探討半競爭風險資料下之不同特徵點比例風險模式動態預測表現

Comparison of dynamic prediction of different landmark proportional hazard model under semi-competing risks data

葉憲周(Hsien-Chou Yeh)

指導教授 ： 張淑惠

國立臺灣大學/公共衛生學院/流行病學與預防醫學研究所/碩士(2019年)

https://doi.org/10.6342/NTU201903431

摘要

在實際臨床應用上，準確地預測病患預後是一非常重要的議題。例如癌症病患的復發情況等生物標誌(biological marker)，為隨時間變動的資訊，為臨床介入治療以及預測未來風險有用的指標。本文在不同特徵時間點(landmark time)下建構一系列以Cox為基礎型式的特徵點模式(landmark model)，並加入隨時間變動標誌資訊進行未來存活機率之動態預測(dynamic prediction)。本文特別之處為，在不同特徵時間點下，建構以Cox為基礎型式之特徵點模式時，針對病患進入研究後至發生標誌事件之標誌時間(marker time)轉化為時間分段函數，更運用此隨時間改變之資訊納入模式，此想法對於在進行未來存活機率預測時，針對時間依賴性資訊處理會更為貼切。在模擬部分，考慮三種標誌時間及存活時間之間不同的相關性結構，並以文中所建構Cox為基礎型式的特徵點模式進行動態預測，比較其表現。最後文中也以大腸直腸癌及阿茲海默症實際資料為例，探討本文所考慮之特徵點模式其動態預測的表現。

關鍵字

動態預測 ； 特徵點模式 ； 多階段模式

並列摘要

An important issue in clinical practice is to accurately predict the prognosis of patients in order to aid clinical decision-making. Biological markers, for example, recurrences in cancer patients, often serve as time-dependent information in the need of clinical intervention and the usefulness of the prediction of future risk. We consider several landmark Cox-type models at a sequence of landmark times to incorporate the time-dependent marker information for dynamic prediction of future survival probabilities. In particular, a piecewise function of the time to a marker included in the Cox-type model at each landmark time may be more adaptable to use the time-dependent marker information for predicting future survival probabilities. In simulation study, we consider three different correlation structures between the marker and survival times to assess the performance of dynamic prediction based on different landmark Cox-type models. Finally, we use colon cancer and dementia data to explore the dynamic prediction abilities under these landmark Cox-type models.

並列關鍵字

Dynamic prediction ； landmark model ； multi-state model

參考文獻

Cortese, G., Gerds, T. A., and Anderson, P. K. (2013). Comparing predictions among competing risks models with time-dependent covariates. Statistics in Medicine, 32, 3089-3101.

Dancourt, V., Quantin, C., Abrahamowicz, M., Binquet, C., Alioum, A. and Faivre, J. (2004). Modeling recurrence in colorectal cancer. Journal of Clinical Epidemiology, 57, 243-251.

Dartigues, D-F., Gagnon, M., Barberger-Gateau, P., Letenneur, L., Commenges, D., Sauvel, C., Michel, P. and Salamon, R. (1992). The paquid epidemiological program on brain ageing. Neuroepidemiology, 14-18.

Etzioni, R., Pepe, M., Longton, G., Hu, C. and Goodman, G., (1999) Incorporating the time dimension in receiver operating characteristic curves: A case study of prostate cancer. Medical Decision Making, 19, 242-251.

Graf, E., Schmoor, C., Sauerbrei, W. and Schumacher, M. (1999). Assessment and comparison of prognostic classification schemes for survival data. Statistics in Medicine, 18, 2529-2545.