The breast cancer “plunge” after initial publication of the WHI results: An alternative explanation

Maturitas 66 (2010) 277–284

Contents lists available at ScienceDirect

Maturitas journal homepage: www.elsevier.com/locate/maturitas

The breast cancer “plunge” after initial publication of the WHI results: An alternative explanation James A. Simon a,b,∗ , Gerard G. Nahum c , Harold Stanislaw d , Tatiana Gaines b a

Obstetrics and Gynecology, George Washington University School of Medicine, Washington, DC, United States Women’s Health & Research Consultants® , Washington, DC, United States Global Clinical Development, U.S. Women’s Healthcare, Bayer HealthCare Pharmaceuticals, Montville, NJ, United States d Department of Psychology, California State University, Stanislaus, Turlock, CA, United States b c

a r t i c l e

i n f o

Article history: Received 22 January 2010 Received in revised form 1 March 2010 Accepted 9 March 2010

Keywords: Breast cancer Hormone therapy Estrogen–progestin therapy Estrogen-only therapy HRT HT ERT ET WHI study SEER 9 database

a b s t r a c t From 2002 to 2003, the breast cancer incidence in the United States, as reported by the National Cancer Institute’s Surveillance Epidemiology and End Results (SEER 9) database, appeared to decrease by 6.7%. This phenomenon has been attributed to a reduction in the use of menopausal hormone therapies after the initial publication of the Women’s Health Initiative (WHI) study results in July of 2002. However, attempts to draw a causal association between the use of menopausal hormone therapies and the incidence of breast cancer have not accounted for the facts that prescriptions of estrogen-plus-progestin menopausal therapies, which are associated with increased rates of breast cancer, fell by 53% from 2002 to 2003, while prescriptions of estrogen-only therapies fell by only 27%. To address this issue, we analyzed the effects of the higher rate of discontinuation of estrogen-plus-progestin menopausal therapies relative to estrogen-only treatments during the 2002–2003 time period, based upon the effects of different types of menopausal hormone therapies on breast cancer incidence as determined by the WHI interventional hormone trials. This approach demonstrates that the relative persistence with menopausal estrogen-only therapies – as compared to estrogen-plus-progestin therapies – can explain the reduction in breast cancer incidence from 2002 to 2003. In addition, we point out the incompatibility of the breast cancer incidence rates found in the two WHI interventional hormone trials and the rates reported in the SEER 9 database. Based on these findings, we conclude – as previously demonstrated in the estrogen-only arm of the WHI interventional hormone trials – that menopausal estrogen-only use is not responsible for increasing the risk of breast cancer in menopausal women and may, in fact, be protective. Additional studies are still needed to better define the relationship between different types of menopausal hormone therapies and the incidence of breast cancer. © 2010 Elsevier Ireland Ltd. All rights reserved.

“Discoveries of fundamental importance may be in store for those who are adventurous enough to leave the safe ground of conventional assumptions and search for newer aspects of reality.” Dr. Alexis Carrel (June 28, 1873 to November 5, 1944)1 1. Introduction The relationship between menopausal hormone therapy (HT2 ) and breast cancer has been studied extensively. In 1997, the Collab-

∗ Corresponding author at: Women’s Health and Research Consultants® , Washington, DC, United States. Tel.: +1 202 293 1000. E-mail addresses: [email protected], [email protected] (J.A. Simon). 1 Alexis Carrel was a French surgeon and biologist who was awarded the Nobel Prize for Physiology and Medicine in 1912. 2 In this article, the generic term “hormone therapy” (HT) refers to all forms of postmenopausal estrogen therapy, including those with the addition of a progestin 0378-5122/$ – see front matter © 2010 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.maturitas.2010.03.011

orative Group on Hormonal Factors in Breast Cancer re-analyzed approximately 90% of the worldwide epidemiological data available regarding more than 160,000 women [1]. The relative risk (RR3 ) of breast cancer was 1.023 (95% CI 1.011–1.036); current and recent HT users were 2.3% more likely to have a diagnosis of breast cancer for each year of HT use. The cumulative RR over a 5-year period of use was 1.35 (95% CI 1.21–1.49), but there was no difference in breast cancer incidence 5 or more years after discontinuing HT. Similar trends were observed among the 1,084,110 UK women surveyed between 1996 and 2001 in the “Million Women Study”; current HT users were more likely than never users to develop breast cancer at an adjusted RR of 1.66 (95% CI 1.58–1.75) [2].

(Estrogen + Progestin [E + P] therapy) [typically used only in non-hysterectomized women] and those without a progestin (Estrogen-only [E-only] therapy) [typically used in hysterectomized women]. 3 Relative risk (RR) and hazard ratio (HR) both pertain to the concept of risk ratios,

278

J.A. Simon et al. / Maturitas 66 (2010) 277–284

Fig. 1. Kaplan–Meier estimates of cumulative hazards for invasive breast cancer. (Left) Estrogen plus progestin therapy associated with an increased risk of breast cancer in the WHI [4]; (right) estrogen-only therapy associated with a decreased risk of breast cancer in the WHI [5].

But the most highly publicized findings regarding breast cancer and postmenopausal HT came from the Women’s Health Initiative (WHI) in the United States [3,4]. The WHI study included a large, prospective, randomized, double-blind, placebo-controlled clinical investigation of more than 16,000 women, aged 50–79 years, who used Estrogen + Progestin (E + P) from 1993 to 2002 [4]. This NIH-sponsored trial was terminated on May 31, 2002, after the previously unknown and non-validated “global index” – which included the incidence of invasive breast cancer – exceeded a prespecified stopping boundary for the active treatment group. The hazard ratio (HR3 ) for invasive breast cancer was initially reported as 1.26 after an average of 5.2 years of use (a 26% increased risk of breast cancer) (adjusted 95% CI 0.83–1.92, Fig. 1 left) [4]. A second, estrogen-only (E-only) interventional study of the WHI involving hysterectomized postmenopausal women was terminated early in 2004. At that time, a statistically non-significant 23% decrease in the incidence of invasive breast cancer was noted for women treated with estrogen alone (HR 0.77, 95% CI 0.59–1.01, Fig. 1 right) [5]. Further analysis showed that the incidence of breast cancer associated with E-only treatment was decreased by 20% (HR 0.80, 95% CI 0.62–1.04) [6], with a statistically significant 33% decrease for a subgroup of women who were adherent to 80% or more of their medication (HR 0.67, 95% CI 0.47–0.97, Fig. 2) [6].

The seemingly contradictory results from the WHI studies can be summarized as follows (Fig. 2): • For Estrogen-only therapy – no increased risk of invasive breast cancer in postmenopausal women without a uterus, at least over a mean of 7.1 years of treatment; the risk was statistically significantly decreased if women were compliant in taking at least 80% of their medication [5,6]. • For Estrogen + Progestin therapy – a small statistically significant increased risk of invasive breast cancer after a mean of 5.6 years of treatment in women with a uterus. This might have been attributable to the progestin component rather than the estrogen component of the treatment [7], prior use of hormone therapy [4,8], or other factors.

where the risk of an event occurring in one group (i.e., in this case an “exposed group”) is compared to the risk of the same event occurring in another group (i.e., in this case an “unexposed group”) in a ratio of the type: RR(or HR) =

P1 P0

where P1 is the probability of the event occurring to the members in the exposed group, P0 the probability of the event occurring to the members in the unexposed group. For the results of the WHI, the HR can be used to estimate the RR because the relative risk appears constant (the risk functions are linear) over the study interval that was investigated. Because of this linearity, the RR can be considered constant and equal to the HR over any time span up to the duration of the WHI studies. In addition, since the baseline probability of the event of interest (breast cancer) is much less than 10% (i.e., <1-in-1000), the RR (and HR) is also approximated by the associated odds ratio (OR) for the occurrence of the event [35–37].

Fig. 2. Comparison of the change in breast cancer incidence for postmenopausal women in the Estrogen-only (E-only) versus the Estrogen + Progestin (E + P) arms of the WHI trial [All Subjects (those incorporated in the Intent-to-Treat analysis) and Adherent Subjects (those incorporated in the Per Protocol analysis)] (Graphic courtesy of Maida Taylor, MD).

J.A. Simon et al. / Maturitas 66 (2010) 277–284 Table 1 Change in annual age-adjusted breast cancer incidence in 2004 compared to 2001 (%) [10]. Group

Breast cancer incidence (%)

95% CI (%)

Women of all ages <50 years 50–69 years ≥70 years

−8.6 +1.3 −11.8 −11.1

−6.8 to −10.4 −3.1 to +5.8 −9.2 to −14.5 −7.9 to −14.2

2. “The plunge” in breast cancer incidence following WHI’s “bad news” Shortly after the WHI findings were published, the Surveillance Epidemiology and End Results (SEER 9) database maintained by the National Cancer Institute (NCI) showed a drop in the reported incidence of breast cancer in the United States [9]. The age-adjusted rate of breast cancer was −6.7% lower in 2003 than in 2002 [10]. The decrease was evident only in women 50 years of age or older. A comparison between the breast cancer rates in 2001 and 2004 showed an overall decrease of −8.6% in the annual age-adjusted incidence; for subgroups, there was an increase of +1.3% for women under 50 years old, a decrease of −11.8% for women age 50–69 years, and a decrease of −11.1% for women 70 years of age and older (Table 1) [10]. Concurrently, beginning in 2002 and particularly in 2003, there were steep declines in sales of the two most commonly prescribed forms of HT in the U.S. (Table 2): CEE (Conjugated Equine Estrogens) [Premarin] and CEE + MPA (Medroxyprogesterone Acetate) [Prempro] [11]. These are the same two products that were investigated in the WHI and they comprise the bulk of prescriptions filled for all forms of E-only and E + P therapies in the United States during 2001–2004 (Table 3) [11]. The very rapid decrease in breast cancer incidence (which occurred within 6 months of the initial publication of the WHI results) was of both practical and theoretical interest. Scientific evidence points to a minimum progression time of at least 5 years for a first breast cancer cell to become a clinically detectable lesion of ∼1 cm in dimension [12]. Therefore, these data were felt to be most consistent with an effect of HT on pre-existing pre-clinical breast cancer, but this did not rule out possible additional contributions from concurrent changes in usage patterns involving

279

screening mammography. An analysis by the NCI’s Cancer Intervention and Surveillance Modeling Network concluded that the effects of discontinuing HT are complex and probably depend on several mechanisms that could be differentially affected by various forms of therapy [13]. In a parallel study using data from a large prepaid U.S. health plan, breast cancer rates were compared with the use of screening mammography and HT prescriptions dispensed between 1980 and 2006 [14]. The overall adjusted breast cancer incidence rose +25% from the early 1980s, increased an additional +15% during 2000–2001, fell −18% from 2003 to 2004, and increased slightly thereafter. These patterns were largely restricted to women ≥45 years old and to ER+ breast cancers; rates for ER− breast cancers were relatively stable from 1980 to 2001 and fell significantly from 2003 to 2006. The rates of mammographic screening increased from 1980 to 1993 and then plateaued, with 75–79% of women ≥45 years old undergoing mammography at least once every 2 years from 1993 to 2006. Dispensing of HT, particularly E + P, increased from 1988 to 2002, then fell dramatically after 2002 (Tables 2 and 3). The group concluded that qualitative and quantitative trends in breast cancer incidence, particularly ER+ tumors, paralleled the changing patterns of mammographic screening and HT use. 3. Alternative explanations previously offered for the breast cancer “plunge” Peer-reviewed responses concerning the 2003 decrease in breast cancer incidence cited discrepancies that challenged a direct correlation between the reduced use of menopausal HT and the breast cancer rate “plunge” [15]. These included: • Inconsistent findings during the same time in Canada, Norway, and Sweden, where breast cancer rates remained stable despite sizeable decreases in HT prescription rates [16,17]. • Several studies that found no significant decrease in breast cancer risk when HT use was discontinued, even up to 10 years later (Collaborative Group 1997) [1]. • An American Cancer Society study that found a −10.6% overall decline in breast cancer incidence between 1999 and 2002, when

Table 2 Premarin (CEE) and Prempro (CEE + MPA) total prescription units sold in the U.S. during the period 2001 through 2004, with changes per year (in %) [11]. Total prescription units sold

Premarin Prempro Total Premarin % of total prescription units sold

2001

2002

2003

2004

66,905,104 28,883,214 95,788,318 69.8%

58,472,721 21,317,740 79,790,461 73.3%

38,868,700 7,848,442 46,717,142 83.2%

29,053,913 6,079,575 35,133,488 82.7%

Change (%) 

Premarin Premprob a b

02/ 01



−12.6 −26.2

a

03/ 02



−33.5 −63.2b

04/ 03

−25.3 −22.5

a

Premarin (CEE) change in total Rx units sold from 2002 to 2003: (38,868,700–58,472,721)/58,472,721 × 100% = −33.5%. Prempro (CEE + MPA) change in total Rx units sold from 2002 to 2003: (7,848,442–21,317,740)/21,317,740 × 100% = −63.2%.

Table 3 Reductions in total oral E-only and E + P prescription units sold by 6-month period from 2002 through 2004 (% change from pre-WHI prescriptions) [11]. 1st half 2002 Total oral estrogen-only market Total oral Estrogen + Progestin market

a

Baseline (=100%) Baselinea (=100%)

2nd half 2002

1st half 2003

2nd half 2003

1st half 2004

2nd half 2004

−17% −37%

−27% −53%

−37% −63%

−45% −67%

−49% −70%

a Baseline = 100% = prescriptions for the period 6 months immediately prior to the initial publication of the WHI results = volume of E-only or E + P prescription sales for 1st half of 2002, including all oral formulations (exclusive of transdermal products).

280

J.A. Simon et al. / Maturitas 66 (2010) 277–284

Fig. 3. Percentage of female Kaiser Permanente Northwest (KPNW) members who were dispensed at least one unopposed estrogen or estrogen-plus-progestin prescription per year, 1988–2006, by age group (modified from Glass et al. [14]).

HT use was relatively stable (Fig. 3), and a −5% drop in breast cancer incidence in 2002, before the WHI findings were announced and HT use began to drop [18]. • The WHI itself, which found no significant decrease in breast cancer risk after the end of the E + P trial (i.e., when the “plunge” occurred). In fact, a decline in breast cancer incidence did not occur until more than 3 years following the end of the E + P trial [19]. Of particular interest is that 45–50-year-old women were most numerous in the group that discontinued HT after 2002, but this age group showed no decrease in breast cancer incidence in the SEER 9 database. Conversely, women over 70 years old – who rarely use HT – had a −11.1% reduction in risk [10,20]. Furthermore, HT use continued to decline after 2003, but this was not associated with a further decline in breast cancer incidence [10].

4. Analysis of the inconsistencies in breast cancer incidence between the WHI interventional studies and the SEER 9 epidemiologic database Despite the obvious appeal of attempting to attribute the breast cancer “plunge” of 2003 to a decline in menopausal HT use, the invasive breast cancer rates reported by the WHI interventional hormone studies cannot be reconciled with those in the SEER 9 database. To demonstrate this, a Cox proportional hazards regression analysis was performed to compare the incidence of invasive breast cancer among placebo users in the WHI E + P study (0.33% annualized risk) [8] and the WHI E-only study (0.34% annualized risk) [5,6]. In spite of minor demographic differences among the women in the two WHI hormone study groups, this analysis showed – as would be expected – that the women in the two placebo groups had identical invasive breast cancer risks (HR = 0.98, 95% CI = 0.86–1.10, p = 0.69). Thus, having undergone a hysterectomy does not affect the incidence of invasive breast cancer in postmenopausal women; placebo users have a uniform annualized invasive breast cancer risk of 0.33% (as determined by a life table analysis of the combined data from both placebo groups). This confirms that the demographic differences between the (non-hysterectomized) women in the WHI E + P study and the (hysterectomized) women in the WHI E-only study did not have a significant impact on the intrinsic risk of breast cancer for the two groups of women; the breast cancer risk was identical for women treated with placebo in both groups.

Next, we assumed (as was done implicitly in the initial analysis and interpretation of the SEER 9 breast cancer trends [10]) that E + P (i.e., Prempro) cessation immediately lowers a nonhysterectomized woman’s risk of invasive breast cancer from 0.41% (employing the WHI data [8]) to the placebo rate of 0.33%. Similarly, we assumed that E-only (i.e., Premarin) cessation immediately raises a hysterectomized woman’s risk of invasive breast cancer from 0.28% (as reported by the WHI [6]) to the placebo rate of 0.33%. Other authors neglected this latter effect, which – importantly – contradicts the notion that any reduction of postmenopausal HT use reduces the incidence of invasive breast cancer. Finally, we assumed that the ratio of actual users to prescription sales units sold is the same for E-only therapies (i.e., Premarin) as for E + P (i.e., Prempro). This enabled us to use the prescription sales data (Table 2) to estimate the ratio of E-only to E + P users as follows: Calendar year

Ratio of prescription sales of E-only to E + P

2001 2002 2003 2004

66,905,104:28,883,214 = 2.32 58,472,721:21,317,740 = 2.74 38,868,700:7,848,442 = 4.95 29,053,913:6,079,575 = 4.78

These figures show what is already widely appreciated: Premarin (E-only) use is – and always has been – far more widespread than Prempro (E + P) use among menopausal HT users. What is much less well appreciated, however, is that these ratios render the WHI breast cancer rates fundamentally incompatible with those of the SEER 9 database. This incompatibility exists for every year, but can be illustrated by initially considering 2001. If we assume that approximately 30% of all eligible women used HT in 2001 (which is likely a high estimate) [14], then 9.1% of all postmenopausal women used E + P, 20.9% used E-only (consistent with the ratio of 2.32 E-only users for every E + P user), and the remaining 70.0% used no HT products at all. When the invasive breast cancer rates from the WHI interventional studies are combined with these percentages, the projected rate of invasive breast cancer for 2001 is: (0.091 × 0.41%) + (0.209 × 0.28%) + (0.700 × 0.33%) = 0.327% However, the actual SEER 9 invasive breast cancer incidence in 2001 was 0.385%, which is 17.7% higher than the WHI interventional study data predict. Prescription sales data demonstrate that E + P use fell far more than E-only use in subsequent years (Table 2). If 9.1% of women

J.A. Simon et al. / Maturitas 66 (2010) 277–284

used E + P in 2001, the prescription data imply that only 1.9% of women did so in 2004, while 9.1% of women used E-only in 2004; the remaining 89.0% used no HT products at all. These percentages, when combined with the WHI findings, lead to a projected annualized invasive breast cancer incidence of 0.327% – precisely the same rate as for 2001 – while the SEER 9 incidence for 2004 was 0.340% (4.0% higher). In fact, the annualized invasive breast cancer incidence that is projected by combining the WHI findings with the prescription sales data from 2001 to 2004 is virtually identical for all 4 years (0.327% for 2001, 0.326% for 2002, 0.326% for 2003, and 0.327% for 2004), despite very significant declines in overall HT usage over that period. In the unlikely event that the percentage of eligible women who used menopausal HT in 2001 was greater than 30% (which would also increase the percentage of HT users in subsequent years), the projected risk of invasive breast cancer would actually decrease, because the increased rate of breast cancer associated with E + P use would be more than offset by the decreased rate of breast cancer associated with E-only use, which has always been far more prevalent. In fact, the highest possible projected breast cancer incidence occurs when it is assumed that no menopausal HT products are used at all by the at-risk population. The projected incidence of invasive breast cancer is then 0.33% – the WHI placebo rate. However, the SEER 9 annualized breast cancer rate for women over 50 years of age has been above 0.33% since 1985 [21]. This rate peaked in 1999 at 0.396%, declined to 0.385% in 2001, and diminished further to 0.340% in 2004. Thus, there is a major – and fundamentally irreconcilable – discrepancy between the breast cancer incidences found in the WHI interventional hormone studies and the SEER 9 database. This discrepancy calls into question the breast cancer incidences reported by one or the other of these two sources. The women in the WHI interventional studies and SEER 9 database differ in their demographic characteristics, but – importantly – these differences mask the true discrepancy between the breast cancer rates from the two sources; i.e., they operate in the opposite direction from what would be necessary to explain the discrepant breast cancer rates. For example, in 2002 the SEER database showed a higher annualized incidence of breast cancer for white women than for other races (e.g., the incidence was 0.39% for whites, compared to 0.33% for Blacks and 0.25% for Asian/Pacific Islanders, in women aged ≥50 years) [21]. The percentage of white women was higher in the WHI hormone studies (81% in the E-only arm and 89% in the E + P arm, including Hispanics) [4,5] than in the SEER 9 database (71%) [22]. Therefore, the racial differences between the women in the WHI studies and the SEER 9 database create a bias toward a higher breast cancer incidence in the WHI studies. Despite this, the reported breast cancer rates are lower in the WHI studies than in the SEER 9 database. Similarly, women entered the WHI hormone studies at a slightly younger age (a mean of 64 years in both the E + P and E-only arms) [4,5] compared with the SEER 9 database (65 years) [23]. However, because the SEER 9 data are adjusted to match a constant age profile [21], the women under observation did not age. By contrast, the WHI participants were followed longitudinally for an average of 5.6 years (E + P arm) [4] or 7.1 years (E-only arm) [5], so that the WHI participants were, on average, 4–6 years older than the SEER 9 women at the end of observation. The risk of breast cancer increases with advancing age; for example, in the SEER 9 database itself, in 2002 the annualized breast cancer incidence increased from 0.32% for women aged 50–64 years, to 0.44% for women aged 65–74 years, to 0.45% for women ≥75 years old [21]. Thus, the age disparity between the women in the WHI hormone studies and the SEER 9 database – like the racial differences noted above – should bias the results in favor of a higher annualized breast cancer incidence in the (older) WHI women relative to the (younger) SEER 9 women.

281

However, the two data sources exhibit just the opposite pattern with regard to breast cancer risk. As a consequence of these data, if adjustments were made to correct for the demographic differences between the WHI interventional hormone studies and the SEER 9 database, the breast cancer rates predicted from the WHI findings would be even lower, further increasing their disparity with the reported SEER 9 rates. Based on these findings, demographic differences cannot explain the discrepancy in breast cancer incidence between the WHI interventional studies and the SEER 9 database. The WHI trials were prospective, randomized, blinded, and controlled. These characteristics would ordinarily render its findings more reliable than those of the SEER 9 database, which is an observational source. However, enrollment in the WHI studies was selective, with multiple exclusion criteria, and many of the women were concurrently enrolled in a dietary modification study [24]. These additional factors may help to explain the otherwise marked discrepancy between the two sources of data regarding the breast cancer incidence in the United States.

5. The SEER 9 database findings in isolation: Is persistent estrogen-alone exposure a plausible explanation for the “plunge” in the breast cancer incidence in 2003? Despite the irreconcilable differences between the SEER 9 data and the widely accepted results from the WHI interventional studies, if we nevertheless accept the two sets of findings in isolation, there is an alternative explanation for the SEER 9 breast cancer “plunge.” The most precipitous reduction in HT use occurred in exposure to E + P therapies, rather than E-only treatments. This is evident from hormone prescription sales data (Tables 2 and 3) and published analyses of menopausal hormone treatment persistence [25–27]. E-only therapies are associated with a reduced risk of invasive breast cancer, while E + P therapies are associated with an increased risk. Thus, we may posit that it is the continuation of Eonly therapies, not the discontinuation of E + P therapies, which can be used to explain the majority of the “plunge” in breast cancer incidence. By way of explanation, consider the following: The WHI findings imply that, if Prempro and Premarin were used equally often, the incidence of breast cancer would be higher than if no women used HT products. For example, if 10,000 women used Prempro and another 10,000 women used Premarin, 41 cases of invasive breast cancer would be expected each year among the Prempro users and 28 cases would be expected among the Premarin users. The total number of breast cancer cases for the two combined “hormone therapy” groups would then be 41 + 28 = 69 cases, as compared with only 33 + 33 = 66 total cases had no hormone therapy been used at all; the overall rate of breast cancer for the two “hormone therapy” groups combined would then be 0.345%, which is higher (by 4.5%) than the baseline rate of 0.33% that was observed for the women exposed to placebo in the WHI studies. However, this is true only if both products are used to the same extent; the projected breast cancer incidence drops to the placebo rate when the ratio of E-only users to E + P users is 1.6-to-1, and it falls below the placebo rate when this ratio is exceeded. Prescription sales data show that E-only users outnumbered E + P users by 2.32-to-1 in 2001. Thus, the invasive breast cancer risk should have been below the background “placebo rate” for postmenopausal women even before the “plunge” of 2003 began. By 2004, the ratio of E-only to E + P users more than doubled, to 4.78to-1. This change should have reduced the invasive breast cancer incidence downward even further toward the rate associated with E-only usage. Thus, if the 2003 “plunge” in breast cancer incidence can be attributed to changes in the use of menopausal HT, the primary

282

J.A. Simon et al. / Maturitas 66 (2010) 277–284

source of this change appears to be with the relatively high continued use of E-only treatments (e.g., Premarin) in the face of the much more rapidly decreasing rates – but the numerically much smaller drop in terms of absolute numbers – of E + P use (e.g., Prempro). It is important to note that this holds true even if the precise values of breast cancer risk estimated by the WHI studies are questioned. Since these studies imply that E-only use reduces the incidence of invasive breast cancer below the placebo rate while E + P use results in an increase above the placebo rate, and since the prescription data demonstrate clearly that E-only use greatly exceeds E + P use – especially after the WHI findings were released – the breast cancer rates after 2002 were affected far more by any favorable effects resulting from the continued use of E-only products than by negative effects that resulted from prior E + P use. 6. An important additional consideration: The possibly non-identical effects of different estrogens and progestins on breast tissue The WHI studies investigated only one estrogen (CEE) and only one progestin (MPA). Both were studied at a single dose (0.625 mg/d for CEE and 2.5 mg/d for MPA) and in a single oral dosing schedule (once daily). Despite this, it has been assumed that the results are generalizable to other types, doses, dosing regimens, and delivery systems for menopausal estrogens and progestins, although there are currently no clear data to confirm this. Besides CEE – which contains well over a hundred different estrogenic substances, many of non-human type – there are many other estrogens available for the treatment of menopausal women, including several that are biologically very similar to the estrogens that are produced endogenously by the ovaries of younger women during their reproductive years, such as estradiol and estrone. In industrialized countries other than the United States, CEE products are used only rarely by menopausal women and other types of estrogens are used more routinely (e.g., estradiol, estradiol valerate, ethinyl estradiol, etc.). Although the therapeutic benefits of several different types of estrogens administered in different doses, regimens, and routes of delivery have been shown to be similar to CEE, it is not known whether the risks associated with their use are equivalent. The same is true for progestins other than MPA. Both progesterone and other synthetic progestins can exert a diverse spectrum of biologic activities in different tissues that are divergent from MPA. Besides their basic progestogenic actions, different progestins can exert agonist or antagonist effects on other steroid receptors, resulting in differential biological impacts on different target organs, including the breast and endometrium [28]. MPA has recently been compared to another progestin with regard to its relative breast and endometrial effects in in vivo animal studies [29]. These studies showed marked differences in tissue-specific biologic effects, supporting the hypothesis that not all progestins are the same and that some may be more endometrial-specific than others. Given that progestins are used and titrated in menopausal hormone therapies to exert appropriate anti-estrogenic effects on the endometrium, these tissue-specific differential affinities for breast tissue are potentially important. The results of the recent E3N cohort study from France suggest that such differential effects of progestins may apply to the breast effects that are seen in postmenopausal women who use E + P therapies [30]. While the U.S. Food and Drug Administration assumes that the adverse effects of all estrogens and progestins in postmenopausal women are the same [31],4 potential differences in the effects of

4 U.S. FDA Class Labeling recommended in the Guidance document entitled “Noncontraceptive Estrogen Drug Products for the Treatment of Vasomotor Symptoms

different estrogens and progestins at varying doses and in different regimens and delivery systems may exist with respect to their breast effects. To address this, additional clinical trials are needed to investigate the differential effects of the various estrogens and progestins that may be used by postmenopausal women. These types of differences have been recognized by the WHI investigators themselves [4], and may help to explain the disparate effects of the changes in HT usage patterns observed in the United States and other countries, as well as the inconsistent changes in breast cancer incidence that have been seen in different parts of the developed world. 7. Conclusions The invasive breast cancer incidences for menopausal women who were enrolled in the prospective, randomized, and controlled interventional arms of the WHI studies are fundamentally incompatible with those reported by the observational SEER 9 database. This raises questions as to the intrinsic accuracy of the SEER 9 data, as well as how applicable the WHI findings are to the women who are tracked using the SEER 9 database (i.e., the general population of the United States). Of particular note is that the demographic differences between the women in the WHI interventional hormone studies and the SEER 9 database cannot be used to explain the inconsistencies in the breast cancer incidence between the two: both the age and racial differences between the two populations of women operate in the opposite direction from what would be required to explain the differences in the reported incidence of breast cancer. Epidemiologic models – such as the ones used to evaluate the observational data collected by the SEER 9 database – are designed to elucidate potential associations that can exist between proposed predictor variables (e.g., in this case, certain types of hormonal exposures in postmenopausal women) and particular outcomes of interest (e.g., in this case, the incidence of breast cancer). But these models are – by their very nature – non-interventional and insufficient to establish the veracity of suspected causal relationships. The implications of the analyses presented in this article are clear: Using the same types of data and governing assumptions, different types of models can be constructed using different levels of data granularity, resulting in disparate results and different conclusions concerning the relationship of postmenopausal estrogen use to breast cancer risk. In our analyses, we included the WHI study’s own quantitative estimates of the reported opposing effects of CEE + MPA (an increase) and CEE-alone (a decrease) on the incidence of breast cancer, as well as the quantitative changes in the prescription sales of CEE + MPA (E + P) versus CEE-alone (E-only) during the time frame of interest. By assuming an instantaneous effect on breast cancer incidence (as was assumed in previous analyses of the association between menopausal HT use and breast cancer incidence [10]), it is possible to conclude that the entire effect size on the reduction of breast cancer incidence reported by the SEER 9 database (a decrease in breast cancer of −6.7%) [10] is explicable on the basis of the protective effects of estrogen-only therapy against the development of breast cancer in postmenopausal women. This finding is consistent with other studies in the scientific literature: for example, in a large community-based cohort, Kerlikowske et al. reported an 8% decrease (95% CI −16% to 0%) in breast cancer incidence

and Vulvar and Vaginal Atrophy Symptoms” (CDER Draft Guidance, November 2005 [Revision 4]) currently states: “Other doses of oral conjugated estrogens with medroxyprogesterone acetate, and other combinations and dosage forms of estrogens and progestins were not studied in the WHI clinical trials and, in the absence of comparable data, these risks should be assumed to be similar.” [31].

J.A. Simon et al. / Maturitas 66 (2010) 277–284

among postmenopausal women using estrogen alone for more than 5 years, as compared to those not using hormones at all [32]. As a caveat, it is important to recognize that the integrity of the data concerning the incidence of breast cancer, both in the WHI interventional studies and in the SEER 9 database, depends on assuming that all breast cancers are accurately identified. Unfortunately, there is a well-known false negative rate associated with all known screening tests for breast cancer (e.g., manual examination by medical practitioners, periodic breast self-examinations, screening mammography, MRI, etc.), meaning that not all such lesions will be detected. In addition, postmenopausal hormone therapy may have an influence on breast density, which may in turn influence the sensitivity and specificity of screening tests, including mammography. Thus, we (and others) can only comment on the breast cancer data that have been reported by the current studies, all of which make an implicit assumption concerning the accuracy of screening methods for the identification of breast cancers. However, we recognize that the correlation between “what actually exists” and “what can be identified by normal screening methods” in women’s breasts is imperfect. Although this is an important limitation, it exists for all such studies that examine the incidence of breast cancer, not just the current ones. In this article, we have presented evidence that includes a detailed analysis of human populations to show that the reduction in breast cancer incidence following the initial publication of the WHI findings in July 2002 was primarily due not to postmenopausal women discontinuing their HT, but to hysterectomized (and mostly oophorectomized) women continuing to use their estrogen-only therapies more consistently than their non-hysterectomized counterparts who used regimens that contained both an estrogen and a progestin. This result is in contrast to numerous prior observational studies that have suggested that there is an increase in breast cancer incidence associated with postmenopausal HT use in general and estrogen use alone [1,2]. Our findings should cause all medical practitioners to take pause. Research findings that reveal small effect sizes – such as those found by the WHI interventional hormone studies – are much more likely to be false than true. Many such research findings are simply measures of the prevailing investigator and study design biases [33], which may both be inadvertent. So, leaping to conclusions concerning potential cause-and-effect relationships between “hormones” of various types and the incidence of breast cancer in postmenopausal women merely as a result of the associations promulgated by specialized epidemiologic models should never be regarded as conclusive. Perhaps the late Dr. Trudy Bush, an accomplished epidemiologist, summarized this conclusion most eloquently when she said: “No single study has a monopoly on the truth.” [34]

283

Ingelheim (Ingelheim, Germany), FemmePharma (Wayne, PA), GlaxoSmithKline, Nanma/Tripharma/Trinity, Novartis (Basel, Switzerland), and Proctor and Gamble (Cincinnati, OH). He has also served on the speaker’s bureaus of Ascend, Barr, Bayer, GlaxoSmithKline, KV Pharmaceutical Co., Merck, Novartis, Novogyne, Sciele, Warner Chilcott, and Wyeth. Dr. Nahum is the Head of Global Clinical Development U.S. in Women’s Healthcare at Bayer HealthCare Pharmaceuticals in Montville, New Jersey, where he has also served as the Senior Director of Medical Affairs for Women’s Healthcare in the United States. Previously, he was a Medical Officer within the Office of New Drugs at the U.S. Food and Drug Administration in Rockville, Maryland and an Associate Professor of Obstetrics and Gynecology at the Duke University School of Medicine in Durham, North Carolina. Bayer HealthCare Pharmaceuticals (and its predecessor organization, Berlex Laboratories) manufactures and distributes hormonal products for the management of menopausal symptoms and menopause-associated conditions, including the treatment of moderate-to-severe vasomotor symptoms associated with the menopause, vulvar and vaginal atrophy associated with the menopause, and prevention of postmenopausal osteoporosis. Dr. Stanislaw is Professor and prior Chairman in the Department of Psychology, California State University, Stanislaus and has nothing to disclose. Tatiana Gaines, MD, is a research associate in the offices of James A. Simon, MD, PC and has nothing to disclose. Competing interest The author, James A. Simon, MD, CCD, FACOG, declares that neither he nor his any immediate family member has a current financial arrangement or affiliation with any organization(s) that may have a direct interest in the subject matter of the indicated CME program. Funding None. Provenance and peer review Not commissioned, externally peer reviewed. Acknowledgement The authors would like to thank Anja Schmidt, Ph.D., for her invaluable research assistance. References

Contributors Dr. Simon has served as a consultant or on the advisory boards of Allergan (Irvine, CA), The Alliance for Better Bone Health (Cincinnati, OH), Ascend Therapeutics (Herndon, VA), Barr (Pomona, NY), Bayer (Leverkusen, Germany), BioSante (Lincolnshire, IL), Corcept Therapeutics, Inc. (Menlo Park, CA), GlaxoSmithKline (Philadelphia, PA), KV Pharmaceutical Co. (St. Louis, MO), Meditrina Pharmaceuticals (Ann Arbor, MI), Merck (Whitehouse Station, NJ), Merrion Pharmaceuticals (Wilmington, NC), Nanma/Tripharma/Trinity (Glen Arm, MD), Novo Nordisk A/S (Bagsvrerd, Denmark), Novogyne (East Hanover, NJ), Pear Tree Pharmaceuticals (Cambridge, MA), QuatRx Pharmaceuticals (Ann Arbor, MI), Roche (Basel, Switzerland), Sciele (Atlanta, GA), Solvay (Marietta, GA), Warner Chilcott (Rockaway, NJ), and Wyeth (Madison, NJ). He has received grant/research support from BioSante, Boehringer

[1] Collaborative Group on Hormonal Factors in Breast Cancer. Breast cancer and hormone replacement therapy: Collaborative reanalysis of data from 51 epidemiological studies of 52,705 women with breast cancer and 108,411 women without breast cancer. Lancet 1997;350:1047–59. [2] Beral V, Million Women Study Collaborators. Breast cancer and hormonereplacement therapy in the Million Women Study. Lancet 2003;362:419–27. [3] Schwartz LM, Woloshin S. The media matter: a call for straightforward medical reporting. Ann Intern Med 2004;140:226–8. [4] Rossouw JE, Anderson GL, Prentice RL, et al., Writing Group for the Women’s Health Initiative Investigators. Risks and benefits of estrogen plus progestin in healthy postmenopausal women: Principal results from Women’s Health Initiative randomized controlled trial. JAMA 2002;288:321–33. [5] Steering Committee for the Women’s Health Initiative. Effects of conjugated equine estrogen in postmenopausal women with hysterectomy: The Women’s Health Initiative randomized controlled trial. JAMA 2004;291:1701–12. [6] Stefanick ML, Anderson GL, Margulis KL, et al. Effects of conjugated equine estrogens on breast cancer and mammography screening in postmenopausal women with hysterectomy. JAMA 2006;295:1647–57. [7] Simon, JA. Management issues. NAMS Menopause e-Consult. Available at http://www.menopause.org/ME0707.pdf [accessed October 20, 2008]; July 2007;3:3–6.

284

J.A. Simon et al. / Maturitas 66 (2010) 277–284

[8] Chlebowski RT, Hendrix SL, Langer RD, et al. Influence of estrogen plus progestin on breast cancer and mammography in healthy postmenopausal women: the Women’s Health Initiative Randomized Trial. JAMA 2003;289:3243–53. [9] Ravdin PM, Cronin KA, Howlander N, Chlebowski RT, Berry DA. A sharp decrease in breast cancer incidence in the United States in 2003, 29th Annual San Antonio Breast Cancer Symposium. Breast Cancer Res Treat 2006;100(Suppl. 1):S6. [10] Ravdin PM, Cronin KA, Howlander N, et al. The decrease in breast-cancer incidence in 2003 in the United States. N Engl J Med 2007;356:1670–4. [11] Information derived by Bayer HealthCare from data provided by Wolters Kluwer Health, Source (R) Retail data; 2001–2008. [12] Dietel M, Lewis MA, Shairo S. Hormone replacement therapy: pathobiological aspects of hormone-sensitive cancers in women relevant to epidemiological studies on HRT: A mini-review. Hum Reprod 2005;20:2052–60. [13] Berry DA, Cronin KA, Plevritis SK, et al. Effect of screening and adjuvant therapy on mortality from breast cancer. N Engl J Med 2005;353:1784–92. [14] Glass AG, Lacey JV, Carreon JD, Hoover RN. Breast cancer incidence, 1980–2006: combined roles of menopausal hormone therapy, screening mammography, and estrogen receptor status. J Natl Cancer Inst 2007;99(15):1152–61. [15] Ravdin PM, Cronin KA, Chlebowski RT, et al. A decline in breast-cancer incidence. N Engl J Med 2007;357:509–13. [16] Kliewer EV, Demers AA, Nugent ZJ. A decline in breast-cancer incidence. N Engl J Med 2007;357:509–10. [17] Zahl P-H, Maehlen J. A decline in breast-cancer incidence. N Engl J Med 2007;357:510–1. [18] Jemal A, Siegel R, Ward E, Murray T, Xu J, Thun MJ. Cancer statistics, 2007. CA Cancer J Clin 2007;57:43–66. [19] Heiss G, Wallace R, Anderson GL, et al. Health risks and benefits 3 years after stopping randomized treatment with estrogen and progestin. JAMA 2008;299:1036–45. [20] SEER Cancer Statistics Review. Available at http://seer.cancer.gov/csr/ 1975 2005/ [accessed October 20, 2008]; 1975–2005. [21] SEER Cancer Statistics. URL: http://seer.cancer.gov/faststats/selections.php? series=race. [22] SEER Cancer Statistics. URL: http://seer.cancer.gov/registries/data.html. [23] SEER Cancer Statistics. URL: http://seer.cancer.gov/stdpopulations/stdpop. 19ages.html. [24] Women’s Health Initiative Study Group. Design of the Women’s Health Initiative clinical trial and observational study. Control Clin Trials 1998;19:61–109.

[25] Simon JA, Wysocki S, Brandman J, Axelsen K. A comparison of therapy continuation rates of different hormone replacement agents: a 9-month retrospective, longitudinal analysis of pharmacy claims among new users. Menopause 2003;10:37–44. [26] Hersh AL, Stefanick ML, Stafford RS. National use of postmenopausal hormone therapy: annual trends and response to recent evidence. JAMA 2004;291(1):47–53. [27] Haas JS, Kaplan CP, Gerstenberger EP, Kerlikowske K. Changes in the use of postmenopausal hormone therapy after the publication of clinical trial results. Ann Intern Med 2004;140(3):184–8. [28] Schindler AE, Campagnoli C, Druckmann R, et al. Classification and pharmacology of progestins. Maturitas 2003;46S1:S7–16. [29] Otto C, Fuchs I, Altmann H, et al. Comparative analysis of the uterine and mammary gland effects of drospirenone and medroxyprogesterone acetate. Endocrinology 2008;149:3952–9. [30] 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. [31] FDA Guidance for Industry. Noncontraceptive estrogen drug products for the treatment of vasomotor symptoms and vulvar and vaginal atrophy symptoms – recommended prescribing information for health care providers and patient labeling. Available at http://www.fda.gov/cder/guidance/6932dft.pdf [accessed February 9, 2009]. [32] Kerlikowske K, Miglioretti DL, Ballard-Barbash R, et al. Prognostic characteristics of breast cancer among postmenopausal hormone users in a screened population. J Clin Oncol 2003;21:4314–21. [33] Ioannidis JPA. Why most published research findings are false. PLoS Med 2005;2(8):e124. [34] Lobo R. Evidence-based medicine and the management of Menopause [evidence-based approach to menopause]. Clin Obstet Gynecol 2008;51(3):534–8. [35] Zhang J, Yu KF. What’s the relative risk? A method of correcting the odds ratio in cohort studies of common outcomes. JAMA 1998;280:1690–1. [36] McNutt LA, Wu C, Wue X, Hafner JP. Estimating the relative risk in cohort studies and clinical trials of common outcomes. J Epidemiol 2003;157:940–3. [37] Simon S. Understanding the odds ratio and the relative risk. J Androl 2001;22:533–6.