# The risk prediction models for sarcopenia in older adults: a systematic review and critical appraisal

## TL;DR

This review evaluates existing models for predicting sarcopenia in older adults and finds most have high bias and limited clinical use.

## Contribution

The first systematic review and critical appraisal of sarcopenia risk prediction models, highlighting methodological flaws and design limitations.

## Key findings

- Twenty-six sarcopenia prediction models were identified, mostly with high risk of bias.
- Most models are diagnostic, not prognostic, limiting their ability to predict future sarcopenia risk.
- Common predictors include age, BMI, calf circumference, and gender.

## Abstract

Reliable sarcopenia risk prediction models are essential for identifying older adults who are currently non-sarcopenic but at risk of developing sarcopenia in the future, thereby enabling early and personalized prevention strategies. However, the prediction models for sarcopenia have not yet been systematically evaluated. This systematic review aimed to conduct a comprehensive overview and critical appraisal of current sarcopenia risk prediction models.

We conducted a systematic search across MEDLINE, Embase, Cochrane Library, and SCI-EXPANDED. Eligible primary studies on sarcopenia prediction models were identified based on the CHARMS checklist (CHecklist for critical Appraisal and data extraction for systematic Reviews of prediction Modelling Studies). The Prediction model Risk Of Bias Assessment Tool (PROBAST) was applied to evaluate risk of bias and clinical applicability.

Twenty-six sarcopenia prediction models were identified, mostly targeting community-dwelling older adults or patients. Twenty-three studies developed diagnostic prediction models, while only three studies established sarcopenia prognostic models. Age, BMI, calf circumference and gender were most frequently utilized predictors. Despite reported discriminative performance ranging from moderate to excellent (AUC > 0.70), 96.1% of prediction models exhibited high risk of bias due to significant methodological shortcomings, suggesting that model performance might be overestimated. Moreover, most existing prediction models were diagnostic study design, limiting their ability to predict the future risk of sarcopenia development.

Most existing sarcopenia prediction models demonstrated moderate to high discriminatory performance. However, due to their predominantly diagnostic study design and high risk of bias, these models cannot yet be broadly recommended for routine clinical application in the early identification of high-risk older adults with sarcopenia. Future studies are needed to develop and externally validate practical, accurate prognostic sarcopenia models to fulfill sarcopenia early prevention.

The protocol has been registered on the Open Science Framework (10.17605/OSF. IO/BFDK6).