Scholarlink

Background <p>The optimal timing of a treatment within therapy lines can maximize its benefit while minimizing risks and side effects. Because patients differ in clinical and biological characteristics, personalized treatment planning can support physicians’ decision-making. However, multistate disease progression complicates the comparison of treatment strategies, motivating the development of a user-friendly web tool for multistate disease progression simulation and treatment decision support. Microsimulation is a natural approach for comparing prespecified treatment-timing strategies, where it generates disease trajectories for a large number of virtual patients matched to an individual patient’s observed data. In our framework, microsimulation is performed using a multistate model to capture state-to-state transition hazards.</p> Results <p>We develop TxMicroSim, a web-based interface built with R Shiny, to compare treatment-timing strategies by estimating model-projected restricted mean survival times (RMSTs). The application consists of three main steps: defining the multistate structure, building the multistate model, and performing microsimulation. First, the user specifies the multistate structure, with the treatment of interest as one of the states and an absorbing state representing the time-to-event endpoint. The user also identifies transitions that lead into the treatment state, defining the strategy-driven transitions used to compare treatment-timing strategies. Next, to build clock-reset transition-specific flexible parametric proportional hazards models, the user can either upload data for model fitting or manually enter covariate hazard ratios and baseline hazard parameters. Two baseline hazard specifications are supported: a piecewise constant function and a natural cubic spline. Finally, the user enters patient-specific information, including covariates, the initial disease state at diagnosis, and the treatment start time for strategy-driven transitions. Microsimulation is used to compute RMST up to a prespecified time horizon under each strategy. Throughout the app, interactive visualizations are provided for the multistate structure, estimated hazard ratios and baseline hazards, and RMST comparisons across treatment strategies.</p> Conclusions <p>TxMicroSim is a flexible web tool that compares treatment-timing strategies and quantifies survival impact using interpretable summaries and interactive visualizations.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Toolkit for multistate disease progression simulation and treatment decision-making aid

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月22日

查看全文
  • Can Xie,
  • Ruosha Li,
  • Wael Saber,
  • Jorge E. Cortes,
  • Nicholas J. Short,
  • Xuelin Huang

摘要

Background

The optimal timing of a treatment within therapy lines can maximize its benefit while minimizing risks and side effects. Because patients differ in clinical and biological characteristics, personalized treatment planning can support physicians’ decision-making. However, multistate disease progression complicates the comparison of treatment strategies, motivating the development of a user-friendly web tool for multistate disease progression simulation and treatment decision support. Microsimulation is a natural approach for comparing prespecified treatment-timing strategies, where it generates disease trajectories for a large number of virtual patients matched to an individual patient’s observed data. In our framework, microsimulation is performed using a multistate model to capture state-to-state transition hazards.

Results

We develop TxMicroSim, a web-based interface built with R Shiny, to compare treatment-timing strategies by estimating model-projected restricted mean survival times (RMSTs). The application consists of three main steps: defining the multistate structure, building the multistate model, and performing microsimulation. First, the user specifies the multistate structure, with the treatment of interest as one of the states and an absorbing state representing the time-to-event endpoint. The user also identifies transitions that lead into the treatment state, defining the strategy-driven transitions used to compare treatment-timing strategies. Next, to build clock-reset transition-specific flexible parametric proportional hazards models, the user can either upload data for model fitting or manually enter covariate hazard ratios and baseline hazard parameters. Two baseline hazard specifications are supported: a piecewise constant function and a natural cubic spline. Finally, the user enters patient-specific information, including covariates, the initial disease state at diagnosis, and the treatment start time for strategy-driven transitions. Microsimulation is used to compute RMST up to a prespecified time horizon under each strategy. Throughout the app, interactive visualizations are provided for the multistate structure, estimated hazard ratios and baseline hazards, and RMST comparisons across treatment strategies.

Conclusions

TxMicroSim is a flexible web tool that compares treatment-timing strategies and quantifies survival impact using interpretable summaries and interactive visualizations.