For years, one of the most controversial questions surrounding mammography has been deceptively simple:
How often does screening find a breast cancer that would never have harmed the woman if it had remained undiscovered?
This phenomenon—known as overdiagnosis—is real. But estimates of its frequency have varied enormously, from almost zero to nearly 50% in different analyses. OUP Academic
Now researchers have gone back to the major randomized mammography trials and analyzed them within a single framework.
Their conclusion is striking.
The trial evidence appears compatible with an overall breast-cancer overdiagnosis rate of less than 5%, substantially below some widely cited estimates of 30–50%. OUP Academic
First, What Exactly Is Overdiagnosis?
Overdiagnosis doesn’t mean that a mammogram produces a false-positive result.
Those are two different things.
A false positive occurs when screening raises suspicion of cancer but subsequent testing shows that cancer isn’t present.
With overdiagnosis, the cancer really is there.
The problem is that the tumor would never have progressed enough to cause symptoms or threaten the woman’s life during her lifetime.
Because doctors generally can’t know with certainty which individual screen-detected cancer will remain harmless, such a cancer may still lead to surgery, radiation, endocrine therapy or other treatment.
That’s why determining the true frequency of overdiagnosis matters.
Previous Estimates Were All Over the Map
The scientific literature has produced remarkably different answers.
Earlier reviews reported estimates ranging from essentially zero to nearly 50%, depending heavily on how investigators defined and calculated overdiagnosis. NCBI
One influential Cochrane analysis estimated approximately 30% overdiagnosis and overtreatment among women undergoing screening. PubMed
Other analyses produced considerably smaller estimates.
So why could competent researchers examining mammography reach such radically different conclusions?
The new study argues that much of the disagreement comes from how the trial data were analyzed.
Researchers Went Back to Eight Major Trials
The new analysis was published September 14, 2026, in the Journal of the National Cancer Institute.
Researchers examined all eight major randomized mammography screening trials:
New York Health Insurance Plan, Malmö, Two-County, Edinburgh, Canadian National Breast Screening Study, Stockholm, Gothenburg and UK Age. OUP Academic
Rather than simply comparing the final number of breast cancers in screened and unscreened groups, they examined what happened to cancer incidence over time.
That distinction turns out to be crucial.
Mammography Moves Some Diagnoses Forward in Time
Imagine two women who have biologically identical cancers.
One receives regular mammograms.
Her tumor is detected in 2026.
The other isn’t screened.
Her tumor becomes noticeable and is diagnosed in 2029.
If researchers compare cancer incidence in 2026, the screened population appears to contain an “extra” cancer.
But wait until 2029 and the control population begins catching up.
That earlier diagnosis wasn’t necessarily an overdiagnosis.
It may simply have been earlier diagnosis.
This phenomenon is closely related to lead time, and failing to allow sufficient follow-up can make screening appear to produce more excess cancers than it ultimately does.
The “Compensatory Drop” Matters
After organized screening ends, researchers would expect cancer incidence in the previously screened population to fall relative to the comparison population.
Why?
Because screening already found some cancers earlier.
Those cancers aren’t available to be diagnosed again later.
The new analysis specifically considered these post-screening compensatory drops when interpreting the trials. OUP Academic
And this is one place where the researchers believe previous analyses could mistake earlier detection for overdiagnosis.
Another Problem: The Control Groups Didn’t Always Stay Unscreened
Randomized trials sound simple:
one group gets mammography;
the other doesn’t.
Reality was messier.
In several trials, women assigned to control groups were eventually screened themselves—sometimes at the end of the trial or after population screening programs were introduced.
The new researchers therefore treated control groups receiving an exit screen as screened at that point rather than continuing to consider them completely unscreened. OUP Academic
That seemingly small methodological decision can substantially alter the apparent difference in cancer incidence between the groups.
They Also Accounted for the Number of Screens
Another overlooked variable was simply:
How many opportunities did each group have to find cancer?
A population offered several rounds of mammography naturally has more opportunities for cancers to be discovered earlier than one offered fewer rounds.
The researchers therefore considered three major differences across the historical trials:
whether control participants eventually received screening,
how many screening rounds participants received,
and how long researchers continued following them afterward. sdu
Once those differences were considered consistently, a surprisingly coherent pattern appeared.
The Researchers Used Denmark as a Reference
The team compared the randomized-trial patterns with population screening experience from Denmark.
Specifically, they used observations from the screening program in Funen, where researchers could observe how cumulative breast-cancer incidence changed between birth cohorts invited and not invited to screening.
The Danish pattern provided a reference for what the incidence curves might look like when overdiagnosis is low.
Then the researchers asked:
Do the randomized trials behave similarly?
Across 52 follow-up timepoints, they examined 73 relative-proportion observations involving invasive breast cancer, both with and without ductal carcinoma in situ (DCIS). OUP Academic
The agreement was striking.
The Pattern Was Consistent With Less Than 5% Overdiagnosis
Most confidence intervals around the observed trial estimates overlapped the values expected from the Danish reference pattern.
The researchers concluded that, when the trials are analyzed consistently while accounting for screening timing and follow-up, their incidence patterns are compatible with very low overall overdiagnosis—below 5%. OUP Academic
That’s dramatically different from the highest historical estimates.
But there’s an important nuance.
The researchers did not somehow count every overdiagnosed tumor individually and establish that the exact universal mammography overdiagnosis rate is, for example, 3.7%.
Their conclusion is that the randomized-trial patterns are consistent with an overdiagnosis rate below 5% under their analytical framework.
That’s a more careful—and scientifically important—distinction.
Why Earlier Studies Sometimes Reached 30–50%
Suppose screening finds 110 cancers while the control group has detected only 80 at a particular point.
It would be tempting to interpret those additional cancers as overdiagnosis.
But some of those 30 women may simply have had their cancers diagnosed earlier.
Continue following both groups and some cancers that appeared early in the screened population may subsequently appear in the comparison population.
That’s why overdiagnosis can’t necessarily be calculated simply by looking at the temporary excess number of cancers after screening.
The length and timing of follow-up matter enormously.
Historical reviews have acknowledged that differences in methodology—including comparison populations, assumptions about lead time and even which denominator researchers use—can produce very different overdiagnosis estimates. NCBI
This Doesn’t Mean Overdiagnosis Has Disappeared
The new findings shouldn’t be interpreted as:
“Mammograms never find cancers that wouldn’t have caused harm.”
They can.
The question is how often.
And the new paper doesn’t erase previous research reaching different estimates. Rather, it argues that methodological inconsistencies explain much of the extraordinary variation between those estimates.
For example, an influential 2016 analysis of U.S. tumor-size trends concluded that mammography had substantially increased detection of small tumors beyond the decrease observed in larger tumors, interpreting much of that difference as overdiagnosis. New England Journal of Medicine
That illustrates why the issue has remained scientifically contentious.
Different methods can produce very different answers.
DCIS Is Particularly Important
One reason this discussion becomes complicated is ductal carcinoma in situ, or DCIS.
DCIS consists of abnormal cells contained within breast ducts and hasn’t invaded surrounding breast tissue.
Some DCIS lesions may eventually become invasive cancers.
Others may never progress enough to cause harm.
Unfortunately, doctors cannot always reliably predict the future behavior of an individual lesion.
Screening increases detection of DCIS, making it an important component of the overdiagnosis discussion.
The new analysis examined results both including and excluding DCIS and still found patterns compatible with low overall overdiagnosis. OUP Academic
Screening Still Has Other Potential Harms
Even if the new estimate proves closer to reality, overdiagnosis isn’t the only downside associated with screening.
Mammograms can produce false-positive results.
A suspicious image can lead to additional imaging and sometimes biopsy before cancer is ruled out.
Earlier research has shown that repeated screening gives women multiple opportunities to experience a false-positive result over time. New England Journal of Medicine
Screening decisions therefore still involve balancing several different considerations rather than focusing on one number.
And Mammography Has Changed Since Many of These Trials
There’s another important complication.
The classic randomized mammography trials were conducted across very different eras of imaging technology.
Modern digital mammography—and increasingly digital breast tomosynthesis—differs substantially from some of the technology used in the earliest trials.
Breast cancer treatment has changed dramatically too.
So historical randomized trials remain enormously valuable, but they don’t perfectly recreate the screening environment women encounter today.
That’s another reason precise estimates should be interpreted carefully.
Why the New Study Matters
Overdiagnosis is unusual because it can’t usually be identified directly in an individual patient.
Once a cancer has been discovered and treated, we can’t rewind time and observe what would have happened had it remained undiscovered.
Researchers therefore have to infer overdiagnosis at the population level.
That makes methodology enormously important.
A change in assumptions about:
follow-up,
lead time,
screening frequency,
or what happened to control participants
can substantially change the final estimate.
The new paper’s contribution is essentially to place the major randomized trials into one consistent analytical framework rather than interpreting each using different assumptions. OUP Academic
From “Up to Half” to Potentially “Less Than One in Twenty”
That’s the number people are likely to remember.
Some previous discussions placed mammography-associated overdiagnosis as high as 30–50%.
This new analysis finds the randomized trial evidence compatible with less than 5%. OUP Academic
If further research supports that interpretation, it would substantially change how one of mammography’s major potential harms is communicated.
But the researchers themselves frame the findings around improving balanced clinical and policy discussions, not claiming that screening is free of harm. OUP Academic
The Bigger Picture
The debate over mammography has never simply been about whether screening can detect breast cancer.
Of course it can.
The harder question is whether finding cancer earlier produces enough benefit to outweigh false positives, additional testing, overdiagnosis and treatment of cancers that might never have become dangerous.
For decades, the estimated size of that last problem has varied wildly.
This new analysis offers a provocative explanation:
some of the apparent excess cancer attributed to mammography may have been early diagnosis mistaken for overdiagnosis because researchers weren’t comparing screening exposure and follow-up consistently.
That doesn’t close the debate.
But it changes an important part of it.
And for millions of women making decisions about breast-cancer screening, getting that number right matters enormously. OUP Academic
For more intriguing scientific discoveries, check out how scientists mapped cat cancer genetics on an unprecedented scale, or see what had people talking about Usha Vance’s return to the spotlight and Madonna’s recent VMAs appearance.
