shows first the distribution of age, gender, and ethnicity across the low, normal and high ABI groups. Both low and high ABI groups were somewhat older than the normal group, and women were more common in the low and men more common in the high ABI group. Compared to NHWs, African-Americans were more likely to be in the low ABI group, while Hispanics and Chinese were less likely. Chinese were also less likely to be in the high ABI group. Because of these differences, the baseline risk factor levels in the remainder of were adjusted for age, gender, and ethnicity.
Education and income levels were lower in the low ABI group. With the exception of BMI, traditional CVD risk factors were uniformly significantly higher in the low ABI group, while only BMI and diabetes were higher in the high ABI group. Among the six novel biomarkers, hs-CRP, Il-6, and homocysteine were all significantly higher in both the low and high ABI groups. Fibrinogen and D-dimer were higher, and eGFR borderline lower, in the low ABI group.
There were 317 first CVD events during follow-up, 226 CHD events, 89 strokes (including 7 who also had CHD events), and 9 were “other” atherosclerotic CVD deaths. CVD event rates per 1000 person-years in men for NHW, African-Americans, Hispanics, and Chinese were 14.9, 14.5, 16.1, and 10.2 respectively, and for women 8.4, 7.4, 6.4, and 3.7 respectively. shows Kaplan-Meier event curves illustrating the similarly increased hazard of CVD events for both a low and a high ABI over time compared to a normal ABI. shows the results of the proportional hazards analysis. Adjusted for age, gender, and ethnicity (model 1), a low ABI was associated with a hazard ratio of 2.30 (p<.001) for CVD events. Adjustment for traditional CVD risk factors in addition (model 2) attenuated the HR to 1.78 (p<.001). Further adjustment (model 3) for newer biomarkers produced no change (HR = 1.77, p<.001). Finally, adjustment for multiple subclinical CVD measures, CAC, both common and internal CIMT, and major ECG abnormalities attenuated the HR somewhat to 1.46, but the association remained significant (p=.021).
Kaplan-Meier Event Curves for Incident CVD by ABI Categories
Risk of Incident Cardiovascular Events by ABI level
A high ABI was also associated with an increased risk, and after multiple adjustment (model 3) the HR was 1.85, p=.050, a similar effect size as for the low ABI. Additional adjustment CAC, common and internal CIMT, and major ECG abnormalities attenuated the HR somewhat to 1.69, p= .099. Tests of the above models for gender-ABI interactions were not significant. After model 2 adjustment for the categorical ABI analysis the AUC was 0.78 without and 0.79 with the ABI, p=0.022 for the difference, and the IDI index showed that the ABI contributed significantly (p= 0.003) to improving classification of events and non-events. shows that results were similar for analyses with ABI as a continuous variable (excluding ABIs ≥ 1.40), and a generalized additive model fit to examine potential non-linearity indicated that after covariate adjustment there was little deviation from a linear inverse relationship across the ABI values.
illustrates the risk of CVD events by ABI group within strata of CAC. In those without any CAC, incidence rates were quite low irrespective of ABI group. In the groups with non-zero CAC, ABI showed a U-shaped association within CAC groups. As a continuous variable, ABI was inversely and significantly related to the CVD event rate among those with any CAC (data not shown). Models limited to “hard” CVD events; i.e., excluding angina and “other” CVD deaths, showed similar results except for reduced statistical power (data not shown).
Unadjusted Rates of Incident CVD per 1000 Person Years at Risk by Categories of ABI and Coronary Artery Calcium
Additional models evaluated the association of the ABI defined continuously (excluding ABIs ≥ 1.40) within ethnic groups, adjusted for age and gender. The hazard ratios were quite similar: 0.74 for Whites; 0.80 for Hispanics; 0.83 for Chinese; and 0.87 for African-Americans. Although power was insufficient to analyze the ABI categorically within ethnic groups, an additional Cox model was run including an interaction term for ABI group × ethnicity. The p value for the interaction term was 0.71, confirming the similar effect size for ABI within ethnic groups.
Models were also run for CHD and stroke events separately, using model 2 adjustments. For CHD endpoints, the low and high ABIs again showed similar hazard ratios, 1.87 (p=.001) and 2.15 (p=.029). However, the results appeared to differ for stroke, where high ABI was e associated with a HR of 2.69, p=.06, while low ABI showed a weaker association, HR 1.56, p=.10.