Eight mammography trials, re-read against Danish data, put screening overdiagnosis below 5 percent
A meta-analysis in JNCI Journal of the National Cancer Institute reanalysed all eight mammography screening trials against two Danish regional programmes and found that the excess of detected cancers matches the Danish level, where overdiagnosis is put below 5 percent. Sisse Helle Njor of the University of Southern Denmark and her co-authors argue that earlier estimates of 30-50 percent were built on trial data that had not matured; the work generates no new screening data of its own.
Science··Morning
Eight trials read against a Danish yardstick
The study, published in JNCI Journal of the National Cancer Institute and announced on 13 September, comes from a team led by Sisse Helle Njor of the University of Southern Denmark and Lillebælt Hospital, with co-authors at Queen Mary University of London, the University of Copenhagen and the American Cancer Society. It pools every trial in the field that compared women offered mammography with women who were not: the New York Health Insurance Plan, Malmö, Two-County, Edinburgh, the Canadian National Breast Screening Study, Stockholm, Gothenburg and UK Age. Overdiagnosis means finding a cancer that would never have given symptoms or shortened a woman's life; estimates drawn from those trials have ranged as high as 30-50 percent of screen-detected cancers, and those figures shaped guidelines and the information women receive before screening. As a reference the researchers used two Danish regional programmes, where organised screening began 17 years earlier in some regions than in others, so the course of diagnoses after screening starts can be followed over the long term. The analysis counted invasive breast cancer together with ductal carcinoma in situ, the non-invasive form confined to the milk ducts.[1], [2]
Three adjustments turned a 30-50 percent estimate into under 5 percent
When screening is introduced, diagnoses first rise because cancers are found earlier than they would have been; over time that rise should be followed by a fall, as the same cancers would otherwise have been diagnosed later. Elsebeth Lynge of the University of Copenhagen, a co-author, says the pattern is distorted when women in either arm keep being screened after a trial ends, which was common, and that an analysis ignoring this can mistake the early rise for overdiagnosis. The team therefore adjusted for three things: whether women in the comparison arm were later screened, for example once a routine programme began; how many screening rounds each arm was offered, since more rounds give more chances to detect a cancer early; and how long each arm was followed, since a longer follow-up gives cancers detected early in the screened arm time to surface in the other. With those differences accounted for, the excess of detected cases in the trials closely matched what the Danish programmes would predict, where overdiagnosis is estimated at below 5 percent.[1]
What the authors say it means for women invited to screening
Matejka Rebolj of Queen Mary University of London says some of the high estimates that influenced screening guidelines rested on evidence gathered before the trial data had matured, and that read in their full temporal context the trial data are consistent with overdiagnosis of less than 5 percent rather than close to 50 percent. Njor adds that most women will not develop breast cancer, and that in her reading the benefit of early detection and of preventing premature death outweighs the small risk of unnecessary treatment; the authors hope the work offers a framework for telling invited women more realistically what screening can and cannot do. The analysis is a re-reading of existing trials and produces no new screening data of its own. Funding came from the Novo Nordisk Foundation for co-author Casper Urth Pedersen and from Cancer Research UK for Rebolj.[1], [2]
