Diagnostic value of tongue manifestations for primary Sjögren’s syndrome: construction and validation of disease screening model

TL;DR

This study shows that tongue features from traditional Chinese medicine can help screen for primary Sjögren’s syndrome with high accuracy.

Contribution

A novel non-invasive screening model for pSS based on tongue manifestations and demographic factors is developed and validated.

Findings

The model achieved an AUC of 0.89 in training and 0.93 in testing with high sensitivity and specificity.

Key predictors included bluish-purple tongue, thin tongue, and specific tongue fur characteristics.

The model was validated using bootstrap and test set analysis, showing strong discrimination and calibration.

Abstract

Tongue manifestations (tongue fur and tongue body) of traditional Chinese medicine is specific for primary Sjögren’s syndrome (pSS). We investigated the value of tongue manifestations in the diagnosis of pSS and established a model for screening or diagnosing pSS.

A total of 400 individuals attended at the China-Japan Friendship Hospital were included in this study. 200 patients with pSS and 200 healthy individuals were randomly divided into training and test sets (7:3 ratio). The training set was used to build models based on combined tongue, tongue fur, and tongue body, respectively, and the optimal model was selected through model discrimination and calibration. The optimal model is internally validated using the bootstrap method and further performed on the test set to assess transportability and generalisability. Receiver operating characteristic curve, calibration curve, Brier score, and decision curve analysis (DCA) were used to evaluate the performance of the model.

In the training set, we identified an optimal model which included sex, age, bluish-purple tongue, thin tongue, fissured tongue, thin fur, and peeling fur, after evaluating and comparing three models. The model demonstrated an area under the curve (AUC) of 0.89, and the lowest Brier score of 0.13, with well-fitted calibration curves. The optimal cut-off value for the model is 0.347, with sensitivity and specificity of 87.86 and 76.43%, respectively. The model also shows well discrimination in internal validation. In the test set, this model demonstrated an AUC of 0.93 and with a Brier score of 0.11, and sensitivity and specificity of 93.33 and 71.67%, respectively. A nomogram for pSS screening was developed based on the model.

The non-invasive, reliable and convenient model, established based on age and sex of the patient and specific tongue manifestations can serve as a clinical reference tool for screening or diagnosing of pSS.

Figures