Non–ST-segment elevation acute coronary syndrome (NSTEACS) is a heterogeneous condition regarding both the prognosis and the underlying pathophysiology. Persistent or transient thrombotic occlusion at the site of plaque rupture or erosion (1) or microembolizations from the thrombus (2) lead to unpredictable episodes of myocardial ischemia, which, in turn, may lead to irreversible myocardial damage. Ischemia and myocardial damage cause transient and permanent increase in wall tension and myocardial stretch. The B-type natriuretic peptide (BNP) and the N-terminal part of its prohormone (NT-proBNP) are mainly released from myocytes in the cardiac ventricles in response to increased stretch and wall tension (3). Therefore, these markers are useful for detection of left ventricular systolic and diastolic dysfunction (4- 5). However, the natriuretic peptides might also rise in response to a number of other stimuli, among them ischemia per se and cytokines (6- 7).

The BNP and NT-proBNP rise early after the onset of symptoms in patients with NSTEACS, and it has recently been shown that the degree of elevation is associated with both short- and long-term mortality (8- 10). However, so far the long-term temporal changes of natriuretic peptide levels and the relation between the changes of natriuretic peptide levels, preexisting diseases, markers of myocardial necrosis and inflammation have not been examined after an episode of NSTEACS. Therefore, we collected serial observational data in association with the Fragmin and fast Revascularisation during InStability in Coronary artery disease (FRISC)-II trial with the following aims:

The FRISC-II trial was a prospective, randomized multicenter study in which 3,489 patients with NSTEACS were randomized in a factorial design to an early noninvasive or invasive strategy and to three months treatment with dalteparin or placebo (11- 12). Patients with a history of previous open heart surgery, or who were of advanced age, or poor general health, and those included after completion of recruitment of patients to the invasive versus noninvasive arm were not randomized regarding invasive versus noninvasive strategy. These patients were treated primarily noninvasively and randomized only regarding extended treatment with dalteparin or placebo. Both chest pain and signs of ischemia (ST-segment depression ≥0.10 mV or T-wave inversion ≥0.10 mV or raised biochemical markers) were required for inclusion. Written informed consent was a prerequisite for inclusion. All local ethics committees approved the protocol.

Samples of EDTA-plasma were obtained at randomization and at 48 h, 6 weeks (range 4 to 7 weeks), 3 and 6 months after randomization at selected sites. All samples were frozen and stored in aliquots at −70°C and sent for central analyses.

Plasma NT-ProBNP was determined at all time points using Elecsys proBNP sandwich immunoassay on an Elecsys 2010 (Roche Diagnostics, Mannheim, Germany). The analytical range extends from 20 to 35,000 ng/l. The upper reference level (97.5th percentile) in men and women aged 40 to 65 years is 184 and 268 ng/l, respectively, and age 66 to 76 years, 269 and 391 ng/l, respectively (13). The total coefficient of variation was 3.3% (n = 21) at a level of 209 ng/l; NT-proBNP was analyzed in 1,352; 1,273; 1,189; 1,141; and 1,096 subjects at inclusion; day 2; 6 weeks; 3 months and 6 months, respectively, and at all five time points in 961 patients.

Plasma cardiac troponin T (cTnT) was determined at randomization by the third generation cTnT assay on an Elecsys 2010 (Roche Diagnostics, Mannheim, Germany). The minimum detectable concentration is 0.01 μg/l, and the 99th percentile in healthy individuals is <0.01 μg/l. Plasma C-reactive protein (CRP) was measured at randomization with a high sensitive assay (Immulite CRP, Diagnostic Products Corporation, Los Angeles, California). The detection limit is 0.1 mg/l. Serum creatinine was determined locally at each center. The creatinine clearance rate was calculated using the Cockroft and Gault equation with correction for gender (14).

Standard 12-lead electrocardiogram readings were obtained at inclusion and were interpreted at a central core laboratory. Echocardiograms were obtained in 1,186 of the 1,352 patients (88%) before discharge and always before any invasive procedure. The left ventricular ejection fraction (LVEF) was visually assessed, and a depressed systolic left ventricular function was considered present when the ejection fraction was below 0.45. In patients randomized to an early invasive treatment, coronary angiography was performed in median 3 (25th to 75th percentile, 2 to 5) days after randomization. The operator assessed the presence and grade of any stenosis and the Thrombolysis In Myocardial Infarction (TIMI) flow grade according to a detailed evaluation form.

The patients randomized to an invasive versus a noninvasive strategy were followed for two years regarding occurrence of death and myocardial infarction (MI). The cohort of patients eligible only for randomization to long-term dalteparin versus placebo was followed up to six months. Analyses regarding mortality and occurrence of MI in relation to NT-proBNP levels were, therefore, performed only in the cohort randomized to an invasive versus a noninvasive strategy.

Continuous variables are summarized as median and 25th and 75th percentiles. Because NT-proBNP levels were not normally distributed, nonparametric tests were applied. Differences in levels of NT-proBNP over time were evaluated by the Wilcoxon signed rank test or the Friedman test as appropriate. Differences in median levels of NT-proBNP between two groups were evaluated by the Mann-Whitney U test. Differences in proportions were evaluated by the chi-square test.

The kinetics of NT-proBNP was studied by modeling it as a response variable on time. A logarithmic transformation of the NT-proBNP values was applied. The time variable was ordinalized as 0, 1, 2, 3, and 4 for the 5 consecutive visits (baseline, 48 h, 6 weeks, 3 months, and 6 months). Hence, changes of NT-proBNP per time unit should be interpreted as changes between these consecutive visits. The temporal change of each patient's ln(NT-proBNP) values per visit could be approximated by a linear function, and, therefore, for each patient with at least three measures of NT-proBNP, a linear regression model was fitted. The distribution of the estimated intercept and slope was then associated to background characteristics of the patients. After that, a linear mixed effects model, with individual intercept and slope as random effects, was formulated. The covariates entered as fixed effects were: age (centered at median and scaled to a unit of 10 years), gender, diabetes, treatment (dalteparin/placebo), strategy (not randomized/randomized to invasive strategy/randomized to noninvasive strategy), history of heart failure, history of MI, ST-segment depression, CRP, cTnT, creatinine clearance, and ongoing treatment on admission with an angiotensin-converting enzyme inhibitor, diuretic, long-acting nitrate, and beta-blocker.

To compare the predictive value of NT-proBNP measurements at different time points, receiver-operating characteristic curves were generated, and the area under the curves calculated. The adjusted odds ratios for two-year mortality of a doubling of NT-proBNP level at different time points were calculated by logistic regression analysis. Adjustment was made for age, gender, diabetes, previous MI, ST-segment depression, creatinine level, cTnT, and CRP elevation.

In all tests, a p value <0.05 was considered statistically significant. The data analyses were performed using the SPSS system 11.5 (SPSS Inc., Chicago, Illinois) and the statistical program R (R Development Core Team [2003], Vienna, Austria).

Of the patients included in the study, 71% were men, 14% had diabetes mellitus, 29% had previously had an MI, and 5% had a history of congestive heart failure (CHF). The admission electrocardiogram showed an abnormal Q-wave in 21% and ST-segment depression in 48%. The median age was 68 years (25th to 75th percentile, 59 to 74 years).

Randomization took place in median 38 h (25th to 75th percentile, 26 to 53 h) from onset of the last episode of chest pain. A total of 539 and 544 of the patients were randomized to an invasive strategy and a noninvasive strategy, respectively. A total of 269 additional patients not eligible for randomization to the invasive study were randomized to long-term dalteparin or placebo treatment.

The serial changes of NT-proBNP in the 961 patients with NT-ProBNP analyzed at all time points were characterized by an initial phase with a rapid decrease and then a secondary phase with a slower decrease continuing throughout the six-month observational period (Figure 1A). The median NT-proBNP level at inclusion was markedly elevated, 529 ng/l (25th to 75th percentile, 207 to 1,237 ng/l), and at six months the median level had been more than halved, 238 ng/l (25th to 75th percentile, 114 to 519 ng/l). Very similar results were found if all patients with NT-proBNP measurements at each time point were included in the analysis (Figure 1B). However, although the NT-proBNP levels, on average, decreased by time substantially, the pattern varied considerably between individual patients.

A large number of variables were associated with the baseline levels and rate of change of NT-proBNP in univariate analysis (Table 1). However, in the multiple linear mixed effects model (Table 2) only increasing age, female gender, previous MI, elevated cTnT and CRP, creatinine clearance below the median, and ST-segment depression on admission were independently associated with a higher baseline level. Elevated cTnT, CRP, and female gender were associated with significantly higher reduction rates. In contrast, high age, diabetes, previous MI, treatment with diuretics and long-acting nitrates on admission were associated with significantly lower reduction rates. The combined effect of cTnT and CRP levels at inclusion on NT-proBNP levels at the different time points is illustrated in (Figure 2).

Separate analyses, treating cTnT as a continuous variable, and excluding those without a measurable value of cTnT (<0.01 μg/l) revealed that higher cTnT levels were associated with higher NT-proBNP levels at baseline, as well as at six months. There was a negative correlation between the cTNT level and the rate of change of NT-proBNP (i.e., the higher cTnT level, the higher reduction rate of NT-proBNP). Similarly, there was a negative correlation between the CRP level and the rate of change of NT-proBNP. There was no significant difference between patients randomized to an invasive versus a noninvasive strategy in baseline levels, rate of change, or six-month levels of NT-proBNP.

A total of 16% of the patients had an LVEF below 0.45 before discharge. In patients with baseline levels of NT-proBNP in the lower tertile, approximately 8% had an estimated LVEF below 0.45 regardless of the rate of decrease of NT-proBNP between randomization and day 2. However, in patients with baseline levels in the upper two tertiles, the proportion with depressed left ventricle was related to the reduction rate of NT-proBNP (Figure 3A). In those with baseline levels in the third tertile and no, or very slight, decrease, 38% had a depressed LVEF compared with only 21% in those with a high reduction rate.

There was a decreasing proportion of patients with TIMI-3 flow in the coronary arteries by increasing baseline levels of NT-proBNP: 67%, 46%, and 44% (p < 0.001) in the first, second, and third tertiles, respectively. However, in those with a high reduction rate of NT-proBNP between randomization and day 2, the proportion of patients with TIMI flow grade 3 were similar regardless of baseline levels (Figure 3B).

The median (25th to 75th percentile) baseline NT-proBNP level was markedly elevated in nonsurvivors compared with survivors, 1,140 ng/l (359 to 2,523 ng/l) versus 511 ng/l (197 to 1,184 ng/l), p < 0.001. However, the rate of change of NT-proBNP between randomization and day 2 was not associated with the two-year mortality; the median (25th to 75th percentile) percentage decrease was 39 (19 to 58) and 31 (−4 to 54), p = 0.16, in nonsurvivors and survivors, respectively. Neither was the rate of change of NT-proBNP between randomization and six months associated with mortality (data not shown).

The NT-proBNP level at each separate time point was predictive of mortality. In receiver operating characteristic analysis regarding subsequent death, the area under the curve (95% confidence interval) increased by time and was 0.67 (0.59 to 0.75), 0.66 (0.57 to 0.75), 0.74 (0.65 to 0.83), 0.82 (0.73 to 0.91), 0.81 (0.70 to 0.93) at baseline, day 2, 6 weeks, 3 months, and 6 months, respectively. Also, the adjusted odds ratio increased for each time point (Figure 4). Furthermore, comparing cut-off levels at certain specificity for subsequent death revealed that the corresponding cut-off levels decreased considerably by time. The cut-off values at a specificity of 60% were 722, 482, 398, 301, and 264 ng/l at baseline, day 2, 6 weeks, 3 months, and 6 months, respectively.

B-type natriuretic peptide and NT-proBNP become markedly elevated early in acute coronary syndromes (6). The NT-proBNP level peaks at 14 to 48 h from the onset of an acute MI (15), which roughly coincides with the time the baseline sample was obtained in the present study. The median level at baseline of NT-proBNP was 529 ng/l, a level exceeding the levels seen in populations of patients with stable CHF with moderately depressed LVEF (4). However, the NT-proBNP levels in patients with NSTEACS are subject to marked dynamic changes. In the present study, the median NT-proBNP levels were found to decrease throughout the whole sampling period, rapidly in the early phase followed by a more gradual decline.

A number of factors were independently associated with higher baseline levels and/or a more subtle decrease of NT-proBNP levels over time. Patients with a previous MI often have chronically impaired left ventricular function, which gives a higher baseline level of NT-proBNP and also limits the possibility for normalization of the level. Patients with diabetes are also known to more often have impaired left ventricular systolic and diastolic function, even in the absence of coronary artery disease (16- 17). High age and female gender are associated with higher baseline levels of NT-proBNP in healthy populations (13).

Elevated cTnT and CRP were independently associated with higher baseline levels together with a more pronounced decrease of NT-proBNP levels over time. The cTnT level is correlated to the infarct size (18). However, most of the elevations of cTnT in the present study were only minor and the associated necrosis too small to cause a significant permanent reduction in global left ventricular function (19). Thus, the infarct size cannot be the sole explanation for the association between cTnT elevation and the marked elevation of NT-proBNP at baseline. A second explanation might be ischemia, which may not only cause necrosis, but also transient ventricular dysfunction leading to elevation of NT-proBNP. Furthermore, experimental studies have shown that ischemia might induce elevation of NT-proBNP even without any changes in loading conditions (6). ST-segment depression, a marker of ischemia, was strongly associated with higher baseline levels of NT-proBNP in the present study. Microembolies originating from the site of the ruptured plaque, a common reason for minor elevations of troponin in NSTEACS (20), might offer a third explanation. Experimental studies have shown that microembolization causes a transient depression of the left ventricle, which is mediated by a massive activation of cytokines, such as tumor necrosis factor-alpha and interleukin-6 (21). Both tumor necrosis factor-alpha and interleukin-6 possess a negative inotropic effect on the myocytes (22). Furthermore, interleukin-6 might also directly increase the levels of BNP and NT-proBNP by increasing BNP gene expression (23). Because interleukin-6 also regulates the production of CRP, this mechanism might also explain the association between CRP and NT-proBNP levels. Thus, myocardial damage and the associated inflammatory reaction (24- 25) seem to be very important for the dynamic changes of NT-proBNP in NSTEACS. Consequently, there were stable levels of NT-proBNP over time in the absence of cTnT elevation (Figure 2) in the present study.

Unfortunately, we could not evaluate the effect on the rate of change of NT-proBNP of nonrandomized treatments started after enrollment. However, other studies have shown that long-term treatment with angiotensin-converting enzyme inhibitors (26), beta-blockers (27), and diuretics (28) decrease BNP levels. We cannot, therefore, exclude that initiation of such treatments might have contributed to the decreasing NT-proBNP levels, especially during the later part of the measurement period.

The NT-proBNP has been shown to be a powerful marker of systolic function and discriminates well between normal and impaired left ventricular systolic function in the general population (29). However, these results may not be applicable in NSTEACS patients because also among those with a baseline level of NT-proBNP as high as above 1,034 ng/l, only a minority had impaired LVEF in the present study (Figure 3A). This was particularly evident in those with a rapid decrease of NT-proBNP, indicating that, in those with a transient increase in NT-proBNP, the peak level poorly predicts the resulting left ventricular function. Furthermore, in acute MI, NT-proBNP has been found to be a complementary predictor to LVEF of major adverse events (30), rather than a substitute for the LVEF, as in CHF (31- 32).

There were decreasing proportions of patients with TIMI flow grade 3 in the coronary arteries by increasing baseline levels of NT-proBNP. However, this was not seen in the subgroup with a rapid decrease where the proportions of patients with TIMI flow grade 3 were similar regardless of baseline levels (Figure 3B). This finding suggests that maintaining or re-establishing an adequate coronary blood flow is essential in NSTEACS in order to rapidly decrease the NT-proBNP levels.

The baseline level of NT-proBNP in the FRISC-II trial was highly predictive of subsequent mortality as previously reported (33) and in accordance with previous studies of NT-proBNP or BNP in acute coronary syndromes as well as in CHF (8- 10,29,31- 32). However, the rate of decline of NT-proBNP was not associated with the two-year mortality. Neither did the degree of early rise of NT-proBNP predict subsequent death in a previous study (9). It might seem contradictory that patients with a high baseline level and a rapid decrease did not have a better prognosis than patients with high baseline levels and no or only minor decrease, bearing in mind that these patients less often had impaired systolic left ventricular function and more often TIMI flow grade 3. However, that was probably counterbalanced by a more unstable lesion in the coronary arteries in those with a transient marked elevation of NT-proBNP, as indicated by a higher proportion with cTnT elevation (34).

The NT-proBNP level at each separate time point was predictive of subsequent death and, as indicated by the areas under the curve as well as the adjusted odds ratios, increasingly so for each time point. One possible explanation for this might be that the risk of death after the initial unstable phase with a high risk of reinfarction (of which NT-proBNP is not predictive 10) is more closely related to the left ventricular systolic function, and that the NT-proBNP level in stable phases better reflects the left ventricular systolic function.

The dynamic changes of NT-proBNP in patients with ACS also have important implications for the choice of decision limits. An NT-proBNP level above 264 ng/l at six months had the same specificity for subsequent death as a level above 722 ng/l at randomization, an almost three-fold difference.

The present study shows that the initial rise of NT-proBNP in patients with NSTEACS is mainly reversible. After the early rapid decrease, NT-proBNP levels continue to decrease for at least six months. Factors associated with less reversibility of NT-proBNP are related to chronically impaired left ventricular function, and factors associated with greater reversibility are related to the acute myocardial damage.

The NT-proBNP level is predictive of subsequent death irrespective of when it is measured. However, the NT-proBNP level measured during a chronic, relatively stable phase seems to be an even stronger predictor of mortality than during an acute unstable phase. Furthermore, the absolute levels of NT-proBNP associated with a certain specificity are vastly different in stable and unstable phases. Thus, the clinical setting and timing of measurement will be important to consider when using NT-proBNP for risk assessment and when defining optimal decision limits for NT-proBNP.