N-terminal fraction of pro-b-type natriuretic peptide versus clinical risk scores for prognostic stratification in chronic systolic heart failure

Transcription

1 Full research paper N-terminal fraction of pro-b-type natriuretic peptide versus clinical risk scores for prognostic stratification in chronic systolic heart failure European Journal of Preventive Cardiology 2018, Vol. 25(8) ! The European Society of Cardiology 2018 Reprints and permissions: sagepub.co.uk/journalspermissions.nav DOI: / journals.sagepub.com/home/ejpc Chiara Arzilli 1, Alberto Aimo 2, Giuseppe Vergaro 1,2, Andrea Ripoli 1, Michele Senni 3, Michele Emdin 1,2 and Claudio Passino 1,2 Abstract Background: The Seattle heart failure model or the cardiac and comorbid conditions () scores may help define patient risk in heart failure. Direct comparisons between them or versus N-terminal fraction of pro-b-type natriuretic peptide () have never been performed. Methods: Data from consecutive patients with stable systolic heart failure and data were examined. A subgroup of patients had the Seattle heart failure model data available. The endpoints were one year all-cause or cardiovascular death. Results: The population included 2023 patients, aged years, 75% were men. At the one year time-point, 198 deaths were recorded (10%), 124 of them (63%) from cardiovascular causes. While areas under the curve were not significantly different, displayed better reclassification capability than the score for the prediction of one year all-cause and cardiovascular death. Adding to the score resulted in a significant improvement in risk prediction. Among patients with Seattle heart failure model data available (n ¼ 798), the area under the curve values for all-cause and cardiovascular death were similar for the Seattle heart failure model score (0.790 and 0.820), (0.783 and 0.803), and the score (0.770 and 0.800). The combination of the score and displayed a similar prognostic performance to the Seattle heart failure model score for both endpoints. Adding to the Seattle heart failure model score performed better than the Seattle heart failure model alone in terms of reclassification, but not discrimination. Conclusions: Among systolic heart failure patients, levels had better reclassification capability for all-cause and cardiovascular death than the score. The inclusion of to the and Seattle heart failure model score resulted in significantly better risk stratification. Keywords Chronic heart failure, natriuretic peptides, risk stratification, clinical scores Received 16 January 2018; accepted 2 March 2018 Introduction Heart failure (HF) is a frequent health problem with high morbidity and mortality, increasing prevalence, and escalating healthcare costs. Predicting the clinical course of individual patients is difficult, but accurate estimation of prognosis is important for many reasons. Patients are concerned about their probability of future events. Physicians may use prognosis estimates to decide the appropriate type and timing of additional tests or therapies, including heart transplantation and 1 Division of Cardiology and Cardiovascular Medicine, Fondazione Toscana Gabriele Monasterio, Italy 2 Institute of Life Sciences, Scuola Superiore Sant Anna, Italy 3 Division of Cardiology, Ospedali Riuniti ASST Papa Giovanni XXIII, Italy Corresponding author: Claudio Passino, Division of Cardiology and Cardiovascular Medicine, Fondazione Toscana Gabriele Monasterio, Via G. Moruzzi Pisa, Italy. passino@ftgm.it

2 890 European Journal of Preventive Cardiology 25(8) mechanical circulatory support. Accurate prognostic assessment may prevent delays in the appropriate treatment of high-risk patients or overtreatment of low-risk patients. Finally, reliable risk prediction also facilitates research, for example in the design of randomised trials and the assessment of subgroup effects. 1,2 In the past three decades, investigators have developed many models to predict adverse outcomes in patients with HF. A 2013 systematic review on this topic identified 20 event-free survival prediction models in ambulatory patients with HF. 1 Only a few scores were validated in more than two independent cohorts, mostly displaying modest to poor discrimination. 1 By far the most extensively characterised score was the Seattle heart failure model (SHFM), which was validated in 14 cohorts (16,057 patients), demonstrating limited discriminative ability despite the large number of variables considered: age, gender, ischaemic aetiology, New York Heart Association (NYHA), ejection fraction, systolic blood pressure, therapy with mineralocorticoid receptor antagonists, statin use, allopurinol use, haemoglobin, lymphocyte count, uric acid, sodium, cholesterol and diuretic dose/kg. 3 Other simpler models have been proposed in recent years, such as the cardiac and comorbid conditions () score, considering the following variables: NYHA class III IV, left ventricular ejection fraction (LVEF) less than 20%, no beta-blocker, no renin angiotensin system inhibitor, severe valve heart disease, atrial fibrillation, diabetes with small and/or large vessel disease, renal dysfunction, anaemia, hypertension and more advanced age. 4 The score was validated for the prediction of one year mortality, performed well in both the original population and in a validation cohort, and could be easily calculated in everyday clinical practice. 4 Nonetheless, as with many other scores, it has not been considered after its introduction in Natriuretic peptides are established predictors of outcome in chronic HF, 5,6 and their assessment is now widespread. Nevertheless, they were not included in either the SHFM nor the scores. In the present study we sought to examine for the first time whether the addition of N-terminal fraction of pro-b-type natriuretic peptide () to these scores would result in more accurate risk stratification, considering both all-cause and cardiovascular mortality. Methods Study population Data from 2368 consecutive patients enrolled in a prospective registry of patients with chronic systolic HF in a tertiary referral hospital in Pisa, Italy, from 1999 to 2016 were retrospectively examined. The inclusion criteria were: HF duration 3 months or greater and clinical stability; diagnosis of HF with reduced ejection fraction (HFrEF; LVEF < 40%) or HF with mid-range ejection fraction (HFmrEF; LVEF 40 49%), in agreement with latest European Society of Cardiology Guidelines; 5 follow-up data available; assay and all variables needed to calculate scores (see above) available. One-hundred and twelve patients were excluded due to HF duration less than 3 months, 27 patients due to lack of complete follow-up data, and 206 due to missing variables to compute at least the score. The final study population included 2023 patients. All patients underwent complete clinical assessment, 12-lead electrocardiogram, transthoracic echocardiography and biohumoral characterisation within 1 week. Blood samples were collected at 8 a.m. after an overnight fasting period and a 20-minute supine rest, as previously described. 3 Renal function was estimated from plasma creatinine based on the chronic kidney disease epidemiology collaboration formula. 7 was measured with the ECLIA monoclonal method using the Cobas e411 platform (Roche Diagnostics Italia, Monza, Italy). 8 The variables included in the SHFM and scores have been listed in the Introduction. The study protocol was performed according to the Declaration of Helsinki and was approved by the local ethics committee, and all subjects provided informed consent at time of inclusion in the prospective registry. Follow-up and study endpoints All patients received guideline-recommended therapy for HF. The follow-up period continued until study termination (November 2016). Independent interviewers obtained information from electronic health records or from patients, cardiologists or general practitioners in charge of the patient. The primary endpoint was all-cause death, and the secondary endpoint was cardiovascular death. The risk estimation of the SHFM and the scores was evaluated at one year (i.e. the time-point for which both scores were validated) as previously described. 4,9 Patients with follow-up duration shorter than one year were censored. Statistical analysis Statistical analysis was performed using IBM SPSS Statistics (version 22, 2013) and R ( version 3.2.3, 2015). Normal distribution was assessed through the Kolmogorov Smirnov test; variables with normal distribution were presented as mean standard deviation, and those with non-normal distribution as median and interquartile range.

3 Arzilli et al. 891 Mean differences among groups were evaluated through the unpaired two-sided Student s t test. Discrete variables were compared by the 2 test with Yates correction. The strength of correlation among, SHFM and was quantified through Spearman s r. In the entire analysis, levels were lntransformed to account for skewed distribution, whereas transformation was not possible for the and SHFM scores because of the possibility of negative values. The prognostic accuracy of SHFM, and was evaluated by calculating the area under the curve (AUC) values of receiver operating characteristics curves for each variable; optimal cut-offs were established by Youden s J statistic. C-statistics summarised the prognostic discrimination, consisting in the ability to distinguish two outcomes correctly; the AUC values were compared through the De Long s test. The Table 1. Population characteristics. All patients (n ¼ 2023) SHFM subgroup (n ¼ 798, 39%) Age (years) Gender (n, % men) 1518 (75) 612 (77) BMI (kg/m 2 ) 26.4 ( ) 26.5 ( ) Ischaemic aetiology 954 (47) 408 (51) (n, %) LVEF (%) 35 (28 42) 37 (30 45) LVEF 40% (n, %) 712 (35) 348 (44) egfr (ml/min/1.73 m 2 ) 70.0 ( ) 70.4 ( ) Hypertension (n, %) 1150 (57) 507 (64) Diabetes (n, %) 657 (33) 286 (36) COPD (n, %) 374 (19) 165 (21) Anaemia (n, %) 213 (11) 108 (14) (ng/l) 1441 ( ) 1485 ( ) BB (n, %) 1820 (90) 745 (93) ACEi/ARB (n, %) 1733 (86) 675 (85) MRA (n, %) 1296 (64) 508 (64) CRT (n, %) 136 (7) 45 (6) ICD (n, %) 215 (11) 96 (12) (%) 4 (2 18) 4 (2 16) SHFM (%) 3 (2 8) (n, %) 198 (10) 75 (9) CV death (n, %) 124 (6) 50 (6) : cardiac and comorbid conditions; ACEi: angiotensin-converting enzyme inhibitor; ARB: angiotensin receptor blocker; BB: beta blocker; BMI: body mass index; COPD: chronic obstructive pulmonary disease; CRT: cardiac resynchronisation therapy; CV: cardiovascular; egfr: estimated glomerular filtration rate; ICD: implantable cardioverter defibrillator; LVEF: left ventricular ejection fraction; MRA: mineralocorticoid receptor antagonist; : N-terminal fraction of pro-b-type natriuretic peptide; NYHA: New York Heart Association; SHFM: Seattle heart failure model. D Agostino Nam version of the Hosmer Lemeshow calibration test was used to calculate 2 values as measure of calibration. The continuous net reclassification improvement (NRI) and the integrated discrimination improvement were calculated to assess reclassification, which consists in the ability to stratify patients across risk categories (<10%, 10 30% and >30%) correctly. P values less than 0.05 were considered significant. Results Study population Patient characteristics in the whole population (n ¼ 2023) and in the subgroup with available SHFM data (n ¼ 798, 39%) are summarised in Table 1. Patients were aged years, 75% were men, 47% had HF of ischaemic aetiology, and median LVEF was 35 (28 42). Patients with HFrEF were 1311 (65%). Compared with patients with HFmrEF, those with HFrEF were younger, were more often men, had lower body mass index, worse renal function and higher natriuretic peptide levels; furthermore, they were more often on drug and device therapy (Supplementary Table 1). All patients had data available; the variables considered in this score are reported in Supplementary Table 2. Patients included in the subgroup with SHFM data had a similar clinical profile to the whole population (Table 1 and Supplementary Table 3). At the one year time-point, 198 deaths were recorded (10%), 124 of them (63%) from cardiovascular causes. One year mortality was similar in the subgroup with available SHFM (Table 1). No patient was lost at the one year follow-up. Whole population: vs. ; vs. þ In the whole study population, plasma levels (ln-transformed) were modestly correlated with the score (r ¼ 0.589, ). While AUC values were not significantly different (Supplementary Figure 1), displayed better reclassification capability than the score for the prediction of one year all-cause death, and (at least for NRI) for one year cardiovascular death (Table 2). Adding to the score resulted in a significant improvement in risk prediction, as demonstrated by metrics of both discrimination and reclassification (Figure 1 and Table 2). SHFM subgroup: SHFM vs. vs. Among patients with SHFM data available (n ¼ 798), a fair correlation between the two scores was found

4 892 European Journal of Preventive Cardiology 25(8) Table 2. Whole population: combination of the score and N-terminal fraction of pro-b-type natriuretic peptide (). Index þ Discrimination AUC ( ) ( ) P ¼ ( ) P ¼ Calibration Hosmer Lemeshow , P ¼ , P ¼ , P ¼ Reclassification IDI Reference (95% CI ) P ¼ NRI Reference (95% CI ) P ¼ Discrimination AUC ( ) ( ) P ¼ (95% CI ) (95% CI ) ( ) P ¼ Calibration Hosmer Lemeshow , P ¼ , P ¼ , P ¼ Reclassification IDI Reference (95% CI ) P ¼ NRI Reference (95% CI ) P ¼ (95% CI ) (95% CI ) levels were ln-transformed. : cardiac and comorbid conditions; AUC: area under the curve; IDI: integrated discrimination improvement; NRI: net reclassification improvement. Whole population AUC AUC Sensitivity AUC AUC AUC AUC Specificity Figure 1. Improvement in discrimination following the addition of N-terminal fraction of pro-b-type natriuretic peptide (NTproBNP) to the cardiac and comorbid conditions () score. Area under the curve (AUC) values are reported, and the whole study population is considered (n ¼ 2023). The endpoints are all-cause death (left), and cardiovascular death (right). levels are ln-transformed. (r ¼ 0.533, ), whereas the correlation between either score and was weaker (SHFM vs. : r ¼ 0.347, ; vs. NTproBNP: r ¼ 0.359, ). The AUC values for all-cause death were similar for the SHFM score (0.790), (0.783, and the score (0.770), all comparisons being not significant (Figure 2). Similar findings were observed for

5 Arzilli et al. 893 SHFM subgroup Sensitivity SHFM AUC AUC AUC SHFM AUC AUC AUC Specificity Figure 2. N-terminal fraction of pro-b-type natriuretic peptide (), the cardiac and comorbid conditions () and Seattle heart failure model (SHFM) scores. Area under the curve (AUC) values are compared in the subgroup of patients with available SHFM data. The endpoints are all-cause death (left), and cardiovascular death (right). levels are ln-transformed. cardiovascular death, AUC values being for the SHFM score, for and for the score (Figure 2). SHFM subgroup: SHFM vs. þ ; SHFM vs. SHFM þ The combination of the score and performed better than the score alone in terms of reclassification, but not discrimination (Supplementary Table 4 and Supplementary Figure 2). The same combination displayed a similar prognostic performance to the SHFM score for both endpoints (Table 3). Finally, adding to the SHFM score improved reclassification over SHFM alone (Table 3). Discussion In a cohort of patients with stable systolic HF, levels performed better than the score for the prediction of all-cause and cardiovascular death, in terms of patient risk reclassification (which is more relevant than discrimination in prognostic studies). 10 Furthermore, adding to the caused a significant improvement in all metrics of risk prediction. In the subgroup with SHFM data available, the SHFM and scores and displayed similar AUC values, in line with previous studies. 4,9,11 The combination of the score and had a similar prognostic performance to the SHFM score, while adding to the SHFM improved reclassification over the SHFM score alone. Many prognostic scores have been developed in order to assist the clinician in the challenging task of risk stratification. Overall, these scores have demonstrated mostly modest to poor predictive capacity, possibly because of the difficulty in summarising the complexity of patient characterisation effectively. 12 The most promising scoring system is currently the SHFM, which includes a large number of variables. 12 Calculating the SHFM score is a cumbersome, time-consuming process, especially when computerised data repositories are not available. Furthermore, missing data are frequent in the real-world setting of outpatient care, especially when considering variables such as the percentage of lymphocytes or uric acid, possibly affecting the prognostic performance of the SHFM score. 13 In an attempt to overcome these limitations, simpler scores have been proposed, such as the. 4 In the present study only 34% of patients enrolled in our prospective registry (n ¼ 2368) had all variables available for SHFM calculation (compared to 85% for the ), confirming the poor applicability of SHFM in the outpatient setting. Interestingly, both scores do not include B-type natriuretic peptides, or have been validated against them. Indeed, only one study has investigated the addition of B-type natriuretic peptide to the SHFM score, demonstrating a small increase in AUC values for a one year composite endpoint. 13

6 894 European Journal of Preventive Cardiology 25(8) Table 3. SHFM subgroup: the scores and N-terminal fraction of pro-b-type natriuretic peptide () for risk prediction. Index SHFM þ SHFM þ Discrimination AUC ( ) ( ) P ¼ Calibration Hosmer Lemeshow , P ¼ , P ¼ Reclassification IDI Reference (95% CI ) P ¼ NRI Reference (95% CI ) P ¼ Discrimination AUC ( ) ( ) P ¼ Calibration Hosmer Lemeshow , P ¼ , P ¼ Reclassification IDI Reference (95% CI ) P ¼ NRI Reference (95% CI ) P ¼ ( ) P ¼ , P ¼ (95% CI ) P ¼ (95% CI ) P ¼ ( ) P ¼ , P ¼ (95% CI ) P ¼ (95% CI ) P ¼ ( ) P ¼ , P ¼ (95% CI ) P ¼ (95% CI ) ( ) P ¼ , P ¼ (95% CI ) P ¼ (95% CI ) levels were ln-transformed. AUC: area under the curve; IDI: integrated discrimination improvement; NRI: net reclassification improvement; SHFM: Seattle heart failure model. Circulating levels of B-type natriuretic peptide or reflect the severity of HF disease. 14 These peptides are strong prognostic indicators for HF patients in all stages of disease and are better prognostic predictors than many traditional markers, such as LVEF or estimated glomerular filtration rate. 14,15 Sampling for either B-type natriuretic peptides or for risk stratification has been recommended by HF guidelines, 5,6 and is now widespread in current clinical practice. Moreover, a systematic review has proposed that improves risk stratification when added to a generic base prognostic model. 16 Our findings provide a further confirmation of the strong prognostic value of plasma. Indeed, in the whole population plasma displays similar discrimination and better reclassification capability for all causes of death than the score, regardless of patient characteristics. The combination of and the score significantly improved risk prediction over the score alone, and improved metrics of reclassification over the SHFM score when added to that score. Notably, natriuretic peptide levels should always be considered together with the patient s history and clinical status, which the scores try to recapitulate. Therefore, when a thorough patient characterisation is performed, the overall prognostic accuracy of natriuretic peptides may approach that observed in the present study resulting from the combination of and a prognostic score. On the other hand, the finding of broadly similar prognostic performances of and risk scores suggests that the latter can usefully replace natriuretic peptide assessment whenever this last is not available. Finally, the performance of natriuretic peptides added to other patient variables for risk stratification deserves consideration also in disease settings other than HF, as suggested by several studies The main limitation of this preliminary, hypothesisgenerating study is the presence of two population subgroups. As all variables needed for SHFM calculation were available only for patients with more recent assessment, the SHFM score could be computed only in 39% of patients. As our institution is a tertiary referral centre for HF, this point further confirms the limited applicability of the SHFM in a real-world setting. In addition, further studies should verify whether our conclusions can be extended beyond the less than 50% LVEF range, i.e. to patients with HF and preserved ejection fraction. In conclusion, among patients with chronic systolic HF, levels had better reclassification

7 Arzilli et al. 895 capability for all-cause and cardiovascular death than the score. The inclusion of to the score improved metrics of both discrimination and reclassification. In a subgroup with SHFM data, the SHFM and scores and yielded similar AUC values. Adding to the SHFM improved reclassification over the SHFM score alone. Author contribution CA, GV, CP, MS and ME contributed to the conception, design, and interpretation of data for the work. CA, AA and AR contributed to the acquisition and analysis of data for the work. CA, GV, AA and AR drafted the manuscript. MS, ME and CP critically revised the manuscript. All authors gave final approval and agree to be accountable for all aspects of the work ensuring integrity and accuracy. Declaration of conflicting interests The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. Funding The author(s) received no financial support for the research, authorship, and/or publication of this article. References 1. Alba AC, Agoritsas T, Jankowski M, et al. Risk prediction models for mortality in ambulatory patients with heart failure: a systematic review. Circ Heart Fail 2013; 6: Allen LA, Yager JE, Funk MJ, et al. Discordance between patient-predicted and model-predicted life expectancy among ambulatory patients with heart failure. JAMA 2008; 299: Emdin M, Passino C, Prontera C, et al. Cardiac natriuretic hormones, neurohormones, thyroid hormones and cytokines in normal subjects and patients with heart failure. Clin Chem Lab Med 2004; 42: Senni M, Parrella P, De Maria R, et al. Predicting heart failure outcome from cardiac and comorbid conditions: the score. Int J Cardiol 2013; 163: Ponikowski P, Voors AA, Anker SD, et al. Authors/Task Force Members ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure: the Task Force for the Diagnosis and Treatment of Acute and Chronic Heart Failure of the European Society of Cardiology (ESC): developed with the special contribution of the Heart Failure Association (HFA) of the ESC. Eur Heart J 2016; 37: Yancy CW, Jessup M, Bozkurt B, et al. American College of Cardiology Foundation; American Heart Association Task Force on Practice Guidelines ACCF/AHA Guideline for the management of heart failure: a report of the American College of Cardiology Foundation/ American Heart Association Task Force on Practice Guidelines. J Am Coll Cardiol 2013; 62: e147 e Levey AS and Stevens LA. Estimating GFR using the CKD Epidemiology Collaboration (CKD-EPI) creatinine equation: more accurate GFR estimates, lower CKD prevalence estimates, and better risk predictions. Am J Kidney Dis 2010; 55: Prontera C, Zucchelli GC, Vittorini S, et al. Comparison between analytical performances of polyclonal and monoclonal electrochemiluminescence immunoassays for. Clin Chim Acta 2009; 400: Levy WC, Mozaffarian D, Linker DT, et al. The Seattle Heart Failure Model: prediction of survival in heart failure. Circulation 2006; 113: Pencina MJ, D Agostino RB and Vasan RS. Statistical methods for assessment of added usefulness of new biomarkers. Clin Chem Lab Med 2010; 48: Oremus M, Don-Wauchope A, McKelvie R, et al. BNP and as prognostic markers in persons with chronic stable heart failure. Heart Fail Rev 2014; 19: Ferrero P, Iacovoni A, D Elia E, et al. Prognostic scores in heart failure critical appraisal and practical use. Int J Cardiol 2015; 188: 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 2007; 100: Doust JA, Pietrzak E, Dobson A, et al. How well does B-type natriuretic peptide predict death and cardiac events in patients with heart failure: systematic review. BMJ 2005; 330: Rahimi K, Bennett D, Conrad N, et al. Risk prediction in patients with heart failure: a systematic review and analysis. J Am Coll Cardiol 2014; 2: Don-Wauchope AC, Santaguida PL, Oremus M, et al. Incremental predictive value of natriuretic peptides for prognosis in the chronic stable heart failure population: a systematic review. Heart Fail Rev 2014; 19: Wei XB, Liu YH, He PC, et al. Prognostic value of N-terminal prohormone brain natriuretic peptide for in-hospital and long-term outcomes in patients with infective endocarditis. Eur J Prev Cardiol 2017; 24: Mjelva ØR, Po nitz V, Bru gger-andersen T, et al. Long-term prognostic utility of pentraxin 3 and D-dimer as compared to high-sensitivity C-reactive protein and B-type natriuretic peptide in suspected acute coronary syndrome. Eur J Prev Cardiol 2016; 23: Courand PY, Harbaoui B, Be` cle C, et al. Plasma mirrors the deleterious cardiovascular and renal continuum in hypertension. Eur J Prev Cardiol 2017; 24: