Overdiagnosis and mammography screeningBMJ 2009; 339 doi: http://dx.doi.org/10.1136/bmj.b1425 (Published 09 July 2009) Cite this as: BMJ 2009;339:b1425
- H Gilbert Welch, professor of medicine
- 1VA Outcomes Group and the Dartmouth Institute for Health Policy and Clinical Research, White River Junction, VT 80302, USA
The NHS recently scrapped its leaflet inviting women to undergo mammography in response to criticisms that it failed to mention the major harm of screening—overdiagnosis.1 In the linked systematic review (doi:10.1136/bmj.b2587), Jørgensen and Gøtzsche provide evidence that screening has led to overdiagnosis of breast cancer not only in the United Kingdom, but also in Canada, Australia, Sweden, and Norway.2
Overdiagnosis refers to the detection of abnormalities that will never cause symptoms or death during a patient’s lifetime. Overdiagnosis of cancer occurs when the cancer grows so slowly that the patient dies of other causes before it produces symptoms or when the cancer remains dormant (or regresses). Because doctors don’t know which patients are overdiagnosed, we tend to treat them all. Overdiagnosis therefore results in unnecessary treatment.
With the advent of widespread efforts to diagnose cancer earlier, overdiagnosis has become an increasingly vexing problem. Overdiagnosis is a widely recognised problem in prostate cancer screening, and it has also been reported in other cancers, including neuroblastoma, melanoma, thyroid cancer, and lung cancer. Some degree of overdiagnosis is likely to be the rule rather than the exception in cancer screening.
Jørgensen and Gøtzsche’s results are consistent with a growing body of observational evidence that screening mammography is associated with sustained increases in the incidence of breast cancer in women of screening age, with little or no subsequent decrease in incidence in older women.3 4 5 6 One cohort study concluded that some invasive breast cancers detected by screening must spontaneously regress.7
But legitimate questions can always be raised about the role of confounding in inferences based on observational data. The most compelling evidence to date, therefore, is the long term follow-up of the randomised controlled trial by Zackrisson and colleagues that was published in the BMJ three years ago.8 At the end of the 10 year trial, 150 more women were diagnosed with breast cancer in the mammography group than were diagnosed in the control group. Such an excess is expected—for mammography to work, it must advance the time of diagnosis for some women and lead to more women being diagnosed in any discrete period after its initiation. The researchers followed the women for another 15 years, during which time both groups received the same amount of mammography, so that cancers in the control group would have had the chance to “catch up.” But after a total of 25 years, there were still 115 extra women diagnosed in the group originally randomised to mammography. Unless mammography itself causes cancer, this persistent excess is strong evidence for overdiagnosis.
The question is no longer whether overdiagnosis occurs, but how often it occurs. Jørgensen and Gøtzsche conclude that about one in three of all breast cancers detected represent overdiagnosis. The corresponding number from Zackrisson and colleagues’ study is one in six.9 But these may not be the most useful numbers from the users’ perspective.
Mammography is one of medicine’s “close calls”—a delicate balance between benefits and harms—where different people in the same situation might reasonably make different choices. Mammography undoubtedly helps some women but hurts others. No right answer exists, instead it is a personal choice.
To inform that choice, women need a simple tabular display of benefit and harms—a balance sheet of credits and debits (see table⇓ for a draft version).
The cumulative risk of a false positive mammogram result varies widely on the basis of geography, but women largely accept this risk.11 We do not know how women feel about being diagnosed at a younger age without this influencing their prognosis (those destined to die still do, those destined to survive would have done just as well if diagnosed later).
The information that will probably influence most women’s choice will be data on the trade-off between the number of deaths from breast cancer avoided and the number of cancers overdiagnosed. More research is needed to confirm or dispute this assertion and to determine how sensitive women’s choices are to various estimates of the trade-off.
Equally important are the estimates themselves. Zackrisson and colleagues reported 62 fewer deaths from breast cancer and 115 women overdiagnosed—a ratio of one death avoided to two women overdiagnosed. Recently, Gøtzsche and colleagues argued in the BMJ that the ratio is one to 10.12 For many women, the tipping point may be within this range. Careful analyses that explicitly lay out their assumptions and methods, which will improve the precision of these estimates, are sorely needed.
Finally, researchers need to do more than just describe the problem, they need to work towards mitigating it. The amount of overdiagnosis is a function of the mammographer’s threshold to recommend biopsy. The time has come for a randomised controlled trial to test higher thresholds, such as only recommending biopsy for breast masses larger than a certain size.
Cite this as: BMJ 2009;339:b1425
Competing interests: None declared.
Provenance and peer review: Commissioned; not externally peer reviewed.