Body Composition in the General Population: Whole-body MRI-derived Reference Curves from Over 66 000 Individuals

Area of Science:

• Radiology
• Cardiovascular Medicine
• Metabolic Health

Background:

• Body composition (BC) is linked to cardiometabolic risk, but its predictive value is often confounded by body size and age.
• Traditional risk factors may not fully capture an individual's future disease susceptibility.
• Novel methods are needed to refine BC assessment for improved risk prediction.

Purpose of the Study:

• To develop age-, sex-, and height-normalized body composition (BC) metrics using MRI data from a large population.
• To evaluate the prognostic capability of these normalized BC metrics for cardiometabolic outcomes and mortality, independent of traditional risk factors.

Main Methods:

• A retrospective analysis of whole-body MRI scans from over 66,000 individuals in the UK Biobank and German National Cohort.
• Calculation of age-, sex-, and height-specific BC z-scores for various tissues (adipose, muscle) using a deep learning framework.
• Multivariable Cox regression analysis to assess the association between BC z-score categories and incident diabetes, major adverse cardiovascular events, and all-cause mortality.

Main Results:

• Age-, sex-, and height-specific BC z-scores were successfully calculated for 66,608 individuals.
• High visceral adipose tissue (VAT) z-scores were associated with a 2.26-fold increased risk of incident diabetes.
• High intramuscular adipose tissue (IMAT) z-scores correlated with a 1.54-fold increased risk of major adverse cardiovascular events, and low skeletal muscle (SM) z-scores with a 1.44-fold increased risk of all-cause mortality.

Conclusions:

• Whole-body MRI-derived, normalized BC z-scores effectively identify individuals at risk for adverse cardiometabolic outcomes and mortality.
• These advanced BC metrics offer prognostic value beyond conventional risk factors.
• An open-source z-score calculator is available to facilitate broader clinical application.