Flowchart of participants included in the analysis. ABI indicates ankle-brachial Index; LEAD, lower extremity arterial disease.
Frequency distribution of ankle-brachial index (ABI) decline over 6 years in 2289 men and women. Gray bars represent subjects with incident lower extremity arterial disease based on a decline of more than 0.15 and to a value of 0.9 or less. Black bars represent noncases of decline. Eight participants whose values changed by more than 0.7 U were excluded from these data.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Kennedy M, Solomon C, Manolio TA, et al. Risk Factors for Declining Ankle-Brachial Index in Men and Women 65 Years or Older: The Cardiovascular Health Study. Arch Intern Med. 2005;165(16):1896–1902. doi:10.1001/archinte.165.16.1896
An ankle-brachial index (ABI) of less than 0.9 is a noninvasive measure of lower extremity arterial disease and a predictor of cardiovascular events. Little information is available on longitudinal change in ABI or on risk factors for declining ABI in a community-based population.
To assess risk factors for ABI decline, we studied 5888 participants in the Cardiovascular Health Study cohort (men and women 65 years or older). We measured ABI in 1992-1993 and again in 1998-1999. At baseline, we excluded individuals with an ABI less than 0.9, ABI greater than 1.4, or confirmed symptomatic lower extremity arterial disease (n = 823). The group with ABI decline included 218 participants with decline greater than 0.15 and to 0.9 or less. The comparison group comprised the remaining 2071 participants with follow-up ABI.
The percentage of participants with ABI decline was 9.5% over 6 years of follow-up. The mean ± SD decline was 0.33 ± 0.12 in cases of ABI decline and 0.02 ± 0.13 in noncases. Independent predictors of ABI decline, reported as odds ratios, were age, 1.96 (95% confidence interval [CI], 1.42-2.71) for 75 to 84 years and 3.79 (95% CI, 1.36-10.5) for those older than 85 years compared with those younger than 75 years; current cigarette use, 1.74 (95% CI, 1.02-2.96); hypertension, 1.64 (95% CI, 1.18-2.28); diabetes, 1.77 (95% CI, 1.14-2.76); higher low-density lipoprotein cholesterol level, 1.60 (95% CI, 1.03-2.51), and lipid-lowering drug use 1.74 (95% CI, 1.05-2.89).
Worsening lower extremity arterial disease, assessed as ABI decline, occurred in 9.5% of this elderly cohort over 6 years and was associated with modifiable vascular disease risk factors.
Lower extremity arterial disease (LEAD) is a manifestation of systemic atherosclerosis. As such, it shares its natural course of insidious and gradual progression. The disease usually becomes evident in older adults who have symptoms of intermittent claudication. Lower extremity arterial disease can be defined as a disease spectrum that can range from asymptomatic reduction in leg arterial blood pressure to symptomatic claudication, encompassing both atypical leg pain and classic claudication progressing to critical limb ischemia.
The prevalence of LEAD in the general population depends on the definition used and is substantially underestimated, if one considers only individuals who exhibit typical symptoms of claudication. The ankle-brachial index (ABI) reflects the ratio of the systolic arterial pressure in the posterior tibial or the dorsalis pedis arteries to that in the brachial artery measured by Doppler ultrasonography. A low ABI is a valid, noninvasive indicator of LEAD1 and is a much more frequent manifestation of LEAD than intermittent claudication.2,3 Several epidemiological studies4-8 report that a low ABI, representing primarily asymptomatic LEAD, is associated with prevalent cardiovascular disease and predicts future cardiovascular disease and death. Decline in ABI is used for postoperative surgical outcome assessment of patients who have had lower extremity artery bypass.9 Little data exist on ABI decline in community-based populations10 or on the incidence rate of abnormal ABI over time. To our knowledge, risk factors for declining ABI in the absence of symptoms have not been reported.
To assess the natural course of asymptomatic LEAD, we characterized longitudinal decline in ABI in a community-based elderly cohort and identified risk factors for ABI decline over 6 years of follow-up. We hypothesized that ABI decline would be common and that traditional cardiovascular risk factors and baseline presence of cardiovascular disease in other vascular beds would be associated with ABI decline.
The Cardiovascular Health Study is an observational study of risk factors for cardiovascular disease among 5888 men and women 65 years or older living in 4 communities (Forsyth County, North Carolina; Sacramento County, California; Washington County, Maryland; and Pittsburgh, Pa). Most of the first 5201 participants who were enrolled in 1989-1990 were white. In 1992-1993, an additional 687 African Americans were enrolled. For both enrollment periods, we used random samples of Medicare eligibility lists to recruit participants. All gave informed consent for participation, and study methods were approved by local institutional review boards. A detailed description of the recruitment and examination methods has been published elsewhere.11 The enrollment examination included medical history, physical examination, laboratory testing, and assessment for the presence of cardiovascular disease. For follow-up, participants were seen yearly and interviewed by telephone every 6 months.
Baseline and follow-up ABIs were measured using a standard protocol.12 After an individual rested for 5 minutes (lying flat on an examination table), a standard arm blood pressure cuff was used to measure systolic blood pressure in the right arm. The cuff was then applied to each ankle. After the brachial and posterior tibial arteries were palpated and ultrasound gel applied, a Doppler stethoscope and standard mercury manometer were used to measure systolic blood pressure in the right brachial artery and in each posterior tibial artery in rapid succession. The lower value of the left and right ABIs was used as the baseline index value. An ABI value of 0.9 or less was used to define LEAD.6,12
To determine the change in ABI with similar follow-up time for both enrollment cohorts, we used ABI determination in the index leg from 2289 participants who had an ABI measured in 1992-1993 and in 1998-1999. Mean follow-up time was 6 years. Figure 1 shows the flow of participants included in this analysis. We excluded participants who, in 1992-1993, had an ABI of 0.9 or less, an ABI greater than 1.4, or confirmed symptomatic LEAD. Those with an ABI greater than 1.4 were excluded because these values are consistent with noncompressible, calcified arteries, and including these values might have resulted in misclassification.13 Reasons for missing ABI in 1998-1999 data included death, no clinic visit, or a missing ABI (including a follow-up ABI in the index leg) among participants who came to the clinic. For the remaining 2289 participants with paired ABI determinations, we analyzed ABI decline in 2 ways. For primary analyses, we identified cases and noncases of incident LEAD. We defined “ABI decline” as the difference between ABI at baseline and in the same leg in 1998-1999 (baseline minus follow-up). We then defined cases of incident LEAD as a decline of a participant’s ABI of more than 0.15 and to 0.9 or less. We used a minimum value of decline of 0.15 based on studies in clinical populations9,14-16 and to limit the impact of measurement error in assessing decline. The comparison group for those with incident LEAD consisted of the remaining participants with paired ABI measures. In secondary analyses, we evaluated ABI decline as a continuous variable to help validate the associations of risk factors with incident LEAD.
We used data from the 1992-1993 examinations to define baseline risk factors and clinical diagnoses. We defined hypertension as the use of antihypertensive medications, systolic blood pressure of 140 mm Hg or higher, or diastolic blood pressure of 90 mm Hg or higher. We defined diabetes as use of antidiabetic medications or by the 1997 American Diabetes Association criteria,25 and clinical LEAD at enrollment was defined as a history of leg artery revascularization, diagnosed claudication, or lower extremity angioplasty. Clinical LEAD during follow-up included events of hospitalization with International Classification of Disease codes 440.2 or 443.9 that were validated by a committee. These events were used to update the first enrollment cohort’s LEAD status in 1992-1993.
We used Stata 7.0 software (StataCorp, College Station, Tex) to perform analyses and used χ2 tests or analysis of variance to compare baseline risk factors between those with or without incident LEAD. We calculated odds ratios and 95% confidence intervals (CIs) using logistic regression adjusted for age, race, sex, and baseline ABI. We modeled categorical variables as indicator variables and used multivariate regression using variables significant in unadjusted analysis to determine independent risk factors for incident LEAD. In secondary analyses, we used linear regression to assess predictors of the ABI decline as a continuous variable. Statistical tests were considered significant at P<.05. We used adjustment for baseline ABI to address regression to the mean and possible confounding by the baseline value.
Of 2289 eligible participants with ABI measured in the same leg in 1992-1993 and 1998-1999, 218 participants (9.5%) had incident LEAD. Figure 2 shows the distribution of ABI decline. The mean ± SD decline was 0.33 ± 0.12 among cases of incident LEAD and 0.02 ± 0.13 in noncases. If the lowest ABI of the 2 legs at follow-up were used to define incident LEAD, the incidence was higher at 13.8%, and most risk relationships were similar (data not shown).
Table 1 shows the baseline characteristics of the case and noncase groups of LEAD. The case group was slightly older than the noncase group; included more African Americans; and had higher frequencies of cigarette use, hypertension, diabetes, lipid-lowering drug use, and history of myocardial infarction and stroke. The individuals in the case group also had a lower baseline ABI and higher total and low-density lipoprotein cholesterol, fibrinogen, and white blood cell counts. The distributions of sex, body mass index, and regular aspirin use were similar in both groups. In those who developed incident LEAD, a positive Rose questionnaire response for claudication was uncommon (1.4%). Table 1 also includes characteristics of participants eligible to become case subjects but who had a missing ABI in 1998-1999. Their mean baseline ABI was intermediate between that of the case and noncase groups. Between 1989-1990 and 1992-1993, 783 of these subjects had an ABI determined, and the mean ABI increased slightly (0.02 ± 0.12). The frequency of cigarette use in those with a missing follow-up ABI was similar to the noncase group, but some other vascular risk factors were more common.
Table 2 shows the frequency of incident LEAD according to baseline risk factors and the association of baseline characteristics with incident LEAD and ABI decline as a continuous variable, adjusted for age, race, sex, and baseline ABI. Without adjustment for baseline ABI, all associations in Table 2 were similar (data not shown). Participants 85 years or older and those with a history of stroke were most likely to develop LEAD. The correlates of ABI decline were similar using the 2 outcome variables, development of incident LEAD, and change in ABI as a continuous variable. However, statistical significance was not always met in analysis of the continuous variable, and smoking was more strongly correlated with the continuous variable. Factors significantly associated with both definitions of ABI decline were older age; hypertension; use of lipid-lowering drugs; and higher total and low-density lipoprotein cholesterol, fibrinogen, and white blood cell count. Sex was not associated with incident LEAD, but women had greater decline than men. A positive response to the Rose claudication questionnaire, while rare, was associated with an odds ratio of 3.93 (95% CI, 0.88-17.4) for incident LEAD. For every standard deviation higher value of baseline ABI (0.1) the odds ratio for incident LEAD was 0.49 (95% CI, 0.42-0.57). As a continuous measure, the ABI declined by 0.036 for every 0.1 higher level of baseline ABI.
Table 3 shows the independent risk factors for both incident LEAD and ABI decline in multivariate models. Factors associated with the highest odds of incident LEAD were an age of 85 years or older (compared with ages 65-74 years), history of stroke, current tobacco use, diabetes, and lipid-lowering drug use with odds ratios of 3.79, 2.12, 1.77, and 1.74, respectively. Male sex and history of myocardial infarction were associated with the continuous outcome of decline but not with incident LEAD. White blood cell count had a similar association with LEAD as fibrinogen (data not shown).
The population-attributable risk percentages (which reflect the fraction of cases of incident LEAD that could be attributed to modifiable risk factors) were 26% for hypertension, 8% for diabetes, and 5% for current cigarette use.
A 6-year decline of ABI to a clinically relevant value, termed “incident LEAD” herein, occurred in 9.5% of participants with paired ABI measurements in this elderly cohort. For comparison, the percentage of participants who experienced a first myocardial infarction within the same 6-year period was 6.4%. There was little change in ABI in those individuals not classified as LEAD cases, in agreement with other findings in subjects without claudication.10 Risk factors for incident LEAD were older age, hypertension, current cigarette use, diabetes, higher low-density lipoprotein concentration, and use of lipid-lowering drugs at baseline. The latter association probably reflects hyperlipidemia as a risk factor. We are not aware of other data on risk factors for decline in ABI in individuals without symptomatic LEAD. Published studies15,17 to date focus on populations with clinically symptomatic LEAD.
In this study, risk factors for subclinical LEAD were similar to those for clinical LEAD reported in other studies. Thus, our findings extend those of other studies, supporting a view that LEAD is a spectrum from subclinical to symptomatic disease. Risk factors for clinical LEAD include cigarette use and diabetes,18,19 with hypertension and hypercholesterolemia playing lesser roles.20 Although other studies21 have indicated associations of elevated fibrinogen with progression of clinical LEAD, we did not observe an association of fibrinogen with ABI decline after adjustment for other risk factors. Our findings show that incident LEAD was not more likely to occur in men than in women, which is similar to results in a study15 of a vascular laboratory population; however, women had a greater 6-year decline in ABI than men, as was recently observed in another study.10 A lack of association of male sex with lower ABI was also noted in a cross-sectional analysis of the Cardiovascular Health Study,12 which identified smoking, diabetes, older age, and nonwhite race as factors correlated with ABI of less than 0.9. Although the prevalence of LEAD is higher in African Americans23 after adjusting for other known risk factors, African Americans were not more likely than white patients to develop incident LEAD. The association of stroke with ABI decline may be the result of shared risk factors and pathophysiologic traits, with both stroke and LEAD representing “peripheral” vascular disease. This is supported by the lack of clear association of myocardial infarction with incident LEAD.
In our analysis, a higher baseline ABI was associated with a lower risk of developing incident LEAD, whereas a higher baseline ABI was associated with a larger decline in ABI as a continuous variable. Because we required a case to decline by more than 0.15 U and to 0.9 U or less, those with a higher baseline ABI were much less likely to become case subjects than those with a lower baseline ABI. In our analysis of decline in ABI as a continuous variable, however, regression to the mean probably resulted in a higher ABI, predicting more decline. This was reported in a study15 of a vascular laboratory population and a study10 of men and women aged 55 to 74 years followed for 12 years.
The limitations of our study require consideration. Missing data for the follow-up ABI could introduce bias. The rate of missing paired ABI here was lower than in the only other similar published study.10 Participants without a follow-up ABI had characteristics that were similar to our case group in some respects but similar to the noncase group in other respects. Cigarette use, one of the strongest risk factors for LEAD, was not more common in those with a missing follow-up ABI. Besides missing data, there are 4 other potential sources of bias. First, 800 participants died before their ABI could be measured in 1998-1999. Because individuals with progressive atherosclerosis were likely overrepresented in this group, any bias for our analyses is likely to be conservative. Second, there may be bias related to the measurement of ABI among diabetic patients and exclusion of those with an ABI greater than 1.4 at baseline. Because an ABI greater than 1.4 is associated with stiff, calcified arteries (predominantly among diabetic patients) and the ABI of diabetic patients might rise over time as a result of leg artery calcification, we may not have adequately assessed LEAD progression in some patients with diabetes. This may have led to underestimation of the risk of ABI decline associated with diabetes. Third, we were not able to differentiate between the influences of calcification and atherosclerosis on the incidence of LEAD. These processes could occur simultaneously, and we may have underestimated the true incidence of LEAD by not being able to adequately address rising ABI in some participants. Finally, because of the method of the ABI measurement, which did not include a postexercise value, we could have missed a contribution to the incidence of LEAD related to aortoiliac disease.
Our data support a strong role for modifiable risk factors, especially hypertension, smoking, diabetes, and hyperlipidemia in LEAD progression. There are known benefits of risk factor control in clinical LEAD.24 Because LEAD is associated with substantial morbidity and mortality, our data support the need to investigate the impact of risk factor intervention on LEAD prevention. This is particularly relevant for the elderly population, who have a high incidence and benefit in other ways from risk factor control.
Our findings, documenting risk factors for ABI decline and demonstrating that ABI decline can be quantified over time, provide a basis for further study of the associations of ABI decline with cardiovascular morbidity and mortality. Given its high incidence, ABI decline might also represent a valuable surrogate for research studies of risk-factor interventions for reducing the burden of vascular disease. Because a low ABI is related to the risk of other cardiovascular events, interventions to prevent ABI decline should be investigated for potential to improve overall vascular health outcomes in the elderly population.
Correspondence: Mary Cushman, MD, MSc, Department of Medicine, University of Vermont, 208 Park Dr, Suite 2, Colchester, VT 05446 (firstname.lastname@example.org).
Accepted for Publication: April 28, 2005.
Group Members: A full list of participating investigators and institutions can be found at http://www.chs-nhlbi.org.
Financial Disclosure: None.
Funding/Support: The Cardiovascular Health Study was funded by contracts N01-HC-85079–N01-HC-85086, 01-HC-35129, and N01-HC-15103 from the National Heart, Lung, and Blood Institute. This project was also supported by grants T32 HL 07594 (Dr Kennedy) and HL-03618 (Dr Cushman) from the National Heart, Lung, and Blood Institute.
Previous Presentation: This study was presented in part at the 43rd Annual Conference of the American Heart Association Council on Epidemiology and Prevention; March 5, 2003; Miami, Fla.
Acknowledgment: We thank the many Cardiovascular Health Study investigators and staff for their contributions to this study.