Imputation of Truncated Tumor Growth Data Via Bayesian Mixed Modeling
M Bleile1, 2*, Yixun Xing 2,3, Dan Nguyen 2,5, Casey Moore 4, Debabrata Saha 5, Benjamin Chen 5, Michael Story 5, Robert Timmerman5, Steve Jiang 2,5 Daniel F. Heitan 2,6,(1) Department of Statistical Science, Southern Methodist University, Dallas, TX (2) Medical Artificial Intelligence and Automation (MAIA) lab, University of Texas Southwestern Medical Center, Dallas, TX (3) Department of Advanced Data Analytics, University of North Texas, Denton, TXs(4) Department of Pathology, University of Texas Southwestern Medical Center, Dallas, TX (5) Department of Radiation Oncology, University of Texas Southwestern Medical Center, Dallas, TX, (6) Department of Population & Data Sciences, University of Texas Southwestern Medical Center, Dallas, TX
Presentations
PO-GePV-M-69 (Sunday, 7/10/2022) [Eastern Time (GMT-4)]
ePoster Forums
Purpose: In a longitudinal study, dropout occurs when an experimental unit becomes unavailable for further evaluation prior to the planned end of follow-up. A common instance of dropout is the removal from a tumor xenograft experiment of mice who die either naturally or by sacrifice for morbidity. Dropout is problematic because it can lead to bias and inefficiency in analysis. One solution is to impute, i.e. fill in, values of the lost observations with values predicted under a statistical model. We describe and compare various methods for imputing the lost tumor volume values, and present an R package that implements the methods and will be made publicly available. We apply a multiple imputation scheme to the analysis of a tumor xenograft experiment that studied the optimal timing of radio-immunotherapy.
Methods: Based on data of measured tumor volumes of mice at multiple prescribed times, we modeled the series of log tumor volumes for each mouse as a set of observations from a multivariate Gaussian distribution. We estimated this model using a Bayesian approach, and sampled values from the posterior distribution of the mean and variance. Using these, for each mouse with dropout we impute the counterfactual tumor volumes, conditional on its observed sequence. We tested the performance of the method via simulation, and created an implementation available in an R package.
Results: Plots of the imputed values reveal consistency with the trend of the observed data.Simulations reveal that model-based imputation uses the data efficiently without inducing bias.
Conclusion: We have adapted established statistical methods for use in analysis of tumor xenograft experiments with dropout, and demonstrated the effectiveness of these methods via simulation. We have additionally created a computational tool which allows users to impute counterfactual values in such experiments.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email