Effect of Body Mass Index on Diagnostic and Prognostic Usefulness of Amino-Terminal Pro–Brain Natriuretic Peptide in Patients With Acute Dyspnea FREE

The prevalence of obesity has reached unprecedented proportions. Nearly 40 million Americans are obese, and 300 000 deaths are attributed to obesity annually in the United States.¹ Obesity is associated with an increased risk for a broad range of fatal and nonfatal cardiovascular events, including the development of heart failure (HF),² which can be a challenging diagnosis to secure in an obese patient. Clinical examination, electrocardiography, and chest radiography lack sensitivity and specificity in these patients, and an accurate diagnosis may only be established with the help of other tests, such as echocardiography, magnetic resonance imaging, or measurement of natriuretic peptide levels. For the clinical diagnosis of acute HF, use of tests for plasma levels of brain natriuretic peptide (BNP) and amino (N)–terminal pro-BNP (NT-proBNP) are well established^3- 4 and represent a potentially attractive option for the diagnostic evaluation of the obese patient with dyspnea. However, for reasons that remain unexplained, plasma levels of these cardiac peptides appear to be inversely associated with body mass index (BMI) in subjects with and without HF,^5- 6 possibly owing to suppression of synthesis or release of natriuretic peptides from cardiomyocytes in obese subjects.^7- 8

In the setting of acute breathlessness and a high BMI in the emergency department, a correct and prompt diagnosis may be even more relevant for immediate triage and therapy. In a previous analysis, the age-related NT-proBNP cut points of 450, 900, and 1800 ng/L for patients younger than 50, aged 50 to 75, and older than 75 years were found to be optimal to identify acute HF, and the age-independent cut point of 300 ng/L was valuable to exclude acute HF.⁴ Concentrations of NT-proBNP were also powerfully prognostic during the first 2 months after presentation among those with acute HF. Recent results demonstrated that the prognostic implications of elevated NT-proBNP concentrations in dyspneic patients extend to at least a full year after presentation.⁹ However, given the inverse relationship between BMI and NT-proBNP,¹⁰ it remains unclear whether BMI affects the utility of the NT-proBNP concentration for diagnosis and prognosis in acutely dyspneic patients. To answer these questions, we evaluated in a prospective population of patients with acute dyspnea (1) whether previously defined NT-proBNP diagnostic cut points retain their diagnostic utility across the clinical strata of normal, overweight, and obese patients and (2) whether the association between plasma NT-proBNP concentration and long-term mortality is preserved across these weight categories.

Information regarding the International Collaborative of NT-proBNP (ICON) Study, which included participation of 4 centers worldwide (Christchurch, New Zealand; Barcelona, Spain; Boston, Mass; and Maastricht, the Netherlands), has been previously published.⁴ All data sources had compatible inclusion and exclusion criteria and obtained similar clinical information, including standard demographics, medical history, drug therapy, presenting symptoms and signs, results of physical examination and laboratory testing, chest radiography information, and electrocardiography results; and similar criteria were used for diagnosing acute HF.⁴ In brief, HF in the Christchurch cohort was diagnosed according to the European Society of Cardiology guidelines for HF by study physicians blinded to the NT-proBNP concentrations. For patients from the Barcelona cohort, the diagnosis of HF was assigned by a panel of physicians using all available clinical data pertaining to each subject. For patients in the Boston cohort, the final diagnosis for each patient was assigned by study cardiologists blinded to the NT-proBNP concentrations, using available clinical data from presentation to the 60-day follow-up visit. For patients in the Maastricht cohort, all had acute HF and dyspnea of varying severities.

Of the total 1256 patients in the ICON study with acute dyspnea presenting to the emergency department, BMI was available in 1103 subjects; of these, 583 subjects had acute HF, and the rest had dyspnea of noncardiac origin. The final diagnosis for each patient was assigned by study cardiologists blinded to NT-proBNP concentrations using all available clinical data pertaining to each subject.

Follow-up for vital status was completed in all subjects through 1 year after presentation to the emergency department (baseline). Vital status was ascertained via review of medical records and contact with the patients or their caregivers and was complete in all subjects.

Blood was collected into EDTA tubes and processed immediately, and NT-proBNP concentrations were determined by an immunoelectrochemiluminisence method (Elecsys; Roche Diagnostics, Basel, Switzerland). This assay has less than 0.001% crossreactivity with bioactive BNP and, in the constituent studies in this report, the assay had interrun coefficients of variation ranging from 0.9% to 5.5%.⁴

Weight and height were measured during admission, and BMI was calculated as weight in kilograms divided by height in meters squared. Subjects in the ICON study were stratified according to baseline BMI into the following 3 groups: lean (BMI, <25.0), overweight (BMI, 25.0-29.9), and obese (BMI, ≥30.0).¹¹ Because the death rates and associations with NT-proBNP concentrations were nearly identical for those with a BMI lower than 18.5 and those with a BMI of 18.5 to 24.9, and given the relatively few individuals with a BMI lower than 18.5 (n = 48), we collapsed these 2 groups into 1 category of a BMI lower than 25.0.

Correlations between BMI and NT-proBNP values were performed first by log-transforming both variables, followed by the Spearman correlation. Multivariate linear regression analysis was performed to address the independent association of log-transformed NT-proBNP concentrations with other factors. Comparisons of NT-proBNP concentrations in varying weight categories were performed using the Wilcoxon rank sum test. The utility of the NT-proBNP concentration for the diagnosis of acute HF was assessed using receiver operating characteristic curves; the area under each curve with 95% confidence intervals was calculated in each weight category and compared with those of the other weight categories using z testing. The value of diagnostic (rule in) cut points (of 450, 900, and 1800 ng/L for patients younger than 50, aged 50-75, and older than 75 years, respectively) as previously described⁴ were evaluated using positive likelihood ratios, whereas the utility of the previously elucidated⁴ age-independent NT-proBNP cut point for excluding acute HF (rule out) was evaluated using negative likelihood ratios.

To identify independent predictors of death at 1 year after presentation with dyspnea, age- and sex-adjusted Cox proportional hazards models were performed, with 1-year mortality as the dependent variable. Because an NT-proBNP cut point of 986 ng/L was previously suggested to be optimal for stratifying risk in patients with dyspnea,⁹ this cut point was used in the multivariate model. Given a possible inverse relationship between BMI and survival in those with and without acute HF, BMI was included in the Cox analyses as a categorical variable (ie, <25.0, 25.0-29.9, and ≥30.0) and as a continuous variable in separate models. Significant predictors of mortality at 1 year are expressed with hazards ratios and the 95% confidence interval.

Kaplan-Meier survival curves were plotted and the groups were compared using the log-rank test. Statistical analyses were performed with the use of SPSS (SPSS Inc, Chicago, Ill) and Stata (version 8SE; StataCorp, College Station, Tex) statistical software, whereas receiver operating characteristic curve analyses were performed using Analyse-it software (Analyse-it Ltd, Leeds, England). Data are presented as medians with interquartile ranges for nonnormally distributed variables and as mean ± SD for all other continuous variables. A 2-sided P<.05 was considered significant.

Table 1 shows the baseline characteristics for the study population stratified by BMI category. A BMI lower than 25.0 was found in 37.4% of the studied population; of 25.0 to 29.9, in 34.0%; and 30.0 or higher, in 28.6%. A higher BMI was associated with younger age at presentation, more prevalent hypertension or diabetes mellitus, and less prior ischemic heart disease or HF. Patients in the highest BMI group were also more likely to have a higher left ventricular ejection fraction, higher blood glucose levels, and lower troponin T levels. Similarly, patients with a BMI of 30.0 or higher were more likely to have lower NT-proBNP concentrations.

In Spearman rank correlation, we found an inverse association between log-transformed NT-proBNP values and BMI for patients with (r = −0.27; P<.001) and without (r = −0.10; P=.02) acute HF. Indeed, linear regression analysis of the variables associated with log-transformed NT-proBNP values at presentation demonstrated that BMI is an independent inverse determinant of NT-proBNP values (Table 2).

Figure 1 shows NT-proBNP concentrations by diagnosis, separately for each BMI group. Within each of the 3 BMI strata, NT-proBNP concentrations were significantly higher in patients with HF compared with those without HF (all P<.001), as expected.

Among patients with acute HF, the NT-proBNP concentrations were lower in overweight and obese patients. The difference in NT-proBNP concentrations between lean and overweight patients with acute HF wasstatistically significant (P = .005), whereas the difference between concentrations in overweight and obese patients was not significant. Similarly, among patients without acute HF, NT-proBNP concentrations were also significantly lower in overweight and obese patients. In this group, the differences in NT-proBNP values between lean, overweight, and obese patients were all significant (all P<.05; Figure 1).

To examine whether the lower NT-proBNP concentration found in obese patients affected the diagnostic value of the marker in those with acute dyspnea, diagnostic receiver operating characteristic curves for NT-proBNP values were plotted, stratified by BMI. As shown in Figure 2, the NT-proBNP concentration remained highly diagnostic for acute HF across the 3 BMI strata (all areas under the curve, ≥0.94; each P<.001).

Table 3 shows the utility of previously defined age-adjusted diagnostic NT-proBNP cut points⁴ as a function of BMI. The positive likelihood ratios for the age-adjusted rule-in cut points were 5.3 for a BMI lower than 25.0, 13.3 for a BMI of 25.0 to 29.9, and 7.5 for a BMI of 30.0 or higher, all indicating a moderate to high or conclusive likelihood for disease. The negative likelihood ratio for an NT-pro-BNP concentration of 300 ng/L was highly significant in all 3 BMI strata (0.02, 0.03, and 0.08 for the respective increasing BMI categories; Table 3), in each case indicating a large and conclusive decrease in the likelihood of disease below this cut point, irrespective of BMI.

We then evaluated the effects of BMI on the frequency of an intermediate natriuretic peptide concentration that, in the context of NT-proBNP testing, would be between the rule-out concentration of 300 pg/mL and the respective age-adjusted rule-in concentration.^12- 13 Overall, an intermediate concentration for NT-proBNP was observed in 202 (18.3%) of our subjects; of these, 105 (9.5%) had acute HF, whereas 97 (8.8%) did not. The median NT-proBNP concentration among these subjects was 727 ng/L; using this cut point, we observed a positive likelihood ratio of 2.34 in this range, with a corresponding negative likelihood ratio of 0.44. However, using the recommended age-stratified cut point strategy⁴ across BMI categories, the rates of intermediate NT-proBNP concentrations were 17.2%, 20.0%, and 17.7% for those BMIs lower than 25.0, 25.0 to 29.9, and 30.0 or higher, respectively.

During the 1-year follow-up, 211 patients died; the crude rates of death among dyspneic subjects were 24.6% among those with a BMI lower than 25.0, 17.6% for those with a BMI of 25.0 to 29.9, and 13.9% for those with a BMI of 30.0 or higher (P<.001 for trend). Similar trends across BMI strata were observed in those with and without acute HF.

Figure 3 illustrates NT-proBNP concentrations as a function of mortality status in the 3 BMI categories. The NT-proBNP concentration was significantly higher among decedents in all 3 BMI categories (P<.001). Likewise, among survivors and deceased patients, the NT-proBNP concentrations were significantly lower in overweight and obese patients (P<.001).

A previously published NT-proBNP prognostic cut point of 986 ng/L⁹ was applied to our population to validate its utility. In Cox proportional hazards analyses adjusted for age, sex, and BMI, an NT-proBNP cut point of greater than 986 ng/L remained strongly prognostic across all 3 BMI categories (Table 4). Despite the effects of increasing BMI on the NT-proBNP concentrations, we observed no statistically significant interaction between BMI category and NT-proBNP concentration in prediction of 1-year mortality. These results were nearly identical whether we included or excluded those with a BMI lower than 18.5 in the lowest BMI category. The hazard associated with elevated NT-proBNP concentrations remained largely the same across the BMI strata when examined as a function of the presence or absence of a diagnosis of acute HF.

In univariate analyses, higher BMI was an independent predictor of survival, with a 4% reduction in the risk of death with each increase of 1 BMI unit (95% confidence interval, 0.94-0.99; P=.002). However, a higher BMI was not significantly associated with 1-year mortality once age was added to the model.

Kaplan-Meier curves (Figure 4) demonstrate that the risk associated with elevated NT-proBNP concentrations in patients with or without acute HF was present early and sustained in all 3 BMI categories to a full year after presentation (log-rank P<.001 for all comparisons).

The finding of lower circulating natriuretic peptide levels in overweight and obese patients has cast doubt about the utility of these novel biomarkers in patients with a high BMI, a population of patients in whom evaluation is often challenging owing to the effects of their weight on clinical histories and particularly physical examination findings. This study examined the influence of BMI on the utility of the NT-proBNP assay for diagnosis and prognosis of dyspneic patients attending the emergency department. Our data suggest that, despite the fact that NT-proBNP concentrations were relatively lower in overweight and obese patients with dyspnea, NT-proBNP concentrations remained useful to diagnose or exclude acute HF, and elevated concentrations of NT-proBNP strongly predicted higher mortality at all levels of BMI.

Reports from previous investigators found lower natriuretic peptide levels with higher BMI in subjects with chronic established HF or with acute HF and in those without HF.^{5,8,10,14- 15} Our study confirms that overweight and obese dyspneic patients with acute HF (BMI ≥25.0) have significantly lower NT-proBNP concentrations compared with lean patients (BMI <25.0) with HF. Indeed, linear regression analysis showed that BMI was inversely associated with NT-proBNP concentrations, even after adjustment for all other significant covariates. Likewise, NT-proBNP concentrations were lower in overweight and obese dyspneic patients without acute HF, although in patients without acute HF, median NT-proBNP concentrations were typically below the established rule-out cut point of 300 ng/L⁴ in all BMI categories, emphasizing the utility of this cut point in patients of all weights.

More importantly, this study confirms the usefulness of the proposed age-specific NT-proBNP cut points for diagnosing HF in patients with acute breathlessness. The previously recommended optimal cut points for identifying acute HF (450, 900, and 1800 ng/L in subjects younger than 50, aged 50-75, and older than 75 years, respectively)⁴ performed with strong diagnostic accuracy and specificity in this population of subjects with a wide range of BMIs. Despite slightly lower sensitivity at higher BMIs, these cut points demonstrated similar positive predictive values across all BMI strata, including the most obese subjects. The age-independent cut point of 300 ng/L for excluding acute HF also performed well across all BMI strata. The use of these cut points allows for dual functionality of NT-proBNP testing, that is, confident identification (rule in) of acute HF when the NT-proBNP concentration is above the age-adjusted rule-in cut points, as well as confident exclusion (rule out) of the diagnosis when the concentration is below 300 ng/L, despite BMI. Although it would be desirable to have a single cut point for both diagnosis and exclusion of acute HF, the recognition that natriuretic peptides exist as continuous variables with considerable overlap between those with and without acute HF (owing to other diagnoses that lead to elevation of BNP and NT-proBNP concentrations, such as atrial fibrillation, pulmonary embolism, or acute coronary ischemia) renders such a dual approach necessary for both BNP¹² and NT-proBNP concentrations.¹³ The intermediate (or gray-zone) NT-proBNP values between the cut points optimal for excluding or confirming the disease were recently shown to be of importance in this cohort,¹³ imparting an intermediate risk for short-term mortality irrespective of diagnosis. Thus, although we have defined useful discrete NT-proBNP cut points for ease of everyday use, we emphasize the importance of considering NT-proBNP concentration most appropriately as a continuous rather than a dichotomous variable. For the physician using NT-proBNP concentrations to evaluate acute dyspnea, it is worthwhile to emphasize that increasing BMI did not appear to appreciably increase the likelihood of an intermediate NT-proBNP concentration; nonetheless, when this circumstance occurs, we recently reported that the use of factors from clinical history and physical examination are most often useful to ascertain the correct diagnosis.¹³

In addition to assisting in emergency department diagnosis and triage, NT-proBNP concentrations at presentation are predictive of long-term mortality in patients with dyspnea. A recent report found that an NT-proBNP concentration above 986 ng/L predicted 1-year mortality after emergency department presentation with dyspnea.⁹ In our series, NT-proBNP concentrations in deceased patients during the 1-year follow-up were significantly higher in lean patients than in overweight or obese patients, but the prognostic NT-proBNP cut point of greater than 986 ng/L remained strongly prognostic at all BMI strata. The finding is potentially important for therapeutic decision making, such as more intensive diagnostic and therapeutic maneuvers. Indeed, several studies are ongoing to test whether treatment of HF based on NT-proBNP concentrations appears to reduce cardiovascular events compared with clinically guided treatment, as shown by the pilot study of Troughton et al.¹⁶

The link between obesity and low natriuretic peptide levels is not yet fully elucidated. Some authors postulate that the inverse relationship observed between BMI and the BNP concentration may be due to increased expression of the natriuretic peptide clearance receptor by adipose tissue, resulting in increased clearance of BNP in obese subjects.^17- 18 However, the association between higher BMI and lower NT-proBNP levels suggests that nonclearance mechanisms must be considered, because NT-proBNP is an inactive peptide unlikely to be actively cleared by specific receptors. More recently, it has been suggested that decreased release of natriuretic peptides from the heart, rather than increased clearance, may be responsible for the association between higher BMI and lower natriuretic peptide levels.^7- 8,19 A novel lipolytic and potential lipid-mobilization effect of these peptides has been identified, thus suggesting that decreasedcirculating natriuretic peptide levels may not be a consequence of excess adiposity, but rather may be a causative factor that leads to the development of obesity. In sum, reduced release likely explains the lower NT-proBNP concentrations in overweight and obese subjects, whereas reduced release and possible increased clearance may affect BNP values. In a previous study, we reported NT-proBNP concentration to be more sensitive than BNP in overweight subjects (but not obese subjects¹⁰), a finding that may be explicated on the basis of reduced release and increased clearance of BNP relative to reduced release of NT-proBNP.

Although obesity is clearly associated with cardiovascular disease,² recent studies have focused on an apparent paradox regarding the relationship between obesity and subsequent cardiovascular prognosis. Several studies^20- 22 have found a strong inverse relationship between indices of obesity (BMI) and subsequent clinical prognosis in patients with HF. In our study of patients presenting to the emergency department with acute dyspnea, we also found that a higher BMI was associated with lower rates of death in those with and without acute HF. Although BMI was significantly and inversely associated with survival in the univariate analysis, this association was attenuated with the addition of age to the model. Thus, although cachexia and wasting may be independent predictors of increased mortality in advanced HF,^23- 24 our data support the contention of Lavie and Milani²⁵ that this apparent paradox in HF represents an association that is unlikely to be causal.

With the knowledge that a higher BMI affects both BNP and NT-proBNP concentrations, the physician is left with the uncertainty as to whether obesity would affect the prognostic power of these markers. Collectively, our data (derived from a large multinational pooling of patients with and without acute HF) are reassuring in that, across wide BMI strata, NT-proBNP concentrations were well able to identify or exclude acute HF in most of the patients, as well as identify those at highest risk for death.

Our study has limitations worth mention. Symptoms of HF in obese patients may be due to a combination of milder forms of HF together with obesity-related deconditioning or restrictive lung disease. In this study we did not conduct pulmonary function tests to rule out additional lung disease. However, our data indicate that the NT-proBNP concentration retains its diagnostic and prognostic capacity across all BMI categories, no matter what is the cause of breathlessness in a patient attending the emergency department. Also, those subjects in the lowest BMI category were more likely to have more incident myocardial infarction and lower left ventricular ejection fractions compared with those with a higher BMI. Thus, the inverse relationship between BMI and NT-proBNP concentrations could simply reflect less prevalent ischemic heart disease or left ventricular dysfunction in the heaviest subjects. This is unlikely, as several analyses have now shown that this inverse relationship between BMI and natriuretic peptide concentrations is independent of ischemic heart disease and ventricular function.^10,17 Finally, although we previously showed improved diagnostic utility for NT-proBNP concentration by using an age-dependent stratification schema,⁴ we analyzed prognostic cut points and used an age-independent approach. Age was an independent variable in the model for mortality at 1 year in the study from which the NT-proBNP cut point of 986 ng/L was identified⁹; thus, it is tempting to speculate whether age stratification might sharpen the prognostic value of NT-proBNP concentrations and could be the focus of future analyses.

Correspondence: James L. Januzzi, Jr, MD, Department of Cardiology, Massachusetts General Hospital, Yawkey 5984, 55 Fruit St, Boston, MA 02114 (jjanuzzi@partners.org).

Accepted for Publication: November 16, 2006.

Author Contributions: Dr Januzzi had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Bayes-Genis, Lloyd-Jones, Lainchbury, Richards, Ordoñez-Llanos, Santaló, and Januzzi. Acquisition of data: Bayes-Genis, van Kimmenade, Lainchbury, Richards, Ordoñez-Llanos, Santaló, Pinto, and Januzzi. Analysis and interpretation of data: Bayes-Genis, Lloyd-Jones, van Kimmenade, Lainchbury, Richards, Ordoñez-Llanos, Pinto, and Januzzi. Drafting of the manuscript: Bayes-Genis, Lloyd-Jones, van Kimmenade, Pinto, and Januzzi. Critical revision of the manuscript for important intellectual content: Bayes-Genis, Lloyd-Jones, van Kimmenade, Lainchbury, Richards, Ordoñez-Llanos, Santaló, Pinto, and Januzzi. Statistical analysis: Bayes-Genis, Lloyd-Jones, and Januzzi. Administrative, technical, and material support: van Kimmenade, Lainchbury, Richards, Ordoñez-Llanos, Santaló, Pinto, and Januzzi. Study supervision: Richards, Ordoñez-Llanos, and Januzzi.

Financial Disclosure: Drs Bayes-Genis, van Kimmenade, Richards, Ordoñez-Llanos, Santaló, Pinto, and Januzzi have received speaking and research grants from Roche Diagnostics; Drs Richards, Ordoñez-Llanos, and Januzzi, consulting grants from Roche Diagnostics; and Dr Januzzi, speaking, consulting, and research grants from Dade-Behring and Ortho Diagnostics.