LETTERS TO THE EDITOR
European Journal of Cardio-Thoracic Surgery 48 (2015) 176–178
N2 involvement in lung cancer: the Danaïdes’ barrel Marc Riqueta,b,*, Alex Aramea,b and Elizabeth Fabrea,b a Thoracic Surgery Department, Georges Pompidou European Hospital, Paris, France b Oncology Department, Georges Pompidou European Hospital, Paris, France
Association for the Study of Lung Cancer descriptive classification in zones. Interact CardioVasc Thorac Surg 2010;11:260–4. * Corresponding author. Thoracic Surgery, Georges Pompidou European Hospital, 20 rue Leblanc, 75015 Paris, France. Tel: +33-1-56093450; fax: +33-156093380; e-mail: [email protected] (M. Riquet). doi:10.1093/ejcts/ezu369 Advance Access publication 4 October 2014
Received 21 July 2014; accepted 18 August 2014 Keywords: Lung cancer • N2 In the article ‘Time to refine N2 staging’ [1], Matsunaga et al. investigated the relationship between the prognosis and the location of the primary tumour and nodes (LN) involved. They suggested a classification of these patients to either cN2α, which consisted of an upper lobe tumour with clinical positive ipsilateral upper mediastinal LNs (station: 2R and 4R on the right and 2L, 4L, 5 and 6 on the left), or cN2β, which was a lower lobe tumour with clinical positive ipsilateral mediastinal LNs (station: 7, 8 and 9 on both sides) [1]. They found that survival was better in the cN2α group regardless of the number of LN stations involved. The following points need to be considered: (i) cN2 diagnosis based on tomodensitometry (TDM) and positron emission tomography (PET), available in only 10 patients, lacks sensitivity and specificity, and comparing cN2 in this manner without considering pN2 may result in reflecting a Will Rogers ‘phenomenon’. (ii) Upper or lower mediastinal LN involvement related to tumour location was considered an important prognostic factor, which is not demonstrated on a pathological basis [2]. (iii) Single- and multiple-station involvements were mixed as visible in Figure 1 [1]: two stations could be located in the same anatomical LN chain in cN2α, and in two different LN chains in cN2β. Such a mixture might shift the results in favour of a non-difference between single- and multiple-station involvements. Refining N2 staging in this manner is only adding one more paper to many already published papers during the last 30-year period (at least 1850 available in PubMed), which sought a solution to understanding the N2-involvement effects on prognosis, classification and grouping (stations, zones and lobespecific drainage) and management (sampling, radical or extended lymphadenectomy or nothing). Lymphatic drainage of tumours follows an anatomical pathway. Therefore, returning to anatomical landmarks may be crucial to a better understanding of lung cancer lymphatic spread [3]. This could avoid the perpetual filling of an N2 Danaïdes’ barrel and permit the escape of a vicious cycle impeding any consistent progress.
REFERENCES [1] Matsunaga T, Suzuki K, Takamochi K, Oh S. Time to refine N2 staging? cN2α and cN2β based on local regional involvement provide a more accurate prognosis in surgically treated IIIA non-small-cell lung cancer than N2 alone or the number of node stations involved. Eur J Cardiothorac Surg 2014;46:86–91. [2] Riquet M, Rivera C, Pricopi C, Arame A, Mordant P, Foucault C et al. Is the lymphatic drainage of lung cancer lobe-specific? A surgical appraisal. Eur J Cardiothorac Surg 2015;47:543–9. [3] Riquet M, Arame A, Foucault C, Le Pimpec Barthes F. Prognostic classifications of lymph node involvement in lung cancer and current International
Reply to Riquet et al. Takeshi Matsunaga, Kenji Suzuki*, Kazuya Takamochi and Shiaki Oh Department of General Thoracic Surgery, Juntendo University School of Medicine, Tokyo, Japan Received 16 August 2014; accepted 18 August 2014 Keywords: Lung cancer • Mediastinal lymph node • Stage IIIA • N2 We would like to thank Riquet et al. for their interest in our paper [1, 2]. Firstly, our classification is based on the location of the primary tumour and nodes involved, regardless of whether the mediastinal lymph node metastasis is single station or multiple station. We defined cN2α as the involvement of an upper mediastinal lymph node in a main tumour located in the upper lobe or as that of a lower mediastinal lymph node in a main tumour located in the lower lobe. cN2β was defined as other situations. For example a main tumour located on the lower right lobe that involved Station #7 was classified as cN2α, and a main tumour located on the upper left lobe that involved Station #7 was classified as cN2β [2]. Sensitivity and specificity of diagnosis for cN2 using computed tomography (CT) only, as in this study, were decreased compared with those using positron emission tomography-computed tomography (PET-CT) as Riquet et al. pointed out. Although, in these 4 years, we have always staged patients with lung cancer using not only CT but also PET–CT, we have considered the definition of mediastinal metastasis as unclear, causing it to differ among institutions or radiologists. There were two reasons why our discussion was mainly on clinical factors but not pathological factors in our paper even though the Will Rogers’ phenomenon might occur when cN2 is compared without considering pN2. The first is about N2 involvement in lung cancer: the Danaïdes’ barrel was clinical N2 and not resected pathological N2. Adjuvant chemotherapy was recommended for lung cancer patients with resected pathological N2. On the other hand, treatment for clinical N2 lung cancer is controversial although, in general, radical chemoradiotherapy is recommended for cN2 Stage IIIA lung cancer patients. Secondly, there is report on prognosis based on the location of the primary tumour and the pathologically involved node. As such, we considered clinical status to be important in deciding the treatment strategy, and therefore investigated mainly the clinical status. Our classification of mediastinal lymph node was based on the concept of the anatomical pathway of lymphatic drainage in lung cancer. We therefore suggested that prognosis of patients with N2 non-small-cell lung cancer was poor as the distance between the primary tumour and the lymph node involved was farther in the same mediastinal lymph node. Moreover, in our paper, the 5-year disease-free survival rate for cN2α was significantly better than that for cN2β (29.6 vs 0%, P < 0.001). We assumed that our result was supported by Riquet et al. [3]. In their report, the patterns of pN2 distribution in the right and left lobectomies showed that the incidence of upper mediastinal involved lymph nodes was much higher than that of lower mediastinal involved lymph nodes in upper lobectomy. Moreover, the incidence of lower mediastinal involved lymph nodes was much higher than that of upper mediastinal involved lymph nodes in lower lobectomy. We believe that the anatomical pathway of lymphatic drainage could provide a clue as we resolve N2 involvement in lung cancer: the Danaïdes’ barrel.
© The Author 2014. Published by Oxford University Press on behalf of the European Association for Cardio-Thoracic Surgery. All rights reserved.
Letters to the Editor / European Journal of Cardio-Thoracic Surgery
[1] Riquet M, Arame A, Fabre E. N2 involvement in lung cancer: the Danaïdes’ barrel. Eur J Cardiothorac Surg 2015;48:176. [2] Matsunaga T, Suzuki K, Takamochi K, Oh S. Time to refine N2 staging? cN2α and cN2β based on local regional involvement provide a more accurate prognosis in surgically treated IIIA non-small-cell lung cancer than N2 alone or the number of node stations involved. Eur J Cardiothorac Surg 2014;46:86–91. [3] Riquet M, Rivera C, Pricopi C, Arame A, Mordant P, Foucault C et al. Is the lymphatic drainage of lung cancer lobe-specific? A surgical appraisal. Eur J Cardiothorac Surg 2015;47:543–9. * Corresponding author. Department of General Thoracic Surgery, Juntendo University School of Medicine, 1-3 Hongo 3-chome, Bunkyo-ku, Tokyo 1138431, Japan. Tel: +81-3-38133111; fax: +81-3-58000281; e-mail: kjsuzuki@ juntendo.ac.jp (K. Suzuki). doi:10.1093/ejcts/ezu374 Advance Access publication 4 October 2014
External validation of the EuroSCORE II risk stratification model in the USA Daniel Hernández-Vaqueroa,*, Rocío Díaza, Blanca Meanaa and César Morísb a Department of Cardiac Surgery, Heart Area, Central University Hospital of Asturias, Oviedo, Spain b Department of Cardiology, Heart Area, Central University Hospital of Asturias, Oviedo, Spain Keywords: Risk prediction • Decision-making • Coronary artery bypass grafting • Aortic valve replacement • Transcatheter aortic valve implantation The accuracy of any predictive model depends largely on the homogeneity between the population used to create the system and the study population. For this reason, the article by Osnabrugge et al. [1] analysing the performance of EuroSCORE II in a large US database is of great relevance. Furthermore, these authors investigated the implications of this performance for the selection of patients as candidates for transcatheter aortic valve implantation (TAVI). The external validation of any predictive system is usually measured in terms of calibration and discrimination [2]. Calibration is the agreement between model predictions and observed event incidences. For instance, if a model predicts a 10% mortality risk (expected mortality), the observed mortality incidence should be 10 of 100 patients for a perfect calibration. For patients who underwent aortic valve replacement (AVR), these authors found that the calibration of EuroSCORE II was almost perfect with an observed/expected ratio of 0.96, indicating that, overall, the predicted mortality was very close to reality. Interestingly, this model under-predicted in low-risk patients and over-predicted in high-risk patients (Fig. 3; [1]). Discrimination, which is a measure of how well the model can separate those who will develop the event from those who will not, is of most interest when classification into groups is the goal, for example low, intermediate or high risk for surgical treatment [2]. Therefore, discrimination is crucial when defining which patients are candidates for TAVI. The area under the receiver operator curve (AUC) is the most popular measure of discrimination and indicates the probability that a randomly selected patient who will develop the event has a higher risk score than a randomly selected patient who will not develop the event. Subsequently, an AUC of 0.5 indicates no predictive ability, whereas an AUC of 1.0 represents perfect discrimination. Although Osnabrugge et al. [1] showed a good calibration for patients who underwent AVR, discrimination measured by the AUC was only 0.71. This means that, 29% of the time (1–0.71), this system gives a lower predicted risk for a patient who will develop the event than for a patient who will not [2]. Under this circumstance, we believe that the ability of the EuroSCORE II to identify patients at low or high risk is only moderate and, therefore, decisionmaking should not be based on this model in the USA. Moreover, the discrimination of the Society of Thoracic Surgeons-Predicted Risk of Mortality (STS-PROM) is only slightly superior [1], and so this should be taken into
account when analysing the results of some clinical trials that have used this model to select patients for transcatheter procedures [3].
REFERENCES [1] Osnabrugge RL, Speir AM, Head SJ, Fonner CE, Fonner E, Kappetein AP et al. Performance of EuroSCORE II in a large US database: implications for transcatheter aortic valve implantation. Eur J Cardiothorac Surg 2014;46: 400–8. [2] Cook NR. Use and misuse of the receiver operating characteristic curve in risk prediction. Circulation 2007;115:928–35. [3] Adams DH, Popma JJ, Reardon MJ, Yakubov SJ, Coselli JS, Deeb GM et al. Transcatheter aortic-valve replacement with a self-expanding prosthesis. N Engl J Med 2014;370:1790–8. * Corresponding author. Department of Cardiac Surgery, Heart Area, Central University Hospital of Asturias, Avenue of the Central University Hospital of Asturias, Oviedo, Spain. Tel: +34-985-108000; e-mail: [email protected] (D. Hernández-Vaquero). doi:10.1093/ejcts/ezu460 Advance Access publication 4 December 2014
Reply to Hernández-Vaquero et al. Ruben L. Osnabruggea, Stuart J. Heada, A. Pieter Kappeteina,* and Jeffrey B. Richb a Department of Cardio-Thoracic Surgery, Erasmus University Medical Center, Rotterdam, Netherlands b Mid-Atlantic Cardiothoracic Surgeons, Sentara Heart Hospital, Norfolk, VA, USA Keywords: Risk prediction • Decision-making • Coronary artery bypass grafting • Aortic valve replacement • Transcatheter aortic valve implantation We thank Hernández-Vaquero et al. [1] for their insightful comments on our manuscript that was recently published in the European Journal of CardioThoracic Surgery [2]. Hernández-Vaquero et al. elaborate on the concepts of calibration and discrimination with respect to our analyses of patients who underwent aortic valve replacement (AVR). Firstly, the authors of the letter discuss calibration. The overall observed : expected (O : E) ratio for the EuroSCORE II in AVR patients was good (0.96, 95% confidence interval 0.79–1.13), whereas the EuroSCORE II overpredicted in low-risk patients undergoing AVR, and underpredicted in high-risk patients (Fig. 3 and Table 3 [2]). Especially in high-risk patients, risk prediction is challenging as the lower numbers of patients make the estimates unstable (Conference discussion [2]). Indeed, these issues should be kept in mind when interpreting the data. Secondly, Hernández-Vaquero et al. questioned the discriminative abilities of the EuroSCORE II (AUC = 0.71; 95% CI, 0.67–0.75) and STS score (AUC = 0.74; 95% CI 0.69–0.78). The results are comparable with a recently published meta-analysis [3]. Nevertheless, we agree with the authors that users of the risk scores should be aware that they are not perfect. Patient selection in clinical settings and for inclusion of randomized trials should not solely be based on risk scores, but rather be the result of multidisciplinary heart team discussions [4–6].
REFERENCES [1] Hernández-Vaquero D, Díaz R, Meana B, Morís C. External validation of the EuroSCORE II risk stratification model in the USA. Eur J Cardiothorac Surg 2015;48:177. [2] Osnabrugge RL, Speir AM, Head SJ, Fonner CE, Fonner E, Kappetein AP et al. Performance of EuroSCORE II in a large US database: implications for transcatheter aortic valve implantation. Eur J Cardiothorac Surg 2014;46: 400–8; discussion 408. [3] Biancari F, Juvonen T, Onorati F, Faggian G, Heikkinen J, Airaksinen J et al. Meta-analysis on the performance of the EuroSCORE II and the Society of Thoracic Surgeons Scores in Patients Undergoing Aortic Valve Replacement. J Cardiothorac Vasc Anesth 2014;28:1533–9.
LETTERS TO THE EDITOR
REFERENCES
177
European Journal of Cardio-Thoracic Surgery 48 (2015) 176–178
N2 involvement in lung cancer: the Danaïdes’ barrel Marc Riqueta,b,*, Alex Aramea,b and Elizabeth Fabrea,b a Thoracic Surgery Department, Georges Pompidou European Hospital, Paris, France b Oncology Department, Georges Pompidou European Hospital, Paris, France
Association for the Study of Lung Cancer descriptive classification in zones. Interact CardioVasc Thorac Surg 2010;11:260–4. * Corresponding author. Thoracic Surgery, Georges Pompidou European Hospital, 20 rue Leblanc, 75015 Paris, France. Tel: +33-1-56093450; fax: +33-156093380; e-mail: [email protected] (M. Riquet). doi:10.1093/ejcts/ezu369 Advance Access publication 4 October 2014
Received 21 July 2014; accepted 18 August 2014 Keywords: Lung cancer • N2 In the article ‘Time to refine N2 staging’ [1], Matsunaga et al. investigated the relationship between the prognosis and the location of the primary tumour and nodes (LN) involved. They suggested a classification of these patients to either cN2α, which consisted of an upper lobe tumour with clinical positive ipsilateral upper mediastinal LNs (station: 2R and 4R on the right and 2L, 4L, 5 and 6 on the left), or cN2β, which was a lower lobe tumour with clinical positive ipsilateral mediastinal LNs (station: 7, 8 and 9 on both sides) [1]. They found that survival was better in the cN2α group regardless of the number of LN stations involved. The following points need to be considered: (i) cN2 diagnosis based on tomodensitometry (TDM) and positron emission tomography (PET), available in only 10 patients, lacks sensitivity and specificity, and comparing cN2 in this manner without considering pN2 may result in reflecting a Will Rogers ‘phenomenon’. (ii) Upper or lower mediastinal LN involvement related to tumour location was considered an important prognostic factor, which is not demonstrated on a pathological basis [2]. (iii) Single- and multiple-station involvements were mixed as visible in Figure 1 [1]: two stations could be located in the same anatomical LN chain in cN2α, and in two different LN chains in cN2β. Such a mixture might shift the results in favour of a non-difference between single- and multiple-station involvements. Refining N2 staging in this manner is only adding one more paper to many already published papers during the last 30-year period (at least 1850 available in PubMed), which sought a solution to understanding the N2-involvement effects on prognosis, classification and grouping (stations, zones and lobespecific drainage) and management (sampling, radical or extended lymphadenectomy or nothing). Lymphatic drainage of tumours follows an anatomical pathway. Therefore, returning to anatomical landmarks may be crucial to a better understanding of lung cancer lymphatic spread [3]. This could avoid the perpetual filling of an N2 Danaïdes’ barrel and permit the escape of a vicious cycle impeding any consistent progress.
REFERENCES [1] Matsunaga T, Suzuki K, Takamochi K, Oh S. Time to refine N2 staging? cN2α and cN2β based on local regional involvement provide a more accurate prognosis in surgically treated IIIA non-small-cell lung cancer than N2 alone or the number of node stations involved. Eur J Cardiothorac Surg 2014;46:86–91. [2] Riquet M, Rivera C, Pricopi C, Arame A, Mordant P, Foucault C et al. Is the lymphatic drainage of lung cancer lobe-specific? A surgical appraisal. Eur J Cardiothorac Surg 2015;47:543–9. [3] Riquet M, Arame A, Foucault C, Le Pimpec Barthes F. Prognostic classifications of lymph node involvement in lung cancer and current International
Reply to Riquet et al. Takeshi Matsunaga, Kenji Suzuki*, Kazuya Takamochi and Shiaki Oh Department of General Thoracic Surgery, Juntendo University School of Medicine, Tokyo, Japan Received 16 August 2014; accepted 18 August 2014 Keywords: Lung cancer • Mediastinal lymph node • Stage IIIA • N2 We would like to thank Riquet et al. for their interest in our paper [1, 2]. Firstly, our classification is based on the location of the primary tumour and nodes involved, regardless of whether the mediastinal lymph node metastasis is single station or multiple station. We defined cN2α as the involvement of an upper mediastinal lymph node in a main tumour located in the upper lobe or as that of a lower mediastinal lymph node in a main tumour located in the lower lobe. cN2β was defined as other situations. For example a main tumour located on the lower right lobe that involved Station #7 was classified as cN2α, and a main tumour located on the upper left lobe that involved Station #7 was classified as cN2β [2]. Sensitivity and specificity of diagnosis for cN2 using computed tomography (CT) only, as in this study, were decreased compared with those using positron emission tomography-computed tomography (PET-CT) as Riquet et al. pointed out. Although, in these 4 years, we have always staged patients with lung cancer using not only CT but also PET–CT, we have considered the definition of mediastinal metastasis as unclear, causing it to differ among institutions or radiologists. There were two reasons why our discussion was mainly on clinical factors but not pathological factors in our paper even though the Will Rogers’ phenomenon might occur when cN2 is compared without considering pN2. The first is about N2 involvement in lung cancer: the Danaïdes’ barrel was clinical N2 and not resected pathological N2. Adjuvant chemotherapy was recommended for lung cancer patients with resected pathological N2. On the other hand, treatment for clinical N2 lung cancer is controversial although, in general, radical chemoradiotherapy is recommended for cN2 Stage IIIA lung cancer patients. Secondly, there is report on prognosis based on the location of the primary tumour and the pathologically involved node. As such, we considered clinical status to be important in deciding the treatment strategy, and therefore investigated mainly the clinical status. Our classification of mediastinal lymph node was based on the concept of the anatomical pathway of lymphatic drainage in lung cancer. We therefore suggested that prognosis of patients with N2 non-small-cell lung cancer was poor as the distance between the primary tumour and the lymph node involved was farther in the same mediastinal lymph node. Moreover, in our paper, the 5-year disease-free survival rate for cN2α was significantly better than that for cN2β (29.6 vs 0%, P < 0.001). We assumed that our result was supported by Riquet et al. [3]. In their report, the patterns of pN2 distribution in the right and left lobectomies showed that the incidence of upper mediastinal involved lymph nodes was much higher than that of lower mediastinal involved lymph nodes in upper lobectomy. Moreover, the incidence of lower mediastinal involved lymph nodes was much higher than that of upper mediastinal involved lymph nodes in lower lobectomy. We believe that the anatomical pathway of lymphatic drainage could provide a clue as we resolve N2 involvement in lung cancer: the Danaïdes’ barrel.
© The Author 2014. Published by Oxford University Press on behalf of the European Association for Cardio-Thoracic Surgery. All rights reserved.
Letters to the Editor / European Journal of Cardio-Thoracic Surgery
[1] Riquet M, Arame A, Fabre E. N2 involvement in lung cancer: the Danaïdes’ barrel. Eur J Cardiothorac Surg 2015;48:176. [2] Matsunaga T, Suzuki K, Takamochi K, Oh S. Time to refine N2 staging? cN2α and cN2β based on local regional involvement provide a more accurate prognosis in surgically treated IIIA non-small-cell lung cancer than N2 alone or the number of node stations involved. Eur J Cardiothorac Surg 2014;46:86–91. [3] Riquet M, Rivera C, Pricopi C, Arame A, Mordant P, Foucault C et al. Is the lymphatic drainage of lung cancer lobe-specific? A surgical appraisal. Eur J Cardiothorac Surg 2015;47:543–9. * Corresponding author. Department of General Thoracic Surgery, Juntendo University School of Medicine, 1-3 Hongo 3-chome, Bunkyo-ku, Tokyo 1138431, Japan. Tel: +81-3-38133111; fax: +81-3-58000281; e-mail: kjsuzuki@ juntendo.ac.jp (K. Suzuki). doi:10.1093/ejcts/ezu374 Advance Access publication 4 October 2014
External validation of the EuroSCORE II risk stratification model in the USA Daniel Hernández-Vaqueroa,*, Rocío Díaza, Blanca Meanaa and César Morísb a Department of Cardiac Surgery, Heart Area, Central University Hospital of Asturias, Oviedo, Spain b Department of Cardiology, Heart Area, Central University Hospital of Asturias, Oviedo, Spain Keywords: Risk prediction • Decision-making • Coronary artery bypass grafting • Aortic valve replacement • Transcatheter aortic valve implantation The accuracy of any predictive model depends largely on the homogeneity between the population used to create the system and the study population. For this reason, the article by Osnabrugge et al. [1] analysing the performance of EuroSCORE II in a large US database is of great relevance. Furthermore, these authors investigated the implications of this performance for the selection of patients as candidates for transcatheter aortic valve implantation (TAVI). The external validation of any predictive system is usually measured in terms of calibration and discrimination [2]. Calibration is the agreement between model predictions and observed event incidences. For instance, if a model predicts a 10% mortality risk (expected mortality), the observed mortality incidence should be 10 of 100 patients for a perfect calibration. For patients who underwent aortic valve replacement (AVR), these authors found that the calibration of EuroSCORE II was almost perfect with an observed/expected ratio of 0.96, indicating that, overall, the predicted mortality was very close to reality. Interestingly, this model under-predicted in low-risk patients and over-predicted in high-risk patients (Fig. 3; [1]). Discrimination, which is a measure of how well the model can separate those who will develop the event from those who will not, is of most interest when classification into groups is the goal, for example low, intermediate or high risk for surgical treatment [2]. Therefore, discrimination is crucial when defining which patients are candidates for TAVI. The area under the receiver operator curve (AUC) is the most popular measure of discrimination and indicates the probability that a randomly selected patient who will develop the event has a higher risk score than a randomly selected patient who will not develop the event. Subsequently, an AUC of 0.5 indicates no predictive ability, whereas an AUC of 1.0 represents perfect discrimination. Although Osnabrugge et al. [1] showed a good calibration for patients who underwent AVR, discrimination measured by the AUC was only 0.71. This means that, 29% of the time (1–0.71), this system gives a lower predicted risk for a patient who will develop the event than for a patient who will not [2]. Under this circumstance, we believe that the ability of the EuroSCORE II to identify patients at low or high risk is only moderate and, therefore, decisionmaking should not be based on this model in the USA. Moreover, the discrimination of the Society of Thoracic Surgeons-Predicted Risk of Mortality (STS-PROM) is only slightly superior [1], and so this should be taken into
account when analysing the results of some clinical trials that have used this model to select patients for transcatheter procedures [3].
REFERENCES [1] Osnabrugge RL, Speir AM, Head SJ, Fonner CE, Fonner E, Kappetein AP et al. Performance of EuroSCORE II in a large US database: implications for transcatheter aortic valve implantation. Eur J Cardiothorac Surg 2014;46: 400–8. [2] Cook NR. Use and misuse of the receiver operating characteristic curve in risk prediction. Circulation 2007;115:928–35. [3] Adams DH, Popma JJ, Reardon MJ, Yakubov SJ, Coselli JS, Deeb GM et al. Transcatheter aortic-valve replacement with a self-expanding prosthesis. N Engl J Med 2014;370:1790–8. * Corresponding author. Department of Cardiac Surgery, Heart Area, Central University Hospital of Asturias, Avenue of the Central University Hospital of Asturias, Oviedo, Spain. Tel: +34-985-108000; e-mail: [email protected] (D. Hernández-Vaquero). doi:10.1093/ejcts/ezu460 Advance Access publication 4 December 2014
Reply to Hernández-Vaquero et al. Ruben L. Osnabruggea, Stuart J. Heada, A. Pieter Kappeteina,* and Jeffrey B. Richb a Department of Cardio-Thoracic Surgery, Erasmus University Medical Center, Rotterdam, Netherlands b Mid-Atlantic Cardiothoracic Surgeons, Sentara Heart Hospital, Norfolk, VA, USA Keywords: Risk prediction • Decision-making • Coronary artery bypass grafting • Aortic valve replacement • Transcatheter aortic valve implantation We thank Hernández-Vaquero et al. [1] for their insightful comments on our manuscript that was recently published in the European Journal of CardioThoracic Surgery [2]. Hernández-Vaquero et al. elaborate on the concepts of calibration and discrimination with respect to our analyses of patients who underwent aortic valve replacement (AVR). Firstly, the authors of the letter discuss calibration. The overall observed : expected (O : E) ratio for the EuroSCORE II in AVR patients was good (0.96, 95% confidence interval 0.79–1.13), whereas the EuroSCORE II overpredicted in low-risk patients undergoing AVR, and underpredicted in high-risk patients (Fig. 3 and Table 3 [2]). Especially in high-risk patients, risk prediction is challenging as the lower numbers of patients make the estimates unstable (Conference discussion [2]). Indeed, these issues should be kept in mind when interpreting the data. Secondly, Hernández-Vaquero et al. questioned the discriminative abilities of the EuroSCORE II (AUC = 0.71; 95% CI, 0.67–0.75) and STS score (AUC = 0.74; 95% CI 0.69–0.78). The results are comparable with a recently published meta-analysis [3]. Nevertheless, we agree with the authors that users of the risk scores should be aware that they are not perfect. Patient selection in clinical settings and for inclusion of randomized trials should not solely be based on risk scores, but rather be the result of multidisciplinary heart team discussions [4–6].
REFERENCES [1] Hernández-Vaquero D, Díaz R, Meana B, Morís C. External validation of the EuroSCORE II risk stratification model in the USA. Eur J Cardiothorac Surg 2015;48:177. [2] Osnabrugge RL, Speir AM, Head SJ, Fonner CE, Fonner E, Kappetein AP et al. Performance of EuroSCORE II in a large US database: implications for transcatheter aortic valve implantation. Eur J Cardiothorac Surg 2014;46: 400–8; discussion 408. [3] Biancari F, Juvonen T, Onorati F, Faggian G, Heikkinen J, Airaksinen J et al. Meta-analysis on the performance of the EuroSCORE II and the Society of Thoracic Surgeons Scores in Patients Undergoing Aortic Valve Replacement. J Cardiothorac Vasc Anesth 2014;28:1533–9.
LETTERS TO THE EDITOR
REFERENCES
177
