Baseline characteristics
A total of 1028 patients were included in the final analysis. For better comparison patients were divided into tertiles of NT-proBNP, as shown in Table 1. Patients of the higher NT-proBNP tertiles were significantly older (p < 0.001), more often of female sex (p < 0.001) but had lower BMI levels (p = 0.043). Rates of type 2 diabetes mellitus (T2DM, p < 0.001) and arterial hypertension (p < 0.001) were higher in the higher tertiles, but active smoking rates (p < 0.001) and rates of hyperlipidemia were lower (p = 0.047). Despite these differences in rates of traditional CV risk factors levels of low-density lipoprotein cholesterol (LDL-C, p = 0.247), HbA1c (p = 0.157) did not differ between the groups, while systolic blood pressure measurements were significantly higher (p < 0.001). Patients with the highest tertile showed higher rates of previous myocardial infarction (p < 0.001), while stroke rates differed only in the lowest vs. the highest tertile with higher rates in the latter (p = 0.015). Patients of the higher tertiles showed worse renal function, as measured by estimated glomerular filtration rate (eGFR, p < 0.001). While similar rates of previous PAD diagnosis (p = 0.318) were seen, rates of critical limb ischemia were significantly higher in the upper tertiles (p < 0.001). Similarly, ankle-brachial index (ABI, p < 0.001) and toe-brachial index (TBI, p < 0.001) measurements were worse in the higher tertiles. All patients had high treatment rates for CV medication but rates for beta-blockers (p < 0.001), calcium channel blockers (p = 0.376), and diuretics (p < 0.001) were higher in the higher tertiles, while renin angiotensin aldosterone system (RAAS)-blockade rates did not differ (0.223). Rates for statin treatment were lower in the higher tertiles (p = 0.003).
PAD severity and associations with lesion site
Patients with critical limb ischemia (CLI) had significantly higher NT-proBNP levels than intermittent claudication (IC) patients (log-transformed NT-proBNP: 6.33 ± 1.55 vs 5.12 ± 1.30, p < 0.001). NT-proBNP correlated with worse limb perfusion parameters, as significant inverse correlations between ABI (r = − 0.165, p < 0.001) and TBI (r = − 0.276, p < 0.001) were found.
Highest NT-proBNP levels were found in patients with target lesions at below-the-knee (BTK) level (6.31 ± 1.72) followed by multisite target lesions (5.69 ± 1.58). The lowest NT-proBNP levels were found in patients with iliac target lesions (5.07 ± 1.39), while SFA (superficial femoral artery)/popliteal lesions were the third highest with 5.32 ± 1.35. A graphical overview can be seen in Fig. 1. ANOVA between the four defined categories was significant (p < 0.001). Post-hoc LSD analyses were significant between all categories (iliac vs. femoral/popliteal p = 0.037, iliac vs. BTK p < 0.001, iliac vs. multisite p < 0.001, femoral/popliteal vs. BTK p < 0.001, femoral/popliteal vs. multisite p < 0.001, BTK vs. multisite p = 0.003).
Violin plots for log NT-proBNP shown for four lesion sites (iliacal, femoral, BTK, multisite). ANOVA was significantly different between the groups (p < 0.001) and NT-proBNP levels were the highest in BTK endovascular repair (LDS post-hoc, iliac vs. femoral p = 0.037, iliac vs. BTK p < 0.001, iliac vs. multisite p < 0.001, femoral vs. BTK p < 0.001, femoral vs. multisite p < 0.001, BTK vs. multisite p = 0.003).
A binary logistical regression model for NT-proBNP and stenosis location with iliac and femoral/popliteal vs. BTK and multisite stenosis was performed to adjust for site-specific risk factors. Both univariate (OR 1.29, 95% CI 1.17–1.43) and multivariable-adjusted (age, sex, LDL-C, active smoking, HbA1c, arterial hypertension, eGFR and Fontaine stage) binary logistic regression analyses showed significant associations for each increase of one-unit of logarithmically transformed NT-proBNP increase (OR 1.14, 95% CI 1.01–1.30). An overview is presented in Table 2.
Outcome analyses
During the observation period with a median of 4.6 years (25th percentile 3.2 years, 75th percentile 6.2) a total of 336 deaths were registered (overall event rate 32.7%, calculated event rate per year 7.1%). Furthermore, a total of 157 events were classified as CV-deaths (overall CV-death rate 15.3%, calculated event rate per year 3.3%). A clear-cut significant association (log-rank p < 0.001) between tertiles of NT-proBNP and both all-cause and CV-death was seen in Kaplan–Meier curves with the highest event rates in the highest NT-proBNP tertile (Fig. 2).
KM-graphs for tertiles of NT-proBNP. (A) Event rates for higher NT-proBNP tertiles were significantly higher for all-cause mortality (p < 0.001). (B) Similar results were seen for CV-mortality (p < 0.001).
For further analyses, Cox-regression analyses for a continuous increase of one unit logNT-proBNP showed a significant association with all-cause death in crude regression analysis (hazard ratio 1.91, 95% confidence interval 1.77–2.07) and a multivariable-adjusted model (age, sex, diabetes mellitus, LDL-C, arterial hypertension, active smoking, eGFR, history of heart failure, Fontaine stage and BMI) for traditional CV risk factors and heart failure (HR 1.68, 95% CI 1.51–1.87). Likewise, results for CV death showed a significant association in a multivariable-adjusted model (HR 1.86, 95% CI 1.55–2.15).
Similar associations were found in patients without known heart failure (crude HR 1.94, 95% CI 1.77–2.13; multivariable-adjusted HR 1.65, 95% CI 1.47–1.86) as well as those with known heart failure (crude HR 1.92, 95% CI 1.55–2.38, multivariable-adjusted HR 2.03, 95% CI 1.49–2.77). Regarding CV-mortality associations remained significant both for sub analyses of known heart failure (HR 1.88, 95% CI 1.23–2.77) and without known heart failure (HR 1.83, 95% CI 1.52–2.21).
Likewise, significant associations were found in patients with IC (crude HR 1.79, 95% CI 1.61–2.00, multivariable-adjusted HR 1.62, 95% CI 1.42–1.86). Hazard ratios were even higher in patients with CLI (n = 212/252, crude HR 1.82, 95% CI 1.59–2.08, multivariable-adjusted HR 1.79, 95% CI 1.50–2.14). Analyses for CV mortality showed similar associations (IC HR 1.71, 95% CI 1.37–2.13; CLI HR 1.95, 95% CI 1.52–2.51).
A further sub analysis for PAD patients without other known CV comorbidities (excluding known myocardial infarction, known coronary artery disease, and previous stroke) revealed similar results (crude HR 1.94, 95% CI 1.72–2.19, multivariable-adjusted HR 1.69, 95% CI 1.44–1.99). The strongest HR was found in a sub analysis for CV mortality (HR 2.22, 95% CI 1.65–3.00) An overview of the results is shown in Table 3.
With Youden calculation a theoretically optimal cut-off point regarding all-cause mortality with a specificity of 67.8% and sensitivity of 73.1% for NT-proBNP was determined at 212.1 pg/ml. Furthermore, the added predictive power of NT-proBNP was calculated in receiver operating curve (ROC) analyses. The area under the curve (AUC) for the same multivariable model as used in the Cox-regression analyses increased from 0.75 (0.71–0.78) to 0.80 (0.77–0.83). with addition of NT-proBNP (Fig. 3).
ROC graphs for the predictive power of all-cause mortality for addition of NT-proBNP. The AUC improved from 0.75 (0.71–0.78) in the baseline model (age, sex, diabetes mellitus, LDL-C, arterial hypertension, active smoking, eGFR, history of heart failure, Fontaine stage and BMI) with the addition of NT-proBNP 0.80 (0.77–0.83).