The Biology of Vegetables and Cancer

Eating vegetables will prevent cancer. Scientific progress is beginning to dismantle this familiar myth. On the other hand, surprisingly, vegetables seem to help prevent dementia. Based on recent research, we will re-examine the effects of vegetables.

Chapter 1: Conclusions Change Based on "Research Methods"

There are many ways to investigate the relationship between vegetables and disease, but the reliability of the conclusions varies greatly depending on the method. Therefore, I will first explain research methods.
The first is the difference between "observational studies" and "interventional studies (randomized controlled trials)." Observational studies investigate and compare the health status of people who eat a lot of vegetables and those who do not. However, differences in lifestyle are significant; there are major gaps not only in vegetable intake but also in exercise, smoking, drinking, and stress. These are called "confounders," and their presence makes it difficult to accurately evaluate the "effect of vegetables."
To eliminate confounding, "human experiments" using people are necessary. This is an "interventional study," where subjects are divided into two groups, and an "intervention" is performed where only one group increases their vegetable intake. Because subjects are assigned by "chance," the variation in conditions other than vegetables becomes almost identical. Therefore, if a difference appears between the two groups, it can be said that the cause is the vegetables. This is the design that medicine trusts the most.
The second is the classification of observational studies, which includes "retrospective studies" and "prospective studies." Retrospective studies (case-control studies) gather people who have already become ill and healthy people and ask them to recall their past diet. While easy to conduct, the problem is how well they remember their past eating habits, and this "recall bias" distorts the results.
On the other hand, prospective studies (cohort studies) record the diets of healthy people and follow them for many years to see who becomes ill. Because "recall bias" is less likely to occur, the results are reliable.
With an understanding of these differences, let's look at each study. You will see the terms "odds ratio" and "hazard ratio," but please consider that if either is less than 1, the risk is reduced.

Chapter 2: Do They Prevent Cancer?

The EPIC study that followed 500,000 people

This is a massive prospective study that gathered 500,000 people across 10 European countries. It began in 1992, and by carefully interviewing healthy people about their diets and tracking those who developed cancer through 2000, researchers calculated the risk by statistically subtracting the effects of smoking, drinking, and other factors.
The results were disappointing. Even eating 200 grams more fruit and vegetables per day resulted in only a 3% reduction in the overall risk of cancer. Moreover, while the risk decreased for smoking-related cancers, there was almost no difference for cancers unrelated to smoking. This makes it impossible to tell whether "vegetables prevent cancer" or if "confounding from smokers simply remained." European countries even used TV commercials to advocate for "5 servings of vegetables a day," but the gap between expectation and reality was so large that experts criticized the campaign as being "disproportionate to such a small benefit."

What is "confounding"?

Confounding is the most difficult part of interpreting epidemiological studies (observational studies). The problem with the EPIC study is that there is a "third factor" related to both the factor being focused on (vegetables) and the result (cancer). That is smoking. Let's confirm these connections.
First, there is the relationship that the more people smoke, the less they eat vegetables. Smoking correlates with lifestyle habits such as drinking, lack of exercise, low income, and indifference to health, and is inversely correlated with vegetable intake. In other words, the "group that does not eat vegetables" is the "group with many smokers."
Second, smoking is a powerful cause of cancer, especially lung, laryngeal, and esophageal cancer. The group that did not eat vegetables had more cancer, but it is unclear whether that was due to not eating vegetables or due to smoking. This is the effect of a confounding factor. Therefore, it is necessary to perform statistical processing to adjust for the effects of smoking.

Is the adjustment perfect?

However, it is difficult to completely adjust for confounding. Adjustments are made using "rough" classifications such as "non-, former, and current smokers." However, the harm of smoking varies greatly depending on the number of cigarettes, years smoked, age of onset, years since quitting, and depth of inhalation. Even for the same "current smoker of 20 cigarettes," the impact of smoking is completely different between someone who has been inhaling deeply since their teens and someone who has been smoking lightly since their 30s.
However, it is impossible to investigate and adjust for hundreds of thousands of people in such detail. In addition, there are problems such as smoking being self-reported and often under-reported, and the fact that even after quitting, the risk lingers for decades, meaning "former smokers" include a variety of people, making the adjustment incomplete.
Thus, even if one tries to adjust for confounding, "residual confounding" remains. If it is small, the practical problem is small, but if it is large, it will significantly distort the results.

When residual confounding is large

The reason smoking is the greatest confounding factor is that its carcinogenic effect is extraordinarily large. The lung cancer risk for smokers is 10 to 20 times higher than for non-smokers. When the effect is this huge, a small bias in smoking alone can lead to a large difference in cancer rates.
For example, suppose the "current smokers" in the group that eats many vegetables smoke an average of 15 cigarettes/day, and in the group that eats few vegetables, they smoke an average of 20 cigarettes/day. This difference of 5 cigarettes has not been adjusted for. Since lung cancer risk increases steeply with the number of cigarettes, this residual of 5 cigarettes alone creates an "apparent vegetable effect" with a hazard ratio of around 0.9. Even if vegetables have no effect, it looks numerically like "vegetables reduce cancer by 10%." This is the effect of residual confounding.

Differences by "type" of cancer

The biggest problem with the EPIC study is the difference by type of cancer. There was an effect on smoking-related cancers (0.92 per 200g increase in vegetables), but almost no effect on cancers unrelated to smoking (0.98). If vegetables truly prevented cancer, they should reduce all cancers. This is exactly the pattern predicted by residual confounding.
People who eat vegetables smoke less, but if this adjustment is incomplete, the flow of "people who eat vegetables -> smoke less (insufficient adjustment) -> less cancer" remains. And the survey results look exactly like that. That is precisely why these results "cannot be definitively attributed to the influence of the vegetables themselves."

Vegetarians have less cancer

If you want to look at the relationship between vegetables and cancer, you should study vegetarians. Many people think so. The UK's EPIC-Oxford studied this, and as a result, vegetarians had 12% less total cancer, and in the North American Adventist study, the hazard ratio for total cancer was 0.88. At first glance, it looks like a victory for vegetarianism, but looking closely, it is not that simple.
Adventists have 30% less total cancer than the US average, but the difference due to vegetarianism compared to non-vegetarian Adventists was only 12%, and the rest was the effect of lifestyle habits such as not smoking or drinking for religious reasons. Moreover, when adjusted for BMI, the effect disappeared. The result is that it is not that "vegetables have an effect," but that there is less cancer because they are "less likely to be obese." Furthermore, there was no difference in mortality between the vegetarian and meat-eating groups.
Here too, the effect is not from vegetables, but can be considered to be the entire pattern of life and diet, such as quitting smoking, not drinking, low body weight, and avoiding meat.

Interventional studies of vitamin supplements—the most famous "tragedy" in history

So, what happens in interventional studies? The results of interventional studies that many people paid attention to became the most tragic episode in the history of nutrition.
"The vegetable pigment beta-carotene has antioxidant properties. People with higher blood concentrations have less lung cancer. If we make it into a pill and have them take it, we should be able to prevent lung cancer." This hypothesis, based on observational studies, was tested by two large-scale interventional studies in the US, CARET and Finland's ATBC.
The design of CARET is clear. Over 18,000 people at high risk for lung cancer due to smoking history or asbestos exposure were randomly divided into two groups, one given beta-carotene and vitamin A daily, and the other given a placebo, and followed. The results were the exact opposite of expectations. In the intervention group, lung cancer increased by 28%, and deaths increased by 17%. Seeing this, the study was stopped. In the ATBC study, which targeted 29,000 male smokers, beta-carotene also increased lung cancer. Moreover, even years after stopping the administration, the risk lingered.
Why did isolating and giving large amounts of a component that looked "effective" in observational studies turn out to be harmful instead? The effectiveness resides in "whole food as vegetables, low dosage, and multiple components," and it is considered to be different from pills with extracted components. It was also an incident that showed how fragile the intuition that "it's natural, so it's safe" is.
By lining these two up, I think you can understand the collapse of the myth that if you eat vegetables, you won't get cancer. Current science places the leading roles in cancer prevention not on vegetables themselves, but on avoiding obesity, moderating alcohol, quitting smoking, and dietary fiber.

Chapter 3: Do They Prevent Dementia?

Next is whether vegetables have a preventive effect on dementia. Unlike cancer, quite clear results have emerged regarding this.

A method for summarizing research called "meta-analysis"

Since individual studies vary in scale and subjects, meta-analysis is used to statistically bundle multiple studies to provide an overall picture. A 2017 meta-analysis that bundled 9 studies (31,000 subjects) found that for those with high fruit and vegetable intake, the odds ratio for dementia and similar conditions was 0.80. This means the risk for the group that eats plenty of fruits and vegetables is 20% lower.
In a larger 2022 analysis of 16 studies (64,000 subjects), vegetables alone showed a one-quarter lower risk (0.75), while the effect of fruit was weaker at 0.83, and the significant difference disappeared for Alzheimer's disease alone. It is interesting that vegetables and fruits, which are often grouped together in the context of cancer and heart disease, show different faces when it comes to the brain.

The Hisayama Study

Japan has a study that it can be proud of on the world stage: the survey in Hisayama Town, Fukuoka Prefecture. This is a prospective cohort study that followed 1,071 residents aged 60 and over without dementia for a long period of 24 years. The strength of this town lies in the extremely high rate of autopsies performed on those who passed away, meaning there are almost no misdiagnoses. The accuracy of the diagnosis gives the conclusion special weight.
The results showed that the group that ate the most vegetables had a 27% lower risk of dementia and a 31% lower risk of Alzheimer's disease compared to the group that ate the least. However, there was no difference in vascular dementia. Another large-scale study (JPHC, 43,000 people) also confirmed that people with higher dietary vitamin C intake have a lower risk of dementia requiring nursing care. However, no study has confirmed that vitamin C tablets prevent dementia.

Latest Research Paper

The 2026 paper by Yuan et al. (American Journal of Clinical Nutrition) reinforces this trend, bundling many studies to redraw the relationship between "vegetables/fruits and dementia." The results show that people who eat a healthy plant-based diet have a lower risk of dementia (hazard ratio of 0.71 for the highest group), and the effects were greatest for vegetables, nuts, tea, coffee, and legumes. Conversely, plant-based diets biased toward refined grains and sugar increased the risk. The conclusion is that it is not a matter of "if it is plant-based, it is good," but rather "which plant-based foods" are consumed.
However, these are all observational studies. There is the confounding factor that "people who are health-conscious to begin with also eat more vegetables," and the reverse causality that is particularly troublesome in dementia—that is, changes in the brain before the onset of the disease change eating habits first, leading to a relationship where it is not "dementia because vegetable intake decreased," but "vegetable intake decreased because dementia was starting." There is no intervention trial yet that proves vegetable intake reduces dementia.

Chapter 4: What is the true nature of the effect?

Assuming vegetables have an effect, is the protagonist "phytochemicals" or "dietary fiber and gut bacteria"? Both work in different places and have different strengths of evidence.

The Phytochemical (Plant Chemical) Theory

Vegetables contain physiologically active substances such as sulforaphane (broccoli), various flavonoids, carotenoids, and polyphenols. In test-tube studies and animal experiments, these certainly work. For example, sulforaphane pushes a switch in cells called "Nrf2," simultaneously activating groups of genes for detoxification enzymes and antioxidant enzymes. Anti-inflammatory effects have also been confirmed.
However, there is a pitfall. The simple explanation that "antioxidants eliminate the body's rust (active oxygen) and prevent cancer" was completely refuted by tests on β-carotene. In other words, phytochemicals are not "fire extinguishers that directly eliminate active oxygen."
A more persuasive explanation is "hormesis," a phenomenon where a substance or stimulus shows beneficial effects on a living organism at low doses, while showing harmful effects at high doses. For plants, many phytochemicals are "poisons" to repel insects and microorganisms; they are, so to speak, natural pesticides created by plants.
Biochemist Bruce Ames pointed out that the overwhelming majority of chemicals we consume daily are not artificial chemicals like pesticides or additives, but natural substances created by plants. When we continue to eat those trace amounts of "mild poison," the body perceives it as a stimulus and activates defense and repair systems like Nrf2 in advance. The view is that what is effective is not the component itself, but our own adaptive response that is activated in response to it.
This theory neatly explains the mysteries so far. At low doses and in many varieties, it promotes an adaptive response and is beneficial, but if isolated and taken at high doses, the toxicity comes to the fore and it becomes harmful. That is why vegetables are good for the body, but β-carotene tablets were not. This paradox is the core of the phytochemical theory.
This way of thinking is attractive, and there is a correlation that people with higher blood levels of carotenoids and flavonoids have fewer diseases. However, there is still no evidence showing that phytochemicals reduce cancer or dementia. A fair assessment is that it is promising but unproven.

The Dietary Fiber and Gut Bacteria Theory

The other protagonist is dietary fiber and gut bacteria. This area has stronger evidence. Dietary fiber increases stool volume, dilutes carcinogens, and shortens intestinal transit time, reducing contact between harmful substances and the intestinal wall. However, the real action happens beyond that. When gut bacteria ferment dietary fiber, large amounts of short-chain fatty acids such as acetic acid, propionic acid, and butyric acid are produced.
Among them, butyric acid has an effective duality. Butyric acid is a major energy source for large intestine cells. In situations where there is sufficient energy, butyric acid changes its point of action, promotes apoptosis (programmed cell death) of cells with damaged DNA, and suppresses the proliferation of cancer cells. This "butyric acid paradox" is considered the core of the anti-cancer effect of dietary fiber. In addition, short-chain fatty acids maintain the intestinal barrier and have the function of regulating immunity. The World Cancer Research Fund evaluates the relationship between dietary fiber and colorectal cancer prevention as the most important. This is the most solid pathway among the many claims of "cancer prevention by vegetable-derived components."
However, caution is needed here as well. Giving "supplements" of dietary fiber for several years could not prevent the recurrence of colorectal polyps. This points in the same direction as the lesson of β-carotene. What is effective is continuing a diet that includes diverse dietary fiber, and the ecosystem itself woven by dietary fiber and gut bacteria.

The Gut-Brain Axis Theory

So what about dementia? Here, the two hypotheses merge. The dementia prevention effect likely passes through a pathway that protects blood vessels. Folic acid and vitamins contained in vegetables lower homocysteine, nitrates dilate blood vessels, and anti-inflammatory effects protect the small blood vessels in the brain. The fact that effects were seen in total dementia and Alzheimer's disease rather than vascular dementia in Hisayama Town suggests something more than mere vascular protection.
The leading candidate for that "something" is the currently trending gut-brain axis. Short-chain fatty acids and other metabolites produced by gut bacteria send signals to the brain via immunity, inflammation, or the vagus nerve. The possibility that dietary fiber and the intestinal environment affect the brain through this pathway is being studied. However, this is still at the animal experiment stage and is not an established causality in humans.

Conclusion

Regarding cancer, the possibility that "vegetables prevent it" is considered quite low. Regarding dementia, vegetables (not fruits) have a preventive effect. However, the limitations of observational studies, "confounding and reverse causality," are always present.
As for the mechanism of the effect, the pathway of dietary fiber and gut bacteria (short-chain fatty acids) is a candidate for colorectal cancer, and vascular protection and the gut-brain axis are candidates for dementia. Phytochemicals may be contributing in the form of hormesis (activation of adaptive responses by mild toxins) rather than simple antioxidant effects.
And the failure of β-carotene tablets and the lack of results from dietary fiber supplements tell us the same thing. The effectiveness disappears when turned into a tablet. It resides in a "diet" where diverse components are mixed at low doses and work over decades while interacting with the ecosystem of gut bacteria.
Rather than looking for a miracle pill, have a plate of vegetables at today's dinner table. That is the conclusion.