PSY 318

Understanding the mammogram controversy

During the fall of 2009, a government-organized committee recommended a change in policy regarding the use of mammograms to screen for breast cancer.  The government recommendation prior to that time was that women should receive a mammogram every year beginning at age 40.  After reviewing the research on the costs and benefits of screening healthy non-symptomatic women yearly beginning at age 40, the 2009 committee recommended that screening of non-symptomatic women not begin until age 50, and that women in their 50's only have a mammogram every two years instead of every year.

These recommendations produced an immediate, strong, and negative reaction from a variety of groups, and even the president disavowed the recommendation and it did not become the government-recommended policy.

The purpose of the discussion of this controversy in this class is not to provide health-related advice, but rather, to help you better understand this issue in the context of phenomena discussed in the class, particularly the base rate effects that you have just learned about.


What are the costs and benefits of mammogram screenings?

Undeniably, some women's lives are saved by mammogram screenings between ages 40 and 49.  At that same time, there are undeniable costs associated with mammogram screening.  These costs include: 

Money. Screening a very large number of women each year costs a lot of money.   Even if some women’s lives are saved by the screening, one needs to consider the possibility that even more lives would be saved if that money were redirected to some other use.   Clearly, in order to make a judgment like this it is important to have a good estimate of the number of lives saved by the screening.

Anxiety.  Most women find the procedure at least mildly unpleasant and anxiety provoking.  However, the more significant anxiety cost occurs in the case of women who receive a false positive reading, that is, women who do not have breast cancer but whose mammogram reading suggests that they do have breast cancer.  In those cases, the unnecessary anxiety that the false positive produces is significant.

Pain and discomfort and harmful unnecessary (or non-beneficial) treatments.  Again, the focus here is on those women who receive a false positive reading.   In some cases, what follows from that will involve expensive and stressful and painful invasive procedures (e.g., a biopsy) whose purpose is to find out if the woman really does have breast cancer or not.  In some cases, the biopsy will be followed by treatments (possibly including chemotherapy and/or surgery) that do not actually benefit the woman.

Given that there are these not-trivial costs associated with mammogram screening, the next question to consider is:

How accurate are mammograms at detecting breast cancer?

"Hit rate" refers to the probability that if a woman really does have breast cancer, the mammogram reading will indicate that she has cancer.   Of course, different studies have produced different estimates of the hit rate for mammograms (when screening otherwise non-symptomatic women), but a reasonable average figure is about 85%.  In other words, mammograms do a good, but not perfect, job of detecting breast cancer in non-symptomatic women who have really have breast cancer.   "False alarm rate" refers to the probability that if a woman does NOT really have breast cancer, the mammogram reading will nonetheless indicate that she has cancer.  False alarms are errors, and again, estimates of the false alarm rate for mammograms vary, but seem to average about 5% (but higher than that for women in their 40's).  In other words, when screening women in their 40's who do not have breast cancer, more than 1 in 20 will receive a false positive mammogram reading.

Given those estimates of the hit rate and false alarm rate, we now need to ask a subtly different question:

What percentage of women who receive a positive reading (signaling breast cancer) actually have cancer? 

Even though the hit rate is about 85% and the false alarm rate is about 5%, the answer is that – MOST cases of positive readings are false alarms!!

How is that possible?

Because -- the actual base rate of breast cancer is low enough (particularly among women in their 40's) that the huge majority of women who are screened DO NOT have breast cancer.  As a result, the number of false alarms that occur is much higher than the number of hits.

The relevant calculation is below:

~ 1.5% of women develop breast cancer in their 40's.
During one yearly screening, ~ .2% real new cases. 

If 100,000 women in their 40's are screened every year, there will be approximately:

200 new cases each year (.2%), and 99,800 cases of women who do not have breast cancer.

Of the 200 new cases, about 85% are correctly detected by the mammogram screening = 170 correct positive readings (hits).
Of the 99,800 women who do not have breast cancer, 1/20 will receive a false positive reading = 4,999 women (false alarms)

Thus, out of the total of 5,169 women who would be expected to receive a positive reading, only 200 are women who really have breast cancer.  In other words, of the total number of positive readings (5,169) only ~ 3% are hits, and 97% are false alarms!

[Although I generated these figures myself based on estimates of hit rates, false alarm rates, and breast cancer rates that I read in various articles, when I have seen figures published regarding the actual hit and false alarm numbers that follow from widespread screening of non-symptomatic women, the percentage of positive readings that are false alarms has been in the range from 85 - 97% -- but those figures are for all ages, not just women in their 40's, and it is known that the false alarm rate is significantly higher for women in their 40's than for older women.]

In other words, given current knowledge of breast cancer and current technologies, a non-symptomatic woman in her 40's who receives a positive mammogram reading (signalling breast cancer) on a yearly mammogram test is many times more likey to not have breast cancer than to have breast cancer. 

How many women really benefit from mammogram screening?

This question is more complex than you might imagine.  Most people assume that every woman who has had cancer accurately detected through a routine mammogram screening has benefitted (perhaps in a life-saving way) by her decision to undergo screening.   That is not actually the case!

Of those asymptomatic women who are screened and receive a positive reading and who really do have breast cancer, which women do NOT benefit from learning at that time that they have breast cancer?   There are four subcategories of women who receive an accurate mammogram reading that they have breast cancer but who do not benefit from the screening.

(A) SOME women who receive an accurate cancer-indicating mammogram -- would never have become symptomatic at all (that is -- they would have lived, and never become symptomatic even without any treatment). Not all breast cancers are lethal, and currently, medical science does not know how to distinguish between those kinds of breast cancers that would be first detected through a yearly mammogram screening that will prove lethal if untreated and those that might never grow to the point of even bothering the woman at all.  Some of these women will undergo serious treatments (e.g., chemotherapy or perhaps surgery) that, unbeknownst to them and their doctors, are totally unnecessary.   Of course, all these women (and their doctors) are likely to end up thinking that the mammogram screening saved the woman's life.

(B) SOME women who receive an accurate cancer-indicating mammogram -- would have become symptomatic and their cancer would then have been detected and their treatment would have been successful -- even if they had never undergone the mammogram screening.  Again, it is very likely that these women, and their doctors, will believe that the screening saved the women's  lives because it was through the screening that their breast cancer was first detected.  However, there is no way to know in any individual case whether everything would have turned out essentially the same even without the screening – because for some women, at some point the woman would have become symptomatic, and then diagnosed, and then treated successfully. 

(C) SOME would have become symptomatic and their cancer would then have been detected, but they end up dying soon thereafter of something else.  This is essentially the argument against screening women above a given age.  If someone is so old that their life expectancy is quite short anyway, then it is unlikely that anything is to be gained by detecting a cancer that wouldn’t kill them for a time that is longer than their expected lifespan.

(D) SOME will die of breast cancer anyway even though it was detected very early.  If the treatment is not successful (and sometimes it isn’t) then nothing has been gained by the early discovery of the cancer.

So -- in that case, which women DO benefit from mammogram screenings?  That is -- which women's lives are actually saved by the screening?  The women whose lives are truly saved by the screening are the women who:

(E) Would have died of breast cancer without the screening (that is, if they had waited until they were symptomatic and had their cancer detected then, it would have been too late to treat them successfully) but who, because of the screening, are able to receive treatment early enough for the treatment to be successful. THESE are the only women who actually have benefitted from the screening (even though almost all of the A and B cases are likely to attribute their successful treatment to the early mammogram screening).

How many E cases are there?

Based on research in Sweden, the absolute risk reduction associated with yearly mammogram screenings for women of all ages =  ~ one-tenth of one percent.  The figure for women in their 40's would be significantly smaller.  In other words, what the Swedish study found was that yearly mammogram screenings beginning at age 40 save the life of about one woman in 1,000 from dying of breast cancer.   Thus,  yearly mammogram screening of asymptomatic and otherwise not high risk women in their 40's provides a very large benefit (life saving) to a very small number of women at a very high price (lots of false alarms and some women treated who did not need the treatment).  And that is why reasonable people can differ about whether it is a good idea.

This general way of analyzing the costs and benefits of mammogram screenings for breast cancer can, of course, be applied to the analysis of the costs and benefits of other forms of cancer screening.  For example, a problem with the use of PSA tests for screening men for signs of prostate cancer is that the tests result in a lot of men in category "A" (that is, men who really do have prostate cancer, but their cancer would never kill them if it had not been detected by the PSA test).  In addition, treatment for prostate cancer runs the significant risk of leaving a man impotent or incontinent (or both).  As a result, even though there is no question that some men have had their lives saved by routine PSA testing, the scientist who originally developed the test now considers routine use of the test for screening non-symptomatic men to be medically irresponsible (click HERE to read his article in the New York Times).  

The following two articles discuss many of the same points made above, along with a discussion of the very important concept of “lead time bias” which must be understood to properly evaluate data regarding the effects of cancer treatments administered to people whose cancer is detected through routine screenings.

Reading: Mammogram Math (New York Times)
Reading: Dr. Harriet Hall - A Skeptical Look at Screening Tests

The following article reports the findings from a study published in Oct., 2011, the discusses the same issues discussed above.

Reading: Has the power of the mammogram been oversold?

Update: Summer, 2012:  Decision regarding PSA testing for prostate cancer in asymptomatic men

In May, 2012, the Preventive Services Task Force recommended against any use of PSA testing to screen asymptomatic men  [the link takes you to an excellent article on the topic from the LA Times].  A brief articles from the NY Times will also help with your understanding of the issues:  To screen or not to screen. 

If the data are so clear, then why do many men still desire PSA testing?  That rather interesting issue is discussed HERE.  The imp;act of cancer anxiety on decisions to seek/accept treatment is further discussed in this opinion piece from the NY Times.