This study investigates the association between LFP, VAT and T2DM using DXA in 542 adults. We apply penalized regression, generalized additive models (GAM), and causal mediation analysis to explore whether higher LFP reduces diabetes risk and whether this effect is mediated by VAT. Sex-specific analyses further clarify the biological differences in adipose distribution. The findings aim to improve understanding of fat distribution and metabolic health.

T2DM is a major public health concern, and growing evidence suggests that not all body fat confers equal metabolic risk. Beyond total adiposity, the regional distribution of fat strongly influences insulin resistance and diabetes development. VAT is metabolically detrimental, promoting inflammation and lipotoxicity, whereas lower-body subcutaneous fat, particularly in the legs and gluteofemoral region, may exert protective metabolic effects by serving as a "safe storage" depot for excess lipids. However, the causal mechanisms underlying these associations remain incompletely understood, and whether the protective role of LFP differs by sex remains unclear.

This study aims to elucidate the metabolic and causal pathways linking LFP, VAT, and diabetes risk in adults. Using DXA for precise body composition assessment, the study evaluates the associations between LFP, VAT, and T2DM prevalence, with a focus on sex-specific effects. A total of 542 adult participants will be analyzed, including both men and women with and without T2DM. Clinical, biochemical, and anthropometric data will be collected concurrently.

The analytic framework integrates multiple complementary approaches:

1. Penalized regression models (LASSO, ridge, elastic net) will identify the most predictive adiposity components for T2DM while minimizing collinearity.
2. GAMs will examine potential nonlinear relationships between LFP, VAT, and diabetes risk, separately for male and female.
3. Causal mediation analysis will quantify the extent to which VAT mediates the relationship between LFP and diabetes, providing mechanistic insight into adipose redistribution pathways.
4. Inverse-probability-of-treatment weighting (IPTW) will strengthen causal inference by balancing covariates across LFP strata.
5. Two-sample Mendelian randomization (MR) will further test the causal effect of genetically predicted LFP on T2DM risk using summary-level genome-wide association data.

We hypothesize that higher LFP will be associated with lower diabetes risk and that this protective effect will be partially mediated by reduced VAT accumulation, particularly in females.

This study integrates imaging-based body composition, causal modeling, and genetic validation to bridge observational and causal evidence. The findings are expected to improve the understanding of sex-specific fat distribution in metabolic health and support the development of personalized prevention strategies targeting regional adiposity rather than total body fat.

Inclusion Criteria:

Adults (≥18 years old)

Available DXA measurement of fat distribution

Complete fasting glucose or diabetes diagnosis record

Exclusion Criteria:

• Missing key variables (fat distribution, diabetes status)

Severe systemic disease interfering with data integrity

Implausible or inconsistent records

Adults who underwent DXA examination and had complete clinical and biochemical data.