A Review of Statistical Methods for Safety Surveillance

Abstract

The data-mining statistical methods used for disproportionality analysis of drug-adverse event combinations from large drug safety databases such as the FDA's Adverse Event Reporting System (FAERS), consisting of spontaneous reports on adverse events for postmarket drugs, are called passive surveillance methods. However, the statistical signal detection methods for longitudinal data, as the data accrue in time, are called active surveillance methods. A review of the most commonly used passive surveillance statistical methods and the relationships among them is presented with unified notations. These methods are applied to the 2006-2012 FAERS data; the number of drug signals of disproportionate rates (SDRs) detected by each of these methods with the common SDRs from all of these methods, for the adverse event myocardial infarction, are given. Finally, there is a brief discussion on the recently developed active surveillance methods.

A nonparametric method to detect increased frequencies of adverse drug reactions over time

“…Valid test statisticsT for the individual testing problems (7) are standard and easy to obtain. A test statistic maxT for the global null hypothesis can be obtained as the maximum of the individual test statistics.…”

“…In the remainder of this section, we will describe the approach in a bit more detail. An appropriate test statisticT for the individual testing problems (7) compares the observed reporting proportionŝ…”

“…. Hence, for ADR j, there are 2 unknown parameters p j [ref] and p j [rep] , and we are interested in the testing problem (7) as defined in Section 4.…”

“…We use a weighted 2-sample t-test statistic to test (7). Obviously, this is a test statistic for normally distributed data, and the observed monthly rates are not normally distributed data.…”

“…Examples for Bayesian approaches to disproportionality analysis can be found in Bate et al 1 and DuMouchel. 2 More recently, a frequentist likelihood ratio-based approach has been suggested by Huang et al 3 Methods that adjust for exposure or exposure duration were introduced by Schuemie et al 4 and Huang et al 5 For a comprehensive overview over methods for disproportionality analyses, we refer to the excellent review papers by Almenoff et al 6 and Huang et al 7 Less attention has been paid to methods directed at detecting increased frequencies of already identified ADRs. Norwood and Sampson 8 introduced a method that compares the data obtained during a reporting period with those of a reference period.…”