Predicting Biological Warfare Agent Detector Performance

Abstract

Biological warfare agents (BWAs) are inherently dangerous. United States of America public law forbids the release of biological warfare agent into the environment. The test and evaluation community desires realistic open air field tests, but does not want to risk harming test participants or the general public. Neither do they want to break the law by releasing actual BWAs during open air field tests. To test biological defense systems in the field, the test and evaluation community releases relatively harmless substances known as simulants. It is highly desirable that the biological warfare system under test performs identically with the simulant as it does with real biological warfare agent. Unfortunately this never occurs. A new class of simulants known as Agent-Like Organism (ALO) has been developed. An ALO is phylogenetically closely related to its corresponding biological warfare agent. A vaccine strain is an example of ALO. The purpose of this dissertation is twofold: to determine if killed or inactivated ALOs are acceptable simulants for a biological point detection system that uses particle fluorescence and immunoassay technology; and to develop a model based on logistic regression to relate detector simulant performance to detector performance with biological warfare agent. Data for this analysis was obtained from Dugway Proving Ground, Utah. The data set consisted of 2,717 Joint Biological Point Detector System (JBPDS) challenges. The system was challenged with either BWA or ALO simulant. All BWA challenges occurred in a Bio-Level 3 (BL-3) facility. Data was analyzed using logistic regression. The determination as to the acceptability of the ALOs as simulants was based on both statistical analysis and judgment on the impact of differences on the performance in an open air field test. Results are given that show the acceptability of various simulants for particular BWAs with respect to both detection and identification. The simple models examined in this dissertation do not adequately explain the variability that occurs in field open air releases. As a result of unexplained variability in detector performance during open air field releases, even the best predictive model is of minimal utility in predicting detector performance. The best predictor of biological warfare agent detector performance is field trials with killed ALOs. A possible method to improve the acceptability of Killed N ALO and Killed NU ALO as identification simulants is discussed.