Map of life expectancy at birth from Global Education Project.

Friday, February 14, 2014

Cognitive bias

By now most people have heard about the Canadian breast cancer screening trial, just published in BMJ. Hate to say "I told you so," but I did: screening mammography of the general population of women is a bad idea. This ought to be the nail in the coffin, but it won't be.

So here's the 4-1-1. Back in the 1980s, Canadian researchers randomized almost 90,000 women either to get mammographic screening every year for 5 years; or not. Twenty five years later, there was no difference in the mortality rate from breast cancer between the two groups. None. Zip. Zero. In fact, more of the women who were 40-49 who had been screened died than those who had not, although the difference was not statistically significant. But, there were 106 more cancers diagnosed in the screening arm. Doing the math, 22% of diagnosed cancers represented "overdiagnosis," i.e. cancers that would never have become clinically significant, in other words those women underwent unnecessary and in fact very harmful surgery, plus chemotheraphy and/or radiation.

This is a randomized controlled trial with long-term follow up. It is the strongest possible study design. But it confirms what we have already been seeing from epidemiological studies. Now don't get me wrong -- this is a study of indiscriminate screening of the total population of women. The math might be different for women at higher risk but it also might not be. It turns out that with modern treatment, the difference between a lesion detected through screening and one detectable through physical examination might be too small to make a difference, in other words early detection isn't important after all. That's what the evidence seems to be saying.

By coincidence, in the New England Journal of Medicine, appearing the day before, Lisa Rosenbaum, M.D., discusses the meaning of breast cancer within the culture, and particularly among women. She starts off wondering why women seem to be so much more worried about breast cancer than they are about heart disease, while in fact heart disease is overwhelmingly a more common cause of death, and far more prevalent. (We are talking 400,000 deaths every year from heart disease, vs. about 40,000 for breast cancer, in other words 10 to 1.) Then she says this:

In 2009, the U.S. Preventive Services Task Force recommended decreased frequency of mammography for most women younger than 50 years old, noting that the potential harms outweighed the benefits. Although the recommendations were based on an unbiased review of decades' worth of data, a public outcry ensued. The recommendations were criticized as an assault on women's health, and a 2009 USA Today poll found that 84% of women 35 to 49 years of age planned to ignore them.

So intense was the outrage over these evidence-based recommendations that a provision was added to the Affordable Care Act specifying that insurers were to base coverage decisions on the previous screening guidelines. Rather than acknowledge this blatant dismissal of new guidelines, many political leaders, physicians, and advocacy organizations argued that we simply didn't have enough data to justify the new recommendations. But data have shown for years that early mammography screening doesn't save lives, just as data show that preventing heart disease, through certain lifestyle modifications and appropriate use of medications, does. So why do we resist these data?

Her answer is essentially that preoccupation with breast cancer is a symbol of female solidarity and a feminist cause -- it's a marker of tribal identity, in a sense. That's part of it, but it is also the case that the American Cancer Society represents the interests of surgeons, oncologists, radiologists, and pharmaceutical manufacturers -- that's who finances it. And they all make money off of screening and ensuing unnecessary treatment. So they scream and yell that this cannot possibly be true and they lobby congress, successfully.

They also routinely misinform women, as by for example making claims about five year survival being better for cancers detected through screening. As they well know, this is caused by what's called lead time bias and ascertainment bias. Lead time bias means that, if the cancer is going to kill you, of course it will take longer if it's detected earlier; but that doesn't mean you wouldn't have died at the same time anyway, it's just that the diagnosis came earlier -- subjecting you to earlier futile treatment. Ascertainment bias means, as we have already seen, that many of the "cancers" found through screening are harmless, so of course you survive. You would have anyway, and you would have been better off because you never would have been told that you had cancer and you would not have had surgery and been poisoned and irradiated.

It is also true that it's very difficult for most people to perceive that doing less, and even knowing less, can actually be better.  Of course we want to know! Of course early detection is better! But it isn't.

However, this information is not going to shut down the industry. It probably won't make much difference at all. But, I told you so.

No comments: