Characterizing Factors that Impact the Evolution of Neurocysticercosis Cysts: A Cyst-Level Analysis Using New Statistical Methods for Complex Longitudinal Data

点击翻译按钮获取中文摘要

PROJECT SUMMARY Despite accumulating evidence that neurocysticercosis (NC) cyst evolution varies within the brain, the life course of the encysted parasite remains relatively understudied. We understand that cyst evolution differs between parenchymal and extraparenchymal brain locations and possibly by patient age and sex, but little is known about the timing of the transitions through the stages of evolution overall or by cyst and patient characteristics, nor how cyst and patient characteristics impact the effect of anthelminthic treatment such as albendazole (ALB). Our understanding of cyst evolution is hindered by the fact that most studies examine only patient-level aggregate measures of NC cyst burden within the brain. However, individual NC cysts in the same patient can evolve differently, and ALB may kill some parasites but have little effect on others within the same patient; only by following individual cysts can we understand these differences. Therefore, in this project, we aim to disaggregate patient-level data to the cyst-level and use multistate modeling to examine transitions of individual cysts through stages of evolution from the disease progress perspective. Although multistate models can handle multivariate longitudinal data, our NC cysts data pose additional challenges: (1) the data are interval-censored due to pre- specified data collection schedules, (2) the data can be informatively right-censored due to loss to follow-up, which results in missing data that may not be random, (3) the data is left-censored because patients enter a study with pre-existing cysts and the time of infection is unknown, and (4) because multiple cysts can be within the same patient, and even within the same brain location, we have multilevel correlated data. In this study, we propose a selection-model embedded time-homogeneous Markov multistate joint model with nested frailty for the NC cyst data. Inference wise, we will consider the maximum likelihood-based approach. Using these new methods, we propose to conduct cyst-level analysis to identify the cyst and patient characteristics that impact NC cyst evolution and ALB treatment effectiveness using data from a 2001-05 randomized controlled trial conducted in Ecuador that compared ALB treatment to placebo among 178 patients over 24 months follow-up with brain imaging (CT/MRI) conducted at baseline, months 1, 6, 12 and 24. This methodological approach will advance neurocysticercosis research by providing more detailed information on the cyst evolutionary course and factors that modify the impact of ALB on this course. Results from these analyses will increase our understanding of the factors that modify the effectiveness of ALB, first step towards the development of more precise patient treatment options and thereby improved patient outcomes. The proposed statistical methods may also have applications to modeling other diseases that evolve through predefined clinical states and impact multiple body regions when assessed in longitudinal studies with intermittent data collection and various forms of censoring.