The Perils of Medical Blogging – and Reporting on it
Rebecca Goldin, Ph.D, January 8, 2007
ABC reports on a cancer researcher’s blog post to explain a reduction in breast cancer incidence. Uh-oh.

Last year, a new study came out showing that breast cancer incidence (the rate at which women are diagnosed with breast cancer) was going down rather dramatically. This decrease comes three years after it was found that hormone replacement therapy (HRT) was associated with increased breast cancer, or at least the kind of breast cancer that is related to hormones. Many women stopped HRT as a result of this research, prompting some researchers to attribute the decline in breast cancer rates to the decline in the use of HRT.

ABC News threw a wrench into that theory in December: Citing Dr. Len Lichtenfeld, Deputy Chief Medical Advisor for the American Cancer Insitute and his blog site, ABC pointed out that the decrease could also be explained by a decrease in diagnoses. Women may be getting fewer mammograms, which may mean that they are being diagnosed less frequently with breast cancer than they were three years ago, independent of the HRT therapy. Lichtenfeld did not back this conjecture up with any hard numbers, and ABC failed to call him on it or try and pin down any actual numbers.

But perhaps the more egregious flaw is what ABC did not report. To support his claim that a decline is HRT is not sufficient to explain the decline in breast cancer incidence, Lichtenfeld makes a mathematical argument, one that ABC does not even mention. And unfortunately, the math just doesn’t add up.

Lichtenfeld begins by noting that scientific tests of HRT in controlled environments found a very small increased risk rate for breast cancer – he claims on the order of 0.8 percent.

"For every 10,000 women taking combined HRT, there would be an additional 8 breast cancers each year. 

If we flip that estimate around, what the authors are saying that there would be 0.8% fewer breast cancers each year if women on HRT had in fact not taken those medications"

First, Lichtenfeld made a division error: 8/10,000 is .08 percent. But the broader issue is that he confuses percentages of the population with percentages of cancers.

He points out that the decline in breast cancer rates is about 11 percent in certain age groups, about ten times higher than what could be attributed to a complete cessation of HRT.  But comparing an 11 percent decline among those with cancer to a .8 percent increase (which should have been .08 percent) in cancer among the general (female) population is comparing apples to oranges. In order to assess what 8 additional cancers per 10,000 women means as a percentage of cancers, we need to know how many women are diagnosed with cancer every year.

According to the National Cancer Institute’s SEER data, there were approximately 125 diagnoses of breast cancer each year in 2003, for every 100,000 women. This amounts to 12.5 diagnoses per 10,000 women. An additional 8 cancers per 10,000 women is a 64% increase. At the peak in 1999, there were about 140 cancers diagnosed per 100,000 women, or 14 per 10,000. Another way of putting this is that an 11% decrease in cancer incidence would be about 11% of 14, or about 1.5 fewer cancers per 10,000 people, far fewer than the 8 fewer cancers that a complete cessation of HRT might result in.

In other words, Lichtenfeld’s main criticism of the scientific study is that a reduction in use of HRT isn’t significant enough to explain the decline in breast cancer rates. But by our calculation, it certainly is.

While it is laudable for ABC’s to monitor what scientists are blogging (and not just what they are publishing in peer-reviewed journals), ABC failed to ask the obvious questions that Lichtenfeld did not address:

• Did mammogram screening actually go down? The article goes to some pains to discuss some doctor’s impressions about longer waits and decreasing numbers of radiologists. But ABC was not able to provide any corroborating data that might point to whether the theory that decreased screening means decreased diagnoses is likely.
• Would a decrease in mammogram screening be sufficient to explain the decrease in incidence of breast cancer?
• Are there other possible explanations for the decrease in incidence? A decrease in doctor visits, women’s decreased attention to self-examination, or a decrease in health insurance would all explain a decrease in measured incidence of breast cancer while people are still getting the cancer at the same rate – similar to the proposed consequence of reduced mammogram screening. But there could be other reasons as well, from environmental to diet changes, or even a shift in demographics, that are leading to a decrease in breast cancer rates.

And perhaps most importantly, ABC did not explain the erroneous reason that Dr. Lichtenfeld suspects that a decline in HRT use is not responsible for the large decline in breast cancer incidence. ABC should be careful when reporting based on blog-science – even that from a well-respected member of a scientific organization. Journalists still need to “eyeball” the numbers to see if they make sense.

Science has a peer-review process for a reason, and even doctors and researchers have opinions and hunches that might not measure up to scientific scrutiny.