视觉预测检查（VPC）通常用于评估药理学模型。然而，如果结果较差的患者提前退出，他们的表现可能会受到阻碍，这在临床试验中经常发生，尤其是在肿瘤学领域。虽然文献中已经出现了解释辍学的方法，但它们在假设、灵活性和性能方面各不相同，而且它们之间的差异尚未得到广泛理解。本手稿旨在阐明哪些方法可用于处理具有 dropout 的 VPC 以及何时处理，以及使用置信区间的信息更丰富的 VPC 方法。此外，我们建议根据观测数据而不是模拟数据构建置信区间。在 VPC 中纳入 dropout 的理论框架得到开发并应用，提出两种方法：完全方法和条件方法。完整方法是使用参数事件时间模型来实现的，而条件方法是使用参数模型和 Cox 比例风险 (CPH) 模型来实现的。这些方法的实际性能通过对纳武单抗和多西他赛两项癌症临床试验数据的肿瘤生长动力学（TGD）建模的应用进行了说明，其中对患者进行随访直至疾病进展。该数据集由 855 名受试者的 3504 个肿瘤大小测量值组成，由 TGD 模型描述。受试者的退出通过 Weibull 或 CPH 模型来描述。模拟数据集还用于进一步说明 VPC 方法的属性。结果表明，与不调整 drop 的简单方法相比，更熟悉的完整方法可能无法为 TGD 模型评估提供有意义的改进，并且可能优于使用 Weibull 模型或 Cox 比例风险模型的条件方法。总体而言，包括 VPC 中的置信区间应该可以改善解释，条件方法被证明在发生丢失时更普遍适用，非参数方法可以提供额外的稳健性。© 2024。作者，获得 Springer Science Business 的独家许可Media, LLC，隶属于施普林格自然。

Visual predictive checks (VPC) are commonly used to evaluate pharmacometrics models. However their performance may be hampered if patients with worse outcomes drop out earlier, as often occurs in clinical trials, especially in oncology. While methods accounting for dropouts have appeared in literature, they vary in assumptions, flexibility, and performance, and the differences between them are not widely understood. This manuscript aims to elucidate which methods can be used to handle VPC with dropout and when, along with a more informative VPC approach using confidence intervals. Additionally, we propose constructing the confidence interval based on the observed data instead of the simulated data. The theoretical framework for incorporating dropout in VPCs is developed and applied to propose two approaches: full and conditional. The full approach is implemented using a parametric time-to-event model, while the conditional approach is implemented using both parametric and Cox proportional-hazard (CPH) models. The practical performances of these approaches are illustrated with an application to the tumor growth dynamics (TGD) modeling of data from two cancer clinical trials of nivolumab and docetaxel, where patients were followed until disease progression. The dataset consisted of 3504 tumor size measurements from 855 subjects, which were described by a TGD model. The dropout of subjects was described by a Weibull or CPH model. Simulated datasets were also used to further illustrate the properties of the VPC methods. The results showed that the more familiar full approach might not provide meaningful improvement for TGD model evaluation over the naive approach of not adjusting for dropout, and could be outperformed by the conditional approach using either the Weibull model or the Cox proportional hazard model. Overall, including confidence intervals in VPC should improve interpretation, the conditional approach was shown to be more generally applicable when dropout occurs, and the nonparametric approach could provide additional robustness.© 2024. The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature.