AIM: Does screening mammo ignite an epidemic of overdiagnosis?

A study, which breast imaging stakeholders have characterized as underpowered and statistically sloppy, has concluded that 15 to 25 percent of breast cancer cases diagnosed in the Norwegian Screening Program are overdiagnosed, according to research published April 3 in Annals of Internal Medicine. These study results run counter to published data about organized screening programs, which report up to a 40 percent decrease in the death rate among women who participate in screening, Michael N. Linver, MD, director of mammography at X-Ray Associates of New Mexico in Albuquerque, told Health Imaging.

In the study, Mette Kalager, MD, from the department of epidemiology at Harvard School of Public Health in Boston, and colleagues sought to estimate the percentage of overdiagnosis of breast cancer attributable to mammography screening.

“Overdiagnosis creates a substantial ethical dilemma and burdens the patient and the healthcare system,” wrote Kalager and colleagues.

The researchers aimed to refine previous estimates of overdiagnosis, which range from 0 to 54 percent. They leveraged the Norwegian Screening Program, a nationwide breast cancer program implemented in a staggered geographic fashion over a nine-year period, allowing them to compare contemporaneous trends.

The bi-annual screening program was launched in four counties populated by 40 percent of the eligible population (women between the ages of 50 and 69 years) in 1996, and gradually introduced to the entire country by 2005.

Kalager and colleagues limited their focus to women diagnosed with invasive breast cancer and analyzed four groups comprising women older than 29 years and younger than 85 years.

These are: a current screening group comprised of women residing in screening counties between 1996 and 2005, the current nonscreening group comprised of women living in counties that had not yet implemented screening between 1996 and 2005, the historical screening group comprised of women living in counties between 1986 and 1995 that had implemented screening in the following 10 years and the historical nonscreening group comprised of women living in the counties between 1986 and 1995 that did not implement screening in the following 10 years.

A total of 39,888 women with invasive breast cancer were included in the analysis, 7,793 of whom were diagnosed after the screening program started.

The researchers compared current and historical nonscreening and screening groups to account for temporal changes in breast cancer incidence. They used two approaches to adjust for lead-time bias. Approach 1 compared incidence rates in the screening and nonscreening groups for women ages 50 to 79. Approach 2 excluded cancers detected in the first screening round.  

Kalager et al reported breast cancer incidence increased by 18 percent over 10 years among women not invited to screening. Women in the current screening group had a 47 percent increase in incidence compared with women in the historical group.

When the researchers applied their model, they reported the percentage of overdiagnosis to be 18 to 25 percent for approach 1 and 15 to 20 percent in the approach 2 analysis. They also found lower incidence of stage III and IV in current groups than in historical groups. However, the decrease in advanced stage disease was similar in both the screening and nonscreening groups.

“The similar results from our two analytic approaches support the validity of our methods,” wrote Kalager and colleagues.

Linver disagrees. “The data overwhelmingly support screening.” However, he admits, “It’s true that some cancers are overdiagnosed and would never progress. But if the death rate has dropped by 40 percent [in organized screening programs,] obviously radiologists are findings a lot of cancers that would have progressed.”

Linver pointed out the importance of long-term follow-up. “Ten-year follow-up does not compensate for the deficits among women not undergoing screening. Researchers need a minimum of 20 years of follow-up for conclusions to begin to be considered valid.”

Indeed, when Laszlo Tabar, MD, from the departments of mammography, surgery and pathology at Falun Central Hospital in Falun, Sweden, and colleagues, analyzed 29 years’ worth of screening data in Sweden, they reported a sustained 30 percent mortality reduction, with the majority of deaths prevented occurring after the first 10 years of follow-up.

Linver reinforced the value of screening, noting that a re-analysis of the impact of U.S. Preventive Services Task Force guidelines indicated the shift to bi-annual screening beginning at age 50 would result in additional 7,000 deaths per year.

In an accompanying Annals editorial, Joann G.  Elmore, MD, MPH, from University of Washington School of Medicine in Seattle, and Suzanne W. Fletcher, MD, from Harvard Medical School in Boston, offered that overdiagnosis may occur more frequently in the U.S. than in Norway because U.S. radiologists are more likely than their European colleagues to report abnormalities and U.S. women often begin annual screening at age 40.

Elmore and Welch suggested the “ethical responsibility” to alert women to overdiagnosis.

Linver provided different advice. “Radiologists need to have a counterattack and list the reasons why these results are untrue.”