Introduction

The ankle-brachial index (ABI) has been used for many years to confirm the diagnosis and evaluate the severity of peripheral artery occlusion disease [1, 2]. ABI < 0.9 has been well proven to be a useful prognostic parameter in different patient populations, such as patients with chronic kidney disease [3,4,5], patients with acute coronary syndrome [6], older patients [7], and patients after transcatheter aortic valve replacement [8].

Recently, the ABI difference (ABID), calculated as |right ABI–left ABI|, has also been proven to be positively correlated with increased mortality and major adverse cardiovascular events in patients undergoing chronic hemodialysis [9] and patients with acute ischemic stroke [10]. However, no study has assessed the ability of the combination of low ABI and high ABID to predict survival. In the present study, we created an ABI score, which was calculated by assigning 1 point for ABI < 0.9 and 1 point for ABID ≥ 0.17. The aim of this study was to examine the ability of ABI score to predict overall and cardiovascular mortality and compare ABI score and other ABI-related parameters on their ability to predict overall and cardiovascular mortality.

Materials and methods

Study population

The study patients were randomly selected from a group of subjects scheduled for echocardiographic examination at our hospital (Kaohsiung Municipal Siaogang Hospital) from March 2010 to March 2012 due to abnormal electrocardiographic findings, cardiomegaly on chest X ray, abnormal cardiac physical examination, dyspnea and chest pain, heart failure, or complicated hypertension or to assess their preoperative cardiac function. Patients with significant aortic or mitral valve disease, atrial fibrillation, inadequate echocardiographic image visualization, or hemodialysis were excluded. In addition, seven patients with failed ABI measurements due to limb amputation (n = 3) or noncooperation (n = 4) were excluded from the final analysis. We did not include all patients consecutively because ABI must begin to be measured within 10 min after the completion of echocardiographic examination. Finally, 941 patients were enrolled. This study conformed with the principles outlined in the Declaration of Helsinki and was approved by the ethics committee of Kaohsiung Medical University Hospital. All study subjects provided written informed consent.

Assessment of ABI

The ABI value was assessed using an ABI-form device (VP1000, Colin, Aichi, Japan), which automatically and simultaneously measures blood pressure in both arms and ankles by an oscillometric method [11, 12]. ABI was calculated as the ratio of the ankle blood pressure over the higher of the two arms’ systolic blood pressures. The ABI measurement was done once in each patient. After obtaining bilateral ABIs, the lower one was used for later analysis. In addition, ABID was calculated as |right ABI–left ABI|.

Collection of demographic and medical data

Age, sex, body mass index (BMI), estimated glomerular filtration rate (eGFR), and comorbid conditions [13, 14] were acquired from medical records or interviews with study subjects. In addition, patient medications, including angiotensin-converting enzyme inhibitors (ACEIs), angiotensin II receptor blockers (ARBs), β-blockers, calcium channel blockers, diuretics, and antiplatelet agents at enrollment, were obtained from medical records. Laboratory data were collected within 1 month of patient enrollment.

Statistical analysis

We used SPSS 22.0 software (SPSS, Chicago, IL, USA) to perform the statistical analysis. Data are presented as the mean ± standard deviation, percentage, or median (25th–75th percentile) for the follow-up period. Categorical and continuous variables between groups were compared by the chi-square test and independent samples t test, respectively. We input the significant variables from the univariable analysis into the multivariable analysis. The times to all overall and cardiovascular mortality events and the covariates of risk factors were modeled using the Cox proportional hazards model. A Kaplan–Meier survival plot was drawn from baseline to the time of death. The incremental value of ABI-related parameters over conventional parameters to assess the risk for overall and cardiovascular mortality was studied by calculating the improvement in the global chi-square value and integrated discrimination improvement (IDI) using STATA 16 (StataCorp., College Station, TX, USA). To find the optimal cutoff values of ABI and ABID as predictors of overall mortality, we created several models using different cutoff values. The chi-square value was used to select the model with the best performance. All tests were two-sided, and the level of significance was established as P < 0.05.

Results

Among the 941 subjects, mean age was 62 ± 14 years. The prevalence of ABI < 0.9 and ABID ≥ 0.17 was 6.1% (n = 57) and 6.8% (n = 64), respectively. In 64 patients with ABID ≥ 0.17, 46% (n = 31) and 54% (n = 33) of patients had ABI ≥ 0.9 and ABI < 0.9, respectively. Of the 57 patients with ABI < 0.9, 42% (n = 24) and 58% (n = 33) had ABID < 0.17 and ABID ≥ 0.17, respectively. There were 853, 55, and 33 patients with ABI scores of 0, 1, and 2, respectively. Table 1 compares the baseline characteristics according to ABI score. There were significant differences in age, prevalence of diabetes mellitus, BMI, systolic and diastolic blood pressures, heart rate, eGFR, percentages of patients taking ACEIs/ARBs, diuretics, and antiplatelet agents, ABI, ABID, prevalence of ABI < 0.9, and prevalence of ABID ≥ 0.17.

Table 1 Comparison of baseline characteristics according to ABI score
Full size table

The mortality data of the study subjects were collected up to December 2018. Mortality information was acquired from the Collaboration Center of Health Information Application, Ministry of Health and Welfare, Executive Yuan, Taiwan. The median follow-up to mortality was 93 months (25th–75th percentile: 86–101 months). The mortality events recognized during the follow-up period included cardiovascular mortality (n = 87) and overall mortality (n = 228).

Using the chi-square value to select the model with the best performance, the model using ABI < 0.9 and ABID ≥ 0.17 had the best performance in predicting the overall mortality. Table 2 shows the predictors of overall and cardiovascular mortality using the Cox proportional hazards model in the univariable analysis in all 941 study patients. The factors associated with increased overall mortality were increased age, BMI, systolic blood pressure, or heart rate; diuretic use; antiplatelet agent use; diabetes; and decreased cholesterol or eGFR. Of the ABI-related parameters, decreased ABI, increased ABID and ABI score, and the combination of ABI < 0.9 and ABID ≥ 0.17 were associated with increased overall mortality. The factors associated with increased cardiovascular mortality were increased age, BMI, systolic blood pressure, or heart rate; diuretic use; antiplatelet agent use; diabetes; and decreased eGFR. Of the ABI-related parameters, decreased ABI, increased ABID and ABI score, and the combination of ABI < 0.9 and ABID ≥ 0.17 were associated with increased cardiovascular mortality.

Table 2 Predictors of overall and CV mortality using Cox proportional hazards model in the univariable analysis in all 941 study patients
Full size table

Table 3 shows the predictors of overall and cardiovascular mortality from the Cox proportional hazards model in the multivariable analysis in all 941 study patients. The variables used for the multivariable analysis of overall mortality included age, systolic blood pressure, heart rate, total cholesterol, diabetes, eGFR, BMI, antiplatelet agent use, and diuretic use. The variables used for the multivariable analysis of cardiovascular mortality included age, systolic blood pressure, heart rate, diabetes, eGFR, BMI, antiplatelet agent use, and diuretic use. The multivariable analysis showed that all ABI-related parameters, including ABI, ABID, the combination of ABI < 0.9 and ABID ≥ 0.17, and ABI score, were significantly associated with overall and cardiovascular mortality (P ≤ 0.009).

Table 3 Predictors of overall and CV mortality using Cox proportional hazards model in the multivariable analysis in all 941 study patients
Full size table

Figure 1 illustrates the Kaplan–Meier curves for overall survival (Fig. 1A) and cardiovascular mortality-free survival (Fig. 1B) in study patients subdivided according to ABI score (both log-rank P < 0.001).

Fig. 1
figure 1

Kaplan–Meier analyses for overall survival (A) and cardiovascular mortality-free survival (B) in the study patients subdivided according to ankle brachial index (ABI) scores of 0, 1, and 2

Full size image

For overall mortality prediction, the univariable model with ABI score (chi-square value, 175.798) outperformed the univariable model with ABI value (chi-square value, 93.823, P < 0.001), ABI < 0.9 (chi-square value, 140.143, P < 0.001), ABID (chi-square value, 101.551, P < 0.001), or ABID ≥ 0.17 (chi-square value, 111.217, P < 0.001). For cardiovascular mortality prediction, the univariable model with ABI score (chi-square value, 97.820) outperformed the univariable model with ABI value (chi-square value, 67.489, P < 0.001), ABI < 0.9 (chi-square value, 87.564, P = 0.001), ABID (chi-square value, 50.247, P < 0.001), or ABID ≥ 0.17 (chi-square value, 54.355, P < 0.001).

Figures 2 and 3 directly compare ABI-related parameters for overall and cardiovascular mortality prediction in the multivariable model, respectively. The basic model consisted of the significant variables in the univariable analysis except the ABI-related parameters. We added ABI-related parameters to the basic model one by one. Among these models, the basic model + ABI score had the highest predictive value for overall (P ≤ 0.001) and cardiovascular mortality prediction (P ≤ 0.049). We also found that adding ABID ≥ 0.17 to the basic model + ABI < 0.9 provided an extra benefit in the prediction of overall mortality (P < 0.001) and cardiovascular mortality (P = 0.012) when compared with the basic model + ABI < 0.9.

Fig. 2
figure 2

Direct comparison of the basic model + ankle brachial index (ABI), basic model + ABI < 0.9, basic model + ABI difference (ABID), basic model + ABID ≥ 0.17, and basic model + ABI score for overall mortality prediction. The variables in the basic model included the significant variables in the univariable analysis except ABI-related parameters

Full size image
Fig. 3
figure 3

Direct comparison of the3 basic model + ankle brachial index (ABI), basic model + ABI < 0.9, basic model + ABI difference (ABID), basic model + ABID ≥ 0.17, and basic model + ABI score for cardiovascular mortality prediction. The variables in the basic model included the significant variables in the univariable analysis except ABI-related parameters

Full size image

When we performed a subgroup analysis in the 872 patients with 0.9 ≤ ABI < 1.3, we found that ABID value and ABID ≥ 0.17 predicted overall mortality, and only ABI value predicted cardiovascular mortality in the multivariable analysis (P ≤ 0.044) (Supplementary Table S1).

Table 4 shows the IDI of ABI parameters (categorical variables) added to the basic model for overall and cardiovascular mortality prediction. As above, the basic model consisted of the significant variables in the univariable analysis except ABI-related parameters. Compared to the basic model alone, adding each ABI parameter (ABI < 0.9, ABID ≥ 0.17, or ABI score) to the basic model provided an extra benefit in the prediction of overall mortality (P ≤ 0.004). Furthermore, adding ABI score and ABI < 0.9 to the basic model also had an extra benefit in the prediction of cardiovascular mortality (P = 0.035 and 0.040, respectively).

Table 4 The IDI of ABI parameters added into the basic model for overall and CV mortality prediction
Full size table

Discussion

This study aimed to evaluate the ABI score (concurrent consideration of ABI < 0.9 and ABID ≥ 0.17) in survival prediction and compared different ABI-related parameters, including ABI, ABID, ABI < 0.9, ABID ≥ 0.17, and ABI score, in the prediction of overall and cardiovascular mortality. We found that the ABI score significantly predicted overall and cardiovascular mortality in the multivariable analysis. In the direct comparison of the univariable models, ABI score was the best a predicting overall and cardiovascular mortality among the five ABI-related parameters (all P < 0.001). In the direct comparison of the multivariable models, the basic model + ABI score was the best model at predicting overall and cardiovascular mortality among the five ABI-related multivariable models (P ≤ 0.049). In addition, in the subgroup analysis of patients with 0.9 ≤ ABI < 1.3, ABID predicted overall mortality in the multivariable analysis.

ABI < 0.9 is a well-established parameter in predicting long-term overall and cardiovascular mortality in different patient groups, including coronary artery disease [15, 16], diabetes [17], chronic kidney disease [3, 18], and hemodialysis [19, 20]. In addition, ABID ≥ 0.15 is reported to be a useful parameter in the prediction of overall mortality in hemodialysis patients, but it might be biased in the prediction of cardiovascular mortality through an effect of peripheral vascular disease [9]. Recently, Han et al. included 2901 patients with acute stroke to examine the ability of ABID to predict short- and long-term outcomes. They found that ABID was associated with poor short-term functional outcomes, the long-term occurrence of major adverse cardiovascular events, and all-cause mortality [10]. However, no study has evaluated whether concurrent consideration of low ABI and high ABID is useful in survival prediction. In the present study, we created a novel ABI score by assigning 1 point for ABI < 0.9 and 1 point for ABID ≥ 0.17 and found that the ABI score not only was a useful parameter in the prediction of long-term overall and cardiovascular mortality but also had the best predictive value for overall and cardiovascular mortality among the five ABI-related parameters in both the univariable and multivariable models. Of our 64 study subjects with ABID ≥ 0.17, there were 31 patients (46%) without ABI < 0.9. Because ABID ≥ 0.17 was a helpful parameter in survival prediction, only considering ABI < 0.9 might not yield good mortality predictions. On the other hand, of the 57 patients with ABI < 0.9, 42% (n = 24) did not have ABID ≥ 0.17. As above, because ABI < 0.9 was a helpful parameter in survival prediction, only considering ABI ≥ 0.17 might not yield good mortality predictions. We reasoned that an ABI score that takes ABI < 0.9 and ABID ≥ 0.17 into consideration concurrently should be able to exert good value in survival prediction. In fact, compared to the other ABI-related parameters, including ABI, ABID, ABI < 0.9, and ABID ≥ 0.17, the ABI score had the best value for long-term overall and cardiovascular mortality prediction in our study subjects.

Low ABI and high ABID have been associated with the presence of peripheral artery disease (PAD) [21]. However, low ABI was reported to be not sensitive enough to detect asymptomatic PAD in the general population [22]. Recently, a high normal ABI was reported to be associated with renal artery intimal thickening and impaired renal function in chronic kidney disease [23]. Hence, some patients with PAD might not be detected by ABI < 0.9. Because increased ABID is associated with the presence of PAD, ABID might be a useful parameter in the detection of PAD in patients with normal ABI [10, 11]. Therefore, ABI score, concurrently considering ABI < 0.9 and ABID ≥ 0.17, might have the potential to detect more patients with PAD than ABI < 0.9 or ABID ≥ 0.17 alone. Patients with higher ABI scores might have a higher percentage of PAD and concomitant atherosclerosis and thus might have a higher mortality.

Recently, high ABID was found to be associated with increased all-cause mortality in patients with acute ischemic stroke without PAD indicated by ABI ≥ 0.9 [10]. Consistent with that finding, we found that increased ABID was a useful predictor of overall mortality in patients with 0.9 ≤ ABI < 1.3 on multivariable analysis. In addition, in the study of Lin et al. including chronic hemodialysis patients, ABID ≥ 0.15 was not associated with increased mortality in patients with ABI ≥ 0.9 [9]. Our present study similarly found that ABID ≥ 0.17 could not predict mortality in patients with 0.9 ≤ ABI < 1.3 in the multivariable analysis.

Study limitations

There were several limitations to this study. First, our study patients were enrolled from those scheduled for echocardiographic examination, so their baseline characteristics were very heterogeneous. Therefore, our results must be cautiously interpreted when applied to a homogeneous study group, such as chronic kidney disease, diabetes, hypertension, chronic heart failure, and generally healthy patients. Second, information on LDL-C, HDL-C, smoking history, and medications for diabetes and dyslipidemia was lacking, so we could not analyze the impact of such parameters on mortality. Third, we only aimed to evaluate the mortality events, so nonfatal events were not studied. Fourth, because there has been no large-scale study to validate the optimal cutoff value of ABID, we used the chi-square value to select the best cutoff value of ABID in predicting overall mortality. The best cutoff value of ABID in mortality prediction in might be different in different study populations. Fifth, surgery might have affected the patients’ prognosis. Three patients underwent amputation of the leg and were excluded from the study, but we had no the other operation data. Sixth, there might be a selection bias because we excluded patients who could not begin the ABI measurement within 10 min after the completion of the echocardiographic examination. Seventh, patients with insignificant valvular heart diseases, ischemic heart diseases, and dilated and hypertrophic cardiomyopathy were not excluded from our cohort, which might have affected the outcome prediction. Finally, although ABI < 0.9 and ABID ≥ 0.17 make different contributions to mortality prediction, we arbitrarily assigned one point for ABI < 0.9 and one point for ABID ≥ 0.17 when calculating the ABI score. It should be noted that the hazard ratios of ABI < 0.9 (5.926) and ABID ≥ 0.17 (5.033) in predicting overall mortality were similar.

Conclusions

In this study, we found that the ABI score could significantly predict overall and cardiovascular mortality on multivariable analysis. Furthermore, when we directly compared the univariable models, the model with the ABI score was the best at predicting overall and cardiovascular mortality among the five ABI-related parameters. In the direct comparison of the 5 multivariable models, the basic model +ABI score was the best at predicting overall and cardiovascular mortality. Hence, the ABI score, a combination of ABI < 0.9 and ABID ≥ 0.17, should be calculated for better mortality prediction.