Association of Annual N-Terminal Pro-Brain Natriuretic Peptide Measurements With Clinical Events in Patients With Asymptomatic Nonsevere Aortic Stenosis

Study Design and Participants

The SEAS study was a multicenter (173 centers in Northern Europe), double-blind randomized clinical trial investigating the effect of the simvastatin with ezetimibe combination vs placebo. A total of 1873 patients, aged 45 to 85 years, were randomized between January 6, 2003, and March 4, 2004, and the trial was completed April 1, 2008. The main inclusion criterion was an investigator-assessed transaortic maximal flow velocity (Vmax) 2.5 to 4.0 m/s measured by Doppler echocardiography. The original protocol dictated the exclusion of patients with known ischemic heart disease or an indication for statin therapy.^29,30

This post hoc NT-proBNP substudy examined the association of adjusted NT-proBNP concentrations at baseline (enrollment at year 0) and during 1 year with AVEs during the next 2 years of follow-up. We included 1644 patients with NT-proBNP levels ascertained at baseline and after 12 months (year 1) who had at least 1 in-study Vmax measurement and no AVEs before year 1 (eFigure 1 in the Supplement). Data on race were collected but not used in this study, because 99.7% (1639 of 1644) of the population was White. Patients who were excluded were not substantially different from those included in the study (eTable 1 in the Supplement). The ethics committees approved the SEAS trial and adhered to the Declaration of Helsinki³¹ as reflected by written informed consent obtained from all participants.

Echocardiography

Echocardiograms were analyzed according to published guidelines,^30,32 and we report results obtained by the core laboratory's analysis after clinical data blinding.²⁹ Mild and moderate AS were defined by the split of Vmax greater than or equal to 3.0 m/s at baseline, based on the core laboratory evaluation. We used the velocity difference from baseline to year 1 greater than or equal to 0.3 m/s/y as a covariate in multivariable analyses.

NT-proBNP Level Measurement

Blood samples at baseline and year 1 were drawn into tubes containing EDTA, centrifuged, stored at −80 °C, and analyzed in a central laboratory between November 18 and December 2, 2016, with no indication of sample degradation (eFigure 2 in the Supplement). Measurement of NT-proBNP was performed with the Roche immunoassay (Elecsys ProBNP II, on COBAS e 601), with a reported measuring range of 5 to 35 000 ng/L and an intermediate precision between 2.9% and 6.1%. There were 50 missing NT-proBNP measurements at baseline.

An adjusted NT-proBNP level, specific for age and sex, was defined similarly to a previous study on BNP in AS²⁵:

A normal NT-proBNP level was defined as an NT-proBNP ratio less than or equal to 1 (upper limit of normal considered 97.5 percentile specific for age and sex³³) and an increased level was defined as an NT-proBNP ratio greater than 1.

A relative NT-proBNP change was defined as

A relative NT-proBNP change greater than or equal to 1.5 was the primary exposure variable, but spline models also tested the continuous variable.

End Points

An independent clinical end point committee had adjudicated the end points according to a prespecified end point classification manual,²⁹ independently of NT-proBNP results. The primary end point for this NT-proBNP substudy was the adjudicated AVEs, defined as a composite of the first aortic valve replacement, cardiovascular death, or incident heart failure due to AS progression (1 event tallied per participant). The secondary end point was all-cause mortality as the first occurring solitary event. According to contemporary guidelines,^{34,35,36,37,38,39,40,41,42,43} the local heart team decided to refer patients for surgery without knowledge of the NT-proBNP concentrations.

Statistical Analysis

Statistical analysis was performed from February 10 to October 10, 2021. Descriptive statistics were used, including χ², analysis of variance, Wilcoxon, and Kruskal-Wallis tests as appropriate. Analyses were performed separately for mild and moderate AS. Imputation of missing Vmax values at either baseline or year 1 was done for 136 patients, and missing creatinine level values were imputed in 94 patients (eMethods in the Supplement). A multivariable linear regression model examined the correlation between the natural logarithm of NT-proBNP levels at baseline and the following clinically relevant independent variables: aortic valve area, left ventricular ejection fraction (LVEF), left ventricular mass index, Vmax, systolic blood pressure, body mass index, creatinine level, age, and sex.

All analyses started at year 1 as the landmark to account for immortal time bias in the time-to-event outcomes. Cumulative incidence rates per 100 patient-years, presented with 95% CIs, were determined to examine the primary outcome, and the Fine and Gray cumulative incidence function with competing risk model examined the secondary end points for a clinically appropriate 2-year follow-up (year 1 to year 3), concordant with the 1- to 3-year follow-up interval recommended in current guidelines.^1,2 The Gray test compared cumulative incidence curves for the primary end point for normal and increased NT-proBNP levels and NT-proBNP level changes above and below 1.5 during an extended follow-up (year 1 to year 5).

Adjusted risk estimates for outcomes were assessed by univariable and multivariable Cox proportional hazards regression models and are presented as cause-specific hazard ratios (HRs) with 95% CIs. The multivariable model included NT-proBNP level, age, sex, echocardiographic measures, creatinine level, systolic blood pressure, body mass index, and diuretic use, stratified by the center of inclusion. We tested for main effect estimates and potential interactions between NT-proBNP level and NT-proBNP level change and each study outcome, as well for additional interactions with age, sex, and obesity.

Restricted cubic splines based on Cox proportional hazards regression models described the association between NT-proBNP level change from baseline to year 1 and risk for AVEs. Four knots were applied based on visual judgment, which indicated smooth, progressive risk without overfitting.

Two-tailed P values <.05 were considered statistically significant. Statistical analyses were performed in Stata software, version 13.1 (Stata/IC; packages estout, table1_mc, rc_spline; StataCorp LLC).

Baseline Characteristics

Of the 1873 patients in the SEAS trial, we included 1644 patients (87.8%) in the present analysis. Of these, 996 were men (60.6%) and 648 were women (39.4%); mean (SD) age was 67.5 (9.7) years and the mean (SD) LVEF was 65.9% (8.2%). A total of 860 patients (52.3%) had moderate AS with a mean Vmax 3.5 (0.4) m/s, aortic valve area of 1.1 (0.4) cm², mean aortic gradient 28.9 (6.9) mm Hg, and higher left ventricular mass index than patients with mild AS (104.6 [31.3] vs 96.7 [28.7] g/m²) (Table 1).

Adjusted NT-proBNP level was normal in 77.0% of patients (1228 of 1594 [50 missing NT-proBNP values]) at baseline and 70.8% (1164 of 1644) of patients at year 1 (P < .001) (Table 1; eFigure 1 in the Supplement). Increased NT-proBNP levels were measured in 20.5% (n = 161) of patients with mild AS and 37.1% (n = 319) of those with moderate AS at year 1. From baseline to year 1, patients with moderate AS had a median increase of 21% in the NT-proBNP level compared with a 10% increase in those with mild AS (P < .001) (Table 1).

Adjusted NT-proBNP levels at baseline correlated with the aortic valve area, Vmax, left ventricular mass index, LVEF, body mass index, and creatinine level (all P < .05) in a multivariable linear regression analysis (eTable 2 in the Supplement). The change in NT-proBNP levels at year 1 as a continuous variable correlated with a decrease in LVEF between baseline and year 1 (eTable 3 in the Supplement). Patients with an NT-proBNP level change greater than or equal to 1.5 had more frequently abnormal echocardiographic parameters at baseline, including left ventricular mass index, mean aortic gradient, and Vmax (eTable 4 in the Supplement).

Outcomes Associated With NT-proBNP Level

Over a median follow-up of 42.2 months beyond year 1, a total of 510 patients (31.0%) experienced an AVE. The rate of AVEs increased significantly per quintile of NT-proBNP (year 1), especially at the fourth quintile, marking the transition to an abnormal NT-proBNP level (NT-proBNP ratio: 0.71-1.38) (eFigure 3 in the Supplement). Within the 2-year follow-up, the rate of AVEs was lower in patients with normal adjusted NT-proBNP concentrations at year 1 than patients with increased levels for patients with both mild and moderate AS (Figure 1, Table 2). Specifically, the AVE rates per 100 patient-years for normal vs increased adjusted NT-proBNP levels at year 1 were 1.39 (95% CI, 0.86-2.23) vs 7.05 (95% CI, 4.60-10.81) in mild AS (P < .01), and 10.38 (95% CI, 8.56-12.59) vs 26.20 (95% CI, 22.03-31.15) in moderate AS (P < .01). AS. Aortic valve replacement accounted for more than 80% of the observed composite events.

The incidence of AVEs was independently associated with both an increased NT-proBNP level at year 1 and NT-proBNP level change in multivariable Cox proportional hazards regression analysis (eTable 5 in the Supplement) for both patients with mild and moderate AS.

A restricted cubic spline model demonstrated that an increase in NT-proBNP levels greater than or equal to 1.5 was associated with an excess risk for AVEs (Figure 2). An NT-proBNP level decrease from baseline to year 1 or a less than 50% increase in NT-proBNP levels was not associated with higher AVE rates. In a multivariable Cox proportional hazards regression analysis, an increased NT-proBNP level at year 1 was associated with an increased risk of cardiovascular death in both patients with mild AS (HR, 5.28; 95% CI, 1.19-23.40; P = .03) and moderate AS (HR, 4.01; 95% CI, 1.32-12.20; P = .01) (eTable 5 in the Supplement). The all-cause mortality rates per 100 patient-years for normal vs increased adjusted NT-proBNP levels at year 1 were 1.05 (95% CI, 0.61-1.81) vs 4.17 (95% CI, 2.42-7.19) in mild AS (P < .01), and 1.60 (95% CI, 0.99-2.57) vs 4.78 (95% CI, 3.32-6.87) in moderate AS (P < .01). Less than 20% of these deaths were due to a cardiovascular cause (Table 2).

Sensitivity Analyses

Aortic valve event rates were significantly higher among both patients with mild and moderate AS with a 1.5-fold or greater increase in NT-proBNP levels over 1 year (Figure 1C, D). A 1.5-fold or greater increase in NT-proBNP levels was additive to an increased NT-proBNP level at year 1 for all outcomes (all P < .01) (Figure 3), but the statistical interaction was not significant (P = .25). Nonetheless, the highest risk for AVEs, the individual components of the composite outcome and all-cause mortality, was observed in patients with both an increased NT-proBNP level at year 1 and a 1.5-fold or greater increase in NT-proBNP levels during 1 year. The HR for AVEs was 8.12 (95% CI, 3.53-18.66; P < .001) in patients with mild AS and 4.05 (95% CI, 2.84-5.77; P < .001) for patients with moderate AS (Figure 3).

These results were observed during a 2-year follow-up period (year 1 to year 3). Follow-up data through study completion (year 1 to year 5) suggest a continued separation of the event curves (eFigure 4 B, C in the Supplement). The association of the combination of NT-proBNP levels at year 1 plus NT-proBNP level changes from baseline with events was noted across age, sex, and body mass index subgroupings (eFigure 5, A-C; eTable 6 in the Supplement). A normal NT-proBNP level at both baseline and year 1 was associated with AVE rates of less than 5% during 24 months in patients with mild AS and 6 months in those with moderate AS (Figure 1C, D).

Single NT-proBNP Level Measurements

Increased NP levels are well-known risk markers of morbidity and mortality in cardiovascular diseases and severe AS^{18,19,20,21,22,23,24,25,26,27,28,44} but to our knowledge have not been systematically evaluated in patients with nonsevere AS. The European guidelines¹ recommend BNP measurement when considering aortic valve intervention in patients with asymptomatic severe AS. This study explores the association between NT-proBNP levels and AS-related outcomes. For patients with moderate AS, our data are consistent with those reported from a 2020 study²⁷ showing that NT-proBNP levels above 888 pg/mL (to convert to nanograms per liter, multiply by 1) (approximately 1.5 times above normal values for age) were significantly associated with a higher mortality rate after appropriate adjustment.

We used assay-specific NT-proBNP reference values because there is no universal standardized cutoff value for clinical decision-making in the individual patient. Other studies introduced risk scores^45,46 or age- and sex-adjusted ratios^25,28 based on their populations.

The SEAS study population differs from the usual AS outpatient population, which ensured the opportunity to examine the effect of isolated AS on NT-proBNP levels. Hence, the prevalence of normal NT-proBNP levels was probably higher than for unselected patients with AS encountered in clinical practice.

In our study, the association between a normal NT-proBNP level and low event rates was also found after adjusting for several factors, including echocardiographic variables. Therefore, our findings suggest the potential utility of using NPs as an alternative or add-on test to risk stratify a substantial population of patients who are monitored for many years. The low incidence of AVEs associated with a normal NT-proBNP level at year 1 suggests that echocardiography may be safely deferred. Similarly, increased NT-proBNP levels, and especially levels that increase more than 1.5-fold over 1 year, identify patients who likely deserve more frequent or more intense scrutiny. For example, patients with mild AS with an increased NT-proBNP level combined with a 1.5-fold or greater annual NT-proBNP level change had an almost 10% 1-year incidence of AVEs (Figure 1C). In contrast, patients with moderate AS with the same NT-proBNP combination reached the same probability of 10% in less than 6 months (Figure 1D).

Serial NT-proBNP Level Measurements

Natriuretic peptide levels fluctuate with age, sex, and kidney function.⁴⁷ NT-proBNP has a longer half-life⁴⁸ and better stability than BNP.⁴⁹ Therefore, previous research proposed serial NT-proBNP measurements as a good marker for personalized risk prediction.⁵⁰ Other studies used population-derived cutoffs,^21,23,44 with inadequate performance, and therefore proposed that change over time would be a better prognostic marker in moderate AS.⁵¹ To our knowledge, we have evaluated for the first time the association of an annual NT-proBNP level change of 1.5-fold or more as an individual patient-related biomarker in nonsevere AS.

Our analysis demonstrated that an annual NT-proBNP level change of 1.5-fold or more was independently associated with an increased risk of AVEs, a novel finding. This association supports hypotheses generated in previous longitudinal studies.^21,44,52 Still, direct comparison is hampered by the scarce literature on serial NT-proBNP measurements describing the considerable differences between study populations and different cutoff values.^21,23,44 In addition, previous studies failed to account for substantial confounders, such as AS severity, and did not adjust for competing risks and immortal time bias.

To our knowledge, the proposed cutoff value of 1.5 or greater NT-proBNP level change is novel and has not been prospectively tested. Spline curves indicated that a threshold NT-proBNP change value of 1.5 was different from a hazard ratio of 1.0, including its lower 95% CI (Figure 2). Physiological data support a threshold of 1.5, just exceeding the diurnal NT-proBNP level within-subject variability of 30% to 50% in healthy individuals.^51,53

Correlations of NT-proBNP Levels With Clinical and Echocardiographic Measures

Our subanalyses aligned with the expectation that NT-proBNP levels correlate with echocardiographic measures of AS progression, namely, higher LV mass,^22,24 higher mean transaortic gradient,^{20,22,23,24,44} and lower LVEF.^26,28 NT-proBNP crude values and the adjusted NT-proBNP ratio was also associated with outcomes after adjusting for echocardiographic parameters and the other NT-proBNP modifiers, including diuretic therapy.⁴⁷ We could not adjust for latent ischemic heart disease and cardiac amyloidosis that could affect NT-proBNP levels, symptoms, and outcomes. Notwithstanding this limitation, a normal NT-proBNP concentration at year 1 in our study is associated with a good prognosis. At the same time, an increased or increasing NT-proBNP level is a signal for further exploration to identify the potential need for medical or invasive intervention.

NT-proBNP Level as a Risk Marker

For patients with mild AS, a normal NT-proBNP level carries less than 5% AVE risk for at least 24 months, supporting the recommended 2- to 3-year interval for repeated echocardiography. The frequency of echocardiography may need to be individualized, however, for patients with mild AS with increased NT-proBNP levels, especially those with a 50% or greater increase during 1 year. Similar concerns relate to patients with moderate AS.

Our data suggest that annual or biannual NT-proBNP measurements may contribute to risk estimation and inform the timing of serial clinical or echocardiographic examinations. Such a strategy can also address other conditions that affect NP levels and prognosis, including atrial fibrillation, ischemic heart disease, kidney dysfunction, and hypertension. Prospective trials are needed to demonstrate the cost-effectiveness of such a strategy.

Limitations

This study has limitations. The proportion of patients with abnormal NT-proBNP levels may be more frequent in everyday clinical practice than in the present study,²³ but this differing number does not invalidate the favorable prognosis of a normal and stable NT-proBNP level in patients with nonsevere AS. The HRs were also unaffected after adjusting for diuretic treatment. Some patients had unexpectedly high NT-proBNP concentrations, which may have reflected unacknowledged comorbidities at the time of enrollment. Cardiac amyloidosis has been reported to occur in 16% of patients with AS referred for aortic valve replacement.⁵⁴

Although imperfect and often affected by several considerations, aortic valve replacement referral was the best available surrogate marker for AS progression in this study. Criteria for surgical referral have slightly eased since the present data were collected more than a decade ago. Furthermore, the current clinical use of NPs has markedly increased, which is a reason why our study could potentially lead to a more refined strategy regarding serial follow-up. Transcatheter aortic valve implantation was not available during the conduct of this study.

A lack of financial support delayed the blood sample analyses up to 8 years after the end of the SEAS trial. Hence, NP information could not bias any clinical decision-making during the trial. With the contemporary use of NPs in several clinical practices, it may be difficult to repeat the study.