how should science compare the benefits of surgery and surveillance?


U.S. Army photo by Steven Hoover (Photo Credit: U.S. Army)

Last month, several Dutch newspapers wrote that preventive surgery was not needed to improve breast cancer survival among carriers of BRCA gene mutations. The coverage caused a stir among women who had opted for surgery and made me raise questions about the validity of the scientific study on Twitter. This week, Medisch Contact, a Dutch magazine for health professionals, tried to cover the commotion, but made it worse. Here is what I wish they had written to clarify the study and the (limited) scope of its conclusion.

The study sought to investigate whether mastectomy and intensive surveillance have different survival among carriers of the BRCA gene mutations. To answer this question, we ideally want to do a randomized controlled trial in which we randomly assign women to preventive surgery or intensive surveillance of annual mammography, MRI, and breast examination. As such study would never be ethically approved, the observational study is the go-to alternative, but we should never forget that we wanted a randomized trial.

There are two major obstacles in using observational data. The first is that choices that would have been randomized in a trial are not random in real life. Opting to take a DNA test and subsequently choosing between preventive surgery and surveillance are two choices that are not decided on coin tosses. Women may be more likely to opt for a DNA test when their families are more seriously affected by breast cancer, and, within this group, women who choose mastectomy may come from families in which more women died from cancer. As a result, they may be more likely to develop cancer and to have worse prognosis than women who opt for intensive surveillance.

The second is that observational data tend to be ‘messier’. Women enter the study at different ages and may drop out before the study ends. They undergo DNA testing and choose between surgery and surveillance when they are ready, which is at various ages. These variations in follow-up time and timing of interventions complicate the statistical analysis. Researchers have to make choices how to best deal with such limitations of their data.


Here, the researchers used data from a family-based study of hereditary cancer in which women (and men) were eligible to participate if they had a family member who had cancer and undergone DNA testing for suspected BRCA mutations. The study population thus consists of women who 1) had a DNA test and cancer at the start of the study; 2) did a DNA test before some of them developed cancer; or 3) did no DNA test or only after they developed cancer (see Figure).

Participants in the breast cancer survival study.

The only group of women for whom the researchers knew both their DNA status and their preventive strategy were those who had undergone DNA testing before some of them developed cancer. They considered this group as their study population, declared the date of the DNA testing as the study baseline, and removed all other women, including 2000 BRCA carriers who had breast cancer, from the analysis.

To be sure, leaving these 2000 women in the analyses while being unable to include women who never did a DNA test is wrong, it will underestimate the survival in the surveillance group, but that doesn’t automatically make every alternative right.

Excluding women should be justified by checking if those who opted for DNA testing were equally likely to develop cancer as women who didn’t opt for testing, and if those who choose mastectomy were equally likely to develop cancer as those who choose surveillance.

If both groups are comparable, then the observational study comes close to the randomization in the trial that we are unable to perform. If we find that these groups are not comparable (which I think is likely) then the data are not suitable for comparing the survival between surgery and surveillance.

Apart from these practical analytical problems, I also have a more fundamental concern. The probability of surviving breast cancer is a composite of two other probabilities, namely: the probability of getting breast cancer and the probability of not dying from breast cancer once you have it (Figure).

Preventive surgery removes most breast tissue and therewith substantially reduces the risk of getting breast cancer, but what does preventive surgery do to the probability of not dying from BC once the cancer is detected? Surveillance helps detecting cancer in earlier stages, but what does surveillance do for surviving once the cancer is detected?

Survival is a relevant piece of information in choosing between surgery and surveillance, but it misses that the trajectories to survival differ between the two alternatives.

Mastectomy prevents women from developing the disease. Even in this population of young women, approximately 25% of the carriers in the surveillance group developed breast cancer within the ten years of the study, 5–10% of whom died. How many more will get breast cancer and die if the study follows them for another ten years?

Show More

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button

Adblock Detected

Please consider supporting us by disabling your ad blocker