To synthesize research on sarcopenia prediction models for older adults in China and identify key limitations, such as variable selection and methodological rigor, constraining their clinical applicability.
Key Findings:
20 articles encompassing 34 prediction models were identified.
Sarcopenia prevalence across studies ranged from 12% to 54.17%.
Logistic regression and machine learning were the predominant modeling techniques.
Predictor variables per model ranged from 3 to 8, with age, BMI, and sex being the most frequently included.
Models demonstrated acceptable discriminative ability, with AUC values ranging from 0.706 to 0.974. Sensitivity ranged from 0.405 to 0.963, whereas specificity ranged from 0.400 to 0.947.
Interpretation:
Despite the growth of sarcopenia prediction models, deficiencies in variable selection, methodological rigor, and external validation limit their clinical applicability, impacting early identification and intervention.
Limitations:
Lack of methodological quality and clinical applicability of existing models.
Insufficient guidance for clinicians on model selection for specific populations and clinical settings.
Conclusion:
Addressing the identified issues is essential for developing predictive tools that are statistically robust and clinically applicable to China’s aging population.
Analysis of England’s multicancer early detection screening trial found modest, temporary diagnostic delays in participating regions, adding to concerns about health system effects and the evidence base for population-level blood-based cancer screening.
