Software for Efficient Estimation of the Cox Model With Interval-Censored Data

Statacorp Llc, College Station TX

Investigators

Abstract

Software for Ef?cient Estimation of the Cox Model With Interval-Censored Data Project Summary Interval-censored data arise frequently in clinical and epidemiological studies because the time to the develop- ment of an asymptomatic disease (e.g., tumor occurrence, HIV infection) cannot be observed exactly but rather is known to lie in a time interval between two consecutive clinical examinations. Recent theoretical and com- putational advances have made it possible to ?t the well-known Cox proportional hazards model with potentially time-dependent covariates to interval-censored data through ef?cient nonparametric maximum likelihood esti- mation. The broad, long-term objective of this SBIR proposal is to create a suite of commands, along with a companion text, in the widely used commercial software package Stata for performing cutting-edge nonparamet- ric maximum likelihood estimation of the Cox proportional hazards model with time-dependent covariates and interval-censored event times. The goal of the Phase I project is to establish the scienti?c merit and technical feasibility of the proposed research and development effort, with two speci?c aims: (1) to develop and implement numerical methods for performing nonparametric maximum likelihood estimation of the Cox proportional hazards model for a single disease with unrelated subjects under interval censoring; and (2) to certify the correctness of the estimation results of the new command by comparing them to results from published papers and limited re- search code. The Phase II project then will use the procedures developed in Phase I to build a suite of commands for semiparametric regression analysis of interval-censored data that is reliable, robust, user-friendly, and opti- mized for speed and memory ef?ciency. The code also will be expanded substantially to allow non-proportional hazards and to deal with multiple diseases and clustered data under interval censoring. In addition, a compan- ion text will be written to not only document the software itself, but also serve as a substantive reference for researchers new to the ?eld. The software program produced by this SBIR project will be an integral part of the Stata package and thus will be marketed by StataCorp. This powerful new tool will enable biomedical investi- gators to analyze interval-censored data in a statistically ef?cient and unbiased manner. As such, the tool will facilitate the search for effective intervention and prevention strategies for many common diseases (e.g., cancer, HIV/AIDS, diabetes, hypertension), thereby leading to improvements in public health.