Associations of body shape index (ABSI) and hip index with liver, metabolic, and inflammatory biomarkers in the UK Biobank cohort

Associations of liver, metabolic, and inflammatory biomarkers in blood with body shape are unclear, because waist circumference (WC) and hip circumference (HC) are dependent on overall body size, resulting in bias. We have used the allometric “a body shape index” (ABSI = WC(mm)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\,*\,$$\end{document}∗Weight(kg)-2/3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\,*\,$$\end{document}∗Height(m)5/6) and hip index (HIwomen = HC(cm)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\,*\,$$\end{document}∗Weight(kg)-0.482\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\,*\,$$\end{document}∗Height(cm)0.310, HImen = HC(cm)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\,*\,$$\end{document}∗Weight(kg)-2/5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\,*\,$$\end{document}∗Height(cm)1/5), which are independent of body mass index (BMI) by design, in multivariable linear regression models for 121,879 UK Biobank men and 135,559 women. Glucose, glycated haemoglobin (HbA1c), triglycerides, low-density-lipoprotein cholesterol, apolipoprotein-B, alanine aminotransferase (ALT), gamma-glutamyltransferase, and lymphocytes were associated positively with BMI and ABSI but inversely with HI. High-density-lipoprotein cholesterol and apolipoprotein-A1 were associated inversely with BMI and ABSI but positively with HI. Lipid-related biomarkers and ALT were associated only with HI in obese men. C-reactive protein, neutrophils, monocytes, and alkaline phosphatase were associated positively, while bilirubin was associated inversely, with BMI and ABSI but not with HI. Associations were consistent within the clinical reference ranges but were lost or changed direction for low or high biomarker levels. Our study confirms associations with waist and hip size, independent of BMI, for metabolic biomarkers but only with waist size for inflammatory biomarkers, suggesting different contribution of the mechanistic pathways related to body shape.


S1
Associations of body shape index (ABSI) and hip index with liver, metabolic, and inflammatory biomarkers in the UK Biobank cohort Sofia Christakoudi, Elio Riboli, Evangelos Evangelou, Konstantinos K. Tsilidis

Supplementary Tables
Supplementary Table S1 Categories  S5 self-reported NSAID use from Fields [20003-0/47] "Treatment/ medication code2" (see list of medication codes in Supplementary Table S1) (further n=1689). Participants with missing values were assigned the median sex-specific category (No).
Oophorectomy (bilateral) was used to define menopausal status and was based on Field: Answer 1: "Yes"; Pre-menopausal were classified women who had not been defined as postmenopausal above AND had reported pre-menopausal status with Answer 0: "No" to Field [2724-0.0] AND had not reported hysterectomy; Undetermined or missing included the remaining women.

Supplementary
Reportability codes: 1: "Reportable at assay and after aliquot correction, if attempted" -values used as provided.
2: "Reportable at assay but not reportable after any corrections (too low)" -see code 4.
3: "Reportable at assay but not reportable after any corrections (too high)" -see code 5.
4: "Not reportable at assay (too low)" -values were replaced with half the lowest detected level for all except direct bilirubin, which was imputed with quantile regression imputation of left-censored data (QRILC) (imputeLCMD v2.0 package in R), following log-transformation.
5: "Not reportable at assay (too high)" -replaced with the highest detected value.
Biomarker values for participants without attempted measurements were considered missing and these participants were excluded from the analysis of the corresponding biomarker.

Liver function tests
Bilirubin total LIN 115,878 -0.06 (-0.07 to -0.06)** -0.06 (-0.06 to -0.05)** 0.02 (0.01 to 0.02)** 128,819 -0.11 (-0.11 to -0.10)** -0.06 (-0 LIN -estimates for SD difference (95% confidence interval) were obtained from multivariable linear regression models with each biomarker on a continuous scale (sex-specific z-scores, following log-transformation) as an outcome variable and BMI+ABSI+HI on a continuous scale (sex-specific z-scores) as independent variables ( Figure 1); SB1+SB2 -estimates for SD difference (95% confidence interval) were obtained from multivariable linear regression models as in LIN, but replacing one of BMI, ABSI, or HI with restricted cubic splines (knots at -2, 0, and 2) and retaining the other two anthropometric indices on a linear scale. All models were additionally adjusted for height, age at enrolment, weight change within the last year preceding enrolment, smoking status, alcohol consumption, physical activity, Townsend deprivation index, region of the assessment centre, time of blood collection, fasting time, use of nonsteroidal anti-inflammatory drugs, paracetamol use, and in women also menopausal status, oral contraceptives use, hormone replacement therapy use, and age at the last live birth. Covariates are defined in Supplementary Methods. * -p<0.0001 from Wald test for the individual term; ** -p<1* 10 -6 p non-linearity -p-value from a likelihood ratio test comparing a fully adjusted linear model (LIN) nested within the corresponding fully adjusted model including restricted cubic splines for one of the anthropometric indices (SB1+SB2) (p non-linearity <0.0001 are shown in bold).
All models were adjusted for height, age at enrolment, weight change within the last year preceding enrolment, smoking status, alcohol consumption, physical activity, Townsend deprivation index, region of the assessment centre, time of blood collection, fasting time, use of nonsteroidal antiinflammatory drugs, paracetamol use, and in women also menopausal status, oral contraceptives use, hormone replacement therapy use, and age at the last live birth. Covariates are defined in Supplementary Methods. * -p<0.0001 from Wald test for the individual term; ** -p<1*10 -6 (biomarkers with opposite direction of the association in the two groups and p<0.0002 for the small tail-end group are shown in bold).

S23
Supplementary Figure  SD difference (95% CI) -unadjusted estimates for standard deviation difference (95% confidence interval) were obtained from multivariable linear regression models including each biomarker on a continuous scale (sex-specific z-scores, following log-transformation) as an outcome variable and BMI, ABSI, and HI on a continuous scale (sex-specific z-scores) as independent variables without any covariates. Adjusted models are included for comparison (estimates and covariates correspond to Figure 1). Adjustment variables included height, age at enrolment, weight change within the last year preceding enrolment, smoking status, alcohol consumption, physical activity, Townsend deprivation index, region of the assessment centre, time of blood collection, fasting time, nonsteroidal anti-inflammatory drug use, paracetamol use, and in women also menopausal status, oral contraceptives use, hormone replacement therapy use, and age at the last live birth.

S32
ABSI -a body shape index; ALP -alkaline phosphatase; ALT -alanine aminotransferase; ApoA1 SD difference (95% CI) -estimates for standard deviation difference (95% confidence interval) were obtained from multivariable linear regression models including each biomarker on a continuous scale (sex-specific z-scores, following log-transformation) as an outcome variable and BMI, ABSI, and HI on a continuous scale (sex-specific z-scores) as independent variables. All models were adjusted for height, age at enrolment, weight change within the last year preceding enrolment (except for the subgroup with no weight change), smoking status, alcohol consumption, physical activity, Townsend deprivation index, region of the assessment centre, time of blood collection, fasting time, NSAID use (for the subgroups with no weight change and paracetamol use), paracetamol use (for the subgroups with no weight change and NSAID use), and in women also menopausal status, oral contraceptives use, hormone replacement therapy use, and age at the last live birth. Covariates are defined in Supplementary Methods. Participant counts per subgroup are shown in Supplementary Figure S1 and by subgroup and body shape phenotype in Table 1.

S38
Supplementary Figure  SD difference (95% CI) -estimates for standard deviation difference (95% confidence interval) were obtained from multivariable linear regression models including each biomarker on a continuous scale (sex-specific z-scores, following log-transformation) as an outcome variable and BMI, ABSI, and HI on a continuous scale (sex-specific z-scores) as independent variables. All models were adjusted for height, age at enrolment, weight change within the last year preceding enrolment, smoking status, physical activity, Townsend deprivation index, region of the assessment centre, time of blood collection, fasting time, NSAID use, paracetamol use, and in women also menopausal status, oral contraceptives use, hormone replacement therapy use, and age at the last live birth. Covariates are defined in Supplementary Methods.