Table 2.
Regression models relating fasting insulin to 2-h metabolite change in individuals with impaired glucose tolerance (FOS-IGT)
| Predictor(s) | R2adj | P-value | Prediction errora |
|---|---|---|---|
| Δb Leucine/isoleucine | 0.36 | 9E−4 | 6.65 |
| Δ Valine | 0.17 | 3E−2 | 7.74 |
| Δ Lactate | 0.16 | 3E−2 | 7.60 |
| Δ Glycochenodeoxycholic acid | 0.14 | 4E−2 | 7.86 |
| Δ Methionine | 0.14 | 4E−2 | 7.68 |
| Δ β-Hydroxybutyrate | 0.14 | 4E−2 | 7.90 |
| Δ Leucine/isoleucine+Δ glycerolc | 0.54 | 7E−5 | 5.66 |
| PLSd | 0.46 | 1E−4 | 6.89 |
| BMI | 0.33 | 1E−3 | 6.74 |
The prediction error is expressed as the root mean square error of prediction (RMSEP), in micro-international units per milliliter insulin.
Δ denotes log of the 2-h fold change of metabolite levels.
A bivariate model consisting of the 2-h changes in leucine/isoleucine and in glycerol.
Partial least squares based on changes in the 18 validated metabolites.