Author + information
- Received February 18, 2004
- Revision received April 4, 2004
- Accepted April 16, 2004
- Published online August 4, 2004.
- Ramin Shadman, BA⁎,
- Michael H. Criqui, MD, MPH†,⁎ (, )
- Warner P. Bundens, MD‡,
- Arnost Fronek, MD, PhD‡,
- Julie O. Denenberg, MA†,
- Anthony C. Gamst, PhD† and
- Mary M. McDermott, MD§
- ↵⁎Reprint requests and correspondence:
Dr. Michael H. Criqui, Department of Family and Preventive Medicine, University of California San Diego, 9500 Gilman Drive, 352 SCRB, La Jolla, California 92093-0607, USA.
Objectives The objective was to assess the prevalence of subclavian artery stenosis (SS) in four cohorts (two free-living and two clinical populations) and determine both risk factors for this condition and the association with other cardiovascular conditions.
Background The prevalence of SS in the general population is unknown, and its association with risk factors and other cardiovascular diseases is not well-established.
Methods A total of 4,223 subjects (2,975 from two free-living cohorts and 1,248 from two clinical cohorts) were included in this cross-sectional analysis. Subclavian artery stenosis was defined as ≥15 mm Hg interarm pressure difference.
Results The prevalence of SS was 1.9% in the free-living cohorts and 7.1% in the clinical cohorts; SS was significantly (p < 0.05) associated with past smoking (odds ratio [OR] = 1.80), current smoking (OR = 2.61), and higher levels of systolic blood pressure (OR = 1.90 per 20 mm Hg). Higher levels of high-density lipoprotein (HDL) cholesterol were inversely and significantly associated with SS (OR = 0.87 per 10 mg/dl). In regression analyses relating SS to other cardiovascular diseases, the only significant finding was with peripheral arterial disease (PAD) (OR = 5.11, p < 0.001).
Conclusions Significant SS is present in approximately 2% of the free-living population and 7% of the clinical population. Additionally, SS is correlated with current and past smoking histories, systolic blood pressure, HDL levels (inversely), and the presence of PAD. These findings suggest that bilateral brachial blood pressure measurements should routinely be performed in patients with an elevated risk profile, both to screen for SS, and to avoid missing a hypertension or PAD diagnosis because of unilateral pressure measurement in an obstructed arm.
Upper extremity blood pressure differences are usually due to atherosclerotic plaque, resulting in a reduction in blood pressure to one upper extremity, or less commonly, both upper extremities (1). Nonatherosclerotic conditions that can also result in interarm systolic pressure differences, such as Takayasu’s arteritis or radiation-induced vascular disease, are quite rare in North America (2–6). Currently, a bilateral brachial artery blood pressure measurement is the standard for screening for significant obstruction (stenosis/occlusion) of the proximal vasculature supplying the upper extremity. Upper extremity atherosclerotic obstruction is predominantly due to subclavian artery stenosis (SS), although on the right side, where the subclavian artery branches off the innominate artery (as opposed to directly from the aortic arch on the left side), up to a third of the obstructions are actually in the innominate artery (7,8). For convenience, we here refer to a interarm systolic blood pressure (SBP) difference of at least 15 mm Hg as SS, realizing that perhaps as many as 15% of such cases are actually in the innominate artery.
SS in select populations
Clinicians have continued to debate what brachial SBP difference is “clinically significant.” Previous research has shown the interarm difference in individuals without known cardiovascular pathology does not favor one arm and does not vary by gender or age (4,9–12). Furthermore, in a population of subjects (n = 400) with significant cardiovascular risk factors, Lane et al. (13) employed univariate analyses to demonstrate that increased age was the only significant predictor of SBP differences of both >10 mm Hg and >20 mm Hg, while gender, ethnicity, arm circumference, handedness, hypertension, diabetes, and cardiovascular disease were not significantly predictive. Lower extremity peripheral arterial disease (PAD) was not assessed. They reported a prevalence in this selected population of 10% having an interarm pressure difference of >15 mm Hg. In a small angiographic study (n = 52), Gutierrez et al. (14) reported that 41.6% of patients with PAD had stenosis of at least one of the brachiocephalic arteries, and 18.7% had more than a 50% diameter stenosis of the left subclavian artery. Additionally, investigators have reported angiographic prevalences of left SS between 0.5% and 4% in populations of patients with coronary artery disease (CAD) (15). Frank et al. (16) examined the association with various interarm pressure differences and clinical risk factors/CAD/PAD on a small scale (n = 134). Though that study failed to identify a relationship to gender, age, smoking, hypertension, diabetes, or CAD, but did demonstrate a significant relationship between having PAD and a systolic interarm pressure difference of ≥10 mm Hg, ≥15 mm Hg, ≥20 mm Hg, or ≥45 mm Hg.
Despite the suggested correlation between SS and PAD and the implications of SS for clinical diagnostic measurements, there has not been a large-scale study of the prevalence of or risk factors for SS in a free-living population. Furthermore, the relationship between SS and other atherosclerotic diseases, such as PAD, CAD, and stroke has received little attention. In this study we quantified the prevalence of SS in subjects in two free-living populations (cohorts A and B) and patients in two clinical groups (cohorts C and D), analyzed risk factors for SS, and explored the association between SS and the prevalence of three other atherosclerotic diseases: PAD, stroke, and CAD.
A total of 4,223 subjects (composite cohort) were included in the cross-sectional analysis. They represented four distinct cohorts: two (cohorts A and B) were from community population studies (17–19) and two (cohorts C and D) were from clinical studies enriched with vascular patients (20,21). After informed consent, subject information concerning demographic data, risk factors, and previous and current cardiovascular disease was determined variously by standardized subject interview, medical record review, primary care physician questionnaire, and physical examination.
In all studies, patient ethnicity was self-reported and was classified as non-Hispanic white (NHW), African American, Hispanic, or Other. “Systolic hypertension” was defined as SBP >140 mm Hg or previous physician diagnosis. “Pack-years” was defined as the average number of cigarettes smoked per day during the years smoked divided by 20 and multiplied by the number of years the subjects had smoked. Standardized laboratory tests were used to measure cholesterol levels. The concentration of high-density lipoprotein (HDL) cholesterol was determined using a direct enzymatic colorimetric assay in all cohorts (22).
“Coronary artery disease” was defined as having had a myocardial infarction (MI), percutaneous transluminal coronary angioplasty (PTCA), or coronary artery bypass graft (CABG) surgery (standardized electrocardiogram information or a history of angina was not available for all subjects). The definition of “stroke” was a previous physician diagnosis of the condition. Patients were defined as having “PAD” if they had an ankle brachial index (ABI) ≤0.90 in either leg or if they had already undergone surgery/angioplasty for PAD in either leg. This ABI threshold has been shown to be 90% sensitive and 98% specific for angiographically diagnosed PAD (23). For the purposes of this research, “SS” was defined as an interarm pressure difference of ≥15 mm Hg. Participants provided written consent for participation in the studies and the use of their data for future research. Institutional review board approval was obtained at the time of data collection from the University of California, San Diego (Cohorts A, B, and D) and Northwestern University and Catholic Health Partners Hospital (Cohort C). Furthermore, institutional review board approval was obtained for the current research project.
Cohort A consisted of 624 individuals who were all members of a geographically defined population study as part of a Lipid Research Clinics protocol (1978 to 1979) (17,19). The subjects were initially recruited through an introductory letter and telephone call and were all residents of a predominantly white, upper middle class community in southern California. A total of 59 subjects (9.4%) were excluded because of missing data. The remaining population used in the analysis consisted of 45.3% (n = 256) men and 54.7% (n = 309) women with a mean age of 65.9 ± 10.4 (range, 38 to 82) years. This population was all NHW.
Cohort B consisted of 2,410 individuals who were recruited between 1995 to 1999 (18). They were randomly selected within categories of gender, age, and ethnicity from a database of current and retired employees of the University of California, San Diego. This allowed an oversampling of minorities and women to increase power for testing selected hypotheses. The spouse or significant other of each randomly selected participant was also invited to participate in the study. The population consisted of 34.2% (n = 825) men and 65.8% (n = 1,585) women with a mean age of 59.3 ± 11.4 (range, 29 to 91) years. This population was 59.7% NHW, 13.5% African American, 14.6% Hispanic, and 12.2% other ethnicities.
Cohort C was composed of 740 individuals from a patient population in the Chicago area recruited between 1998 and 2000 (20). This cohort consisted of vascular laboratory and general medicine patients. Patients were excluded if they were wheelchair bound, had foot/leg amputations, were nursing home residents, non-English speaking, or had dementia. Additionally, patients with an ABI >1.50 were excluded because of possible arterial stiffness and thereby inaccurate pressure measurement (20). The population consisted of 56.2% (n = 416) men and 43.8% (n = 324) women with a mean age of 70.9 ± 8.4 (range, 55 to 93) years. This population was 78.1% NHW, 17.2% African American, 1.5% Hispanic, and 3.1% other ethnicities. A total of 62.4% of this cohort had PAD.
Cohort D included 508 patients who had visited the San Diego Veterans Administration Medical Center or the University of California, San Diego Medical Center vascular laboratories between 1990 and 1994 (21). The population consisted of 88.2% (n = 448) men and 11.8% (n = 60) women with a mean age of 68.6 ± 9.1 (range, 40 to 100) years. This population was 86.8% NHW, 4.5% African American, 5.5% Hispanic, and 2.8% other ethnicities. A total of 65.9% of the patients in this cohort had PAD.
Blood pressure and ABI measurement
Blood pressure measurement protocols differed slightly among the four cohorts. However, each of the methods used have been shown to produce accurate (and therefore comparable) results. To summarize, subjects in cohorts A and D had sequential measurements of brachial and ankle pressures, using 12 cm pneumatic cuffs (Hokanson, Bellevue, Washington, model SC-12) and either a mercury-in-silastic gauge (cohort A) or a photoplethysmographic sensor attached to the great toe (cohort D). The pressure at the site of the cuff was the pressure measured. The ABI in each leg for these cohorts was calculated using the measured ankle systolic pressure divided by the highest brachial blood pressure (left vs. right). Subjects in cohort B had SBPs measured sequentially in each arm, in addition to two measurements in each posterior tibial artery, using a CW Doppler with a 5 Mhz transducer (Medasonics, Mountain View, California). The ABI was calculated by dividing the average posterior tibial artery pressure from each side by the brachial artery measurement in the arm with the highest pressure. The pressures of subjects in cohort C were measured sequentially using a hand-held Doppler probe (Nicolet Vascular Pocket Dop II, Golden, Colorado). Each site had two sequential measurements taken. The ABI for each leg was calculated by dividing the highest average leg artery systolic pressure (either anterior tibial or posterior tibial) by the highest average brachial pressure (right vs. left).
All analyses were cross-sectional. The population cohorts (A and B) and the clinical cohorts (C and D) were examined separately for prevalence estimates. Multiple logistic regression analyses were conducted as a means of assessing independent effects on brachial blood pressure differences. The variables for gender, race, smoking status, diabetes, PAD, stroke, and MI/CABG/PTCA were included as categorical variables. Systolic blood pressure, age, HDL, and total cholesterol were included as continuous variables. Because systolic pressures were recorded rounded to the nearest 5 mm Hg in cohort A, pressures in cohorts B, C, and D were similarly rounded for uniformity of analyses. All analyses were performed using SPSS version 10.0 for Macintosh (SPSS Inc., Chicago, Illinois).
As shown in Table 1.the population cohorts (A and B) were composed of 2,975 subjects, with a mean (SD) age of 60.5 (11.5) years, of whom 36.3% (n = 1,081) were male. The sample was 67.4% NHW, 10.9% African American, and 11.9% Hispanic. The clinical cohorts (C and D) were composed of 1,248 subjects with a mean (SD) age of 70.0 (8.8). A total of 69.2% (n = 864) of the subjects were male, and 30.8% were female. This sample was 81.7% NHW, 12.0% African American, and 3.1% Hispanic. As expected, the clinical cohorts had a higher prevalence of cardiovascular risk factors than the population cohorts including more diabetes, more systolic hypertension, and more extensive smoking histories (Table 1). Furthermore, the clinical cohorts had a greater proportion of patients with a history of PAD, a previous stroke, or CAD.
Interarm SBP differences, right minus left, in each of the four cohorts are shown in Table 2.The mean interarm pressure difference in the two population cohorts was symmetrically distributed around zero, with the 10th percentile having the identical absolute value as the 90th, and the 5th percentile the same as the 95th, in both populations. In contrast, the two clinical cohorts were skewed significantly toward higher pressures in the right arm (ttest, p < 0.001), with the 90th and 95th percentiles exceeding the 10th and 5th percentile absolute values, respectively, in both cohorts.
Based on previous research by English et al. (15) implying a high specificity at a 15 mm Hg interarm pressure difference threshold, we selected this cut-point for defining SS. Of those in the combined cohort with an interarm difference of at least 15 mm Hg, 66.7% (n = 94) had a greater pressure in the right than in the left arm. Furthermore, Osborn et al. (24) demonstrated in a population of CABG candidates that a ≥15 mm Hg difference identified all patients (n = 59) with at least a 50% subclavian artery narrowing, and none of the patients with a 10 to 14 mm Hg difference has significant subclavian artery narrowing.
As shown in Table 3,the calculated prevalence (95% confidence interval) of SS in the population cohorts was 1.9% (1.8% in males and 1.9% in females). The prevalence of SS increased with age, from 1.4% in subjects below 50 years to 2.7% in subjects over the age of 70 years. As expected, there was a much greater prevalence of SS in the clinical cohorts: 6.0% in males and 9.7% in females. The prevalence of SS in the clinical cohorts increased from 4.3% in subjects aged 50 to 59 to 8.7% in subjects over 70 years of age.
In order to evaluate the appropriateness of combining the four cohorts for multivariate analysis, logistic regression models were run with each of the four cohorts for cardiovascular diseases and risk factors separately (eight models total), as well as for the population cohorts combined and the clinical cohorts combined for cardiovascular diseases and risk factors separately (four additional models). The results in these 12 models were quite consistent, the major difference being a somewhat higher risk estimate for PAD in the population than clinical cohorts. Nonetheless, in all four cohorts, the PAD association with SS was positive, independent, and highly significant.
We examined the independent effect of risk factors and other cardiovascular diseases on the presence of SS using multivariate logistic regression models. We did not analyze a model predicting SS by both cardiovascular risk factors and diseases simultaneously, because risk factors associated with both SS and other cardiovascular diseases would show a spuriously weakened association with SS in such a model (25). The first model addressed risk factors (Table 4.)The second model (Table 5)addressed cardiovascular diseases (PAD, stroke, and CAD).
Table 4shows a significant (p value < 0.05) association of age, SBP, the subject’s smoking history, and HDL levels with SS. For each decade of age, the risk of SS increased 1.21 times, and for a 20 mm Hg increase in SBP, the risk increased 1.90 times. Compared with subjects who had never smoked, those who had only smoked in the past had a 1.80 times greater risk of SS, and current smokers a 2.61 times greater risk. Though total cholesterol was not significantly predictive, increased HDL showed a protective effect: every 10 mg/dl increase in HDL reduced the risk of having SS by 13.1%. In the cardiovascular disease model (Table 5), PAD was shown to be strongly predictive of SS; increasing the risk by 5.11 times. Age was significantly associated with SS, but gender, ethnicity, previous stroke, and CAD were not. The borderline inverse association between CAD and SS (odds ratio = 0.66, p = 0.050) may have been an artifact of the strong association between CAD and PAD, because the univariate association between CAD and SS was positive.
Using angiographic information as a gold standard, English et al. (15) have suggested that an interarm pressure difference of ≥15 mm Hg has a sensitivity of approximately 50% and a specificity of 90% for detecting SS. The relatively low sensitivity/high specificity is likely due either to stenoses that are not significant enough to cause significant blood pressure differences, or to bilateral stenoses. This suggests that the actual prevalence of significant subclavian obstruction is greater than the prevalence based on ≥15 mm Hg interarm pressure difference.
As expected, the prevalence of SS in the clinical population (7.1%) was strikingly higher than the prevalence in the free-living population (1.9%). English et al. (15) reported a prevalence of left SS of 1.5% in patients with no PAD, 4.3% in hypertensive patients, 4.3% in patients with a history of smoking, 6.8% in patients with diabetes mellitus, 7.6% in patients with cerebrovascular disease, and 11.5% in patients with PAD. A reason for the somewhat lower prevalence of SS in the clinical cohorts in our study was the proportion of individuals without PAD (approximately one-third) in these cohorts. Furthermore, in addition to establishing the independent associations of age, current and past smoking history, SBP, and HDL on the interarm pressure difference, the multivariate logistic analyses highlighted the close relationship between SS and PAD.
The chief limitation of this study was that blood pressure measurement is an indirect method of detecting SS. However, in the study by English, et al. (15), the sensitivity of a 15 mm Hg interarm blood pressure threshold for subclavian stenosis was approximately 50%, and the specificity was approximately 90%. This threshold was also supported in a small scale study by Osborne, et al. (24). In conditions with low prevalence, such as subclavian stenosis, at a given level of sensitivity and specificity, the positive predictive value will be lower, and the negative predictive value higher than at higher prevalence levels. One clinical implication of the moderate sensitivity of the interarm pressure difference test is that our data likely underestimate the true prevalence of subclavian stenosis. Another limitation is that the clinical populations included all patients referred to the vascular laboratory, regardless of diagnosis. As a result, the measured prevalence of subclavian stenosis is somewhat lower than the expected prevalence, based on previous studies of PAD patients.
The Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure (JNC VII) emphasizes the importance of bilateral brachial blood pressure measurement (26). One justification for this requirement is that stenosis of the subclavian artery is typically asymptomatic, and, thus, unilateral measurement can result in an inaccurate estimation of systemic blood pressure. Unilateral measurement can be hazardous to the patient, potentially leading to a lack of treatment and/or optimal control of hypertension. Because blood pressure is often higher when the patient first enters the office, it is important for the patient to sit relaxed for a few minutes so that the most accurate bilateral blood pressures can be obtained (26). If an interarm systolic pressure difference of 10 mm Hg or greater is found, we believe this finding should be rechecked later in the visit for verification.
The finding that PAD was the strongest single predictor of SS emphasizes the importance of using the highest brachial pressure for the ABI denominator. A unilateral or average brachial pressure measurement could result in a falsely normal ABI (27) and a missed diagnosis of PAD (28). These findings emphasize the importance of bilateral blood pressure measurements, particularly in patients with recognized risk factors or PAD, to facilitate proper recognition and management of cardiovascular pathology.
Supported for the design and conduct of the study by the NIH and the Stein Institute for Research on Aging Research Fellowship (#NHLBI 5 T35 HL07491) and the UCSD Department of Family and Preventive Medicine Student Research Grant. Support for the data collection from grants #M01RR00827 and #RR00048 (National Center for Research Resources, NIH), #R01HL022255, #R01HL042973, #R01HL053487 (NIH). Additional support for data collection from grants #R01-HL58099 and #R01-HL64739 from the National Heart, Lung, and Blood Institute. Dr. McDermott is recipient of an Established Investigator Award from the American Heart Association. Dr. William Weintraub acted as the guest editor for this paper.
- Abbreviations and Acronyms
- ankle brachial index
- coronary artery bypass graft
- coronary artery disease
- high-density lipoprotein
- myocardial infarction
- non-Hispanic white
- peripheral arterial disease
- percutaneous transluminal coronary angioplasty
- systolic blood pressure
- subclavian artery stenosis
- Received February 18, 2004.
- Revision received April 4, 2004.
- Accepted April 16, 2004.
- American College of Cardiology Foundation
- Carter S.A.
- Cherry K.J.
- Fairbairn J.F.,
- Juergens J.L.,
- Spittell J.A.
- Kappert A.
- Orme S.,
- Ralph S.G.,
- Birchall A.,
- Lawson-Matthew P.,
- McLean K.,
- Channer K.S.
- Fotherby M.D.,
- Panayiotou B.,
- Potter J.F.
- Gutierrez G.R.,
- Mahrer P.,
- Aharonian V.,
- Mansukhani P.,
- Bruss J.
- Criqui M.H.,
- Fronek A.,
- Barrett-Connor E.,
- Klauber M.R.,
- Gabriel S.,
- Goodman D.
- Criqui M.H.,
- Fronek A.,
- Klauber M.R.,
- Barrett-Connor E.,
- Gabriel S.
- Bird C.E.,
- Criqui M.H.,
- Fronek A.,
- Denenberg J.O.,
- Klauber M.R.,
- Langer R.D.
- Sugiuchi H.,
- Uji Y.,
- Okabe H.,
- et al.
- Szklo M.,
- Nieto F.J.