Association of static balance parameters with prediabetes: a cross-sectional study and nomogram development

To investigate the association between static balance function parameters and prediabetes, and to develop a nomogram integrating center-of-pressure path length, Romberg ratios, and sex, thereby evaluating its potential as a candidate functional screening tool.

A cross-sectional observational design was adopted. A total of 97 patients with prediabetes and 44 age-matched healthy controls were consecutively enrolled from January 2024 to June 2025 at the Third People's Hospital Affiliated to Fujian University of Traditional Chinese Medicine. General characteristics and static balance parameters under four sensory conditions were compared between groups. Variables were first screened via Least Absolute Shrinkage and Selection Operator (LASSO) regression, followed by multivariable logistic regression to identify independent associations. A nomogram was then constructed. Model discrimination and calibration were assessed using the receiver operating characteristic curve, Hosmer-Lemeshow test, and 1000-bootstrap internal validation (corrected C-statistic, calibration slope/intercept, Brier score). Effect sizes were quantified using Cohen's d.

Compared with the healthy control group, the prediabetes group exhibited significantly higher values for Y-COP and COP-PL under Condition 1, RRA and RRT under Condition 2/1, as well as Y-COP, COP-EA, and COP-PL under Conditions 3 and 4 (all P < 0.05). LASSO regression retained T1 COP-PL, T3 COP-PL, T4/T3 RRA, and RRT as candidate variables. Multivariable logistic regression showed that female sex (vs. male, OR = 0.208, 95% CI: 0.044-0.973; P = 0.046), T1 COP-PL (per unit increase, OR = 0.990, 95% CI: 0.980-0.999), T3 COP-PL (per unit increase, OR = 1.016, 95% CI: 1.008-1.023), and T4/T3 RRA (per unit increase, OR = 1.013, 95% CI: 1.007-1.020) were independently associated with prediabetes. A nomogram incorporating these variables yielded an AUC of 0.972. After bootstrap internal validation, the corrected AUC was 0.964 (95% CI: 0.941-0.991), with a calibration slope of 0.806, intercept of -0.011, and Brier score of 0.075, indicating good discrimination and calibration. Cohen's dvalues for T1 COP-PL, T3 COP-PL, and T4/T3 RRA were -0.40, 1.45, and 1.39, respectively, reflecting moderate to large effect sizes.

Static balance indices including T1 COP-PL, T3 COP-PL, and T4/T3 RRA are associated with prediabetes. The nomogram combining these measures with sex demonstrated robust performance in internal validation, offering a functional perspective that may aid in the auxiliary screening of prediabetes. Nevertheless, as a cross-sectional exploratory study, these associations do not imply causality, and the model requires external validation before its clinical generalizability can be determined.