A public health paper on breast cancer incidence that does not withstand scrutiny
©CLIMACTERIC The Journal Of the International Menopause Society. December 2010.
*Also published in ©Menopause Live 2010. An e-newsletter of the International Menopause Society.
Winnifred Cutler. Ph.D.,Athena Institute for Women's Wellness, Chester Springs, PA.
Regula Burki, M.D., Berne, Switzerland
Date of release: December 2010, and 23 August, 2010
The recently published report from Ereman and colleagues  was intended to summarize recent trends in breast cancer incidence, mammography screening rates and hormone therapy (HT) use in non-Hispanic, white women (NHW) ≥ 50 years old in Marin County, California. The authors state that Marin County showed an excessively high incidence of breast cancer throughout the 1990s. They assemble tables and graphs that are used to suggest that a large drop in estrogen/progestin hormone therapy (EPHT), attributed to the 2002 media reporting on the Women’s Health Initiative (WHI), by 2004 resulted in a concomitantly reduced incidence of invasive breast cancer with no decline in mammography utilization. Further, they report a concomitant drop in estrogen receptor-positive (ER+) tumors. They also show that the incidence rates appear to have bottomed out and began rising after 2004 without reaching statistical significance.
The prevalence estimates for each HT were supplied from a 2006–2007 survey taken from 1083 Marin County NHW women ≥ 50 years attending mammography screening. A total of 18% reported having undergone a hysterectomy. The women were asked to report what regimen they took now, and remember what their history of HT use was. The authors report that postmenopausal HT use decreased from 1995 to 2006: the use of estrogen hormone therapy (EHT) decreased by 15.4% (from about 25% to 22.4%) and the use of EPHT by 58.9% (from about 21–22% to 6.7%)
The prevalence estimates for mammography derive from a second source. The authors used an online tool on the National Cancer Institute (NCI)’s ‘AskCHIS’ to calculate the mammography prevalence in 2001 and compare it with the prevalence in 2004. This used random-digit dialing to homes (white race not specified) and the women who answered were asked if they ever had had a mammogram, and when most recently. No data were available on first vs. rescreening. The authors report that the prevalence of mammography screen was ≥ 85% in 2001 and 2004, i.e. essentially unchanged in these two groups.
Breast cancer incidence rates are gathered from a third source. The Survey for Epidemiologic and End Results Program (SEER) data banks of the NCI sourced specifically for Marin County, ‘age adjusted data to the 2000 US population standard’. The authors report that invasive breast cancer rates from 1990 to 2007 for NHW ≥ 50 years old reached their absolute peak in 1999: 522/100,000. Translating this to a more readable ‘% of women’ yields incidence rates of 0.52% (95% confidence interval (CI) 0.45–0.60%) in 1999; 0.34% (95% CI 0.29–0.41%) in 2004; and 0.45% (95% CI 0.39–0.52%) in 2007. They further state ‘These results also were paralleled specifically by the ER+ tumors’.
Combining information from these three sets of data, the authors suggest that, if the population estimates for given years are correct, then the incidence declines in parallel with the EPHT decreases, and that breast cancer incidence seems now to be rising again, although such a trend would need to continue to achieve statistical significance.
We find that each of the three disparate sets of data lack rigor and derive from unrelated groups of data inappropriately linked to each other. The data themselves appear to be unreliable, which is very disturbing.
First, the way the data on HT use were gathered violates standard scientific practice by relying on recall, during a mammogram visit, of what products the women are taking now and what products they took up to 10 years ago. Nor did the authors report any attempt to show women pictures of different products that might serve as a memory aid. The estrogen + progestin data presented make no attempt to distinguish between regimens (sequential vs. continuous combined) or between chemical composition (progesterone, antiandrogenic like medroxyprogesterone acetate, or androgenic like norethisterone). Such good scientific practice has revealed important differences in breast cancer incidence among different products and regimens [2,3]. These differences also reflect the balance between apoptosis and cell proliferation in breast cancer cell studies in vitro .
Next, the random-digit dialing survey of mammogram utilization is problematic. Neither the NCI nor these authors acknowledge that, by 2004, major technological changes had shifted phone usage toward cell phones and Caller ID has enabled household members to avoid interruptions from unknown callers . We do not believe the people answering the phone and the interviewer can be assumed to be representative of NHW women ≥ 50 years in Marin County.
Most serious are the errors in calculating the incidence of breast cancer in Marin County. First, the age-specific incidence rates for the two time periods (1990–2001 vs. 2004–2007) were not significantly different, but a graph is displayed in which the cancer incidence appears to be lower in later years, and the descriptive caption fails to correct that misimpression.
Although this paper’s title and introduction imply that Marin County has excessively high breast cancer incidence, an earlier paper, co-authored by this paper’s first and seventh authors, demonstrated that very large errors were discovered in the population projections that formed the denominator from which cancer incidence was calculated . Once the 2000 census became available, the ‘error of closure’ between population projections, which must always use the prior census and either the sophisticated Department of Finance (DOF) data or SEER data, revealed how distorted the population figures actually were for 2000: the figure estimated by the DOF for the age group 15–24 years was overestimated by 40%; for the age group 25–34 years, the overestimation was by 41.4%; for 45–54 years it was underestimated by 31.1%; and for the age group 55–64 years, the figure was underestimated by 32.9%.
These population undercounts erroneously elevated the ‘calculated breast cancer incidence’ in 2000 by 50% in women 45–64 years of age. The 2005 paper expressed alarm at the data’s unreliability: ‘This (2005) study shows these projections are subject to large discrepancies from the actual census data and support a need for restructuring of population estimation procedures’ .
This paper, however, only offers a mild caveat re ‘variability’ of data and only in the discussion : ‘Intercensal population projections… can be substantially incorrect, and have biased breast cancer incidence rate estimation in Marin County in prior decades.’
Likewise, errors in the data presentation of ER+ tumors appear to have similarly distorted the facts. A graphic display in Figure 2 derived from the data in Table 5 shows the incidence of ER+ tumors apparently plunging from 0.39% in 2002 to 0.17% in 2004. Table 5 provides the total number of women with invasive breast cancer each year as well as how many of these were ER+ and ER-. In order to accurately project trends in ER tumor receptor-positive or -negative histology, one needs data from all the invasive cancers. Unfortunately, the numbers in the subgroups do not add up to the total count. For the estrogen receptor data, for example, incomplete/missing numbers might distort ‘trend lines’ and falsely imply a lowered incidence of ER+ tumors when what is really happening is a lowered incidence of data: in 2001, of 210 women with invasive breast cancer, only 21 had missing ER data; in 2002, of 200 women with invasive breast cancer only 21 had missing ER data; in 2004, of 154 women with invasive breast cancer, 53 had missing ER data; in 2005, of 183 women with invasive breast cancer, 78 had missing ER data; in 2006, of 190 women with invasive breast cancer, 78 had missing ER data; and in 2007, of 205 women with invasive breast cancer, 82 had missing ER data.
We conclude that the paper by Ereman and colleagues  has not demonstrated what its title claims. Neither is Marin County a place of proven high breast cancer incidence, nor has a putative high incidence decreased as the WHI media blitz influenced the choices that women make about their HT. Another unsettling point is that the conflict of interest disclosure noted that one author has served as expert witness for plaintiff lawyers preparing hormone therapy litigation. It is our opinion that this paper does not withstand vigorous scrutiny.
Winnifred Cutler, Ph.D.
Athena Institute for Women's Wellness, Chester Springs, PA
Regula Burki, M.D.
1. Ereman RR, Prebil LA, Mockus M, et al. Recent trends in hormone therapy utilization and breast cancer incidence rate in the high incidence population of Marin County, California. BMC Public Health 2010;10:228.
2. Fournier A, Berrino F, Clavel-Chapelon F. Unequal risks for breast cancer associated with different hormone replacement therapies: results from the E3N cohort study. Breast Cancer Res Treat 2008;107:103-11.
3. Lyytinen H, Pukkala E, Ylikorkala O. Breast cancer risk in postmenopausal women using estradiol-progestogen therapy. Obstet Gynecol 2009;113:65-73.
4. Franke H, Vermes I. Differential effects of progestogens on breast cancer cell lines. Maturitas 2003;46(Suppl 1):S55-8.
6. Phipps AI, Clarke CA, Ereman RR. Impact of intercensal population projections and error of closure on breast cancer surveillance: examples from 10 California counties. Breast Cancer Res 2005;7:R655-60.
*Related book by Dr. Cutler: Hormones and Your Health