Table 3. Binary logistic regression models for incident T2DM (n = 605).

A)
Incident Diabetes (overall)	Parameter	Beta	p Value
Model 1	Glucose	17.6	7.33E-13
Model 2	TG	1.2	0.002
Model 3	1-deoxySO	1.02	0.008
Model 4	1-deoxySA	1.54	0.001
Model 5	1-deoxySL	1.22	0.004
Model 6	Glucose	17.34	1.95E-12
	1-deoxySO	0.85	0.034
Model 7	Glucose	17.22	3.04E-12
	1-deoxySA	1.32	0.007
Model 8	Glucose	17.32	2.15E-12
	1-deoxySL	1.04	0.019
Model 9	Glucose	17.42	1.84E-12
	TG	1.12	0.006
Model 10	Glucose	17.32	2.60E-12
	TG	0.85	0.078
	1-deoxySL	0.53	0.312
B)
Incident Diabetes (according to BMI)	Parameter	Beta	p Value
Model 1 (BMI < = 25, n = 99)	glucose	20.48	0.005
	1-deoxySA	3.66	0.009
Model 2 (BMI < = 27, n = 203)	glucose	17.61	2.29E-04
	1-deoxySA	2.47	0.005
Model 3 (BMI < = 30, n = 406)	glucose	14.56	1.31E-06
	1-deoxySA	1.47	0.0165
Model 4 (BMI < = 25, n = 99)	glucose	20.52	0.005
	TG	0.36	0.797
	1-deoxySA	3.41	0.043
Model 5 (BMI < = 27, n = 203)	glucose	18.26	1.66E-04
	TG	1.14	0.194
	1-deoxySA	1.84	0.065
Model 6 (BMI < = 30, n = 406)	glucose	14.62	1.31E-06
	TG	0.69	0.205
	1-deoxySA	1.07	0.121

All models were adjusted for age, gender, BMI, lipid lowering drugs (y/n). Models at baseline were also adjusted for the metabolic syndrome (y/n). All variables were log transformed. Variables in bold are statistically significant (p<0.05). SA, sphinganine; SO, sphingosine; TG, triglycerides