# A gut microbiota-based predictive model for the treatment efficacy of Parkinson’s disease

## TL;DR

This study creates a model to predict when Parkinson’s disease treatments become less effective, using patient data and gut microbiota markers.

## Contribution

A predictive model for Parkinson’s treatment efficacy decline using gut microbiota and clinical markers is developed and validated.

## Key findings

- The random forest model achieved the highest predictive performance (AUC = 0.874).
- Fecal calprotectin, fecal lactoferrin, and E. coli/Lactobacillus ratio were key predictors of treatment decline.
- The model integrates clinical and biological markers for personalized therapeutic strategies.

## Abstract

This study aimed to develop and validate a predictive model for the decline in treatment efficacy among Parkinson’s disease (PD) patients based on clinical characteristics and biological markers, providing a basis for early risk identification and personalized therapeutic strategies.

A retrospective study was conducted on 500 PD patients admitted to our hospital between January 2021 and December 2024. The patients were randomly divided into a training set (n = 350) and a validation set (n = 150) at a 7:3 ratio. Demographic characteristics, clinical rating scales, and biological markers were collected for all patients. In the training set, univariate analysis was performed to screen variables associated with treatment efficacy decline. After variable selection using LASSO regression, multivariate logistic regression analysis was performed to identify independent predictors. Predictive models, including random forest (RF), support vector machine (SVM), and gradient boosting, were constructed using Python 3.8.5 and the scikit-learn library. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), and the optimal model was selected based on key predictor importance.

No significant differences in baseline characteristics were observed between the training and validation sets (all p > 0.05). A multivariate logistic regression analysis identified the total MDS-UPDRS score, MDS-UPDRS II (activities of daily living), MDS-UPDRS IV (motor complications), PDQ-39 score, E. coli/Lactobacillus ratio, fecal lactoferrin, and fecal calprotectin as independent risk factors (all p < 0.05), while total fecal bacterial count was an independent protective factor (all p < 0.05). The RF model demonstrated superior predictive performance (AUC = 0.874, 95%CI: 0.831–0.917) compared to SVM (AUC = 0.806, 95%CI: 0.753–0.859) and gradient boosting (AUC = 0.842, 95%CI: 0.794–0.889).

The RF model incorporating clinical and biological markers effectively predicts decline in treatment efficacy among PD patients, with fecal calprotectin, fecal lactoferrin, and the E. coli/Lactobacillus ratio serving as key predictors.