By Dr. Benjamin Caplan, MD | Board-Certified Family Physician, CMO at CED Clinic | Evidence Watch
A large national electronic health records study of nearly 145,000 gastrointestinal cancer patients found that only 2.4% received an FDA-approved cannabinoid prescription within 90 days of starting chemotherapy, and those who did were more likely to also be on opioids. However, this likely reflects the fact that sicker patients receive both medications, not that cannabinoids increase opioid use, and the study cannot determine whether broader cannabis use might reduce opioid needs in cancer care.
Cannabis Prescriptions Did Not Lower Opioid Use in GI Cancer Patients Starting Chemo, Large EHR Study Finds
A national database of nearly 145,000 patients shows cannabinoid prescribing is rare, regionally variable, and associated with higher, not lower, opioid use, likely due to confounding by symptom severity rather than a true pharmacological effect of cannabinoids on opioid demand.
#72
Strong Clinical Relevance
One of the largest real-world characterizations of cannabinoid prescribing in oncology, with direct implications for how clinicians interpret the cannabis-opioid relationship in cancer care.
Opioid-Sparing Strategies
GI Cancers
Real-World Evidence
Health Disparities
Interest in cannabis as an opioid-sparing tool in cancer care has intensified as medical cannabis laws expand across the United States, yet the field lacks large-scale data on who actually receives cannabinoid prescriptions and whether those prescriptions track with reduced opioid use. Gastrointestinal cancers carry substantial symptom burdens including pain, nausea, and cachexia, making them a natural testing ground for this question. Understanding the real-world landscape of cannabinoid prescribing in this population is a prerequisite for designing interventional trials that could generate causal evidence, and the demographic and geographic disparities this study reveals raise urgent equity questions about access to multimodal symptom management in oncology.
Cannabinoids, particularly dronabinol, nabilone, and pharmaceutical-grade cannabidiol, have FDA approval for specific indications including chemotherapy-induced nausea and vomiting, but their role in broader cancer symptom management remains poorly characterized in routine clinical practice. This retrospective cohort study drew on the Epic COSMOS electronic health records network, a large, multisite, de-identified US database, to examine prescribing patterns of FDA-approved cannabinoids among adults with gastrointestinal cancers who initiated chemotherapy between January 2016 and September 2025. The investigators sought to determine how frequently cannabinoids are prescribed in this setting, which patient and regional factors predict prescribing, and whether early cannabinoid receipt is associated with lower opioid co-prescribing within 90 days of chemotherapy initiation.
Among 144,981 patients, only 3,390 (2.4%) received an FDA-approved cannabinoid prescription. Those who did had a substantially higher rate of opioid co-prescribing (60.6% versus 31.1%, p less than 0.01), a finding that runs counter to the opioid-sparing hypothesis but is most plausibly explained by confounding by indication: patients who receive cannabinoids are those with greater symptom severity, particularly pancreatic cancer patients (adjusted odds ratio 3.26 for cannabinoid receipt), who also have the highest opioid needs. Prescribing rates varied substantially by race (Black patients had 45% higher adjusted odds), sex, census region, and cancer type. Critically, the study captured only pharmacy-dispensed, FDA-approved cannabinoids and could not identify the far larger pool of patients using dispensary-obtained or self-directed cannabis products. The authors appropriately emphasize that the observational design precludes causal inference and that prospective trials with broader cannabis exposure definitions are needed.
This study does something genuinely useful: it holds up a mirror to prescribing behavior across a massive real-world dataset and shows us that FDA-approved cannabinoid prescribing in GI oncology is vanishingly rare. That 2.4% figure alone is noteworthy. But the headline finding that cannabinoid recipients use more opioids tells us far more about who gets cannabinoids than about what cannabinoids do. These are the sickest patients with the most intractable symptoms, and their higher opioid use is entirely predictable. The study simply was not designed to answer whether cannabis reduces opioid needs, and its narrow exposure definition misses the vast majority of real-world cannabis use by cancer patients.
In my own practice, I work with many cancer patients who use cannabis products obtained outside the FDA-approved pipeline, and the conversations I have with them are far more nuanced than any binary prescription variable can capture. When I counsel patients on cannabis for symptom management during chemotherapy, I am thinking about formulation, timing, route, and how it fits alongside their existing antiemetic and analgesic regimen. This study reinforces my conviction that we need prospective trials that capture actual cannabis exposure, not just dronabinol scripts, and that we should be deeply cautious about reading observational prescribing data as evidence for or against opioid-sparing effects.
For oncologists and palliative care clinicians, this study sits at the descriptive, hypothesis-generating stage of the research arc. It characterizes current practice without testing an intervention, which is a necessary step but not one that should change clinical behavior directly. The demographic and regional disparities it reveals are clinically actionable in their own right: if cannabinoid prescribing varies by race, sex, and geography independent of disease characteristics, that suggests access barriers or implicit biases that warrant institutional audit. The finding that prescribing rates declined over time relative to 2017 is also intriguing and may reflect evolving practice norms, though the reasons remain speculative.
