Detecting the Detection Limit

Detection limits are highly controversial in practice despite decades of developmental efforts by analytical scientists and regulators. This high resistance to obtaining effective and accepted detection limit solutions indicates our collective failure in how we traditionally approach this problem. The analytical chemist approaches it from analytical science, the statistician from statistical theory, the regulator from a desire for regulatory needs, the instrument manufacturer from the standpoint of selling instruments, the contract laboratory from the standpoint of selling services, and the regulated from the standpoint of reducing the cost of regulation. This exquisite mix of science, misapplication of science, regulatory needs, and business concerns highlights why serious trouble abounds. Figuring out how to simultaneously address these constraints is not a trivial task. Our journey to navigate the detection limit obstacle course successfully begins with a broad view of the problem focused on developing a conceptual understanding of key detection limit concepts and limitations. Throughout the presentation, key detection limit concepts, including the problems caused by our natural biases, will be explored. Issues in detection limit and rule-set application will be probed because it is at the application level that detection limit standards will succeed or fail. This presentation does not require anything other than a rudimentary knowledge of statistics. Although statistics plays a key role in any good detection limit definition, statistics needs to be the last and final piece of solving the detection limit puzzle. In fact, to succeed, statistical practice must come last. Statistics must not drive the definition process; the definition and the problem environment must drive the statistical process's development.