Abstract
OBJECTIVES: Evaluate the classification accuracy of biomarkers of nicotine exposure for identifying noncompliance in randomized controlled trials (RCTs) of very low nicotine content (VLNC) cigarettes. METHODS: We combined data from 2 studies to evaluate the classification accuracy of biomarkers of nicotine exposure for identifying noncompliance in RCTs of VLNC cigarettes. Using a novel approach that did not require knowledge of each participant's compliance status, we modeled the distributions of total nicotine equivalents (TNE), total cotinine, and anatabine in compliant and noncompliant participants using a mixture model. Estimates of the classification accuracy were derived from the estimated densities. RESULTS: TNE and total cotinine had near-perfect classification accuracy, but TNE had better classification accuracy than total cotinine (p = 0.03) and anatabine (p = 0.014). The classification accuracy of TNE and total cotinine decreased as self-reported study cigarettes increased; anatabine was similar, but the results were more statistically variable. CONCLUSIONS: TNE and total cotinine are better classifiers of compliance than anatabine. These results will be useful in determining biomarker thresholds for identifying noncompliance and will aid in the interpretation of future RCTs of VLNC cigarettes.