Author + information
- Received January 14, 2013
- Revision received April 7, 2013
- Accepted April 22, 2013
- Published online August 6, 2013.
- Kevin T. Nead, MPhil∗,
- John P. Cooke, MD, PhD∗,
- Jeffrey W. Olin, DO† and
- Nicholas J. Leeper, MD‡∗ ()
- ∗Division of Cardiovascular Medicine, Stanford University School of Medicine, Stanford, California
- †Zena and Michael A. Wiener Cardiovascular Institute, Mount Sinai School of Medicine, New York, New York
- ‡Division of Vascular Surgery, Stanford University School of Medicine, Stanford, California
- ↵∗Reprint requests and correspondence:
Dr. Nicholas J. Leeper, Division of Vascular Surgery, Stanford University, 300 Pasteur Drive, Room H3634, Stanford, California 94305-5642.
Objectives The aim of this study was to determine whether use of an alternative ankle-brachial index (ABI) calculation method improves mortality risk prediction compared with traditional methods.
Background The ABI is used to diagnose peripheral arterial disease (PAD) and to identify those at risk for cardiovascular events. Traditionally, the ABI is calculated with the higher of the dorsalis pedis and posterior tibial ankle arteries. Studies directly comparing calculation methods are limited.
Methods The ABI was calculated at baseline in 1,413 study participants undergoing nonemergent coronary angiography subsequently followed for all-cause and cardiovascular mortality. There were 224 individuals assigned to the traditional-PAD group (ABI <0.90) with the traditional ABI method. Of those remaining, an alternative ABI method using the lower of the 2 ankle pressures assigned 282 patients to the alternative-PAD group. The 862 individuals not assigned to PAD by either method were the no-PAD group.
Results There were 163 mortalities during a median follow-up of 5.0 years. Adjusted Cox regression models showed that the alternative-PAD group had an increased risk for all-cause (hazard ratio [HR]: 1.49; 95% confidence interval: 1.01 to 2.19) and cardiovascular mortality (HR: 3.21; 95% confidence interval: 1.53 to 6.37) versus the no-PAD group. Additionally, in the no-PAD group, there was an 11% (HR: 1.11; 95% confidence interval: 1.05 to 1.17) increased risk of all-cause mortality/1-mm Hg increased difference between the left and right brachial systolic pressures.
Conclusions The implementation of an alternative ABI method and use of the brachial difference identifies individuals at an increased risk for mortality who are currently missed with traditional ABI methods. Current ABI protocols might need to be evaluated.
Peripheral arterial disease (PAD) results from atherosclerosis and narrowing of the arteries in the distal extremities. Estimates of PAD prevalence demonstrate a wide range, from 4% in general populations to 30% among primary care practices (1,2). Individuals with PAD are at an increased risk for myocardial infarction, aortic aneurysm rupture, ischemia, and amputation (3) with a 6-fold increased risk of death from cardiovascular causes (4,5). This cardiovascular morbidity and mortality has been demonstrated even in PAD patients without clinical symptoms (4). Although aggressive medical treatment of PAD has been shown to reduce cardiovascular morbidity and mortality, many patients are not recognized and, as a result, are not adequately treated (2,6).
The ankle-brachial index (ABI) provides an objective clinical measure for the diagnosis of PAD, with normal ABI defined as at least 1.00 to 1.40 and PAD classified as an ABI of <0.90. These criteria have been validated against peripheral angiography to calculate the sensitivity and specificity of the ABI as a diagnostic tool for PAD. Although the specificity has been shown to be reproducibly high (>95%) (7–10), some validation studies have demonstrated low sensitivity (7,8). Therefore, methods to improve the sensitivity of the ABI, without a concomitant reduction in specificity, might improve risk prognostication.
According to current guidelines and a recent Scientific Statement from the American Heart Association (11), the preferred method to calculate the ABI requires measuring the systolic pressures of the left and right brachial arteries in addition to the dorsalis pedis and posterior tibial arteries in each leg. Then, for each leg, the higher of the dorsalis pedis and posterior tibial pressures is divided by the higher brachial pressure, and the lower overall ABI of the 2 legs is used to determine PAD status. Here we assess the implementation of an alternative measure of ABI, by instead using the lower of the dorsalis pedis and posterior tibial pressures and examine whether individuals reclassified into the PAD group with this alternative measure are at an increased risk of mortality. Additionally, we determine whether there is evidence to support the direct evaluation of the brachial difference in risk prediction among individuals undergoing ABI testing.
The GenePAD (Genetic Determinants of Peripheral Arterial Disease) study comprises individuals (n = 1,755) who underwent an elective, nonemergent coronary angiogram for angina, shortness of breath, or an abnormal stress test at Stanford University and Mount Sinai Medical Centers between January 1, 2004, and March 1, 2008, as previously described (12–14).
Participants were included in the study sample if complete data were available on all covariates (n = 1,413). Individuals without both a dorsalis pedis and posterior tibial systolic pressure recorded in at least 1 leg were excluded (n = 5), because calculation of both a traditional and alternative ABI value would not be possible. Remaining eligible patients with an ABI >1.40 were excluded (n = 40) (15). With these criteria, 1,368 patients were included.
The outcomes of interest in this analysis were death from any cause and death from cardiovascular causes including myocardial infarction, cardiac arrest, stroke, heart failure, or aneurysm rupture. Ascertainment of mortality and cause of death was achieved through phone or postal communication, medical record review, and the Social Security Death Index. New mortalities were identified through March 31, 2012.
The use of lipid-lowering and anti-hypertensive medications was evaluated by direct medication inventory. Diabetes status was classified as self-reported use of insulin or oral hypoglycemic agents. Hypertension was defined as a systolic blood pressure of ≥160 mm Hg or use of anti hypertensive medications. Creatinine and total and high-density lipoprotein (HDL) cholesterol levels were measured at the time of coronary angiography. An experienced cardiologist who was blinded to participant details evaluated coronary angiograms. Hemodynamically significant coronary artery disease (CAD) was defined as >60% stenosis (16,17).
Before the coronary angiogram, posterior tibial, dorsalis pedis, and brachial artery systolic pressures were measured with a 5-MHz Doppler ultrasound. Calculation of the ABI with the traditional and alternative methods is illustrated in Figure 1. The traditional ABI was calculated for each leg by dividing the higher of the posterior tibial or dorsalis pedis pressures by the higher of the left or right brachial systolic pressures. Then, the leg with the lower ABI measurement was recorded as the traditional ABI value. The alternative ABI was calculated for each leg by dividing the lower of the posterior tibial or dorsalis pedis pressures by the higher of the left or right brachial systolic pressures. Then, the leg with the lower ABI measurement was recorded as the alternative ABI value. For each method the index leg was defined as the leg with the lower ABI, and data from the contralateral leg were not included in the analysis.
PAD group classification
Study participant classification into mutually exclusive groups is represented in Figure 2. Patients with an ABI of <0.90 in either leg with the traditional method were classified as having traditional-PAD. Of those remaining (who therefore had an ABI of ≥0.90 in both legs with the traditional method), individuals with an ABI of <0.90 in either leg with the alternative method were classified as having alternative-PAD. Patients without PAD as classified by either method were used as the no-PAD reference group. Among individuals in the no-PAD group, we calculated the brachial difference as the difference between the left and right brachial systolic pressures.
Baseline patient characteristics were calculated for the 3 PAD status groups and compared with a t test or chi-square test. For all survival analyses, the follow-up time was defined as the period between the enrollment interview and the last confirmed follow-up or date of death.
Hazard ratios (HRs) were calculated with adjusted Cox proportional hazard models to compare all-cause and cardiovascular mortality risk in the traditional-PAD, alternative-PAD, and no-PAD groups. Additionally, individuals with severely decreased ABI values of <0.40 have been shown to be at a further increased risk for PAD-related sequelae (18). To examine the ability of the alternative ABI method to stratify mortality risk, we compared the risk of individuals with severely decreased ABIs with the traditional method with the risk of any individual in the full study cohort with an ABI <0.40 only with the application of the alternative method. Similarly, we compared the risk of individuals with mildly to moderately decreased ABIs (0.40 ≤ABI <0.90) with the traditional method with the risk of individuals with moderately decreased ABIs only when using the alternative method. All survival analyses were adjusted for age, sex, race, smoking history (ever or never), BMI, hypertension, use of lipid-lowering medications at enrollment, diabetes, total cholesterol, HDL cholesterol, creatinine and CAD. Proportional-hazards assumptions were evaluated by Schoenfeld's residuals tests.
Next, we calculated the C-index to quantify the integrated sensitivity and specificity of the traditional- and alternative-PAD classifications in predicting all-cause and cardiovascular mortality. Therefore, the traditional and alternative ABI methods were separately applied to the full study cohort to ascertain the respective discriminatory capacity of each method in all comers. In survival analysis, the C-index interpretation is equivalent to the area under the ROC curve or c-statistic, while allowing for censored data (19).
We further examined whether the brachial difference provided additional prognostic information in predicting mortality risk among those with no detectable PAD by either the traditional or alternative method. This variable was calculated as the difference between the left and right brachial systolic pressures and examined as a continuous variable and as a binary variable of <10 mm Hg or ≥10 mm Hg. The brachial difference was calculated in 200 patients in the no-PAD group after excluding individuals with only 1 brachial pressure due to intravenous placement or amputation.
Tests were considered significant if the 2-sided p value was <0.05. All analyses were performed with Stata (version 12.0, StataCorp, College Station, Texas). Study data were collected and managed with REDCap electronic data capture tools hosted at Stanford University (20).
Study population characteristics
There were 224 individuals in the traditional-PAD group, which consisted of those categorized as having PAD (ABI <0.90) with the traditional ABI method. Among those remaining, the alternative ABI method placed 282 individuals into the alternative-PAD group. The no-PAD group consisted of the remaining 862 participants not classified as having PAD with the traditional or alternative ABI methods. The mean higher and lower ABI for the index leg in each group is presented in Table 1. The prevalence of PAD when applying each method to the full study cohort was 16.4% for the traditional ABI method and 37.0% for the alternative ABI method. There were 163 mortalities (12%), of which 47 were known to be from cardiovascular causes, during a median follow-up period of 5.0 years (interquartile range: 4.0 to 6.3 years).
Enrollment characteristics of the 1,368 participants in this study according to PAD status are displayed in Table 2. The alternative-PAD group significantly differed from the traditional-PAD group with a lower mean age, lower total cholesterol and creatinine levels, and a significantly lower proportion of hypertensive individuals, smokers, diabetic individuals, and individuals with CAD. Conversely, the only significant differences observed between the alternative-PAD group as compared with the no-PAD group were a higher mean age, higher HDL cholesterol, and a greater proportion of women.
For all-cause and cardiovascular mortality, individuals in the traditional-PAD group had significantly increased risk compared with those in the no-PAD group (Table 3). Individuals in the alternative-PAD group (who would have been considered “normal” by the traditional ABI method) also had significantly elevated risk compared with the no-PAD group. Notably, the estimated risk in the traditional- and alternative-PAD groups were equally elevated and did not significantly differ from each other for either all-cause (p = 0.506) or cardiovascular mortality (p = 0.653).
Patients with a severely decreased ABI (<0.40) with the traditional ABI method as well as those with an ABI <0.40 only when using the alternative method had a >3-fold increased risk for all-cause mortality (Table 4). This risk did not significantly differ between ABI calculation methods (p = 0.712). Similarly, patients placed in the mild/moderate ABI category (0.40 ≤ABI <0.90) with the traditional method and individuals moved into this group only with the application of the alternative method were at a significantly increased risk for death from any cause with an HR that did not statistically differ between these 2 groups (p = 0.618).
We then directly compared the discriminatory capacity of the traditional and alternative ABI methods among the entire cohort. The C-index demonstrated that there was no loss of integrated sensitivity and specificity in predicting all-cause or cardiovascular mortality with the implementation of the alternative ABI method to designate PAD status as compared with the traditional method among all comers (Table 5). Because the cardiovascular mortality event rate was so high among those in the alternative-PAD group (Table 3) and they would have been inappropriately classified as low-risk with the traditional method, the classification of PAD with the alternative ABI method resulted in a significantly greater C-index for cardiovascular mortality (p = 0.013) that was maintained even with adjustment for CAD status (p = 0.021).
Among the no-PAD group, who would be classified as lowest risk with both ABI methods, we found that the brachial difference predicted an 11% (95% confidence interval: 1.05 to 1.17; p < 0.001) increase in all-cause mortality risk/1-mm Hg difference between the left and right brachial systolic pressures (n = 200). Furthermore, individuals with a brachial difference of at least 10 mm Hg (n = 16) demonstrated a nearly 4-fold increased risk of all-cause mortality compared with those with a brachial difference of <10 mm Hg (HR: 3.96; 95% confidence interval: 1.21 to 13.02; p = 0.023).
The risk of all-cause and cardiovascular mortality did not significantly differ between participants excluded (n = 387) and included (n = 1,368) in the study-cohort, and tests for interaction showed that the reported results were not significantly different according to sex or racial group. Sensitivity analyses adjusting for clopidogrel and, separately, removing heart failure and aneurysm rupture as causes of cardiovascular mortality, yielded results consistent with our primary data. Schoenfeld's residuals tests demonstrated that the proportional hazards assumption was met for all models.
The ABI is a validated and widely used measure to detect the presence of PAD. Currently, the ABI is recommended as the first-line test for the diagnosis of PAD among individuals with symptoms and clinical findings suggestive of PAD (11). Because PAD is considered a coronary heart disease risk equivalent (21), the ABI is also a simple tool that can identify those at risk of major adverse cardiovascular events (MACE). The current methods used to identify the presence or absence of PAD rely on criteria that might be insensitive (7,8) and have been suggested to falsely classify some at-risk individuals as “normal.” In this study, we compared participants classified as having PAD (ABI <0.90) with the traditional ABI calculation method with patients that were classified as having PAD only when using an alternative ABI calculation method, to see whether we could enhance the prognostic utility of this widely used test.
Individuals who were reclassified as having PAD only when implementing the alternative ABI calculation had a mortality risk that was significantly greater than those without PAD by either method. This is important, because these patients would normally be classified as “free-of-disease” and would not meet criteria to have risk-reducing therapies initiated. Additionally, the risk in this alternative group was as high as the risk for those individuals classified as having PAD with the traditional ABI method. These findings were consistent for both all-cause and cardiovascular mortality as well as when examining categories of ABI severity. Finally, among those without PAD detected by either ABI method, the brachial difference was a significant predictor of all-cause mortality risk, which is consistent with previous studies demonstrating an association between increased brachial difference and markers of subclinical atherosclerosis (22). This finding emphasizes that vascular disease in any anatomic distribution might provide important prognostic information and should not be ignored.
Although a similarly increased risk of mortality was demonstrated for the traditional- and alternative-PAD groups compared with the no-PAD group, examination of baseline characteristics revealed that the alternative-PAD group had far fewer established cardiovascular risk factors, such as significantly lower rates of hypertension, smoking, diabetes, and CAD, than the traditional-PAD group. This is an important finding, because their pre-test probability of disease is low, and it is probable that a “normal” traditional ABI would lead many practitioners to “rule-out” disease, according to Bayesian principles. It is unclear why individuals in this group resemble lower-risk individuals and harbor disease in an alternative anatomic pattern, but it is clear that their risk is significantly elevated and that they are often clinically ignored. Future studies are required to understand whether these individuals possess other unmeasured risk factors (e.g., premature endothelial dysfunction) that accelerate their risk of death or whether isolated/focal tibial disease might reflect a more aggressive variant of PAD.
When applied to all-comers, the implementation of the alternative ABI calculation method will often result in a lower ABI value compared with the traditional method. Therefore, a greater number of individuals will be classified as having PAD with the alternative method, making it a more sensitive predictor of risk. Although this study shows that the use of the alternative method detects a group of individuals at increased risk for mortality that was not previously identified, we wished to determine whether the implementation of this more sensitive measure in the full study cohort resulted in a net loss of integrated sensitivity and specificity for predicting MACE. We evaluated this with the C-index, which provided evidence against any loss of integrated sensitivity and specificity for all-cause mortality. Indeed, because the risk of death was so high in the alternative-PAD group, and this group was falsely classified as normal by the traditional ABI method, we found that the ability to predict future cardiovascular death was significantly increased when using the alternative ABI method to define PAD status in the entire cohort. Thus, our concern that re-defining the interpretation of this test to capture more at-risk individuals would enhance sensitivity at an unacceptable cost in specificity (for MACE) seems unfounded.
The current study differs from previous reports that attempted to identify more sensitive methods to diagnose PAD (23,24) or prognosticate the risk of future events (25). Use of the lower ankle pressure in calculating the ABI has been shown to predict a higher prevalence of PAD while having greater sensitivity for atherosclerotic disease (23,24). Although a recent American Heart Association Scientific Statement does acknowledge that the lower ankle pressure can be used as a prognostic marker for cardiovascular events, it gives this approach a Level C recommendation and raises concerns about “overdiagnosis” and “unnecessary tests” that might be associated with less-specific ABI methods (11). To our knowledge, only 1 previous study has examined the implementation of this alternative ABI calculation in predicting clinical outcomes (25). Their study reported an increased risk of cardiovascular events among those subjects assigned to the “Suspected PAD” group by an alternative ABI method, similar to the findings reported here. In contrast to the current report, however, their study evaluated a significantly younger patient population with fewer comorbidities and did not observe a statistically significant difference in risk for cardiovascular events (death from cardiovascular causes, nonfatal myocardial infarction, or nonfatal stroke) in fully adjusted Cox-regression models. These findings might be explained in part by the fact that both the presence and systemic sequelae of PAD increase significantly with age and thus might not have been fully manifest in the younger cohort from the prior study (26). In the current report, the alternative method correctly prognosticated risk after controlling for all comorbidities and did so without a reduction in net sensitivity and specificity, which has not been previously shown.
Recently published guidelines (11) for calculating the ABI use the higher brachial pressure and higher of the dorsalis pedis and posterior tibial pressures in each leg. The higher brachial pressure is uniformly used to limit inaccurate ABI calculations among individuals with lower brachial pressures in 1 arm due to subclavian or axillary artery stenosis. However, the rationale for using the higher ankle pressure is less clear, and direct comparisons of different methods are limited. It is obvious that single vessel coronary disease is associated with cardiovascular events and mortality in CAD (27) and that the severity of a stenosis does not always predict the likelihood of future plaque rupture (28). It could be argued that single vessel disease in the lower vasculature should not be overlooked, as when using the higher ankle pressure and dismissing the lower pressure recorded in a potentially diseased vessel. Although concerns about overdiagnosis, false positivity, and resource use with more sensitive ABI methods are valid (11), consideration should be given to the potential value of identifying at-risk individuals earlier in the disease process. The classical ABI is currently recommended as the first-line noninvasive test for the diagnosis of suspected PAD as well as a screening test for those considered to be at high risk for cardiovascular disease (11). When taken in the context of prior studies (25), the current findings suggest that the ABI interpreted with the alternative method might have an additional role in the risk stratification of subjects with an indeterminate risk profile, such as patients with few comorbidities but symptoms severe enough to warrant cardiac catheterization.
The current study was undertaken in a cohort of individuals undergoing coronary angiography and might not be generalizable to other populations. Additionally, we were unable to evaluate the accuracy of PAD diagnosis with the traditional and alternative ABI methods against a gold standard such as peripheral angiography. Larger cohorts will be needed to confirm the utility of the alternative ABI method by ABI severity, because this analysis had limited events within these subgroups. Additionally, future studies are needed to examine the additive value of the alternative ABI method compared with established risk models.
Peripheral arterial disease is associated with an increased risk of MACE. The ABI is widely used to diagnose PAD and to inform the medical management of patients. Here we show that the implementation of an alternative ABI calculation method identifies a large proportion of individuals at an increased risk for all-cause and cardiovascular mortality. These individuals would not be identified with traditional ABI methods and otherwise might be ignored due to their clinical risk profile. Additionally, we provide evidence that the evaluation of the brachial difference among patients not classified as having PAD might provide useful data for risk prediction that is not directly evaluated under current guidelines. These methods improved the prognostic utility of the ABI without a net impairment in sensitivity and specificity. These findings should be replicated in lower-risk general populations, and randomized controlled trials will be needed to determine whether the use of the lower ankle pressure in ABI calculations improves clinical outcomes. The results of this study suggest that protocols commonly employed for this widely used diagnostic test might need to be re-evaluated.
This work was funded by grants from the National Institutes of Health (K12HL087746 to Dr. Cooke) and the American Heart Association (10BGIA3290011 to Dr. Leeper). The authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- ankle-brachial index
- body mass index
- coronary artery disease
- high-density lipoprotein
- hazard ratio
- major adverse cardiovascular event(s)
- Received January 14, 2013.
- Revision received April 7, 2013.
- Accepted April 22, 2013.
- American College of Cardiology Foundation
- Selvin E.,
- Erlinger T.P.
- Newman A.B.,
- Shemanski L.,
- Manolio T.A.,
- et al.
- Nead K.T.,
- Zhou M.,
- Caceres R.D.,
- Olin J.W.,
- Cooke J.P.,
- Leeper N.J.
- Fowkes F.G.
- Feigelson H.S.,
- Criqui M.H.,
- Fronek A.,
- Langer R.D.,
- Molgaard C.A.
- Aboyans V.,
- Criqui M.H.,
- Abraham P.,
- et al.
- Sadrzadeh-Rafie A.H.,
- Stefanick M.L.,
- Sims S.T.,
- et al.
- Nead K.T.,
- Zhou M.J.,
- Caceres R.D.,
- et al.
- Creager M.A.,
- Belkin M.,
- Bluth E.I.,
- et al.
- Tonino P.A.,
- Fearon W.F.,
- De Bruyne B.,
- et al.
- Hirsch A.T.,
- Haskal Z.J.,
- Hertzer N.R.,
- et al.
- Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults
- Allison M.A.,
- Aboyans V.,
- Granston T.,
- et al.
- Espinola-Klein C.,
- Rupprecht H.J.,
- Bickel C.,
- et al.
- Criqui M.H.