Skip to main content

Table 4 Difference in type 2 diabetes incidence per 10,000 person-years per 1-point HLS increasea

From: Healthy lifestyle, metabolomics and incident type 2 diabetes in a population-based cohort from Spain

Adjustment models

Difference in type 2 diabetes rates

per 10,000 person-years (95% CI)

Percent

explained, % (95% CI)c

Age, sex, education, prevalent hypertension, total plasma cholesterol, use of lipid lowering medication (Reference Model)

āˆ’8.23 (āˆ’ā€‰16.3, āˆ’ā€‰0.13)

Reference

Reference model + lipoprotein subclasses

āˆ’4.45 (āˆ’ā€‰12.65, 3.75)

45.9 (3.6, 238.7)

Reference model + amino acids

āˆ’4.49 (āˆ’ā€‰13.06, 4.08)

45.4 (7.2, 274.0)

Reference model + fatty acids

āˆ’6.38 (āˆ’15.00, 2.24)

22.5 (āˆ’11.2, 164.1)

Reference model + fluid balance

āˆ’6.12 (āˆ’ā€‰14.58, 2.34)

25.6 (4.4, 223.7)

Reference model + energy

āˆ’5.20 (āˆ’13.93, 3.54)

36.8 (0.3, 285.2)

Reference model + products of bacterial co-metabolism

āˆ’5.10 (āˆ’ā€‰13.82, 3.62)

38.0 (2.8, 266.2)

Reference model + O-phosphoethanolamine

āˆ’6.24 (āˆ’ā€‰14.79, 2.31)

24.2 (2.9, 211.3)

Reference model + phenylpropionate, and medium HDL and small LDL particle concentrations b

āˆ’3.91 (āˆ’ā€‰12.25, 4.44)

52.5 (15.9, 384.3)

  1. Abbreviations: CI, confidence interval
  2. a Aalen additive models with progressive degrees of adjustment; nā€‰=ā€‰830, 51 incident type 2 diabetes cases and 779 non-cases
  3. b Relevant metabolites defined as showing a PIPā€‰>ā€‰20% in BKMR-P regression
  4. c Bias-corrected and accelerated 95% confidence intervals from bootstrap based on 1000 resamplings as conducted by the boot R package