Success by association?
The current issue of the British Medical Journal [BMJ], published July 28th, includes an interesting review study about screening mammography. The study, Breast Cancer Mortality and Screening Mammography, takes a little explaining. Researchers sought to determine whether the availability and widespread use of screening mammography was directly responsible for decreases in mortallity from breast cancer observed over the last few decades in Europe. A summary of this study was published on July 29th at MedicineNet.com. This is a subject I recently reviewed in my post Betty Ford & the Status of Mammography. It is also a subject that raises another thorny and confusing issue with respect to research in general. Any study that seeks to examine the effectiveness of a treatment or screening technique on any disease tends to make a fairly simple type of comparison. Researchers will examine how many people have received a given cancer treatment over a period of time, compare them to a similar number who have not, hopefully using a double-blind technique, determine how many people have improved or not, and finally draw some conclusions about whether the treatment was effective.
The problem is that these types of studies are, of necessity, narrow in their focus, and generally do not take into account other factors that may have influenced whether the study participants improved or not. These other factors may include types of additional treatments the participants also received right before or during the study period — like surgery, for example — as well as other factors that may have influenced disease progression or outcomes, such as the baseline health of the participants, changes in lifestyle or financial circumstances, concurrent health problems, access to quality healthcare, stress, and other factors that can influence health. When such a study concludes that a particular treatment is directly responsible for decreasing disease progression or death, this conclusion may not, in fact, be accurate. Careful researchers may, for example, qualify their conclusions by saying that a given treatment was “associated” with a decrease in mortality. But it might only have been a coincidence, and something else besides the treatment examined in the study may have contributed to or influenced the results. Research on pre-treatment screening is even more problematic.
I am always suspicious when research results are given “Wow” headlines, like “Screening Mammograms Save Lives.” If the headline instead says, “Screening Mammograms Associated with Decreased Breast Cancer Deaths,” I may be less suspicious, but I also wonder what else was going on, and whether it was cancer treatment itself — not detection — that should be credited with saving lives. Fortunately, there are researchers who also wonder about this notion of “success by association.”
Can screening mammography rest on its ‘laurels?’
The researchers in the BMJ study attempted to peel back the veil on screening mammography. They gathered data on the implementation of screening mammograms across Europe and compared that to data gathered by the World Health Organization on breast cancer deaths in those same countries. Then, they paired up individual countries that were similar, but started widespread screening mammograms at different times. The paired countries had to fit these criteria: they had to be geographic neighbors, they had to have similar populations, similar socioeconomic structures, and similar access to comparable healthcare and treatment. Finally, they had to differ in one important way, which is that one country had to have implemented a nationwide mammography screening program by 1990, while the other country had to have implemented a similar program several years later.
The idea was to compare breast cancer mortality between these countries during that period when one country provided screening mammograms and the other did not. The researchers hoped this might confirm whether or not changes in breast cancer mortality during that period could be directly tied to screening mammography. Part of the impetus for the study was the confirmed success that has been documented for cervical cancer screening. Once women began to get regular Pap smears, it has been shown, virtually without question, that Pap smears have had a direct influence on decreasing the number of deaths from cervical cancer over the past several decades. So they wondered if such an influence could be demonstrated for mammograms and breast cancer.
And what they found is that breast cancer mortality had declined over the years they studied, but the decline was no better after one country implemented screening mammography than it was for the same period of time in a similar country that had not yet implemented it. Therefore, the decline observed in breast cancer deaths over this period could not be attributed to screening mammography itself, but perhaps to other factors they had not specifically examined. Below is an illustrative figure from the study, that shows the comparison between Sweden and Norway:
The period of time that starts at the aqua line, when most Swedish women were getting regular screening, to the fuschia line, when Norwegian women were as well, and the years that followed, is what the researchers were interested in, a period from about 1989 to 2005. The blue and dashed-red lines between those two lines show the mortality from breast cancer in these two countries. It is easy to see that these two rates are about the same, and both show a gradual and similar rate of decline. In other words, while most Swedish women were getting mammograms then, when most Norwegian women were not, Swedish women were dying of breast cancer at about the same rate as their counterparts in Norway. Analysis of the statistics for the other two paired countries — the Netherlands and Belgium, and Northern Ireland (part of the UK) and the Republic of Ireland — showed similar results.
If the researchers brought any bias to their study at all, it may have been the hope that screening mammography would demonstrate a similar impact on cancer mortality that was demonstrated in Europe by Pap smears for cervical cancer screening. But it did not. Further research on what has in fact brought about the decline in breast cancer deaths in Europe is indicated, but the researchers, after ruling out other factors like comparative obesity, the decline in use of HRT for menopause, availability of treatment, and the accuracy of available statistics, suggested that perhaps the most plausible conclusion is that increased public awareness of breast cancer, improved healthcare access, and better cancer treatment are the factors that account for the decline in breast cancer mortality in these countries since 1990.
I decided to take another look at breast cancer statistics in the U.S. and see if I could draw some of my own conclusions. I extrapolated some data from a report prepared by the American Cancer Society on Breast Cancer, borrowing from a graph of U.S. breast cancer mortality for the same period covered by the study. Then I placed an overlay of the mortality rates for the same period in Sweden and Norway. It yielded the following:
The Centers for Disease Control offer some numbers on the above. In the U.S., breast cancer deaths between 1997 and 2006 have:
• Decreased significantly by 1.9% per year from 1998 to 2006 among women.
• Decreased significantly by 2.0% per year from 1997 to 2006 among white women.
• Decreased significantly by 1.5% per year from 1997 to 2006 among African American women.
• Decreased significantly by 2.1% per year from 1997 to 2006 among Hispanic women.
• Remained level from 1997 to 2006 among American Indian/Alaska Native women.
• Remained level from 1997 to 2006 among Asian/Pacific Islander women.
It is notable that regular screening mammograms were encouraged in the U.S. earlier than they were in either Sweden or Norway, and that U.S. society, socioeconomic status, and access to healthcare is far less homogeneous than it is in Scandinavia. In addition, U.S. screening mammograms typically involve at least two views per breast, whereas European screening mammograms have typically involved one view and were offered less often than in the U.S. The decline in U.S. mortality has also been reported to have dropped off between 2003 and 2007, after a more significant decline was previously observed that has been attributed to the sharp decrease in HRT use in menopausal women. Meanwhile, the incidence of ER+ breast cancer has increased in these years in women aged 40-49, both in white and African American women, the latter of whom continue to have more incidence than other women of aggressive ER- breast cancer. A full report of these trends can be found here in another article at MedicineNet.com. And snapshot graphs of incidence and mortality trends for the years 1987-2007 can be seen at this link from the National Cancer Institute.
The only overall conclusions one can draw from all of this are perhaps the ones I’ve been pounding away at here for months now: No matter where in this world we live, we need better screening tools, better and earlier detection, and better treatment for metastatic breast cancer. And we better hurry up. This ship is way too slow.
Please click on the post title or the comment link below to post a response.