Low Ankle-Brachial Pressure Index Predicts Increased Risk of Cardiovascular Disease Independent of the Metabolic Syndrome and Conventional Cardiovascular Risk Factors in the Edinburgh Artery Study

The Edinburgh Artery Study recruited 1,592 people (809 men and 783 women) ages 55–74 years in 1988 from a general northern European white population sample. Full details of recruitment, data collection at baseline in 1988–1989, methods of follow-up, and definitions of outcomes have been previously described (19,20). The study population was selected to provide a population-based sample by random identification of eligible individuals in 5-year age bands from 11 general practices across the city of Edinburgh that provide primary care for populations of varying socioeconomic status. The response rate was 65%. Respondents were found to be representative of the wider population, as comparisons with a random sample of 20% of nonrespondents did not detect substantial bias (19). Baseline data were collected during participants’ attendance at a research clinic using a combination of questionnaire and examination data. Complete mortality follow-up was achieved by flagging for deaths at the National Health Service Central Registry in addition to being notified of deaths by general practitioners and family members. For this analysis, follow-up was undertaken until the date of death or for ∼15 years until the end of April 2003. The study was approved by the Lothian Health Board Ethics Committee, and informed consent was obtained from each participant.

Participants completed a self-administered questionnaire at baseline that included the WHO angina and intermittent claudication questionnaires (21). Height and weight were measured using standard methods. Participants were asked to indicate whether they had received a diagnosis of diabetes from a doctor. Systolic blood pressure was measured in the right arm using a random zero sphygmomanometer with the subject in the supine position after a 10-min rest. The ankle systolic blood pressure at the posterior tibial artery was measured where possible in both legs using a random zero sphygmomanometer and a Doppler probe.

A fasting blood sample was collected from the antecubital vein using a tourniquet while the subject was recumbent. Glucose, total cholesterol, HDL cholesterol, and triglycerides were measured using a Cobas Bioanalyser (Roche, Welwyn Garden City, U.K.) and standard kits. Participants received an oral glucose tolerance test in the form of 75 g of glucose in 335 ml of Solripe Gluctoza Health Drink (Strathmore Mineral Water, Angus, U.K.); a further blood specimen was collected 2 h later for the measurement of glucose using the same method as for the fasting sample.

Prevalent CVD at baseline was defined as stroke (recall of a doctor’s diagnosis), angina (WHO questionnaire evidence and electrocardiographic ischemia or recall of a doctor’s diagnosis), or myocardial infarction (two out of three of recall of a doctor’s diagnosis, evidence from responses to the WHO questionnaire, and electrocardiographic evidence). Diabetes at baseline was defined using recall of a doctor’s diagnosis and WHO 1999 criteria (fasting glucose ≥7 mmol/l or 2-h glucose ≥11.1 mmol/l) (8) applied to data collected during the oral glucose tolerance test. Smoking status was defined as self-report of current smoking at baseline.

Criteria used to define cardiovascular outcomes were adapted from international diagnostic criteria developed by the American Heart Association (22) and have been described in detail elsewhere (19). In brief, cardiovascular death was defined as postmortem evidence of acute myocardial infarction, cerebral infarction, or hemorrhage or from death certificates with an underlying cause of death in the range of CVD codes from ICD-9 (codes 410–414, 430–438, and 440–445) and ICD-10 (I21–25 and I60–73). Nonfatal cardiovascular outcomes included myocardial infarction, stroke or transient ischemic attack, angina, intermittent claudication, revascularization, and amputation. All possible cardiovascular outcomes were investigated using medical records, and only those that fulfilled the protocol criteria for a definite event were included in the analysis. Questionnaire and clinic measurement data were checked by clinic staff, coded, and entered onto a DBASE IV database. Double data entry was used and discrepant entries were checked against original records.

A low ABI was defined as ABI <0.9. Participants with an ABI >1.4 were excluded because of possible erroneously high levels due to arterial stiffness.

The ATP III criteria for the metabolic syndrome include the presence of three or more of the following: waist circumference ≥102 cm (men) or ≥88 cm (women); blood pressure 130/85 or treatment for hypertension; triglycerides >1.7 mmol/l (>150 mg/dl); HDL cholesterol <1.03 mmol/l (<50 mg/dl) (men) or <1.29 mmol/l (<50 mg/dl) (women); and fasting plasma glucose ≥6.1 mmol/l (>110 mg/dl) (9). Waist circumference was not available in the Edinburgh Artery Study, and therefore the waist circumference criteria used in the ATP III definition were replaced with cut points for BMI (28.8 kg/m² for men and 26.7 kg/m² for women), as in previous studies (16,22). The cut point for men was identified as being equivalent to a waist measurement of 102 cm in a regression analysis of a cross-sectional study (16). The criterion for women was developed from the Women’s Health Study using the BMI value that corresponded to the same percentile cut point for a waist circumference of 88 cm measured at the same time during follow-up (23).

Statistical analysis was performed using Stata software (Stata, College Station, TX). Geometric means and transformed confidence intervals are presented for triglyceride levels that were positively skewed. ANOVA was used to test for differences in baseline characteristics between groups defined by the presence or absence of a low ABI and the metabolic syndrome, with t tests and χ² tests used to determine differences between groups for continuous and categorical variables. Hazard ratios for CVD were estimated using Cox proportional hazards modeling for mortality and for all CVD with adjustment for age and sex in a basic model and adjustment for various other CVD risk factors in subsequent models. Likelihood ratio tests were used to compare full-adjusted models that included and excluded ABI to determine whether a low ABI contributed to the risk of CVD outcomes.

Of the 1,592 participants in the Edinburgh Artery Study, 8 had missing data for ABI, 13 had an ABI >1.4, 31 had data missing for one or more criteria of the metabolic syndrome, and 2 had data missing for both ABI and the definition of the metabolic syndrome. A further 71 had data missing on diabetes or smoking status at baseline. Data for the remaining 1,467 people were included in our analysis. Of this population, >17% had a low ABI and 25% had the metabolic syndrome, with 6% having both a low ABI and the ATP III–defined metabolic syndrome. Characteristics of the subgroups defined by the presence or absence of a low ABI and the metabolic syndrome are given in Table 1. Subjects with both a low ABI and the metabolic syndrome tended to have a more adverse cardiovascular risk factor profile. During the follow-up period, 554 people died, 226 from CVD. Just over half of the people that died of CVD (118 people) had a nonfatal cardiovascular event before death. A further 462 people had at least one nonfatal confirmed cardiovascular event but did not die from CVD during the study, giving a total number of cardiovascular events of 688.

Tables 2 and 3 show hazard ratios for cardiovascular mortality and for combined fatal and nonfatal CVD derived from Cox proportional hazards models of increasing complexity. A low ABI was a statistically significant risk factor for CVD mortality independent of the metabolic syndrome and conventional cardiovascular risk factors: hazard ratio 1.53 (95% CI 1.13–2.07; P = 0.006). The hazard ratio for the metabolic syndrome for CVD mortality did not achieve conventional levels of statistical significance after full adjustment for other CVD risk factors (1.30 [0.97–1.75]; P = 0.07), and the hazard ratio was further attenuated by the addition of a low ABI to the model (1.26 [0.94–1.70]; P = 0.13). The addition of a low ABI to the model with the outcome of CVD mortality and independent variables of age, sex, metabolic syndrome, LDL cholesterol, baseline CVD, known or newly diagnosed diabetes, and smoking status at baseline significantly improved the fit of the model (P = 0.007, change in −2 log likelihood 7.2 with one degree of freedom [df]). A low ABI was also a statistically significant risk factor for the combined outcome of fatal and nonfatal CVD independent of the metabolic syndrome and conventional cardiovascular risk factors (hazard ratio 1.46 [1.20–1.76]; P < 0.0001). The statistical significance of the metabolic syndrome as a risk factor for all CVD independent of a low ABI and conventional risk factors was borderline (hazard ratio 1.20 [1.00–1.43]; P = 0.05). Adding a low ABI to a model with all CVD as the outcome with adjustment for all other factors mentioned above significantly improved the fit of the model (P = 0.002, change in −2 log likelihood 14.3 with 1 df).

We have shown that a simple measure of preclinical atherosclerosis, a low ABI, is associated with an increased risk for CVD independent of the metabolic syndrome and conventional cardiovascular risk factors. This measurement could be applied in clinical practice to identify patients at high risk of CVD. The excess risk of CVD associated with the metabolic syndrome was of borderline statistical significance after adjusting for conventional cardiovascular risk factors and a low ABI; this may have been due to the limited power of this study to detect an independent effect of the metabolic syndrome.

The IDF recently released a worldwide consensus definition of the metabolic syndrome, stated as a waist circumference ≥94 cm (men) or ≥80 cm (women) and two or more of the following factors: fasting plasma glucose 5.6 mmol/l and the same blood pressure and lipid criteria as for the ATP III definition of the metabolic syndrome (10). In our study, data for the general population ages 55–74 years in the Health Survey for England 1999 were used to generate BMI cut points in place of the waist cutpoints used in the IDF definition for the metabolic syndrome that had similar sensitivity and specificity to those used as proxy measurements for the ATP III criteria (S.H.W., unpublished observations). These values were 26.2 kg/m² for men and 24.2 kg/m² for women. Using the modified IDF criteria, over a third (34%) of the study population had the metabolic syndrome compared with 25% using the modified ATP III criteria. The hazard ratios for fatal and all CVD for a low ABI adjusted for age, sex, metabolic syndrome, and cardiovascular risk factors were similar for the ATP III–and IDF-defined metabolic syndrome (data not shown but available from S.H.W.). The adjusted hazard ratios for fatal and all CVD for the metabolic syndrome were generally lower for the IDF definition than for the ATP III definition.

The advantages of this study were its long follow-up period and the high number of fatal and nonfatal outcomes. Work before the start of this study showed that a single measurement of ABI is suitable for most epidemiological studies (24). The limitations of this study were those common to many cohort studies with lengthy follow-up, such as the use of baseline data only, which does not allow for adjustment for regression dilution bias or changing prevalence of risk factors. It was not possible to validate reports of doctor diagnoses of myocardial infarction, stroke, or intermittent claudication in the baseline data. Self-report of angina or myocardial infarction appears to be reasonably valid in men (25) but less accurate in women (26) when comparisons are made against medical records. In this study, further information from questionnaires and electrocardiographic findings were used to define CVD in addition to self-report of a doctor’s diagnosis.

Patterns of cardiovascular risk factors have changed in recent years, and it is likely that the distribution of the metabolic syndrome components have changed. In a contemporary cross-sectional study, a lower prevalence of hypertension and hypercholesterolemia might be expected than was observed in this population at baseline in 1988–1989. However, the current prevalence of the metabolic syndrome may be similar to the 1988–1989 level because of the increasing prevalence of central obesity and dysglycemia, and there is evidence to suggest that the prevalence of the metabolic syndrome is increasing (27). At present it is not clear whether certain combinations of metabolic syndrome components confer different risks from other combinations. Smoking habits have also shown secular changes, which is likely to affect the prevalence of a low ABI. Although waist circumference measurements were not available for Edinburgh Artery Study data, we were able to use a proxy measure of BMI that has been shown to reflect central obesity in other populations, one of which was a cohort of Scottish men (16,23).

A previous cross-sectional study reported no independent effect of the metabolic syndrome on the prevalence of CVD beyond that associated with the individual components of the syndrome and diabetes (28). This finding may have been influenced by survival bias, and it contrasts with those of a cohort study that found the metabolic syndrome was a stronger predictor of coronary heart disease, CVD, and total mortality than its individual components (13). Moreover, in a prospective study of subjects with newly diagnosed type 2 diabetes, one of us (C.D.B.) has shown that there is a progressive decrease in survival over ∼5 years with the presence at diagnosis of each additional feature of the metabolic syndrome defined by ATP III criteria (29). Based on data from the Framingham offspring cohort study of 3,323 men and women (mean age 52 years) with 8 years follow-up, it has been estimated that the metabolic syndrome (as defined using ATP III criteria) contributes almost half of the population-attributable risk for diabetes and ∼25% of all incident CVD (30).

In summary, a low ABI can be used to identify individuals at increased risk of CVD and thus provides information on risk beyond that provided by a prevalence of conventional cardiovascular risk factors and the presence of the metabolic syndrome (whether defined using ATP III or IDF criteria). ABI can easily be measured in a primary care setting and could be used as part of the assessment of cardiovascular risk in individuals if further research confirms its cost-effectiveness.