The mean age of the population was 71.0 years; 14.6% were over the age of 80 years (). The population was ethnically diverse, with 45.9% of subjects composed of nonwhite racial/ethnic groups.
Population description by baseline A1C categories
The mean A1C was 7.0%. A total of 15% of the population had an A1C <6.0%, 60% had an A1C <7.0%, and 6.5% had an A1C ≥9.0%. The most frequently prescribed medications for glucose control were sulfonylureas (50.6%), followed by metformin (37.7%), and any form of insulin (17.5%). The mean duration of diabetes was 8.3 years. Across A1C categories, patients with lower baseline A1C values tended to be older and more likely to be non-Hispanic white. Patients with lower A1C values also were more likely to have a shorter duration of diabetes, better cholesterol control, and lower GFR, but less evidence of other microvascular complications (e.g., microalbuminuria, laser photocoagulation). Patients with lower A1C levels also were much less likely to be treated with insulin.
The mean follow-up time was 3.1 years. Chronic cardiovascular events had the highest incidence (47.2 per 1,000 person-years), followed by mortality (40.4 per 1,000 person-years), chronic microvascular events (26.7 per 1,000 person-years), and acute metabolic events (1.2 per 1,000 person-years) ().
Baseline A1C, complications, and mortality: overall population results*
The risk of acute metabolic events, based on point estimates, increased steeply with each unit change in A1C above 6.0%; this increased risk became statistically significant above the A1C of 7.0%. The adjusted HR for an A1C of 7.0–7.9% was 2.35 (95% CI 1.31–4.23) and increased to 11.52 (5.166–23.47) at an A1C ≥11.0% (reference: A1C <6.0%). Adjusted HRs were somewhat attenuated compared with unadjusted HRs, but the steep, monotonic pattern persisted.
There was a similar, stepwise relationship between baseline A1C and chronic microvascular events. In the unadjusted model, the increased risk associated with higher glucose levels was significantly different than the reference (A1C <6.0%), starting at A1C 6.0–6.9% (HR 1.56 [95% CI 1.41–1.72]). The A1C–microvascular event relationship was attenuated in the adjusted model but still suggested an increased risk beginning at A1C ≥6.0% (1.11 [0.99–1.25]). The risk became significantly higher than the reference for A1C levels ≥7.0% (1.25 [1.11–1.41]) for A1C 7.0–7.9%.
For chronic cardiovascular events, A1C levels ≥6.0% also were associated with a significantly increased risk in both unadjusted and adjusted models. The risk increased continuously without a clear threshold level, although the risk increase was less steep than that observed for acute metabolic or chronic microvascular events.
Unlike the nonfatal complications, mortality had a U-shaped relationship with baseline A1C. In unadjusted models, the risk of mortality was significantly lower for A1C levels between 6.0 and 8.9% relative to A1C levels <6.0%. After adjustment, this general pattern of lower risk in the midrange of A1C still was observed, although the lower risk for A1C levels between 8.0 and 8.9% was no longer statistically significant. The mortality risk did not differ statistically between reference group (A1C <6.0%) and A1C ≥9.0% in the unadjusted model, although the point estimates for A1C ≥10.0% indicated a somewhat higher risk. In adjusted models only, mortality risk became significantly higher once A1C levels exceeded 10.0% (HR 1.21 [95% CI 1.01–1.45]).
When evaluating “any complication,” we found that the risk rose steadily with each incremental rise in baseline A1C without a clear threshold, although the steepness of the relationship was attenuated in adjusted models. After adding death to this combined end point (i.e., “any complication or death”), the risk increased significantly at A1C ≥6.0% in the unadjusted model. However, after adjustment, the risk increase was not significant until A1C was ≥8.0%.
The epidemiologic patterns observed in the overall population were similar across the three age-groups (60–69, 70–79, and ≥80 years), with some notable exceptions (). All three age-groups had U-shaped mortality curves with the highest risk of death among patients with A1C levels at the extremes (A1C <6.0 and ≥9.0%). For patients aged 60–69 years, the lowest point estimate for mortality risk was observed with A1C levels between 7.0 and 7.9%. For patients aged ≥70 years, the mortality risk was statistically lower across a broader range of A1C categories (e.g., A1C 6.0–7.9% for the aged ≥80 years group) compared with the reference group. For the “any complication” outcome, patients aged 60–69 years had a continuous, positive relationship between A1C and complications with no clear threshold. For older patients, the increased risk (relative to the reference) of any complications was significantly higher at the threshold of A1C ≥7.0%. For the “any complication or death” outcome, there was an increased risk in the outcome for all age-groups when A1C exceeded 8.0%, although for patients aged >80 years, this was statistically significant only for A1C ≥9.0%.
Age-stratified results: adjusted analyses*
In analyses of effect modification, the relationships between A1C and mortality or the combined outcomes were not significantly different for those with differing durations of diabetes (data not shown). In a sensitivity analysis using extended Cox models that accounted for the time-varying nature of A1C, the overall forms of the relationships between A1C and complications did not change from the baseline analyses, although the strengths of the associations weakened, particularly for chronic complication events (data not shown). For extended Cox models of the “any complication” outcome, A1C was significantly associated with a higher risk of events at A1C ≥8.0%, instead of A1C ≥6.0%, whereas for the “any complication or death” outcome, A1C became significantly associated with higher risk at A1C ≥10%, instead of A1C ≥8.0%.