JACC: HEART FAILURE
VOL. 3, NO. 3, 2015
ª 2015 BY THE AMERICAN COLLEGE OF CARDIOLOGY FOUNDATION
ISSN 2213-1779/$36.00
PUBLISHED BY ELSEVIER INC.
Letters TO THE EDITOR
non short mortality risk prediction, it is key to take into account whether the C-statistic was calculated for death as a binary outcome (death at 1 fixed time
Risk Prediction Tools in Patients With Heart Failure
point using logistic regression models) or time-toevent outcome using the Somers Dxy rank correlation, which incorporates information from censored data. The later seems more correct nowadays but might tend to provide lower values, as we observed in the BCN Bio-HF risk calculator (C-statistic was 0.79
We read with interest the papers from Ouwerkerk
when time to event was considered, whereas it was
et al. (1), Rahimi et al. (2), and Levy and Anand (3),
0.83 using logistic regression) (4).
and we would like to comment on mortality risk prediction models in contemporary heart failure (HF) management, with biomarkers already incorporated into routine practice. The 2013 American College of Cardiology Foundation/American Heart Association HF guidelines state that biomarkers are powerful adjuncts to current standards for acute and chronic HF, and we deeply agree with the discussion remark of Ouwerkerk et al. (1): “Developing a model using a systems biology approach, incorporating information from demographic, biomarker, genomic, proteomic, and initial responses to therapy might create a more effective model.” Moreover, Levy and Anand (3) elegantly point out in their editorial comment that HF risk models would improve if: 1) they were properly validated; 2) state-of-the-art discrimination (starting the model with age and sex), calibration, and reclassification analyses were performed to assess the incremental value of adding new variables to the model; and 3) online calculators were made to allow for easy provider use. We recently developed a calculator for HF mortal-
The introduction of novel biomarkers has to be considered in new-generation prediction models. Josep Lupón, MD, PhD Joan Vila, MSc *Antoni Bayes-Genis, MD, PhD *Department of Medicine Autonomous University of Barcelona Hospital Universitari Germans Trias i Pujol Carretera del Canyet s/n 08916 Badalona Spain E-mail: [email protected] http://dx.doi.org/10.1016/j.jchf.2014.10.010 Please note: For previous studies, ST2 assays were performed by Critical Diagnostics (San Diego, California) and high-sensitivity troponin T and N-terminal pro–B-type natriuretic peptide assays were provided by Roche Diagnostics (Barcelona, Spain), which also provided a grant for statistical development and online application of the calculator. Neither had a role in the design of the study or the collection, management, analysis, or interpretation of the data. Dr. Lupón has received lecture honoraria from Roche Diagnostics and holds shares in Critical Diagnostics. Dr. Bayes-Genis has received lecture honoraria from Roche Diagnostics and Critical Diagnostics; has received research funding from Roche; and holds shares in Critical Diagnostics. The BCN Bio-HF calculator has been registered by Drs. Lupón and Bayes-Genis. Mr. Vila has reported that he has received research funding from Roche Diagnostics.
ity risk stratification that, in addition to classical risk factors and treatment, includes N-terminal pro–Btype natriuretic peptide, high-sensitivity troponin T, and soluble ST2. The Barcelona Bio-HF risk calculator (4), derived from a contemporary-treated cohort of HF patients and externally validated in a Boston cohort (5), fulfills the key points raised by Levy and Anand (3). It is available online in 4 languages at a user-friendly website (www.bcnbiohfcalculator.cat) (also www.bcnbiohfcalculator.org). According to Levy and Anand, the addition of biomarkers doubles per unit the improvement in C-statistic obtained with the addition of individual clinical variables. Finally, about the C-statistics reported in the Rahimi et al. paper (2), when comparing C-statistics in
REFERENCES 1. Ouwerkerk W, Voors AA, Zwinderman AH. Factors influencing the predictive power of models for predicting mortality and/or heart failure hospitalization in patients with heart failure. J Am Coll Cardiol HF 2014;2: 429–36. 2. Rahimi K, Bennett D, Conrad N, et al. Risk prediction in patients with heart failure: a systematic review and analysis. J Am Coll Cardiol HF 2014;2: 440–6. 3. Levy WC, Anand IS. Heart failure risk prediction models: what have we learned? J Am Coll Cardiol HF 2014;2:437–9. 4. Lupón J, de Antonio M, Vila J, et al. Development of a novel heart failure risk tool: the Barcelona Bio-Heart Failure risk calculator (BCN Bio-HF calculator). PLoS One 2014;9:e85466. 5. Lupón J, Januzzi JL, de Antonio M, Vila J, Peñafilel J, Bayes-Genis A. Validation of the Barcelona Bio-Heart Failure Risk Calculator in a cohort from Boston. Rev Esp Cardiol (Engl Ed) 2015;68:80–1.
VOL. 3, NO. 3, 2015
ª 2015 BY THE AMERICAN COLLEGE OF CARDIOLOGY FOUNDATION
ISSN 2213-1779/$36.00
PUBLISHED BY ELSEVIER INC.
Letters TO THE EDITOR
non short mortality risk prediction, it is key to take into account whether the C-statistic was calculated for death as a binary outcome (death at 1 fixed time
Risk Prediction Tools in Patients With Heart Failure
point using logistic regression models) or time-toevent outcome using the Somers Dxy rank correlation, which incorporates information from censored data. The later seems more correct nowadays but might tend to provide lower values, as we observed in the BCN Bio-HF risk calculator (C-statistic was 0.79
We read with interest the papers from Ouwerkerk
when time to event was considered, whereas it was
et al. (1), Rahimi et al. (2), and Levy and Anand (3),
0.83 using logistic regression) (4).
and we would like to comment on mortality risk prediction models in contemporary heart failure (HF) management, with biomarkers already incorporated into routine practice. The 2013 American College of Cardiology Foundation/American Heart Association HF guidelines state that biomarkers are powerful adjuncts to current standards for acute and chronic HF, and we deeply agree with the discussion remark of Ouwerkerk et al. (1): “Developing a model using a systems biology approach, incorporating information from demographic, biomarker, genomic, proteomic, and initial responses to therapy might create a more effective model.” Moreover, Levy and Anand (3) elegantly point out in their editorial comment that HF risk models would improve if: 1) they were properly validated; 2) state-of-the-art discrimination (starting the model with age and sex), calibration, and reclassification analyses were performed to assess the incremental value of adding new variables to the model; and 3) online calculators were made to allow for easy provider use. We recently developed a calculator for HF mortal-
The introduction of novel biomarkers has to be considered in new-generation prediction models. Josep Lupón, MD, PhD Joan Vila, MSc *Antoni Bayes-Genis, MD, PhD *Department of Medicine Autonomous University of Barcelona Hospital Universitari Germans Trias i Pujol Carretera del Canyet s/n 08916 Badalona Spain E-mail: [email protected] http://dx.doi.org/10.1016/j.jchf.2014.10.010 Please note: For previous studies, ST2 assays were performed by Critical Diagnostics (San Diego, California) and high-sensitivity troponin T and N-terminal pro–B-type natriuretic peptide assays were provided by Roche Diagnostics (Barcelona, Spain), which also provided a grant for statistical development and online application of the calculator. Neither had a role in the design of the study or the collection, management, analysis, or interpretation of the data. Dr. Lupón has received lecture honoraria from Roche Diagnostics and holds shares in Critical Diagnostics. Dr. Bayes-Genis has received lecture honoraria from Roche Diagnostics and Critical Diagnostics; has received research funding from Roche; and holds shares in Critical Diagnostics. The BCN Bio-HF calculator has been registered by Drs. Lupón and Bayes-Genis. Mr. Vila has reported that he has received research funding from Roche Diagnostics.
ity risk stratification that, in addition to classical risk factors and treatment, includes N-terminal pro–Btype natriuretic peptide, high-sensitivity troponin T, and soluble ST2. The Barcelona Bio-HF risk calculator (4), derived from a contemporary-treated cohort of HF patients and externally validated in a Boston cohort (5), fulfills the key points raised by Levy and Anand (3). It is available online in 4 languages at a user-friendly website (www.bcnbiohfcalculator.cat) (also www.bcnbiohfcalculator.org). According to Levy and Anand, the addition of biomarkers doubles per unit the improvement in C-statistic obtained with the addition of individual clinical variables. Finally, about the C-statistics reported in the Rahimi et al. paper (2), when comparing C-statistics in
REFERENCES 1. Ouwerkerk W, Voors AA, Zwinderman AH. Factors influencing the predictive power of models for predicting mortality and/or heart failure hospitalization in patients with heart failure. J Am Coll Cardiol HF 2014;2: 429–36. 2. Rahimi K, Bennett D, Conrad N, et al. Risk prediction in patients with heart failure: a systematic review and analysis. J Am Coll Cardiol HF 2014;2: 440–6. 3. Levy WC, Anand IS. Heart failure risk prediction models: what have we learned? J Am Coll Cardiol HF 2014;2:437–9. 4. Lupón J, de Antonio M, Vila J, et al. Development of a novel heart failure risk tool: the Barcelona Bio-Heart Failure risk calculator (BCN Bio-HF calculator). PLoS One 2014;9:e85466. 5. Lupón J, Januzzi JL, de Antonio M, Vila J, Peñafilel J, Bayes-Genis A. Validation of the Barcelona Bio-Heart Failure Risk Calculator in a cohort from Boston. Rev Esp Cardiol (Engl Ed) 2015;68:80–1.
