Ankle brachial index screening in asymptomatic older adults

Abstract

Background

Screening for peripheral arterial disease (PAD) by measuring ankle brachial index (ABI) in asymptomatic older adults is currently recommended to improve cardiovascular disease risk assessment and establish early treatment, but it is not clear if the strategy is useful in all populations. We examined the prevalence and independent predictors of an abnormal ABI (<0.90), in an asymptomatic sample of 1,017 adults, 60 to 69 years old, enrolled in the ADVANCE study.

Methods

Baseline data collected between December 2001 and January 2004 among the healthy older controls enrolled in ADVANCE was examined. Frequency distributions and prevalence estimates of an abnormal ABI were calculated, using both standard and modified definitions of ABI. Stepwise logistic regression was used to examine independent predictors of ABI <0.90. Signal detection analysis using recursive partitioning was employed to explore potential demographic and clinical variables related to ABI <0.90.

Results

The prevalence of ABI <0.90 was 2% when using the standard definition and 5% when using a modified definition. ABI prevalence did not differ by gender (P >.05). Compared with subjects who had a normal ABI (0.90–1.39), subjects with an ABI <0.90 were more likely to currently smoke, be physically inactive, have a coronary artery calcium score >10, and an FRS >20% (P ≤.02). Independent predictors of ABI <0.90 when using the standard definition included currently smoking, physical inactivity, and body mass index >30 (all P values ≤.03), and when using the modified definition included currently smoking, physical inactivity, and hypertension (all P values ≤.04). Currently, smoking was the only significant variable for ABI <0.90 derived through recursive partitioning (P = .02), and indicated that prevalence of ABI <0.90 was 1.5% for nonsmokers, while it was 6.6% for current smokers.

Conclusions

ABI screening in generally healthy individuals 60 to 69 years old may result in lower prevalence rates of a positive result than estimates based on studies in clinical populations. The modified definition for calculating ABI captured more asymptomatic adults with suspected peripheral arterial disease. More evaluation of the appropriate role of ABI screening in unselected populations is needed before routine screening is implemented.

Evidence that adults with peripheral arterial disease (PAD) in the lower extremities are at higher risk for a cardiovascular disease (CVD) event, such as myocardial infarction or stroke, is clearly established.¹ The American College of Cardiology/American Heart Association 2005 Practice Guidelines for the management of patients with PAD recommend measurement of ankle brachial index (ABI) in asymptomatic adults ≥50 years old with history of smoking or diabetes, among adults with lower extremity circulation problems, and all adults ≥70 years old to improve CVD risk assessment and establish early treatment.¹ The recent Ankle Brachial Index Collaboration meta-analysis² suggests that measurement of ABI may improve CVD risk prediction beyond the Framingham risk score. On the other hand, the United States Preventive Services Task Force recommends against routine ABI screening in asymptomatic adults, citing that there is insufficient evidence to warrant routine screening and that the added costs in time and resources may exceed benefits.³

Routine screening tests are most useful when they reduce mortality or morbidity. The elements of a good screening test include its ability to detect a subclinical phase of the disease, when early treatment is known to improve patient outcomes, and to be widely accessible, simple to administer, inexpensive, and associated with minimal discomfort and morbidity for the population to be screened.⁴ Further, it is paramount that a screening test has established strong sensitivity and specificity for the disease being screened, for example, an ABI for detecting PAD.^5,6 Reported sensitivity and specificity of ABI <0.90 to detect ≥50% stenosis in the lower extremities using digital subtraction angiography, is 76% and 90%, respectively.⁷

In 2003, the average US cost of performing an ABI was $61 per case.⁸ Yet, ABI screening among adults at higher risk for developing PAD also uses resources such as staff training and time, equipment and supplies. Campbell et al⁹ conducted a study to improve targeted screening efforts for identifying PAD in high risk persons and reported associated costs ranging from 1 to 3 days of staff time per diagnosis.⁹ In addition, these investigators estimated that 15 patients needed to be screened to detect one new patient with PAD.⁹

Reported prevalence of PAD, using ABI <0.90 as the indicator, ranges from 1.2% in a managed care organization’s population of 6.67 million adult members ≥18 years old,¹⁰ up to 29% in the PARTNERS study which enrolled only higher-risk adults, that is, 50 to 69 years old with a history of smoking or diabetes mellitus or adults at least 70 years old.¹¹ It is unknown what the prevalence of PAD (using ABI <0.90) is among healthy adults 60 to 69 years old without documented clinical cardiovascular disease.

Another issue for determining prevalence estimates of PAD using ABI <0.90 as the indicator, are the various methods reported for calculating ABI.¹² The standard method uses the highest ankle pressure for each leg, divided by the highest brachial pressure.^13,14 Recently, other investigators^15,16 have suggested that modifying the ABI calculation method by using the lowest ankle pressure for each leg, divided by the highest brachial pressure, would improve sensitivity.^15,16 This modified method for calculating an ABI leads to higher prevalence estimates, although it may degrade the specificity of the test.¹²

Our aim was to determine the prevalence of an abnormal ABI (<0.90) along with independent predictors, in an asymptomatic sample of 1017 older adults, 60 to 69 years old, enrolled in the ADVANCE study, using both standard and modified methods for calculating ABI.

Methods

Study design

This is a cross-sectional study design examining baseline data (collected between December 2001 and January 2004) among the healthy older controls enrolled in ADVANCE.

Subjects

A total of 1000 older men and women were targeted for enrollment in the study. Male and female members of Kaiser Permanente of Northern California, 60 to 69 years old (as of 01/06/2001), were identified in the health plan’s electronic databases as potentially eligible participants, after excluding those with a major chronic disease or living >50 miles from the research clinic. Details of the recruitment plan have been previously reported.^17,18 After screening and confirming eligibility, a total of 1023 older controls enrolled in the study. Complete data are available for 1017 subjects.

Approval to conduct the study was obtained from the institutional review boards at Stanford University Medical School and the Kaiser Foundation Research Institute. The investigation was carried out according to the principles outlined in the Declaration of Helsinki, including written informed consent from all subjects.

Data collection procedures

A comprehensive, self-administered health survey was mailed to subjects for completion prior to their baseline clinic visit. Survey items included age, gender, marital status, educational level, employment status, household income, birthplace, race/ethnicity, and self-reported medical history. Subjects’ self-reported medical history included previous myocardial infarction, peripheral artery disease, stroke, hypertension, high cholesterol, diabetes mellitus, smoking status, alcohol consumption, major medical conditions, surgical procedures, major depression, and cancer. Subjects were also asked to bring all current medications for review at the baseline visit, and these were recorded by study staff.

Selected clinical measures were obtained at the clinic visit using standard methods and included blood pressure, height, weight and waist circumference, electrocardiogram, heart rate variability, coronary artery calcium score (CAC), and ABI. In addition, fasting blood samples were drawn to obtain DNA for genetic studies and plasma for determination of lipid and lipoprotein levels, glucose, insulin, and C-reactive protein. Plasma and serum were also stored for future studies. Subjects found to have abnormal CVD risk factors, for example, ABI <0.90, were referred back to their primary health care provider for follow-up and treatment. Data were collected between December 2001 and January 2004.¹⁷

Measures

Ankle brachial index

Ankle brachial index was performed after subjects rested for at least 10 minutes in a supine position. Systolic blood pressure was measured using a handheld Doppler (Nicolet Vascular Elite 100; Madison, WI) with a 5-Mhz probe using transducer gel, in both right and left brachial, dorsalis pedis and posterior tibial arteries, unless contraindicated. The arm/ankle ratios were calculated using the standard definition, that is, highest ankle pressure for each leg, divided by the highest brachial pressure.^13,14 In addition, the arm/ankle ratios were calculated using a modified definition, that is, lowest ankle pressure for each leg, divided by the highest brachial pressure.¹⁵ For both definitions, we used the leg with the lowest ABI for analysis.^13,14

Data analysis

All forms were reviewed for accuracy and completeness at the time of the clinic visit. Variable frequency distributions were used to check for extreme values. Descriptive statistics were calculated for all variables. Frequency distributions were calculated and included subject characteristics by ABI score as follows: ABI <0.90, ABI = 0.90–1.39, or ABI ≥1.40.^19,20 Prevalence of PAD was determined using ABI <0.90 as the indicator. In addition, stepwise logistic regression was used to examine independent predictors of ABI <0.90. Finally, signal detection analysis was used to determine, through recursive partitioning, an algorithm that characterizes distinct subgroups of subjects that are mutually exclusive and maximally discriminated from each other with respect to a specific dichotomous outcome, that is, ABI <0.90.^21–23 Signal detection allows for full use of all data available for each variable being evaluated. Data were analyzed using SAS (Version 9.1, SAS Institute Inc, Cary, NC).

The authors are solely responsible for the design and conduct of this study, all study analyses, the drafting and editing of the manuscript and its final contents. All authors had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Results

Subjects were on average 66 years old. Most subjects were married, retired, and white, and 23% had graduated from college (Table I). Despite being selected as a healthy cohort, the subjects had a reasonably high prevalence of hypertension, dyslipidemia, physical inactivity, diabetes, and obesity. Compared with women, men were more likely to be married, have diabetes, a triglyceride/high-density lipoprotein (HDL) ratio ≥3, be more physically active, have a CAC score >10, and a FRS >20% (P <.01) (Table I).

Table I.

ADVANCE older controls with ABI score, baseline characteristics, n = 1017

	Total, n = 1017	Women, n = 381	Men, n = 636	P-value, gender differences
Age, mean years (SD)	65.8(2.8)	65.8(2.8)	65.8(2.9)	.69
Employed part or full-time, %	39.3	36.5	41.3	.16
Married, %	77.3	65.6	84.3	<.01
College graduate, %	23.3	21.1	24.7	.19
Race/ethnicity
White/European, %	67.5	66.9	67.8	.78
Black/African-American, %	8.2	6.8	9.0	.23
Hispanic, %	6.0	6.6	5.7	.59
Asian/Pacific Islander, %	7.8	7.3	8.0	.80
Mixed, Hispanic, %	3.0	2.6	3.3	.58
Mixed, Non-Hispanic, %	7.6	9.7	6.3	.05
Cardiovascular disease risk Factors
Current smoker, %	7.5	8.1	7.1	.53
Diabetes*, %	18.0	11.0	22.2	<.01
Metabolic Syndrome, (n = 989)
Metabolic syndrome with diabetes, %	10.8	7.7	12.7	.01
Metabolic syndrome without diabetes, %	18.9	19.1	18.8	.93
Obesity
BMI ≥ 30 kg/m2, %	32.0	31.0	32.4	.68
Waist ≥ 88cm (women only), %	—	38.8	—	—
Waist ≥ 102cm (men only), %	—	—	36.9	—
Hypertension†, %	66.6	63.3	68.6	.09
Dyslipidemia‡
LDL >160 mg/dL, % (n = 992)	11.8	12.4	11.4	.68
Triglycerides ≥ 150 mg/dL, %, n = 1006	32.3	29.2	34.1	.12
Triglyceride/HDL Ratio ≥ 3, %	36.7	26.3	42.8	<.01
Physical inactivity
SBAS <moderate, %, n = 1015	38.7	38.9	38.6	.94
PAR kcal/kg per day ≤33, %, n = 1009	36.0	45.6	30.2	<.01
Coronary calcium score >10, %, n = 1008	62.4	40.9	75.4	<.01
Framingham Risk Score >20%, %, n = 1006	20.1	3.2	30.0	<.01
ABI score13 <0.90, % (standard definition) (highest DP or PT/highest SBP)	2.2	1.0	2.8	.07
ABI score15 <0.90, % (modified definition) (lowest DP or PT/highest SBP)	5.1	5.0	5.2	1.00
Hx lower extremity circulation problems, %	4.5	3.8	4.9	.43

SBAS, Stanford Brief Activity Survey; PAR, 7-day physical activity recall; DP, dorsalis pedis; PT, posterior tibial; SBP, systolic blood pressure; Hx, history. Bold type indicates P-value ≤ 0.05.

^*Medical history of diabetes, taking insulin or oral hypoglycemic, or fasting blood glucose ≥126 mg/dL.

^†SBP >140 mm Hg, diastolic blood pressure >90 mm Hg, medical history of hypertension or taking anti-hypertensives.

^‡Includes medical history or medications.

Ankle brachial index prevalence

The prevalence of ABI <0.90 in our study was 2% when using the standard definition and 5% when using the modified definition for ABI calculation. There was no statistically significant difference in ABI prevalence by gender (Table I). Compared to subjects with a normal ABI (>0.90–1.39), subjects with an ABI <0.90 were more likely to currently smoke, be physically inactive, have a CAC score >10, and a FRS >20% (P ≤ .02), independent of method used for calculating ABI (Tables II and III). Almost 5% of subjects reported a history of lower extremity circulation problems, a single self-reported medical history item on the study questionnaire. As this single item may or may not represent claudication symptoms associated with PAD, we elected to eliminate these subjects (n = 45) from further analysis.

Table II.

Subject characteristics by ABI score using standard definition for ABI calculation, n = 1017

ABI score frequency, n	ABI score			χ2 [df = 2]	P
	<0.90	0.90–1.39	≥1.40
	22	984	11
	% within ABI Score Category
Cardiovascular disease risk factors
Current Smoker	23	7	18	9.53	<.01
Obesity
BMI ≥ 30 kg/m2	55	32	18	6.22	.04
Waist ≥ 88cm (women only), n = 381	50	39	0	1.48	.47
Waist ≥ 102cm (men only), n = 635	56	37	22	3.56	.17
Diabetes*	41	17	27	8.72	.01
Hypertension†	82	66	55	3.03	.22
Dyslipidemia‡
LDL >160 mg/dL	68	44	36	5.43	.07
Triglycerides ≥ 150 mg/dL	41	32	18	2.07	.35
Triglyceride/HDL Ratio ≥ 3	86	81	91	1.60	.45
Physical Inactivity§	73	38	18	12.80	<.01
Coronary Calcium Score >10	86	51	91	9.50	<.01
Framingham Risk Score >20%	50	19	18	13.96	<.01

Bold type indicates P-value ≤ 0.05.

^*Medical history of diabetes, taking insulin or oral hypoglycemics, or fasting blood glucose ≥126 mg/dL.

^†SBP>140 mm Hg, diastolic blood pressure >90 mm Hg, medical history of hypertension or taking anti-hypertensives.

^‡Includes medical history or medications.

^§Not meeting national recommendations.

Table III.

Subject Characteristics by Ankle-Brachial Index Score using Modified Definition for ABI calculation, n = 1017

ABI Score Frequency, n	Ankle-Brachial Index Score			χ2 [df = 1]	P-value
	<0.90	0.90–1.39	≥1.40
	52	963	2
	% within ABI Score Category, n = 1015
Cardiovascular disease risk factors
Current Smoker	15	7	—	5.12	.02
Obesity
Body Mass Index ≥ 30 kg/m2	42	31	—	2.78	.09
Waist ≥ 88cm (women only), n = 381	37	39	—	0.03	.85
Waist ≥ 102cm (men only), n = 633	49	36	—	1.98	.16
Diabetes*	25	18	—	1.86	.17
Hypertension†	85	66	—	8.00	<.01
Dyslipidemia‡
LDL >160 mg/dL	56	44	—	3.00	.08
Triglycerides ≥ 150 mg/dL	31	32	—	0.07	.79
Triglyceride/HDL Ratio ≥ 3	88	82	—	1.10	.30
Physical Inactivity§	60	38	—	10.01	<.01
Coronary Calcium Score >10	77	62	—	4.98	.03
Framingham Risk Score >20%	33	19	—	5.13	.02

Bold type indicates P-value ≤ 0.05.

^*Medical history of diabetes, taking insulin or oral hypoglycemics, or fasting blood glucose ≥ 126 mg/dL.

^†SBP >140 mm Hg, diastolic blood pressure >90 mm Hg, medical history of hypertension or taking anti-hypertensives.

^‡Includes medical history or medications.

^§Not meeting national recommendations.

Ankle brachial index predictors

To see if we could derive a useful guide to aid practitioners in selecting patients for ABI screening, we examined several demographic and clinical variables as independent predictors of ABI <0.90, including gender, current smoker, ever-smoker, waist circumference, body mass index (BMI) >30, physical inactivity, diabetes, hypertension, and dyslipidemia. Significant independent predictors of ABI <0.90 when using the standard definition included currently smoking, physical inactivity, and BMI >30 (all P ≤ .03). When using the modified definition, these included currently smoking, physical inactivity, and hypertension (all P ≤ .04) (Table IV).

Table IV.

Predictors for Ankle-Brachial Index <0.90 using Stepwise Logistic Regression^*, n = 940^†

Variable‡	Standard definition for ABI <0.90		Modified definition for ABI > 0.90
	χ2 [df = 1]	Odds ratio (95% CI)	χ2 [df = 1]	Odds Ratio (95% CI)
Current smoker	11.08	5.35 (1.77–16.22)	7.16	2.72 (1.20–6.17)
Physical inactivity	4.29	2.98 (1.02–8.73)	4.30	1.92 (1.03–3.58)
BMI>30	7.14	2.97 (1.08–8.17)	—	—
Hypertension	—	—	5.10	2.31 (1.05–5.06)

^*Variables entered into model: current smoker, ever smoker, BMI >30, waist ≥88cm (women only), waist ≥102 cm (men only), diabetes (medical history of diabetes, taking insulin or oral hypoglycemics, or fasting blood glucose ≥126 mg/dL), hypertension (SBP >140 mm Hg, DBP>90 mm Hg, medical history of hypertension or taking anti-hypertensives), Physical inactivity, LDL >160 mg/dL or taking cholesterol lowering medications, triglycerides ≥150 mg/dL, and triglyceride/HDL Ratio ≥3.

^†Subjects with history of lower extremity circulation problems removed from analysis (n = 45).

^‡All variables listed have χ² P-values ≤.04.

Another approach to guide patient selection for screening tests is to use recursive partitioning to derive a clinical decision tool. Recursive partitioning produces a series of dichotomies (yes/no questions) to guide clinical decision-making. This produced only one useful dichotomy, current smoking status (P = .02). In our study, the prevalence of ABI <0.90 was 1.5% for nonsmokers, while it was 6.6% for current smokers. Diabetes was not an independent predictor of ABI <0.90 in this study. This may due to the fact that the prevalence of ABI <0.90 was very low (standard definition = 10/171, modified definition = 13/183) among the diabetics in our study.

Discussion

Peripheral arterial disease is often referred to as an under-diagnosed and under-treated public health problem. Patients with PAD have a significantly increased risk for serious CVD events, and risk factor management in PAD patients would likely reduce this risk. Therefore, routine screening for PAD using ABI has been advocated for all persons ≥50 years old with history of smoking or diabetes, and among all adults with lower extremity circulation problems indicative of claudication symptoms.²⁴ However, the prevalence of abnormal ABI in clinic-based studies may not indicate the performance of this test in less selected populations.^25,26 For example, Wyman et al²⁵ used ABI to screen 493 subjects (mean age = 55 years) for subclinical atherosclerosis. Although subjects in this study had at least two CVD risk factors and a high prevalence of non-occlusive carotid plaque (56%), only one subject had an ABI <0.90.

Screening tests done in epidemiological studies typically involve low-risk populations, requiring large samples with prolonged follow-up, to assess the impact of screening on disease-related outcomes or events.^5,6 In our study (n = 1,017) of healthy older adults (mean age = 66 years), a total of 22 (2.2%) subjects had an ABI <0.90 when using the standard method for calculating the ABI. When using a modified definition for calculating the ABI, the prevalence was 5.5% (n = 52), an absolute increase of 2.9%. However, these estimates are below population estimates for community-dwelling adults 60 years and older,^11,27–30 regardless of the definition used.

Presumably the lower estimated prevalence of ABI <0.90 in our study, compared to other reports, was due to our selection of a healthy cohort. We therefore compared the prevalence rates for smoking and diabetes in our sample to adults in California, as these are considered the two major risk factors for PAD for persons 50 to 69 years old.³¹ In 2004, current smoking among adults ≥18 years was 14.8%, and the prevalence of diabetes was 10–19.8% among adults 45 to 74 years old.^32,33 In our study, 18% of subjects had a medical history of diabetes, were taking insulin or an oral hypoglycemic, or had a fasting blood glucose ≥126 mg/dL; these figures are comparable to the prevalence estimates among similar aged adults in California. However, only 7.5% of subjects were current smokers, which is only half of the reported smoking prevalence among adults in California and may, in part, explain the lower prevalence of ABI <0.90 in our study. However, age-specific smoking prevalence estimates are not available in California; therefore, smoking prevalence among adults 60 to 69 years old may be lower than the reported estimate for adults ≥18 years.

Recently, a large multisite study used a targeted screening strategy²⁶ among patients in primary care settings (n = 717) to identify persons with asymptomatic PAD (ABI <0.90). Persons ≥70 years old without known CVD, and those 50 to 69 years old with at least one CVD risk factor associated with PAD, that is, diabetes, smoking, hypertension, and/or dyslipidemia, were enrolled. Prevalence of ABI <0.90 among persons ≥70 years old was 12.5%, while among those 50 to 69 years old prevalence was only 2.5%, similar to our study. Approximately 72% of the subjects in this study with an ABI <0.90 reported a history of smoking.

It is well established that persons at higher risk for developing PAD include those with known CVD and diabetes.¹ Therefore, to improve targeted screening efforts for identifying PAD, Campbell et al ⁹ conducted a study in three high-risk groups, that is, smokers and those with either hypertension or dyslipidemia, from one large general practice in Scotland. Subjects (n = 343) were on average 70 years old. A total of 24 undiagnosed persons (6.9%) had an ABI <0.90, with 76% of these persons reporting a history of smoking.⁹

In another study, Eason et al³⁴ used a cross-sectional study design in an ethnically diverse population (n = 403) drawn from four primary care clinics to screen for PAD (ABI <0.90). The average age was 64 years and the prevalence of asymptomatic PAD was approximately 6% (n = 25). After controlling for age and gender, the rate of asymptomatic PAD was higher in those with diabetes mellitus (3.8 times), and in those who smoked at least 1 pack of cigarettes/day (2.5 times). Our findings and the above cited studies all support the targeted ABI screening recommendations made by Beckman et al,³¹ in particular, among asymptomatic adults with a history of smoking, and represent a more efficient screening process compared to widespread or routine screening for PAD.

Since our study included a large number of variables, we attempted to discover other factors that might help refine ABI screening criteria and improve case-finding efficiency. We used a recursive partitioning method that is particularly suited to such, and which handles interactions quite well. However, only current smoking status was significant, so we were unsuccessful in this effort.

Study limitations

This large epidemiological study precluded us from doing angiography of lower extremities to confirm extremity occlusion in subjects with an ABI <0.90. In addition, we were unable to follow up subjects with an ABI ≥1.40 to confirm or refute a diagnosis of PAD. Our prevalence estimates were based on a recruited sample of insured adults in Northern California and may not be fully generalizable to other populations and clinical settings. We also do not have sufficient follow-up events to examine the association between abnormal ABI and incidence of clinical cardiovascular events, in this population.

Conclusion

ABI screening in generally healthy individuals 60 to 69 years old may result in lower prevalence rates of a positive result, than estimates derived from clinic-based studies. Targeted screening using Beckman’s recommendations³¹ appears to be more reasonable and efficient for identifying those at high risk for PAD and CVD. Since all diabetics are considered to have a high CVD risk equivalent,³⁵ screening for ABI is unlikely to change their treatment. Thus, it may be most efficient to perform ABI screening only in persons 50 to 69 years old with a history of smoking (former or current), and who have no other reason to implement aggressive preventive measures. In addition, using the modified definition for calculating ABI to capture more asymptomatic adults with suspected PAD is likely more efficient than the standard definition because in other studies, it was equally predictive of future events.^15,16 However, more evaluation of the appropriate role of ABI screening in unselected populations is needed before routine screening is accepted as a standard of care.