A Bayesian promotion time cure rate model with current status data

In medical studies of treatable fatal diseases, some individuals may become cured, causing survival curves to plateau instead of declining to zero. Cure rate models estimate the cure fraction, the distribution of susceptible individuals, and covariate effects. To study the data with a cure fraction, the promotion time cure rate model is favored, owing to its biological interpretability in cancer metastasis. Epidemiological and destructive testing data are often subject to current status censoring, where event status is recorded only once. While the promotion time cure rate model has been extensively studied under various censoring schemes, its Bayesian formulation for current status data remains largely unexplored. Motivated by this, we develop a Bayesian promotion time cure rate model specifically for current status data. The posterior computation is carried out via an adaptive Metropolis-Hastings algorithm. Simulation studies prove our approach’s efficiency, while analyses of lung tumor and breast cancer data illustrate its utility. By integrating prior knowledge with data, this method enhances understanding of disease dynamics and aids in developing strategies to improve cure rates.

在可治疗致命疾病的医学研究中，部分受试者可被治愈，此时生存曲线会趋于平稳而非降至零点。治愈率模型用于估计治愈比例、易感人群的分布以及协变量效应。针对存在治愈比例的数据，促进时间治愈率模型（promotion time cure rate model）因其在癌症转移方面的生物学可解释性而备受青睐。流行病学数据与破坏性试验数据常受现状截尾（current status censoring）影响，即事件状态仅被记录一次。尽管现有研究已针对多种截尾方案对促进时间治愈率模型展开了广泛探讨，但针对现状截尾数据的贝叶斯建模形式仍鲜有研究。基于此，我们针对现状截尾数据构建了专属的贝叶斯促进时间治愈率模型。我们通过自适应Metropolis-Hastings算法完成后验推断计算。仿真实验验证了该方法的有效性，而对肺肿瘤与乳腺癌数据集的分析则展示了其实际应用价值。通过将先验知识与数据相结合，该方法能够加深对疾病动态变化的理解，并助力制定提升治愈率的相关策略。