Breast Cancer

Not so long ago the Agency for Healthcare Research and Quality  created a firestorm of criticism by announcing their new guidelines to clinicians for breast cancer screenings, in particular, mammogram screenings. Check out the full report by clicking on this link.

If you are one of those who were shocked by the recommendation to reduce cancer screenings, after decades of hearing the opposite recommendation from every conceivable source, read on. I intend to explain the science and math behind your reaction.

Finally, here is a news item that justifies my motivation for creating this blog: the science behind the news. The most controversial part of the proposal is the recommendation against mammogram screenings for women between the ages of 40 and 49. The reason is the contention that screening younger women does more harm than good. There is a summary of the findings on the Annals of Internal Medicine site.

For women aged 40 to 49, the recommendation lists under “Benefits” a reduction of cancer deaths of 3% (1% to 6% range).

On the other hand, the recommendation lists many “Harms” including false positives, yielding psychological harm, unnecessary biopsies and unnecessary imaging tests. Other “Harms” are overdiagnosis and unnecessary early treatment. “Overdiagnosis” is defined as finding a cancer that would never have become clinically apparent during a women’s lifetime and “unnecessary early treatment” is defined as treating a cancer that would have become clinically apparent but would not have shortened a women’s life.

Let me boil that down: A benefit is saving up to 6% more lives. Harms include psychological stress and unnecessary tests and treatments, in other words, spending money you didn’t have to.

To me, the benefits and harms are not equivalent in nature — lives versus money. First of all, doesn’t it depend on whose life and whose money? The whole premise of the recommendation is a collectivist one, that is, “what is better for the collective, a mere 3% of lives saved or all that money wasted?” If you are one of the 3% I bet it matters a lot.

To be fair, the recommendation does allow that individual risk factors, such as family history, should be evaluated on a case-by-case basis.

To really understand the issue you have to understand a few statistical concepts. As usual, to really understand them you need a graduate degree in statistics. But you can understand the ideas well enough with a couple of examples to understand why the AHRQ report is driving people nuts.

The first, and easiest, concept to understand is that all tests are prone to error. There are ways to control the size of the errors and the errors may be relatively large or relatively small depending on the test.  But there are always errors. A breast self-exam is not as reliable as a mammogram which is not as reliable as a biopsy. The self-exam is more prone to error than the mammogram which is more prone to error than the biopsy.

The second concept is the difference between false positives and false negatives. As you probably know, a false positive is when the test tells you that you have cancer but you really do not. A false negative is when a test tells you that you are healthy but, really, you are sick. A false positive is a false sick. A false negative is a false healthy.

Another name for false positive errors is Type I errors. Another name for false negatives is Type II errors. Type I is false sick. Type II is false healthy.

For example, if you were manufacturing a little, cheap, kid’s flash light, you might want to simply check if the thing turns on before you ship it. Or you might even ship the toys to stores without testing them at all. If any failed the customer could just bring it back to the store and exchange it. No harm done. You would be operating in the realm of Type II errors, or false healthy.

But what if you were manufacturing a parachute? A failure in the field would be pretty serious. So you would set up your tests with a bias toward failing the test before shipping to minimize failing under use. Then you would get a lot of test failures. You would be getting a lot of  Type I, false sick, errors and hopefully very few false healthy errors.

When you go to the doctor and you are not sure if you have a cold or an allergy or the flu. Often the diagnosis is made by instinct and experience rather than a battery of more accurate, and expensive, tests. So what if the diagnosis is wrong or test inaccurate? You can go back and try another treatment. In that case you can tolerate Type II errors, false healthy. And you do this for financial reasons.  Because to do all the tests necessary to separate flu, colds, bronchitis etc. is more expensive than most people think is worth the result.

But, if you go to the doctor because you may have a time-sensitive, terminal disease, you cannot tolerate a Type II (false healthy) error and live. So, if you want to do your best to guarantee living, you prefer to get Type I, false sick errors. You want the test to be biased to finding disease than finding health even if it generates false positives.

Finally, when a condition is relatively rare, the tests produce a high percentage of false positives. So, if one in a thousand people have a disease,  and the test is 99% accurate, the test will be wrong ten times in a thousand. You will get about ten times more false positives than true diagnosis of sickness.

So, going back to the AHRQ study, to me, the results of the research are not surprising. The younger the woman, the less likely she is to have breast cancer. The earlier in life you screen for breast cancer the more rare it is. The more rare something is, the more likely the test will be in error. If we want the test biased to finding cancer rather than missing cancer, and there are few cancers to find, the test will produce a large number of false positives or false sick results.

Did we really have to do the study? The recommendations are not scientific, even though they are wrapped in data. The recommendations are more like the decision the toy manufacturer makes compared to the decision of the parachute manufacturer. When we conclude that we can “tolerate” 1% or 3% or 6% more deaths in order to do fewer tests we are doing the same thing the parachute manufacturer would be doing if he concluded that he would save a lot of money eliminating the final inspection of the parachutes and only 3% more of his customers would die as a consequence.

I’m not saying that the decision should always be one way or another. You cannot guarantee, even with a thousand tests, that a parachute will always open or that a woman will not die of cancer. What I am arguing is that the AHRQ is making recommendations that amounts to having women change from tolerating Type I errors to tolerating Type II errors and that the fundamental reason is not scientific but economic. The moral issue is that economic decision should be left to the woman at risk, with full cognisance of the percentages and her personal risk factors, not the AHRQ as a blanket recommendation to everyone. Honestly, it would be the same if it were the other way around. A very small number of men contract breast cancer. But we do not routinely screen men for the disease because both the medical community and the majority of men have decided that the expense of the screenings are not worth the tiny risk. On the other hand, for the individual with a family history, the likelihood of disease is high enough that the expense is acceptable.

As more and more people in our country make collectivist arguments for national (government) health care expect more collectivist recommendations for screenings and treatment.


2 Responses to “Breast Cancer”

  1. What collectivism means for health care « Fox Enterprises Limited Weblog Says:

    […] collectivism means for health care Breast Cancer January 10, 2010 by […]

  2. foxenterprises Says:

    Excellent post. You’ve unpacked and explained the problem here quite well.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: