This is a retrospective cohort study of adults (age ≥18 years) with CHD who had ≥2 NT-proBNP measurements in the outpatient clinic with ≥1 year between measurements between January 1, 2003 and December 31, 2023. The patients with unrepaired/palliated cyanotic CHD were excluded. We then divided the patients into 2 groups (derivation cohort and validation cohort) using a random assignment based on the last digit of their research identification number (even vs odd numbers). Within the derivation and validation cohorts, we classified the patients based on physiology as follows: (1) Biventricular physiology with systemic morphologic left ventricle (BiV-SLV), (2) Biventricular physiology with systemic morphologic right ventricle (BiV-SRV), (3) Fontan physiology. CHD severity groups were defined as mild, moderate, and complex CHD based on the criteria stipulated in the guidelines for management of adults with CHD.

(1) Determine the correlates of NT-proBNP assay, and the association between baseline and longitudinal NT-proBNP levels and all-cause mortality in derivation cohort. (2) Compare the prognostic performance of NT-proBNP level to conventional clinical indices (echocardiographic indices, invasive hemodynamic indices, New York Heart Association [NYHA] functional capacity, and aerobic capacity) in derivation cohort. (3) Test the prognostic performance of NT-proBNP levels and derived thresholds in the validation cohort.

The first and second NT-proBNP measurements performed in the outpatient clinic within the study period were designated as NT-proBNP #1 and #2, respectively. The temporal change in NT-proBNP levels was calculated as absolute ∆_NT-proBNP (NT-proBNP #2- NT-proBNP #1) and relative ∆_NT-proBNP ([NT-proBNP #2– NT-proBNP #1] ÷ NT-proBNP #1 × 100), and temporal increase in NT-proBNP level was defined as relative ∆_NT-proBNP >0. Clinical indices, echocardiographic indices, laboratory blood tests, cardiopulmonary exercise test, and cardiac catheterization data obtained within 12 months from the baseline NT-proBNP (NT-proBNP #1) measurement were used to define the baseline characteristics of the cohort.

Data were presented as mean ± standard deviation, median (Q1, Q3), and count (%). The NT-proBNP data were log transformed for all analyses because of skewed distribution. The correlates of baseline NT-proBNP were assessed using linear regression analysis. The variables included in the linear regression models were chosen based on clinical relevance, and included demographic/anatomic indices, comorbidities, biomarkers of hepatorenal function, and echocardiographic indices. The covariates with P < 0.1 on univariable analysis were used to create multivariable models using stepwise backwards selection with P < 0.1 as the criterion for a variable to remain in the model. Subgroup analyses were performed for patients with BiV-SLV, BiV-SRV, and Fontan physiology, as well as patients with cardiac catheterization.

All-cause mortality were assessed as time-to-event outcome and ascertained by review of medical records and Accurint mortality database using the date of NT-proBNP #2 assay as “time 0.” The relationship between NT-proBNP levels (both baseline and relative ∆_NT-proBNP) and all-cause mortality was assessed using Cox regression analysis, using similar criteria for covariate selection as described above. The incremental prognostic value of temporal change in NT-proBNP relative to baseline NT-proBNP was assessed using c-statistic comparison. The prognostic performance of NT-proBNP was compared to the conventional clinical indices such as echocardiographic indices (systemic ventricular global longitudinal strain, systemic ventricular ejection fraction, nonsystemic ventricular free wall strain, lateral E/e’), cardiac catheterization indices (right atrial [RA] pressure, pulmonary artery (PA) mean pressure, PA wedge pressure), functional capacity (NYHA class) and exercise tests indices (peak oxygen consumption).

We used receiver operating characteristic curve (ROC) analysis to determine the optimal cutoff point for baseline NT-proBNP (modeled as log NT-proBNP) to detect all-cause mortality based on Youden index which provided optimal balance between sensitivity and specificity. The robustness of the model was assessed using the area under the curve (AUC). The corresponding NT-proBNP level was rounded to the nearest ten for ease of clinical application. The cumulative incidence of mortality was estimated using Kaplan-Meier analysis and compared using log rank test.

The Cox model from the derivation cohort was used to fit the data from the validation cohort. The discrimination power of the models in both cohorts was assessed by comparing the c-statistics from the derivation and validation cohorts. Missing data were managed using conditional imputation. All statistical analyses were performed with BlueSky Statistics software (version. 7.10; BlueSky Statistics LLC, Chicago, IL), and JMP statistical software (version 17.1.0, JMP Statistical Discovery LLC, NC). P value <.05 was considered to be statistically significant for all analyses.

Supplementary Figure S1 shows a flowchart of patient selection, and Supplementary Figure S2 compares the baseline characteristics of the CHD patients (excluding those with on repair/cyanotic heart disease) with serial NT-proBNP measurement ( n = 4,307) vs those without serial NT-proBNP measurements ( n = 3,309). There was no significant difference in age, sex, body mass index, CHD severity, and CHD physiologic group between the 2 groups.

Of 4,307 patients who met the study inclusion criteria, 2,154 (50%) and 2,153 (50%) patients were assigned to the derivation cohort and validation cohort, respectively. Supplementary Table S1 shows comparison of baseline characteristics and CHD diagnoses between derivation and validation cohorts. Both groups had similar baseline characteristics and CHD diagnoses.

Of 2,154 patients, 1,837 (85%) had BiV-SLV, 144 (7%) had BiV-SRV, and 173 (8%) had Fontan physiology. Table 1 shows comparison of the clinical characteristics and hemodynamic indices across the 3 groups. Compared to the BiV-SLV, patients in BiV-SRV and Fontan groups were younger, more likely to have pacemakers, and had worse hepatorenal function and aerobic capacity (lower peak oxygen consumption). The BiV-SRV and Fontan groups also had worse systemic atrial dilation and dysfunction, worse systemic ventricular systolic dysfunction, and higher cath-derived filling pressures compared to the BiV-SLV groups ( Table 1 ).

Correlates of baseline NT-proBNP levels

The median baseline NT-proBNP was 189 (77; 538) pg/mL and was different across the 3 CHD groups, with the highest level observed in patients with BiV-SRV ( Figure 1 A). Using the BiV-SLV as the reference group, baseline NT-proBNP level was higher in the BiV-SRV (181 [74, 523] vs 326 [134, 778] pg/mL, P <.001) but comparable to the levels in the Fontan group (181 [74, 523] vs 177 [93, 493] pg/mL, P =.48). NT-proBNP levels increased with older age groups (134 [63, 370] in age group <40 years vs 233 [94, 619] pg/mL in age group 40-59 years vs 456 [188, 1087] pg/mL in age group ≥60 years, P <.001) ( Figure 1 B). NT-proBNP levels were also higher in females compared to males (199 [85, 536] vs 170 [63, 513] pg/mL, P = 0.03) ( Figure 1 C).

Scatter dot plot comparing baseline NT-proBNP levels based on CHD groups (BiV-SLV vs BiV-SRV vs Fontan physiology) A, based on age groups (age <40 years vs 40-59 years vs ≥60 years) B, and based on sex (male vs Female), C. Footnote: Line and error bars signify median and interquartile range. P values were derived from comparisons across all groups. In A, pairwise comparison was performed using BiV-SLV as the reference group, and * signifies statistically significant difference between BiV-SLV vs BiV-SRV. In B, pairwise comparison was performed using age group <40 years as the reference group, and *signifies statistically significant difference. BiV-SLV, biventricular physiology with morphologic systemic left ventricle; BiV-SRV, biventricular physiology with morphologic systemic right ventricle; NT-proBNP, N terminal pro B type brain natriuretic peptide.

Supplementary Table S2 show univariable linear regression models assessing correlations between baseline NT-proBNP level and clinical indices. On multivariable analysis, higher baseline NT-proBNP level was associated with older age, female sex, BiV-SRV, atrial fibrillation, hepatorenal dysfunction, and echocardiographic indices (atrial dysfunction and biventricular systolic dysfunction) ( Table 2 ). Similar consistent correlates of baseline NT-proBNP level were observed on subgroup analysis of patients with BiV-SLV, BiV-SRV, and Fontan physiology, as well as patients with cardiac catheterization data ( Table 2 ).

Table 2

Multivariable linear regression models showing correlates of baseline NT-proBNP.

All Patients ( n = 2,154)
	β ± SE	P
Demographic indices
Age (per 5 years)	0.10 ± 0.01	< 0.001
Male sex	−0.09 ± 0.03	0.004
CHD group
BiV-SLV	Reference
BiV-SRV	−0.23 ± 0.06	< 0.001
Fontan	−0.06 ± 0.06	0.28
Comorbidities
Atrial fibrillation	0.22 ± 0.04	< 0.001
Chronic kidney disease	0.24 ± 0.07	< 0.001
Cirrhosis	0.21 ± 0.09	0.02
Laboratory data
MELD-XI score	0.06 ± 0.01	< 0.001
Echocardiographic indices
LA reservoir strain (%)	−0.03 ± 0.01	0.001
Systemic ventricular GLS (%)	−0.10 ± 0.02	< 0.001
Estimated RA mean pressure (mmHg)	0.08 ± 0.03	0.02
Nonsystemic ventricular FWS (%)	−0.05 ± 0.02	0.01
BiV-SLV ( n = 1837)
	β ± SE	p
Demographic indices
Age (per 5 years)	0.08 ± 0.01	< 0.001
Male sex	−0.12 ± 0.01	0.007
Comorbidities
Atrial fibrillation	0.23 ± 0.04	< 0.001
Chronic kidney disease	0.27 ± 0.07	< 0.001
Laboratory data
MELD-XI score	0.05 ± 0.01	< 0.001
Echocardiographic indices
LA (systemic) reservoir strain (%)	−0.03 ± 0.01	0.003
Systemic ventricular GLS (%)	−0.05 ± 0.01	< 0.001
Estimated RA pressure (mmHg)	0.06 ± 0.02	0.008
Nonsystemic ventricular FWS (%)	−0.05 ± 0.02	0.04
BiV-SRV ( n = 144)
	β ± SE	p
Demographic indices
Age (per 5 years)	0.09 ± 0.03	< 0.001
Comorbidities
Atrial fibrillation	0.16 ± 0.09	0.06
Echocardiographic indices
LA (systemic) reservoir strain (%)	−0.07 ± 0.02	0.009
Systemic ventricular GLS (%)	−0.12 ± 0.04	0.02
Estimated RA mean pressure (mmHg)	0.08 ± 0.03	0.03
Fontan ( n = 173)
	β ± SE	p
Comorbidities
Atrial fibrillation	0.39 ± 0.10	0.002
Cirrhosis	0.23 ± 0.10	0.02
Laboratory data
MELD-XI score	0.10 ± 0.03	0.001
Echocardiographic indices
LA (systemic) reservoir strain (%)	−0.03 ± 0.01	0.01
Systemic ventricular GLS (%)	−0.15 ± 0.05	0.001
Patients with Cath data ( n = 530)
	β ± SE	p
Demographic indices
Age (per 5 years)	0.09 ± 0.03	0.002
Comorbidities
Atrial fibrillation	0.22 ± 0.10	0.04
Laboratory data
MELD-XI score	0.08 ± 0.02	0.02
Echocardiographic indices
Systemic ventricular GLS (%)	−0.10 ± 0.02	< 0.001
Cardiac catheterization data
RA (nonsystemic) mean pressure (mmHg)	0.05 ± 0.02	0.003
PAWP, mmHg (mmHg)	0.04 ± 0.02	0.04

Only gold members can continue reading. Log In or Register to continue

Share this:

• X
• Facebook

Related

Related posts:

1. Design and rationale of the GUARD-OAC: A randomized controlled trial evaluating proton pump inhibitor cotherapy for gastrointestinal protection in patients requiring direct oral anticoagulants
2. Low levels of atherogenic lipoproteins and incident atherosclerotic cardiovascular disease: A pooled cohort primary prevention study
3. Comparative performance of large language models on cardiovascular certification simulation exam
4. Accuracy of site versus core laboratory interpretations of right heart catheterization hemodynamics

Stay updated, free articles. Join our Telegram channel

Join