Usefulness of Postexercise Ankle-Brachial Index to Predict All-Cause Mortality

Peripheral arterial disease predicts future cardiovascular events and all-cause mortality. Conventional methods of assessment might underestimate its true prevalence. We sought to determine whether a postexercise ankle-brachial index (ABI), not only improved peripheral arterial disease detection, but also independently predicted death. This was an observational study of consecutive patients referred for ABI measurement before and after the fixed-grade treadmill or symptom-limited exercise component to a noninvasive vascular laboratory from January 1990 to December 2000. The subjects were classified into 2 groups. Group 1 included patients with an ABI of ≥0.85 before and after exercise, and group 2 included patients with a normal ABI at rest but <0.85 after exercise. A total of 6,292 patients underwent ABI measurements with exercise during the study period. Propensity score matching of the groups was performed to minimize observational bias. Overall mortality, as determined using the United States Social Security death index, was the end point. The 10-year mortality rate of groups 1 and 2 was 32.7% and 41.2%, respectively. An abnormal postexercise ABI result independently predicted mortality (hazard ratio 1.3, 95% confidence interval 1.07 to 1.58, p = 0.008). Additional independent predictors of mortality were age, male gender, diabetes, and hypertension. After the exclusion of patients with a history of cardiovascular events, the predictive value of an abnormal postexercise ABI remained statistically significant (hazard ratio 1.67, 95% confidence interval 1.29 to 2.17, p <0.0001). In conclusion, our results have shown that the postexercise ABI is a powerful independent predictor of all-cause mortality and provides additional risk stratification beyond the ABI at rest.

A diagnosis of peripheral arterial disease (PAD) serves as a marker for systemic atherosclerosis. The National Cholesterol Education Panel guidelines have recognized PAD as a disease equivalent to the presence of coronary artery disease (CAD). This serves to highlight the newfound appreciation for a PAD diagnosis in the cardiovascular risk assessment and determining the vigor of preventive strategies. The ankle-brachial index (ABI) is a useful tool to screen for PAD; however, conventional ABI measurements have been obtained in the at rest state and might underestimate the true prevalence of PAD. Although it has been recognized in clinical practice that obtaining a measurement of the ABI after exercise might improve the sensitivity of PAD detection, the postexercise ABI has not thus far been validated as a screening test nor has the association with all-cause mortality been determined in a United States-based population. In the present study, we assessed the prognostic value of the postexercise ABI among patients referred for testing with exercise for clinical reasons. We also sought to determine whether an abnormal postexercise ABI results was an independent predictor of death and could potentially identify a population at greater mortality risk that would otherwise have been missed using the conventional at rest ABI measurements.


Consecutive patients referred to the Cleveland Clinic noninvasive vascular laboratory from January 1990 to December 2000 for either complete pulse volume recordings of the lower extremities or standard ABI measurement before and after a fixed-grade treadmill protocol or symptom-limited exercise test were considered for analysis. The patients had to be >40 years old and residents of the United States with a valid Social Security number. The indication for testing was at the sole discretion of the referring physician. For the purposes of the present study, the patients were divided into 2 groups according to their ABI result: group 1, a normal ABI of ≥0.85 both at rest and after exercise; and group 2, a normal ABI at rest that was abnormal (<0.85) in either limb after exercise. The exclusion criteria were a history of lower limb revascularization, either surgical or percutaneous, the lack of pressure values secondary to completely or partially calcified arteries or an inaudible Doppler signal, unilateral studies, and nonatherosclerotic etiologies for lower extremity occlusive vascular disease.

At testing, the patient information, including demographic data, pulse pressure measurements, and risk factor profile, were stored in an institutional review board-approved vascular laboratory database. Additional data were collected, if required, by reviewing the electronic and paper medical records and adding to the database in a manner that protected patient confidentiality. The criteria for the various risk factors were as follows. CAD was defined as present if the patient had a history of angina, congestive heart failure, or myocardial infarction or had undergone percutaneous coronary intervention or coronary artery bypass grafting. Similarly, a history of a cerebrovascular accident or abdominal aortic aneurysm was defined as a history of stroke, transient ischemic attack, or previous carotid endarterectomy and previous identification on suitable imaging studies or previous surgical repair, respectively. The assessment for hypertension, diabetes, and hyperlipidemia was determined by a combination of questioning at testing, chart review, and medication use. Smoking history was determined by self-reported current use at testing.

Doppler-derived systolic pressures at the level of the brachial artery and posterior tibial and dorsalis pedis arteries were obtained by recording the level at which the first Doppler sound was heard after deflation of the cuff placed over the upper arm and ankle, respectively. The index was calculated using the greater of the 2 ankle pressures (posterior tibial or dorsalis pedis) obtained on each leg, with the greater of the brachial pressures. The patients exercised on a treadmill for either 5 minutes or until the onset of leg pain that limited them from continuing. The exercise was also terminated by the onset of any cardiopulmonary symptoms, especially if these were the principal restriction to exercise. Once the exercise had been completed, the patient returned to the bed in the supine position and both ankle pressures were obtained (first in the symptomatic extremity or the extremity with the lower pressure at rest) followed by the brachial pressure in the arm with the greater pressure at rest. The ABI thus obtained was referred to as the postexercise ABI in the present study. When the patients had undergone more than one test during the study period, only the first was taken. The United States Social Security death index was used to match all subjects to their records according to name and Social Security number. The vital status was determined as of May 2007.

The continuous data are displayed as the mean ± SD. The categorical data are displayed as frequencies and percentages within the population. The continuous data were analyzed using the nonparametric Wilcoxon test and categorical data using the chi-square test. Survival analysis was done using Cox proportional hazard modeling, adjusted for confounding factors. The proportionality assumption was tested with all time-dependent covariates simultaneously. The relative risks (hazard ratio) were estimated and 95% confidence intervals given. Patient age at testing, gender, race, and history of CAD, cerebrovascular accident, smoking, diabetes, hypertension, hyperlipidemia, abdominal aortic aneurysm, and appropriate transformations were included in the modeling.

The patients from group 2 were matched one-to-one to the patients from group 1 using propensity scores before survival analysis to minimize any bias introduced by the observational nature of the present study. The propensity score matching was done in all aspects, except for their postexercise ABI results, using a greedy algorithm. This was accomplished using a nonparsimonious logistic regression model, with all available covariates included to derive a propensity score for patients with an abnormal postexercise ABI result, and then using the propensity scores to match those from group 2 to the patients in group 1. Patients with missing values were excluded from matching and additional analysis. The resulting matched groups were compared for each covariate to confirm the similarity between the 2 groups ( Table 1 ). The propensity score was also included in the Cox proportional hazard model in the analysis. All the analyses were done using the SAS, version 8.2, statistical package (SAS, Cary, North Carolina).

Table 1

Demographic and risk factor profile of patients in groups 1 and 2 after propensity score matching

Variable Group 1 (n = 713) Group 2 (n = 713) p Value
Men 484 (67.9%) 467 (65.5%) 0.3395
White 659 (92.4%) 664 (93.1%) 0.6090
Black 47 (6.6%) 44 (6.2%) 0.7452
Age (years) 64.5 (±9.2) 64.7 (±9.5) 0.6745
40–49 58 (8.1%) 59 (8.3%)
50–59 149 (20.9%) 159 (22.3%)
60–69 302 (42.4%) 259 (36.3%)
≥70 204 (28.5%) 236 (33.1%)
Smoker 21 (3.0%) 25 (3.5%) 0.5488
Hypertension 230 (32.3%) 234 (32.8%) 0.8211
Diabetes mellitus 87 (12.2%) 91 (12.8%) 0.7486
Coronary artery disease 236 (33.1%) 243 (34.1%) 0.6947
Cerebrovascular accident 69 (9.7%) 798 (10.9%) 0.4332
Abdominal aortic aneurysm 39 (5.5%) 41 (5.8%) 0.8180
Hyperlipidemia 126 (17.7%) 129 (18.1%) 0.8358

Group 1, patients with normal ankle-brachial index at rest and after exercise, group 2: patients with normal ankle-brachial index at rest but abnormal after exercise.


From January 1990 to December 2000, 11,295 patients underwent either complete or limited lower extremity pulse volume recording studies with ABI measurement at the Cleveland Clinic noninvasive vascular laboratory. Of these, 6,292 were performed with a fixed-grade treadmill or symptom-limited exercise component. For the purposes of the present study, the patients with an abnormal ABI at rest were not considered. A total of 2,416 patients met the inclusion criteria according to the postexercise ABI results. After matching, 1,426 patients (713 each in group 1 and 2), as defined in the “Methods” section, were included in the present study.

The baseline demographic and risk factor profiles of the matched and unmatched patient groups are listed in Tables 1 and 2 . The mean patient age was 64.1 ± 10.3 years, and 40.8% were women. Notable differences were present between the 2 groups before matching. The subjects in group 2 had a greater percentage of white men and were more likely to be older, have a history of CAD, cerebrovascular accident, or abdominal aortic aneurysm, take medications for hyperlipidemia, hypertension, or diabetes, and be current smokers at testing. This bias was minimized by matching the patients from group 2 with those from group 1 using propensity scores ( Table 1 ).

Table 2

Demographic and risk factor profile of patients in groups 1 and 2

Variable Group 1 (n = 1,700) Group 2 (n = 716) p Value
Men 960 (56.5%) 470 (65.36%) <0.0001
White 1,453 (85.5%) 667 (93.3%) <0.0001
Black 216 (12.7%) 44 (6.2%) <0.0001
Age (years) 63.9 (±10.6) 64.7 (±9.5) 0.0729
40–49 213 (12.5%) 59 (8.2%)
50–59 362 (21.3%) 159 (22.2%)
60–69 601 (35.4%) 262 (36.6%)
≥70 524 (30.8%) 236 (33.0%)
Smoker 27 (1.6%) 28 (3.9%) 0.0005
Hypertension 451 (26.5%) 237 (33.1%) 0.0011
Diabetes mellitus 161 (9.5%) 92 (12.9%) 0.0133
Coronary artery disease 392 (23.1%) 246 (34.4%) <0.0001
Cerebrovascular accident 101 (5.9%) 79 (11.0%) <0.0001
Abdominal aortic aneurysm 64 (3.8%) 43 (6.0%) 0.0145
Hyperlipidemia 194 (11.4%) 131 (18.3%) <0.0001

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

Stay updated, free articles. Join our Telegram channel

Dec 22, 2016 | Posted by in CARDIOLOGY | Comments Off on Usefulness of Postexercise Ankle-Brachial Index to Predict All-Cause Mortality

Full access? Get Clinical Tree

Get Clinical Tree app for offline access