Robert A. Smith has spent thirty-four years at the American Cancer Society telling women to get mammograms. He joined in 1992, led the development of its cancer-screening guidelines, and sat on the international agency panels that told the rest of the world to do the same. This month his name turned up on the paper that shrinks the single strongest argument against the thing he has promoted his entire career.
On September 14, JNCI published a meta-analysis under the reassuring subtitle “separating signal from noise.” Its finding, announced with fanfare, is that overdiagnosis in breast screening, long estimated at anywhere from a fifth to half of all screen-detected cancers, is actually below 5 percent.
Overdiagnosis is not a small thing to reassess. It is the harm that has shadowed mammography for two decades: the cancers screening finds that would never have grown, never spread, never killed, tumors a woman would have died with rather than of, had no one gone looking. Once a mammogram finds one, no one can tell it apart from the kind that does kill, so it gets treated. That means lumpectomies and mastectomies, radiation, sometimes chemotherapy, delivered to women whose cancer was never going to hurt them. It is the reason honest screening counseling started naming a downside at all, the reason a growing number of clinicians began treating a woman’s choice to skip the mammogram as a reasonable call rather than a reckless one. It is the crack in the screening edifice, and this paper sets out to seal it.
Here is how. The team reexamined all eight of the old randomized screening trials the entire debate rests on: the New York Health Insurance Plan study, Malmö, the Two-County trial, Edinburgh, the Canadian National Breast Screening Study, Stockholm, Gothenburg, and UK Age. Some of that data is older than the women it was meant to protect; the earliest trial began enrolling in the 1960s. They gathered nothing new. They reinterpreted what was already there, adjusting for whether control-group women eventually got screened too, for how often screening rounds came, and for how long each trial counted cases before it stopped. Run those adjustments and the excess cancers that once looked like overdiagnosis mostly dissolve into a timing artifact. When screening starts, co-author Elsebeth Lynge explained, diagnoses climb simply because cancers get found earlier than they otherwise would have been; count long enough and the bulge evens out. The trials, the authors argue, mostly never counted long enough.
The number moved. The data did not.
A reanalysis is a legitimate scientific act, and the confounders the authors name have bedeviled overdiagnosis estimates for years. But reinterpreting sixty-year-old trials is an argument about assumptions, not a fresh measurement, and the assumptions here all run one direction: toward vindicating the screening enterprise. Reclassify the control arms, pick which follow-up window counts, and you are making choices, and different choices produced the older, larger estimates. The UK Independent Panel, no den of screening skeptics, landed at roughly 19 percent. This paper says under 5 percent. Same trials, different hands.
Whose hands is what the press releases left out. Smith is the American Cancer Society’s most decorated screening advocate and, as it happens, a sitting member of the editorial board of the journal that published the paper, co-authoring the work that rescues the intervention he has championed since 1992. None of that makes the arithmetic wrong. It prices the enthusiasm. The screening establishment revising its own central harm downward, through its own people, in its own journal, is not the same event as an outside team arriving at the finding cold.
Even a sympathetic reader sees the seams. Kefah Mokbel, chair of breast surgery at the London Breast Institute and himself convinced that overdiagnosis has been overstated, endorsed the reanalysis and then named what it cannot do: it is “a compatibility analysis, not a pooled quantitative estimate,” he wrote, one that shows the trial data are consistent with low overdiagnosis rather than measuring it; everything rests on a single Danish reference population; and the reclassification of the control arms “deserves independent scrutiny” before anyone treats 5 percent as settled. His counsel was for “balanced counselling” that reflects “genuine uncertainty in both directions.” It is a strange thing to have to say about a paper being sold as the end of the argument.
The funding, at least, is disclosed. One co-author’s work was backed by the Novo Nordisk Foundation, the charitable wing of one of the world’s largest drugmakers, another’s by Cancer Research UK. The conflict that goes unmentioned is the one sitting in the author list.
What the paper changes is not the biology of anyone’s tumor. It changes the sentence a doctor is expected to say at the exam. For twenty years, honest informed consent meant telling a woman that screening carries a chance of treating a cancer that would never have harmed her. A number under 5 percent makes that sentence easy to drop. Papers like this one get cited in exactly that direction, in guideline fights and screening-drive press kits, and this one arrives pre-built for the errand. Robert Smith spent thirty-four years making the case for the mammogram. The last thing his life’s work needed was for its biggest liability to survive the archive. It didn’t, and the paper that buried it went out under a subtitle about telling signal from noise.
Sources
- JNCI: Njor, Pedersen, Lynge, Smith, Rebolj, “Breast cancer overdiagnosis in mammography trials: separating signal from noise” (2026)
- OncoDaily: Kefah Mokbel, “Is breast cancer overdiagnosis largely a methodological artefact?”
- EurekAlert: “New analysis points to low overdiagnosis in breast cancer screening” (press release, 2026)
- American Cancer Society: Robert Smith, PhD, researcher biography
- ScienceDaily: “A major risk of breast cancer screening may have been overestimated for decades” (2026)