N-Terminal Pro-Brain Natriuretic Peptide, C-Reactive Protein, and Urinary Albumin Levels as Predictors of Mortality and Cardiovascular Events in Older Adults FREE

Risk stratification in the general population is receiving increasing interest.^1- 3 In particular, C-reactive protein (CRP) has emerged as a possible potent risk marker for cardiovascular disease.^4- 5 Measurements of CRP are now considered to improve risk stratification beyond traditional cardiovascular risk factors in the general population.⁶ The B-type natriuretic peptides have also received increasing interest as potential risk markers. Brain natriuretic peptide (BNP) and the N-amino terminal fragment of the prohormone BNP (NT-proBNP)⁷ are released predominantly from the ventricular myocardium in response to increased ventricular wall stress.⁸ Plasma levels of these peptides are substantially increased in conditions like chronic heart failure^9- 10 and during acute coronary syndromes.¹¹

The prognostic value of BNPs has mainly been established in patients with heart failure¹² and in populations covering the range of acute coronary syndromes.^13- 15 In nonhospitalized individuals, a limited number of studies haveevaluated the prognostic information of the B-type natriuretic peptides.^16- 18 However, one study only included older participants,¹⁸ and no studies have compared the predictive value of BNPs with established cardiovascular risk markers. These studies and data from the Framingham Heart study¹⁹ suggest that the B-type natriuretic peptides can provide important prognostic information beyond traditional cardiovascular risk factors in the general population.

The objective of our study was to investigate the prognostic value of plasma NT-proBNP compared with that of CRP and urinary albumin levels on overall mortality and cardiovascular events in older people from a community in Copenhagen, Denmark.

The participants in this study were recruited from the general population in the municipality of Frederiksberg, in Copenhagen, Denmark. The study population has previously been demonstrated to be representative of the population in the urban community of Copenhagen.^20- 21 A random sample was collected of 1088 persons from the local community who were invited to participate in the study. As we were interested in analyzing a representative sample of high-risk individuals, enrollees had to be aged 50 to 89 years. Exclusion criteria were participants’ inability to cooperate and lack of response to 2 written invitations. A total of 764 individuals (70.2%) responded to the invitations.²⁰ Cardiovascular event rates 1 year before and 1 year after baseline did not differ between those individuals who did not respond and those who did, with age- and sex-adjusted relative risks (RRs) of 1.07 (95% confidence interval [CI], 0.72-1.59; P=.75) and 1.20 (95% CI, 0.76-1.90; P=.43), respectively. However, individuals who did not respond had a higher 1-year mortality rate; those who responded had an age- and sex-adjusted RR of 0.32 (95% CI, 0.16-0.62; P<.001). The mortality of the total invited sample was higher than the background population, with an age- and sex-adjusted RR of 1.56 (95% CI, 1.10-2.23; P=.01). The 1-year mortality rate among those who responded was similar to the background population (RR, 0.98; 95% CI, 0.84-1.14; P=.78).²⁰ Plasma NT-proBNP, CRP, and urinary albumin concentration were measured in 658 participants. Among these, 27 participants were excluded because of prevalent heart failure, 3 because of renal insufficiency (defined as serum creatinine of at least 2.26 mg/dL [≥200 μmol/L]), and 2 because of lack of follow-up data. A total of 626 individuals were finally included in the study (Figure 1).

The population was examined between September 1, 1998, and January 24, 2000. All participants completed a questionnaire providing information on current symptoms and on demographic, behavioral, and lifestyle factors. An extensive medical and drug history was obtained by study physicians, including information on previous hospital admissions. Prevalent cardiovascular disease was defined as previous hospital admission for the primary diagnosis of myocardial infarction, unstable angina pectoris, stroke or transient ischemic attack, or current symptoms of angina pectoris. All participants underwent echocardiography assessing presence of left ventricular hypertrophy and left ventricular ejection fraction (LVEF), which was evaluated off-line in a blinded fashion by 2 experienced and independent observers. The participants were followed up until December 31, 2003, with respect to mortality status and development of cardiovascular events. The median (range) follow-up period was 5.0 (0.17-5.28) years. The study was approved by the central ethics committee of Copenhagen, and all participants provided written informed consent.

Plasma concentration of NT-proBNP was measured using a highly sensitive and specific immunoassay based on double-antibody sandwich technique (Roche Diagnostics, Mannheim, Germany). The intra-assay and interassay coefficients of variation were 1.3% and 4.8%, respectively.²² C-reactive protein was measured with a highly sensitive, latex-particle–enhanced immunoassay (Roche Diagnostics), with a measuring range of 0.1 to 300 mg/L and a lower detection limit of 0.03 mg/L. The intra-assay and interassay coefficients of variation were 1.3% and 6.5%, respectively, for values less than 4 mg/L.²³ Urine samples were collected as first morning–voided urine. Urinary albumin concentration was measured by immunoturbidimetry on a Cobas Bioanalyzer (Roche Products, Basel, Switzerland). Lower detection limit was 1 mg/L and the coefficient of variation was less than 4%. Urinary albumin excretion was determined as the urinary albumin/creatinine ratio; upper limit of normal range was 30 mg/g.

Participants were monitored with respect to mortality status and first major cardiovascular events on a regular basis after enrollment. Major cardiovascular events were recorded by the discharge registry of the Danish National Board of Health, which records all primary hospital discharge diagnoses in Denmark. The register has been described previously.²⁴ All cardiovascular events required hospitalization to meet the outcome definition. The codes of diagnosis for the cardiovascular events were prespecified. Codes were assigned according to the International Classification of Diseases, 10th Revision (ICD-10). All deaths were confirmed by the Danish Personal Register, which records all deaths in Denmark within 2 weeks. Deaths from cardiovascular disease were ascertained from central registers in the Danish National Board of Health and verified by study physicians from death certificates. The members of the Danish National Board of Health and the study physicians were blinded with respect to data on the biomarkers. Major cardiovascular events were analyzed as a combined end point, including nonfatal myocardial infarction, fatal coronary heart disease (CHD), unstable angina pectoris, heart failure, stroke, and transient ischemic attack, defined as hospitalization with the following ICD-10 codes: I20.0-I22, I24, I42.0, I46, I50, I63, I65, and I66. The major cardiovascular events were subsequently analyzed individually as heart failure, stroke, or transient ischemic attack, and as CHD, which included myocardial infarction and unstable angina pectoris.

The participants were divided into tertiles according to their baseline NT-proBNP levels. Comparisons between the groups were performed by 1-way analysis of variance or Kruskal-Wallis test for the continuous variables, according to whether distribution of variables was Gaussian. The χ² test was used for categorical data. Baseline NT-proBNP, CRP, and urinary albumin/creatinine levels were highly skewed and therefore logarithmically transformed in all analyses. Levels of NT-proBNP, CRP, and urinary albumin/creatinine ratio were compared for individuals who died (n = 94) vs those who were alive at the time of follow-up (n = 532), using the Mann-Whitney test. Subsequently, the cumulative survival according to increasing tertiles of NT-proBNP, CRP, and urinary albumin/creatinine ratio was estimated by Kaplan-Meier curves, followed by a trend test. Using Cox proportional hazards regression models, hazard ratios (HRs) for NT-proBNP, CRP, and urinary albumin/creatinine ratio with each outcome were assessed. To evaluate the association between the biomarkers and outcomes, these were considered both as continuous (logarithmically transformed) and as categorical variables. We used the prespecified cut points corresponding to the 80th percentile of each biomarker in the categorical analysis. This cut point was used to facilitate comparisons with previous studies.^4,19,25- 27 Analyses of first major cardiovascular event, stroke or transient ischemic attack, and CHD were restricted to participants without prevalent cardiovascular disease; 537 participants were included in these analyses. Analyses of development of heart failure were restricted to participants with LVEF of more than 50% at baseline, and this subpopulation comprised 598 individuals (Figure 1). Hazard ratios and 95% CIs were calculated in unadjusted analyses, after adjustment for age and sex, as well as in multivariable models with traditional cardiovascular risk factors (age, sex, smoking status, diabetes mellitus, hypertension, serum total cholesterol, and serum creatinine) and variables reflecting severity of cardiovascular disease. In the analyses of overall mortality, additional adjustment was made for presence of ischemic heart disease. In a second analysis, we further adjusted the mortality model for intervening cardiovascular events (incident nonfatal myocardial infarction, unstable angina pectoris, stroke, or transient ischemic attack) to determine whether NT-proBNP was predictive of mortality independent of increased risk for cardiovascular events. Secondary analyses of mortality included adjustments for baseline systolic blood pressure, heart rate, and cardiovascular medication instead of history of hypertension. Analyses of first major cardiovascular event and stroke also included adjustment for atrial fibrillation. Furthermore, all outcome measures were analyzed with additional adjustment for left ventricular systolic dysfunction (LVEF <50%) and left ventricular hypertrophy.

For the Cox proportional hazards regression model analyses, deviation from linearity was tested by comparing models with nonlinear transformations of the covariates (square root terms) with models containing linear terms of the covariates, using likelihood ratio test. The assumption of proportionality with regard to NT-proBNP, CRP, and urinary albumin/creatinine ratio was met. When testing specifically whether the association between NT-proBNP and mortality and first cardiovascular event varied according to age (>65 years), sex, renal function, hypertension, or LVEF, by entering interaction terms in the multivariable models, only interaction with age was found. Accordingly, interaction between age and log-transformed CRP was assessed. Inclusion of interaction terms with age did not significantly alter the results of the main multivariable models with NT-proBNP and CRP, respectively. All P values were 2-sided and P<.05 was considered statistically significant. The statistical software package SPSS version 11.5 (SPSS Inc, Chicago, Ill) was used for all analyses.

Mean (SD) age at baseline was 67.9 (10.6) years. Among the 626 participants, 361 (57.7%) were women, 72 (11.5%) had ischemic heart disease, 149 (23.8%) had a history of hypertension, and 36 (5.8%) had diabetes mellitus. Baseline clinical characteristics according to tertiles of baseline NT-proBNP levels are presented in Table 1. Participants in the highest tertiles were older, more often women, and more likely to have a history of hypertension. Higher levels of NT-proBNP were positively associated with urinary albumin/creatinine ratio (R = 0.33, P<.001) and with serum creatinine (R = 0.14, P<.001) and negatively associated with LVEF (R = −0.23, P<.001).

During the 5 years of follow-up, 94 (15%) of 626 participants died from any cause. A total of 65 (12%) of 537 participants without cardiovascular disease at baseline had a first major cardiovascular event. Of these participants, 23 had a stroke or transient ischemic attack, 17 developed heart failure, 12 had a CHD event, and 32 died of cardiovascular disease. A total of 20 (3%) of 598 patients with normal left ventricular systolic function (LVEF ≥50%) at baseline developed heart failure.

In individuals who died compared with those who were alive at follow-up, baseline median (interquartile range) plasma NT-proBNP was 505.5 (281.8-1138.3) vs 234.9 (134.9-461.5) pg/mL, respectively (P<.001); CRP was 2.98 (1.43-6.95) vs 2.24 (1.05-4.82) mg/L, respectively (P = .03); and urinary albumin/creatinine ratio was 14.0 (7.0-30.0) vs 6.0 (4.0-11.0) mg/g, respectively (P<.001). Mortality risk increased by increasing tertile of NT-proBNP (P for trend <.001) (Figure 2A), with an absolute increase in risk of 18.5% between the lowest and highest tertile. Mortality rates also increased by tertile of CRP (P for trend = .02) (Figure 2B), with an absolute increase in risk of 8.1% between the lowest and highest tertile. For urinary albumin/creatinine ratio, the absolute increase in risk between the lowest and highest tertile was 20.3% (Figure 2C).

Increased levels of NT-proBNP predicted total mortality even after adjusting for cardiovascular risk factors. Participants with NT-proBNP values above the 80th percentile had a 2-fold increase in mortality (HR, 1.96; 95% CI, 1.21-3.19; P = .007) relative to those with values equal to or below, and a 24.5% absolute increase in unadjusted risk. When NT-proBNP was analyzed as a continuous variable, increasing NT-proBNP also was associated with increased mortality, the adjusted HR being 1.43 per 1-SD increase in log-transformed peptide values (95% CI, 1.10-1.86; P = .008) (Table 2). Additional adjustment for left ventricular systolic dysfunction and left ventricular hypertrophy only moderately attenuated the association between NT-proBNP and death; the HR was 1.82 (95% CI, 1.11-2.98; P = .02) for values above the 80th percentile. After adjusting for intervening incident major cardiovascular events, high levels of NT-proBNP continued to predict mortality (HR, 1.69; 95% CI, 1.05-2.76; P = .03) for individuals with NT-proBNP levels above the 80th percentile.

In secondary analyses, systolic blood pressure, heart rate, and the use of cardiovascular medication were included in the model instead of the presence of hypertension. These analyses yielded results similar to those of the main model (HR, 2.65; 95% CI, 1.34-3.70; P<.001) for values above the 80th percentile.

Urinary albumin/creatinine ratio levels were independently associated with an increased risk of death, with an adjusted HR of 1.88 (95% CI, 1.18-2.98; P = .008) for values above the 80th percentile (Table 2). The absolute increase in unadjusted mortality risk was 19.5%. In contrast, increasing levels of CRP were significantly associated with an increased mortality risk only in the unadjusted model, and had an absolute unadjusted increase in mortality risk of 7.8% for values above the 80th percentile vs those equal to or below the 80th percentile. Thus, high levels of CRP did not significantly predict mortality after adjusting for cardiovascular risk factors; adjusted HR was 1.46 (95% CI, 0.89-2.24; P = .14) for values above the 80th percentile. When the biomarker was analyzed as a continuous variable, the results also were not significant (HR, 1.17; 95% CI, 0.95-1.43; P = .12).

A multivariable model including both NT-proBNP and urinary albumin/creatinine ratio and the clinical baseline variables of age, sex, diabetes mellitus, hypertension, ischemic heart disease, current smoking, serum cholesterol, and creatinine demonstrated that both markers were still independent predictors of mortality. Adjusted HRs for levels above the 80th percentile of NT-proBNP and urinary albumin/creatinine ratio were 1.98 (95% CI, 1.22-3.22; P = .006) and 1.90 (95% CI, 1.22-2.96; P = .004), respectively.

Table 3 shows the HRs for a first major cardiovascular event according to baseline levels of all 3 biomarkers. Plasma NT-proBNP and urinary albumin/creatinine ratio levels were significantly associated with the risk of a first major cardiovascular event. Participants with values of NT-proBNP above the 80th percentile had a 3.2-fold (95% CI, 1.80-5.79; P<.001) increase in risk; while for urinary albumin/creatinine ratio, the increase in risk was 2.3-fold (95% CI, 1.33-4.05; P = .003), after adjustment for cardiovascular risk factors, left ventricular systolic dysfunction, and left ventricular hypertrophy. Conversely, CRP levels did not predict an increased risk of a major cardiovascular event, independent of the covariables in the model. Plasma NT-proBNP and urinary albumin/creatinine ratio were still independent predictors of a first cardiovascular event when entered into the same multivariable model, with HRs for values of NT-proBNP and urinary albumin/creatinine ratio above the 80th percentile of 2.63 (95% CI, 1.60-4.32; P<.001) and 1.67 (95% CI, 1.05-2.65; P = .03), respectively.

Plasma levels of NT-proBNP predicted the risk of stroke or transient ischemic attack, with a 3.63-fold (95% CI, 1.15-11.45; P = .03) increased risk of stroke for participants with values above the 80th percentile vs those with values equal to or below the 80th percentile. Urinary albumin levels were also associated with increased risk of stroke, although after complete adjustment this relationship was no longer significant (HR, 2.63; 95% CI, 0.98-7.07; P = .06 for urinary albumin/creatinine ratio levels above the 80th percentile). No association between levels of CRP and ischemic stroke was found (Table 4).

In the 598 participants with normal left ventricular systolic function at baseline, plasma NT-proBNP levels predicted the development of heart failure (n = 19, missing covariate in 1 participant). The adjusted HR was 2.29 (95% CI, 1.29-3.98; P = .004) per 1-SD increment in log NT-proBNP value; the HR for values above the 80th percentile was 3.21 (95% CI, 1.13-9.08; P = .03). In contrast, there were no associations between CRP or urinary albumin/creatinine ratio and heart failure. In age- and sex-adjusted analyses, HRs were 1.24 (95% CI, 0.81-1.87; P = .33) and 1.38 (95% CI, 0.98-1.98; P = .06), respectively, per 1-SD increase in log CRP and log urinary albumin/creatinine ratio.

Only 12 major CHD events were recorded among those individuals initially free of cardiovascular disease, and none of the 3 biomarkers were significantly associated with an increased risk of a first CHD event. After adjustment for age and sex, individuals with NT-proBNP values above the 80th percentile had a nonsignificant trend toward higher risk of CHD event (HR, 2.66; 95% CI, 0.74-9.57; P = .13; n = 12, also in the multivariate models), whereas there was no association between CHD events and high levels of CRP (HR, 1.04; 95% CI, 0.29-3.67; P = .95) and urinary albumin/creatinine ratio (HR, 1.05; 95% CI, 0.29-3.72; P = .94) for values above the 80th percentile of the 2 respective biomarkers.

This is the first study to our knowledge to examine the prognostic value of the cardiac peptide NT-proBNP compared with the established cardiovascular risk markers, CRP, and urinary albumin/creatinine ratio in nonhospitalized individuals. In our study, plasma NT-proBNP levels independently predicted mortality and first major cardiovascular events. The prognostic information of NT-proBNP was independent of traditional cardiovascular risk factors, prevalent cardiovascular disease, as well as left ventricular dysfunction and renal function. NT-proBNP was a better predictor than both CRP and urinary albumin with respect to all of the outcomes, and CRP did not provide prognostic information beyond traditional risk factors.

Our study suggests that CRP levels are not independently associated with mortality or first cardiovascular event in older individuals. Increasing age seems to attenuate the association between plasma CRP levels and the risk of cardiovascular disease. In accordance with our observations, several studies have demonstrated that high CRP levels did not predict mortality or cardiovascular risk beyond traditional risk factors in older persons,^28- 29 although other studies have found an independent relationship between CRP levels and cardiovascular risk.³⁰ The majority of the previous studies that have found an independent association between plasma levels of CRP and cardiovascular outcome or mortality only enrolled middle-aged populations.^4,27,31 Both NT-proBNP and CRP had significant interactions with age, but NT-proBNP continued to be a significant predictor after controlling for age. Furthermore, improving risk stratification beyond traditional risk factors with measurements of CRP has proven difficult.^28,31 This lack of improved predictive ability was further emphasized in a recent meta-analysis of CRP and prediction of coronary heart disease.³²

In contrast to previous studies on the prognostic value of the B-type natriuretic peptides and CRP, measurements of urinary albumin excretion were included in our study. High urinary albumin excretion rate is a well-known risk marker of overall mortality and cardiovascular disease, both in individuals with and without diabetes mellitus.^33- 35 This was supported in our study, as NT-proBNP and urinary albumin/creatinine ratio both emerged as independent predictors of death and a first major cardiovascular event, even when entered into the same model.

The only biomarker that was able to predict an increased risk of stroke was NT-proBNP. This association was remarkably strong, as values above the 80th percentile were associated with a 3.63-fold increase in risk of ischemic stroke in individuals without known cardiovascular disease, even after accounting for known risk factors. This observation supports similar recent data from the Framingham population.¹⁹ The mechanism behind this interesting relationship remains to be elucidated. However, an association between the gene for atrial natriuretic peptide and the risk of stroke has recently been found in humans.³⁶

High NT-proBNP levels were also markedly and independently associated with increasing risk of heart failure, even after exclusion of individuals with impaired left ventricular systolic dysfunction at baseline. Several mechanisms apart from possible undetected diastolic dysfunction at baseline may account for this observation.^37- 38 The B-type natriuretic peptides are released in response to increased ventricular wall stress.³⁹ However, the results of our study suggest that plasma NT-proBNP levels independently predict future mortality and major cardiovascular events, even after adjustment for echocardiographic variables related to left ventricular dysfunction. This observation could reflect the relatively weak association that has been observed between left ventricular systolic dysfunction and increased levels of BNPs in individuals without any symptoms.^40- 41 However, other mechanisms may be important and, interestingly, myocardial ischemia could also be a key stimulus for BNP and NT-proBNP release.⁴² Furthermore, both circulating BNP and NT-proBNP increase after percutaneous coronary intervention, even when the intraventricular filling pressures remain unchanged.^43- 44 Thus, these peptides could be indicators of generalized early cardiac impairment, including asymptomatic myocardial ischemia. A possible reason for no observed significant association between NT-proBNP and a first CHD event could be the limited number of CHD events in our study.

Plasma NT-proBNP values were associated with an increased 2- to 3.5-fold risk at levels found in the upper fifth of the population, which is consistent with the recent study of the predictive value of BNP on mortality and cardiovascular events in the younger Framingham population.¹⁹ Our study extends these findings using BNP with the use of NT-proBNP and adds to the applicability of the test by including a broad range of older individuals. We have previously reported on the short-term prognostic value of NT-proBNP in the present cohort, which suggested that NT-proBNP is a strong predictor of mortality. Unlike our current investigation, this study included individuals with known heart failure and the follow-up period was only 2.2 years.¹⁷ Previous population studies measured BNP and demonstrated the predictive ability of this peptide in younger as well as in older individuals.^16,18 Thus, measurement of both BNP and NT-proBNP are promising risk biomarkers in apparently healthy individuals. To our knowledge, no studies have compared the prognostic value of BNP and NT-proBNP in the same cohort; hence, whether one of these peptides may be better as a risk marker is unknown.

There are several limitations to our study. Given the sampling procedure, the current cohort included only individuals aged 50 to 89 years; thus, the main results should not be extrapolated to younger groups, especially as CRP seems to be a stronger risk marker among those groups. Furthermore, as in the Framingham cohort,¹⁹ our population is homogeneous and almost exclusively white, which may limit the generalizability to other ethnic groups. We did not include lifestyle factors, such as body mass index, diet, and exercise, in the survival analyses because these data were not available. However, given the strength of the results, we consider it unlikely that further adjustment for these factors would change the main results of our study. The small number of individual cardiovascular events requires caution when interpreting our results. Hence, the cut points used in this population may not be applicable in other community cohorts. Future large-scale population studies may provide clinically useful information regarding prognostic cut points for various high-risk populations. Finally, being the first population study examining the predictive ability of NT-proBNP, our findings should be confirmed in other populations from the community.

In conclusion, our results demonstrate that increased plasma NT-proBNP was independently associated with increased mortality and first cardiovascular events even after controlling for traditional cardiovascular risk factors. Urinary albumin/creatinine ratio was also predictive but plasma CRP did not contribute to the risk stratification in older nonhospitalized individuals.