A recently published study in the New England Journal of Medicine suggests that breast cancer screening is causing an increase in overdiagnosis that exceeds the number of actual cancer the process detects. This is not the first study to suggest that there are unintended consequences and hazards to mammography screening, but it does take a relatively straightforward approach to the question that is easier to understand.
In Brief: Screening
Screening is the process of performing a test or a scan on a large population that does not have symptoms of the condition in question. For breast cancer, this means performing a mammography on women who do not have genetic markers like the BRACA mutations, unexpected lumps, a family history, etc.
In Brief: Overdiagnosis
The inherent risk to breast cancer screening is that it is possible to find tumors that, if they had been left undiscovered, would either not grow, go into remission on their own (rare but happens), or grow so slowly that the patient would die of something else before symptoms start appearing. There is no way to tell from a mammogram how a tumor might or might not progress, so they are all treated as potentially dangerous. This can lead to patients undergoing procedures, costs, and risks for no true medical benefit. Although it is known for a fact that overdiagnosis occurs, how often it happens versus “true” cancer detections is a subject of ongoing debate.
The NEJM Study
This recent study is based on a simple premise. The purpose of mammography screening is to find small breast tumors before they can grow into bigger and more dangerous cancers. Therefore, if screening is doing what we think it does, then historical data will show an increase in small-tumor diagnoses and a decrease in large-tumor diagnoses over time.
To accomplish this task, the researchers drew on the Surveillance, Epidemiology, and End Results (SEER) program, which is basically a giant database of disease trends. They took data on breast cancer among women age 40+ from 1975 to 2012. Since widespread mammography use did not start until the early ‘80s, this gave data from both the pre- and post-mammography periods. Tumors were grouped by size in the following categories:
- in situ (basically a proto-cancer, sometimes called stage 0)
- Under 1cm
- 1-1.9 cm
- 2-2.9 cm
- 3-4.9 cm
- 5+ cm
The rates of different tumor types were parsed according to the portion of all diagnosis in their given year. This led to the following observations:
- in situ was 5% of all diagnosis in 1975 and 22% in 2010
- Under 1 cm was 6% in 1975 and 18% in 2010
- 1-1.9 cm was 26% in 1975 and 27% in 2010
- 2-2.9 cm was 26%% in 1975 and 15% in 2010
- 3-4.9 cm was 25% in 1975 and 11% in 2010
- 5 cm or higher was 13% in 1975 and 7% in 2010
- Tumors measuring 2 cm or less or were in situ were considered small, and any over 2 cm were considered large.
Now, looking at these percentages, it seems that screening is having the intended effect. The large category tumors all showed a drop in their share of diagnosis while the small tumors went up. This gets a bit more complicated, however, when you start looking at the underlying numbers.
From the start of screening, the annual number of large tumors dropped from 145 per 100,000 women to 115, or a 30 tumor decrease. The incidence of small tumors, however, rose from 82 per 100,000 women to 244, which is a 162 increase.
Assuming that the behavior of breast cancer is unchanged, these findings were interpreted to mean that, of the 162 small-tumor increase, only 30 would have gone on to grow clinically larger if not detected at the time. Therefore, the remaining 132 are an overdiagnosis.
Additionally, some inspection of the rates of large tumor mortality over the years suggests that treatment improvements, rather than screening, was responsible for about two-thirds of the decrease in deaths.
It’s worth pointing out that these findings aren’t perfect. Certain assumptions and estimations had to be made about the rate of tumor sizes early on when data from SEERS was more limited. There is also a lot more data and a lot more years of data available from the post-screening era than the pre-screening one.
What This Means
The study suggests that, while breast cancer screening can detect tumors that would have otherwise grown larger—and has as a result saved lives—this effect is dwarfed by improvements in treatment and the risk of overdiagnosis. Does this mean mammography screening does more harm than good? No idea, but that is an ongoing debate in which these findings will inevitably play a role.
Welch, H., et. al., “Breast-Cancer Tumor Size, Overdiagnosis, and Mammography Screening Effectiveness,” New England Journal of Medicine, 2016; 10.1056/NEJMoa1600249.