FIGURE 1.

Hazard ratios for cardiovascular events according to continuous hemoglobin A_1c (HbA_1c) levels among participants without known diabetes. Restricted cubic spline models with the inclusion of transformed variables in the Cox model were used to estimate hazard ratios (solid curve) with point-wise 95% confidence intervals (grey shaded area) for (A) cardiovascular disease, (B) stroke, and (C) coronary heart disease. An HbA_1c level of 5.3% (ie, the mean HbA_1c level in people with HbA_1c levels of 5.0–5.5%) was used to estimate all hazard ratios. We chose the number of knots that produced the smallest Akaike Information Criterion. Hazard ratios were adjusted for age, sex, public health center areas, body mass index, smoking status, alcohol intake, sports and physical exercise, systolic blood pressure, non-high-density lipoprotein cholesterol, and high-density lipoprotein cholesterol.