Cancer: The Unasked Questions

A rise in recorded incidence can come from earlier detection. Screening and imaging find small tumors that would otherwise have gone undiagnosed, and the authors list increased detection among the possible drivers. Detection of that kind raises incidence without raising the death rate. For female breast and kidney cancer, incidence rose and early-onset mortality did not. Rising mortality from colorectal and uterine cancer in people under fifty cannot be produced by earlier diagnosis. At those sites more young people are developing the disease and dying of it.

Eighteen cancer types had significantly falling incidence in at least one early-onset age group, so the increases are specific to particular organs. The databases the NCI team used held no risk-factor data, so the study could not identify causes.

The rise appears in countries with different health systems and diets, and in 14 of the 27 it is confined to younger adults. Twenty-three of the 50 registries cover only part of their country, which limits national comparisons.

The colibactin result is an association between a mutational signature and age at diagnosis. Alexandrov described it that way and said proving causation would be very complicated. It does not show how many early-onset cancers colibactin causes, or why exposure to these bacteria would have increased in children born after the mid-twentieth century.

The causes still under investigation include obesity, described in coverage of the NCI study as the leading hypothesis, and exposures in early life acting through the gut microbiome. Changes in antibiotic use and diet since the 1950s fit a birth-cohort pattern, in which risk set in childhood appears decades later as cancer. None of these has been established as a cause.

Answering the question requires a study that links exposures measured in childhood, such as stool samples, antibiotic prescriptions, diet and weight, to cancer diagnosed decades later. The national incidence data held by the CDC and the National Cancer Institute record diagnoses and deaths. They contain no individual exposure history.

Five-year survival for childhood cancer rose from 63 to 85 percent in the same study. Falling mortality reflects improved treatment and does not measure how many children develop cancer, which only the incidence data record.

Part of the rise may reflect better diagnosis, especially for brain tumors after magnetic resonance imaging came into use in the 1980s. The registry counts diagnoses and does not separate detection from a real change in the number of children who develop cancer. SEER holds the incidence data from 1975, and the 2025 analysis ends with diagnoses in 2019.

Chapter 5 records the regulatory history: EPA's most recent cancer evaluation of glyphosate was vacated by the Ninth Circuit in 2022 and has not been replaced (Finding 5.6). Chapter 4 records the safety reviews written in part by Monsanto staff (Findings 4.5 and 4.6). This section concerns the epidemiology that a new evaluation would have to weigh.

The meta-analysis and the cohort reached different results from overlapping data. The meta-analysis pooled the 2018 cohort update with case-control studies of lymphoma patients. The cohort authors described their acute myeloid leukemia result as one that "requires confirmation." Critics of the cohort point to its imputation of missing exposure data for participants who did not complete follow-up, and to a reference group that included low-exposure users. Lianne Sheppard, a co-author of the meta-analysis, served on EPA's 2016 scientific advisory panel on glyphosate.

The lymphoma signal in the human studies is concentrated in applicators with the heaviest use. The litigation that Chapter 5 describes concerned heavy users of that kind. The new EPA–HHS evaluation is the body of work that would weigh these studies against each other, and until it is published the federal government has no completed cancer evaluation of glyphosate that has survived judicial review.

The acceptable daily intake is the dose that regulators set as safe for a person to consume every day for a lifetime. The results have not been independently replicated.

The authors wrote that their results support IARC's conclusion of sufficient evidence of carcinogenicity in experimental animals. A replication under regulatory guideline conditions, with exposure from gestation and the same formulations, would test the result.

The concentrations were low. The highest measured value was 8.13 µg/L, and the industry response noted that this corresponds to less than 0.14 percent of EPA's reference dose. The epidemiological signal in 8.5 comes from applicators with occupational exposure far above these levels. The Ramazzini study in 8.6 tested doses at the European acceptable daily intake. The NHANES analysis reports exposure and no cancer outcomes.

According to MDLinx, which cites NCI and CDC State Cancer Profiles data for 2018–2022, Kentucky and Iowa have the highest age-adjusted cancer incidence of any US state, at 519.0 and 498.8 per 100,000, against 448.6 nationally. The Gazette reports that the Iowa Cancer Registry's 2026 report describes Iowa's rate as the fastest rising in the country. The Daily Bulletin and Review reported a study finding that Iowa farmers had 13 percent fewer cancers than the general Iowa population between 1994 and 2015, and farmers' spouses 10 percent fewer. The article credits lower smoking and drinking among farmers and the healthy-worker effect for the farmer gap. An Iowa researcher quoted in it said, "it isn't one thing here that's causing all the cancer. It's multiple things."

Chapter 7 records how SV40 entered the polio vaccine, the decision in 1961 to screen new lots without recalling existing stock (Finding 7.2), the CDC estimate that 10 to 30 million Americans may have received a contaminated dose (Finding 7.3), and the Institute of Medicine's 2002 conclusion (Finding 7.4). This section concerns why the cancer question remains open.

The laboratory and population studies reached different results. Laboratory studies repeatedly reported SV40 sequences in particular tumors: mesothelioma, non-Hodgkin lymphoma, brain and bone cancers. The López-Ríos study showed that some of those detections came from contamination by plasmids used in the same laboratories. At a September 2003 House hearing, James Goedert of the National Cancer Institute testified that in a nine-laboratory validation study, several laboratories that had previously detected SV40 were unable to do so. The plasmid study examined mesotheliomas only.

The population studies, led by Strickler and colleagues in 1998, found no excess cancer in birth cohorts likely to have received contaminated vaccine. They had no record of which person received which lot, and contamination varied from lot to lot. The Institute of Medicine judged them too flawed to settle causation for that reason.

Two studies would settle the question. One would link individual vaccination records, by lot, to later cancer diagnoses. The other would test tumors from exposed and unexposed people with assays designed to exclude plasmid contamination, across the cancers where SV40 was reported. As Chapter 7 records, CDC's statement that no vaccine used today contains SV40 virus is correct.

Claims of unusually aggressive cancer after COVID-19 vaccination, called "turbo cancer," rest on clinician reports and case reports. A 2025 cohort of 8.4 million Seoul residents reported higher one-year cancer risk after vaccination, but the gap was already present one month after the dose, a timing that points to detection or selection bias. A nationwide interrupted time-series study of 36,024,394 Koreans, published in October 2026, found no detectable change in cancer incidence trends in the 50 months after the rollout. Fifty months is short for solid tumors, which can take years to appear. Long-term effects have not been studied, and national cancer registries hold the data to study them.

Each question in this chapter has a study that would answer it, and in most cases the data to begin it already exist. The table sets out what was found, the study that would settle each question, and who holds the data.

Four of these questions concern two products that federal agencies approved or recommended: glyphosate, registered by EPA, and polio vaccine contaminated with SV40, released under federal licensure from 1955 to 1963. EPA's evaluation of glyphosate is now expected in early 2027, five years after the last one was vacated. The Institute of Medicine left the SV40 question open in 2002.