The Cancer Prevention Advice Parents Aren’t Hearing

What the evidence really says about food, exposure, and childhood cancer and where your effort is better spent.

A companion to —

It usually comes at the end of the visit, hand already on the doorknob. We have spent fifty minutes on her cancer. Then she asks the question that has been waiting underneath the rest: does any of this matter for her kids?

What she is really asking is whether she has already done something wrong. I have watched women carry that question through an entire course of treatment without ever saying it out loud.

Here is where I land. Most of what parents are afraid of is not there. The few things that are there are almost embarrassingly cheap to address. And the intervention most likely to move the needle is not something you can buy, which is probably why nobody is selling it to you.

Childhood cancer is rare, and much of its biology begins early. In most childhood ALL and some AML, the first leukemic genetic event can be traced back to cord blood or a neonatal blood spot. The clock started before anyone knew there was a clock.

But prenatal is not predetermined.

Screen the cord blood of healthy newborns and you will find that first genetic change already sitting there in 1 to 5% of them, depending on the test used. Childhood ALL affects about 1 in 2,000 children. Of the babies born carrying the change, only about 1 in 500 will ever develop leukemia.

Almost everything that decides the difference happens after birth.

Is food giving my child cancer?

There is no food I can point to and tell you it causes childhood cancer. Not one. What exists is a set of signals, three of which come up often enough that I take them seriously.

Cured meat has the mechanism you would invent if you were trying to convince someone. Nitrite plus amines in the stomach forms N-nitroso compounds, the most potent nervous system carcinogens known in animal models. A meta-analysis put maternal cured meat consumption at 68% higher risk of childhood brain tumors, and prenatal vitamins blunted the association, exactly what you would predict if vitamin C were interrupting nitrosation. It is also the only one of the three not confined to pregnancy. A 2025 meta-analysis put children’s own processed meat intake at 72% higher odds of acute leukemia, and their vegetable intake at 39% lower.

Photo by Christina Victoria Craft on Unsplash

Coffee has the larger literature and gets almost none of the attention. Pooling eight studies and 2,552 cases, the international consortium found more than two cups a day at 27% higher odds of ALL, with a dose response.

Alcohol is the third. Across 39 case-control studies and more than 16,000 cases, maternal drinking in pregnancy tracked with childhood AML, from 64% higher odds at moderate intake to more than double at high. Nothing for ALL, which is the far more common disease.

The pattern nobody mentions

Every time researchers stopped asking mothers to remember and started measuring instead, the signal shrank or vanished.

Pesticides.

Pooled across 12 case-control studies, household pesticide use during pregnancy carried about 40% higher odds of childhood ALL, and a 2021 meta-analysis of 55 studies put maternal exposure at nearly double. The risk concentrated on indoor insecticides and on pregnancy rather than later childhood. Then investigators measured 20 actual insecticides in household carpet dust and found nothing.

Photo by heino eisner on Unsplash

Glyphosate, the same. IARC calls it a probable human carcinogen on occupational evidence: farm applicators, high cumulative exposure, adult lymphoma. Even that is contested, and the largest prospective cohort of 54,000 applicatorsfound no association. The childhood question has been asked exactly once with measurement. Researchers tested house dust in the homes of 181 leukemia cases, found glyphosate in 99% of them, and found no relationship to concentration.

Cured meat, the same. A 2023 review of 181 studies found the association in the case-control data and not in the cohorts, where mothers are asked before anyone knows which child will get sick.

Coffee, the same. A Danish cohort of 141,000 pregnancies found nothing.

Why?

• A mother whose child has leukemia remembers her pregnancy differently than a mother whose child is well. That is not dishonesty. It is what memory does after catastrophe.

• The technical name is differential exposure misclassification. If both groups misremember equally, a study only misses real effects. If one group misremembers more, a study can invent an effect that was never there. And parents do misremember. When researchers checked interviews against medical records six years after birth, parents recalled fewer than half the infections and medications their charts had documented. In a study of childhood ALL, the gap between case and control parents only showed up where parents had already heard that exposure blamed.

To be clear, I don’t think the above makes these exposures safe. A 2025 study dosing rats from prenatal life at the European acceptable daily intake found dose-related tumors including early leukemias. But when an umbrella review graded the entire environmental literature for childhood ALL, exactly two exposures were rated convincing: ionizing radiation in early childhood, and pesticide exposure around conception. No food made that tier. A 2025 overview of thirteen systematic reviews put it plainly: the certainty is too low to establish a clear association between any nutritional factor and childhood ALL.

We have spent decades interrogating what mothers ate, and we still do not know.

What if the exposure is the protective part?

Greaves’ model proposes that common B-cell precursor ALL takes two steps. The lesion in utero, then a second event when an ordinary childhood infection meets an immune system that was never trained, because the early microbial exposure never came. Children with leukemia show delayed gut microbiome maturation.

If that holds, the most common cancer of childhood is partly a disease of insufficient exposure. Not too much of something. Too little.

The evidence is mixed. In the UK childhood cancer study, formal day care in the first year carried about half the odds of ALL against population controls and about 30% lower against a stricter comparison group, with a dose response either way. Then the same Danish registry that produced the breastfeeding finding followed 1.1 million children and found no protection from childcare whatsoever. It remains a hypothesis and not a settled one.

This holds a lot of similarity to the peanut story from the first piece, where a decade of telling parents to withhold allergens was causing the allergy it was meant to prevent. The protective instinct, remove the challenge, produced the harm. I have watched that pattern hold in allergy, in the microbiome, and in exercise physiology. I would not bet against it here.

Can food change my child’s genes?

Probably not in the same way we all think.

People conceived during the Dutch Hunger Winter still had less methylation of the IGF2 gene six decades later than their own unexposed siblings, and later work showed those differences statistically mediated the famine’s effect on adult BMI and triglycerides. Early nutrition does leave durable molecular marks in humans.

But, let’s keep it in proportion. The exposure was starvation, not a mediocre diet. The effect was largest around conception. The differences are a few percentage points, and there is a serious argument that part of the signature reflects which embryos survived rather than which genes were reprogrammed. None of it tells us what an ordinary diet does.

It does reach one generation further. A small follow-up found higher adult BMI among the children of men exposed to famine in utero. Intriguing, and not proof of transgenerational inheritance: the father’s germ-cell precursors were physically present during that exposure. The grandchild was already in the room.

What about soy?

Almost every version of this question comes from a family with breast cancer in it, and the fear runs precisely backward from the data.

Asian American women who ate the most soy as children have about 60% less breast cancer as adults than those who ate the least, and childhood is the window where the association is strongest. That number rests on mothers of 99 cases recalling their daughters’ diets, so hold the size of it loosely. But nothing in humans shows soy foods harm girls, and the signal is reasonably consistent across Asian and Asian American populations. A Taiwanese study of children’s own diets found bean curd running the same protective direction.

One of many meals in Japan showcasing the incredible variety of fermented soy, from miso and natto to tofu and beyond.

Tofu, edamame, and miso are food. Soy protein isolate in a bar is a formulation. It frustrates me that one conversation covers both.

What cancer risk are we not talking about?

Everything above concerns childhood cancer, which is rare. The larger lever points at adult cancer decades out, through body fat. Childhood BMI predicts adult obesity-related cancer across 46 cohorts. But the UK Biobank result is the one I repeat to parents: risk was determined mostly by adult BMI. Childhood sets the trajectory. It does not seal it.

Which means the mother at the door is worrying about the wrong window.

What about red meat and sugar?

The right window is longer and quieter. It opens in childhood too, but nothing in it announces itself for decades, which is why I almost never get these questions from a parent. I get them from the woman sitting in front of me, after her own diagnosis.

Meat

Dr. Leigh Erin Connealy has recently written on eating for cancer prevention, and I agree with most of it. The studies on red meat and cancer risk are sloppy, food frequency questionnaires are a weak instrument, and the whole dietary pattern matters far more than any single food in it. I would go further than she does on processing. Nitrites, preservatives and oxidized fats are almost certainly doing damage, and a hot dog and a grass-fed steak should never have been counted as the same exposure in the first place.

Where I would push a little is the idea that the risk lives entirely in the processing. Five mechanisms get proposed. Three are about what is done to the meat: nitrites from curing, polycyclic aromatic hydrocarbons from smoking, heterocyclic amines from high-heat cooking. She is right about all three. The other two are in the meat itself. Heme iron catalyzes N-nitroso formation and lipid peroxidation in the gut, and Neu5Gc, a sugar humans cannot make, incorporates into our own tissue and provokes an inflammatory response. Neither cares how the animal was raised.

Fed to ileostomy patients, unprocessed red meat raised gut N-nitroso compounds roughly fourfold, and in the Black Women’s Health Study unprocessed red meat carried 33% higher colorectal cancer risk per 100 g a day while processed meat carried none.

The Maasai comparison I would hold loosely too. Kenyan cancer registries cover less than 20% of the population and African cancer registration carries a well documented urban bias, so Maasai cancers are largely uncounted rather than absent. Cancer is also mostly a disease of old age, so a population that dies younger will look protected whether it is or not. None of which makes a steak dangerous. Pooled across cohorts, unprocessed red meat raises the risk of these cancers by roughly 10 to 25%. Smoking multiplies lung cancer risk roughly twentyfold. And those cohorts did not record how the animal was raised, so what they measured is the beef supply as it actually is, which is overwhelmingly conventional. That gap is real, and it is why the heme evidence carries the weight here, because heme is the one thing that does not depend on the question nobody asked. Grass-fed does not remove it. If anything it adds a little, since grass-finished beef runs consistently higher in iron than grain-finished.

Sugar

In NutriNet-Santé, 101,279 adults followed prospectively, top-quartile sugar intake carried 51% higher breast cancer risk, and it held after adjusting for weight gain. But most studies of total sugar find nothing. The signal sits in added sugars and sugary drinks, which is what you would expect if the pathway runs through insulin rather than through the molecule. Same argument as children’s processed fruit snacks. The type of sugar and delivery matters!!

The two sections of Connealy's piece I would put in front of any cancer patient are the ones on sugar and on starving cancer. She takes apart the idea that cutting carbohydrates starves a tumor, and she is right. Restrict glucose and the body manufactures its own from muscle, while the tumor shifts to fatty acids and glutamine. That is a different question from the one above, and the two get lumped together constantly. Sugar intake over decades tracking with getting cancer is not the same claim as glucose restriction treating one.

Which brings it back to children. The Nurses’ Health Study II followed 44,231 women who had reconstructed their high school diets. Those who ate the most red meat as adolescents had 43% higher premenopausal breast cancer risk. Breast tissue between puberty and a first pregnancy is undifferentiated and dividing fast, which is exactly when a carcinogen does the most damage. Same vulnerable window as the leukemia story, different tissue.

One more thing, missing from nearly every version of this conversation …including mine is alcohol. In a diet-wide analysis of 542,778 women, alcohol had the strongest positive association with cancers of the 97 dietary factors examined, and calcium the strongest protective one. Before we argue about whether to eat grass-fed meat or not, we should spend the worry where the evidence is.

So where is the effort actually worth it?

The data are mixed. I think we should act anyway. My threshold is simple: if an intervention is cheap, reversible, independently good for you, and lands in a window that closes, I am willing to act on weaker evidence. If it is expensive, permanent, or displaces something better, I want much more.

Cured meat in pregnancy clears that bar easily. Nine months, no cost, no nutritional loss, coherent mechanism. Same for keeping coffee to a cup or two, which has better established pregnancy reasons anyway. Same for minimizing indoor insecticides and unnecessary outdoor pesticides. The signal is imperfect but persistent, and other than inconvenience, there is little downside to reaching for alternatives first.

Alcohol and sugary drinks clear that bar faster than anything else on this list. Alcohol is a known carcinogen with the strongest signal of the 97 dietary factors examined, and sugary drinks are the delivery system behind most of the sugar association. Neither has a nutritional argument to defend. If you were going to change one thing this week, it should be one of those, not the beef.

Fruit and vegetables clear the bar for the opposite reason. Maternal fruit intake of two or more daily servings is associated with 29% lower odds of childhood ALL, and pooled data put vegetables, legumes, and fish in the same direction. Maybe some of that is confounding. I still lose nothing by acting on it. That asymmetry is why I spend far more clinic time on what to add than what to remove.

A typical farmers market haul: seasonal vegetables, minimally processed whole grains, and as much fresh variety as we can carry home.

Adolescence deserves the same attention as pregnancy and almost never gets it. The advice itself is no different from any other age, which is the unglamorous part. Vegetables, fiber, soy, not much processed meat. What changes is the timing. Breast tissue between puberty and a first pregnancy is proliferating and undifferentiated, so the same diet does more work at fourteen than it will at thirty. It is also when the first drinks happen and when sugary drink intake peaks, and those are the two things I would actually intervene on.

Breastfeeding has some of the strongest prospective evidence on the list. In a Danish cohort of 309,000 children, exclusive breastfeeding for three months or more was associated with a 38% lower risk of B-cell precursor ALL. That is a relative reduction in an already rare disease, and breastfeeding is neither possible nor the right choice for every family. It is a signal, not a mandate.

Organic is where I am less consistent. I buy it for the highest-residue produce. Organic diets consistently reduce measured pesticide exposure; what has never been shown is that this translates into less childhood cancer. The wellness industry routinely collapses those two statements into one. I will not. I simply choose the lower-exposure option when it is practical.

For children I limit processed meat, and I do not make red meat a daily staple either, because the heme signal does not disappear with better sourcing. Neither is a food to fear. Both are foods I would rather see three times a week than seven. But population-level risk is not individual causation. Nothing in these data can tell a parent that something they fed their child caused their leukemia. That distinction matters enormously when you are looking backward for something to blame.

So yes, the effort matters. Just not all of it, and not equally. Most weight on pregnancy, because that window is short, cheap, and closes. Real weight on adolescence, because it is the one nobody is watching. And the most weight of all on the twenty years after that: sleep, movement, a plant-rich diet, and enough exposure to the world to build a competent immune system. Those are the habits that shape a metabolic trajectory long after childhood cancer risk has become vanishingly small.

---

Next: the chemical environment, what the evidence actually supports, and where I think the fear is miscalibrated, including plastics, pesticides, and EMFs.

---

Ksenia Malarkey, ND is a naturopathic physician specializing in integrative oncology, hormones, and metabolic health.

She provides medical care for Washington and Oregon residents and offers global health coaching.

---

To schedule a free 15-minute consult, visit drkseniamalarkey.com