Research article

N-terminal pro-B-type natriuretic peptide complements the GRACE risk score in predicting early and late mortality following acute coronary syndrome

Sohail Q. Khan, Hafid Narayan, Kelvin H. Ng, Onkar S. Dhillon, Dominic Kelly, Paulene Quinn, Iain B. Squire, Joan E. Davies, Leong L. Ng

Abstract

The GRACE (Global Registry of Acute Coronary Events) risk score has been shown to offer predictive power with regard to death and AMI (acute myocardial infarction) in patients with ACS (acute coronary syndromes). NT-proBNP (N-terminal pro-B-type natriuretic peptide) has also been found to be useful in predicting mortality following ACS. In the present study, we sought to investigate the use of the GRACE score and NT-proBNP levels at predicting risk of early and late deaths following ACS. We studied 1033 patients (740 men, mean age 66.5±12.7 years) with AMI. Blood was drawn once within 24 h following the onset of chest pain. The plasma concentration of NT-proBNP was determined using an in-house non-competitive immunoassay. Patients were GRACE risk scored. The 30-day mortality was 3.7% and the 6-month mortality was 7.8%, and all were related to higher GRACE risk scores (P=0.001 for trend). Higher NT-proBNP levels were also related to increased mortality (P<0.0001). In a Cox proportional hazards model, independent predictors of 30-day and 6-month mortality included NT-proBNP levels and the GRACE risk score. The receiver-operating curve for the GRACE risk score was complemented by NT-proBNP levels for prediction of 30-day mortality [AUC (area under the curve), 0.85] and 6-month mortality (AUC, 0.81). NT-proBNP gives complementary information to the GRACE risk score for predicting early and late mortality. The inclusion of the NT-proBNP blood test is useful in risk-stratifying patients after ACS.

  • acute coronary syndrome
  • Global Registry of Acute Coronary Events (GRACE) score
  • myocardial infarction
  • N-terminal pro-B-type natriuretic peptide (NT-proBNP)
  • prognosis

INTRODUCTION

Risk stratification is an important and integral part of the management of patients following ACS (acute coronary syndrome). A number of models have been developed to help the clinician in risk stratification such as the TIMI (Thrombolysis in Myocardial Infarction), PURSUIT (Platelet glycoprotein IIb/IIIa in Unstable angina: Receptor Suppression Using Integrilin Therapy) and GRACE (Global Registry of Acute Coronary Events) scores [1,2]. Of these, the GRACE scoring system is probably the most robust clinical risk stratification tool as it was based on patients seen in routine clinical practice as opposed to selected patients who were enrolled in clinical trials [3]. A GRACE score has been developed predicting in-hospital mortality and 6-month mortality and, recently, variables have been defined that predict death or AMI [acute MI (myocardial infarction)] at 6 months [46]. The three major scoring systems that have been developed have been shown to have prognostic power at predicting early adverse events. However, all three have been shown to have prognostic value at predicting adverse events out to 1 year [7]. The GRACE score had better predictive power than TIMI and PURSUIT. The GRACE score also has good predictive capacity out to 4 years [8].

In parallel with this, biomarkers, particularly BNP (B-type natriuretic peptide) and its more stable counterpart NT-proBNP (N-terminal pro-BNP) [9], have been shown to have a vigorous response following an AMI [10], and plasma levels are related to short- and long-term mortality [1115]. BNP is able to provide additional prognostic information over and above that provided by left ventricular dysfunction following an AMI [12]. NT-proBNP also has similar prognostic power and levels measured in the sub-acute phase, between days 2 and 4, have prognostic power to predict left ventricular function and 2-year survival [14]. This has been shown across the entire ACS spectrum [16]. There are some results to suggest that the natriuretic peptides may also have prognostic power at predicting future MI [13,17].

The aim of the present study was to compare the performance of the GRACE score with NT-proBNP and to investigate whether a combination of NT-proBNP with the GRACE score would give increased predictive power of short-term (30-day) and long term (6-month) mortality.

MATERIALS AND METHODS

Study population

We studied 1033 consecutive subjects admitted to the Coronary Care Unit of Leicester Royal Infirmary with ACS. The study complied with the Declaration of Helsinki, was approved by the local ethics committee and written informed consent was obtained from patients.

AMI was diagnosed if a patient had a plasma creatine kinase-MB elevation greater than twice the normal level, or a cardiac troponin I level>0.1 ng/ml with at least one of the following: symptoms of ischaemia, ECG changes indicative of ischaemia or development of Q-waves on the ECG [18]. AMI was sub-categorized into STEMI (ST-elevation MI) or NSTEMI (non-STEMI).

Exclusion criteria were known malignancy, or surgery in the previous month. Patients with ST-segment elevation of >0.1 mV in two contiguous ECG leads received thrombolytic therapy (tissue plasminogen activator or streptokinase) if they presented within a suitable time frame. Patients underwent assessment for the GRACE risk score for ACS as described previously [4,5], and were grouped into tertiles, low-, intermediate- and high-risk groups.

Plasma samples

Blood samples were drawn on one occasion within 24 h after the onset of chest pain for determination of plasma NT-proBNP. After at least 15 min of bed rest, 20 ml of blood was collected into tubes containing EDTA and aprotinin. All plasma was stored at −70 °C until assayed in a single batch.

Calculating the GRACE score

We used the GRACE score derived by Granger et al. [4] for analysis of in-hospital mortality and the GRACE score developed by Eagle et al. [5] for analysis of 6-month mortality. The variables in the two GRACE scores are shown in Table 1.

View this table:
Table 1 Details of the GRACE risk score for calculating 30-day and 6-month mortality

NT-proBNP assay

The NT-proBNP assay used in the present study was based on a non-competitive assay as previously described [16]. Sheep antibodies were raised to the N-terminal of human NT-proBNP and monoclonal mouse antibodies were raised to the C- terminal. The N-terminal IgG was affinity-purified and biotinylated. Samples or NT-proBNP standards were incubated in C-terminal IgG-coated wells with the biotinylated antibody for 24 h at 4 °C. Detection was with MAE (methyl-acridinium ester)-labelled streptavidin. The lower limit of detection was 0.3 pmol/l. There was no cross-reactivity with atrial natriuretic peptide, BNP or C-type natriuretic peptide. Inter- and intra- coefficients of variation were 2.3% and 4.8% respectively. The results from this in-house assay are highly correlated (r=0.90, P<0.0001, n=86; κ=0.80) to those obtained in the NT-proBNP assay marketed by Roche Diagnostics.

End points

We assessed the value of NT-proBNP for the prediction of early (30-day) and late (6-month) mortality. End points were identified through routine data provided to the hospital from the Office of National Statistics Registry and review of the medical notes. There was a minimum 6-month follow-up of all surviving patients. No patient was lost to follow-up.

Statistical analysis

Statistical analyses were performed using SPSS Version 14 and Stata version 10. Comparisons of continuous variables were made using the Mann–Whitney U test. The relationship of baseline variables with death was assessed using Cox proportional hazards analysis by multivariable analysis. An epidemiological approach was taken and factors thought to be important for the end point were entered in multivariable analyses. The factors entered into the model for predicting 30-day mortality were baseline patient characteristics as part of the GRACE score, gender, past history of AMI, territory of infarction, STEMI or NSTEMI, therapy with ACEIs (angiotensin-converting enzyme inhibitors), ARBs [AT1 (angiotensin II type 1) receptor blocker] and β-blockers, and the peptide marker NT-proBNP. The factors entered into the model for predicting 6-month mortality were baseline patient characteristics as part of the GRACE score, gender, territory of infarction, STEMI or NSTEMI, Killip class, therapy with ACEIs, ARBs and β-blockers, and the peptide marker NT-proBNP. Kaplan–Meier cumulative survival curves were constructed and compared by the log-rank test and the log-rank test for trend. To compare the accuracy of NT-proBNP and GRACE score, ROC (receiver-operating characteristic) curves were generated and the AUC (area under the curve) was calculated. Comparisons between ROC curves was by the method of Hanley and McNeil [19]. Reclassification tables were constructed as a further measure to assess any incremental value for NTproBNP in improving the risk classification afforded by the GRACE score [20]. Levels of NT-proBNP were normalized by log10 transformation. Thus HRs (hazard ratios) refer to a 10-fold rise in the levels of these markers. A P value below 0.05 was considered statistically significant.

RESULTS

Patient characteristics

The demographic features of the patient population are shown in Table 2. Median length of follow-up was 503 days with a range of 1–1059 days. No patient was lost to follow-up. During follow-up, 122 (11.6%) patients died. There were 151 (14.4%) re-admissions with AMI.

View this table:
Table 2 Characteristics of patients in the present study

Values are means±S.D. or numbers (%). CK, creatine kinase.

NT-proBNP levels in patients

Median NT-proBNP was 1106.6 pmol/l (range 0.3–34135 pmol/l). NT-proBNP was higher in patients who died [median (range): survivors, 782.5 (0.3–14109.1) pmol/l compared with dead, 4159.1 (9.0–34135) pmol/l; P<0.0001].

GRACE score and NT-proBNP as predictors of early mortality (30-days)

30-Day mortality was 3.7% (38 patients) and was related to higher GRACE risk scores (P<0.001 for trend; Figure 1). When clinical and demographic characteristics (as listed in the Materials and methods section) were entered into a multivariable Cox proportional hazards model, the independent predictors of death included NT-proBNP {HR, 2.24 [95% CI (confidence interval), 1.07–4.71]; P=0.034} and GRACE risk score [HR, 1.02 (95% CI, 1.01–1.03); P<0.001] (Table 3). When NT-proBNP was added to the individual components of the GRACE score, NT-proBNP displaced many of the factors in the GRACE score (Table 4).

Figure 1 Relationship between higher GRACE score and increased 30-day mortality

Patients in tertiles of low-, intermediate- and high-risk GRACE groups.

View this table:
Table 3 Cox regression analysis for 30-day mortality
View this table:
Table 4 Cox regression analysis for 30-day mortality using GRACE predictors and NT-proBNP

The Kaplan–Meier survival curve revealed a significantly better clinical outcome in patients with NT-proBNP below the median compared with those with NT-proBNP above the median (log rank 27.10, P=0.0001). Kaplan–Meier curves for quartiles of GRACE score and NT-proBNP are shown in Figures 2 and 3 respectively. The combined Kaplan–Meier survival curve related to whether the patients had a below- or above-median GRACE score and below- or above- median NT-proBNP level showed an improved risk prediction (pooled over strata, log rank 98.72, P=0.0001; Figure 4).

Figure 2 Kaplan–Meier analysis for GRACE quartiles
Figure 3 Kaplan–Meier analysis for NT-proBNP quartiles
Figure 4 Kaplan–Meier analysis for NT-proBNP levels (< or >median) predicting the end point of death in patients stratified by a GRACE score < or >median

Group 1=NT-proBNP<median, GRACE<median; Group 2=NT-proBNP>median, GRACE<median; Group 3=NT-proBNP<median, GRACE>median; Group 4=NT-proBNP>median, GRACE>median.

The ROC curve for NT-proBNP yielded an AUC of 0.79 (95% CI, 0.72–0.86; P<0.001), for the GRACE risk score the AUC was 0.84 (95% CI, 0.78–0.89; P<0.001). The combination of GRACE risk score and NT-proBNP improved risk prediction for mortality [AUC, 0.85 (95% CI, 0.85–0.90); P<0.001], which significantly exceeded that of NT-proBNP (P=0.007), but not the GRACE score (P=0.20). Patients who had an NT-proBNP level above the median (above 1100 pmol/l) and a GRACE score above 149 were at especially high-risk of death.

GRACE score and NT-proBNP as predictors of 6-month mortality

6-Month mortality was 7.8% (82 patients) and was related to higher GRACE risk scores (P<0.001 for trend). When clinical and demographic characteristics were entered into a multivariable Cox proportional hazards model the independent predictors of death included NT-proBNP [HR, 1.84 (95% CI, 1.09–3.13); P=0.023] and the GRACE risk score [HR, 1.02 (95% CI, 1.02–1.03); P<0.001] (Table 5). When NT-proBNP was added to the individual components of the GRACE score, NT-proBNP displaced many of the factors in the GRACE score (Table 6). The Kaplan–Meier survival curve revealed a grading of mortality related to whether the patients had a below- or above-median GRACE score and a below- or above-median NT-proBNP level (pooled over strata, log rank 103.0, P=0.0001). The ROC curve for NT-proBNP yielded an AUC of 0.76 (95% CI, 0.70–0.82; P<0.001), and for the GRACE risk score the AUC was 0.79 (95% CI, 0.74–0.83; P<0.001). The combination of the GRACE risk score and NT-proBNP improved risk prediction for mortality [AUC, 0.81 (95% CI, 0.86–0.86); P<0.001], which significantly exceeded that of NT-proBNP (P<0.001) and the GRACE score (P=0.035).

View this table:
Table 5 Cox regression analysis for 6-month mortality
View this table:
Table 6 Cox regression analysis for 6-month mortality using GRACE predictors and NT-proBNP

PCI, percutaneous coronary intervention.

Patients who had an NT-proBNP above 1100 pmol/l and a GRACE score above 120 were at especially high-risk of death.

Reclassification analysis for added predictive ability of NT-proBNP over the GRACE score

Net reclassification improvement [20] was used to assess the incremental prediction improvement afforded by NT-proBNP over the GRACE score for both 30-day and 6-month mortality. Individuals were categorized according to their GRACE risk scores (as low-, intermediate- and high-risk as described on the GRACE site (http://www.outcomes-umassmed.org/grace/grace_risk_table.cfm). Table 7 shows the net reclassification improvement of adding NT-proBNP to the GRACE model for both 30-day and 6-month mortality. The improvement for predicting 30-day mortality was 24.4% (P<0.0001), whereas at 6 months it was 10.4% (P<0.04).

View this table:
Table 7 Net reclassification improvement of mortality risk using the GRACE model alone and the GRACE model with NT-proBNP

DISCUSSION

The aim of the present study was to compare the prognostic utility after ACS of the GRACE risk score to that of NT-proBNP. The results of the present study show that the GRACE risk score and NT-proBNP are good predictors of adverse events in our unselected cohort of patients. We have shown that using a combination of the GRACE risk score with NT-proBNP can give further prognostic information with regard to both short-term and long-term mortality.

GRACE score and NT-proBNP: 30-day mortality

The GRACE score predicting in-hospital mortality considers eight independent factors. In the paper by Granger et al. [4] the GRACE score had an area under the ROC curve of 0.83 at predicting in-hospital death. In our present cohort of patients the area under the ROC curve for GRACE was 0.84. The similarities are probably because of our unselected patient population being similar to that from which the GRACE database is derived. We have recruited consecutive patients with no restriction or overt exclusion criteria (the mean age of our cohort of patients was 65 compared with 66 years in the GRACE study).

The additive value of NT-proBNP is over and above the factors of the GRACE score which include robust factors such as Killip class and serum cardiac markers. Indeed we have shown that when NT-proBNP is included with the standard GRACE variables it out-performs many of the variables as a strong independent predictor of 30-day mortality. Using net reclassification improvement analysis, the contribution of NT-proBNP to improving risk prediction from the GRACE score alone was also very significant.

GRACE score and NT-proBNP: 6-month mortality

The GRACE score for predicting late mortality has nine factors [5]. The 6-month mortality in our cohort was greater than that of the GRACE study (7.8% compared with 4.8% in GRACE). In the cohort used in the present study the main reperfusion strategy employed was thrombolysis which could well account for some of the differences seen in mortality. Having said this, the area under the ROC curve showed that NT-proBNP had a similar predictive power to the GRACE score, and both were similar to that found in the GRACE study (0.81). NT-proBNP improved upon the accuracy of mortality prediction of the GRACE score on both ROC AUC analysis and net reclassification improvement analysis. The mortality at 12 months was similar to that seen in the OPERA (Observatoire sur la Prise en charge hospitalière, l'Evolution à un an et les caRactéristiques de patients présentant un infArctus du myocarde avec ou sans onde Q) study, a large French registry [21]. When NT-proBNP is added to the standard GRACE variables in a multivariable Cox model it remains a strong independent predictor of mortality. The outcome of patients following an AMI is still poor [5,21]; for this reason risk stratification is important as it can select patients who are at risk of adverse events including mortality. Also those at greatest risk are more likely to derive benefit from pharmacological or reperfusion strategies [22]. The present study is the first study comparing GRACE scoring with NT-proBNP, both of which have strong prognostic potential [46,14,15,23,24]. We have shown that there is equivalence of the GRACE risk score and measurement of NT-proBNP levels. The GRACE risk score is the most accurate risk score available for predicting outcomes after ACS. However it is a complicated scoring system. As NT-proBNP is equivalent in terms of prognosis it may in the future be possible to replace this with the GRACE score.

The two combined however give additional prognostic information, as shown by the Kaplan–Meier curve. A raised NT-proBNP above the median and a high GRACE score identify a particularly high-risk cohort of patients following ACS.

Other studies have investigated the natriuretic peptides and compared them with GRACE and even the TIMI score. When the GRACE score was compared with BNP the two were found to provide complementary information [25]. BNP also provides incremental prognostic information over and above the TIMI score in STEMI patients treated with primary angioplasty [26,27]. Similarly NT-proBNP is able to provide prognostic information in NSTEMI patients over and above the TIMI score [28]. The present study however has a larger population than the other studies that have investigated natriuretic peptides and clinical scores.

Prognostic information may help guide treatment in the future; a study which utilized NT-proBNP for risk stratification showed that patients with raised levels of the peptide derived more benefit from an invasive treatment strategy than a conservative treatment strategy [29]. With the availability of rapid point-of-care assays for NT-proBNP, this marker could easily be part of the armamentarium of the clinician and could readily be combined with the GRACE score to identify individuals at high risk of adverse events.

Limitations and strengths

A major strength of the data set in the present study is that it has an approximately equal weighting of STEMI and NSTEMI patients. We have to acknowledge that we have not included patients with unstable angina, this is due to the admission policies in place. We also utilized plasma samples in the acute phase of the AMI. A limitation is that the reperfusion strategy that was used for STEMI patients was predominantly thrombolysis; however, this is still the major reperfusion strategy in many parts of the world. We have a smaller number of females than males in the present cohort. We have also utilized an in-house assay for measuring NT-proBNP, but this is a well-published robust assay. It would be necessary to confirm our findings with the Roche Diagnostics Elecys NT-proBNP assay in an independent population before appropriate cut-offs can be implemented for risk scoring. One of the overwhelming strengths of the present study is the inclusion of consecutive patients with no restrictions on age. Our cohort of patients is similar to that which is encountered by clinicians around the world.

Conclusions

The present large single-centre study reveals that, following ACS, NT-proBNP gives complementary and independent information to the GRACE risk score for predicting early and late mortality. The inclusion of the NT-proBNP blood test is useful in risk-stratifying patients after ACS. A simple blood test has the ability to enhance a well-validated clinical risk score.

FUNDING

This work was supported by the British Heart Foundation [grant number FS/03/028/15486 (Junior Research Fellowship to S. Q. K.)].

Abbreviations: ACEI, angiotensin-converting enzyme inhibitor; ACS, acute coronary syndrome; ARB, AT1 (angiotensin II type 1) receptor blocker; AUC, area under the curve; BNP, B-type natriuretic peptide; CI, confidence interval; GRACE, Global Registry of Acute Coronary Events; HR, hazard ratio; MI, myocardial infarction; AMI, acute MI; NT-proBNP, N-terminal pro-BNP; PURSUIT, Platelet glycoprotein IIb/IIIa in Unstable angina: Receptor Suppression Using Integrilin Therapy; ROC, receiver-operating characteristic; STEMI, ST-elevation MI; NSTEMI, non-STEMI; TIMI, Thrombolysis in Myocardial Infarction

References

View Abstract