By Dr. Benjamin Caplan, MD | Board-Certified Family Physician, CMO at CED Clinic | Evidence Watch
A nationwide study of nearly 145,000 gastrointestinal cancer patients found that only 2.4% received an FDA-approved cannabinoid prescription after starting chemotherapy, and those who did were actually more likely to also receive opioids. This pattern almost certainly reflects sicker patients receiving more medications overall, not a failure of cannabinoids to help with pain, highlighting the limitations of observational data for answering whether cannabis can reduce opioid reliance in cancer care.
Cannabis Prescriptions Did Not Lower Opioid Use in GI Cancer Patients Starting Chemotherapy
A large EHR study of 144,981 gastrointestinal cancer patients in the Epic COSMOS database reveals that FDA-approved cannabinoid prescribing is uncommon after chemotherapy initiation, and cannabinoid recipients used more opioids than nonrecipients, a finding most plausibly driven by confounding by indication in sicker patients rather than any harmful pharmacological effect of cannabinoids themselves.
#72
Strong Clinical Relevance
The largest real-world description of cannabinoid prescribing patterns in GI oncology, though causal conclusions about opioid-sparing effects remain out of reach.
Opioid Prescribing
GI Oncology
Supportive Cancer Care
Real-World Evidence
Gastrointestinal cancers carry some of the highest symptom burdens in oncology, with chemotherapy-related nausea, cachexia, and cancer-related pain driving significant opioid utilization. Whether cannabinoids can meaningfully reduce opioid dependence in these patients remains one of the most clinically pressing and politically charged questions in supportive oncology. Understanding how often oncologists actually prescribe FDA-approved cannabinoids, and to whom, is an essential first step before trials can be designed to test opioid-sparing hypotheses. This study provides the largest window yet into those prescribing patterns and the stark disparities that shape them.
The opioid crisis has intensified interest in alternatives for cancer pain management, and cannabinoids have attracted particular attention as potential opioid-sparing agents. Preclinical work and small clinical series suggest that cannabinoids may reduce nausea, stimulate appetite, and provide analgesic effects through CB1 and CB2 receptor modulation, offering a plausible mechanism for reducing opioid reliance. This retrospective cohort study used the Epic COSMOS de-identified electronic health record database, which aggregates data from hundreds of US health systems, to examine FDA-approved cannabinoid prescribing (dronabinol, nabilone, or cannabidiol) within 90 days of chemotherapy initiation in 144,981 adults diagnosed with GI cancers between January 2016 and September 2025.
Only 3,390 patients (2.4%) received a cannabinoid prescription meeting the study’s threshold of more than two documented days. Cannabinoid recipients were far more likely to also receive opioids (60.6% versus 31.1%, p less than 0.01). In adjusted models, Black patients (adjusted odds ratio 1.45, 95% CI 1.30 to 1.61), women (aOR 1.11), and patients with pancreatic cancer (aOR 3.26) were significantly more likely to receive cannabinoids. Prescribing varied by region and declined over time after 2017. The authors acknowledge that confounding by indication is the most likely explanation for the higher opioid use among cannabinoid recipients, since sicker patients with greater symptom burden are more likely to receive both classes of medication. They note that the study captures only EHR-documented FDA-approved agents, completely missing dispensary and recreational cannabis, which represent how the majority of cancer patients actually access cannabinoids. Prospective trials with validated cannabis exposure measurement are needed to determine whether cannabinoids genuinely reduce opioid utilization.
This study does something valuable by quantifying just how rarely FDA-approved cannabinoids show up in oncology EHRs, and the demographic patterns it uncovers deserve serious attention. But the finding that cannabinoid recipients use more opioids is almost certainly telling us about patient severity, not about cannabinoid failure. Patients prescribed dronabinol or nabilone alongside chemotherapy for pancreatic cancer are, by and large, patients who are suffering more. This is classic confounding by indication, and the study design simply cannot untangle it. The bigger blind spot is that this analysis only captures the tiny fraction of cannabis use that flows through FDA-approved prescriptions. In my experience, most cancer patients who use cannabis are obtaining it from dispensaries or other sources that leave no trace in Epic.
In practice, I routinely discuss cannabis with my oncology patients, and many are already using it before I bring it up. What I see clinically is that thoughtful cannabinoid integration, using specific formulations and careful titration, can sometimes allow a meaningful reduction in opioid doses for patients whose primary symptoms are nausea, poor appetite, or diffuse discomfort rather than acute somatic pain. But I never frame cannabis as an opioid replacement. I frame it as one tool in a broader analgesic and palliative strategy, and this study reinforces why that distinction matters.
This study sits squarely in the descriptive phase of the research arc on cannabinoids in oncology supportive care. It establishes the baseline prevalence and sociodemographic predictors of FDA-approved cannabinoid prescribing in a large, nationally representative cohort. These data are useful for informing trial design, identifying enrollment disparities, and benchmarking institutional practice. However, clinicians should be careful not to interpret the opioid co-prescribing data as evidence that cannabinoids fail to reduce opioid use, because the study was not designed or powered to answer that question. The absence of an o
