Table 4. Trend associations across groups divided by frequency of hypoglycemia and changes in IMT.

Variable	Regression coefficient (95% Confidence Interval)	P value
Change in mean IMT-CCA
Model 1	0.047 (0.019–0.075)	<0.001
Model 2	0.048 (0.021–0.075)	<0.001
Model 3	0.043 (0.016–0.070)	0.002
Model 4	0.050 (0.012–0.088)	0.01
Model 5	0.054 (0.015–0.093)	0.007
Model 6	0.052 (0.013–0.092)	0.01
Model 7	0.047 (0.005–0.088)	0.027
Model 8	0.044 (0.014–0.074)	0.005
Change in right max IMT-CCA
Model 1	0.052 (−0.014–0.118)	0.12
Model 2	0.064 (0.001–0.128)	0.045
Model 3	0.058 (−0.005–0.122)	0.073
Model 4	0.046 (−0.046–0.138)	0.33
Model 5	0.049 (−0.046–0.143)	0.31
Model 6	0.049 (−0.046–0.143)	0.31
Model 7	0.054 (−0.045–0.153)	0.28
Model 8	0.069 (−0.002–0.139)	0.058
Change in left max IMT-CCA
Model 1	0.077 (0.019–0.134)	0.009
Model 2	0.078 (0.026–0.130)	0.004
Model 3	0.070 (0.018–0.123)	0.009
Model 4	0.089 (0.014–0.164)	0.02
Model 5	0.093 (0.016–0.169)	0.017
Model 6	0.088 (0.012–0.164)	0.024
Model 7	0.082 (0.002–0.161)	0.045
Model 8	0.069 (0.011–0.127)	0.021

Model 1: Trend estimation for linear trends across quintiles is based on linear regression analysis for continuous variables unadjusted (Model 1), adjusted for model 1 plus baseline IMT (Model 2), model 2 plus age, gender, baseline IMT and the original treatment group (Model 3), model 2 plus body mass index and current smoking (Model 4), model 4 plus HbA1c, total cholesterol, high density lipoprotein-cholesterol, triglyceride and systolic blood pressure (Model 5), model 4 plus estimated glomerular filtration rate, use of angiotensin-converting enzyme inhibitors/angiotensin II receptor blockers, use of statins and use of anti-platelets (Model 6), model 6 plus the use of sulfonylurea, the use of glinides and the use of α-glucosidase inhibitors (Model 7), model 7 plus C-peptide, high-sensitivity C-reactive protein and interleukin (Model 8). CCA, common carotid artery; IMT, intima-media thickness.