Table 3.

Results of the multivariate linear mixed models for the change in fasting glucose level

	Coefficient (β)	Standard error	p
Variables at baseline
Sex (male)	0.111	0.002	< 0.001
Age (years)	0.003	< 0.001	< 0.001
Baseline fasting glucose (mmol/L)	− 0.764	0.001	< 0.001
Repeatedly measured variables at NHIS health examination
Systolic blood pressure (per 10 mmHg)	0.028	< 0.001	< 0.001
Body mass index (kg/m2)	0.023	< 0.001	< 0.001
Current smoker	0.021	0.002	< 0.001
Alcohol consumption, frequency per week (times)
< 1	Ref
1–2	0.034	0.002	< 0.001
3–4	0.094	0.002	< 0.001
≥ 5	0.116	0.003	< 0.001
Exercise, days per week (days)
< 1	Ref
1–4	− 0.005	0.001	< 0.001
≥ 5	− 0.011	0.002	< 0.001
Household income
Q1, low	Ref
Q2, middle	− 0.004	0.002	0.015
Q3, high	− 0.011	0.002	< 0.001
Time period from baseline to each health examination (year)	0.035	< 0.001	< 0.001
PDC by any statina	0.093	0.007	< 0.001
PDC by fibrateb	0.022	0.025	0.387
PDC by ezetimibe	0.046	0.045	0.314
anDDD by any statinc	0.119	0.009	< 0.001

Data are derived from the linear mixed models for the change in fasting glucose (mmol/L). These variables were included as fixed-effect variables

PDC proportion of days covered by statins during the time period from baseline to serial health examination (ranged from 0 to 1, treated as a continuous variable)

^aValue is the estimated change in fasting glucose when whole time period is covered by statins (PDC = 1)

^bBezafibrate, ciprofibrate, etofibrate, fenofibrate, and gemfibrozil

^cWhen ‘anDDD by any statin’ were included in the multivariate linear mixed model instead of ‘PDC by any statin’