Let’s start with the source of this claim. It was a paper published in the October issue of Archives of Internal Medicine, led by Jiyoung Ahn, Ph.D., at the National Cancer Institute in Bethesda, Maryland.
The researchers tabulated answers to an AARP survey sent out in 1995-1996, asking its senior members to remember their measurements, diets and lifestyle habits through their lives. Women were also asked the age of their first menses, reproductive history and hormone use. The researchers didn’t verify any data in the AARP Study database, and had only this unreliable self-reported retrospective information.
The AARP had mailed out 3.5 million questionnaires to members living in just six states (California, Florida, Pennsylvania, New Jersey, North Carolina and Louisiana) and in two metropolitan areas (Atlanta, Georgia and Detroit, Michigan). This represents only a fraction of its 38 million members. Only 132,942 women, average age of 63 and mostly Caucasian, completed and returned the risk factor questionnaire used in this study. These 0.3% of AARP members are not likely representative of women aged 50-71 in the United States, let alone representative of AARP members. To examine breast cancer risks, these researchers went a step further and unexplainably eliminated the questionnaires from nearly 34,000 additional women: more than 9,000 who had a personal history of cancer, 4,445 who were still premenopausal, and 20,420 who reported extremes of height or weight. This left 99,039 women (0.23% of its members). They then matched these women with state Cancer registries and identified 2,110 cases of breast cancer. That’s our non-Trojan number. This study was looking at a nonrandom sample of only 2,110 women, not the 99,039 reported.
The elderly women in their sample differed considerably from postmenopausal in the U.S., beyond the fact that over 90% were White. Fifty-five percent were still taking hormones at the time of the survey. In contrast, according to the Centers for Disease Control and Prevention, 1988-1994 NHANES data showed only 22% of postmenopausal women had been currently taking hormones, mostly those of higher socioeconomic status. The ‘obese’ women in their sample were also over three times as likely to be African American (6.8% versus 2.1%) and more than 25% less likely to be taking hormones compared to the ‘normal’ weight women.
The researchers divided their sample according to BMIs at various ages, looking for relationships to postmenopausal breast cancer risk. Keeping in mind the caveats about the quality of the data and its applicability to most women, let’s look at what they found.
Among the 1,162 women who had been using hormones at the time of the survey and developed postmenopausal breast cancer, the very fattest (BMIs ≥ 40) had the same breast cancer risks as the ‘normal’ or ‘overweight’ women. And those who gained weight through adulthood, even over 50 pounds, had the same risks as those whose weights had remained stable. “No associations with weight gain were observed in current [hormone] users,” said the authors.
And looking at each of the periods of the women’s lives, those using hormones who had been the most ‘obese’ at age 50, and at age 35, and at age 18, actually had up to 35% lower risks for developing postmenopausal breast cancer than women who were ‘overweight’ or ‘normal’ weight.
In fact, noted the authors: “BMI at age 18 years was inversely associated with breast cancer risk regardless of [hormone] use.”
Among the 948 women not using hormones at the time of the survey, those who had been ‘obese’ at ages 18 had 52% lower risks for developing postmenopausal breast cancer than the ‘normal’ weight women. Those ‘obese’ at age 35 even had lower risks than the ‘overweight’ women.
We’ve now covered 98% of the women in their cohort, all but 40 women, and there’s no tenable correlation between breast cancer and fatness.
The only higher relative risk for postmenopausal breast cancer of any note that they were able to derive was a 2.0 associated with older women with BMIs ≥ 40 and not using hormones at the time of the survey. But this relative risk was too small to be tenable. More disturbingly, it was derived with a statistical sleight of hand — it was not in reference to ‘normal’ weight women as is being claimed, but only the thinnest women (BMIs of 18.5-22.4). They made up a new definition for a ‘normal’ weight. The real relative risks among those not taking hormones, based on their numbers, were only 75% - 89% higher for the women with BMIs ≥ 40 and 25% to 36% higher among most ‘obese’ women, compared with most ‘normal’ weight women — noncredible correlations and not beyond statistical error or random chance. Remember, these relative risk are percentages of an actual incidence of breast cancer for women of 0.119% according to U.S. Cancer Statistics (USCS). But to tell ‘obese’ women not taking hormones that, based on iffy data, their actual risks could go from 0.119% to 0.14 - 0.22% wouldn’t sound nearly as worrisome.
Considering the shakiness of the questionnaire data being used and the confounding factors not considered, these computer-derived correlations are simply not meaningful. For all of the women in this survey, weight wasn’t demonstrated to matter.
In fact, the odd finding of slightly higher risks among nonhormone users regardless of their weight or weight change, conflicts with the literature, according to the American Cancer Society, which suggests that hormonal use might be associated with a higher risk, not lower. There are countless confounding factors not accounted for in this study, which make any credible interpretations impossible and the idea that weight is the cause especially unsupported. Considerably higher relative risks (up to 4-fold), for example, were associated with menarche at age 15 years or older.
It’s so easy to believe that high-sounding relative risks derived from data dredges mean something. But any authors can search hard enough and put enough data into a computer model to come up with all sorts of meaningless associations. Remember that 2005 study of a random sample of 12,178 women published in the British Medical Journal that found a relative risk for breast cancer of 2.41 among left-handed women with BMIs of 25 or under? We can drive ourselves nuts worrying about every correlation that can be dredged up, most of which are never supported in clinical trials.
Was this study a sincere attempt to find a cure for breast cancer, or another attempt to scare women into losing weight? The insinuations we’ve heard in the news and the press release issued by the American Medical Association have been that this study showed that natural weight gain with aging is something to avoid and that losing weight might lower one’s cancer risk.
There’s one big problem with that. It’s the opposite of what the authors found:
Adult weight loss was unrelated to breast cancer compared with stable weight. The lack of association of weight loss with breast cancer risk was consistent across the entire lifetime. Associations between weight loss and breast cancer did not differ based on [hormone] use... To confirm that weight loss due to preclinical disease did not account for the null associations observed, we repeated all the analyses after excluding women who were diagnosed as having cancers of any type during the first two years of follow-up. Risk estimates remain unchanged.
The same AARP data has shown that women with ‘normal’ BMIs of 20.9 and lower have higher risks for premature death than women who are ‘overweight’ or even ‘obese’ with BMIs to 35. In fact, the same AARP data has previously shown that the heaviest women (BMIs ≥ 40) have lower mortality rates from any and all causes than the thinnest women (BMIs < 18.5). An analysis of the AARP data through 2005, led by Dr. Kenneth Adams, Ph.D., found that with aging, mortality rates also dropped for those who were ‘overweight’ and ‘obese’.
And concerning dieting or any intentional weight loss efforts, Adams and colleagues concluded in an August 2006 issue of the New England Journal of Medicine, that it was especially ill advised for women and men, writing:
Among both men and women 65 years of age or older, weight loss after the age of 50 years was more strongly associated with the risk of death than was weight gain.
JFS readers have seen time and again that what is written in an abstract or press release often differs from what a study’s data actually found. In fact, it can be the complete opposite, especially when it contradicts “what everybody knows that being fat is bad.” That’s why it is increasingly necessary to look carefully at the study data because the media or peer reviewers aren’t going to.
© 2007 Sandy Szwarc