Table 3. Subgroup analysis: Generalized Linear Models of fasting glucose levels and verapamil use among various subgroups of adults with diabetes.

	Fasting glucose (mg/dL)		Fasting glucose (mg/dL)
a. Restricted to those not on any anti-diabetic medications*
	Verapamil users, n = 15		Non-users of any CCB, n = 646
	LS-mean	(95% CI)	LS-mean	(95% CI)	Mean difference	P-value for difference
Model	153.8	132.2–175.6	155.0	150.8–159.1	1.1	0.91
b. Restricted to those only on oral anti-diabetics
	Verapamil users, n = 116		Non-users of any CCB, n = 2049
	LS-mean	(95%CI)	LS-mean	(95%CI)	Mean difference	P-value for difference
Model	129.8	120.7–138.9	135.7	133.1–138.4	6.0	0.19
c. Restricted to those on insulin (± oral anti-diabetics)
	Verapamil users, n = 43		Non-users of any CCB, n = 799
	LS-mean	(95% CI)	LS-mean	(95% CI)	Mean difference	P-value for difference
Model	131.7	109.1–154.4	155.9	149.0–162.7	24.1	0.039
d. Restricted to those only on insulin
	Verapamil users, n = 15		Non-users of any CCB, n = 319
	LS-mean	(95% CI)	LS-mean	(95% CI)	Mean difference	P-value for difference
Model	118.8	80.0–157.7	156.3	145.5–167.0	37.4	0.06

Abbreviations: ACR—albumin to creatinine ratio, CCB—calcium channel blocker, CHD—coronary heart disease; LS-mean—least square mean. Model adjusts for age, race, education, systolic blood pressure, total and HDL-cholesterol, log of ACR, baseline CHD, waist circumference, medication adherence, smoking, heart rate and hypertension (only variables significant in bivariate comparisons).

^*Antidiabetic medication defined by self-report or pill bottle review.