A Machine Learning-Driven Multidimensional Evaluation of Tandospirone Formulations Based on Real-World Evidence

Author(s)

Linke Zou, Master1, Xingwei Wu, PhD2, Ming Hu, PhD1.
1West China School of Pharmacy, Sichuan University, Chengdu, China, 2Sichuan Provincial People’s Hospital, Chengdu, China.
OBJECTIVES: To compare the effectiveness and safety of different formulations of tandospirone, enabling precision therapeutic decision-making in patient subgroups.
METHODS: The data was sourced from anxiety patients treated with tandospirone in a tertiary medical institution in Sichuan Province from December 2018 to July 2023. Regular expressions were employed to extract unstructured text. Machine learning algorithms were implemented to estimate predictive models for mild, moderate, and severe anxiety. AUC and Accuracy were used to select the optimal model. The missing cases was input into this model, which the predicted value corresponding to the highest probability as the prediction result for the missing anxiety scale. Efficacy was calculated as the proportion of cases with improved scale scores after 60±20 days of treatment (improved cases/assessable population), while safety was evaluated primarily through adverse drug reaction (ADR) incidence rates.According to subgroups, efficacy and safety were ranked by calculated and ranked per subgroup.
RESULTS: Data from 12,265 inpatient and 144,483 outpatient cases was involved in this study, 95% data processing accuracy was achieved with an anxiety prediction model AUC >0.9. Tandospirone capsules and tablets showed comparable efficacy rates at 60±20 days (93.44% vs 92.87%). In efficacy evaluations, capsules outperformed tablets in 70.94% of subgroups (vs 7.30% for tablets). Safety analyses revealed no significant difference in ADR rates between formulations (1.53% vs 1.63%, p=1), demonstrating superiority in 82.69% of subgroups (363/439) the incidence of ADR with capsules compared to tablets (3.87%, 17/439).
CONCLUSIONS: This study revealed significant variations in the efficacy and safety profiles of tandospirone capsules versus tablets across patient subgroups with distinct clinical characteristics, while demonstrating machine learning's utility in data extraction and missing value imputation. This finding underscore the imperative of personalized treatment strategies in clinical practice, and established a methodological foundation for health outcome evaluation.

Conference/Value in Health Info

2025-09, ISPOR Real-World Evidence Summit 2025, Tokyo, Japan

Value in Health Regional, Volume 49S (September 2025)

Code

RWD269

Topic Subcategory

Health & Insurance Records Systems

Disease

SDC: Mental Health (including addition)

Your browser is out-of-date

ISPOR recommends that you update your browser for more security, speed and the best experience on ispor.org. Update my browser now

×