Another unusual “side effect” has been observed with semaglutide — the drug better known as Ozempic and Wegovy. Originally developed for type 2 diabetes, it proved effective for weight reduction, and now appears to benefit treatment of certain cancers. A University of California San Diego study reports that among people with colon cancer who were concurrently taking semaglutide, five-year mortality was 15.5%, whereas among those not taking it the figure was 37.1% — more than twice as high.
These conclusions come from a new UC San Diego analysis published in Cancer Investigation, which used the University of California Health system’s Data Warehouse to follow 6,871 patients with primary colon cancer.
This is not the first paper on the topic. In 2024, a large study in JAMA Network Open showed that, compared with patients treated with insulin, those receiving GLP-1 receptor agonists had lower incidence of several obesity-associated cancers, including colorectal cancer. GLP-1 receptor agonists are medicines that mimic the action of the eponymous hormone: produced in the intestine, it helps the pancreas secrete insulin, lowers blood glucose, and influences appetite and body weight. Put simply, the statistics suggest that the GLP-1 system in the body is in some way linked to tumor development.
Not a “panacea,” but promising
Indeed, a 2025 review in JCI summarized meta-analytic data and found that people taking GLP-1 receptor agonists had a lower risk of colorectal cancer than those receiving other classes of antidiabetic drugs. In other words, at the level of large datasets, it looks as though these medicines may truly reduce the likelihood of intestinal cancer.
But the authors emphasize:
– the relationship between GLP-1 and cancer depends on the specific tumor type;
– simply measuring how many GLP-1 receptors a tumor expresses is not sufficient to predict outcomes in colon cancer — there is no straightforward linear association.
Laboratory (basic and translational) studies outline plausible mechanisms: GLP-1 signaling may dampen chronic inflammation, reprogram tumor metabolism, and disrupt growth-signaling pathways in colorectal cancer models. In experiments using colon cancer cell lines and mouse models, activating the GLP-1 receptor directly inhibited tumor growth.
Should semaglutide and related agents already be regarded as anticancer drugs? No. This study is observational – investigators examined outcomes among people who were already receiving different treatments. Designs like this always carry risks of bias — patients on GLP-1 therapy may have differed from others at baseline, e.g., body weight, comorbidities, access to care, and part of the apparent survival benefit could reflect weight loss and improved metabolic health rather than a direct antitumor effect.
The authors attempted to account for these issues — adjusting analyses for disease severity and other covariates. Still, they note that the strongest effect appears in people with severe obesity, where the metabolic benefits of GLP-1 therapy are greatest. GLP-1 receptor agonists remain medicines for metabolic disorders (diabetes, obesity) that may confer an “oncologic bonus.” Only randomized clinical trials can determine whether they exert a direct antitumor effect.
A side effect — on the positive side
This is not the first time when drugs designed for other purposes have found a place in anticancer therapy. More than a century ago, in 1921, the tuberculosis vaccine (BCG) entered clinical use. Seven decades later, the very same medication received FDA approval — not as a vaccine, but as a treatment for bladder cancer.
A randomized Southwest Oncology Group trial enrolled patients at high risk of non–muscle-invasive bladder cancer and assigned them to intravesical BCG therapy or no BCG. In those who received BCG, the median time to tumor recurrence was roughly twice as long. Subsequent reviews and clinical guidelines established maintenance BCG as the standard of care for patients with intermediate- and high-risk disease.
How this works at the biological level remains incompletely understood. The leading concept is so-called trained immunity: BCG “retunes” the innate immune system via epigenetic changes, prompting a more forceful response to tumor cells in the bladder wall. The official U.S. prescribing information for TICE BCG still notes that the precise mechanism has not been fully defined. But medicine sometimes proceeds this way: when well-designed clinical trials convincingly demonstrate benefit, a therapy can enter routine practice before science has fully mapped the exact mechanisms.
When a chemical weapon became a medicine
The story of how sulfur mustard became a chemotherapy agent began with physicians’ observations during World War I: in victims of sulfur mustard poisoning, leukocyte counts plummeted and the bone marrow was profoundly suppressed. Later, chemists and clinicians recognized that nitrogen mustards — related compounds selectively damage rapidly dividing hematopoietic cells.
A tragic episode in the port of Bari in 1943 provided clinical impetus: survivors of the explosion of a ship carrying a secret cargo of mustard showed marked leukopenia and destruction of lymphoid tissue. Pathologist Stewart Alexander linked these injuries specifically to mustard, and his findings confirmed that the agent profoundly inhibits proliferation of immune cells.
In parallel at Yale School of Medicine, the team of Goodman and Gilman, working under a classified OSRD contract, was already studying nitrogen mustards to treat lymphosarcoma. The first documented patient with disseminated lymphoma received experimental infusions of a compound designated “substance X,” and the tumor did shrink — transiently, but substantially.
This led to the development of mechlorethamine, a nitrogen mustard derivative adapted for intravenous use. On March 15, 1949, it became the first FDA-approved anticancer drug.
The emblem of a pharmaceutical blunder became a cornerstone of multiple myeloma treatment
Another example is thalidomide. First marketed in the late 1950s as a “gentle” sedative and antiemetic, the drug caused catastrophic congenital malformations in more than 10,000 children whose mothers took thalidomide during pregnancy. It was urgently withdrawn, and its story became a central driver for tightening drug-approval rules and rethinking medication safety in pregnancy.
The story did not end there. In 1999, a NEJM study showed that in patients with refractory multiple myeloma — a malignancy of bone-marrow plasma cells no longer responsive to standard therapy — thalidomide produced clinically meaningful responses: tumors regressed and a subset of patients achieved remission. A drug once synonymous with a pharmacological catastrophe unexpectedly demonstrated potent antitumor activity and opened the way to a new class of immunomodulatory agents, the so-called IMiDs.
The next step was lenalidomide — an “improved relative” of thalidomide. In phase III randomized trials it outperformed dexamethasone alone, providing longer disease control and higher response rates, becoming a key component of myeloma regimens and entering international standards. Subsequent studies of earlier intervention showed that, in patients at high risk of progression from smoldering myeloma (a pre-symptomatic stage), lenalidomide delayed evolution to active disease and improved survival.
As a result, thalidomide and its derivatives traveled the arc from a “nightmare drug” that devastated thousands of families to an important tool in the treatment of hematologic malignancies.
AI — a powerful engine for repurposing
The space of possible “old drug–new disease” combinations is far too large for clinicians and scientists to sort through by hand. Hence the turn to AI. Large-scale algorithms stitch together data on drugs, their targets, biochemical pathways, and diseases to construct vast networks of relationships. Machine learning then helps prioritize which drug–disease pairs look promising enough to warrant time and resources in the lab and clinic. Increasingly, these models incorporate not only molecular and chemical descriptors but also real-world clinical records.
In oncology, such platforms analyze tumor gene-expression changes, protein–protein interaction networks, and off-target binding profiles. Rigorous reviews in Frontiers in Oncology describe available toolkits and their limitations. The output is a short list of candidates to test first in preclinical models and, if supportive, in clinical trials. There are already examples in which algorithm-nominated agents have shown encouraging effects in cancer models. AI does not “discover a cure for cancer” on its own — it markedly accelerates the triage of weak ideas and the selection of plausible hypotheses, while definitive confirmation still happens in clinical settings.
The GLP-1/colon cancer story currently sits at exactly this intermediate stage. Survival signals look convincing enough to justify serious, well-designed trials, but clearly insufficient to change standards of care today. If large studies confirm that GLP-1 receptor agonists improve survival (and not solely via weight loss), oncology would gain a new adjunct — a “metabolically tuned” therapy. If not, the field will at least gain a clearer view of how metabolism shapes cancer behavior.
