Novel study reveals that a surprising number of pregnant people are using cannabis — and need to be informed of its risks
Breaking research shows that rates of cannabis use during pregnancy are far higher than previously thought, a finding that could improve efforts to identify pregnant cannabis users and inform them of potential risks.
This study will be presented at ADLM 2024 (formerly the AACC Annual Scientific Meeting & Clinical Lab Expo), along with a second study on a machine learning model that predicts the duration of opioid use after surgery.
Cannabis is rapidly becoming legalized in more and more states across the U.S. and its recreational use is skyrocketing, but its effects on the developing fetus are not fully understood. Current recommendations advise against cannabis use and exposure during pregnancy due to its association with negative outcomes, such as preterm birth, fetal growth restriction, low birth weight, and developmental deficits.
A research team from NMS Labs led by Dr. Alexandria Reinhart sought to investigate the prevalence of cannabinoid exposure in utero during 2019-2023 by analyzing umbilical cord samples submitted for testing. Of the 90,384 samples tested, 44% were positive for at least one of approximately 60 analytes included in the testing panel, and cannabinoids accounted for 59%-63% of all positive results, making them the most common drug found.
As the effects of cannabinoids on health continue to be studied, the clinical laboratory should be vigilant in testing for them in pregnant individuals, according to Dr. Reinhart. This, in turn, will enable clinicians to educate these patients about the potential harm that cannabis can do to a fetus.
Dr. Hunter Miller along with colleagues from the University of Louisville and a researcher from ARUP Institute for Clinical and Experimental Pathology, evaluated whether machine learning models could predict postoperative hydrocodone use duration in patients who had undergone orthopedic surgery. They developed two different models — a fast and frugal tree (FFTree) and a second that used an extreme gradient boosting approach (xgBoost) — that incorporated patient demographics, genetic test results, concurrently prescribed medications, and other clinical laboratory test results.
The researchers evaluated the two models by using them to predict the duration of hydrocodone use for 79 patients for whom they already had hydrocodone use duration information. Both models demonstrated good to excellent performance when classifying patients as either “short” or “long” duration users. Specifically, the FFTree model classified patients with 0.80 sensitivity and 0.76 specificity, while the xgBoost model achieved a sensitivity and specificity of 0.87 and 0.63, respectively.
