“Mammograms Validated as Key in Cancer Fight” said the New York Times last Thursday. But the Chicago Tribune said “Mammography in Question.” Both articles covered the same study. I wondered who was off-message. So I read the study. In my opinion, the Tribune had it right.
The study starts from the observed decline in the breast cancer death rate between 1990 and 2000. The researchers asked how mammograms compared to chemotherapy and tamoxifen in contributing to the decline. According to their statistical models, mammograms accounted for 15% of the reduction in the death rate while chemical therapy accounted for 18%. Remember those numbers.
Now, while that’s not nothing, it’s also not much of a payoff for the cost, especially in the effect on the quality of a woman’s life. 90% of positive mammograms are false positives, which means that 90% of women with positive mammograms don’t have cancer but will have to go through a bunch of invasive diagnostic procedures and carry the burden of stress that those procedures create and the stress created by the uncertainty itself.
It gets worse when you go beyond the study’s narrow focus. For example, the study equated “screening” to “mammography” and equated “therapy” to “chemotherapy and tamoxifen.” HELLO! Heard of breast self-exams? A highly regarded Canadian study showed that self-exams are at least if not more effective than mammograms in reducing breast cancer deaths. And who said that chemotherapy and tamoxifen are the only therapies?
I know something about creating statistical models. One of the things I know about is called the misspecification error: if you don’t put the right variables into the model, you won’t get valid results. How many breast cancer deaths were avoided because someone felt a lump during a routine breast self-exam? You won’t find out from this study because it doesn’t have the whole picture.
It also doesn’t have the whole picture because it looks only at breast cancer deaths. Critics of mammographic screening argue that the mammogram itself and the diagnostic procedures that follow can disrupt a tumor, promoting its proliferation to other organs. A woman might survive breast cancer only to be taken by another cancer. Or something else.
Something else I know about statistics is called explained and unexplained variation. Remember the 15% and 18% reduction in the death rate attributed to mammograms and chemical therapy? That’s explained variation. Totalling 33% or 1/3. Which leaves 67% or 2/3 of the reduction in the breast cancer death rate unexplained. For my money, I want to know what’s in that 67%.
The cancer establishment’s “mammograms save lives” campaign seems intended quite frankly to create fear. Don’t you want to know? If we act early, we have a better chance to save your life. Don’t you want that? Yet researchers in the middle of this very study acknowledged that early detection might have no effect at all on death rates. They also acknowledged that 30%, again about 1/3 of the tumors discovered by early detection resolve themselves without any intervention at all.
Perhaps mammograms do save lives. I’m not convinced. But what we do know is that positive mammogram results scare the hell out of women and put them through a bunch of invasive procedures. Most of it unnecessarily.
What concerns me is not screening, but overdiagnosis and overtreatment from mammograms. Screening with self-exams is an entirely different matter because that’s about paying attention and using good sense. And that’s what really saves lives.
The issues in this article are developed (with references) in issue #2 of the Progressive Health Observer in the article titled “Too Few Mammograms.”