Composite likelihood based tests with applications in pharmacovigilance and clustered data
Data mining tools for systematically searching safety signals in surveillance databases have been developed recently. The reporting rates of certain vaccine-event combinations vary largely across reporting years. Current methods seldom consider these variations in signal detection. Instead, the reports of different years are collapsed together when conducting safety analyses. Therefore, we proposed a random effect model to test the heterogeneity of reporting rates for a given vaccine-event combinations across reporting years. This method has been shown to have a great power in detecting the variation in reporting rates across years. Our findings on the temporal trend of the reporting rates may reveal the impact of vaccine updating on adverse events and provide evidence for further investigations. The application of the proposed likelihood ratio test is illustrated by investigating vaccine FLU3 in Vaccine Adverse Event Reporting System (VAERS) database. The zero inflation of number of counts is commonly encountered in surveillance reporting system such as VAERS. Existing method rarely consider this important feature in the procedure of safe signal detection. Here, we proposed a random effect model to test the heterogeneity of reporting rate of vaccine-event combination among calendar years while accounting for the zero-inflation observations. Furthermore, a ranking procedure based on shrinkage estimator of reporting rate for each year using Empirical Bayes method is demonstrated to reveal the temporal trend of reporting rate among the reporting years. Standard methods for two-sample tests such as the t-test and Wilcoxon rank sum test may lead to incorrect type I errors when applied to longitudinal or clustered data. Recent alternatives of two-sample tests for clustered data often require certain assumptions on the correlation structure and/or noninformative cluster size. In this paper, based on a novel pseudo-likelihood for correlated data, we propose a score test without knowledge of the correlation structure or assuming data missingness at random. The proposed score test can capture differences in the mean and variance between two groups simultaneously. We use projection theory to derive the limiting distribution of the test statistic, in which the covariance matrix can be empirically estimated. We conduct simulation studies to evaluate the proposed test and compare it with existing methods. To illustrate the usefulness proposed test, we use it to compare self-reported weight loss data in a friends’ referral group, with the data from the internet self-joining group.
Cai, Yi, "Composite likelihood based tests with applications in pharmacovigilance and clustered data" (2016). Texas Medical Center Dissertations (via ProQuest). AAI10250177.