This conventional medical wisdom has been turned upside down by a reassessment of the world’s best mammography-screening trials. After a thorough analysis of each trial, two Danish researchers concluded that the trials do not provide reliable evidence that regular mammograms will reduce mortality. What’s more, mammography screening actually causes more harm than good. The findings, summarized last year in The Lancet (10/20/01), are an update of a review the same authors published nearly two years ago, also in The Lancet.

Ole Olsen and Peter C. Gotzsche, MD, of the Nordic Cochrane Centre in Copenhagen, Denmark, have explored something traditionally ignored in studies of screening tests. Mammography screening trials emphasize the number of breast cancer deaths among the participants. But Olsen and Gotzsche contend that that is not enough; you must also look at deaths from other causes as well. They found that many more women given regular mammograms are treated for breast cancer, and the treatment itself can cause fatalities.

A review like this, one that reverses so much accepted wisdom, will meet a lot of resistance. While there is disagreement over the value of mammography screening for women in their 40s, there has long been a consensus among medical organizations in the U.S. and Canada that regular mammograms can prevent some breast cancer deaths in women, aged 50-69 years. The widely quoted 30% breast cancer mortality reduction shown for this age group comes from the combined results of the same seven trials that were reviewed by Olsen and Gotzsche. (Since two of the seven trials found no benefit to mammography, pooled results are a way of resolving contradictory findings.) The seven trials, initiated in North America and Europe between 1963 and 1980, are regarded as the “gold standard” of medical research because they randomly assigned women to receive regular mammograms, or not. How can the Danish reviewers come to such opposing conclusions when drawing from the same clinical trials that so firmly established mammography screening as lifesaving? Naturally, the reviewers and their methods should be scrutinized.

Olsen and Gotzsche went well beyond what other reviewers would normally do. They analyzed the design and methods of each trial, and they found biases that favored mammography. Specifically, there were flaws in the randomization process and how the causes of death were determined. Here’s an example of the latter. Trial records show that it is not unusual for elderly women to have more than one cancer; and trial investigators might be less likely to attribute the cause of death to breast cancer in a study participant known to have had her tumor found early on a mammogram. Mistakes in identifying cause of death, even if they occur in a minority of cases, can skew results in a trial showing a 30% benefit to screening.

After looking at all the important elements of a good clinical trial, Olsen and Gotzsche concluded that five of these trials had so many flaws that their breast cancer mortality reduction conclusions could not be substantiated. None of the seven were judged to be perfect, but the two given the highest methodological ratings were conducted in Malmö, Sweden and in Canada. These two are the only trials that did not find a lower breast cancer death rate among mammography-screened women. The Canadian trial has been criticized repeatedly in the U.S. media, ever since it announced this unpopular finding in 1992. Still, Olsen and Gotzsche found it to be of much higher quality than the five that purportedly proved mammography’s life-saving benefit.

Should we trust this new review? The peer review process, which is employed by the top medical journals, is intended as a guard against the publication of substandard work. Two or three peers, or experts in the same field, are asked to critique a review or a study to see if it should be accepted for publication. The authors must, in turn, defend their work and correct any errors.

The peer-review process that Olsen and Gotzsche underwent is arguably more rigorous than that of the leading medical journals. They belong to a highly respected international non-profit organization called the Cochrane Collaboration. It started in 1992 with a mission to review all relevant randomized controlled clinical trials to determine whether a treatment, or a screening procedure, is effective. Today, the Cochrane Collaboration includes over 6,000 researchers, physicians, statisticians, consumer advocates, and epidemiologists in over 60 countries. Most are based at universities and medical schools.

What makes the new review different from all others is the philosophy of the Cochrane Collaboration. It encourages reviewers to look at results other than those intended by the people who designed the trials. When Olsen and Gotzsche compared the number of treatments in the screened and the unscreened groups, they found many more treatments were given to the mammography-screened women. They had, for example, about 20% more mastectomies. Many of these treatments, including radiation therapy, were unnecessary, as Olsen and Gotzsche observed that screening identifies some slow-growing tumors that would never develop into a lethal cancer in a woman’s remaining lifetime. Even the microscopic mammography-detected lesions called ductal carcinoma in situ are treated with lumpectomy plus radiation, or a mastectomy, though the majority would never become life threatening had they gone undetected.

The higher number of treatments among mammography-screened women would be acceptable if it led to many more lives saved. But this isn’t what Olsen and Gotzsche found. Some trials actually showed an increase in deaths in the women, aged 40-49, who had received mammography screening. Olsen and Gotzsche suspect that the deaths could be treatment related, for example, heart damage as a result of radiation therapy. To support their hypothesis, they cite an overview of 40 clinical trials involving women who had been treated for early breast cancer 10-20 years earlier (Lancet 2000; 355:1757-70). The overview, conducted by the Early Breast Cancer Trialists’ Collaborative Group, showed that radiation therapy reduced the annual death rate for early breast cancer by 13%, but increased the death rate from other causes by 21%. Much of the latter death rate was attributed to cardiovascular and cerebrovascular causes.

The new review by Olsen and Gotzsche has ramifications for all other screening tests, since none are as thoroughly studied as mammography. It raises the most obvious question about the standard cancer-screening advice to the public: Are we better off not subjecting ourselves to screening tests and, instead, promptly seeking treatment once a symptom appears?

Maryann Napoli, Center for Medical Consumers