Curing cancer statistically via mammography

One of the papers that we studied during our Harvard Medical School “big data” course in February was “National Expenditure For False-Positive Mammograms And Breast Cancer Overdiagnoses Estimated At $4 Billion A Year” (Health Affairs 34:4, 2015). The researchers used a data set of 4 billion insurance claims to see what was going on in the U.S. population. We learned that screening mammograms are not helpful compared to waiting for a lump to show up. There are a lot of things that look bad on a mammogram that aren’t, in fact, bad.

Americans fell in love with mammograms:

Why do we love them so much? It turns out that the five-year survival rates for breast cancer were improved after women en masse got put through the mammography industry. Why would anyone want to stop doing something that improved five-year survival rates?

It turned out that the statistical cure for breast cancer because of mammography was due to the fact that women who did not have cancer were being treated for cancer. They hadn’t been killed by cancer five years later because… they never had cancer to begin with.

So we wrapped ourselves around the axle with data that we weren’t smart enough to comprehend.

(Separately, we learned during this medical school class that it takes approximately 17 years for an identified “best practice” to be adopted by physicians nationwide. Thus we can expect Americans to back off on their love for mammography perhaps in 2032.)


17 thoughts on “Curing cancer statistically via mammography

  1. One of the premises of Obamacare was that increased spending on preventative care would actually save money in the long run, by reducing costs to treat later diseases. Unfortunately it appears to be false.

  2. You can see the same thing with the prostate specific antibody (PSA) test for prostate cancer (yes, it does detect prostate cancer, but it also causes men to undergo invasive treatment that might not have ever been necessary since the PSA test can’t reliably discriminate between prostate cancer that will kill you and prostate cancer that is so slow growing you’ll be dead from old age long before it would be a problem.

    The health care industry has spent the last ~50 years programming the populace to believe that early detection is the best weapon we have against cancer; it’s going to take along time to deprogram that out (particularly when most doctors have a strong bias towards taking positive action (they can charge for tests and treatment, and they’re more likely to be sued for not testing/treating in the event of a rare disease than they are to be sued for testing/treating when it wasn’t necessary in the first place)).

  3. “It turned out that the statistical cure for breast cancer because of mammography was due to the fact that women who did not have cancer were being treated for cancer.”

    A positive mammogram doesn’t lead to treatment, it leads to more invasive (and more accurate) diagnostic procedures.

  4. Having experienced prostate cancer, nobody gets treated based on PSA results. They get a biopsy, which, if positive, is graded as to aggressiveness. The tricky part is deciding what to do with the mild-to-midrange aggressive tumors. Both doctors and patients tend to err on the side of treatment, largely overriding the grading results. It is very difficult to decide to just live with cancer if I am under age 75. The biopsies may be just too good, but if they are not done the survival rate for tumors detected by digital (finger up the) rectal exams is poor. Cancer is nasty and statistical hindsight is, let’s say, annoying. If I have a 65 percent chance of survival, it’s a 100 percent death sentence if I’m in the bad 35 percent.

    For mammograms, the issue is actual FALSE positives rather than “gradient positives”, but the dilemma is the same: waiting for a palpable tumor is certain to kill some women who would benefit from imaging detection.

    The whole sickcare sector needs a thoroughgoing cost reduction at the procedure level, not necessarily a broad procedure reduction.

  5. While I have the soapbox, it is hard to get worked up by mammograms while we are putting every airline passenger through a scan. How many wasted $Billions there, when the assumption going in is nobody (or vanishingly few) is positive?

    I know its whataboutism, but hey that’s 2018.

  6. Speaking as a doctor, it’s an uphill battle to convince my patients that the harms of some screening outweigh the benefits. Naturally every patient assumes they will be one of the rare few to benefit and not one of the many to incur harm. There is some validity in this thinking at the individual level, even if my personal calculation would be different.

    In any case, 73% of well people surveyed would rather have a full-body CT scan than be given $1000 cash ( ). I presume my patients are broadly similar to that sample. Why pick that hill to die on?

  7. Mammograms may or may not be useful and cost effective for screening breast cancer, but that isn’t the subject of this post. The central claim of this post is that “we wrapped ourselves around the axle with data that we weren’t smart enough to comprehend”, but the post presents no argument (convincing or otherwise) in support of that claim.

    I rather suspect that the problem is not that the folks developing breast cancer screening protocols were innumerates whose mistake was to make decisions without spending 20 minutes with the oracle philg to make sense of their data. I suspect the “problem” (if there was a “problem”) was that decisions were of necessity made with incomplete data and then deployed into a system which was/is not uniform and was/is changing rapidly in ways which could impact the effectiveness of those protocols. In short, the problem domain was/is much much harder than, say, building a photo sharing website.

  8. Blame this on marketing and the hype that sales and advertisement bring with it.

    If you ever watch nightly news on ABC, CBS and NBC, you will see that 50% of the advertisement is all about some magic pills that will make our live much better. Watch how the person looks so happy in those adds but keep watching where in the last 2 seconds of the add there is a warning message about the side effects and how this may not work for everybody. You see the same message about scans to catch early diagnostics.

    Americans are so consumed in quick fixes and false hopes. The irony of it all is that we ask and let government take over our lives (regulations et. al.) but yet the government is the wild west in all this: they let the market run wild.

    So will any of this “big” data help? The answer is “no” and a definite “no” simply because by the time we wrap our heads around the data, a new set of variables is thrown at it or discovered that makes the data unusable. The data is nothing but historical data and a possible projection into the future — just like the weather data is.

  9. “It turned out that the statistical cure for breast cancer because of mammography was due to the fact that women who did not have cancer were being treated for cancer. They hadn’t been killed by cancer five years later because… they never had cancer to begin with.”

    I don’t see anything in the source article that asserts this fact. Did I miss it? (The article is not easy to digest). If this is a deduction, what is it based on?

    If I were creating a cancer survival statistic I would base it on diagnosed cancer survival rates – not on total positive results that cause the patient to be further tested to see if it is cancer. (Of course money might influence how statistics are calculated!)

  10. Seventeen years that is, Phil- the magic number the health services research community I belong to cites at every meeting.

    To you and your readers, I would highly recommend this book by a Dartmouth researcher Gelbert Welch to provide further context into pitfalls of early detection and diagnosis :

    For folks interested to further drill into the mammography controversy, the text by Goetzsche is a highly recommended read:

  11. Natalia: This post isn’t about the “mammography controversy”. It is about how this controversy demonstrates that “Obama”, “Hillary Clinton”, and “the central planners behind Obamacare” are idiots and the entire project of government involvement in healthcare is idiocy. Of course, in order to make this argument philg misleadingly focuses in on one aspect of a massive and complex system. The argument in this post is cherry picking under the cover of data.

    We start by ignoring that this year 35,000 people who would have died at the 1950 breast cancer fatality rate will instead survive. We ignore that absent a system for socializing the cost of their care, most of those people (and most of those who would have been saved by 1950 level care) could not afford it. Of course we ignore that in a free market, any system for socializing costs will require government action. In the most ironic twist, we ignore that absent the demand for care generated by the system for socializing costs, much of the absolutely massive investment in research, infrastructure, and personnel required to achieve those improved breast cancer survival rates would not of occurred. Thus, without a system for socializing the costs of care, even people who could afford care without a system for socializing costs (e.g. the women in philg’s family) would likely find themselves facing inferior treatment options than exist with it.

  12. Neal: I may not get your point right , then please correct me, it appears though that you contradict yourself. On one hand you talk about government involvement in healthcare being ‘idiocy’ , on the other- you are for socializing the healthcare costs, which is what European nations have indeed- the direct government involvement in healthcare.

  13. Natalia: My point was that the original posting implies that government involvement in healthcare is idiocy. I was not advocating that idea myself. The U.S. does socialize healthcare costs albeit less completely and less efficiently than in Europe. I think over half of all care is paid for (medicare and medicaid) or provided directly (Veterans Administration, state/county hospitals) by the government. Employer based medical insurance (another mechanism used to socialize healthcare costs) is also heavily regulated by states and by the Federal government via provisions of the tax code.

  14. Neal: I guess the focus is in the eyes of the beholder here. The main take-away from this post for me is not only the waste of healthcare resources but also harms of overdiagnoses. While the $$ costs are important, the harm of going through more and more invasive procedures after a positive mammogram/PSA test etc are tremendous and readers of these blog may be interested to learn more about

Comments are closed.