Predicting Changes in Cardiovascular Risk Factors in Type 2 Diabetes in the Post-UKPDS Era: Longitudinal Analysis of the Swedish National Diabetes Register

Table 2

Estimates of GMM for risk factors from NDR data.

	(1)		(2)		(3)		(4)		(5)
Risk factors	HbA1c^b		Systolic BP^c (mmHg)		TC : HDL		LDL (mmolL^−1)		BMI (Kg/m^2)
	Coefficient	Long-term effect	Coefficient	Long-term effect	Coefficient	Long-term effect	Coefficient	Long-term effect	Coefficient	Long-term effect
Constant	3.055***	—	5.546**	—	1.647**	—	1.938***	—	6.832*	—
Ln (diabetes duration)	0.182***	0.384	-0.050	−0.094	0.099	0.216	−0.059***	−0.091	0.106***	0.561
Year 1^a	−0.144*	−0.304	NA	—	NA	—	NA	—	NA	—
Age at diagnosis	−0.008***	−0.017	0.021*	0.039	−0.009**	−0.020	−0.007***	−0.011	−0.022**	−0.116
Female	−0.034*	−0.072	−0.064**	−0.120	−0.144***	−0.314	0.060***	0.093	0.134	0.709
Smoking	0.192	0.405	0.094***	0.176	0.071	0.155	0.091	0.140	−0.370***	1.958
BMI	0.015***	0.032	0.017***	0.032	0.021***	0.046	0.018**	d	—	—
BMIsquared	NS	—	NS	—	NS	—	−0.0003**	d	—	—
Lag of HbA1c	0.526***	—	NA	—	NA	—	NA	—	—	—
Lag of SBP	NA^e	—	0.466**	—	NA	—	NA	—	—	—
Lag of TC : HDL	NA	—	NA	—	0.541***	—	NA	—	—	—
Lag of LDL	NA	—	NA	—	NA	—	0.352***	—	—	—
Lag of BMI	NA	—	NA	—	NA	—	NA	—	0.811***	—
𝑁 (patients)	4450		4158		2274		2068		4284
Person years	20699		20144		10157		9536		25447
Hansen test	0.053		0.588		0.654		0.461		0.126
AR1^f	<0.001		<0.001		<0.001		<0.001		<0.001
AR2	0.356		0.095		0.219		0.310		0.078

^***,**,*Denote significance level at the 1, 5, and 10%, respectively; Hansen test on overidentifying restrictions; AR1 and AR2 show the test on first and second order autocorrelation, respectively. ^aYear 1: 1 if diabetes duration = 1 year, 0 otherwise. ^bIn HbA_1c equation, the smoking was an endogenous variable. ^cSystolic BP values transformed as systolic BP/10. ^dAs the BMI squared is significant, the long-term effect will be different for different levels of BMI. ^eThe covariate was not included in the estimation. ^fThe strong evidence against null hypothesis of no autocorrelation in the first-differenced errors is expected in the model.