Unraveling the Links: How a Healthy Lifestyle Index Influences MRI Body Composition in the UK Biobank Study

Study population

The UK Biobank represents a substantial prospective cohort investigation aimed at exploring the factors influencing significant diseases associated with middle and old age²¹. A comprehensive study protocol can be accessed online (https://www.ukbiobank.ac.uk/media/gnkeyh2q/study-rationale.pdf). In summary, from 2006 to 2010, individuals enrolled with the National Health Service (NHS) of the United Kingdom and aged between 40 and 69 years were invited to join the study, with recruitment occurring at 22 assessment centers throughout the UK. Ultimately, roughly 500,000 individuals from the general population (5.5% response rate) were enrolled and took part in the initial visit. Informed written consent was secured from participants upon enrollment. The UK Biobank received approval from the North West Multi-Centre Research Ethics Committee, the National Information Governance Board for Health and Social Care in England and Wales, and the Community Health Index Advisory Group in Scotland. The study was executed adhering to the Declaration of Helsinki, and an independent Ethics and Governance Council continuously oversees the UK Biobank’s compliance with the Ethics and Governance framework.

UK Biobank MRI study

Initiated in 2014, the UK Biobank imaging substudy extension aimed to collect body scans from 100,000 participants across four imaging centers located in Newcastle, Manchester, Reading, and Bristol²². Information regarding the imaging acquisition and analysis techniques has been previously published^11,23. Scanning of participants was performed using a Siemens MEGNETOM Aera 1.5-T MRI scanner (Siemens Healthineers, Erlangen, Germany) with the dual‐echo Dixon Vibe protocol, while body composition assessment utilized AMRA Profiler™ (AMRA AB, Linköping, Sweden)^13,24. The body composition metrics incorporated in this study included volumetric (in liters, L) assessments of abdominal visceral adipose tissue (VAT), abdominal subcutaneous adipose tissue (ASAT), total trunk adipose tissue (TTAT), which combines VAT and ASAT, total adipose tissue (measured from the bottom of the thigh muscles to the top of the vertebrae T9) (TAT), total lean tissue (measured from the bottom of the thigh muscles to the top of the vertebrae T9) (TLT), and total thigh fat-free muscle (TTFM); in addition, muscle fat infiltration (%) (MFI), abdominal fat ratio (calculated as total abdominal fat volume/total abdominal fat volume plus thigh muscle volume) (%) (AFR), and weight to muscle ratio (defined as body weight/total thigh fat-free muscle volume) (WMR). The MRI data utilized in the current analysis were gathered between April 2014 and March 2020.

Healthy lifestyle index

During the initial visit, socio-demographic, lifestyle, and health-related attributes of the study participants were gathered through questionnaires and/or direct measurements (http://www.ukbiobank.ac.uk/resources/). The lifestyle factors included in the HLI were selected based on the recommendations of the American Heart Association and encompassed sleep duration, alcohol consumption, smoking status, diet quality data, physical activity, and body mass index (BMI)²⁵. Self-reported sleep duration was based on participants’ responses to the question: “About how many hours of sleep do you get in every 24 h (h)? (Please include naps)”. Individuals reporting < 3 h or > 12 h were asked for confirmation of their responses. Sleep duration was grouped into 3 categories (< 7, 7- < 9, and ≥ 9 h). Validated questionnaires were employed to gather data on the average frequency of consumption (per week) for various food groups (meat, fruit, vegetable, and grain) over the preceding year. For bread types and cereals, portion sizes were noted. Reported alcohol consumption frequency (drinks/day) was classified as: never or only on special occasions, ≥ once/month—≤ four times/week, and daily or almost daily, while smoking history was categorized as current, former, and never. Total physical activity over the last 4 weeks was computed by multiplying the minutes/week (min/wk) of each physical activity type by the metabolic equivalent (MET) of that activity and summing these outcomes; this resulted in a categorized variable: < 600, 600- < 3000, and ≥ 3000 MET min/wk²⁶. Anthropometric measurements were carried out by trained personnel following standardized procedures²⁷. BMI was calculated by dividing weight (kg) by height squared (m²), and categorized as ≥ 30.0, 25.0– < 30.0, and 18.0- < 25.0 kg/m². Waist circumference (cm) was measured at the narrowest part of the torso and categorized into 3 groups (≥ 102, 94- < 102, and < 94 cm for men, and ≥ 88, 80- < 88, and < 80 cm for women). Each individual lifestyle component was scored independently, with the healthiest categories receiving the highest scores, and these were summed to create a combined HLI score (Supplementary table 1).

Statistical analysis

As of July 2023, MRI-derived body composition measurements were available for 40,178 individuals (51.8% women), with the exception of TAT and TLT, for which data were accessible for only 8,522 individuals. Of this cohort, 33,002 (50.1% women) had data on all HLI components, forming the analytical group for this research (6,939 of these individuals possessed data on TAT and TLT) (Supplementary Fig. 1). We compared the initial characteristics of the analytical group with those of the remaining UK Biobank cohort using the ANOVA test to assess means and standard deviations for continuous variables, and the Chi-squared test for percentages of categorical variables. The baseline characteristics of the analytical group were summarized by HLI quartiles and analyzed with similar statistical methods. Additionally, we explored correlations between body composition measurements by calculating Pearson’s correlation coefficients.

Body composition measurement ranges differed by sex; hence, all analyses were carried out separately for men and women. Multivariable linear regression was utilized to compare mean or standard deviation levels of body composition measurements across HLI quartiles, treating the lowest quartile as the referent category. Trend tests in associations across quartiles were performed by assigning each quartile median HLI values and modeling them as continuous variables. Selection of variables included in the final model as potential confounders was based on their correlation with both exposure and outcome, which consisted of age at baseline, interval between recruitment and MRI examination, race/ethnicity (White, Black, Asian, mixed, and others), education (< / ≥ college), employment status (employed, retired, caregiver, disabled, unemployed, volunteer, student, and not reported), height (cm), socio-economic status (Townsend deprivation coefficient), and (for women) menopausal status. Women were designated as post-menopausal if they had undergone natural menopause, were at least 53 years old, did not disclose their menopausal status, or had received a bilateral oophorectomy prior to this age²⁸. Among women, distinct analyses were executed based on menopausal status. Since BMI is significantly associated with body composition measurements, we examined the relationship of HLI, calculated without BMI, with various outcomes, while BMI was included separately in the model. Additionally, we assessed whether the HLI-body composition relationship was modified by BMI levels. For this purpose, a different HLI index was created using waist circumference categories instead of BMI (HLI_wst), and analyses were stratified by BMI categories. Finally, we explored the association of individual HLI components with body composition parameters. For this, we utilized the categories of each lifestyle factor as presented in Supplementary table 1, except for diet (9-point score) for which we formed 4 categories (< 0.75, 0.75–1.25. > 1.25–1.50, and > 1.50).

All statistical evaluations were conducted using Stata version 18 (StataCorp LLC, College Station, TX). All p values were bilateral and deemed statistically significant for p values < 0.05.