Label testing: sample, method and denominator

A 2017 study tested 84 CBD products bought online in 2016: 26 were within its label-agreement definition, 36 contained more CBD than labeled and 22 contained less. Its historical online sampling frame and ±10% definition belong beside those counts. It did not test clinical benefit or establish today’s market prevalence.

The original publisher’s correction restored funding and sponsor-role disclosures, including support and unpaid board relationships. The corrected source should be consulted rather than interpreting missing disclosures in an earlier version as “none.”

The THC study used different materials and a ±15% threshold. Read its separate guide rather than pooling it with this historical CBD sample.

The aggregate result from another flower study is held because the stated tolerance method and published rounded-table arithmetic do not align. We do not publish either the paper’s aggregate or a reviewer-derived alternative as a corrected prevalence estimate. The fictional diagram below shows why a denominator must be stated; it is not a reanalysis of that study.

A separate study tested 118 seized nonretail cannabis samples for targeted drugs. It reported detections in 25 samples; individual drug counts overlap. This enforcement-selected material is not a probability sample of ordinary retail purchases. The detection fraction cannot be recast as a general retail contamination rate.

Fentanyl was not detected in those samples under the reported assay. That observation does not establish that all cannabis lacks fentanyl or other contaminants. Main-text validation did not supply every analyte-specific limit; unreviewed supplementary detail remains unverified. No human exposure, treatment outcome or causal contamination mechanism was measured.

The seized-sample study reported university funding and no conflicts. Its sampling limitations do not become a vendor-quality score; report tools do not certify current-lot safety.