Mammograms on Trial

It’s been medical gospel for years: All women 50 years and older–and younger women in high-risk groups for breast cancer–should undergo mammography once every year or two. But recent reviews of the screening procedure have disagreed about whether routine mammograms help prevent breast cancer deaths. In the face of that scientific uncertainty, what’s a woman to do? There’s even disagreement among doctors and the groups that make policy recommendations.

Researchers performing the recent reviews reanalyzed data from often-cited mammography trials. Some of these teams concluded there isn’t enough evidence to support the claim that routine mammography has lessened deaths from breast cancer. Other reviewers of these same trials came to the opposite conclusion–that mammograms do in fact save lives. One group recently expanded its recommendation to include women 40 and older, not just women 50 and over, in the group that should get regular screening.

“Everyone agrees that mammograms detect early tumors, smaller tumors, and give women more treatment options,” says Peter Greenwald, director of the division of cancer prevention at the National Cancer Institute in Bethesda, Md. “The question is whether it saves more lives in the long run.”

The screening procedure has a variety of financial and psychological costs, although the X-ray exposure of a mammogram isn’t considered a significant danger. Mammography screening’s overall expense in the United States–including follow-up procedures–runs into the billions of dollars. During screening, 5 to 10 percent of mammograms show potential abnormalities, requiring doctors to order additional mammograms, ultrasound imaging, biopsies, and other tests. However, about 97 percent of women in their 40s and about 86 percent of women age 50 and over who have a suspicious initial mammogram turn out not to have breast cancer.

Although the debate over screening has splashed onto the front pages of national newspapers and magazines several times in the past 6 months, the controversy goes back decades. Questions about studies have simmered since the beginnings of routine screening by mammography in the 1970s. The heat intensified over the past 2 years, after Danish researchers published a troubling report in the British medical journal The Lancet. Their article challenged the value of screening with mammograms.

This view has been hotly contested by other researchers, federal agencies, the World Health Organization, and advocacy groups such as the American Cancer Society.

In the United States, breast cancer now strikes about 192,200 women and kills about 40,600 women each year. That reflects a 15 percent drop in breast cancer mortality over the past decade, Greenwald says, “so we’re cautious about pulling out something [screening with mammography] we think contributes to that benefit.”

As the debate rages over specific issues of statistics, physicians struggle to figure out what to tell their patients, and national organizations are rethinking their recommendations. The trouble is, more science may not be the answer.

“There’s an assumption in American society that if you just do enough scientific studies and you try hard enough, you can come up with definitive answers,” says Barron Lerner of Columbia-Presbyterian Medical Center in New York. “In the world of breast cancer, we’ve learned that even trying as hard as you can, you can’t necessarily get answers.”

Old data, new concerns

Both the previous recommendations and the current uncertainty surrounding mammography emerge from all or some of the same large trials conducted since 1963: four in Sweden and one each in the United States, Scotland, and Canada, though because the Canadians separated data from women in their 40s and women in their 50s, that trial is sometimes considered as two. Each study randomly assigned women to receive either mammogram screening or not, the latter being the control group. The scientists statistically analyzed their findings to see whether differences in breast cancer deaths between groups were large enough to show that the screening, rather than chance, was responsible for the overall result.

Two trials found that screening led to significant drops in the number of women dying of breast cancer, others found more-modest benefits that didn’t meet scientists’ customary standard for significance. Two trials even came up with apparent increases in breast cancer deaths among women receiving mammography, but these findings weren’t statistically significant and are considered simply the result of chance.

Even these large trials may not enroll enough women to get to the bottom of the matter. Out of a thousand healthy women, only a few will develop breast cancer and die from it. So, very large numbers of healthy women are needed in studies to detect significant differences between screened and unscreened women. One way to deal with the relative rarity of deaths would be to consider several trials together. For example, in a recent reanalysis of several Swedish trials, 511 of 129,750 women (0.4 percent) who received routine mammography died of breast cancer, compared with 584 of 117,260 women (0.5 percent) who were not screened. These data were statistically significant, the scientists calculated.

However, combining data poses difficulties because each trial was designed in a different way. Some trials included manual breast exams along with mammography. Some rescreened women every 12 months, while others checked them every 28 months. The number and angles of X rays taken during the screening also varied across the different trials. Thus the tests’ capability to detect cancers in different parts of the breast may have varied. Finally, some of the control groups got no breast exams during the study, while others were routinely examined by physicians.

There are statistical techniques, however, for pooling and comparing trials in what scientists call a metanalysis. Such techniques originally put the value of mammography screening at about a 30 percent overall reduction in breast cancer deaths. That’s the figure that most breast cancer specialists and national health associations have used over the years.

That general agreement was thrown into turmoil in 2000 when Peter C. Gtzsche and Ole Olsen of the Nordic Cochrane Center in Copenhagen published their article.

They argued that troubling inconsistencies in the trials were enough to make them useless for recommending mammography as a routine screen for breast cancer. It wasn’t that the trials proved that there is no benefit, they said, but merely that the quality of evidence available is insufficient to prove a benefit.

Gtzsche and Olsen had a variety of specific objections to most of the studies. For example, the researchers pointed out that in several trials, women who unknowingly had cancerous tissue when the study began were removed from the trial when their first screening showed their tumors. However, some of the unscreened group probably also had preexisting breast cancer, which often would not have been detected early, and so these women would have remained in the trial. This difference would be expected to increase the apparent number of deaths from breast cancers recorded for the unscreened group, and so make the mammograms look more effective than they really are.

Raising another objection, the Danish scientists eliminated any study in which they found even a small difference in average age between the control and screened groups. Such differences, they said, might reflect an imperfect procedure for randomly assigning women to the groups. Their concern is that unknown differences among the women, rather than the presence of screening, might underlie findings of fewer deaths from breast cancer in the screened group.

For these and other reasons that scientists continue to debate, the Danes threw out data from 80 percent of the women followed in the studies. Only the Canadian and one Swedish study–neither of which showed a significant effect of routine mammography–met the Danes’ standards.

A flurry of letters disagreeing with the researchers’ paper was published in the Feb. 26, 2000 Lancet: Responding to those critiques, Gtzsche and Olsen reaffirmed their conclusions in the Oct. 20, 2001 Lancet.

In that response, they added yet another problem that they had detected in several of the trials: When researchers in the original studies were uncertain about the cause of a woman’s death, they were more likely to ascribe it to breast cancer if the woman was in the control group than if she was in the screening group. As an example, the death of a woman who dies of a tumor in her lung might be attributed to the spread of breast cancer if she was in the unscreened group but to unrelated lung cancer if she had been regularly screened with mammograms.

In the Feb. 6 Journal of the National Cancer Institute, other researchers point out similar problems in assigning the cause of death in the breast cancer trials and in tests of screening for lung and colon cancer.

More voices

Prompted in part by the work of Gtzsche and Olsen, a U.S. group also reevaluated the large studies since 1963. The U.S. Physicians Data Query (PDQ) Screening and Prevention Editorial Board is an independent body that evaluates the quality of evidence on medical-screening topics. It posts this information on the National Cancer Institute’s Web site, although it isn’t an advisory board to the agency. Previously, the group had noted a benefit of screening but said that the magnitude of that benefit wasn’t clear. At a meeting last month, the board indicated that its most recent review raises uncertainty about whether there’s a benefit to mammogram screening. The PDQ report can be found at http://www.nci.nih.gov/cancer_information/doc_pdq.aspx?version=provider&viewid=b906d0d0-63ac-4d55-ac29-2ae992440adf.

The board’s change in position comes in part from some of the same problems that Gtzsche and Olsen raised, says Donald Berry of the University of Texas M.D. Anderson Cancer Center in Houston and a member of the board. However, the PDQ board highlighted some additional concerns about how mammography has gained favor. Many doctors, for instance, are convinced of the value of mammograms by data showing that the screening prolongs the life of women who have breast cancer, says Berry.

He raises concerns about errors that could creep into a screening trial and undermine measurements of how long women with breast cancer survive. “Since women whose breast cancers are found by mammography [appear to] do so much better, there is a tendency to attribute the benefit to mammography,” Berry says. “This is a mistake that leads to profound misconceptions.”

Consider two women in a trial whose cancer begins on the same day and who both die 10 years later. If one of them has been part of a control group, her cancer may not be detected until near the time she dies. But if the other is screened regularly with mammography, her disease will probably have been detected much earlier, so it will appear as if the screening has prolonged her survival with cancer.

Moreover, Berry says, mammography can detect cancers that would never have killed a woman. If two women with such cancers die of other causes, the trial would probably detect the breast cancer of the one being screened but not that of the woman in the control group. This skews the statistics on length of survival, Berry says. The woman whose cancer was found may also undergo unnecessary treatments with their accompanying side effects.

Another group, the U.S. Preventive Services Task Force, began to review the data on mammography’s benefits in 1999. When members of this independent body that advises the federal government on preventive health issues met this year to continue their review, they considered the criticisms that Olsen and Gtzsche had raised.

“We found the same flaws,” says Janet D. Allan, vice-chair of the task force and dean of the School of Nursing at the University of Texas Health Science Center in San Antonio.

“They interpreted the flaws as being fatal flaws,” she says. “We did not interpret the flaws as fatal . . . and concluded that the studies were still valid and that mammography screening reduces deaths from breast cancer.” The task force’s reanalysis of the mammography trial data, which hasn’t yet been published, shows a 23 percent reduction in breast cancer deaths among the women who were screened. The group’s report and recommendations are available at http://www.ahrq.gov/clinic/3rduspstf/breastcancer/.

Another new analysis of the Swedish data–adding information from the country’s large death-registry system–has considered information on each of the women in most of the trials and from some additional women.

Among 247,000 women who were 40 to 74 years old when they entered the trials, and roughly half of whom were regularly screened, regular mammography led to a 21 percent reduction in deaths from breast cancer in the more than 20 years since the trials began, says Lennarth Nyström of Ume University in Sweden. The data are published in the March 16 Lancet.

Dealing with uncertainty

Right now, most policy groups–including the National Cancer Institute, the American Cancer Society, and the American College of Obstetricians and Gynecologists–continue to recommend regular mammograms for women over 40.

Because large-scale trials to test breast cancer screening each take more than a decade and cost millions of dollars, policy makers can’t wait for scientists to collect additional data. Further, both supporters and skeptics of mammogram screening say that new studies may not be feasible. It would now be difficult, if not unethical, to prevent women in the control groups from getting mammograms.

In the near term, three things can shed light on mammograms’ value for screening, Berry says. First, many researchers have called for independent audits of the Swedish data, some of which is controversial. Second, a U.S. survey of changes in overall breast cancer mortality is now under way to pin down benefits of screening. It’s intended to apportion the relative contributions of mammograms and new therapies on recent improvements in breast cancer survival. Third, new molecular techniques to identify which tumors are likely to progress slowly may reduce overtreatment of cancers detected early, thereby changing the balance of benefits and risks from screening.

There’s one sure point of agreement among investigators who support screening and those who don’t: Mammograms should not be the main thrust of future breast-cancer research. There are already new breast-screening methods coming into use. It would be better, researchers say, to focus on ways to prevent breast cancer from occurring and to develop more early ways of detecting and appropriately treating this disease.

.

.

.

.

.

.

.