I keep my phone on the pillow. I know better, and I do it anyway, because for years the official line was reassuring enough to let me: no solid evidence that the thing pressed against my skull all night touches the cells underneath it. That line got a lot quieter this week, and the reason is a piece of arithmetic so blunt I had to read it three times.

On July 29, the journal Environmental Health published a commentary by Ronald Melnick, Joel Moskowitz, and Paul Héroux taking apart a pair of studies that regulators have been holding up as proof your phone is safe. Melnick is not a bystander here. He designed the original U.S. government experiment that started this whole fight. And his argument is not that the new studies found the wrong answer. It is that they were built, from the first animal cage, in a way that made finding anything nearly impossible.

Back up to what he built. Between 2016 and 2018, the National Toxicology Program ran the most exhaustive animal study of cell-phone radiation ever attempted, a project north of $30 million. Rats were exposed whole-body to 900 MHz signals, both GSM and CDMA, at three dose levels, starting in the womb and running two years, with 90 males and 90 females in each group. The result the NTP called “clear evidence”: male rats grew malignant schwannomas in the heart and gliomas in the brain. The strongest single signal came in the highest CDMA group, where 6 of 90 males developed heart schwannomas against 0 in controls, a difference that cleared statistical significance.

Then the FDA, which had asked NTP for the study in the first place, put out a statement dismissing it. The agency’s device chief said the findings “should not be applied to human cell phone usage” and that “the existing safety limits for cell phones remain acceptable.” Existing limits are fine. Do not read this as a human risk. And ever since, the establishment position has needed one thing to hold: a replication that comes back clean.

Enter Japan and Korea. Their government labs ran a two-year follow-up from 2020 to 2022, and the Japanese results landed in Toxicological Sciences in January 2026: no reproducible carcinogenic effect. Australia’s radiation regulator promptly ran the headline that Korean and Japanese studies “agree” RF exposure doesn’t cause cancer in rats. Case closed, tuck your phone back under the pillow.

Here is where the arithmetic stopped me. To “validate” a study built with three doses and 180 animals per group, the follow-up used one modulation, one dose, one sex, and 70 males per group. They picked 4 W/kg, a number that matters mostly because it sits inside the international exposure guidelines, and they dropped the 6 W/kg CDMA condition entirely, which is precisely the dose where the original heart-tumor signal was strongest. So the “replication” quietly omitted the exact spot where the thing it was replicating lived.

And the tumors NTP was chasing are rare. A rare event needs a lot of animals before its absence means anything. Melnick and his co-authors run the numbers: to reliably detect the heart schwannoma at the incidence NTP saw, you would need about 258 animals per group for 80 percent statistical power, and 323 for 90 percent. The follow-up ran 70. A study that small does not test the hypothesis, it launders a null result. As the commentary puts it, weaker experiments with lower doses, fewer groups, and reduced power “should not be used to test the validity of more robust studies.” You cannot debunk a finding with an instrument too dull to register it.

Wait, I kept asking, why would a radio signal grow a tumor in a rat’s heart at all? It is not a random place for one to land, and that is the part worth chasing. Schwannomas come from Schwann cells, the insulation wrapped around peripheral nerves. Gliomas come from glial cells, the same supporting-cell family up in the brain. NTP saw both tumor types climb in the RF-exposed males, two branches of one nerve-support lineage lighting up together. That is what makes a toxicologist lean in, because when an exposure hits one branch of a cell family and its sibling branch as well, that is the opposite of noise. The long-standing reassurance from industry and regulators has been that RF is non-ionizing, that it can only gently heat tissue, and heat alone does not do this. A biological pattern this tidy is exactly the kind of thing that reassurance was never built to explain.

And the punchline the “agree” headlines skipped: even the underpowered Korean study still turned up 2 heart schwannomas among its 70 exposed males, against 0 in the sham-exposed group and 0 in the cage controls. Same tumor, same direction, in a design too small to call it. The Korean authors said plainly their project was “not designed as a complete replication” of NTP, and the Japanese team said their aim was never to replicate the findings quantitatively. The regulators reporting it as a clean replication either did not read that far or did not want to.

You can see who this arrangement serves. A clean null keeps the guidelines untouched, keeps the ICNIRP limits and the WHO reviews on their current footing, and keeps the telecom rollout free of the one word that would complicate it. The tell isn’t that the follow-up found nothing. It’s that it was engineered to.

I won’t tell you your phone gives you cancer, because the honest read is narrower and, frankly, more damning: the strongest animal evidence we have points somewhere real, and the “debunking” was a magic trick performed on people who wanted to be fooled. So I moved my phone off the pillow this week, and it’s staying on the far nightstand. That costs me nothing, and I’d rather not be the control group for a study nobody bothered to power to protect me.

Sources

  1. Environmental Health – Melnick, Moskowitz & Héroux, commentary on design deficiencies in the RF follow-up validation studies (2026)
  2. Electromagnetic Radiation Safety (Joel Moskowitz, UC Berkeley) – NTP cellphone carcinogenicity study: the follow-up studies (2026)
  3. RF Safe – technical critique of the statistical power and dose design of the Japan-Korea studies, with the Korean schwannoma counts and the power calculation
  4. Microwave News – the Japan/Korea NTP RF replication project
  5. Toxicological Sciences – International Collaborative Animal Study, the Japanese RF carcinogenicity results (2026)
  6. ARPANSA – regulator statement that Korean and Japanese studies “agree” RF does not cause cancer in rats
  7. FDA – statement from Jeffrey Shuren dismissing the NTP report and defending existing cell-phone safety limits (2018)
  8. NIEHS / National Toxicology Program – cell phone radiofrequency radiation study factsheet
  9. Electromagnetic Radiation Safety – NTP final reports on cell phone radiation (2018)