We recruited previously healthy children ages 1 mo to 2 y who were admitted to Children's Hospital Colorado with a diagnosis of bronchiolitis from 2013 to 2015; Recreational marijuana use became legal in Colorado on 1 January 2014. We obtained urine samples from the child within 24 h of admission; these were placed in a −80 °C freezer. Families were given a $25 gift card for participation. For this secondary analysis, only samples from children whose parents consented for future research were used, and the data and samples were anonymized prior to analysis. At the time of the secondary analysis, samples were shipped on dry ice to the laboratory at the US Centers for Disease Control and Prevention. The initial study and secondary analysis were both approved by the Colorado Multiple Institutional Review Board; since families had consented for future research we did not need to reconsent.

After completing informed consent, parents had completed a questionnaire about their child's health, and demographics. Since birth-dates were removed from the secondary dataset to ensure anonymity, we obtained patient age from the family census. Parent report of tobacco smoke exposure was assessed with the question: “Does anyone who lives in your home or who cares for your child smoke tobacco (yes or no)?” Parent report of marijuana smoke exposure was assessed after October 2014 with the question “Does anyone who lives in your home or who cares for your child smoke marijuana (yes or no)”.

Analysis for marijuana biomarkers was performed using the method by Wei et al.(12) Urine samples were equilibrated with isotopically labeled internal standards, deconjugated by enzymatic and alkaline hydrolysis, and extracted using C18 sorbent. The final extracts were concentrated and 10 μl of each sample was injected into an ultrahigh performance liquid chromatograph and analyzed using tandem mass spectrometry under electrospray ionization mode (ESI-MS/MS). This method has limits of detection of 0.005, 0.015, and 0.009 ng/ml for the three metabolites tested: total urinary THC, COOH-THC, and CBD, respectively.

Urinary cotinine (COT) was analyzed using a modification of the method developed by Bernert et al.(13) After adding isotopically labeled internal standards to 200 μl of urine, the samples were hydrolyzed with β-glucuronidase, purified using supported liquid extraction, concentrated, and then taken up in 100 μl water. A 5 μl aliquot of this solution was injected into a ultrahigh performance liquid chromatograph and analyzed using atmospheric pressure chemical ionization tandem mass spectrometry. The limits of detection for COT was 0.030 ng/ml. We used a cutoff of 2.0 ng/ml to identify children with likely household contact exposure, based on Matt et al.(8) Laboratory blank and quality control samples were simultaneously processed and analyzed to assure the quality of the analytical results (14).

Chi square tests were done to assess bivariable differences, and all analyses were done using SAS, (SAS Institute, Cary, NC). For 10 subjects (23%), there was more than one family member under 2, making identification of the subjects' exact age difficult, so we performed two sensitivity analyses in which we assumed all of the missing participants were (i) in the younger age group and (ii) in the older age group; our results did not change significantly.

There are several limitations to this study. As we were working with limited samples remaining from a prior study, we have a very small sample size. We had insufficient power to be able to determine additional associations, such as with illness severity, or to complete more quantitative associations, and thus it has limited external validity. Since this was a secondary analysis on previously collected samples, we were unable to assess additional demographic, socioeconomic status, and other factors that may have been associated with marijuana use. Since we could only analyze the samples from those with consent for future research, it is possible that there was bias towards non-marijuana using families in the sub-sample, and thus the true proportion of exposed children could be higher. We did not assess for medical use of marijuana in these children specifically, but since we enrolled only previously healthy children, medicinal marijuana use would be highly unlikely. We are also unable to assess whether the exposure was from secondhand marijuana smoke vs. neonatal transmission or breastfeeding; however the fact that our exposure rate was similar in the <1-y-old age group as the 1–2-y-old age group would suggest that the detectable COOH-THC is more likely from the secondhand smoke. Finally, this is a sample of children hospitalized for bronchiolitis in Colorado, and the findings may not be generalizable to other populations.

Marijuana metabolites are detectable in young children with exposure to secondhand marijuana smoke. These children are exposed to the psychoactive compounds in marijuana, and are potentially at risk for negative health effects. More than half of children with tobacco smoke exposure were also exposed to marijuana. Understanding the health consequences of marijuana smoke exposure in children is critical for providers, parents, and policymakers to best protect children from harmful exposures.