Peripheral Artery Disease and Cardiovascular Risk Factors in the Elderly
The Honolulu Heart Program
Peripheral vascular disease as measured by the ankle/brachial blood pressure index (ABI) is associated with increased risk of mortality and morbidity. Few sources of data on the relationship of risk factors to ABI are available for the elderly, especially those >80 years of age, and minority populations. ABI measurements from the Honolulu Heart Program's fourth reexamination of 3450 ambulatory, elderly Japanese American men indicate that the prevalence of an abnormal ABI, defined as a ratio of <0.9, was 13.6%, increasing from 8.0% in those 71 to 74 years of age to 27.4% in those 85 to 93 years. Associations that were U or J shaped were present for a number of risk factors (higher rates of abnormality [ABI<0.9] in those in the lowest and highest risk factor quintiles) in a cross-sectional analysis. Risk factors measured at baseline were also predictive of an abnormal ABI 25 years later, even after adjustment for multiple risk factors. The odds ratio (OR) for an ABI<0.9 at the 80th percentile of cholesterol compared with that at the 20th percentile was 1.4; the OR for 1-hour postload glucose was 1.3, and for alcohol intake 1.2. The OR associated with hypertension was 1.8 and that for smoking, 2.9 (P<.05 for all ORs). These findings are consistent with ABI being a marker for generalized atherosclerotic disease in old and very old Japanese American men.
- Received January 30, 1996.
- Revision received May 1, 1996.
Atherosclerotic vascular disease is a process that frequently involves the entire vascular system in an individual. Those who exhibit symptoms secondary to atherosclerosis in one organ system, such as the brain, heart, peripheral vascular system, or kidneys, often have evidence of the process in the arteries of other organ systems. However, there are relatively few noninvasive diagnostic tests that can reveal evidence of atherosclerosis before the development of symptoms. One such simple, low-cost diagnostic test is the comparison of BP measurements in the ankle and arm.1 Disproportionately low ankle BPs are frequently secondary to atherosclerosis in the arteries of the lower extremities. The ABI, a ratio of the ankle and arm pressures (also referred to as ankle/arm index [AAI] and ABPI) is believed to be highly correlated with PVD or lower extremity arterial disease. Elderly adults are known to have increased risk of vascular disease of the lower extremity arteries,2 3 and this condition is associated with increased morbidity and mortality.4 5 6 However, few data are available on risk-factor relationships for this condition in the old and very old individuals, who represent an increasing proportion of many clinicians' practices today. Even less is known about this condition in minority populations.
The examination of the elderly survivors of the original 8006 Japanese American men in the HHP cohort in 1991 through 1993 provided the opportunity to ascertain the range of ABI as well as to examine the relationships of ABI with current and past risk factors in a population of old (>70 years) and very old (>85 years) Japanese American men.
The HHP is a longitudinal study of coronary heart disease and stroke. More recently, other diseases associated with aging have been studied. In 1965 through 1968, 8006 Japanese American men between the ages of 45 and 68 years living on the island of Oahu, Hawaii, were enrolled in the HHP.7 8 These men have been followed up through hospital surveillance and intermittent examination for 30 years. A fourth examination of the 4700 surviving men, then between the ages of 71 and 93, was performed between 1991 and 1993.9 Although secondary follow-up through hospital surveillance had been continuous, the last examination of the full cohort was completed in 1974. As of 1990, 62 of the 8006 men in the original cohort had moved off the island of Oahu, resulting in an outmigration rate of about 1 per 1000 subjects per year. The current survival status is unknown in 4 of the original 8006 men.
The fourth examination was a comprehensive examination that included psychosocial and medical history questionnaires and blood tests as well as various physical assessments, including a measurement of the ankle and arm BP with a handheld Doppler device. A standard protocol was followed. The means of two measurements in the right arm and in each ankle were used to calculate an ABI for both the right and left sides. A ratio of 1 represents equality of the ankle and arm pressures. Ratios of <1 are due to relatively low ankle pressures and may indicate varying degrees of atherosclerosis in lower extremity arteries. In the population in general, the lower the ABI, the higher the prevalence of peripheral artery disease.1
Due to disability and other problems associated with aging, a significant portion of men were examined in their homes or nursing homes or had telephone interviews. The group included in this report represents those ambulatory men examined in the clinic, who represented >65% of the population still alive at the time of the examination. Another 15% were examined in their homes or in nursing homes, and many were not ambulatory, as was the case for many of the men not examined at all. Thus, the men included in these analyses represent the majority of ambulatory men and should be relatively representative of elderly ambulatory Japanese American men.
The prevalence of ABIs above and below cut points reported to represent abnormally low (<0.8 or 0.9), very low (<0.5), and high (>1.3) values was calculated for the whole population and for each 5-year age group. Due to small numbers, the last two 5-year age groups were combined to form an 85- to 93-year-old age group.
To determine whether various risk factors were independently associated with low ABI in the cross-sectional data, risk factor levels for the population with ABI measurements were divided into quintiles, and the prevalence of ABI<0.9 was determined within each quintile. The significance of trends in these relationships was tested with appropriate risk factors treated as continuous variables in logistic regression models10 with the dependent variable being ABI<0.9. When the variable was not continuous in nature (categorical), such as in the case of hypertension, indicator variables were made to test for differences between the categories. To adjust for potential confounders, age, hypertension, BMI, serum cholesterol, HDL-C, serum triglycerides, fibrinogen, fasting glucose, smoking, alcohol, and physical activity were included in the models. Separate models in which BP and diabetes status were substituted for hypertension and glucose, respectively, were also examined. Analyses were also conducted in which individuals with prevalent cardiovascular disease were excluded. Logistic models were used to produce ORs that compared the odds of ABI <0.9 at the 20th and 80th percentiles and the 95% CIs around those ORs. Models that substituted those risk factors available from the HHP baseline examination 25 years previously for those from the fourth examination were used to examine the long-term predictive value of such factors for ABI.
To illustrate the relative magnitude of the effect of each risk factor on the ABI, the difference of the mean ABI value at the 80th percentile of risk factor minus the mean ABI at the 20th percentile of the risk factor was calculated by using general linear models. The absolute difference was expressed as a percent of the SD of the ABI for the total population. Here ABI was modeled as a continuous variable. This approach helps to describe the kinds of absolute differences in ABI that could be expected to exist according to common risk factor differences.
After exclusion of the 129 men with previous vascular surgery of the lower extremities or abdomen, 3450 men were included in these analyses. Table 1⇓ shows the number of men examined by 5-year age group, the mean ankle-arm BP, and the prevalence of ABI above and below specific cut points often used in clinical or research settings in each age group.1 A substantial number of those examined (351) were >85 years of age. The mean ABI decreased from 1.067 in the 71- to 74-year-old group to 0.988 in the 85- to 93-year-old men. The prevalence of a low ABI (<0.9) increased from 8.0% in those 71 to 74 years of age to 27.4% of those >85 years old. The percent with ABI<0.8, another standard cut point, had a similar pattern, ranging from 3.7% of those 71 to 74 years old to 15.4% of those ≥85 years. Severely decreased ABIs (<0.5) were relatively rare, ranging from 0.6% in those 71 to 74 years old to 4% in those ≥85 years. The prevalence of those with marked elevations of the ABI according to the clinically used cut point of >1.30 ranged from 1.7% in those 71 to 74 years of age to 4.6% in those ≥85 years.
Table 2⇓ shows the rates of ABI measurements <0.9 by quintile of risk factor or other appropriate division. A number of these relationships appear to be U or J shaped, with higher abnormality rates in those with the lowest and highest quintiles of the risk factor than the rates seen in the midrange. Such relationships appear to be present for serum cholesterol, SBP, DBP, fasting glucose, BMI, waist/hip ratio, fasting insulin, and 2-hour postload insulin. This nonlinear relationship, as determined by the addition of a squared term to the regression model, was significant (P<.01) for cholesterol, SBP, DBP, and BMI in a logistic model that adjusts for age only. The relationships for these four variables are illustrated in Fig 1⇓. Serum glucose was of borderline significance (P=.059). Among all variables with an appearance of a nonlinear relationship, except for DBP and BMI, the highest rates of abnormality were seen in the upper quintile (those with the highest levels of the risk factor). On the other hand, a positive stepwise increase in rate of ABI<0.9 is seen with increasing levels of fibrinogen and pack-years of smoking; these trends are significant (P<.01) in a logistic model adjusting for multiple risk factors. For physical activity and HDL-C the rates decrease with increasing level of the risk factor (P<.01). Current smoking was associated with very high levels of low ABI compared with past and never smokers (P<.0001). Those with hypertension (SBP≥160 or DBP≥95 or on antihypertensive medications) also had higher rates than those with normal BP (P<.0001). Those with diabetes had higher levels than those with IGT (both conditions were assessed according to World Health Organization standards) or normal subjects (P<.001); however, the lowest rates were seen in subjects with IGT.
Table 3⇓ shows the values of various potential risk factors for atherosclerosis at their 80th and 20th percentiles, the absolute difference in the ABI between these two percentiles, and, in the lower half of the table, between presence or absence of dichotomous traits. The percent of 1 SD of the ABI for the total population that this difference represents is also shown. For example, a difference in serum cholesterol at the 80th minus the 20th percentile (54 mg/dL) would be associated with a difference of 0.023 ABI units between the population mean ABI at these two cholesterol cut points; this change in ABI represents 12.4% of the ABI SD. The greatest difference in ABI between these two risk factor cut points was associated with current smoking; hypertension was next greatest. Serum cholesterol, fibrinogen, past smoking, and BMI are also associated with relatively large differences in ABI.
The first part of Table 4⇓ shows the ORs and 95% CIs for ABI<0.9 at the 80th percentile of a risk factor compared with that at the 20th percentile of the risk factor. In the bottom part of the table similar ORs are shown for the presence or absence of risk factors that are not continuous in nature. These ORs are shown for univariate models that included only age and the risk factor and multiple logistic models that included all other appropriate risk factors. These models were used to examine the data for cross-sectional relationships of the risk factor to ABI at the fourth HHP examination and for the longitudinal relationships between the risk factors measured at the first HHP examination approximately 25 years before and ABI at the fourth examination.
As the cross-sectional analyses show, all variables considered except alcohol intake and IGT from the fourth examination were significantly associated with the difference in ABI at the specified levels of the risk factor. For the rest of the variables the univariate and multivariate ORs tended to be similar, and their CIs did not include 1. Of particular note is the OR of 4.32 for current smoking even after multivariate adjustment. BMI, low physical activity, and HDL all had significant negative associations with ABI even in the multivariate model. Thus, having higher levels of physical activity was associated with a decreased risk of having a low ABI cross sectionally. Only for DBP was there an indication of a nonlinear association when ABI was examined in this analysis.
Examination of the longitudinal associations from the first examination revealed that serum cholesterol, 1-hour postload glucose, alcohol intake, hypertension, and current smoking were significantly and positively associated with ABI 25 years later in stepwise multivariate analyses (Table 4⇑). In addition, history of diabetes was associated with a low ABI. However, the CIs for this relationship included 1 in the multivariate model. There was no significant relationship with physical activity. HDL, insulin, glucose tolerance, and fibrinogen were not measured at the first examination. The relationships seen for cholesterol and hypertension were similar for the cross-sectional and longitudinal analyses. Current smoking, while having a somewhat smaller OR in the longitudinal analyses, was the strongest among the risk factors tested in both examinations. One-hour postload glucose and alcohol intake were also significantly associated with ABI in the longitudinal analysis. For the variables examined in this longitudinal analysis, there were few indications of nonlinear relationships when the data were examined univariately and graphically (not shown), and only DBP had a significant nonlinear relationship with ABI.
We also conducted analyses to examine the association between ABI and prevalence of intermittent claudication by using the Rose Questionnaire. The overall prevalence of intermittent claudication was 1.5% and was over four times greater among men with an abnormal ABI than in men with a normal ABI (5.1% versus 1%, data not shown). However, nearly 60% of the men with intermittent claudication by questionnaire had normal ABI measurements. Similarly, men with an abnormal ABI had a significantly higher prevalence of coronary heart disease and stroke (Fig 2⇓).
Atherosclerosis is the common root of both coronary heart disease and PVD. It remains unclear whether the patterns of risk associated with various risk factors are the same for the two conditions,11 12 and there have been few population-based studies of the predictors of PVD. Symptomatic intermittent claudication and the absence of palpable peripheral arteries are relatively insensitive tools for the diagnosis of atherosclerosis in the peripheral arteries and are often present only in the late stages of such disease. Only recently have noninvasive methods, such as measurement of ankle and arm BPs, allowed the investigations of less advanced disease. The ABI offers a potentially important low-cost, noninvasive tool for assessment of generalized atherosclerotic disease as well as for PVD specifically.
The current study reports the prevalence of and risk factors for abnormal ABI in elderly individuals. The largest numbers of subjects with this condition are reported among the group that is increasing most rapidly in the US population, those over 80 and 85 years of age. It is the only report on elderly Asian Americans. The Cardiovascular Health Study, the only other report with significant numbers of individuals >80 years of age that uses similar methodology, has much smaller numbers of men and minority (mostly black) individuals in those age groups compared with the present study.13 However, in both studies the prevalence rates of low ABI increase with age and are roughly similar in magnitude, and although the exact numbers are not reported in the Cardiovascular Health Study, the prevalence appears to be slightly higher in corresponding age categories in that study. In a slightly younger population (38 to 82 years) and using somewhat more complex methods to determine disease status, Criqui et al14 appear to have found a somewhat higher prevalence of large-vessel peripheral artery disease in corresponding age categories. There have been few longitudinal studies of the predictors of abnormal ABI. Schroll and Munck15 have reported a 10-year follow-up of 50-year-old Danish men and women, but no one has reported follow-up as long as the 25 years in the current study, nor such results among older individuals. Long-term follow-up studies of nonfatal end points are subject to bias since individuals surviving to be examined may not be a representative sample of those examined at baseline. Risk factor status may also change over time as the population ages, and we cannot fully assess such changes in this study. However, if there was an effect, the bias introduced by such temporal changes would be likely to weaken the apparent associations between the risk factor and ABI and thus underestimate the OR. Despite these potential areas of bias, the results of the cross-sectional and longitudinal analyses in the Danish study15 and the current study are remarkably consistent.
Results have conflicted concerning the association of several of the major risk factors with ABI. However, like the present study, most studies indicate that smoking is the most important risk factor for PVD.12 13 15 16 17 In a small study, Criqui et al,16 however, report no association between PVD diagnosed noninvasively and smoking in women. In contrast, smoking was a strong independent risk factor for low ABI among the elderly women in the Multicenter Study for Osteoporotic Fractures.17 Smoking was also a significant predictor of cross-sectional, longitudinal, and low ABI associations in middle-aged Danish men but apparently not in women.15 Smoking was also associated with ABI and low ABI (≤0.9) in the elderly participants in the Cardiovascular Health Study report by Newman et al.13 That study reports the relationships only in individuals free of clinical cardiovascular disease. In the current analyses there was little difference in the results for smoking or other risk factors when prevalent cardiovascular disease was excluded. Smoking has also been strongly related to ABI in Scottish men and women 55 to 74 years of age.12
In the current study, an increased risk of having a lower ABI was associated with a number of other factors associated with increased risk of atherosclerotic disease in previous studies. Such ABI–risk factor relationships have been seen for all these factors in at least some studies, but few have shown consistent relationships for all factors. Increased risk in men or combined populations of men and women have been reported for increased serum cholesterol,12 13 15 decreased HDL-C,12 13 hypertension or increased SBP,12 13 15 16 17 decreased BMI,13 17 increased fibrinogen,13 diabetes or increased glucose,13 16 fasting insulin,13 increased alcohol intake,16 and decreased physical activity.17 It should be noted that in the current study the cross-sectional analyses indicate that those with lower levels of physical activity are more likely to have an abnormal ABI. However, this may simply reflect individuals with lower ABIs being more likely to have symptomatic disease and thus secondarily lower physical activity.
The U- or J-shaped associations of risk factors with a low ABI seen here for a number of these variables do not seem to have been reported before. In many cases previous analyses did not adjust for as many of the potential confounders as we did, nor were the data displayed in a manner in which differences in the ABI relationship at different points in the risk factor distribution could be easily determined. For example, findings in the current study indicate a consistent, moderately strong, and significant relationship between increasing serum cholesterol level and decreased ABI. This relationship is present in both cross-sectional and longitudinal analyses. However, ABI<0.9 was actually significantly associated with both high and low cholesterol. Although this phenomenon has not been emphasized in reports of the ABI-cholesterol relationship, these cross-sectional findings are consistent with other reports of increased levels of atherosclerotic diseases associated with low levels of cholesterol in older individuals.18 This has generally been thought to be a marker of overall disease burden. Similar cross-sectional and statistically significant U- or J-shaped relationships with ABI<0.9 (Fig 1⇑) were found for SBP, DBP, BMI, physical activity, HDL-C, serum glucose, and fasting insulin, all of which might be influenced by general disease burden and debility. For DBP, BMI, physical activity, and HDL the highest rates of abnormality were actually in the first quintile, or those with the lowest levels of the risk factor. Fibrinogen level, which might be expected to be more directly related to total chronic disease burden in the elderly, had a strong and highly significant direct association with abnormal ABI. Only for DBP was a significant nonlinear relationship between a risk factor and ABI as a continuous variable seen. Fasting glucose had a significant linear relationship with low ABI cross sectionally, and there was a suggestion of a J-shaped relationship, although the squared term was not quite significant in a similar model (P=.059).
A number of studies have not found significant relationships for variables that showed significant associations in the current study. Some of the differences in reported risk factor relationships with PVD may be due to nonlinear relationships. For example, while there is obviously significant variation in the prevalence of low ABI across the distribution of DBP (Table 2⇑), were one to look only at the coefficient for linear trend, one would conclude that DBP was unrelated to having a low ABI instead of having a significant U-shaped relationship. This relationship may be the result of comorbid conditions that both lower the ABI in their later stages and are reflected in a lower DBP. However, the relationship with SBP is slightly different than that for DBP. It is possible that the relationship with a low DBP relates to a decrease in cardiac output in elderly individuals with athersclerotic disease of the heart or other comorbidities. Due to the vascular stiffness that is thought to lead to a widened pulse pressure and isolated systolic hypertension in older individuals, the SBP may be less influenced by comorbid conditions than the DBP.
Most other studies to date appear to have assumed linearity and simply report a nonsignificant relationship.12 13 15 16 17 It would seem prudent for future analyses of risk factor associations with PVD to keep this phenomenon in mind, especially in elderly populations.
The findings in this study are consistent with ABI being a marker for generalized atherosclerotic disease in the body. Asymptomatic low ABI measurements were relatively common in these older Japanese American men. Cross-sectional risk factor relationships with ABI were similar to those seen for coronary disease in this and other studies and are more consistently so than in many reported studies. While there is potential for bias in the longitudinal analyses, they too are remarkably consistent for the major risk factors. The findings that both high and low risk factor levels are associated with low ABI are consistent with the similar associations reported for other atherosclerotic diseases in older individuals. It is likely that this phenomenon is present in other elderly populations in which such relationships have not yet been examined. Further follow-up will be necessary to document the implications of these findings.
Selected Abbreviations and Acronyms
|BMI||=||body mass index|
|DBP||=||diastolic blood pressure|
|HHP||=||Honolulu Heart Program|
|IGT||=||impaired glucose tolerance|
|PVD||=||peripheral vascular disease|
|SBP||=||systolic blood pressure|
This study was supported by contract N01-HC-05102 from the National Heart, Lung, and Blood Institute.
Orchard TJ, Strandness DE. Assessment of peripheral vascular disease in diabetes: report and recommendations of an international workshop sponsored by the American Diabetes Association and the American Heart Association. Circulation. 1993;88:819-828.
Brand FN, Abbott RD, Kannel WB. Diabetes, intermittent claudication, and risk of cardiovascular events: the Framingham Study. Diabetes. 1989;38:504-509.
Kagan A, Harris BR, Winkelstein W Jr, Johnson KG, Kato H, Syme SL, Rhoads GG, Gay ML, Nichaman MZ, Hamilton HB, Tillotson JL. Epidemiologic studies of coronary heart disease and stroke in Japanese men living in Japan, Hawaii and California: demographic, physical, dietary and biochemical characteristics. J Chronic Dis. 1974;27:345-364.
Curb JD, Petrovitch H, Rodriguez BL. Fourth examination. In: Kagan A, ed. The Honolulu Heart Program: An Epidemiologic Study of Coronary Heart Disease and Stroke. Amsterdam, Netherlands: Harwood Academic Press; 1996.
Snedecor GW, Cochran WG. Statistical Methods. Ames, Iowa: Iowa State University Press; 1967:246-248.
Fowkes FGR, Housley E, Riemersma RA, Macintyre CCA, Cawood EHH, Prescott RJ, Ruckley CV. Smoking, lipids, glucose intolerance, and blood pressure as risk factors for peripheral atherosclerosis compared with ischemic heart disease. Am J Epidemiol. 1992;135:331-340.
Newman AB, Siscovick DS, Manolio TA, Polak J, Fried LP, Borhani NO, Wolfson SK. Ankle-arm index as a marker of atherosclerosis in the Cardiovascular Health Study. Circulation. 1993;88:837-845.
Criqui MH, Fronek A, Barrett-Conner E, Klauber MR, Gabriel S, Goodman D. The prevalence of peripheral arterial disease in a defined population. Circulation. 1985;71:516-521.
Criqui MH, Browner D, Fronek A, Klauber MR, Coughlin SS, Barrett-Connor E, Gabriel S. Peripheral arterial disease in large vessels is epidemiologically distinct from small vessel disease: an analysis of risk factors. Am J Epidemiol. 1989;129:1110-1119.
Vogt MT, Cauley JA, Kuller LH, Hulley SB. Prevalence and correlates of lower extremity arterial disease in elderly women. Am J Epidemiol. 1993;137:559-568.