. 2019 May 24;160(7):1731–1742. doi: 10.1210/en.2019-00165

Table 2.

Top Hits From Logistic Regression Analysis

Metabolite^a	Estimate^b	SE	z Value	P Value	Adjusted P Value
Histidine	−0.188	0.074	−2.529	0.011	0.4274
Glutamine	−0.175	0.069	−2.525	0.012	0.4274
DHA.FA	0.195	0.081	2.393	0.017	0.5214
FAw3.FA	0.170	0.075	2.272	0.023	0.6203
M.HDL.P	−0.151	0.072	−2.085	0.037	0.7646
M.HDL.L	−0.149	0.072	−2.076	0.038	0.7695
DHA	0.153	0.075	2.029	0.043	0.7975
M.HDL.PL	−0.145	0.072	−2.020	0.043	0.8016
M.HDL.C	−0.139	0.071	−1.941	0.052	0.8513
M.HDL.CE	−0.138	0.072	−1.929	0.054	0.8589
M.HDL.PL	0.141	0.074	1.898	0.058	0.8756

Abbreviations: DHA, docosahexaenoic acid; HDL, high-density lipoprotein; DHA.FA, ratio of docosahexaenoic acid to all fatty acids; FAw3.FA, ratio of ω-3 fatty acids to total acids; M.HDL.P, concentration of medium HDL particles; M.HDL.PL: phospholipids in medium-sized HDL particles.

^{^a}

Logistic regression with single metabolite measure, sex, age, BMI, smoking status, T2DM status, fasting status, and cohort as covariates.

^{^b}

The estimates are the fitted β coefficients from the logistic regression model. As the input metabolite data were scaled, the estimates can be interpreted as follows: the OR for developing pancreatic cancer in a case with a typical low metabolite score of 1 SD below the average z score (= −1) would amount to 1.22 for β of −0.1 and 1.49 for β of −0.2. The z value mentioned in the table is the test statistic from the logistic regression models.