Commentary: The Year in Basic Science: Update of Estrogen Plus Progestin Therapy for Menopausal Hormone Replacement Implicating Stem Cells in the Increased Breast Cancer Risk

Abstract

This transcript is based on my The Year in Basic Science lecture at ENDO 2008. I reviewed current data surrounding hormone replacement therapy and the relationship between systemic estrogen plus progestin (E+P) treatment and increased breast cancer risk, and I explored the hypothesis that women who develop breast cancer while on E+P had occult, undiagnosed disease before they started therapy. Beginning with recent hormone replacement therapy data focusing on E+P and its association with breast cancer to set the stage, the lecture then reviewed our newly published data that progestins expand breast cancer stem cells. Finally, the issues of occult or undiagnosed breast cancer in presumably healthy women, and of tumor dormancy in breast cancer survivors, were brought to bear on the discussion. Taken together, these apparently disparate themes allowed me to suggest the idea that systemic progestins have the ability to reawaken cancers that were presumed to be either nonexistent or cured. To avoid this potentially devastating outcome while retaining the benefits of E+P, I advocated the use of local P delivery methods, rather than the currently popular systemic routes.

AS PART OF The Year in Basic Science lectures, I was asked to present my views on the latest developments in hormones and cancer. Given the overwhelming scope of this field, I narrowed the focus to a topic of particular interest to me: hormone replacement therapy (HRT) and its possible association with an increased risk of breast cancer. Most of the presentation focused on papers published in 2007–2008; however, earlier data relevant to the issue at hand were also discussed. First, I reviewed the question of how HRT, in particular the combination of estrogen and progestin (E+P), became associated with breast cancer. Next, I explored the new field of breast cancer stem cells and how it might impact the HRT story, focusing on our recently published data that progestins reactivate breast cancer stem cells. Finally, I proposed a hypothesis linking HRT, undiagnosed breast cancer and tumor dormancy, and P effects on cancer stem cells that tie these concepts together.

HRT AND BREAST CANCER

In 2002, the first Women’s Health Initiative (WHI) data were published highlighting the link between E+P and breast cancer (1). The estrogen (E)-only data were published in 2005 (2). The WHI trial was a prospective, randomized, placebo-controlled study of US women who, on average, had started HRT 13 yr after menopause. The WHI protocol sought to evaluate the risks and benefits for women taking E only, or an E+P combination, compared with no-hormone controls. After 7 yr of E alone (2) there was an increased incidence of uterine cancers and pulmonary emboli and stroke, but E was protective on bone as evidenced by decreased fractures and was possibly protective with regard to cardiovascular disease and breast and colon cancer although the latter results were not significant. The slight decrease in breast cancer incidence was a welcome observation. Five years of the E+P combination (1) exhibited protective effects on bone, the uterus, and colon cancer. However, the incidence of breast cancers, thromboembolic events and stroke, and cardiovascular disease increased. After weighing the benefits of E+P against the harmful effects, it was decided to abort this arm of the study at 5 yr because, on balance, the therapy was doing more harm than good. The E-only arm (2) was continued for another 2 yr with the possible beneficial effects on breast cancer of considerable interest, despite the increased incidence of strokes.

Major criticisms of the WHI study have been extensively explored: Briefly, there was concern about the average age (66 yr) of participants at the start of therapy and the gap since menopause; there was concern about the hormones tested, namely an oral mixture of conjugated estrogens isolated from pregnant mare urine (Premarin; Wyeth-Ayerst Laboratories, Inc., Philadelphia, PA), and the P in the form of medroxyprogesterone acetate (MPA; Provera), which has mixed steroidal properties. Importantly, the statistical approaches of the study were criticized, especially the fact that early publications failed to note that the absolute risk of increased incidence in any of the relevant conditions, was actually quite small (2). For example, the absolute risk of breast cancer amounted to an increase of eight cases per 10,000 women-years of E+P use (3).

In 2008, Heiss et al. (4) reappraised the data from the perspective of 3 yr of E+P discontinuation. In general, all of the apparent harmful effects on myocardial infarction, stroke, deep vein thrombosis, and pulmonary emboli observed during the trial dissipated after hormone cessation. Termination of HRT also negated harmful effects on breast and colorectal cancers and dissipated beneficial effects on bone and endometrial and colon cancers. Thus cessation of E+P reversed any effects, whether beneficial or harmful.

The United Kingdom’s Million Women Study, published in 2003, generally corroborated results of the WHI trial: namely, that the incidence of breast cancer in women taking various E+P regimens was significantly increased (19 excess breast cancers per 1000 women) compared with women taking E alone (five excess breast cancers per 1000 women) despite the clearly protective effects of E+P on the uterus (5). In this study, the women participating were, at an average age of 55.9 yr, about a decade younger than the women in the WHI trial. The study was based on questionnaire data completed by participants undergoing mammography, which were integrated with the United Kingdom’s extensive cancer registry data. Like the WHI study, the increased risk of HRT was small, revealing few excess cancers compared with the total population.

Despite the statistical uncertainty surrounding both trials and others like them, it clearly became necessary for physicians and patients to carefully consider the risks and benefits of HRT and weigh its potential advantages against the potential deleterious effects. Consequently, the number of HRT prescriptions plummeted dramatically after 2002, as shown by data analyzed on approximately 9% of U.S. menopausal women (6). Logically, then, if E+P was responsible for the increased risk of breast cancer, once the number of prescriptions for these hormones dropped, so too should the cancer incidence.

Glass et al. (7) addressed this hypothesis in a 2007 paper based on pharmacy and tumor-registry data of more than 500,000 women insured by Kaiser Permanente Northwest who were taking E+P (Premarin and MPA before 2005; estradiol and MPA after 2005) from 1980 through 2006, during which 7386 cases of invasive breast cancers were diagnosed. The data show an apparent dramatic rise in the incidence of breast cancer that began in the mid-late 1980s; this rise has been attributed to the proportionate rise in the number of mammographies that were being performed. During this time, the number of women undergoing mammography rose from 5–50%, and then to 75% by the late 1990s. These mammograms were detecting early cancers (especially lymph node-negative small tumors localized to the breast), ones that would previously have gone undetected. This apparently accounted for the initial rapid, followed by subsequent slower, rise in the breast cancer-incidence rate to the year 2000. Incidentally, this relationship also demonstrated the value of mammography for early detection. Starting in 2002 however (7), the incidence of invasive breast cancers began to decline, especially in women over 45 yr of age. Significantly, the number of mammographies being performed during this period remained steady. The authors concluded that the recent drop in breast cancer incidence must be explained by the drop in E+P use.

A 2007 study by Clarke and Glaser (8) concurs with this assessment but with an important additional detail. Data from 3 million white, non-Hispanic California women aged 50–74 yr, showed the same abrupt drop in breast cancer incidence around 2002, just after publication of the WHI’s E+P results. Importantly, though, the decline was restricted to estrogen receptor (ER) and progesterone receptor (PR)-positive, hormone-dependent breast cancers. Hormone-independent cancers showed no change in incidence. Here again, there were no substantive changes in mammography screening patterns during the relevant time period so that widespread cessation of HRT (particularly of E+P) was considered to be the likeliest explanation for the decline in invasive breast cancer rates.

TYPES, ROUTES, AND GAP TIMES OF P ADMINISTRATION

Results of the above studies pointed to progestins as the likeliest culprits for the deleterious effects of E+P on women’s breasts. How, then, to safely manage postmenopausal symptoms in women with an intact uterus, and thus susceptible to the harmful endometrial effects of E alone, while guarding against breast cancer? Interest after 2002 focused on the type of P, natural vs. synthetic, on sequential vs. continuous regimens, and on routes of administration. Also of intensive interest was the time after menopause, called the “gap time” or “critical therapeutic window,” when HRT should be started.

Types of P and Breast Cancer Risk

The E3N-EPIC study authored by Fournier et al. in 2008 (9) assessed breast cancer risk for different HRT preparations, routes, and duration vs. never users in about 500,000 teachers in France covered by health insurance. An E-only regimen of less than 2 yr showed a barely statistically significant risk of increased breast cancer that was insignificant with longer duration. Indeed, after 6 yr, any such effect had dissipated. Combining E with natural progesterone or the pure synthetic P, dydrogesterone, likewise did not increase risk significantly. However, various other synthetic progestins that had mixed steroidal properties were all associated with increased risk even in regimens of less than 2 yr. Moreover, a statistically significant trend for increased risk with increased duration was observed. Results of two earlier studies support these conclusions. The first, based on the 2003 United Kingdom Million Women Study (10) showed increased breast cancer risk with increased duration of treatment, regardless of the type or regimen, sequential vs. continuous, of oral P. Also in 2003, the Three-County Washington State Study (11) based on 975 breast cancer cases and 1007 controls in the 2 yr between 1997–1999, reported that compared with never users, women taking any type of E alone experienced no significant increase in breast cancer. However, ever-users of E+P, regardless of type or regimen of synthetic P, had an increased risk of hormone-dependent but not of hormone-independent breast cancers. This addresses an important biological principle: only those cancers capable of responding to progestins, namely ones that express ER and PR, are increased by E+P regimens. It seems that oral synthetic progestins, regardless of regimen, are associated with increased breast cancer risk.

Alternative Routes of P Delivery

Alternate methods of hormone delivery, therefore, and perhaps use of natural progesterone, have become increasingly of interest. Most of the commonly used synthetic progestins [MPA, Levonorgestrel (LNG), and norethisterone acetate] exhibit mixed steroidal properties; they do not simply target PR. Whereas they all exhibit progestational properties, except for dydrogesterone, they are also androgenic, glucocorticoid, and/or mineralocorticoids. Moreover, when taken orally, these hormones traverse the gut wall and pass directly to the liver where they are metabolized, conjugated, and otherwise metabolically altered such that the end product and concentrations that enter the circulation and reach the breast, the uterus, and other sites pertinent to this discussion are not easily identifiable or quantifiable. That is, the systemic bioavailability of these hormones is highly varied. The advantage conferred by synthetic progestins over natural progesterone, however, is the ease with which they are orally absorbed. Natural progesterone is too poorly absorbed from the gastrointestinal tract to be effective although micronization improves this property, as in the formulation called Prometrium (Solvay Pharmaceuticals, Inc., Marietta, GA).

Transdermal administration, which bypasses the liver, is under extensive study. However, like the oral route, absorption through the skin also leads to the systemic distribution of hormones and targets the breast and other organs (12). Local hormone delivery methods, such as vaginal suppositories in which the first pass is into the uterus, or direct intrauterine devices (IUDs) that place the progestins precisely where they are needed, may therefore be preferable (13). Unfortunately, not much is known about effects of local P delivery with regard to breast cancer risk, and the possibility exists that even this route releases progestins into the systemic circulation (14). Clearly, local P delivery protects the uterus against the deleterious effects of estrogens. Maruo (15) examined effects of LNG delivered directly to the uterus via an IUD (Mirena; Bayer, Schering Pharma, AG, Berlin, Germany) on endometrial proliferative activity. With E alone, high proliferative activity was observed in both the endometrium and stroma, which was significantly reduced by 3 months of the LNG-releasing IUD. Interestingly, local delivery of LNG showed even stronger suppressive effects than oral MPA (16). In this study, the ability of LNG to protect against the hypertrophic effects of E by causing uterine atrophy, the major indication for using a P, was nearly double that of MPA. After 12 months, 96% of women on E plus the IUD experienced P-induced endometrial atrophy compared with 55% on E plus oral MPA. Moreover, after more than 5 yr, there were no cases of endometrial hyperplasia in 800 women on the LNG IUD in combination with different types and routes of estrogens.

So, does local uterine P delivery influence the rate of breast cancer? Unfortunately, such a study has not yet been done. Backman et al. (17) though, analyzed the breast cancer incidence in premenopausal Finnish women who were using the LNG IUD for contraception (17). Regardless of the age group, ranging from 30–54 yr, there was no difference in the breast cancer incidence in women with the implanted P-releasing IUD compared with average Finnish women of the same age not using the hormone. These data are encouraging and underscore the need for more studies on local delivery of HRT and breast cancer risk.

Gap Time for HRT Start

Another issue surrounding HRT is when to initiate it: Should a woman start right at menopause or wait? And if the latter, how long should she wait? A 2008 reanalysis of the E-only arm of the WHI trial assessed breast cancer risk based on years of Premarin use against the years after menopause when HRT was begun (18). When gaps of less than 5 yr, 5–15 yr, and greater than 15 yr were examined, it was observed that women who initiated E therapy soon after menopause were at greatest risk of developing breast cancer with possibly even decreased risk after longer gaps. However the data were not statistically significant, and the authors concluded that overall, they fail to provide clear evidence for either a reduction or increase in breast cancer risk.

A similar reanalysis of the E+P arm of the WHI trial, also by Prentice et al. (19) in 2008 examined whether gap time after menopause affected breast cancer risk, and here some of the data were statistically significant. As with E alone, E+P increased breast cancer risk if it was initiated within 5 yr of menopause. Waiting more than 5 yr was less harmful.

These data were further broken down (19) into hazard ratios for gap times of less than 5 yr, 5–15 yr, and greater than 15 yr, together with duration of E+P use, and the trend was clear: shorter gap times were more detrimental than longer ones. Moreover, as the duration of time on E+P expanded from less than 2 yr, to 2–5 yr, to more than 5 yr, the data grew progressively more statistically significant with the clear conclusion that a short gap time associated with long duration of hormone use was most deleterious. Thus, younger women seem to be at higher risk than older women, especially after long-term hormone use. Paradoxically, the prevalence of older women, which was heavily criticized in the original WHI publications, may, if anything, have masked some of the harmful effects of E+P.

However, this information does not simplify the life of a prescribing physician. Rossouw et al. (20) in a different 2007 reanalysis of the WHI data, reported an opposite trend for HRT gap time with respect to cardiovascular disease. Although not statistically significant, these data pointed to a short gap time if the cardiovascular benefits of HRT were to be realized. The relative risks and confidence intervals for E alone, E+P, and all HRT, showed a nonsignificant trend for reduced cardiovascular risk if HRT was initiated close to menopause. In summary, as a general rule, it would seem that HRT should be started in younger women to derive its beneficial effects, but delayed to avoid its harmful ones. In other words, if for example, a woman exhibits high risk for osteoporosis or cardiovascular disease, HRT should be initiated at or soon after menopause. If breast cancer is a major concern, however, a significant gap time should be considered before HRT is begun.

The biological foundations for this dilemma remain unclear but it is likely that the hormone-responsive target tissues contain ER and PR, and what is needed is a precise understanding of the role of E and P at each of those sites. Erroneous assumptions, like those made in the past that if progestins are protective in the uterus, they must also be protective in the breast, should be discarded in favor of accurate mechanistic data that define harmful vs. beneficial effects at specific targets. At present, to avoid breast cancer, use of the lowest possible hormone doses is recommended, although as yet there are insufficient data to support that premise. There is considerable interest in the use of transdermal systems even for estrogens, because this route may minimize endometrial hyperplasia compared with oral estrogens. Data suggest that synthetic progestins with mixed steroidal activities should be avoided in favor of pure progestins, and minimizing or eliminating systemic progestins by delivering them either through the vagina or directly to the uterus may retain their major protective function while reducing systemic exposure of other target organs (for updated recommendations, see Ref. 21). Finally, it is my considered opinion that HRT should not be given to breast cancer survivors as I discuss below with regard to cancer stem cells and disease dormancy.

PROGESTINS AND BREAST CANCER STEM CELLS

The cancer stem cell theory holds that tumors originate from mutated but mitotically quiescent cells that have the capacity to divide asymmetrically, leading to self-renewal through one daughter cell, while the second daughter cell spawns more highly proliferative, committed progenitor cells capable of differentiating and expanding into the clinically significant tumor mass. In this scenario, rare, slowly dividing cells with stem-like properties would always be found among the more abundant, rapidly dividing differentiated cells that make up the mass of the tumor (for a review, see Ref. 22). Current data indicate that the normal human breast contains ER−, PR− multipotent stem cells that generate lineage-restricted progenitors that are ER+, PR+ (23,24). It has been hypothesized that breast cancers originate in either the stem cells or the progenitor cells through dysregulation of the normally tightly regulated process of stem cell self-renewal (24). In 2003 Al-Hajj et al. (25) defined human breast cancer stem cells as expressing high levels of the marker CD44, a cell surface glycoprotein, and low or absent levels of CD24, a cell adhesion molecule.

We became interested in cancer stem cells as an offshoot of studies we performed trying to understand the role of progesterone in breast cancers. Although the importance of PRs as markers of hormone-dependent breast cancers is undeniable (26), the function of the hormone that binds these receptors in the diseased breast is far from clear. We have persistently found that at physiological concentrations, progestins have only transient effects on human breast tumor cell proliferation in vitro (27) and no effect on long-term tumor growth in vivo (28). For example, in human ER+ and PR+ breast tumor xenografts, only E expanded tumor growth. Progestins, either alone or in combination with E, had no effect on growth. As many have shown, E is the major mitogen in ER+, PR+ breast cancers explaining why ERs are key therapeutic targets in hormone-dependent disease.

In an effort to explain the role of progestins we analyzed solid tumors derived from ER+, PR+ human breast cancer cells that were growing in ovariectomized nude mice replenished with estradiol alone, with progesterone or MPA alone, or with estradiol plus one or the other P. Global gene expression profiling was used to search for P-regulated genes and signaling pathways (28). Among the major P-regulated genes was cytokeratin 5 (CK5), a protein associated with stem cells and supposedly restricted to ER−, PR− breast cancers. What was CK5 doing in models of ER+, PR+ human breast cancer? We found that CK5 was expressed in less than 0.1% of the E-treated tumor cells, but expanded to 5–10% of cells when either progesterone or MPA was added to E (28). To define the properties of this cell subpopulation, delicate laser capture microdissection was used to isolate only the CK5+ cells, which were then resubjected to a second round of expression profiling (29). Surprisingly, the major genes expressed in these P-regulated cells were associated with cancer stem cells. In addition to CK5, these included CD44, epidermal growth factor receptors, TGFß1, and the breast cancer-related protein, ABCG2/BCRP, an ATP-binding cassette membrane transporter originally defined by its ability to confer drug resistance, a hallmark of cancer stem cells that protects them from endogenous and exogenous toxins. Analysis of the relationship among CK5, CD44, and PR showed that all CK5+ cells were also CD44+, but many CD44+ cells were CK5−. Importantly, despite coming from ER+, PR+, P-responsive tumors, all the CK5+ cells were ER− and PR− (29).

We then asked whether CK5+ cells in tumor xenografts have the stem-like property of clonogenicity, i.e. whether they have the capacity to form tumors from a single or a few cells. At present, this is difficult to answer directly because CK5 is a cytoplasmic protein that can only be detected immunologically in permeabilized cells. Clearly, dead cells cannot be used to define clonogenicity! Instead, flow cytometry was used to separate untagged living normal mouse from ZsGreen-fluorescently tagged (30) living human breast cancer cells, and human CD44+ cells from CD44− cells, all resusupended from solid tumor xenografts. The growth potential of the two human CD44 populations was then tested in three-dimensional clonogenicity assays. Only the CD44+ cells, representing approximately 1% of the population, formed colonies. After 14 d in culture, two to three CD44+ cells expanded into colonies of 100 or more cells. Serial sections of these colonies showed that every one of the 100+ cells was CD44+ and CD24^−/low. All these cells could not possibly be stem cells! Instead, we found that within each colony, two or three of the approximately 100 cells were CK5+ and CD44+ (29). We believe that these rare cells are the true stem/progenitor cells and that CD44 is necessary, but alone is not sufficient to define stemness. Also of interest was PR expression in the colonies. Despite the fact that the colonies were derived from ER+, PR+ tumors, within each colony there was a large swath of PR− cells that included the rare CK5+ cells, adjacent to an equally large group of PR+ cells. We surmised that perhaps the progenitor cells lacked PR. Dual labeling showed that indeed, CK5+ cells were invariably PR−. Detailed analysis of numerous colonies ranging in size from approximately three cells to more than 100 cells, dual labeled for CK5 and PR, showed that young colonies of 10 cells or so tend to be mostly PR−, with two to three of the cells CK5+. But, as colonies expanded, the number of PR+ cells increased linearly (every new cell acquired PR), yet, regardless of colony size, the number of CK5+ cells remained constant at two to three cells per colony. This pattern is a hallmark of stem cells. We conclude that the colony-initiating cells are CD44+, CK5+, PR− (and also ER−) and that as the colony expands, the number of stem-like precursors remains constant whereas the rapidly differentiating progeny acquire steroid receptors (29).

Importantly, treatment of colonies for 24 h with progesterone or MPA consistently increased the number of CK5+ cells to 20–30% of the population, and in young colonies this number approached 100%. This was an exclusive property of progestins. Estrogens were not required for CK5 expansion if cells constitutively expressed PR. Similarly, solid ER+, PR+ tumors growing in response to E in nude mice had one or two cells that were CK5+ in each low-power microscopic field, but treatment of mice for 24 h with MPA dramatically increased the percent of CK5+, stem-like cells. Again, this effect of MPA was especially widespread in young tumors. As tumors enlarged and aged, the percent of P-induced CK5+ cells declined (29). We speculate that incipient, young tumors are most susceptible to stem cell reactivation by progestins. This is important with regard to the discussion of occult, noninvasive breast cancers and dormant micrometastases, below. I caution that, until we learn how to tag and expand them without killing them, it remains to be proven that CD44+, CK5+, ER−, PR− cells possess all of the properties of luminal breast cancer stem/progenitor cells.

In summary, I propose that: 1) ER+, PR+, CK5− luminal breast cancers contain a minor tumorigenic ER−, PR−, CK5+ stem cell-like subpopulation. 2) Progestins act on the bulk ER+, PR+, CK5− differentiated tumor cells to rapidly (24 h) and extensively expand the ER−, PR−, CK5+ cell population, which exhibits stem-like features. This does not require proliferation. Estrogens are also not required, unless they are needed to induce PR. 3) Small, nascent tumors or cell clusters are more sensitive to stem cell reactivation by progestins than large, mature tumors. 4) Without hormones, the reactivated stem cells and their progeny expand into the tumor bulk and reacquire steroid receptors. These more differentiated cells can then be targeted by estrogens. This hypothesis thus relegates to progestins the function of stem cell reactivation and expansion without concomitant proliferation, while leaving to estrogens their well-known mitogenic/proliferative properties. Receptor-negative CK5+ stem cells are always lurking in the tumor however, as can be shown by immunohistochemical analysis of ER+, PR+ luminal breast cancers taken from patients (29).

HYPOTHESIS: E+P IN HRT EXPANDS CANCER STEM CELLS WHEN THERE IS “JUST A LITTLE BIT OF CANCER”

I postulate that the ability of progestins to expand the number of breast cancer stem cells, especially in nascent, ER+, PR+ disease, explains the increased risk of breast cancer in women taking E+P for menopausal symptoms. I argue that these women already had breast cancer at the start of HRT but were unaware of it. What evidence is there for such undiagnosed disease? A 1997 review by Welch and Black (31) addressed the issue of healthy women and the extent of occult disease. The authors reviewed seven breast autopsy studies of women who had died of non-breast cancer-related causes: car accidents, heart attacks, etc., and who had exhibited no evidence of breast disease during life. Detailed histological examinations of their breasts found that, on average, 8.9% had undiagnosed ductal carcinoma in situ, an early noninvasive stage of breast cancer, and 1.3% had undiagnosed invasive breast cancer. Remarkably, it was estimated that 82% of these microtumors would have been mammographically undetectable (31,32). It is critically important to note that evidence for such a reservoir of occult disease was restricted to women over 40, women at an age to be candidates for HRT. The authors of the study concluded that a substantial reservoir of minimal disease is undiagnosed during life.

I postulate that this undiagnosed microdisease, if it falls into the ER+, PR+ luminal subtype, is susceptible to cancer stem cell reactivation and expansion by exposure to hormones. This hypothesis would also explain why the increased breast cancer risk observed with E+P is restricted to ER+, PR+ disease. And, if my hypothesis is correct, it is inaccurate to say that women who harbor preexisting minimal disease that is activated by hormones “develop” breast cancer while on HRT. If my ideas are correct, it follows that women must be carefully evaluated for preexisting breast disease by the most sensitive technologies available, before they are prescribed HRT.

These ideas also have direct impact on the use of HRT in breast cancer survivors, because any residual disease would be subject to the same hormonal effects. In this regard, the HABITS study published in 2008 (33) by a European consortium examined the issue of HRT safety in 447 breast cancer survivors of early-stage, mostly lymph node-negative disease (Stage 0 to Stage II), who had completed treatment and were considered to be disease free. They were randomly assigned to HRT or no HRT in 1997; an interim analysis of their status was done in 2002 and another in 2003. At that point it was observed that the hazard ratio for disease recurrence in women on HRT exceeded a statistically significant 1.8 (95% confidence interval, 1.03–3.1), and the trial was stopped. The paper (33) reports that after 5 yr, there were 39 recurrences among 221 women on HRT, compared with 17 recurrences among 221 no-hormone controls. They had taken a variety of hormones depending on local practice, and no regimen was significantly different from another.

Were all the participants of the HABITS study (33) truly disease free at the start of HRT? Another study published in 2008 (34) showed that at the time of primary breast cancer surgery, some patients already harbor undiagnosed micrometastases. The authors demonstrated presence of dissemintated tumor cells in bone marrow, for example, for which no simple screening methods exist. Similarly, tumor cells undetectable by the commonly used hematoxylin and eosin stains, were found in sentinel lymph nodes, if other markers like epithelial-specific antigen, were used for screening (35,36). Indeed the notion that equates breast cancer survivorship with a disease-free state may be highly questionable. A recent study (37) analyzed the blood of breast cancer survivors believed to be disease free for 8–22 yr, for the presence of circulating tumor cells. Survivors were compared with age-matched women with no history of breast cancer. Remarkably, circulating tumor cells were found in 30% of survivors (13 of 36) compared with one, possibly false-positive, among 26 control women. Because the half-life of circulating tumor cells is no more than 1–3 h, the authors speculate that they are shed into the blood from nonexpanding dormant micrometastases in which proliferation rates and cell death are in equilibrium. It therefore seems likely that at least one third of breast cancer survivors, believed to be disease free, retain undiagnosed, dormant tumor cells. Can these cells be awakened by exposure to hormones in HRT?

CONCLUSIONS

Based on my analysis of published data and the research of my laboratory, I have concluded that systemic progestins in HRT can reactivate and expand a subpopulation of preexisting cancer cells with stem cell-like properties. I postulate that these cells form the nidus for activation or resumption of tumor growth in women with undetected, occult disease, and in dormant cells of breast cancer survivors (38). The data, despite all their weaknesses, may explain why systemic E+P increases the risk of breast cancer. I propose that progestins do not cause cancer. Rather, they reactivate a reservoir of occult, silent, preexisting ER+, PR+ disease by expanding its stem cells in woman who are unaware that they have disease. Recall that small, nascent disease is especially susceptible to activation by progestins. There is considerable evidence for the presence of undetected, possibly undetectable, microdisease in a relatively substantial population of women, both in women who have no evidence of disease, as well as in women who have had breast cancer and are believed to be disease free. I believe that systemic HRT, especially the P component, could be dangerous to such women.