2 childhood cancers were higher where PFOS was in the water, but most were not

What did this study compare?

Two lists of children. 6,448 diagnosed with cancer and 13,044 not, sorted by one thing, whether a birth address sat inside a water district where PFAS had turned up. Two cancers came back higher where PFOS was detected, neuroblastoma and retinoblastoma. Cancer as a whole did not.

Nobody knocked on a door. Nobody drew a child’s blood, tested a kitchen tap, or asked one family what they drank. If you want to know what any child in it swallowed, nobody can tell you.

Natalie Binczewski and Verónica Vieira ran it out of UC Irvine, with colleagues at UC Berkeley and USC, and every child in it was born in Los Angeles or Orange County between 2000 and 2019. Cancer cases were followed through December 2021. Each child who got cancer was matched against two born the same year who didn’t. Their paper ran in the Journal of Exposure Science and Environmental Epidemiology on 5 March 2026.

Here’s where you’ve to slow down. They took a round of federal water testing done between 2013 and 2015, marked 17 water systems as detect or non-detect for two PFAS, drew each district’s service boundary on a map, then dropped every birth address into whichever district held it.

So exposure here is a yes or a no attached to a line on a map. Not a glass of water. Not one year of drinking.

Hardly any children fell inside one, either. PFOS got assigned to 12.31 percent of cases against 10.63 of controls. PFOA ran 7.29 against 6.45. Nine children in ten lived outside those districts entirely.

Which cancers came back higher, and which did not?

Two rare ones. Neuroblastoma and retinoblastoma where PFOS had been found, retinoblastoma again where PFOA had, and nothing else anywhere in their grid.

One of those starts in an eye. The other starts in nerve tissue. Most parents have never heard either name, and that’s the first thing worth knowing here.

Read that Cases column before you read anything else. Both results that cleared the line rest on 316 children, and on 180, across two counties and twenty years. 180 is a very small group to cut in half.

Now read those top two rows, because that’s where most of these children are. All cancers together came back at 1.07, on a spread running straight through 1. Leukemia is the biggest single group here, 2,070 children, and it came back at 0.97 for PFOS and 0.93 for PFOA. Nothing moved. In the one childhood cancer nearly everybody pictures first. If leukemia is the word that brought you here, that row is your answer.

Binczewski and colleagues do that arithmetic on themselves. Their work is limited by “multiple testing for each cancer group, which may contribute to finding false positive associations,” they write. Test enough groups and one of them crosses by chance.

One small thing we owe you. That eye cancer figure prints as 1.83 in their abstract and 1.84 in Table 2, on one identical spread. We used the table, since that’s where model output lives. A hundredth changes nothing here, and we’d rather say so out loud than pick quietly.

Why did water carrying 39 parts per trillion pass as clean?

Because in 2014 a laboratory wasn’t asked to report anything under 40.

Two districts, one boundary between them. One reports 42 parts per trillion of PFOS and goes into this study as contaminated. Its neighbor reports 39, gets filed as a non-detect, and every child born there joins the comparison group. If you’d lived in that second district, this study counted your water as clean.

That’s the era it was done in. Water figures here come from testing run between 2013 and 2015, whose cutoffs were 40 parts per trillion for PFOS and 20 for PFOA. Districts “with detectable but non-reported concentrations of PFOS/PFOA may have been misclassified,” the authors write, and they’re the ones who raised it.

Today’s enforceable limit is 4.0 parts per trillion for each. EPA set it in April 2024. So a district counted as contaminated in this work was running ten times over what’s now allowed, with reported detects between 42 and 53 parts per trillion.

Testing has got far finer since, and a current round reports down to 4 parts per trillion for both chemicals. But a reporting level was never a safety line, and EPA says so plainly about its own, that those numbers “are not associated with contaminant health effects information.” They describe what a laboratory can reliably see, which is a different question from what’s safe to drink.

We came at this expecting clean water against dirty. It was never that. It set badly contaminated districts against everywhere else, and that second group covers an enormous amount of ground.

Does this mean PFAS caused those cancers?

No. And the people who did the work say it first.

Their word for it is exploratory. Their full sentence carries its own reasons, so here it is whole. This result “is informative, but should be considered exploratory given that the exposure assessment is limited to a narrow timeframe of public water measurements, without information on actual water consumption behaviors,” they write.

Their conclusion cuts both ways too. PFOS and PFOA “did not fully explain observed geographic variations in risk,” says the abstract. Patches of that map stayed lit after they adjusted for water and for mothers. Binczewski and colleagues point at concentrated poverty and income inequality in northern Orange County and South Los Angeles, and at “unmeasured spatial confounders.” They had nothing on family history of cancer and nothing on household income.

No mechanism has been proposed, either. “Mechanistic evidence for the carcinogenicity of PFAS in nerve tissue is lacking,” they write. Neuroblastoma grows in nerve tissue, so that gap opens directly under one of their two findings.

None of which makes PFAS harmless, and hazard evidence lives somewhere else. IARC classed PFOA as carcinogenic to humans, Group 1, in February 2025, on sufficient evidence in animals plus strong mechanistic findings. Human evidence was limited, covering cancer of the testis and renal cell carcinoma. PFOS went into Group 2B, possibly carcinogenic.

A Group 1 rating says a substance can cause cancer somewhere, in a species, at a dose. It says nothing about the dose in your kitchen, and that panel has never claimed it does.

What does the law let into your tap?

An average under 4.0 parts per trillion of PFOA, and under 4.0 of PFOS. Enforceable since April 2024.

Run your kitchen tap. That limit is what stands behind it, and every child in this study was born before it existed. Systems have until 2027 to finish a first round of testing, and until 2029 to fix anything above that line.

Four more rules arrived alongside those two, covering three chemicals at 10 parts per trillion each, and one covering mixtures. A health-based goal of zero stands behind PFOA and PFOS as well, which is a target rather than something a utility gets cited for missing.

Both limits have been reopened twice since. EPA said in May 2025 it would keep PFOA and PFOS and scrap the rest. Two proposals followed on 18 May 2026, one handing systems two more years to comply, out to April 2031, and one scrapping those four. Any system with a sample at or above 12 parts per trillion would still have to act during that delay. Comments closed on 20 July 2026. Neither proposal is final.

Five kinds of figure have gone past you here and they’re not the same animal. 4.0 parts per trillion is a limit a utility answers for. Zero is a goal with nothing enforcing it. A reporting level is what a laboratory can see. An odds ratio of 1.84 is one comparison in two counties. Group 1 says a hazard exists somewhere. Stack them together and you get a fright that was never about you.

How much American water is over that line?

Roughly 8.0 percent of public systems, on federal testing about 95 percent finished.

If your water comes from a big utility, that’s the group that came off worst, which surprises people. Systems reporting a full set of results put 12.0 percent of the large ones, serving more than 10,000 people, over the PFOS limit, against 6.4 percent of medium systems and 6.0 percent of small. PFOA ran 10.9, 5.7 and 4.4 percent across the same three.

Nobody has explained that pattern and we’re not going to guess at it. Where a utility draws its water and how many people it serves travel together, for reasons with nothing to do with how well it’s run.

That federal summary prints one caution twice. These results don’t settle whether a system is complying or not. Compliance gets decided by a running annual average of quarterly samples filed with a state, so a snapshot settles nothing, and a final release due in fall 2026 will move every one of these percentages again.

Now set 8.0 percent beside this study for a second. Those Southern California districts driving its entire exposure variable were among the worst in this country in 2013. Most American water is nowhere near them.

How do you find out what is in your own water?

Two free documents, and neither one needs a test kit.

One is a webpage. Look your system up in the UCMR 5 Data Finder, then read the Consumer Confidence Report your utility already owes you once a year.

Systems have to tell customers those results exist within 12 months of knowing them. Community systems have to print them in that annual report wherever a contaminant was found, with an average and a range from the year before. So if you’re on public water, you can read your own figures instead of a national percentage.

Two limits on that. If you’re on a private well you appear in neither, because this monitoring covers public systems only. And a system granted that proposed two-year extension would still have to notify whoever it serves, under a rule nobody has finalized.

We’re printing no price for treating it, because no federal source publishes one for a household. Cost turns on which chemical, how much of it, whether you treat one tap or a whole house, and how fast your water kills a cartridge. Federal money exists, $1 billion of it under the Infrastructure Investment and Jobs Act, going to states, water systems and well owners. Not a rebate you claim at a store.

What Binczewski, Vieira and their colleagues handed us is narrower than the coverage suggests and more useful than nothing. Two rare cancers turned up at higher odds where 2013-era testing had found PFOS or PFOA above a coarse line. The largest group of children in their own table showed nothing at all. And their paper says plainly that water didn’t explain that map.

Your own water is a different question. It has an answer. Those figures are published, they’re free, and they run to this year.