Semiparametric time to event models in the presence of error-prone, self-reported outcomes—With application to the women’s health initiative

Abstract

The onset of several silent, chronic diseases such as diabetes can be detected only through diagnostic tests. Due to cost considerations, self-reported outcomes are routinely collected in lieu of expensive diagnostic tests in large-scale prospective investigations such as the Women’s Health Initiative. However, self-reported outcomes are subject to imperfect sensitivity and specificity. Using a semiparametric likelihood-based approach, we present time to event models to estimate the association of one or more covariates with a error-prone, self-reported outcome. We present simulation studies to assess the effect of error in self-reported outcomes with regard to bias in the estimation of the regression parameter of interest. We apply the proposed methods to prospective data from 152,830 women enrolled in the Women’s Health Initiative to evaluate the effect of statin use with the risk of incident diabetes mellitus among postmenopausal women. The current analysis is based on follow-up through 2010, with a median duration of follow-up of 12.1 years. The methods proposed in this paper are readily implemented using our freely available R software package icensmis, which is available at the Comprehensive R Archive Network (CRAN) website.

Bayesian variable selection for high dimensional predictors and self-reported outcomes

“…In this paper, we incorporate a BVS approach into a likelihood-based model proposed by Gu, X. et al (2015). This allows us to conduct variable selection in high dimensional data while accounting for the imperfect observation of a time-to-event outcome.…”

“…Other related work includes that proposed in the context of HPV studies [ 6 ], where the authors accommodate misclassification by incorporating ideas of binary generalized linear models with outcomes subject to misclassification [ 11 ]. A formal likelihood framework was proposed to accommodate sequentially administered, error-prone self-reports or laboratory based diagnostic tests for modeling the association of a targeted set of covariates with the time-to-event outcome of interest [ 12 ]. While a rich literature exists to handle estimation and hypothesis testing in the presence of error-prone survival outcomes, none of these approaches can be applied directly to variable selection in high-dimensional data, in which the number of features ( p ) far exceeds the number of subjects ( n ).…”

Study design for non‐recurring, time‐to‐event outcomes in the presence of error‐prone diagnostic tests or self‐reports

“…In this setting of error-prone diagnostic procedures such as self-reports, standard regression methods for interval censored outcomes are rendered invalid. In recent work, Gu et al [11] present a likelihood-based approach to estimate the association of specific covariates with a time-to-event outcome that is observed through periodic, imperfect self-reports. In this work, the authors show through simulations that analysis procedures that ignore the error inherent in imperfect diagnostic procedures can produce results that are substantially biased.…”

“…When analysis proceeds by ignoring the error in the diagnostic procedure, the resulting estimates of the regression coefficients of interest can be biased as significantly as 90% towards the null. However, when the regression analysis is based on a likelihood procedure that incorporates the error-prone nature of the diagnostic test, this bias is eliminated [11].…”

“…In this paper, we extend the semi-parametric likelihood-based approach described in Gu et al [11] to obtain power and sample size calculations. We consider studies aimed at estimating the association of one or more covariates with a time-to-event outcome that is observed using a repeatedly administered, error-prone diagnostic procedure such as a self-reported questionnaire.…”