Selecting patients for heart transplantation: Comparison of the Heart Failure Survival Score (HFSS) and the Seattle Heart Failure Model (SHFM)

Transcription

1 Selecting patients for heart transplantation: Comparison of the Heart Failure Survival Score (HFSS) and the Seattle Heart Failure Model (SHFM) Ayumi Goda, MD, PhD, a,b Paula Williams, BS, a Donna Mancini, MD, a and Lars H. Lund, MD, PhD c From the a Division of Cardiology, College of Physicians and Surgeons, Columbia University, New York, New York, USA; b Cardiology Department, Kyorin University, Tokyo, Japan; and c Department of Cardiology, Section for Heart Failure, Karolinska University Hospital, Stockholm, Sweden. KEYWORDS: chronic heart failure; heart transplantation; prognostic model; Heart Failure Survival Score; Seattle Heart Failure Model BACKGROUND: The Heart Failure Survival Score (HFSS) risk-stratifies patients with chronic heart failure (CHF) referred for heart transplantation using 7 parameters, including peak VO 2. The Seattle Heart Failure Model (SHFM) is a 2-variable model that combines clinical, laboratory and therapeutic data. Although both models have excellent accuracy, only the HFSS was derived and validated in patients referred for transplantation, and the HFSS and SHFM have not been directly compared. METHODS: We tested the accuracy of the SHFM and compared the HFSS and SHFM in 715 patients referred for heart transplantation. RESULTS: Over a follow-up of days, 354 patients died or received an urgent heart transplantation or a ventricular assist device. One-year event-free survival was 89%, 72% and 6%, respectively, for the low-, medium- and high-risk HFSS strata, and 93%, 76%, and 58%, respectively, for the low-, medium- and high-risk SHFM strata. The HFSS and SHFM were modestly correlated (R.48, p.1). In receiver operating characteristic curve analysis, areas under the curves (AUCs) for the HFSS and SHFM were comparable (1 year:.72 vs.73; 2-year:.7 vs.74, respectively) and incremental to New York Heart Association class. The 1- and 2-year combined HFSS SHFM AUCs were.77 and.76, respectively, significantly better than the HFSS or SHFM alone. CONCLUSIONS: The HFSS and SHFM provide accurate and comparable risk stratification in CHF patients referred for transplantation. Combining the HFSS and SHFM improves predictive ability. J Heart Lung Transplant 211;3: International Society for Heart and Lung Transplantation. All rights reserved. Reprint requests: Lars H. Lund, MD, PhD, Department of Cardiology, Section for Heart Failure, Karolinska University Hospital, N35, Stockholm , Sweden. Telephone: Fax: address: lars.lund@alumni.duke.edu /$ -see front matter 211 International Society for Heart and Lung Transplantation. All rights reserved. doi:1.116/j.healun Chronic heart failure (CHF) is associated with high mortality, but risk may be difficult to assess, ranging from 5% to 75% mortality per year. 1 Therefore, assessing mortality risk becomes a critical component in the evaluation of a candidate for heart transplantation, 2 especially under the current circumstances of severe donor organ shortage. New York Heart Association (NYHA) class correlates with prognosis, but it is subjective. Peak oxygen consumption (VO 2 ) is used in transplant selection but has limitations when used alone. 3 Therefore, we developed the Heart Failure Survival Score (HFSS), which effectively risk-stratifies patients under evaluation for heart transplantation using 7 parameters, including peak VO 2. 4 The HFSS has been validated and found to be more accurate than peak VO 2 alone in numerous settings. 5 1 The Seattle Heart Failure Model (SHFM) was derived from the PRAISE I clinical trial database 1 and has been validated in numerous settings However, 98% of events in the SHFM derivation and validation databases were death, rather than transplantation or left ventricular assist device (LVAD) implantation. 11,15 The SHFM provides risk strata, an estimation of 1-, 2- and 5-year survival rates, a mean life expectancy and an estimated survival curve, using 2 commonly obtained clinical, pharmacologic, device and laboratory parameters, but

2 Goda et al. Comparison of HFSS and SHFM 1237 with NYHA class rather than peak VO 2 as a measure of functional capacity. 1 Although both models have been broadly validated and have excellent accuracy, they were derived and validated in very different populations. The aim of this study, specifically in patients referred for heart transplantation, was to: (1) assess the prognostic accuracy of the SHFM; and (2) compare the HFSS and SHFM. Outcomes Outcome events were defined as death, urgent transplantation (United Network of Organ Sharing [UNOS] Status 1) or LVAD implantation. Patients who were transplanted as non-urgent (UNOS Status 2) were censored alive on the date of the transplant. Vital status of patients lost to clinical follow-up was assessed using the Social Security Death Index. Statistics Methods Study patients and data collection Seven hundred fifteen consecutive patients with systolic heart failure referred to the Columbia University Medical Center for heart transplant evaluation underwent cardiopulmonary exercise testing and collection of variables in the HFSS and SHFM. Clinical characteristics are listed in Table 1. Review of the data was approved by the local human investigations committee. The HFSS includes 7 parameters: resting heart rate (HR); mean blood pressure (mbp); left ventricular ejection fraction (LVEF); serum sodium; presence or absence of ischemic heart disease; presence or absence of intraventricular conduction defect (IVCD); and peak VO 2. Peak VO 2 was determined during maximal treadmill exercise using a modified Naughton protocol and a metabolic cart (Medical Graphics, Minneapolis, MN). LVEF was determined using echocardiography or contrast/radionuclide ventriculography. The presence of IVCD was defined as QRS interval of 12 milliseconds due to left or right bundle branch block, non-specific intraventricular conduction delay or ventricular-paced rhythm. Dichotomous variables were coded as: 1 present and absent. The HFSS was derived in each patient from the 7 clinical parameters. Each variable for the continuous and dichotomous variables was multiplied by a model coefficient, derived from a proportional hazard model. The 7 products were summed and the absolute value determined according to the following equation: HFSS [(.216 * resting HR) (.255 * mbp) (.464 * LVEF) (.47 * serum sodium) (.546 * peak VO 2 ) (.68 * presence or absence of IVCD) (.6931 * presence or absence of ischemic heart disease)]. 4 For the HFSS, risk strata were defined as a low risk ( 8.1), medium risk (7.2 to 8.9) or high risk ( 7.19), using previously described cut-offs. 4 The SHFM score was derived in each patient from 2 variables, including clinical characteristics (age, gender, NYHA class, weight, LVEF, systolic blood pressure [sbp], ischemic etiology), medications (angiotensin-converting enzyme inhibitor, angiotensin receptor blocker, -blocker, statin, aldosterone blocker, loop diuretic equivalent dose, allopurinol), device therapy (implantable cardioverter-defibrillator, cardiac resynchronization therapy) and laboratory data (lymphocyte percentage and serum sodium, hemoglobin, uric acid, total cholesterol), as previously described. 1 We used the electronic medical record to collect data on all variables required to calculate the SHFM score. Missing continuous variables were imputed as the mean for all patients in the data set. The SHFM score was rounded to the nearest integer between and 4 (patients with scores were considered to have a score of ). Risk strata were defined as low risk (score ), medium risk (score 1) or high risk (score 2). Baseline characteristics for patients with and without events were compared by chi-square tests (categorical variables) and unpaired t-tests (continuous variables) (Table 1). Pearson s correlations were calculated between the HFSS and SHFM (Figure 1A) and also between the peak VO 2 alone and the SHFM (Figure 1B) (but not between the peak VO 2 and the HFSS, because the peak VO 2 is a heavily weighted component of the HFSS). Event-free survival rates for the different HFSS and SHFM risk strata were determined using the Kaplan Meier method and compared by log-rank test (Figure 2). Components of the HFSS and SHFM were entered into Cox regressions as single variables (univariate) (Table 2) or in combination (multivariate) (Table 3). The 1- and 2-year AUC receiver operator characteristic curve (AUC of ROC) was calculated for the HFSS and SHFM separately and in combination (Figure 3), and also for NYHA in isolation. Statistical significance between AUC values was tested by the method of Hanley and McNeil. 16 To evaluate the predictive ability of a combined HFSS SHFM, a new score was created. Both variables were entered into a Cox regression model (as continuous variables). Both variables were multiplied by its associated -coefficient and the products were summed to determine a patient s risk score. All analyses, except for the comparison between AUCs, were performed using SPSS, version 11. (SPSS, Inc., Chicago, IL). Statistical comparisons were considered significant at p.5. Results Baseline characteristics and outcomes The clinical characteristics and outcomes are listed in Table 1. The mean HFSS was and the mean SHFM score was The HFSS and SHFM were modestly correlated (R.48, p.1; Figure 1A), but more so than the peak VO 2 alone vs SHFM (R.36, p.1; Figure 1B). During a mean follow-up of days, 354 outcome events (49.5%) occurred; 17 patients underwent urgent heart transplantation, 148 patients died, 36 received LVAD implantation, 35 patients underwent elective transplant, and the remaining 326 patients were alive without transplant at last follow-up. Table 1 shows a comparison of the clinical characteristics between patients with and without events. The Kaplan Meier event-free survival curves stratified by low-, medium- and high-risk HFSS and SHFM strata are

3 1238 The Journal of Heart and Lung Transplantation, Vol 3, No 11, November 211 Table 1 Baseline Clinical Characteristics Characteristic All patients (n 715) Patients without events (n 361) Patients with events (n 354) p SHFM components Clinical Age (years) Males/females (n) 464/ / / NYHA class BMI a Resting sbp (mm Hg) Medications ACE inhibitors 564 (78.9%) 276 (76.5%) 288 (81.4%).337 -blockers 54 (7.5%) 281 (77.2%) 223 (63.5%).1 Aldosterone blockers 231 (32.3%) 111 (3.7%) 12 (33.9%).521 Statin 218 (3.5%) 134 (37.1%) 84 (23.7%).1 Allopurinol 43 (6.%) 2 (5.5%) 23 (6.5%).753 Angiotensin receptor blockers 14 (2.%) 14 (3.9%) (%).1 Loop diuretic equivalent (mg) Laboratory data b Hemoglobin (g/dl), n Lymphocytes percentage, n Total cholesterol (mg/dl), n Uric acid (mg/dl), n Devices.34 ICD 244 (34.1%) 11 (3.5%) 134 (37.8%) CRT 3 (4.2%) 12 (3.3%) 18 (5.1%) CRT-D 18 (15.1%) 65 (18.%) 43 (12.1%) HFSS components Rest HR (bpm) Rest mbp (mm Hg) Peak VO 2 (ml/min/kg) Presence of IVCD 41 (56.1%) 182 (5.4%) 219 (61.9%).3 Common to SHFM and HFSS LVEF (%) Ischemic/non-ischemic etiology (n) 284/ /227 15/ Sodium (meq/liter) Scores HFSS SHFM Outcomes Follow-up, days Death 148 (2.7%) LVAD 36 (4.5%) UNOS Status 1 transplant 17 (23.9%) UNOS Status 2 transplant 35 (4.9%) Alive 326 (46.%) Data presented as mean SD for continuous variables or n and (%) for categorical variables. ACE, angiotensin-converting enzyme; BMI, body mass index; HFSS, Heart Failure Survival Score; HR, heart rate; IVCD, intraventricular conduction defect; LVAD, left ventricular assist device; LVEF, left ventricular ejection fraction; mbp, mean blood pressure; NYHA, New York Heart Association; sbp, systolic blood pressure; SHFM, Seattle Heart Failure Model; VO 2, oxygen uptake. a The SHFM utilizes weight but BMI is presented here. b Missing continuous data for the SHFM data were imputed as the mean for all patients. One or more continuous variable was missing in 272 patients. shown in Figure 2A and B. Event-free survival differed markedly by all strata (all overall and pairwise p.1). One-year event-free survival was 89%, 72% and 6%, respectively, for the low-risk ( 8.1), medium-risk (8.9 to 7.2) and high-risk ( 7.19) HFSS strata, and 93%, 76% and 58%, respectively, for the low-risk (), medium-risk (1) and high-risk ( 2) SHFM strata. Predictors of events Table 2 presents univariate predictors of events. The SHFM (hazards ratio [HR] 1.89, 95% confidence interval [CI] 1.7 to 2.12, p.1) and the HFSS (HR.52, 95% CI.46 to.59, p.1) were highly predictive, as were numerous individual components, including NYHA class.

4 Goda et al. Comparison of HFSS and SHFM 1239 A SHFM HFSS survival rates of low high risk (HFSS 8.1 and SHFM ) and high low risk (HFSS 8.1 and SHFM ) were comparable (81% and 85%, p.556). One-year event-free survival of high high risk (HFSS 8.1 and SHFM ) was 63%, considerably worse than all other groups (all: p.5). In the multivariate analysis that included HFSS and SHFM (Table 3, Part A), both were significantly and independently predictive. The new risk score, combining HFSS SHFM, was determined according to the following equation: combined HFSS SHFM [(.427 * HFSS) (.461 * SHFM)]. The AUCs of combined HFSS SHFM at 1 and 2 years were.77 and.76, respectively (Figure 3), and improved significantly compared with HFSS or SHFM alone (all: p.5). B 4 3 Discussion SHFM Table 3 presents multivariate analysis data. Both the HFSS and SHFM were strong and significant predictors of events independent of one another (p.1). Among components of the SHFM and HFSS, NYHA class, resting sbp, loop diuretic dose equivalent, lymphocytes percentage, presence of IVCD and hemoglobin remained significant independent predictors of events. In ROC curve analysis, AUC results for 1- and 2-year event-free survival for HFSS and SHFM are shown in Figure 3. AUCs for the HFSS and SHFM at 1-year (.72 vs.73, p.26) and 2-year (.7 vs.74, p.54) follow-up were comparable. AUCs for NYHA were.68 and.69, respectively, at 1 and 2 years. Combined HFSS and SHFM 2 Peak VO 2 Table 4 shows the number of the patients in the low- vs medium high-risk HFSS and SHFM strata. Kaplan Meier event-free survival curves stratified by this 4-group combined HFSS and SHFM are shown in Figure 4. One-year event-free survival of low low risk (HFSS 8.1 and SHFM ) was 96%, which was considerably higher than any other groups (all: p.1). One-year event-free 3 4 (ml/min/kg) Figure 1 Pearson s correlations between: (A) HFSS and SHFM (R.48, p.1); and (B) peak VO 2 and SHFM (R.36, p.1). We have presented two novel findings in a CHF population referred for heart transplantation: (1) the SHFM and HFSS are similarly strong; and (2) combining the SHFM and A Event-free Survival B Event-free Survival Over all: p<.1 Low vs Medium: p<.1 Medium vs High: p=.6 High vs Low: p<.1 HFSS stratification 1-year event-free survival 89% 72% 6% Time (Day) SHFM stratification 1-year event-free survival 93% 76% 58% Over all: p<.1 Low vs Medium: p<.1 Medium vs High: p<.1 High vs Low: p<.1 Low (N=313) Medium (N=259) High (N=125) SHFM (N=265) SHFM 1 (N=276) SHFM 2 (N=164) Time (Day) Figure 2 Kaplan Meier curves of survival free from urgent transplant or LVAD implantation, stratified by previously defined low-, medium- and high-risk strata of (A) the HFSS and (B) the SHFM.

5 124 The Journal of Heart and Lung Transplantation, Vol 3, No 11, November 211 Table 2 Univariate Predictors of Events The HFSS Variables HFSS improves predictive power. We have also shown that the SHFM is valid for transplant selection, which was previously not established. 11,17 Heart transplant selection Hazard ratio 95% CI Chi-square p SHFM a HFSS a NYHA a Rest sbp a Loop diuretic dose a equivalent (mg) Peak VO a Sodium a LVEF a Lymphocytes a percentage Presence of IVCD a Uric acid a Hemoglobin a Total cholesterol a Statin a Rest HR a Spironolactone a BMI a -blockers a Age Male gender Ischemic heart disease ACE inhibitors Angiotensin receptor blockers Allopurinol CI, confidence interval. See Table 1 for abbreviations and units of measure. Variables presented in descending order of significance. a Statistically significant. Heart transplantation is an effective treatment option for patients with advanced CHF. An increasing number of ambulatory patients are placed on transplant waiting lists while the supply of donor organs remains limited and is increasingly allotted to urgent transplantation. Therefore, accurate identification of patients most likely to benefit from transplantation is imperative, and risk stratification becomes a critical component of the transplant candidate selection process. 2 Peak VO 2 is a powerful prognostic predictor of survival in CHF patients. 3 However, peak VO 2 may be influenced by several confounding factors such as age, gender, motivation, anemia, body weight and muscle deconditioning, and should not be used as a sole criterion for listing. 2 Therefore, multi-marker scores for risk stratification have been derived and validated. The HFSS provides both better discrimination (large and significant differences in event-free survival between different risk strata) and calibration (appropriate cut-offs for transplant listing) than the peak VO 2 alone, in patients receiving -blockers, 5 for serial risk stratification, 6 in different genders, 7 ages, 9 ethnic origins, 8 and in the modern era of resynchronization and defibrillator therapy. 1 However, the HFSS is limited in that it requires cardiopulmonary exercise testing with measurement of peak VO 2. Some patients cannot perform the exercise test because it is cumbersome and requires specialized equipment. Interpretation of the test is clouded by confounders such as effort and pulmonary dysfunction, and thus requires careful interpretation, including assessment of the respiratory exchange ratio and ensuring that the anaerobic threshold and maximum effort have indeed been achieved. The SHFM The SHFM includes 2 readily available clinical variables and uses NYHA class as a surrogate for peak VO 2 and functional capacity. Because it was developed and validated in outpatient participants with CHF from four clinical trials and two observational registries, 1,13 the utility of the SHFM specifically in a population referred for heart transplantation Table 3 Multivariate Predictors of Events: (A) SHFM and HFSS and (B) Components of SHFM and HFSS Hazard ratio 95% CI Chi-square p (A) Variables SHFM a HFSS a (B) Variables NYHA a Rest sbp a Loop diuretic a dose equivalent (mg) Peak VO Sodium LVEF Lymphocytes percentage a Presence of IVCD a Uric acid Hemoglobin a Total cholesterol Statin Rest HR Spironolactone BMI blockers CI, confidence interval. See Table 1 for abbreviations and units of measure. a Statistically significant.

6 Goda et al. Comparison of HFSS and SHFM 1241 is less well established. Kalogeropoulos et al 11 and Gorodeski et al 17 tested the SHFM in patients with advanced heart failure referred for transplantation, but found that it underestimated risk. In a recent study, the SHFM was useful for evaluation of patients with advanced CHF who were being considered for LVAD implantation. 12 Table 4 Strata Number of Patients in SHFM HFSS Combined Risk SHFM : low risk SHFM 1 to 2: medium to high risk Total HFSS 8.1: low risk HFSS and : medium to high risk Total The SHFM was rounded to nearest integer (as indicated by ). Medium- to high-risk strata were combined for purposes of creating combined strata (here and in Figure 4). Total does not add to 715. The HFSS was not calculated in 18 patients because of missing data. The SHFM was not calculated in 1 patients because of missing categorical data. Missing continuous variables for the SHFM (laboratory data) were imputed as the mean for all patients. In contrast to Kalogeropoulos et al and Gorodeski et al, our findings suggest that the SHFM also appropriately assesses risk in patients referred for heart transplantation, and performs well in both discriminating different risk strata and in calibration of risks, facilitating selection for transplantation. Comparing HFSS and SHFM Our findings also demonstrate that the accuracy of both 1- and 2-year risk prediction was comparable between the HFSS and SHFM. In previous reports, the HFSS had a 1-year ROC for death/lvad/transplantation of.73 to.8 vs.68 to.81 for the SHFM. 1,13,18,19 The AUCs determined from our population are similar to those found previously. Because the HFSS does not require more rare parameters, such as lymphocytes percentage, it may be more available in patients specifically evaluated for transplantation; however, because the SHFM does not require peak VO 2, often available only in specialized referral cen- HFSS+SHFM stratification Figure 3 AUC of the ROC for the HFSS, SHFM, combined HFSS SHFM and clinical control (NYHA class) for: (A) 1-year event-free survival (HFSS vs SHFM: p.26; HFSS vs HFSS SHFM: p.3; SHFM vs HFSS SHFM: p.1; HFSS vs NYHA: p.71; SHFM vs NYHA, p.2; HFSS SHFM vs NYHA: p.1); and (B) 2-year event-free survival (HFSS vs SHFM: p.54; HFSS vs HFSS SHFM: p.3; SHFM vs HFSS SHFM: p.1; HFSS vs NYHA: p.263; SHFM vs NYHA: p.2; HFSS SHFM vs NYHA: p.1). Event-free Survival 1-year event-free survival 96% HFSS:Low-SHFM:Low, n=171 85% 81% 63% Over all: p<.1 Low-Low vs Others: all p<.1 Low-High vs High-Low: p=.556 High-High vs Others: all p<.5 36 HFSS:High-SHFM:Low, n=87 HFSS:Low-SHFM:High, n=142 HFSS:High-SHFM:High, n= Time (Day) 18 Figure 4 Kaplan Meier event-free survival curves stratified by combined HFSS SHFM groups.

7 1242 The Journal of Heart and Lung Transplantation, Vol 3, No 11, November 211 ters, the SHFM may be more easily obtained and useful as an initial screening for selecting patients who should be referred for complete transplant evaluation. The SHFM may also be useful if peak VO 2 cannot be obtained, due to the inability to exercise, or if peak VO 2 is unreliable, due to low respiratory exchange ratio. NYHA class is a component of the SHFM, but even in isolation it was an independent predictor of prognosis. The HFSS and SHFM were incremental to NYHA class, which nevertheless continues to maintain a robust value in clinical discrimination. This suggests that the multiple factors that go into a clinician s assigning an NYHA class are also important. There are some reports that addition of other variables, such as natriuretic peptides and renal function, may improve the predictive ability of peak VO 2, HFSS and SHFM, 13,18,2 23 but the utility of these risk markers for transplant selection remains to be determined. Combining HFSS and SHFM The HFSS and SHFM were strong predictors independent of one another. Not surprisingly, they were complementary. The AUC for a combination of both was higher than either alone. One-year event-free survival in patients with low-risk HFSS ( 8.1) and low-risk SHFM ( ) was 96% in our cohort, higher than that in low-risk HFSS (89%) and lowrisk SHFM (93%) separately. Therefore, low-risk scores on both provide additional reassurance that deferring transplantation is safe. Furthermore, the addition of one score to the other may be especially useful in medium-risk patients, for whom transplant listing decisions are the most difficult. Limitations Our study has limitations. It was a retrospective analysis of clinical and cardiopulmonary exercise data collected at a single center. Patients unable to exercise because of respiratory disorders, arrhythmias, angina, musculoskeletal disease, neurologic disorders or frailty were excluded. The HFSS and the SHFM are derived in stable patients, but most transplants are now urgent. However, most patients who deteriorate and are transplanted urgently or receive an LVAD have at some point been evaluated and listed for transplantation, and we propose that these models are useful in this selection process. Furthermore, non-urgent (UNOS II) transplantation is still common outside the USA. Predictors of outcome after transplantation would also be of interest, but we analyzed only the prognosis in heart failure per se, as a tool for determining which patients were most in need of transplantation. Finally, determining heart transplantation and LVAD candidacy requires consideration of numerous factors in addition to these scores. In conclusion, among patients referred for heart transplant evaluation, the prognostic accuracy of the HFSS and SHFM is strong and comparable. In addition, combining the HFSS and SHFM improves predictive ability. We propose that both should be integral parts of the transplant evaluation process. Disclosure statement This work was supported by grants from the Stockholms Läns Landsting and the Swedish Heart Lung Foundation, Stockholm, Sweden (to L.H.L.); the Division of Research Resources, General Clinical Research Centers Program, National Institutes of Health (5 MO1 RR645), Bethesda, MD; the Foundation for Cardiac Therapies (FACT Fund), New York, NY; and the Altman Fund, New York, NY (to D.M.). None of the authors have any conflicts of interest to disclose. References 1. Levy WC, Mozaffarian D, Linker DT, et al. The Seattle Heart Failure Model: prediction of survival in heart failure. Circulation 26;113: Mehra MR, Kobashigawa J, Starling R, et al. Listing criteria for heart transplantation: International Society for Heart and Lung Transplantation guidelines for the care of cardiac transplant candidates 26. J Heart Lung Transplant 26;25: Mancini DM, Eisen H, Kussmaul W, et al. Value of peak exercise oxygen consumption for optimal timing of cardiac transplantation in ambulatory patients with heart failure. Circulation 1991;83: Aaronson KD, Schwartz JS, Chen TM, et al. Development and prospective validation of a clinical index to predict survival in ambulatory patients referred for cardiac transplant evaluation. Circulation 1997; 95: Lund LH, Aaronson KD, Mancini DM. Predicting survival in ambulatory patients with severe heart failure on beta-blocker therapy. Am J Cardiol 23;92: Lund LH, Aaronson KD, Mancini DM. Validation of peak exercise oxygen consumption and the Heart Failure Survival Score for serial risk stratification in advanced heart failure. Am J Cardiol 25;95: Green P, Lund LH, Mancini D. Comparison of peak exercise oxygen consumption and the Heart Failure Survival Score for predicting prognosis in women versus men. Am J Cardiol 27;99: Goda A, Lund LH, Mancini DM. Comparison across races of peak oxygen consumption and heart failure survival score for selection for cardiac transplantation. Am J Cardiol 21;15: Parikh MN, Lund LH, Goda A, et al. Usefulness of peak exercise oxygen consumption and the heart failure survival score to predict survival in patients 65 years of age with heart failure. Am J Cardiol 29;13: Goda A, Lund LH, Mancini D. The Heart Failure Survival Score outperforms the peak oxygen consumption for heart transplantation selection in the era of device therapy. J Heart Lung Transplant 211; 3: Kalogeropoulos AP, Georgiopoulou VV, Giamouzis G, et al. Utility of the Seattle Heart Failure Model in patients with advanced heart failure. J Am Coll Cardiol 29;53: Levy WC, Mozaffarian D, Linker DT, et al. Can the Seattle Heart Failure Model be used to risk-stratify heart failure patients for potential left ventricular assist device therapy? J Heart Lung Transplant 29; 28: May HT, Horne BD, Levy WC, et al. Validation of the Seattle Heart Failure Model in a community-based heart failure population and enhancement by adding B-type natriuretic peptide. Am J Cardiol 27;1: Mozaffarian D, Anker SD, Anand I, et al. Prediction of mode of death in heart failure: the Seattle Heart Failure Model. Circulation 27;116:392-8.

8 Goda et al. Comparison of HFSS and SHFM Allen LA, Yager JE, Funk MJ, et al. Discordance between patientpredicted and model-predicted life expectancy among ambulatory patients with heart failure. JAMA 28;299: Hanley JA, McNeil BJ. A method of comparing the areas under receiver operating characteristic curves derived from the same cases. Radiology 1983;148: Gorodeski EZ, Chu EC, Chow CH, et al. Application of the Seattle Heart Failure Model in ambulatory patients presented to an advanced heart failure therapeutics committee. Circ Heart Fail 21;3: Zugck C, Kruger C, Kell R, et al. Risk stratification in middle-aged patients with congestive heart failure: prospective comparison of the Heart Failure Survival Score (HFSS) and a simplified two-variable model. Eur J Heart Fail 21;3: Koelling TM, Joseph S, Aaronson KD. Heart failure survival score continues to predict clinical outcomes in patients with heart failure receiving beta-blockers. J Heart Lung Transplant 24;23: Pfister R, Diedrichs H, Schiedermair A, et al. Prognostic impact of NT-proBNP and renal function in comparison to contemporary multimarker risk scores in heart failure patients. Eur J Heart Fail 28;1: Giamouzis G, Kalogeropoulos AP, Georgiopoulou VV, et al. Incremental value of renal function in risk prediction with the Seattle Heart Failure Model. Am Heart J 29;157: Kallistratos MS, Dritsas A, Laoutaris ID, et al. Incremental value of N-terminal pro-brain natriuretic peptide over left ventricle ejection fraction and aerobic capacity for estimating prognosis in heart failure patients. J Heart Lung Transplant 28;27: Rothenburger M, Wichter T, Schmid C, et al. Aminoterminal pro type B natriuretic peptide as a predictive and prognostic marker in patients with chronic heart failure. J Heart Lung Transplant 24; 23: