Predictive Value of Low Muscle Mass and High Fat Mass Phenotype for Prediabetes Risk in Chinese Adults: A Retrospective Cohort Study

Technical Summary: Predictive Value of Low Muscle Mass and High Fat Mass Phenotype for Prediabetes Risk in Chinese Adults

Problem Statement
Prediabetes (PreDM) represents a critical, reversible stage of metabolic dysregulation affecting approximately 35.2% of Chinese adults. While traditional anthropometric indicators like BMI and waist circumference are standard screening tools, they often fail to identify high-risk metabolic phenotypes such as "normal-weight obesity" or "sarcopenic obesity." Although the low muscle mass and high fat mass (LMM-HFM) phenotype is recognized as a marker for metabolic abnormalities, its incremental predictive value for PreDM risk beyond conventional indicators remains unclear. Furthermore, existing literature largely focuses on diabetes mellitus (DM) as an endpoint or relies on correlation analyses without integrating body composition phenotypes into predictive modeling frameworks.

Methodology
This retrospective cohort study analyzed data from 910 adults (aged ≥18) who underwent health examinations at the Affiliated Hospital of Southwest Medical University between 2021 and 2025. Participants were free of PreDM or diabetes at baseline and had at least one follow-up examination with a gap of ≥1 year.

• Exposure Definition: Body composition was assessed via Bioelectrical Impedance Analysis (BIA). Participants were categorized into four phenotypes based on Skeletal Muscle Mass Index (SMI) and Body Fat Percentage (BF%):
  • Normal
  • Isolated Low Muscle Mass (SMI < 7.0 kg/m² for males, < 5.7 kg/m² for females)
  • Isolated High Fat Mass (BF% > 25% for males, > 35% for females)
  • Combined LMM-HFM Phenotype
• Outcome: New-onset PreDM, defined by fasting plasma glucose (5.6–6.9 mmol/L) or HbA1c (5.7%–6.4%) during follow-up.
• Statistical Analysis: The study employed multivariate logistic regression (adjusted for age, sex, BMI, waist circumference, lipids, uric acid, and medical history), Restricted Cubic Spline (RCS) analysis for dose-response relationships, and SHapley Additive exPlanations (SHAP) based on XGBoost models to evaluate feature importance. Sex-stratified and sensitivity analyses were also conducted.

Key Results

• Phenotype Distribution: Among the 910 participants, 51.6% were normal, 20.7% had isolated low muscle mass, 23.4% had isolated high fat mass, and 4.3% presented with the combined LMM-HFM phenotype. Significant intergroup differences were observed in age, BMI, waist circumference, and metabolic markers (glucose, lipids, uric acid).
• Association with PreDM:
  • Isolated High Fat Mass: Remained an independent risk factor for PreDM even after full adjustment (OR=1.72, 95% CI: 1.01–2.96, P=0.047).
  • Combined LMM-HFM: The association with PreDM was primarily driven by the high fat component. After full adjustment, the combined phenotype lost statistical significance in the overall population (OR=1.18, P=0.681).
  • Sex Stratification: A significant interaction was observed. In males, the combined LMM-HFM phenotype was an independent predictor of PreDM (OR=2.52, 95% CI: 1.10–6.03, P=0.032). In females, only the isolated high fat mass phenotype showed a significant association (OR=3.26, P=0.003); the combined phenotype was not significant.
• Dose-Response: RCS analysis revealed a nonlinear inverse association between relative muscle mass index (rMPI) and abnormal glucose risk, suggesting a threshold effect where risk increases sharply below a certain muscle mass level.
• Feature Importance: SHAP analysis ranked body composition phenotypes lower in predictive importance than traditional indicators such as gamma-glutamyl transferase (GGT), waist circumference, and HbA1c. GGT emerged as the most important feature.

Key Contributions

1. Decoupling Muscle and Fat Effects: The study clarifies that the metabolic risk associated with the LMM-HFM phenotype is predominantly driven by high fat mass rather than low muscle mass in the general population.
2. Sex-Specific Risk Profiles: The research identifies that the combined LMM-HFM phenotype holds independent predictive value specifically for males, whereas high fat mass alone is the primary driver of risk in females.
3. Predictive Utility Assessment: By integrating SHAP analysis, the study demonstrates that while BIA-derived phenotyping offers insights, it provides limited incremental predictive value over routine, low-cost clinical indicators (e.g., GGT, waist circumference) for general PreDM screening.

Significance and Claims
The authors conclude that the high fat phenotype is a robust, independent risk factor for prediabetes. However, they modestly claim that the incremental predictive value of adding BIA-derived LMM-HFM phenotyping to standard screening models is limited for the general population.

The study suggests that routine PreDM screening should prioritize accessible indicators such as GGT, waist circumference, and uric acid. The authors propose that BIA-derived phenotyping may serve a supplementary, targeted role, specifically for identifying high-risk males who might benefit from individualized interventions focusing on muscle enhancement and fat reduction. The paper acknowledges limitations including its retrospective design, single-center nature, and the lack of muscle strength measurements, calling for future multicenter prospective studies to validate these findings and explore the specific thresholds of muscle mass on glucose metabolism.