Table 1. List of metabolites independently associated with the VFM diet score (P<0.01 in backward linear regression), their association with VFM and the proportion of the association of the VFM diet score with VFM that is mediated by the VFM diet score association with the metabolites (P<5.56 × 10−3).
| Metabolite name |
VFM diet score stepwisea |
VFMb |
Diet R2 no metabolitec | Diet R2 with metabolited | % Association through metabolite | ||
|---|---|---|---|---|---|---|---|
| Beta (s.e.) | P-value | Beta (s.e.) | P-value | ||||
| Hippurate | −0.45 (0.10) | 2.15 × 10−5 | -0.081(0.012) | 1.33 × 10−11 | 0.0312 | 0.0222 | 28.8% |
| Alpha-hydroxyisovalerate | 0.38 (0.10) | 9.60 × 10−5 | 0.050 (0.013) | 1.65 × 10−4 | 0.0270 | 13.5% | |
| Butyrylcarnitine | 0.33 (0.10) | 8.54 × 10−4 | 0.072 (0.013) | 5.86 × 10−8 | 0.0267 | 14.4% | |
| Bilirubin (Z,Z) | −0.31 (0.10) | 1.76 × 10−3 | −0.049 (0.013) | 1.88 × 10−4 | 0.0258 | 17.3% | |
| Indolepropionate | −0.33 (0.11) | 2.21 × 10−3 | −0.030 (0.012) | 1.40 × 10−2 | |||
| 1-Arachidonoylglycerophosphocholine* | 0.27 (0.10) | 5.20 × 10−3 | 0.031 (0.012) | 1.07 × 10−2 | |||
| EPA (20:5n3) | −0.75 (0.10) | 1.13 × 10−13 | 0.020 (0.012) | NS | |||
| Threonate | −0.32 (0.11) | 2.59 × 10−3 | −0.016 (0.012) | NS | |||
| X-11793—Oxidized bilirubin* | 0.33 (0.11) | 2.63 × 10−3 | −0.004 (0.012) | NS | |||
Abbreviations: BMI, body mass index; EPA, eicosapentaenoate; NS, not significant; VFM, visceral fat mass.
NS: P>0.05.
The asterisk (*) on the end of the metabolite name indicates the metabolite identity has not been confirmed by accurate mass data. The proportion of the variance in VFM explained by the VFM diet score after taking into account all covariates as in (a) and adjusting for the metabolite.
Thirty metabolites significantly associated with the VFM diet score (Table 2) were adjusted for covariates (batch effects, age, BMI and sex) and fitted into a backward stepwise linear regression to predict the VFM diet score using P<0.01 as the threshold cutoff.
Nine metabolites independently associated with the VFM diet score were tested for their association with VFM adjusted for covariates (age, batch effects, BMI, total fat, sex, height2 and family relatedness). Associations passing the Bonferonni cutoff were considered significant (P<5.56x10−3).
The proportion of the variance in VFM explained by the VFM diet score after taking into account all covariates (age, sex, BMI, height2 and batch effects).
The proportion of the variance in VFM explained by the VFM diet score after taking into account all covariates as in (c) plus the metabolite.