The Semi-parametric Stochastic Model for Assessing Three-state Disease Progression

• Main Authors: Chou kai-pei, 周楷沛
• Other Authors: Shu-Hui Chang
• Format: Others
• Language: en_US
• Published: 2001
• Online Access: http://ndltd.ncl.edu.tw/handle/39585233301341660605

碩士 === 國立臺灣大學 === 流行病學研究所 === 89 === Backgorund: The application of semiparametric method to multi- state stochastic process, particularly interval-censored data with hidden transition, has not been fully addressed. Objectives: The aim of this study was to develop a semiparametric stochastic model for assessing the effects of covariate on three-state disease progression. Methods: The three-state stochastic process plus the proportional hazard model and partial likelihood method was applied to assess the effect of covariates on different state transitions. Two thorny issues including the relaxation of Markov assumption and the correlation between different state transitions were also tackled by the extension of partial likelihood and the application of the random-effect model. The semi-Markov method was also proposed to model the covariate effect by the embedded Markov chain and holding time distribution taking competing risk problems into account. The permuted method and interval-censored method were adopted to tackle interval-censored data with hidden state transitions in three-state stochastic process. Illustrations: Two illustrations were given in this study, including adenoma-invasive carcinoma-death for data with the exact transition time known and normal-the PCDP-clinical progression of breast cancer phase for interval-censored with hidden state transitions. Conclusion: The present study proposed the semiparametric stochastic model for assessing the effects of covariates on three-state disease progression taking interval-censored data type and other issues into account.