Does a Child’s Positive Urine CBD Metabolite Prove THC Exposure or Behavioral Harm? An 81-Pair Study

Short answer: No. In an 81-pair Colorado cohort analysis, cannabinoid metabolites detected in a child’s urine at age five were associated with several parent-reported behavioral scores; a smaller 42-child subset also showed differences on two computerized cognitive tests. But the children’s positive urine findings were CBD metabolites, not measured parent THC or THC-COOH. The observations do not establish how exposure occurred, whether the child had been intoxicated, or whether the detected substances caused a behavioral condition. They also cannot diagnose ADHD or determine a disputed child’s individual prognosis. These are the findings and limits of Moore and colleagues’ 2023 Healthy Start primary study.

What question did the Healthy Start investigators actually study?

The investigators asked whether laboratory-classified prenatal and childhood cannabinoid exposure was associated with cognition and behavior at approximately five years of age. That is different from asking whether a particular child’s behavior was caused by smoking, secondhand cannabis smoke, a CBD product, or a precisely timed dose of THC.

The parent Healthy Start cohort enrolled 1,410 pregnant participants in Colorado during 2010–2014. This analysis used a convenience sub-study with stored maternal urine collected at approximately 27 weeks’ gestation and child urine obtained during a five-year follow-up. The investigators analyzed exposure markers from those two different dates and assessed behavior and cognition at age five. The study was observational; investigators did not administer cannabis or randomly assign children to exposure.

The central forensic issue is thus what a qualitative laboratory exposure category proves. The category was constructed for statistical comparison. It was not a medical diagnosis, a reconstruction of absorbed THC dose, or a measurement of an individual’s functioning on the day of a later incident.

How many mother–child pairs were actually included?

Several different denominators appear in the paper and must not be combined. Of 1,410 participants in the parent cohort, 199 entered the stored-urine sub-study. After excluding 20 preterm or low-birth-weight deliveries and 86 records missing specified covariate information, 93 remained eligible. Twelve of those did not complete the Child Behavior Checklist (CBCL), leaving 81 pairs for behavior analyses. Forty-seven eligible children did not complete the NIH Toolbox cognition battery, leaving only 42 for cognitive analyses.

Among the 81 behavior-analysis pairs, maternal mid-gestation urine classified six as prenatally exposed and 75 as not exposed. Age-five child urine classified ten as exposed and 71 as not exposed. Two children were positive at both time points. For cognition, the corresponding groups were five prenatal-positive versus 37 negative, and four childhood-positive versus 38 negative. A cognitive association involving four exposed children is not a result from 81 exposed children—or from the entire 1,410-person parent cohort.

What did the laboratories measure in the mother’s urine and the child’s urine?

The investigators used a liquid-chromatography/tandem-mass-spectrometry (LC–MS/MS) procedure on stored urine. The analytical panel included parent delta-9 THC, 11-hydroxy-THC, THC-COOH, THC and THC-COOH glucuronides, CBD and CBD metabolites, and several other cannabinoids. A finding in maternal urine collected during pregnancy and a finding in child urine five years later are separate laboratory events in different people.

Maternal specimens included detectable THC-related analytes. For example, the source’s Table 2 reported parent THC in three maternal samples, THC-COOH in four, and THC-COOH glucuronide in six; these detection counts overlap and cannot be summed to count exposed mothers. By contrast, none of the age-five child specimens had detected parent THC, 11-hydroxy-THC, THC-COOH, or the listed THC glucuronides. Child detections consisted of CBD glucuronide in nine samples and CBD-COOH in one.

That distinction is decisive for interpretation. A child categorized as cannabinoid-positive in this dataset should not automatically be described as a child with confirmed urinary THC or confirmed recent marijuana intoxication. The parent analyte, metabolite, reporting threshold, specimen, and sampling date all matter. A targeted urine result also cannot establish whether a substance originated from an ingested product, environmental contact, secondhand exposure, or another pathway without additional evidence.

What do the urine measurements and study denominators show?

The paper describes detection and quantification thresholds that vary by analyte; its methods and Table 2 use threshold headings that warrant specimen-level verification before any case-specific comparison. The reported group classification is not equivalent to a known ingested milligram dose or to a venous whole-blood THC concentration. Where the source measured an analyte, it is more precise to name that analyte than to replace it with the broad word “cannabis.”

What parent-reported behaviors differed among the child-urine groups?

The five-year behavior assessment was the preschool Child Behavior Checklist, completed by mothers about the child’s recent behavior. Its T-scores are standardized ratings, not clinical diagnoses independently confirmed by a physician. The investigators fitted covariate-adjusted models that included maternal age, education, race/ethnicity, maternal psychiatric-diagnosis history, the child’s age and sex, and maternal and child urine cotinine classifications for tobacco exposure.

Comparing the ten age-five urine-positive children with the 71 negative children, the authors reported higher adjusted scores for aggressive behavior (difference 3.2 T-score points; 95% CI 0.5 to 5.9; p = .02), DSM-oriented attention-deficit/hyperactivity problems (8.0 points; 95% CI 2.2 to 13.7; p < .01), and oppositional/defiant problems (3.2 points; 95% CI 0.2 to 6.3; p = .04).

Those findings should be interpreted alongside the nulls. The broader externalizing composite had an adjusted difference of 3.5 (95% CI −3.4 to 10.4; p = .32), which was not statistically significant. The separate attention-problems syndrome scale had a confidence interval from −0.1 to 4.0, crossing zero, despite a printed p-value of .05. Individual syndrome scales, DSM-oriented ratings and broadband composites are related but nonidentical outcomes; selective quoting of a positive subscale can misrepresent the fuller pattern.

Which cognitive-test scores differed, and how small were the groups?

The NIH Toolbox subset included only four children with age-five urine metabolite detection and 38 without. The investigators reported lower adjusted cognitive-flexibility scores on the Dimensional Change Card Sort task (difference −15.6 T-score points; 95% CI −30.0 to −1.2; p = .03) and lower receptive-language scores on Picture Vocabulary (difference −9.7 points; 95% CI −19.2 to −0.3; p = .04).

Not every cognitive measure differed significantly. Inhibitory-control performance on the Flanker task showed an adjusted difference of −9.3 points (95% CI −19.1 to 0.5; p = .06), so the confidence interval included no difference. For the prenatal urine classification, none of the three cognition comparisons was statistically significant: cognitive flexibility p = .72, inhibitory control p = .24, and receptive language p = .07. A small positive or negative estimate is not itself proof of a biological effect.

What did the significant and nonsignificant comparisons look like together?

These source-reported p-values accompany numerous tested outcomes. The published analysis applied a conventional p < .05 rule and did not report a family-wide multiplicity correction. With small exposed groups and many related tests, isolated nominal findings may be unstable. They deserve consideration, not conversion into a certain predictive rule.

Did prenatal urine positivity show the same pattern?

No. The six prenatal-positive mothers formed a different exposure comparison from the ten age-five urine-positive children. Prenatal urine positivity was associated with lower maternal reports of childhood internalizing problems (adjusted difference −10.2; 95% CI −20.3 to −0.2; p = .04) and fewer somatic complaints (−5.2; 95% CI −9.8 to −0.6; p = .03). The prenatal group did not show statistically significant differences in the three NIH Toolbox cognition tasks.

It would be scientifically unsound to call this evidence that cannabis during pregnancy protects children from emotional problems. The prenatal-positive group comprised only six pairs, exposure ascertainment was limited to one mid-gestation urine sample, participants were selected through multiple exclusions and missing-data steps, and substantial home-environment and longitudinal exposure factors were not measured. The study itself discussed selection bias and unexpected directions of association. Observational inverse associations are not recommendations about prenatal cannabis use.

Can child urine CBD metabolites distinguish smoke exposure from ingestion?

Not on the evidence reported here. The age-five child urine detections were CBD metabolites, and the investigators did not establish a product-by-product source or quantify a child’s inhaled, swallowed, or dermally transferred dose. CBD-metabolite detection indicates an analytical finding under the test’s reporting rules. It does not independently identify a smoke plume, a specific consumer product, an adult caregiver, environmental contamination, or the route of administration.

Even when a cannabinoid exposure route is otherwise established, a urine sample integrates elimination over a period determined by the analyte and the person’s physiology. It does not function as an event-time THC impairment meter. For a case-specific examination, a reviewer would need the original report, specimen collection and storage record, analyte-level identification and reporting limits, possible products and routes, and independent medical or environmental evidence. See the site’s published guides to specimen collection and confirmatory analytical testing.

Why does a five-year urine result not establish exposure throughout childhood?

The investigators collected urine at only two life stages: maternal urine around 27 gestational weeks and the child’s urine around the fifth birthday. The study did not measure every trimester, every postnatal month, household air concentrations, or serial blood THC concentrations. A positive sample therefore cannot reconstruct a continuous five-year exposure history. A negative at that one age likewise cannot rule out every earlier episode.

The measured outcome and the age-five urine result were contemporaneous. A caregiver’s use patterns, a child’s environment, other household factors, or behavioral difficulties could precede or accompany the observed exposure classification. The temporal direction cannot be established from these measurements alone. The underlying models adjusted for several plausible variables including cotinine, but adjustment is not the same as randomization or elimination of all confounding.

What selection and statistical limitations affect causal interpretation?

The analysis shrank from 1,410 initially enrolled pairs to 199 in the stored-urine sub-study, then 81 with CBCL data and 42 with cognition data. Excluding preterm or low-birth-weight children may reduce variation in important developmental pathways while making the final sample less representative of all children. The source reported prenatal urine positivity of 13% in the 199-pair sub-study but 7% in the 81-pair analytic sample; corresponding childhood detection percentages were 19% and 12%. Loss to follow-up and missing information can alter effect estimates.

The exposure definition also differed by time point, with maternal detection/quantification rules and child urine classification subject to analyte-specific thresholds. One measured maternal or child urine sample provides incomplete information about frequency, duration and route of exposure. Household socioeconomic circumstances, medical history, parental behavior and developmental factors may remain incompletely captured even after covariate adjustment. The tiny cognitive exposed group and uncorrected multiple comparisons further limit causal specificity.

Does this study establish that cannabis caused ADHD, aggression, or cognitive disability?

No. The study reported associations between a laboratory-derived urine-positive category and age-five rating scales or tests. It did not establish clinical diagnoses of ADHD or oppositional defiant disorder through a diagnostic evaluation, nor determine the onset of such disorders or demonstrate that measured CBD metabolites caused them. It offered no randomized exposure intervention and no established child-specific injury threshold or individual causation model.

In litigation, the legally relevant question may concern duty, source, exposure amount, timing, diagnosis, alternative causes, or whether an alleged exposure more likely than not contributed to a particular harm. Those questions cannot be resolved simply by copying a cohort regression coefficient. A defensible analysis must connect the actual child’s specimen and clinical course with the appropriate evidence, while explicitly considering alternative explanations and the study’s narrow inferential scope.

How should a forensic reviewer approach a disputed child cannabinoid finding?

Begin with the complete original laboratory report—not a shorthand label such as “cannabis-positive.” Identify whether the finding was parent THC, THC-COOH, THC glucuronide, CBD-COOH, CBD glucuronide or another compound; confirm the matrix and collection date; and distinguish a screen from specific identification. In this study, the primary child findings were CBD conjugate/metabolite signals, not urinary THC signals. That distinction may materially change the hypothesis being tested.

Then review sample handling, laboratory quantification limits, known household products, possible alternative sources, serial results if any, parental and child medical history, timing of measured developmental concerns, and the assessment instrument used. Finally, separate scientific questions: evidence of exposure, route and time of exposure, dose, clinical diagnosis, and causal contribution are different propositions. The site’s biological-matrices guide explains why a valid urine result should not be silently translated into a blood result or a contemporaneous behavioral effect.

Frequently asked questions

Did ten children test positive for THC at age five?

No. Ten were classified as cannabinoid-positive in age-five urine, but the reported detected child analytes were CBD glucuronide in nine and CBD-COOH in one; the listed THC analytes were not detected. The study did not establish contemporaneous THC intoxication in those ten children.

Were 81 children included in the cognitive-test comparison?

No. There were 81 children in the CBCL behavior analysis, but only 42 in NIH Toolbox cognition analyses. Just four of the 42 had positive age-five urine under the study definition.

Did the study prove that cannabis causes ADHD?

No. An adjusted association with a mother-completed DSM-oriented attention-deficit/hyperactivity rating is not a clinical diagnosis or proof of causation. The small sample and other sources of uncertainty matter.

Can CBD-glucuronide in urine establish that a child inhaled secondhand smoke?

No. It identifies a reported analyte under the assay’s conditions. The article did not validate which route, product, exposure event, or caregiver caused the finding.

Were the prenatal cannabis results harmful across all outcomes?

No. The paper reported inverse associations for some mother-rated behavior measures and nonsignificant cognition comparisons. Those inverse estimates also do not demonstrate that prenatal cannabis was protective.

Could a negative single urine test exclude all cannabis exposure during early childhood?

No. One specimen at age five does not cover every preceding exposure opportunity, and its interpretation depends on analytes, detection limits, timing and collection conditions.

Can the regression coefficients predict impairment or disability in an individual lawsuit?

No. The coefficients describe adjusted associations in selected research groups, not a validated prediction for one person’s impairment, diagnosis, developmental trajectory or incident causation.

Primary study and scope

Original primary research: Moore BF, Salmons KA, Hoyt AT, et al. Associations between Prenatal and Postnatal Exposure to Cannabis with Cognition and Behavior at Age 5 Years: The Healthy Start Study. International Journal of Environmental Research and Public Health. 2023;20(6):4880. https://doi.org/10.3390/ijerph20064880. This article reports the study’s actual maternal/child urine matrices, denominators, behavioral ratings, cognitive tests and limitations; it does not provide an individualized medical or legal causation opinion.