Agreement between a person’s account and urine testing depends on what each measurement covers. A historical validation analysis of methamphetamine self-report illustrates both the value of a short recall window and the danger of turning disagreement into a claim about honesty.
What was the study unit?
Rowe and colleagues analyzed 327 visits from 90 adults enrolled in a pharmacotherapy trial. This was a secondary validation analysis within a randomized trial, not an experiment randomly assigning people to give accurate or inaccurate accounts. Individuals contributed up to four monthly observations.
Self-report came from audio computer-assisted interviews. Recent use was defined as any use during the preceding three days. Weekly urine samples were screened with VERDICT-II assays. The reference was urine screening under the study’s procedures, not continuously observed ingestion.
What agreement was found?
For three-day self-report compared with urine toxicology, sensitivity was 86.7%, 95% confidence interval 81.4%–91.4%; specificity was 85.3%, interval 77.7%–91.3%. Positive predictive value was 91.5%, interval 86.9%–94.8%, and negative predictive value was 78.0%, interval 69.4%–86.1%. Kappa was 0.71.
Those measures answer different conditional questions. Sensitivity concerns reports among urine-positive visits. Negative predictive value concerns urine results among visits with a report of no recent use. Neither is simply the percentage of people who tell the truth.
Why do repeated visits matter?
Four observations from one person are not equivalent to observations from four independent people. The investigators accounted for repeated observations through participant-level resampling for relevant confidence intervals and longitudinal regression methods.
Calling the sample “327 methamphetamine users” would exaggerate its breadth. The visit count increases information about repeated reporting, while the 90-person denominator defines the number of individuals represented. Both numbers belong in the account.
Did a longer recall window give the same result?
For reports of no use over the longer interval between monthly interviews, negative predictive value was 70.6%, with a wide 95% interval of 48.0%–85.7%. That analysis used intervening weekly urine results and different eligibility rules, so it was not an identical test with only one variable changed.
Weekly sampling also left gaps between urine detection opportunities. A set of negative weekly results could not verify every day of the month. A longer self-report interval and intermittent biological testing therefore require careful temporal alignment.
Can discordance establish dishonesty?
No. Recall, interpretation of questions, assay characteristics, exposure timing and detection windows can all matter. This study measured correspondence between two operational definitions; it did not adjudicate each participant’s intent.
The study population was also selected. Entry required acknowledged methamphetamine use and a positive urine test, and participants knew urine testing would occur. The authors identified these conditions as limits on generalizing the findings to other settings.
What is the useful conclusion?
Recent self-report showed substantial agreement with urine screening in this treatment-research setting. Validity differed across recall periods, follow-up and some participant characteristics, but those group associations cannot diagnose an individual’s credibility. For scientific or forensic use, preserve the person-versus-visit denominator, the recall window and the actual biological reference before assigning meaning to agreement or disagreement.
Source
Rowe C, Vittinghoff E, Colfax G, Coffin PO, Santos GM. Correlates of Validity of Self-Reported Methamphetamine Use among a Sample of Dependent Adults. Substance Use & Misuse. 2018;53(10):1742–1755. DOI: 10.1080/10826084.2018.1432649. Full text; bibliographic record.