The REVEAL Registry Risk Score Calculator in Patients Newly Diagnosed With Pulmonary Arterial Hypertension

Transcription

1 CHEST Original Research PULMONARY VASCULAR DISEASE The REVEAL Registry Risk Score Calculator in Patients Newly Diagnosed With Pulmonary Arterial Hypertension Raymond L. Benza, MD ; Mardi Gomberg-Maitland, MD, FCCP ; Dave P. Miller, MS ; Adaani Frost, MD, FCCP ; Robert P. Frantz, MD ; Aimee J. Foreman, MA ; David B. Badesch, MD, FCCP ; and Michael D. McGoon, MD, FCCP Background: In pulmonary arterial hypertension (PAH), survival predictions can be important for optimization of therapeutic strategies. The present study aimed to validate a quantitative algorithm for predicting survival derived from the Registry to Evaluate Early and Long-term PAH Disease Management (REVEAL Registry) and develop a simplified calculator for everyday clinical use. Methods: Prospectively collected data from patients with newly diagnosed (, 3 months) World Health Organization group I pulmonary hypertension enrolled in the REVEAL Registry were used to validate a predictive algorithm for 1-year survival. Model calibration was evaluated by comparing algorithm-predicted survival with observed Kaplan-Meier estimates for the overall validation cohort and for five risk groups. Similarly, the risk discriminators for the simplified calculator were compared with those of the quantitative algorithm. Results: The validation cohort comprised 504 individuals with mean SD 6-min walk distance m, and 61.5% were functional class III. The proportion of patients surviving 1 year fell within the range predicted by the model (95.1%, 91.5%, 84.6%, 76.3%, and 58.2%, respectively) among patients in the low (predicted survival 95%), average (90% to, 95%), moderate (85% to, 90%), high (70% to, 85%), and very high (, 70%) risk strata. Predicted and observed 1-year survival were similar across risk stratum, and the c-index indicated good discrimination for both the full equation (0.726) and the simplified risk calculator (0.724). Conclusions: The REVEAL Registry predictive algorithm and simplified risk score calculator are well calibrated and demonstrate good discriminatory ability in patients with newly or previously diagnosed PAH. Trial registry: ClinicalTrials.gov; No.: NCT ; URL: CHEST 2012; 141(2): Abbreviations: 6MWD 5 6-minute walk distance; D lco 5 diffusing capacity of lung for carbon monoxide; NYHA 5 New York Heart Association; PAH 5 pulmonary arterial hypertension; REVEAL Registry 5 Registry to Evaluate Early and Long-term Pulmonary Arterial Hypertension Disease Management; RHC 5 right-sided heart catheterization; WHO 5 World Health Organization Pulmonary arterial hypertension (PAH) is characterized by progressive fibrotic and proliferative changes in the small pulmonary arteries leading to a gradual increase in pulmonary vascular resistance, right ventricular failure, and premature death. 1 PAH is a rare condition with estimates of annual incidence ranging from 1.1 to 7.6 cases per million people and estimates of the prevalence ranging from 15 to 26 cases per million people. 2,3 For this reason, PAH has been difficult to study, and much of what is known about the condition is derived from patient registries. The largest registry of patients with Pulmonary Arterial Hypertension established to date is the multicenter, observational US-based Registry to Evaluate Early and Long-term Pulmonary Arterial Hypertension Disease Management (REVEAL Registry). 4 Assessment of prognosis in patients with PAH guides therapeutic decision making, and one of the prespecified objectives of the REVEAL Registry was to identify predictors of short- and long-term survival in the context of current treatment and clinical variables. The development of a quantitative algorithm for predicting 354 Original Research

2 survival in PAH using prospectively collected data from 2,716 patients enrolled in the REVEAL Registry database has been previously reported. 5 The objective of the current analysis was to prospectively validate the prognostic equation in a cohort of patients with newly diagnosed PAH and to develop a simplified risk calculator based on the prognostic equation. Materials and Methods A comprehensive description of the design of the REVEAL Registry, including inclusion and exclusion criteria and statistical methods, has been published, as has a description of the 2,716 patients included in the model development cohort and the methods used to develop the predictive survival algorithm. 4,5 Variables independently associated with increased mortality in the multivariable survival model included a collection of etiologic factors and parameters that can be identified by physical examination or laboratory tests: men aged. 60 years, PAH associated with portal hypertension, PAH associated with connective tissue disease, family history of PAH, modified New York Heart Association (NYHA)/World Health Organization (WHO) functional class III or IV, renal insufficiency, resting systolic BP, 110 mm Hg, heart rate. 92 beats/min, mean right atrial pressure. 20 mm Hg, 6-min walk distance (6MWD), 165 m, brain natriuretic peptide (BNP). 180 pg/ml, pulmonary vascular resistance. 32 Wood units, % predicted diffusing capacity of lung for carbon monoxide (D lco ) 32%, and the presence of pericardial effusion on echocardiogram. 5 The four variables associated with increased 1-year survival were modified NYHA/WHO functional class I, 6MWD 440 m, BNP, 50 pg/ml, and % predicted D lco 80%. An individual patient s predicted probability of survival through 1 year is (S 0 (1)exp(Z 9 b g ) ), where S 0 (1) , g , and Z 9 b is the sum of the patient s individual characteristics multiplied by the b coefficients for each of the 19 parameters in the model. 5 Development of Risk Calculator Based on the multivariable Cox proportional hazard model, a risk calculator was derived assigning weighted values for variables that were independent prognosticators of survival in the Manuscript received March 17, 2011; revision accepted May 18, Affiliations: From The Gerald McGinnis Cardiovascular Institute, Allegheny General Hospital (Dr Benza), Pittsburgh, PA; the Department of Medicine, Section of Cardiology, University of Chicago Medical Center (Dr Gomberg-Maitland), Chicago, IL; Statistical Analysis, ICON Late Phase & Outcomes Research (Mr Miller and Ms Foreman), San Francisco, CA; Department of Medicine, Section of Pulmonary, Critical Care, and Sleep Medicine, Baylor College of Medicine (Dr Frost), Houston, TX; Division of Cardiovascular Diseases, Mayo Clinic (Drs Frantz and McGoon), Rochester, MN; and Department of Medicine, Division of Pulmonary Sciences and Critical Care Medicine, and Cardiology, University of Colorado (Dr Badesch), Denver, CO. Funding/Support: Funding for the REVEAL Registry is provided by Actelion Pharmaceuticals US, Inc. Correspondence to: Raymond L. Benza, MD, The Gerald McGinnis Cardiovascular Institute, Allegheny General Hospital, 320 E North Ave, 16th Floor, South Tower, Pittsburgh, PA 15212; RBENZA@wpahs.org 2012 American College of Chest Physicians. Reproduction of this article is prohibited without written permission from the American College of Chest Physicians ( site/misc/reprints.xhtml ). DOI: /chest model development cohort ( Fig 1 ). Two points were assigned for each indicator that was associated with at least a twofold increase in the instantaneous hazard for mortality, and one point was assigned for smaller, statistically significant effects. The b coefficient associated with a twofold increase is Thus, the corresponding model b coefficients for variables with hazard ratios below and above a twofold increase were approximately 0.5 and 1.0. Integer scores of 1 and 2 were used to facilitate a simpler integer scale rather than a half-point scale. Points were subtracted from the total score for indicators associated with an increased likelihood of survival. For a given test, a point value of zero was assigned if it was either unmeasured or if it did not fall above or below the threshold used in the risk calculator. This is consistent with the way that missing values were included as part of the reference group in the model development. Kaplan- Meier estimates were produced for strata based on the final risk score, and a risk calculator was created for easy tabulation of the risk score. Validation Cohort The REVEAL Registry protocol was amended to supplement the primary enrollment cohort of patients with newly and previously diagnosed PAH with an additional cohort made up exclusively of patients with newly diagnosed disease. The model development cohort was derived from the primary enrollment cohort. The validation cohort consisted of consecutive patients with newly diagnosed PAH enrolled via the amendment into the supplemental cohort. Patients in both cohorts were eligible regardless of prior treatment, and newly diagnosed PAH was defined as enrollment within 3 months of a diagnostic right-sided heart catheterization (RHC). Patients of all ages were enrolled, excluding only infants diagnosed at 3 months of age. 4 The protocol was approved by the ethics board at each participating center (e-appendix 1), and all patients provided written informed consent before enrolling in the study. Statistical Methods Kaplan-Meier curves were used to evaluate survival. To avoid immortal time bias 6 and to ensure that risk factors were assessed prior to follow-up, the first day at risk was the day of enrollment rather than the day of diagnosis. The performance of the prognostic model was evaluated for calibration and discrimination. Model calibration was evaluated by comparing survival as predicted by the algorithm with the observed Kaplan-Meier estimates for the validation cohort. We used the calibration plot as a means of showing agreement between predicted and observed 1-year survival. Kaplan-Meier estimates of survival were reported for the overall validation cohort and for five groups stratified by predicted risk as defined in the model development phase of the project. 5 Discrimination for the prognostic equation (algorithm) and simplified calculator-derived risk score were assessed by the c-index, a measure of how well the predictive model differentiates between patients who die and those who survive. 7 Given a pair of patients, one who dies and one who survives, the c-index is the probability that the patient who died had the higher predicted chance of death. In this case, the higher the c-index (ie, closer to the maximum value of 1.0) the better the model discriminates between patients likely to die and those likely to survive. For binary outcomes, the c-index is also sometimes known as the area under the receiver operating characteristics curve, which captures the trade-off between sensitivity and specificity. The interpretation is similar in the survival setting; however, discrimination is not limited to two groups. In assessing discrimination, patients who die are compared not only to survivors but also to patients who die earlier or later. CHEST / 141 / 2 / FEBRUARY,

3 Figure 1. REVEAL Registry PAH risk score calculator. Calculated risk scores can range from 0 (lowest risk) to 22 (highest risk). If N-terminal probnp is available and BNP is not, listed cut points are replaced with, 300 pg/ml and pg/ml. APAH 5 associated pulmonary arterial hypertension; BNP 5 brain natriuretic peptide; BPM 5 beats per minute; CTD 5 connective tissue disease; DLco 5 diffusing capacity of lung for carbon monoxide; FPAH 5 familial pulmonary arterial hypertension; HR 5 heart rate; mrap 5 mean right atrial pressure; NYHA 5 New York Heart Association; PAH 5 pulmonary arterial hypertension; PoPH 5 portopulmonary hypertension; PVR 5 pulmonary vascular resistance; REVEAL Registry 5 Registry to Evaluate Early and Long-term Pulmonary Arterial Hypertension Disease Management; SBP 5 systolic BP; WHO 5 World Health Organization. Results The validation cohort comprised 504 individuals with newly diagnosed PAH who were enrolled between September 2007 and December 2009 and who met the traditional hemodynamic criteria (pulmonary capillary wedge pressure 15 mm Hg). The characteristics of these patients and of those in the model development cohort are shown in Table 1. At the time of enrollment, the mean age of patients in the validation cohort was 53 years, 74% were women, and 69% were white. Compared with the model development cohort, the validation cohort had lower mean 6MWD (308 m vs 370 m), mean serum BNP level was higher (490 pg/ml vs 286 pg/ml), mean D lco 356 Original Research

4 Table 1 Characteristics of Patients at Time of Enrollment in REVEAL Registry Characteristic Validation Cohort (n 5 504) Development Cohort (n 5 2,716) P Value Patients with newly diagnosed PAH 504 (100.0) 367 (13.5),.001 Age, y Female 373 (74.0) 2,135 (78.6).022 Race.012 White 347 (68.9) 1,969 (72.5) Black 85 (16.9) 327 (12.0) Other/unknown 72 (14.3) 420 (15.5) WHO group I PAH subgroup,.001 Idiopathic PAH 239 (47.4) 1,262 (46.5) Familial PAH 13 (2.6) 79 (2.9) Associated PAH Congenital heart disease 24 (4.8) 319 (11.8) Connective tissue disease 156 (31.0) 648 (23.9) Portopulmonary hypertension 34 (6.8) 138 (5.1) Other 38 (7.5) 270 (9.9) Modified NYHA/WHO functional class, I 12 (2.9) 210 (8.5) II 85 (20.3) 936 (37.8) III 257 (61.5) 1,194 (48.2) IV 64 (15.3) 136 (5.5) 6MWD, m ,.001 Right atrial pressure, mm Hg ,.001 Mean pulmonary artery pressure, mm Hg Pulmonary vascular resistance, Wood units Pulmonary capillary wedge pressure, mm Hg Serum BNP level, pg/ml ,.001 N-terminal probnp level, pg/ml Carbon monoxide diffusing capacity, % predicted ,.001 Systolic BP, mm Hg ,.001 Heart rate, beats/min Pericardial effusion, yes 105 (30.2) 532 (25.3).054 Renal insufficiency 25 (5.0) 109 (4.0).33 Values are expressed as the mean SD or No. (%). 6MWD 56-min walk distance; BNP 5 B-type natriuretic peptide; NYHA 5 New York Heart Association; PAH 5 pulmonary arterial hypertension; REVEAL Registry 5 Registry to Evaluate Early and Long-term Pulmonary Arterial Hypertension Disease Management; WHO 5 World Health Organization. was lower (50.1% vs 59.9%), and a higher proportion of patients (76.8% vs 53.7%) were classified as being in NYHA functional class III or IV. A similar proportion of patients in each cohort had idiopathic PAH (47.4% vs 46.5%). With regard to the cause of associated PAH, a higher proportion of patients in the validation cohort had connective tissue disease (31.0% vs 23.9%), and a lower proportion had congenital heart disease (4.8% vs 11.8%) compared with the development cohort ( Table 1 ). The median duration of follow-up for survivors in the validation cohort was 18 months (interquartile range: months). The risk score was tested using the validation cohort ( Fig 2 ). The risk calculator has a minimum risk score of 0, a maximum of 22, and is designed for clinical use. In the REVEAL Registry model development cohort, the mean risk score was 7.4 and ranged from 1 to 15 at the time of enrollment. One-year survival curves for patients in the validation cohort using the prognostic equation and the risk calculator are presented in Figures 3A and 3B, respectively, for each of the five risk strata. Predicted 1-year survival is 95% to 100% in the low-risk group, 90% to, 95% in the average-risk group, 85% to, 90% in the moderately high-risk group, 70% to, 85% in the high-risk group, and, 70% in the very high-risk group. The calibration plots for both the prognostic equation ( Fig 4A ) and the risk calculator ( Fig 4B ) show close agreement between the predicted and observed 1-year survival rates in each of the risk strata. The c-index for the prognostic equation in the overall validation cohort was (95% CI, ) and was similar in the subgroup of patients with idiopathic PAH (0.707; 95% CI, ). In patients with PAH associated with connective tissue disease, the c-index was (95% CI, ). The bias-corrected c-index for the risk calculator in the model development cohort was based on bootstrap sampling and was calculated to be 0.735, indicating that it had similar discriminatory ability to the complete prognostic equation. The c-index for the risk CHEST / 141 / 2 / FEBRUARY,

5 Figure 2. Twelve-month Kaplan-Meier survival estimate for the REVEAL Registry development cohort with predicted risk score. Risk strata are indicated by the lines: predicted 1-year survival is 95% to 100% in the low-risk group, 90% to, 95% in the averagerisk group, 85% to, 90% in the moderately high-risk group, 70% to, 85% in the high-risk group, and, 70% in the very high-risk group. See Figure 1 legend for expansion of abbreviation. calculator in the validation cohort was (95% CI, ). As an additional sensitivity analysis, the subset of 158 patients who were enrolled within 3 days of diagnosis was analyzed separately. One-year survival was similar for this subgroup and the remainder of the cohort (82.9% vs 86.4%) as was the c-index for the prognostic equation (0.729 vs 0.724) and the c-index for the risk calculator (0.736 vs 0.719). Because the validation cohort consisted exclusively of patients with newly diagnosed disease, a secondary analysis was conducted using the second year of follow-up among all patients enrolled in the REVEAL Registry. This allowed for a secondary validation among patients with previously diagnosed disease. Observed survival for patients predicted to have low, average, moderately high, high, or very high risk were 97.2%, 92.6%, 84.6%, 73.3%, and 52.3%, respectively, further demonstrating good discrimination and calibration. Discussion PAH is a comparatively rare condition, which has made it historically difficult to study. For this reason, large registry studies are useful in better defining the condition and for providing insight into factors that influence survival and prognosis. When compared with previous registry studies, 8-15 The REVEAL Registry is the largest and most comprehensive registry to date. Indeed, the size of the validation cohort the subject of the present analysis is as large as or larger than previous general registries established for the study of PAH. 8,9,11,12,14,16,17 The results of this analysis demonstrate that the predictive equation developed with data from 2,716 patients with previously diagnosed and newly diagnosed disease in the model development phase is well calibrated and has good discrimination. During the development phase, five strata were delineated on the basis of predicted 1-year survival, 5 and although the overall survival was somewhat lower in the validation cohort than in the larger development cohort, the observed survival of patients in each stratum fell within the range predicted by the equation. One potential limitation of any analysis of the REVEAL Registry is that it is subject to survivor bias. This is because the majority of patients included in the registry (86.5%) had established disease at the time of recruitment. 5 However, it must be stressed that recent diagnosis was not a significant predictor of mortality in the model,5 and, importantly, all patients included in the validation cohort had a new diagnosis of PAH, which was defined as having been established by RHC within the 3 months preceding enrollment. The calibration plot demonstrates very close agreement between predicted and actual survival in each of the five strata, suggesting that the prognostic equation fully accounts for the survivor bias such that the model developed on a pre dominantly prevalent cohort is also a good model for patients with newly diagnosed disease. It is noteworthy that only patients with newly diagnosed disease were used for the validation, avoiding the need to use advanced methods for estimating time to diagnosis using left truncation. 18 The date of diagnosis and the date of enrollment do differ slightly, as it was not practical to obtain consent from every patient on the exact day of diagnosis, so day of enrollment is considered the first day at risk. The sensitivity analyses suggest that there were not major differences in survival between those with a short delay between diagnosis and follow-up and those who were enrolled immediately upon diagnosis. The true impact of survivor bias is a high proportion of lower-risk patients among the prevalent cohort, as demonstrated by better functional class and longer 6MWD. The differences in these distributions are most likely a combined result of response to therapy in some patients and higher attrition among the highest-risk patients. Nonetheless, even though the predicted survival percentages at 12 months are very accurate, there is a striking dip in the survival curves for the highest risk groups in the first 3 months of follow-up. By comparison, the survival curve in the model development cohort was relatively flat in the first 3 months of follow-up. A limitation of the prognostic equation is that it does not incorporate disease trajectory, and, in fact, even if it did include disease trajectory, that trajectory could not be measured in patients with newly diagnosed disease. Disease may be more likely to be diagnosed in patients at a point when the disease has begun to progress more rapidly. By contrast, some of the very sickest patients in whom disease was previously diagnosed may have moved into convalescent care, effectively moving outside of the population that could be enrolled. 358 Original Research

6 Figure 3. A, One-year survival in the validation cohort stratified according to estimated probability of surviving 1 year. Predicted 1-year survival is 95% to 100% in the low-risk group, 90% to, 95% in the average-risk group, 85% to, 90% in the moderately high-risk group, 70% to, 85% in the high-risk group, and, 70% in the very high-risk group. B, One-year survival in the validation cohort stratified according to risk score. The average predicted 1-year survival is 95% to 100% (low risk) for patients with risk scores of 1 to 7. Similarly, the ranges specified for average risk, moderately high risk, high risk, and very high risk correspond to risk scores of 8, 9, 10 to 11, and 12. The first prospective registry established for patients with PAH was the Patient Registry for the Characterization of Primary Pulmonary Hypertension supported by the National Heart, Lung, and Blood Institute. 8,9 Mortality was associated with right ven tricular hemodynamic function. The predictive equation included just three variables (mean pulmonary artery pressure, mean right atrial pressure, and cardiac index) and had a c-statistic of In comparison, the predictive equation from the REVEAL Registry includes more variables that complement the RHC data, including both positive and negative predictors of survival, and has better discrimination (c-index ), and should, therefore, be useful for the prediction of survival in patients presenting with PAH. The discrimination of the predictive equation developed from the REVEAL Registry is comparable to those calculated for other well-established, widely used, and validated contemporary heart failure risk models. The c-indices for the Seattle Heart Failure Model (0.725) 19 and for the Heart Failure Survival Score (0.690) 20 are slightly lower than those calculated in the present analysis (0.726). Two previous PAH registries also developed novel risk equations. The French registry reported an exponential equation with the factors of sex, 6MWD, and cardiac output (c-index ) from a multivariable Cox proportional hazards model used to predict survival in 190 patients with idiopathic, familial, or anorexigenassociated PAH. 16 The Pulmonary Hypertension CHEST / 141 / 2 / FEBRUARY,

7 Figure 4. Model calibration plots. A, Observed 1-year Kaplan- Meier estimates vs 1-year survival predicted by the prognostic equation. The Xs correspond to the observed survival in the five risk strata based on the prognostic equation. Predicted 1-year survival is 95% to 100% in the low-risk group, 90% to, 95% in the average-risk group, 85% to, 90% in the moderately high-risk group, 70% to, 85% in the high-risk group, and, 70% in the very high-risk group. B, Xs correspond to the observed survival in the five risk strata based on the risk score. Risk score ranges are 1 to 7 in the low-risk group, 8 in the average-risk group, 9 in the moderately high-risk group, 10 to 11 in the high-risk group, and 12 in the very high-risk group. Connection registry reported an exponential survival equation, which suggested that, surprisingly, the hemodynamic variables of the former National Institutes of Health equation carried the strongest weight when all types of clinical data were analyzed together. 17 Still, neither the French registry nor the Pulmonary Hypertension Connection registry equations have been validated or tested as repeated measures. The present analysis has not only validated the predictive equation established with data from the model development cohort but also has translated the predictive equation into a simple risk score calculator that can be incorporated into the work flow of any busy clinic. Assigning integer point values of 1 or 2 does not appear to lead to an appreciable loss of predictive power compared with using the b coefficients from the original model. On the other hand, an automated version of the tool might reasonably use the multidigit coefficients in place of integers. The predictive equation has been developed to apply to both patients with newly diagnosed disease and patients with previously diagnosed disease and includes a missing data indicator. The missing data indicator allows a clinician to calculate the 1-year probability of survival for a patient even if data are unavailable for all of the predictive factors in the equation. High-risk PAH subgroups were assigned additional points rather than developing separate models for each subgroup. This enhances the simplicity of the tool and produced a more stable model than causespecific models that relied on smaller numbers of patients; however, it is conceivable that new cause-specific risk factors may be identified over time by other researchers. This calculator may be useful in everyday clinical practice, once its predictive power over time is validated. A recent single-center study independently validated the predictive risk score for 1- and 5-year mortality of 484 patients with group I pulmonary hypertension. 21 Only 16% of patients in this study by Kane et al 21 were also enrolled in the REVEAL Registry, and they were excluded from the REVEAL Registry score validation. In this study, the REVEAL Registry score mirrored both 1-year and 5-year mortality, with a c-index of The score was developed and validated in the present analysis only for a 1-year time horizon. It remains possible that the score may not work as well over a longer time horizon. Furthermore, it is unlikely that a patient s score would remain constant year after year, in which case a reassessment of risk would be warranted. The REVEAL Registry prognostic equation and simplified risk calculator could help clinicians decide when to implement therapies to relieve symptoms and potentially individualize the type and number of therapies according to the patient s level of risk, prioritize patients for lung transplantation, and facilitate discussions regarding the disease and prognosis. For example, findings from the study by Kane et al 21 suggest that those patients with PAH who were functional class II to III but had a REVEAL Registry score 8 (n 5 115) had an equivalent mortality risk as those patients presenting in functional class IV (n 5 77), and patients in class IV with a REVEAL Registry score, 8 (n 5 21) had better survival than patients with functional class III symptoms (n 5 268). 21 Incorporating the REVEAL Registry score into patient assessment would provide a useful tool to help identify those patients with disconnects between subjective PAH assessments and objective risk. Although expert opinion is a powerful predictor itself, Lee and colleagues 22 have illustrated that evidence-based models can outperform expert opinion with respect to accurate prediction of risks. Although extensive data have been collected on treatments being administered to patients enrolled 360 Original Research

8 in REVEAL Registry, 4 specific therapies were not included as candidate predictors of survival in the equation, and the score has not been evaluated as a tool to monitor treatment response. Evaluation of therapies for PAH is best done in well-powered, head-to-head randomized trials. The predictive equation has the ability to help clinical scientists in the design of prospective randomized trials of therapies for PAH. The equation can better define target populations on the basis of five risk strata and, thus, help with both targeting therapy and statistical planning. In addition, the model will continue to be refined to include the trajectory of key variables. In conclusion, the REVEAL Registry prognostic equation and simplified risk calculator, when applied to a cohort of recently enrolled patients with newly diagnosed PAH from the REVEAL Registry study, is accurate, well calibrated, and easy to use. The risk calculator has the potential to support decision making by clinicians and patients in everyday clinical practice and in future clinical research endeavors. Further research is needed, however, to prospectively assess the application of the tool in real-world clinical management. Acknowledgments Author contributions: Dr Benza: contributed to the study design; collection, analysis, and interpretation of data; and drafting and critical review of the manuscript and has seen and approved the final version. Dr Gomberg-Maitland: contributed to the study design; collection, analysis, and interpretation of data; and drafting and critical review of the Mr Miller: contributed to the study design; collection, analysis, Dr Frost: contributed to the study design; collection, analysis, and interpretation of data; and drafting and critical review of the manuscript and has seen and approved the final version. Dr Frantz: contributed to the study design; collection, analysis, Ms Foreman: contributed to the study design; collection, analysis, Dr Badesch: contributed to the study design; collection, analysis, Dr McGoon: contributed to the study design; collection, analysis, Financial/nonfinancial disclosures: The authors have reported to CHEST the following conflicts of interest: Dr Benza has received honoraria from Actelion, Gilead, and United Therapeutics and has received or is pending receipt of grants from Actelion, the American Heart Association, Bayer, the National Institutes of Health/National Heart, Lung, and Blood Institute (NIH/NHLBI), Novartis, Pfizer, and United Therapeutics. Dr Benza has received honoraria for his service on the REVEAL Registry Steering Committee, which is supported by Actelion. Dr Gomberg-Maitland has served as a consultant/participant on data safety monitoring boards/steering committees for clinical trials for Actelion, Gilead, Medtronic, and Pfizer. Actelion, Gilead, Lilly/Icos, Pfizer, Novartis, and United Therapeutics have provided funding to the University of Chicago to support Dr Gomberg-Maitland s conduct of clinical trials. Dr Gomberg-Maitland has a patent filed entitled Compositions and Methods for Treating Pulmonary Hypertension for the use of sorafenib in pulmonary hypertension, WO/2007/ Mr Miller is employed by ICON Late Phase & Outcomes Research, a company that receives research support from Actelion and other pharmaceutical companies. Dr Frost serves as a consultant for Actelion and Gilead. Dr Frost has received honoraria from Actelion, Gilead, and Pfizer and has provided expert testimony on diet pill litigation. She has also received (through Baylor) funds for IRB-approved research from Gilead, Actelion, United Therapeutics, Eli Lilly, Pfizer, Novartis, and Bayer. Dr Frost has received honoraria for her service on the REVEAL Registry Steering Committee, which is supported by Actelion. Dr Frantz has received support for traveling to REVEAL Registry meetings from Actelion. Dr Frantz s institution has received consulting fees or honorarium, grants/grants pending, and has been in consultancy relationships with Actelion, Gilead, and United Therapeutics. Ms Foreman is employed by ICON Late Phase & Outcomes Research, a company that receives research support from Actelion and other pharmaceutical companies. Dr Badesch has received honoraria for service on Steering Committees and/or Advisory Boards for Actelion, Arena, Bayer, Ikaria, Gilead, Encysive Pharmaceuticals, Pfizer, GlaxoSmithKline, Lung Rx, United Therapeutics, Eli Lilly & Co, Biogen Idec, and mondobiotech. Dr Badesch has received grants from Actelion, Gilead, Encysive Pharmaceuticals, Pfizer, United Therapeutics, Lung Rx, Eli Lilly & Co and the NIH/NHLBI. Dr Badesch has previously been deposed in appetite suppressant litigation, serving as both a treating physician and an expert witness. Dr Badesch has received honoraria for his service on the REVEAL Registry Steering Committee, which is supported by Actelion. Dr McGoon serves as a consultant with Actelion, Gilead, Lung Rx, and Medtronic. Dr McGoon has received grants from Gilead and Medtronic. Dr McGoon has received honoraria for his service on the REVEAL Registry Steering Committee, which is supported by Actelion. Role of sponsors : Preparation of this manuscript was supported by Actelion Pharmaceuticals US, Inc. The REVEAL Registry is sponsored by Actelion Pharmaceuticals US, Inc. Other contributions: We thank Blair Jarvis and Jennifer M. Kulak, PhD, of in Science Communications, a Wolters Kluwer business, for editorial support. We also thank the principal investigators and their study coordinators for their participation in REVEAL Registry. A list of these contributors can be found in e-appendix 2. Additional information: The e-appendixes can be found in the Online Supplement at / 141/2/354/suppl/DC1. References 1. Champion HC, Michelakis ED, Hassoun PM. Comprehensive invasive and noninvasive approach to the right ventriclepulmonary circulation unit: state of the art and clinical and research implications. Circulation ;120(11): Escribano Subias P, Barberà Mir JA, Suberviola V. Current diagnostic and prognostic assessment of pulmonary hypertension. Rev Esp Cardiol ;63(5): Frost AE, Badesch DB, Barst RJ, et al. The changing picture of pulmonary arterial hypertension patients in the United States: how the REVEAL registry differs from historic and non-us contemporary registries. Chest ;139(1): McGoon MD, Krichman A, Farber HW, et al. Design of the REVEAL registry for US patients with pulmonary arterial hypertension. Mayo Clin Proc ;83(8): Benza RL, Miller DP, Gomberg-Maitland M, et al. Predicting survival in pulmonary arterial hypertension: insights from the Registry to Evaluate Early and Long-Term Pulmonary Arterial Hypertension Disease Management (REVEAL). Circulation ;122(2): CHEST / 141 / 2 / FEBRUARY,

9 6. Suissa S. Immortal time bias in observational studies of drug effects. Pharmacoepidemiol Drug Saf ;16 (3 ): Pencina MJ, D Agostino RB. Overall C as a measure of discrimination in survival analysis: model specific population value and confidence interval estimation. Stat Med ; 23 ( 13 ): Rich S, Dantzker DR, Ayres SM, et al. Primary pulmonary hypertension. A national prospective study. Ann Intern Med ;107 (2 ): D Alonzo GE, Barst RJ, Ayres SM, et al. Survival in patients with primary pulmonary hypertension. Results from a national prospective registry. Ann Intern Med ; 115 ( 5 ): Rich S, Rubin L, Walker AM, Schneeweiss S, Abenhaim L. Anorexigens and pulmonary hypertension in the United States: results from the surveillance of North American pulmonary hypertension. Chest ;117 (3 ): Stricker H, Domenighetti G, Popov W, et al ; Swiss Group for Severe Pulmonary Hypertension. Severe pulmonary hypertension: data from the Swiss Registry. Swiss Med Wkly ;131 (23-24 ): Humbert M, Sitbon O, Chaouat A, et al. Pulmonary arterial hypertension in France: results from a national registry. Am J Respir Crit Care Med ;173 (9 ): Walker AM, Langleben D, Korelitz JJ, et al. Temporal trends and drug exposures in pulmonary hypertension: an American experience. Am Heart J ;152 (3 ): Jing ZC, Xu XQ, Han ZY, et al. Registry and survival study in Chinese patients with idiopathic and familial pulmonary arterial hypertension. Chest ;132 (2 ): Thenappan T, Shah SJ, Rich S, Gomberg-Maitland M. A USA-based registry for pulmonary arterial hypertension: Eur Respir J ;30 (6 ): Humbert M, Sitbon O, Chaouat A, et al. Survival in patients with idiopathic, familial, and anorexigen-associated pulmonary arterial hypertension in the modern management era. Circulation ;122 (2 ): Thenappan T, Shah SJ, Rich S, Tian L, Archer SL, Gomberg- Maitland M. Survival in pulmonary arterial hypertension: a reappraisal of the NIH risk stratification equation. Eur Respir J ;35 (5 ): McLaughlin VV, Suissa S. Prognosis of pulmonary arterial hypertension: the power of clinical registries of rare diseases. Circulation ;122 (2 ): Levy WC, Mozaffarian D, Linker DT, et al. The Seattle Heart Failure Model: prediction of survival in heart failure. Circulation ;113 (11 ): Aaronson KD, Schwartz JS, Chen TM, Wong KL, Goin JE, Mancini DM. Development and prospective validation of a clinical index to predict survival in ambulatory patients referred for cardiac transplant evaluation. Circulation ;95 (12 ): Kane GC, Maradit-Kremers H, Slusser JP, Scott CG, Frantz RP, McGoon MD. Integration of clinical and hemodynamic parameters in the prediction of long-term survival in pulmonary arterial hypertension. Chest ;139(6): Lee KL, Pryor DB, Harrell FE Jr, et al. Predicting outcome in coronary disease. Statistical models versus expert clinicians. Am J Med ;80 (4 ): Original Research