A false-positive statistic changes meaning when its denominator changes. In a frequently cited bupropion study, 41% described medication histories among unconfirmed positive screens. It was not the incidence of false positives among everyone taking the medicine.
Reconstructing the tested population
Casey and colleagues’ 2011 retrospective hospital study examined 10,011 urine drug screens. Syva EMIT II Plus amphetamine screening was positive in 362; GC–MS confirmed 234, while 128 failed to confirm. The study counted dextroamphetamine, methamphetamine or MDMA detection as confirmation. Recorded bupropion use occurred in 53 of those 128 unconfirmed results (41%), and in three confirmed results.
The investigators reviewed records after a positive amphetamine screen. They did not establish the total number of bupropion-treated patients among all screen-negative patients. Medication lists could be incomplete, adherence was unverified and the population came from one urban trauma-center setting. The findings were specific to the evaluated testing process.
Two different conditional probabilities
The observed proportion answers: among unconfirmed amphetamine-positive screens, how often was bupropion recorded? The patient-facing question is different: among people taking bupropion, how often will screening produce an unconfirmed positive?
Answering the second question requires the complete denominator of exposed people who underwent the relevant testing, including those with negative screens. A sample selected on a positive screening result cannot supply that missing denominator. Reversing the conditional probability would turn the study into an incidence estimate it did not calculate.
Does the prescription prove interference in a particular specimen?
No. A documented prescription identifies a plausible analytical question. It does not establish ingestion, the relevant drug or metabolite concentration, or the cause of an antibody response in a specific sample. The confirmatory result and the exact assay remain essential parts of the explanation.
The three confirmed results among records noting bupropion also matter. A medication history does not make a positive result automatically false. A particular person can have the medication recorded and a target compound confirmed; the two observations are not mutually exclusive.
Nor should the hospital proportion be transferred to another immunoassay platform without supporting evidence. A reported interference concerns a particular analytical process. Its relevance depends on the actual method used in the case being reviewed.
A defensible use of the paper
The study supports taking an unexpected amphetamine screen seriously as a question for confirmation and method review. It does not establish an individual’s drug source, impairment or credibility from medication history alone. Keeping the 53/128 denominator attached to the 41% figure prevents a useful historical observation from becoming a misleading personal risk estimate.
Source
Casey ER, Scott MG, Tang S, Mullins ME. Frequency of false positive amphetamine screens due to bupropion using the Syva EMIT II immunoassay. J Med Toxicol. 2011;7:105–108. doi:10.1007/s13181-010-0131-5. Full text.