For years, the debate over cannabis and cancer has burned on—a haze of anecdotes, conflicting studies, and a stubborn federal classification that still ranks marijuana as dangerous and without medical use. But a new study, the largest of its kind, cuts through the smoke with surprising clarity.

The analysis reviewed over 10,000 studies, finding “a shocking degree of consensus” that cannabis not only helps manage cancer symptoms like nausea and appetite loss, but may also fight the disease itself.

“It wasn’t 55-45, it was 75-25,” said Ryan Castle, research director at the Whole Health Oncology Institute and lead author of the study. “That’s a shocking degree of consensus in public health research, and certainly more than we were anticipating for a topic as controversial as medical cannabis.”

A Data-Driven Reassessment of Medical Cannabis

Medical cannabis research has been hampered by two problems: patchy data and politics. Because cannabis remains a Schedule I drug in the U.S., it’s been difficult for researchers to study it with the rigor they would apply to conventional medicine.

The new meta-analysis, led by Castle and colleagues at the Whole Health Oncology Institute, used computational sentiment analysis to bypass the traditional bottlenecks. By scanning the language of thousands of scientific articles, the algorithm could detect trends in how cannabis is discussed—whether it’s viewed positively, negatively, or ambiguously in relation to cancer.

It’s not a perfect method. Scientific writing is famously cautious, and machine-learning tools can struggle to interpret nuance. But by using two separate measures—keyword frequency and dominant sentiment—the researchers created a double-layered filter. Both pointed in the same direction.

“The aggregated correlation strength of cannabis across all cancer topics indicates that support for medical cannabis is 31.38 times stronger than opposition to it,” the authors wrote in their study.

The analysis broke the literature down into three broad categories: health metrics (like inflammation), cancer treatments (like chemotherapy), and cancer dynamics (like tumor growth or remission). In each, support for cannabis outpaced skepticism by a wide margin.

One of the strongest signals came from studies examining pain and nausea—two of the most common side effects of cancer treatment. Cannabis was found to be more than twice as likely to be associated with positive sentiment than negative or unclear sentiment. When it came to chemotherapy, studies were 134% more likely to support cannabis than oppose it.

Even more striking was the category of “therapeutic” use—how cannabis performs in a broader healing context. Here, support was particularly strong. Preclinical studies suggest cannabinoids can inhibit tumor growth, promote cancer cell death, and reduce metastasis. The meta-analysis found that support for cannabis as an “anticarcinogenic” agent significantly outweighed skepticism.

Of course, cell culture studies and mouse models aren’t the same as clinical trials. But the authors argue that the volume and consistency of the findings—even when controlling for study design—signal a direction worth following.

Essentially, three out of four studies supported cannabis in a positive light, both as a palliative aid and as a possible anticancer agent.

“Studies investigating therapeutic cannabis use are overwhelmingly likely to present supported sentiments,” the authors noted.

“Our goal was to determine the scientific consensus on the topic of medical cannabis, a field that has long been dominated by a war between cherrypicked studies,” Castle added in an interview with The Guardian.

A Path Forward, If Politics Allows

In the U.S., cannabis remains a Schedule I drug, alongside heroin and LSD. That label has long throttled scientific research, requiring federal approvals so burdensome that many researchers simply walk away.

Castle hopes this new consensus will help break that deadlock.

“We wanted to analyze not just a handful, but nearly every major medical cannabis study to find the actual points of scientific agreement,” he said.

As more jurisdictions around the world legalize medical cannabis, physicians and regulators are being asked to make decisions based on evidence. Until recently, that evidence has been patchy, fragmented, or contradictory.

By using machine learning, the new study attempts to make sense of an otherwise incoherent body of literature. The technique doesn’t replace randomized controlled trials, but it does offer a way to quantify scientific consensus across a vast data landscape.

The study itself acknowledges that sentiment analysis—while powerful—can misread scientific caution as neutrality. The authors also point out that cannabis formulations vary widely in potency and chemical composition, making direct comparisons across studies difficult.

Moreover, the authors stress that cannabis is not a one-size-fits-all treatment. Different cancers, different stages, and different patient responses all influence outcomes. The efficacy of cannabinoids also depends on delivery method, dosage, and the interplay between compounds like THC and CBD.

Still, the strength of the signal is hard to ignore.

Medical cannabis has always walked a tightrope between science and stigma. Now, with tools that can parse thousands of studies at once, the weight of evidence may finally tip the balance.

The findings appeared in the journal Frontiers in Oncology.