Prognostic Value of Brain Natriuretic Peptide in Noncardiac Surgery: A Meta-analysis

The methods for this meta-analysis conform to those proposed by Stroup et al.  6 

The primary objectives of this analysis were to assess whether elevated preoperative plasma BNPs predict short-term MACE, cardiac mortality, and all-cause mortality in patients undergoing noncardiac surgery. The secondary objectives were to determine the relationship between preoperative BNPs and longer-term major adverse cardiovascular event (MACE) and all-cause mortality. On the basis of the range of timepoints in published studies, we defined short-term outcomes as less than 43 days after surgery, whereas longer-term outcomes were defined as 6 months or more after surgery. MACE was defined as cardiac death or nonfatal myocardial infarction. Cardiac death required evidence of myocardial infarction, cardiac arrhythmia, or congestive cardiac failure, and it was adjudicated by the authors of individual studies.

Studies were included in this analysis if they had prospectively collected data on perioperative mortality or MACE in adult patients undergoing noncardiac surgery with either preoperative BNP or NT-proBNP measurement. Study authors were contacted to confirm outcomes data, allowing the creation of a 2 × 2 table based on BNP and NT-proBNP levels (normal and elevated) and outcome (mortality or MACE).

We searched MEDLINE (1966–2008) and EMBASE (1996–2008) databases, using the search terms “brain natriuretic peptide,”“b-type natriuretic peptide,”“BNP,”“NT-proBNP,” and “surgery.” Abstracts from the scientific sessions of the American Heart Association, American College of Cardiology, European Society of Cardiology, British Cardiac Society, American Society of Anesthesiology, and European Society of Anaesthesiology (2004–2008) were searched manually, as were the references of individual papers. The last search was performed on June 11, 2008.

One author (Dr. Ryding) performed the electronic and manual searches and listed the trials that were eligible for inclusion in the study. Study selection was initially performed by review of title. Candidate abstracts were then reviewed and selected for data retrieval. Only prospective studies were considered eligible for inclusion. Two authors (Dr. Ryding and Dr. Kumar) independently reviewed each study for quality assessment and extracted data on studies and patient characteristics, as well as outcomes, using standardized extraction forms. Data quality was verified by an additional reviewer (Dr. Worthington), and disagreements were resolved by discussion. Because no standardized quality scoring system exists for quality assessment of observational studies, the components of the quality review were derived largely from the Cochrane quality checklist for prognostic studies.7Studies were assessed for the presence of ten features: type of study, description of patient sample characteristics, description of inclusion and exclusion criteria, potential selection bias, a priori  definition of study outcomes, objectivity of outcomes, completeness of follow-up, missing data, evidence of selective reporting of data, and completeness of follow-up.

For each study, the following individual data were extracted: general data (study design), patients (number of included patients, mean age, gender, type of surgery, β-blocker usage), BNP assays (name of the assay, type of BNP measured [NT-proBNP or BNP], cutoff level, timing of determination, and overall BNP-positive patients), deaths (cardiac or all cause mortality), and MACE among BNP-positive or -negative patients.

Unadjusted data were used exclusively in all meta-analyses. Outcomes are reported using random-effects models to allow for interstudy variability. Cochrane chi-square test and the I²test for heterogeneity were used to assess between-study heterogeneity. Statistically significant heterogeneity was considered present at P < 0.10 and I²> 50%. Pooled odds ratios (ORs) were reported with 95% confidence intervals (CI). Publication bias was assessed visually by the use of funnel plots by using the method of Sterne and Egger.8 

Separate analyses were performed for MACE, all-cause mortality, and cardiac mortality, BNP, or NT-proBNP and duration of follow up. The association between cutoff and the OR was assessed using meta-regression.9Analyses were performed with Review Manager (RevMan) Version 5.0.12 (The Cochrane Collaboration, 2008, The Nordic Cochrane Centre, Copenhagen, Denmark) and meta-regression was performed with MIX software (Kisato Clinical Research Centre, Sagamihara, Kanagawa, Japan).10The authors had full access to and take full responsibility for the integrity of the data. All authors have read and agree to the manuscript as written.

A total of 792 articles and 350 abstracts were retrieved with the search strategy. From these, 757 papers and 343 abstracts were excluded on the basis of the title (fig. 1); 35 papers and 7 abstracts were identified that studied BNPs in the context of noncardiac surgery; 21 papers and 1 abstract did not report mortality data, and these were excluded; 4 abstracts were subsequently published in full and were therefore not included; 1 paper was excluded because it was a retrospective analysis.11Therefore, 15 publications, involving 4,856 individual patients, were selected for analysis.

The characteristics of the studies are described in table 1 and 2. The demographic data of the study populations were similar in terms of age and gender, except for three studies that involved predominantly elderly female patients.12–14Mixed noncardiac surgery was reported in seven studies,14–19fractured neck of femur surgery was reported in one,13and vascular surgery was reported in seven.18,20–25Two studies looked at patients undergoing emergency surgery,12,13and the remainder were nonemergent. Reported perioperative β-blockade use ranged between 0% and 77% (table 1 and 2). Follow-up varied among publications; 11 reported short-term mortality data (in hospital or 2–42 days after operation),13–21,253 reported only longer-term data (6 months or greater),22,24,26and 1 reported both short-term and longer-term data.12 

Three different assays were used to measure BNP or NT-proBNP (table 3). Most studies measured BNPs the day before surgery (range: day of surgery to 24 days before surgery), and one reported both preoperative and postoperative measurements.24Analyses were restricted to preoperative BNP data only. In general, optimal cutoff values were derived by the authors of the studies from the receiver operator curve characteristics for MACE. There was wide variation in cutoff values between studies, with those for NT-proBNP being higher than for BNP (NT-proBNP range 201–791 pg/ml vs.  BNP range 35–255 pg/ml).

We only included prospective observational studies. All studies were conducted in a blinded fashion, except one in which the BNP values were known to the clinicians treating the patients: this led to cancellation of surgery in 2% of patients.17Furthermore, systematic screening for asymptomatic postoperative cardiac events was not carried out, which may have led to bias in this study. Otherwise, there was no evidence of selective reporting of data or systematic bias in the other studies.

A funnel plot was used to assess for publication bias. This is a scatter plot of treatment effect (OR) on the horizontal axis against study size (variance) on the vertical axis. Publication bias (i.e. , exclusion of negative studies from the literature) or systematic bias is suggested when the scatter plot is unevenly distributed within the 95% CI funnel. No evidence of publication bias was found (fig. 2).

Data for short-term MACE was available for 10 studies comprising 11 patient cohorts (table 1). MACE occurred in 283 of 862 patients (32.83%; 95% CI 29.78–36.04) with elevated BNPs compared to 119 of 3,049 patients (3.9%; 95% CI 3.27–4.65) that did not (OR 19.77; 95% CI 13.18–29.65; P < 0.0001) (fig. 3). There was no significant heterogeneity (P = 0.13; I²= 30%). Sensitivity analysis did not significantly alter the result. The association was consistent for BNP (OR 25.45; 95% CI 12.46–51.97; P < 0.00001) and NT-proBNP (OR 15.65; 95% CI 10.39–25.37; P < 0.00001). We sought to determine whether study differences in cutoff values for BNP and NT-proBNP were associated with the outcome because higher cut offs might be related to worse outcomes. No significant association was found using meta-regression (P = 0.2182).

Only two studies reported longer-term data: MACE occurred in 76 of 202 patients (37.62%; 95% CI 31.23–44.49) with elevated NT-proBNP compared to 18 of 351 who did not have NT-proBNP elevation (OR 17.70; 95% CI 3.11–100.80; P < 0.00001) (fig. 4).

Short-term all-cause mortality data were available for five studies (table 1). Overall, 26 of 221 patients (11.76%; 95% CI 8.17–16.69) with elevated BNPs died compared to 4 of 494 (0.81%; 95% CI 0.3–32.1) who did not. Elevated BNPs were associated with an increased risk of death (OR 9.28; 95% CI 3.51–24.56; P < 0.0001) with no evidence of heterogeneity (P = 0.49; I²= 0%) (fig. 5).

Longer-term all-cause mortality data were available for four studies (table 2). Of 294 patients with elevated BNPs, 71 (24.15%; 95% CI 19.61–29.36) died compared to 31 of 503 (6.16%; 95% CI 4.38–8.62) who had normal BNP levels. Elevated BNPs were associated with an increased risk of all-cause mortality (OR 4.72; 95% CI 2.99–7.46; P < 0.00001) with no evidence of heterogeneity (P = 0.45, I²= 0%) (fig. 6). Serial exclusion of each study did not significantly alter the result.

Short-term cardiac mortality data were available for six studies comprising seven patient cohorts. Of 482 patients with elevated BNPs, 45 (9.34%; 95% CI 7.06–12.27) suffered cardiac death compared to 3 of 1,905 (0.16%; 95% CI 0.06–0.46) who did not. Elevated BNPs were associated with an increased risk of cardiac death (OR 23.88; 95% CI 9.43–60.43; P < 0.00001) with no evidence of heterogeneity (P = 0.44, I²= 0%) (fig. 7). Sensitivity analysis by sequential exclusion of individual studies yielded similar results. The association was consistent for both BNP (OR 30.17; 95% CI 11.16–81.53) and NT-proBNP (OR 39.07; 95% CI 11.63–131.25) (fig. 5). No data were available for longer-term cardiac mortality.

The main conclusions of our meta-analysis are that preoperative plasma BNP elevation is associated with a greatly increased risk of short-term MACE, cardiac mortality, and all-cause mortality, as well as longer-term MACE and all cause mortality after major noncardiac surgery. These findings are consistent for both BNP and NT-proBNP. Although these conclusions echo those of individual studies, this meta-analysis pools data from over 4,000 patients and, therefore, provides a greater insight into the strength of the associations. Furthermore, the analysis of mortality data in this context is novel. As such, measurement of BNPs may be a simple method of risk stratifying patients before noncardiac surgery.

To put the risk of adverse events in context, the relative risk of short-term all cause mortality in patients with elevated BNPs in this study is 6.9, which is similar to studies of myocardial infarction.27Likewise, the relative risk of longer-term all-cause mortality is 3.8, which is concordant with studies of stable heart failure or coronary artery disease.28,29This suggests that our estimates of risk are plausible.

The precise reason for the strong association between BNPs and postoperative outcome is not addressed by this analysis, but it is likely that BNP elevation identifies patients with impaired cardiac function or a significant ischemic burden who may not be able to withstand the hemodynamic and pro-inflammatory stresses of general anesthesia and major surgery. This is consistent with our observation that preoperative BNP measurement predicted short-term cardiac death more strongly than all-cause mortality.

An important question is whether BNP measurement has any incremental value over existing methods of preoperative assessment. Some studies have compared BNP and NT-proBNP to clinical risk scores, echocardiography and dobutamine stress echocardiography. However, insufficient data are available to allow comparison to be made in this analysis.

A limitation of this study is that the optimal cutoff values for plasma BNP and NT-proBNP cannot be defined because there was wide variation between studies. In the absence of definitive cutoffs for use in the noncardiac surgery setting, it may be reasonable to apply standard thresholds defined from heart failure studies. Generally accepted normal values for BNP (less than 100 pg/ml) and NT-Pro-BNP (less than 300 pg/ml)5would have a negative predictive value of at least 95% for short-term MACE if applied to these studies as a whole, and more than 99% excluding two studies that had particularly low cutoffs.16,26In other words, preoperative BNP measurement could be used to rule out the need for further cardiac investigation, unless there were other clinical grounds to strongly suspect a significant cardiac disorder. On the other hand, a raised BNP level should prompt further investigation to assess left ventricular function and or inducible ischemia. Unfortunately, none of the studies included in this analysis actually addressed the issue of how to use BNP levels in practice, so it remains to be proven whether triage of patients by this method can improve patient outcomes, and if so, how best to do this.

Certain factors may have affected the results of our analyses. Although we sought to minimize the possibility of publication bias by using a comprehensive search strategy, it is possible that this may be present because we did not identify any negative studies. There were differences between studies in terms of population characteristics that might affect BNP levels as well as outcomes, such as left ventricular function, age, gender, medication, renal function, obesity, and hemodynamic stress.5,30–32Adjusted odds ratios were reported by some studies for MACE, but mortality data were not, so we chose to use unadjusted outcome data for our analyses because this was available for all studies. It is possible that confounders may have influenced postoperative outcomes, although the magnitude of this effect is likely to be small because the adjusted and unadjusted OR were very similar where these were reported for individual studies.16,17,21Furthermore, in the absence of specific cutoffs for mortality outcomes, we used cutoffs derived from MACE data. Although these may not be optimal for predicting mortality outcomes, studies that have examined MACE and death found similar cutoffs for both,22,26suggesting that this approach is valid. Finally, the use of study-specific cutoff points introduces bias favoring a positive result, which is reflected in the very low heterogeneity in each analysis. For these reasons, it is difficult to be certain whether a particular BNP or NT-proBNP level actually carries the estimated risk reported in this study. This uncertainty can only be overcome with patient-level data for every study. Nevertheless, there is remarkable consistency between studies, lending credence to the notion that BNP levels are predictive of postoperative mortality and MACE.

In conclusion, preoperative plasma brain natriuretic pepide elevation identifies a subgroup of patients at high risk of death and MACE from major noncardiac surgery; as such, it has promise as a preoperative risk stratification tool. Current data are insufficient to define the exact role of BNP and NT-proBNP testing in preoperative patients, and large adequately powered prospective studies are required to clarify the best assay, the optimal cutoff, and the incremental value compared to standard preoperative assessment tools.