The aim of this study was to determine whether measurement of pre-operative brain natriuretic peptide can significantly improve risk stratification of vascular surgical patients. The study endpoint was postoperative raised troponins. Net reclassification improvement was determined for risk categories based on the Revised Cardiac Risk Index. Two reclassifications were conducted based on either the optimal discriminatory point or brain natriuretic peptide tertiles. Two hundred and sixty-seven patients were studied of whom 36 (13.5%) had raised postoperative troponin. The Revised Cardiac Risk Index score and the pre-operative brain natriuretic peptide were independent predictors of postoperative troponin elevation (p = 0.02 and p = 0.001, respectively). Reclassification based on the optimal discriminatory point significantly improved risk stratification (net reclassification improvement 38.3% (95% CI 9.3–67.3%), p = 0.01 for the entire cohort and 70.3% (95% CI 27.1–113.6%), p = 0.002 for intermediate risk patients). The brain natriuretic peptide tertiles only improved stratification of intermediate risk patients (net reclassification improvement 50.0% (95% CI 16.7–83.3%), p = 0.01). We have shown that measurement of pre-operative brain natriuretic peptide is relevant in the context of risk assessment in this cohort of patients.
It is important to evaluate new clinical biomarkers accurately to establish their role in risk prediction and clinical practice. The American Heart Association (AHA) has suggested a six-stage approach to the evaluation of novel biomarkers for cardiovascular risk prediction . These include: (i) proof of concept, where biomarker levels differ between subjects with and without an adverse cardiac outcome; (ii) prospective validation of subsequent cardiovascular events in subjects with a positive biomarker; (iii) whether or not the biomarker adds incremental value to existing risk indices; (iv) the clinical utility of the biomarker in changing recommended therapy; (v) whether or not changing practice according to biomarker results improves clinical outcomes; and (vi) cost-effectiveness.
Currently, B-type natriuretic peptides (which includes brain natriuretic peptide (BNP) and N-terminal pro-B-type natriuretic peptide ), when used for cardiovascular risk prediction for vascular surgery, fulfil the first two of these recommendations [3, 4]. Recently, it has been shown that BNP adds ‘incremental value’ to the Revised Cardiac Risk Index  in patients undergoing major non-cardiac surgery, by significantly improving the area under the receiver-operating characteristic (ROC) curve . There are no peri-operative BNP data addressing the last three AHA recommendations. The aim of this study was to determine whether measurement of pre-operative BNP can sufficiently alter the risk category classification of elective vascular surgical patients so as to change the recommended pre-operative management. As the Revised Cardiac Risk Index assesses peri-operative cardiac risk , an objective marker of myocardial necrosis (troponin elevation) was used in this study; troponin elevation is now a prerequisite for the diagnosis of myocardial infarction .
Ethical approval was granted by the Biomedical Ethics Committee of the Nelson R Mandela School of Medicine. This study was registered in the South African National Clinical Trial Register (unique registration number: DOH-27-0810-3320), and is a sub-cohort of a Medical Research Council (MRC) study that is currently still recruiting patients. All elective vascular surgical patients between February 2008 and July 2010 were eligible for recruitment and written informed consent was obtained from all subjects.
We prospectively collected data in a customised registry, including patients’ characteristics, surgical procedure conducted and the clinical risk predictors identified in the Revised Clinical Risk Index . The data set was complete for all patients recruited into the study, and the data were reviewed by one author (BB) for accuracy. All patients had troponin I and BNP levels measured pre-operatively and subsequently had troponin I levels measured on the first three postoperative days. The primary endpoint was a troponin I level above the upper reference limit (0.1 ng.ml−l) within the first three postoperative days.
The choice of anaesthetic technique was left to the discretion of the attending anaesthetist. There was no study protocol for the management of elevated troponin, and individual patient management was determined by the anaesthetic and surgical team. The attending clinicians were not blinded to the BNP or troponin I results.
The samples for BNP and troponin I were collected in EDTA and serum separator tubes (Greiner Bio-One, Frickenhausen, Germany), respectively. All samples were centrifuged and analysed immediately on the Advia Centaur Xp (Siemens Medical, Deerfield, IL, USA), utilising chemiluminescent technology. The analytical range for BNP is 0.58–1445 pmol.l−1 (2.01–5000 pg.ml−1) with a coefficient of variation of 3.5% and 3.8% at 500 pmol.l−1 (1730.1 pg.ml−1) and at 131 pmol.l−1 (453.3 pg.ml−1), respectively. The analytical range for troponin I is 0.006–50 ng.ml−1 with a coefficient of variation of 11.5% and 8.7% at 0.61 ng.ml−1 and at 5.45 ng.ml−1,respectively.
Categorical data were analysed using Fisher’s exact test or chi-squared test where appropriate. Continuous data were compared using independent samples t-test or Mann–Whitney U-test. Unadjusted and adjusted logistic regressions were conducted for age, risk factors and pre-operative BNP. The latter two were entered into the multivariate analysis using binary logistic regression analysis, to keep the event per variable ratio above 10 and hence minimise bias associated with the estimate of risk . A backward stepwise modelling technique was used, based on likelihood ratios, with entry and removal probabilities set at 0.05 and 0.1, respectively.
The optimal discriminatory point for BNP was determined using a ROC curve for postoperative troponin I elevation above the upper reference range. Tertiles for pre-operative BNP were also established, following natural log normalisation of the pre-operative BNP data.
Reclassification statistics were performed; the three AHA risk categories, that is low risk (0 risk factors), intermediate risk (1–2 risk factors) or high risk (≥ 3 risk factors) [5, 9] were used. The consensus in the literature is that patients with ≥ 3 risk factors may require further risk stratification and risk modification before surgery, while patients with no risk factors can proceed to surgery [9, 10]. Patients were reclassified into a higher or lower risk category according to whether the pre-operative BNP was above or below the optimal discriminatory point, respectively. Patients were also reclassified into low, intermediate or high risk categories, respectively, according to whether they were in the lowest, intermediate or highest pre-operative BNP tertiles.
The success of the reclassification by the addition of pre-operative BNP levels is described by the net reclassification improvement. This is the difference between the proportion of patients correctly and incorrectly reclassified  according to subsequent troponin elevation above the upper reference range. The equation for net reclassification improvement is shown in the Appendix. The upper range of the 95% CI is 200%, as this comprises potentially 100% for correctly reclassifying patients into a lower risk category and 100% for correctly reclassifying patients into a higher risk category (M. Pencina, personal communication). Data analysis was with spss 15.0 for Windows (IBM, Chicago, IL, USA) and sas 9.1 (SAS Institute Inc, Cary, NC, USA).
During the study period, 758 patients were scheduled for elective vascular surgery, of whom 590 patients consented to the study. A total of 297 of these patients were recruited into the BNP biomarker sub-study; there was no bias in the selected cohort for this study (p = 0.823). Due to the high rate of troponin elevation in patients who underwent supra-inguinal vascular surgery (11 out of 30 patients, 37%), compared with the rest of the vascular patients (36 out of 267, 14%), a post-hoc decision was taken to analyse only patients who underwent infra-inguinal surgery (267 of the 297 patients, Table 1).
Table 1. Baseline characteristics of patients with and without postoperative troponin elevation. Data are expressed as number (proportion) or median (IRQ [range]).
Of the 267 patients, six (2%) died within 30 days of surgery (four non-cardiac deaths and two cardiac deaths). Ten (4%) of the 267 patients had a pre-operative troponin I above the upper reference limit (> 0.1 ng.ml−1). Nine (90%) of these patients with an elevated pre-operative troponins also had a BNP above the optimal discriminatory point.
The relationship between pre-operative BNP levels and postoperative troponin elevation is shown in Table 1. The area under the curve (AUC) for postoperative raised troponin was 74% (95% CI 65–83%) for pre-operative BNP and 66% (57–76%) for the number of risk factors (Fig. 1). The optimal discriminatory point (> 69 pg.ml−1) for pre-operative BNP had a sensitivity of 75% and specificity of 64%.
The unadjusted (univariate) and adjusted predictors of postoperative raised troponin are shown in Table 2. The multivariate analysis of the Revised Cardiac Risk Index score and pre-operative BNP is shown in Table 3. Both the Revised Cardiac Risk Index and the pre-operative BNP were independent predictors of postoperative troponin elevation.
Table 2. Unadjusted and adjusted odds ratios for postoperative raised troponin. Values are OR (95% CI).
Table 3. Result of the multivariate analysis for postoperative raised troponin. Values are OR (95% CI).
BNP, brain natriuretic peptide.
Revised Cardiac Risk Index (per unit risk factor increase)
BNP (per pg.ml−1 increase)
Reclassification statistics using the optimal discriminatory point for BNP or BNP tertiles according to risk categories are summarised in Table 4. The use of the optimal discriminatory point of pre-operative BNP increased the proportion of patients correctly classified as low or high risk patients from 0.36 to 0.74. The BNP tertile model improved the proportion of patients correctly classified as low or high risk patients from 0.36 to 0.56. All the models, with the exception of the use of BNP tertiles applied to the entire cohort, significantly improved the pre-operative risk classification (as reflected by the net reclassification improvement). However, the model with the trend to the best predictive performance used the optimal discriminatory point for pre-operative BNP to reclassify only intermediate risk vascular surgical patients (70% net reclassification improvement, p = 0.002).
Table 4. Reclassification for postoperative raised troponin based on Revised Cardiac Risk Index categories. Values are proportion (95% CI).
Reclassification improvement; %
Net reclassification improvement
Positive likelihood ratio
Negative likelihood ratio
BNP, brain natriuretic peptide.
BNP above the optimal discriminatory point
Intermediate risk category only
Intermediate risk category only
This study suggests that the measurement of pre-operative BNP serum levels to reclassify intermediate risk vascular surgical patients into low or high risk categories has significant clinical utility. This has implications for clinical practice; patients reclassified as low risk could potentially proceed to surgery, while patients reclassified into the high risk category would require further investigation and/or further cardioprotective strategies before surgery. This could be considered as a reasonable extension to the current AHA pre-operative cardiovascular evaluation algorithm .
A recent study highlighted the need to evaluate the utility of pre-operative B-type natriuretic peptides in peri-operative risk stratification . We believe that we have fulfilled the fourth requirement proposed by the AHA in the evaluation of pre-operative BNP for elective vascular surgical risk prediction whereby the use of pre-operative BNP would significantly change the risk classification of a patient and hence the clinical therapy required . This is an important finding as the clinical utility of the Revised Cardiac Risk Index when used for vascular surgery alone is poor. A recent meta-analysis showed an AUC of 64% (95% CI 61–66%), with a positive likelihood ratio (LR) of 1.56 (1.42–1.73) and a negative LR of 0.55 (0.53–0.82) when patients were split into two groups (< 2 and ≥ 2 risk factors) .
The use of ROC curves has limitations when assessing the utility of prognostic biomarkers such as BNP, as they merely assess discrimination, which is the ability to separate cases from controls . In risk prediction, it is the calibration of the model, which is an analysis of how closely the predicted probabilities agree with the observed outcomes , that is more important. As has been seen in previous medical models, the addition of BNP has resulted in significant improvements in risk prediction that were not identified using the ROC curve alone, where the 95% CI overlapped [15, 16], as was the case in our study. However, the addition of BNP significantly improved the predicted proportion of patients who went on to suffer postoperative raised troponins, which has also been shown in medical models using BNP [15, 16].
Reclassification statistics are performed on a pre-existing risk classification , such as the Revised Cardiac Risk Index, which has been adopted by both the AHA and the European Society of Cardiologists in their proposed cardiovascular evaluation algorithms [9, 10]. The most appropriate clinical application for pre-operative BNP appears to be in the reclassification of intermediate risk category patients (defined as the presence of one or two risk factors). This may be because the Revised Cardiac Risk Index has poor clinical performance in intermediate risk patients (LR of 0.34 and 2.72) . However, while BNP significantly improved the risk classification of these patients, the small sample size does not allow us to determine reliably which approach to BNP reclassification is optimal due to overlapping confidence intervals. It is, however, inappropriate to consider pre-operative BNP reclassification in all vascular surgical patients. There are potential patients whose clinical risk category should not change irrespective of their pre-operative BNP level. While the net reclassification improvement of the entire cohort still showed significant improvement in risk stratification based on the optimal discriminatory point, it was not as impressive as when it was only applied to the intermediate risk patients, and indeed when the risk stratification was based on BNP tertiles for the entire cohort, it resulted in an insignificant improvement. This may be explained by the good clinical and statistical performance of the Revised Cardiac Risk Index in patients with no clinical risk factors . It is therefore unlikely that pre-operative BNP would be an appropriate risk stratification tool in patients without risk factors. Indeed, in our study, 92 patients had no clinical risk factors, of whom five had postoperative raised troponin (5%) compared to 17% in the patients with clinical risk factors (p = 0.005). In this group of patients it may be inappropriate to consider further risk stratification using pre-operative BNP.
In addition, the use of BNP tertiles to identify low, intermediate and high risk categories also appears to be an inappropriate strategy to re-stratify an entire vascular surgical cohort. We assessed BNP tertiles, in addition to the use of the optimal discriminatory point, as extremes of pre-operative BNP serum levels may be associated with a significantly higher odds ratio for postoperative raised troponins, which may justify moving patients across two risk categories (for example from low risk to high risk based on an extremely high BNP). As there is a dose–response relationship with cardiovascular risk factors, reclassification across more than one risk category based on the extent of BNP elevation appears reasonable . However, the use of BNP tertiles resulted in an insignificant improvement in risk stratification for the entire cohort, and a trend to a worse performance than in the adoption of the optimal discriminatory point when applied to the intermediate risk category. Another problem with the use of BNP tertiles for risk stratification of the intermediate risk category is that 50 (34%) of the original intermediate risk patients were not reclassified as they were found to be in the middle BNP tertiles, as opposed to the use of the optimal discriminatory point that significantly improves the reclassification of all intermediate risk patients.
Recent reports have suggested a very high prevalence of pre-operative elevated troponins in vascular surgical patients [19, 20]. Elevated pre-operative troponins were found in 34% of patients presenting for critical limb ischaemia with acute symptomatology or unstable coronary syndromes  and in 7% of patients undergoing leg amputation for chronic peripheral arterial vascular disease. Ten (4%) of the patients in this cohort had pre-operative troponin levels above the upper reference limit. The presence of elevated pre-operative troponins in our elective patients is therefore consistent with these other vascular surgery studies. Certainly, this finding is associated with an adverse outcome, and the utility of pre-operative troponin evaluation for risk stratification in vascular patients needs further investigation. Pre-operative BNP did, however, identify 90% of the patients with an elevated pre-operative troponin level.
A limitation of our data is that it contains a very small sample of patients undergoing supra-inguinal vascular surgery and hence these patients were not analysed. It would be appropriate therefore in the future to analyse supra-inguinal patients separately to determine appropriate BNP cut-off points for these patients. Unfortunately, the sample size of this study is too small to address this question.
The optimal discriminatory (or ‘cut-off’) point requires further investigation. In vascular surgical patients the optimal discriminatory points range from 38 to 100 pg.ml−1 for 30-day outcomes for BNP, and appear to be influenced by the extent of the surgery . Currently there are no data regarding the optimal discriminatory point for BNP. Furthermore, although there are data supporting pre-operative BNP as an intermediate term predictor of major adverse cardiac events [3, 21] there are no data on the clinical utility of the pre-operative BNP and risk reclassification and subsequent longer term outcomes.
It is appropriate now to validate this model prospectively before considering the fifth recommendation of the AHA proposal on evaluating clinical biomarkers, which will involve determining whether responding to the risk category reclassification of elective vascular surgical patients shown in this study improves patient outcomes .
In conclusion, the use of pre-operative BNP to reclassify elective vascular surgical patients into higher or lower risk categories has significant clinical utility for pre-operative risk stratification.
We would like to thank Ms Tonya Esterhuizen and Dr Michael Pencina for statistical assistance. This study was funded by a research grant from the Medical Research Council of South Africa, who had no role in the design of the study, data collection, data analysis, data interpretation or the writing of the report. No competing interests declared.
Appendix Equation for net reclassification improvement used in the analysis .
Where ‘NRI’ represents the net reclassification improvement; ‘up’ the reclassification into a higher risk category; ‘down’ the reclassification into a lower risk category; ‘Pr’ the proportion; ‘case’ the patients with postoperative raised troponin; ‘control’ the patients without postoperative raised troponin.