A meta-analysis of the utility of pre-operative brain natriuretic peptide in predicting early and intermediate-term mortality and major adverse cardiac events in vascular surgical patients

Summary

We conducted a meta-analysis of the utility of pre-operative B-type natriuretic peptide (BNP) and N-terminal-pro B-type natriuretic peptide in predicting early (< 30 days) and intermediate (< 180 days) term mortality and major adverse cardiac events (cardiac death and nonfatal myocardial infarction) in patients following vascular surgery. A Pubmed Central and EMBASE search was conducted up to January 2008. Of 81 studies identified, seven prospective observational studies were included in the meta-analysis representing five patient cohorts: early outcomes (504 patients) and intermediate-term outcomes (623 patients). A B-type natriuretic peptide or N-terminal-pro B-type natriuretic peptide above the optimal discriminatory threshold determined by receiver operating characteristic curve analysis was associated with 30-day cardiac death (OR 7.6, 95% CI 1.33–43.4, p = 0.02), nonfatal myocardial infarction (OR 6.24, 95% CI 1.82–21.4, p = 0.004) and major adverse cardiac events (OR 17.37, 95% CI 3.31–91.15, p = 0.0007), and intermediate-term, all-cause mortality (OR 3.1, 95% CI 1.85–5.2, p < 0.0001), nonfatal myocardial infarction (OR 2.95, 95% CI 1.17–7.46, p = 0.02) and major adverse cardiac events (OR 3.31, 95% CI 2.1–5.24, p < 0.00001). B-type natriuretic peptide and N-terminal-pro B-type natriuretic peptide are potentially useful pre-operative prognostic tests in vascular surgical patients.

Vascular surgery is associated with major adverse peri-operative cardiac events. Unfortunately, the currently used pre-operative diagnostic tests for these patients are not statistically robust enough to accurately predict these events [1]. This was recently confirmed in a meta-analysis of six pre-operative tests for vascular surgical patients, which included: ambulatory ECG, exercise ECG, radionuclide ventriculography, myocardial perfusion scintigraphy, dipyridamole stress echocardiography and dobutamine stress echocardiography [2]. Dobutamine stress echocardiography had a trend to the best performance of the six tests for the prediction of major adverse cardiac events (defined as peri-operative cardiac death and nonfatal myocardial infarction (MI)) within 30 days of surgery [2], with a positive likelihood ratio (LR) and negative LR of 2.8 and 0.21 respectively. As statistically good discrimination requires a LR of < 0.2 and > 10, there are clearly clinical limitations to the utility of these pre-operative tests for vascular surgical patients [1].

There has been recent interest in brain natriuretic peptide (BNP) and N-terminal pro-brain natriuretic peptide (NT-pro-BNP) as prognostic biomarkers of death and major cardiovascular events, even after control for other cardiovascular risk factors [3]. BNP is an integrated marker of multiple aspects of cardiac dysfunction, including both myocardial stretch and ischaemia, and is clinically useful in identifying patients with acute cardiac failure [4, 5]. Recent reports suggest that pre-operative BNP levels may be a better prognostic test than dobutamine stress echocardiography for both 30-day [6] and long-term [7] major cardiac morbidity following vascular surgery.

The aim of our meta-analysis was therefore to determine the performance of pre-operative BNP and NT-pro-BNP as a diagnostic test for the prediction of mortality and major cardiac morbidity following vascular surgery. In addition, it was a prerequisite that this meta-analysis included a 30-day major adverse cardiac events (MACE), defined as a composite of cardiac death and nonfatal myocardial infarction (MI). The reason being that Kertai et al. [2] used this outcome in their meta-analysis of pre-operative diagnostic tests in vascular surgical patients. Thus, presenting the same outcome in this meta-analysis would allow for the comparison of the performance of pre-operative BNP or NT-pro-BNP, with the performance of the diagnostic tests presented by Kertai et al. [2].

Methods

We conducted a meta-analysis of the utility of BNP and NT-pro-BNP in predicting early (< 30 days) and intermediate (< 180 days) postoperative mortality and MACE in patients following vascular surgery. The Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines were adhered to in conducting and reporting this meta-analysis [8].

Study identification and selection

On the 9th August 2007, BB conducted a search of Pubmed Central from 1966 to 2007 and a search of EMBASE from 1988 to week 40 of 2007. The terms used in the search strategy were: ‘brain natriuretic peptide’ and ‘vascular surgery’ and ‘human’ and ‘adults' for the Pubmed Central search; ‘brain natriuretic peptide’ and ‘vascular surgery’ and ‘human’ for the EMBASE search. The search was repeated and updated on 10th January 2008 by BB. We then manually searched the reference lists of all studies and reviews identified.

We included observational peri-operative studies of pre-operative BNP and/or NT-pro-BNP that reported mortality (all-cause and/or cardiac) and cardiac morbidity (nonfatal MI) in patients undergoing vascular surgery. Studies that reported on BNP or NT-pro-BNP in noncardiac surgery but not specifically in vascular surgery were included if the authors were able to supply the specific data of the vascular surgical patients. We excluded studies published only in abstract form, nonhuman studies, cardiac surgical studies, paediatric studies and in-vitro or nonoperative human studies.

Data extraction and quality assessment

LP and RR independently screened citations, abstracted data and assessed methodological quality, using a standardised data extraction sheet. A third reviewer (BB) resolved any disagreements. If data needed clarification or was not presented in the publication, we contacted the original authors.

Data analysis

The quality of each study was assessed according to the Newcastle Ottawa Quality Assessment Scale for cohort studies (NOS) [9]. The NOS scale and the definitions used for this meta-analysis are shown in Table 1. Comparability of the cohorts from the included studies were analysed using an ANOVA test for age or a chi-square test for gender, ischaemic heart disease (IHD), congestive heart failure (CHF), diabetes mellitus (DM), renal dysfunction and pre-operative medical therapy (graphpad instat version 3.06 (2003); GraphPad Software Inc., San Diego, CA, USA).

Table 1. The Newcastle Ottawa Quality Assessment Scale for cohort studies (NOS) [9] as adapted for use in this meta-analysis.

Category	Criteria	Maximum number of stars
Selection	1. Is the cohort representative of vascular surgical patients?	1
	2. Are the patients with a low or normal BNP from the same vascular surgical cohort?	1
	3. Is it possible to ascertain the plasma BNP or NT-pro-BNP in all patients?	1
	4. Is it possible to exclude the presence of major cardiac morbidity at the start of the study? For example, through examination of pre-operative troponin levels.	1
Comparability	1. The data analysed in this meta-analysis must only include vascular surgical patients.	1
	2. The age, gender and prevalence of IHD, DM and renal dysfunction should be comparable between studies. Exclusion of comorbidities was undesirable.	1
Outcome	1. Assessment must be independent blinded, record linkage or secure records.	1
	2. Follow-up must be long enough for outcome to occur. Defined as 30 days and/or 180 days.	1
	3. Follow-up must be adequate. Defined as complete follow-up, or description of drop outs to ensure that they were not secondary to any of the outcomes being investigated.	1

• BNP, brain natriuretic peptide; IHD, ischaemic heart disease; DM, diabetes mellitus.

From each study, we extracted data on mortality (all-cause and cardiac), nonfatal MI and MACE (defined as cardiac death or nonfatal MI). We examined these outcomes in the peri-operative period (< 30 days) and in the intermediate-term (< 180 days). The meta-analysis was conducted using revman version 4.2.8 software (The Nordic Cochrane Centre, Kobehavn, Denmark). Heterogeneity between studies was assessed using univariate chi-square analysis. Random or fixed effects models were used based on the presence or absence of significant heterogeneity between studies, respectively. Pooled dichotomous outcomes were reported as the odds ratio (OR) and the 95% confidence intervals (CI).

A funnel plot was not constructed for the outcome of MACE at 30-days due to the significant heterogeneity between the studies. The funnel plot would have investigated whether the outcomes reported were biased in a particular direction by publication bias. However, when there is significant heterogeneity between exposure (a high BNP) and outcome, then this will lead to asymmetry in the funnel plot [10].

Results

Included studies

We identified 81 studies between 1966 and January 2008, of which five patient cohorts reported in seven studies [6, 7, 11–15] fulfilled our inclusion criteria (Figure 1). All the included studies were prospective, observational studies. Additional, unpublished data was obtained from the authors of all the included studies.

Figure 1

Flow chart of included and excluded studies.

Three studies were of vascular surgical patients only [6, 13, 15], while the authors of the two noncardiac surgical studies included [12, 14] extracted the vascular surgical patients data for this meta-analysis. All patients underwent either elective aortic or peripheral vascular disease surgery. One study also included patients undergoing peripheral limb amputations [13]. The nature of the vascular surgery, demographics of the patients and the reported pre-operative medical therapy are shown in Table 2.

Table 2. Patient demographics of studies included in the meta-analysis.

First author [reference]	Type of surgery	Patient (N)	Age	Male, n (%)	CAD, n (%)	CCF, n (%)	Diabetes, n (%)	Renal failure, n (%)	Pre-operative statin, n (%)	Pre-operative beta blocker, n (%)
Yeh [12]	Major elective vascular surgery	5	71 [6–91]	3 (60)	3 (60)	0 (0)	2 (40)	2 (40)	0 (0)	3 (60)
Berry [13]	Major elective vascular surgery and amputations	41	NR	27 (66)	17 (41)	NR	11 (27)	NR	NR	6 (15)
Feringa [6]	Major elective vascular surgery Short-term cohort	170	59 (13)	120 (71)	69 (41)*	39 (23)	31 (18)	6 (4)	87 (51)	109 (64)
Feringa [7]	Major elective vascular surgery Intermediate cohort	335	62.2 (12.4)	256 (76)	165 (49)	57 (17)	68 (20)	20 (6)†	180 (54)	236 (70)
Cuthbertson [11, 14]	Major elective vascular surgery	70	70.1 (10.4)	48 (69)	31 (44)	1 (1)	18 (26)	0 (0)	48 (69)	22 (31)
Mahla [15]	Major elective vascular surgery	218	70 (9)	170 (78)	57 (26)*	10 (5)	78 (36)	0 (0)‡	78 (36)	167 (77)

• Values are number (proportion), mean (SD) or median (range).
• SD, standard deviation; CAD, coronary artery disease; CCF, congestive cardiac failure; NR, not reported.
• *Previous myocardial infarction.
• †Creatinine > 177 μmol.l⁻¹.
• ‡Patients with creatinine > 123μmol.l⁻¹ were excluded (n = 29).

The NOS assessment [9] for all included studies are listed in Table 3. All the studies were given one star for comparability, which was for providing data only for vascular surgical patients. No study was awarded a second star for comparability, as the proportion of patients with IHD, CHF, DM, renal dysfunction, and receiving pre-operative statin or beta-blocker were all significantly different between the studies (p < 0.01 for all comparisons). Only the reported proportion of male patients was similar between the included studies (p = 0.24).

Table 3. Characteristics of the studies included in the meta-analysis.

First author [reference]	Follow-up period	B-natriuretic molecular form	Optimal discriminatory threshold (pg.ml^−1)	NOS Score [9]
				Selection	Comparability	Outcome
Short-term studies
Yeh [12]	Short term (to discharge < 30 days)	NT-pro-BNP	450	****	*	***
Berry [13]	Short term (48 h)	BNP	100	****	*	***
Cuthbertson [14]	Short term (30 days)	BNP	40	****	*	***
Feringa [6]	Short term (30 days)	NT-pro-BNP	533	****	*	***
Mahla [15]	Short term (30 days)	NT-pro-BNP	280	****	*	***
Intermediate-term studies
Cuthbertson [11]	Intermediate term (180 days)	BNP	35	****	*	***
Feringa [7]	Intermediate term (180 days)	NT-pro-BNP	319	****	*	***
Mahla [15]	Intermediate term (180 days)	NT-pro-BNP	280	****	*	***

Three studies used NT-pro-BNP [6, 12, 15] and two studies measured plasma BNP [13, 14]. All the studies reported early (< 30 day) outcomes (504 patients) and three studies [7, 14, 15] collected data on intermediate-term (< 180 day) outcomes (623 patients). The optimal discriminatory threshold for BNP or NT-pro-BNP was determined by the receiver operating characteristic curve analysis in the original studies. This allowed for determination of the optimal sensitivity and specificity for adverse outcomes according to the pre-operative BNP or NT-pro-BNP level. The optimal discriminatory threshold was determined in vascular surgical cohorts in three studies [6, 13, 15] and in mixed noncardiac surgical cohorts in two studies [12, 14].

Although, the outcome data in this meta-analysis is categorised according to the optimal discriminatory threshold presented in the original studies, the postoperative time at which the threshold was determined varied between studies: from outcomes recorded at 48 h [13], three days [14], until discharge [12], 30 days [6], a mean of 14 months [7], a median of 654 days [11], to a median of 826 days [15]. The optimal discriminatory thresholds and follow-up periods used in the meta-analysis are listed in Table 3.

Outcomes

Outcomes were categorised according to the optimal discriminatory threshold for BNP presented in each individual study. It was possible to determine all cause mortality for all the intermediate-term studies [7, 14, 15], and for three of the five short-term outcome studies [12, 14, 15]. With the exception of one study [6], cardiac death and nonfatal MI were reported in all other short-term studies. Cardiac death and nonfatal MI were only reported in one intermediate-term study [15]. MACE was reported in all short-term studies and two [7, 15] of the three intermediate-term studies. There was complete follow-up in all studies for all these outcomes reported.

A high BNP or NT-pro-BNP was associated with a significantly increased risk of early (< 30 day) cardiac mortality, nonfatal MI and MACE. It was also associated with a significantly increased risk of intermediate-term (< 180 day) all-cause mortality, nonfatal MI and MACE. Short-term all-cause mortality and intermediate-term cardiac death were only associated with a trend towards an adverse outcome with an elevated BNP (2-7). The single study reporting cardiac mortality and nonfatal myocardial infarction at 180 days [15] showed a trend towards cardiac mortality (OR 2.82, 95% CI 0.66–12.13, p = 0.16) and significantly increased nonfatal myocardial infarction associated with a high BNP (OR 2.95, 95% CI 1.17–7.46, p = 0.02).

Figure 2

Pre-operative BNP and NT-pro-BNP and 30-day all cause mortality.

Figure 3

Pre-operative BNP and NT-pro-BNP and 30-day cardiac mortality.

Figure 4

Pre-operative BNP and NT-pro-BNP and 30-day nonfatal myocardial infarction.

Figure 5

Pre-operative BNP and NT-pro-BNP and 30-day major adverse cardiac events.

Figure 6

Pre-operative BNP and NT-pro-BNP and 180-day all cause mortality.

Figure 7

Pre-operative BNP and NT-pro-BNP and 180-day major adverse cardiac events.

Discussion

The predictive value of BNP/NT-pro-BNP for MACE following vascular surgery

The results presented in this meta-analysis allow the reader to compare the performance of pre-operative BNP or NT-pro-BNP with meta-analytical data presented for ambulatory ECG, exercise ECG, radionuclide ventriculography, myocardial perfusion scintigraphy, dypridamole stress echocardiography and dobutamine stress echocardiography in vascular surgical patients [2]. Dobutamine stress echocardiography was reported to have a sensitivity and specificity of 85% (95% CI 74–97%) and 70% (95% CI 62–79%) respectively for the prediction of 30-day MACE in vascular surgical patients [2]. The sensitivity and specificity for BNP or NT-pro-BNP for the same outcome in vascular surgical patients presented in this paper is 83% (95% CI 69–91%) and 73% (95% CI 68–77%). The positive and negative LR for dobutamine stress echocardiography and BNP (or NT-pro-BNP) were 2.8 and 0.21, and 3.1 and 0.23, respectively. This meta-analysis could not confirm, therefore, that BNP or NT-pro-BNP performs better as a pre-operative diagnostic test than dobutamine stress echocardiography in vascular surgical patients for 30 day cardiac outcomes, as the 95% CI of both tests overlap for both sensitivity and specificity. However, it does suggest that BNP and NT-pro-BNP are at least as predictive as the best pre-operative diagnostic tests of major cardiac events in vascular surgical patients.

Does this meta-analysis present the optimal diagnostic performance of BNP or NT-pro-BNP for MACE following vascular surgery?

The optimal discriminatory threshold for BNP or NT-pro-BNP used in the studies included in this meta-analysis may be incorrect for prediction of 30 day MACE following vascular surgery. Firstly, in two studies the threshold was derived from mixed noncardiac surgery cohorts [12, 14], as opposed to vascular surgical cohorts alone. Including nonvascular surgical patients in the derivation of an optimal discriminatory threshold may alter the sensitivity and specificity of the test. Indeed an analysis of 30 day MACE of only the vascular surgical cohorts reveals a higher OR associated with an adverse outcome in the high BNP group than when the mixed cohorts [12, 14] are included in the analysis (OR 25.9, 95% CI 2.8–240.8 and OR 17.4, 95% CI 3.3–91.2 respectively). However, as the CI overlap in both these analyses, it is impossible to confirm the suggestion that the sensitivity and specificity would be improved by only including vascular surgical cohorts.

Secondly, the derivation of the optimal discriminatory threshold is also affected by the event time horizon. Shorter outcome periods are associated with higher discriminatory thresholds for BNP. This is seen in the study of Berry et al. [13] which reported outcomes at 48 h. Similarly, this is evident in the cohort presented by Cuthbertson et al. [11], where the optimal discriminatory threshold increased from 35 pg.ml⁻¹ for medium term mortality to 40 pg.ml⁻¹ for short term mortality [14]. It should be noted that of the five studies included in the 30-day outcome analyses, three had determined the optimal discriminatory threshold at a time period shorter than 30-days [12–14], one study at 30 days [6], and one study at a median of 826 days [15]. For the intermediate-term outcomes, the optimal discriminatory threshold was determined at a time period longer than 180 days for all three studies analysed [7, 11, 15].

Thirdly, although renal dysfunction (defined as a serum creatinine > 177 μmol.l⁻¹) is an important predictor of major adverse cardiac events in noncardiac surgery [16], decreasing renal function is associated with a lower specificity of NT-pro-BNP for adverse cardiac events [17, 18]. The utility of NT-pro-BNP as a predictive marker is highly dependant on the severity of the renal dysfunction, with the best performance in patients with a glomerular filtration rate (GFR) ≥ 90 ml.min⁻¹.1.73 m⁻² [17]. With a GFR < 30 ml.min⁻¹.1.73 m⁻² the prognostic ability of NT-pro-BNP is questionable in vascular surgical patients [17]. Of the three studies that reported on NT-pro-BNP, only one study excluded patients with a serum creatinine exceeding 123 μmol.l⁻¹ [15]. Thus, it is likely that inclusion of patients with renal dysfunction in the other two studies [6, 12] may have decreased the specificity of the test [17]. Although exclusion of patients with renal dysfunction in the study by Mahla et al. [15] may have decreased the adverse cardiac event rate secondary to a decrease in cardiac clinical risk factors [16], this would not have adversely affected the sensitivity or specificity of the test [1] in their study.

Finally, there is an inverse relationship between an increasing body mass index and the optimal discrimination threshold for BNP [19]. This may be an important confounder in vascular patients, as it is well documented that abdominal obesity is an independent predictor of acute MI in medical patients [20].

Thus, it is likely that the optimal discriminatory thresholds presented in Table 3 do not represent the optimal discriminatory thresholds for outcomes at 30 or 180 days postoperatively. In addition, these factors may also partly account for the significant heterogeneity seen in the 30-day MACE outcome (Figure 5). Based on the large difference in the discriminatory thresholds presented in Table 3 and thus used to classify outcomes for this meta-analysis, it would be currently clinically inappropriate to use the data presented in this paper to prognosticate peri-operative outcomes following vascular surgery. This would only be possible once accepted discriminatory thresholds have been established.

Could BNP or NT-pro-BNP theoretically outperform other pre-operative diagnostic tests for vascular surgical patients in the future?

Determining the optimal discriminatory threshold for specific time period (such as 30- or 180-day) outcomes in vascular surgical patients would improve the performance of the test. Furthermore, there are two additional reasons why BNP or NT-pro-BNP may outperform other diagnostic tests in the future. The first is the ability of BNP and NT-pro-BNP to predict heart failure. Although BNP is an independent predictor of all-cause mortality, first cardiovascular events, atrial fibrillation, stroke and coronary heart disease events, it is most predictive of heart failure, even after controlling for numerous cardiovascular risk factors [3]. With the relative increase in peri-operative cardiac morbidity associated with heart failure, in comparison with patients with IHD [21], it is possible that the prognostic performance of BNP as a pre-operative test will increase relative to other pre-operative tests when the complication of postoperative cardiac failure is considered in addition to cardiac death and nonfatal MI.

Secondly, the postoperative NT-pro-BNP levels in vascular surgical patients has been shown to be a better predictor of MACE than the pre-operative NT-pro-BNP level [15]. While this meta-analysis confirms that the pre-operative BNP or NT-pro-BNP level is important in stratifying patients pre-operatively, the rise in BNP or NT-pro-BNP peri-operatively may add significant additional prognostic information and allow for further risk stratification [15]. This is an area which certainly needs more investigation.

Are BNP and NT-pro-BNP of equivalent utility in vascular surgical patients?

Whether BNP or NT-pro-BNP is preferable for vascular surgical patients remains unclear. While NT-pro-BNP may be limited by renal dysfunction [17], there is some evidence to suggest that NT-pro-BNP may be a slightly better predictor of outcomes in medical patients with congestive cardiac failure [22] and ischaemic heart disease [23]. It is possible that the longer half-life of NT-pro-BNP may result in it being less affected by peri-operative haemodynamic shifts, than BNP [15], although we are unaware of any studies examining this phenomenon.

Conclusion

The prognostic performance of BNP and NT-pro-BNP are at least equivalent to the best of the ‘traditional’ pre-operative diagnostic tests [2] of cardiac morbidity for vascular surgical patients. It is possible that the performance of BNP and NT-pro-BNP as pre-operative diagnostic tests may be better than what is presented in this meta-analysis.

Acknowledgements

We would like to thank Drs Berry, Cuthbertson, Feringa, Goei, Kingsmore, Leibowitz, Mahla, Planer, Poldermans, Silva and Yeh for additional trial information.