Diabetes Lowers Survival Across Many Cancer Types, Study Finds

For millions of people living with Type 2 diabetes, a cancer diagnosis brings a double burden—and a new study from Finland suggests that the burden is not shared equally. Researchers analyzing a nationwide cohort of nearly 50,000 cancer patients have found that preexisting diabetes substantially worsens cancer-specific survival for some tumor types, has little effect on others, and may even improve survival in a surprising few. The findings, published in Cancer Reports, offer one of the most detailed pictures yet of how metabolic disease reshapes cancer prognosis across the body, and they challenge the assumption that diabetes uniformly makes every cancer worse.

The study drew on the Finnish CARING Project, a population-based effort that identified all individuals in Finland who purchased insulin or oral antidiabetic medications with reimbursement between 1997 and 2010, along with a reference group matched one-to-one by sex, age, and hospital district. From this foundation, the researchers assembled every incident cancer case in the two groups: 25,899 cancers among patients with diabetes and 23,900 among those without, for a total of 49,799 cases. The mean age at diagnosis was 72.2 years, and 40.9 percent of the cohort were women. Cancer diagnoses came from the Finnish Cancer Registry, a national resource with approximately 97 percent completeness for solid tumors, while cause-of-death information was obtained from Statistics Finland. Follow-up began at cancer diagnosis and ended at death or the end of 2017, whichever came first.

Over a mean follow-up of 3.8 years—188,329 person-years in total—the researchers recorded 13,602 cancer deaths. Their primary outcome was death specifically attributable to the cancer diagnosed at baseline, excluding deaths from other causes. To estimate the effect of diabetes, they used Poisson regression models adjusted for age, sex, calendar year, time since diagnosis, socioeconomic group, and hospital district, with follow-up time as an offset term. Rather than reporting only relative risks, the team used g-computation to derive both mortality rate ratios (MRRs) and mortality rate differences—measures that capture not just whether diabetes changes the risk of cancer death but by how much in absolute terms, a distinction that matters enormously for patients and clinicians weighing prognostic information.

The headline result was a clear gradient of harm across the 13 most common cancer sites. Mortality rate ratios were substantially elevated—above 1.1—for cancers of the stomach, endometrium, and the lympho-hematological system. Endometrial cancer showed the strongest adverse association, with an adjusted MRR of 1.25 (95 percent confidence interval 1.08 to 1.47), meaning diabetic patients faced roughly a 25 percent higher rate of cancer death than their nondiabetic counterparts. Lympho-hematological malignancies followed at 1.19 (1.03 to 1.40), and stomach cancer at 1.12 (1.01 to 1.25). Colon cancer (MRR 1.08) and breast cancer (MRR 1.09) were also modestly but significantly worse among patients with diabetes. The pattern held when the researchers examined Kaplan–Meier survival curves for the whole cohort and for each individual cancer type.

Then came the surprises. For two cancer sites—liver and rectum—patients with diabetes actually fared better. Liver cancer survival showed an MRR of 0.83 (0.74 to 0.94) and rectal cancer an MRR of 0.88 (0.80 to 0.98), both statistically significant advantages for the diabetic group. For the remaining six sites—pancreas, lung, prostate, bladder, kidney, and nonmelanoma skin cancer—the confidence intervals straddled unity, indicating no material difference in cancer-specific survival either way. The researchers also tested for statistical interaction between diabetes and sex and between diabetes and age, finding nothing of consequence, which suggests the pattern is broadly consistent across demographic subgroups.

What could explain such heterogeneity? The authors point to well-characterized metabolic biology. Chronic hyperglycemia and hyperinsulinemia, the hallmarks of Type 2 diabetes, alter the profile of growth factors implicated in tumor progression, including VEGF, EGF, IGF-1, FGF-21, and TGF-β1. These signals converge heavily on the PI3K/AKT/mTOR pathway, a central regulator of cell-cycle progression, apoptosis, and glucose metabolism. In the setting of obesity and insulin resistance, sustained nutrient excess drives chronic mTOR activation, which intensifies insulin resistance while simultaneously creating a metabolic environment permissive to unchecked cellular proliferation, enhanced tumor survival, and altered angiogenesis. Because these pathways operate in many tissues, their downstream consequences for an established tumor can vary dramatically depending on the tumor’s biology, its hormone dependence, and the local microenvironment—plausibly accounting for why endometrial tissue, which is exquisitely insulin-sensitive, shows the strongest diabetes penalty while other tissues show none.

The comparison with prior literature is instructive. A landmark Swedish register study by Liu and colleagues, covering 16,123 cancer patients with Type 2 diabetes followed for a mean of seven years, found significantly reduced cancer-specific survival for 16 of 26 cancer types examined—and, crucially, never found improved survival for any site. The Finnish results are less uniformly grim. The authors attribute this partly to differences in how diabetes was ascertained: the Swedish study captured patients hospitalized with diabetes, likely a more severely affected population, and its data spanned 1961 to 2008, an era before modern diagnostic criteria (including HbA1c ≥ 48 mmol/mol as a diagnostic threshold) and before metformin became the near-universal first-line therapy. Metformin, in particular, has attracted intense interest for potential anti-cancer effects, and today’s Finnish diabetic patients—over 90 percent of whom in this cohort used only oral medications—are likely metabolically healthier than hospitalized diabetic patients of earlier decades.

The liver cancer finding, one of the most counterintuitive results, may have a surveillance explanation. Metabolic dysfunction-associated steatotic liver disease is common in diabetes, which means the livers of diabetic patients are monitored more closely. Early detection of hepatocellular carcinoma in a monitored population could translate into diagnosis at more treatable stages and better cancer-specific survival, even as overall mortality appears worse in other datasets. The authors also note that most previous studies reporting poor liver cancer survival in diabetes were conducted in high-risk Southeast Asian populations diagnosed in the 1980s and 1990s, under far less intensive follow-up than contemporary Finnish healthcare provides.

For endometrial cancer and lympho-hematological malignancies, the Finnish results align closely with prior evidence. Meta-analyses have consistently shown reduced survival among endometrial cancer patients with diabetes, and both non-Hodgkin lymphoma and chronic lymphocytic leukemia—two diagnoses that together account for the majority of lympho-hematological cancers—have previously been linked to inferior outcomes in diabetic patients. The breast cancer finding dovetails with a large Mendelian randomization study restricted to European-ancestry individuals, which likewise found reduced breast cancer survival associated with diabetes, whereas a smaller, more ethnically diverse US cohort found no association. The Finnish population, like the Mendelian randomization study, is relatively homogeneous and treated within a European healthcare system, which may explain the concordance.

The study’s strengths are considerable: its population-based design eliminates much selection bias, its registry linkage provides essentially complete follow-up, and its sheer scale—with more than 700 cancer cases in the diabetes group at each of the 13 sites, and over 1,000 at seven sites—dwarfs most prior analyses, some of which rested on datasets of similar or smaller size than this single cohort. Limitations remain, however. The registers lacked clinical prognostic factors such as tumor stage, grade, treatment, and comorbidities, though the authors argue these may function as mediators rather than confounders in the causal chain from diabetes to cancer death. Detailed diabetes information—glucose levels, HbA1c values, medication types—was unavailable, the register data could not cleanly separate Type 1 from Type 2 diabetes (though the age distribution and overwhelming predominance of oral medication use point strongly to Type 2), and some reference-group members may have had undiagnosed diabetes. The ethnic homogeneity of Finland may also limit generalizability elsewhere.

The clinical implications are nonetheless significant. Diabetes materially worsens prognosis for several major cancers, and there is evidence that poor glycemic control independently predicts worse cancer outcomes, suggesting that tighter metabolic management after a cancer diagnosis could reduce excess mortality. Some diabetes medications, particularly metformin, have been associated with lower cancer mortality and remain under active investigation. The authors call for future research into the biological and metabolic factors that drive cancer survival in diabetic patients—and into how treatment strategies for cancer can be optimized for this large and growing patient group. In the United Kingdom, cancer has already overtaken cardiovascular disease as the leading cause of death among people with diabetes, and demographic projections suggest the intersection of these two diseases will only grow. As this Finnish study makes clear, the intersection is not a single story but thirteen different ones, each written in the language of metabolism.