"Startling" - if true
The New York Times online reported today that breast cancer rates (for "the most common form of breast cancer," not specified) in the US dropped 15% from August 2002 to December 2003, and the only explanation the experts can think of is that use of hormone therapy dropped off sharply after the results of the Women's Health Initiative were released. Here's the press release from the MD Anderson Cancer Center; I couldn't find the context for the 15% but I trust that the author got it from the investigators. (Note: That's one of the tricky things about news reports on unpublished data.)
The report is fascinating. It really is a bunch of experts - evidence-based experts at that - sitting around saying, in effect, "Huh. Yup. I guess that's what it must mean," but tentatively, because, as one of them points out, epidemiology (or non-randomized observational studies) cannot prove causation. Particularly this type of study, of ecologic design, which compares population-wide rates of hormone use with cancer incidence. The drop in hormone use is what some people - environmental scientists, for example - call a "natural experiment." The Women's Health Initiative was the only randomized study (read: experiment) to look at health effects of hormone therapy (Prempro, specifically), and breast cancer was one whose incidence was slightly elevated, along with heart attacks, stroke, and blood clots. The Wyeth spokesperson interviewed for this article was skeptical of the explanation, and said that more studies need to be done; sure (see below), but there is not going to be another study on the scale of the WHI to replicate a causal relationship. Sorry.
A good point made in an AP article is that stopping hormones may in part have slowed the growth of existing tumors, which could still come to light at a later date. Another factor to look at is screening, but you'd really have to see dramatic changes in screening utilization to explain these results. Clearly, there's a lot more interesting work, and closer looks at the numbers, to be done on this finding.
Questions: Is the observed 15% drop comparable to the magnitude of increase of breast cancer incidence seen in the hormone arm of the WHI, given competing causes of breast cancer (which, to be honest, we don't really understand) and given the 50% drop in hormone use? Could tamoxifen use explain some of the decline?
Also, have there been concurrent drops in heart disease and stroke? We don't collect those data as systematically as we do cancer incidence (via the Surveillance, Epidemiology, and End Results system), but I imagine that cohort studies such as the Nurses Health Study will be able to look at these conditions.