than other patients with PAH or their response to therapy may be smaller.
Although PAH is predominantly a female disorder, we demonstrated that men >60 years of age have poorer survival compared with men ≤60 years of age at the time of assessment and compared with female patients regardless of age. Despite the subjective nature of functional class assessment, we identified the full 4-category range of functional class as having discriminatory power. Others have identified elevated BNP as a marker for poor survival, but we found that lower-than-average BNPs (<50 pg/mL) are a marker for better survival.
There are several limitations to this study. Unlike randomized clinical trials, all relevant measurements were not collected at mandated study visits; only 24.2% of patients had data available for all 19 of the possible risk factors and protective factors, and the average patient had data available for 16 of the 19. However, our use of a missing data indicator allowed us to include all of these patients, making the model broadly generalizable to clinical practice. Inclusion of 131 patients ≤18 years of age expands the target cohort for the proposed prognostic equation, even though a model developed exclusively for pediatric patients might have led to different cut points for some components of the model. Among all patients, there are some hypothesized predictors that were not captured in our database such as tricuspid annular plane systolic excursion, serum sodium levels, and quantitative measures of renal and hepatic function; these should be examined in future studies.
Although the study of a predominantly prevalent patient population may be perceived as incurring a potential survival bias, recent diagnosis was not an independent prognostic variable in our analysis. Three important predictors (BNP, percent Dlco, and 6MWD) were available exclusively or predominantly at the time of enrollment and not at the time of diagnosis and were missing more often among newly diagnosed compared with previously diagnosed patients. Because each of these 3 variables was associated with both increased and decreased risk, the absence of these tests for individual patients does not bias the predictions upward or downward. Separate validation of the newly diagnosed cohort (Table 4) suggested nearly identical discriminatory ability compared with the total cohort, and sensitivity analyses using left truncation methodology from the time of diagnosis yielded similar results for the prognostic factors available at both time points (see the Appendix in the online-only Data Supplement). Although predicting survival from the time of diagnosis may be valuable, assessing survival risk at any point of disease progression may be even more useful. Suggesting that the prognostic equation may be used at any time is not equivalent to suggesting that changes in the calculated risk, observed through serial measurements, have independent prognostic value. Serial assessment was not evaluated in our analysis. The present analysis does tell us what to expect on the basis of the most recent data at any point in the patient’s clinical course, but we have not tested the implications of an upward or downward clinical trajectory.
External validation of the model, preferably including serial assessment, may be considered an important step before widespread application of the prognostic equation. However, the bootstrap cross-validation technique is a statistically rigorous approach that is comparable to external validation in that the patient data used to develop the model are not used in the assessment of the discriminatory power of the model. This protects against overfitting and producing spurious findings and provides the most important diagnostic evaluation of model validity.
We identified key predictors of survival for patients with WHO group I PAH and present them in a weighted prognostic equation. We envision that this equation may be used at diagnosis or at any time during a patient’s course. Its potential use as a serial measure may allow regular reassessment of risk and differentiation of patients with stable chronic disease from those with actively progressive disease. By obtaining an evidence-based, global assessment of the patient, clinicians may be better able to individualize and optimize therapeutic strategies to improve survival. Using this equation in future clinical trials will test these hypotheses and guide initial and stepwise therapies. Further investigation with longer follow-up is warranted.
We thank all patients, principal investigators, and study coordinators for their participation in REVEAL. We also thank Sharon Safrin, MD, and Bill Prucka, PhD, for medical and statistical consultative services, as well as Jennifer M. Kulak, PhD, of inScience Communications, a part of the Wolters Kluwer organization, for editorial support. Preparation of this manuscript was supported by Actelion Pharmaceuticals US, Inc.
Source of Funding
Funding for the REVEAL Registry is provided by Actelion Pharmaceuticals US, Inc.
Dr Benza has received honoraria from Actelion, United Therapeutics, and Gilead and has received or is pending receipt of grants from Actelion, United Therapeutics, Gilead, and Lung Rx. Dr Benza has received honoraria for his service on the REVEAL Steering Committee, which is supported by Actelion. D.P. Miller and A.J. Foreman are employed by ICON Clinical Research, a company that receives research support from Actelion and other pharmaceutical companies. Dr Gomberg-Maitland has received research grant support from Actelion, Gilead, Eli Lilly and Co/ICOS, Novartis, Pfizer, and United Therapeutics and has served as a consultant and/or on advisory boards for Biomarin, Gilead, Medtronic, Millenium, and Pfizer. She has a patent filed for the use of sorafenib in pulmonary hypertension entitled “Compositions and Methods for Treating Pulmonary Hypertension,” WO/2007/087575. Dr Frantz has served on advisory boards for Actelion, Gilead, Lung Rx, United Therapeutics, Pfizer, and Medtronic and has served as an investigator for multicenter trials sponsored by these companies. Honoraria for advisory board activities have gone into a Mayo Clinic research account in compliance with Mayo Clinic guidelines for consulting activities when the consultant is an investigator on research studies funded by the company. Dr Frost serves as a consultant for Gilead and Actelion. Dr Frost has received honoraria from Gilead, Actelion, and Pfizer and has provided expert testimony on diet pill litigation. She has also received grants from Gilead and Actelion and grants to Baylor for Institutional Review Board–approved research. Dr Frost has received honoraria for her service on the REVEAL Steering Committee, which is supported by Actelion. Dr Barst serves as a consultant for and has received honoraria from Actelion, Bayer, GeneraMedix, Gilead, Eli Lilly and Co, MondoBIOTECH, and Pfizer. Dr Barst has provided expert testimony on diet pill litigation for the plaintiffs and has also received grants from Actelion, GeneraMedix, Gilead, Eli Lilly and Co, NIH/NHLBI, Novartis, Pfizer, and United Therapeutics. Dr Barst has received honoraria for her service on the REVEAL Steering Committee, which is supported by Actelion. Dr Badesch has received honoraria for service on steering committees and/or advisory boards for Actelion/CoTherix, Gilead/Myogen, Encysive Pharmaceuticals, Pfizer, GlaxoSmithKline, Lung Rx, United Therapeutics, Eli Lilly and Co/ICOS, Biogen Idec, and MondoBIOTECH. Dr Badesch has received grants from Actelion/CoTherix, Gilead/Myogen, Encysive Pharmaceuticals, Pfizer, United Therapeutics, Lung Rx, and Eli Lilly & Co/ICOS, and NIH/NHLBI. Dr Badesch has received honoraria for his service on the REVEAL Steering Committee, which is supported by Actelion. Dr Elliott is employed by Intermountain Healthcare. Intermountain Healthcare, with Dr. Elliott as Principal Investigator, has received grant support during the past 5 years from Actelion, Pfizer, Encysive Pharmaceuticals, and United Therapeutics. Dr Elliott has received honoraria for his service on the REVEAL Steering Committee, which is supported by Actelion. Dr Liou has received grants from the NIH/NHLBI, the Margolis Family Foundation of Utah, and the CF Foundation. He has been the site principal investigator for studies of cystic fibrosis and its treatment for the Therapeutic Development Network of the CF Foundation, Altus, Axcan Scandipharm, Bayer, Boehringer, Genentech, Gilead, Inspire, Kalobios, MPEX, Novartis, and Vertex. Dr Liou has received honoraria for his service on the REVEAL Steering Committee, which is supported by Actelion. Dr McGoon serves as a consultant with Actelion/CoTherix, Gilead/Myogen, Lung Rx, and Medtronic. Dr McGoon has received grants from Gilead/Myogen and Medtronic and honoraria for his service on the REVEAL Steering Committee, which is supported by Actelion. Dr Coffey has nothing to disclose.
D'Alonzo GE, Barst RJ, Ayres SM, Bergofsky EH, Brundage BH, Detre KM, Fishman AP, Goldring RM, Groves BM, Kernis JT, Levy PS, Pietra GG, Reid LM, Reeves JT, Rich S, Vreim CE, Williams GW, Wu M. Survival in patients with primary pulmonary hypertension: results from a national prospective registry. Ann Intern Med. 1991; 115: 343–349.
Galie N, Manes A, Negro L, Palazzini M, Bacchi-Reggiani ML, Branzi A. A meta-analysis of randomized controlled trials in pulmonary arterial hypertension. Eur Heart J. 2009; 30: 394–403.
Sitbon O, Humbert M, Jais X, Ioos V, Hamid AM, Provencher S, Garcia G, Parent F, Herve P, Simonneau G. Long-term response to calcium channel blockers in idiopathic pulmonary arterial hypertension. Circulation. 2005; 111: 3105–3111.
Lin D, Wei L, Ying Z. Checking the Cox model with cumulative sums of Martingale-based residuals. Biometrika. 1993; 80: 557–572.
Shah SJ, Thenappan T, Rich S, Tian L, Archer SL, Gomberg-Maitland M. Association of serum creatinine with abnormal hemodynamics and mortality in pulmonary arterial hypertension. Circulation. 2008; 117: 2475–2483.
Raymond RJ, Hinderliter AL, Willis PW, Ralph D, Caldwell EJ, Williams W, Ettinger NA, Hill NS, Summer WR, de Boisblanc B, Schwartz T, Koch G, Clayton LM, Jobsis MM, Crow JW, Long W. Echocardiographic predictors of adverse outcomes in primary pulmonary hypertension. J Am Coll Cardiol. 2002; 39: 1214–1219.
Wensel R, Opitz CF, Anker SD, Winkler J, Hoffken G, Kleber FX, Sharma R, Hummel M, Hetzer R, Ewert R. Assessment of survival in patients with primary pulmonary hypertension: importance of cardiopulmonary exercise testing. Circulation. 2002; 106: 319–324.
Miyamoto S, Nagaya N, Satoh T, Kyotani S, Sakamaki F, Fujita M, Nakanishi N, Miyatake K. Clinical correlates and prognostic significance of six-minute walk test in patients with primary pulmonary hypertension: comparison with cardiopulmonary exercise testing. Am J Respir Crit Care Med. 2000; 161: 487–492.
Humbert M, Sitbon O, Chaouat A, Bertocchi M, Habib G, Gressin V, Yaici A, Weitzenblum E, Cordier JF, Chabot F, Dromer C, Pison C, Reynaud-Gaubert M, Haloun A, Laurent M, Hachulla E, Simonneau G. Pulmonary arterial hypertension in France: results from a national registry. Am J Respir Crit Care Med. 2006; 173: 1023–1030.
The Registry to Evaluate Early and Long-Term Pulmonary Arterial Hypertension Disease Management (REVEAL Registry) was designed to assess longitudinal clinical course and disease management in the largest cohort of patients with pulmonary arterial hypertension ever monitored. Pulmonary arterial hypertension remains a morbid disease unless well-timed clinical intervention is implemented. Therefore, factors that determine survival in pulmonary arterial hypertension can significantly drive and focus clinical management. We analyzed data from 2716 patients with pulmonary arterial hypertension to derive a multivariable, weighted risk formula that could be used by the practicing clinician at any time in the course of a patient’s disease progression to predict survival. Nineteen independent factors were identified as having an impact on patient survival. A multivariable risk formula comprising all 19 factors provided a more accurate assessment of clinical outcome than each independent variable. These results emphasize the importance of using the full spectrum of clinical data commonly available to the practicing clinician for the assessment of patients with pulmonary arterial hypertension. We believe that the risk stratification provided by this predictive equation will facilitate counseling of patients about their disease and prognosis and will provide a benchmark for prospective evaluation of new therapies.
The online-only Data Supplement is available with this article at http://circ.ahajournals.org/cgi/content/full/CIRCULATIONAHA.109.898122/DC1.