Screening Mammography Use among Current, Former, and Never Hormone Therapy Users May Not Explain Recent Declines in Breast Cancer Incidence

See Editorial on p. 705

This article is featured in Highlights of This Issue, p. 699

Several previous studies have shown a rapid decline in breast cancer incidence following the Women's Health Initiative (WHI) trial publication in 2002 that showed higher breast cancer rates in estrogen + progestin (EPT) users compared with placebo, but found no association among estrogen-only (ET) users (1, 2). After study results were widely publicized, a dramatic decline occurred in both EPT and ET use among women in the United States (3). Declines in annual U.S. breast cancer rates have been reported ranging from 5%–8.5% (4, 5) since the early 2000s; the most recent report suggests a 1.8% annual decrease between 2003 and 2007 (6, 7). These declines have not been consistently shown in other countries, where varying proportions of HT formulations and use occurred before and after WHI, and breast cancer screening programs differed. Germany (8) Switzerland (9), Spain (10) and Canada (11) have all noted declines in breast cancer incidence, whereas New Zealand (12), Norway (13), the Netherlands (9), Scotland (14), and Italy (15) have reported no change.

We have previously described a number of hypotheses that may explain the declining incidence (16), but the leading hypothesis continues to point to the immediate cessation of hormone therapy (HT). Women may also be less inclined to undergo routine screening mammography if they stop taking HT because they may be less concerned about their breast cancer risk or less likely to regularly see their physician. Notably, screening mammography use has declined in the United States by 6.8% between 2000 and 2005 among women aged 50 to 64 years, and the decline is associated with decrease HT use (17). We would expect an immediate decline in early-stage invasive breast cancer with screening mammography declines, because many nonpalpable cancers would go undetected. We recently completed an analysis using data from the Breast Cancer Surveillance Consortium (BCSC) to address the hypothesis that women who used HT were less likely to return for screening mammograms after cessation (4). In that study, we had no information on those women who stopped HT and mammography screening because most BCSC registries only capture HT use on women undergoing mammography.

This study was designed to address gaps in knowledge about the relationship between mammography, HT use, and breast cancer incidence by linking individual women-level data on HT use (current, former, and never by EPT and ET) to both mammography screening and breast cancer rates over time.

We identified a dynamic cohort of women aged 50 to 79 years, with no prior history of breast cancer, who were enrolled in Group Health Cooperative's integrated group practice in western Washington between 1994 and 2009 and linked their breast cancer risk factor data (18) to automated pharmacy dispensing data, enrollment files, and to the western Washington Surveillance Epidemiology and End Results (SEER) registry. Breast cancer risk factors were obtained through a survey taken at the time of the mammogram, and most women were part of the organized screening program that provided regular mailed reminders directly to women who were due for mammograms (19). Institutional Review Board approval was obtained for all study activities.

To ensure sufficient information on prior and current HT use among the cohort, we required 4 years or more of uninterrupted enrollment (“run-in period”). Women started contributing information to the analysis after 4 years of enrollment. We censored women at the first occurrence of any of the following dates: breast cancer diagnosis, disenrollment from Group Health (>60 day lapse in membership), (3, 20) death, 80th birthday, or the end of the study period (September 30, 2009). A total of 741,681 woman-years of analytic follow-up time (median of 3.3 years per woman) were obtained on 163,490 unique women.

HT use was ascertained from our automated pharmacy data system and claims files and captured HT use from Group Health–owned pharmacies and external facilities if a claim was submitted (including Medicare Part D beneficiaries). Nearly all (≥95%) GHC members report filling all or almost all their prescriptions at GHC pharmacies (3, 21). Each dispensing was categorized as EPT or ET using previously developed methods (3). The intended duration of each drug dispensing was based on the number of estrogen pills or patches dispensed. We developed run-out dates for each successive dispensing and bridged gaps in dispensings that were <60 days to allow for some noncompliance with the dispensings (3). Each month, a woman was classified as a current, former, or never user of EPT or ET. We defined the monthly indicators as follows: never users had no HT fills within the past 48 months, current users had an HT dispensing within the past 2 months, and former users had used HT in the last 3 to 48 months, but not in the last 2 months.

We included all mammograms that met the Breast Cancer Surveillance Consortium screening definition (22). Mammography examinations conducted outside Group Health were captured using claims data. Each month, a woman was defined as being screening compliant if she received a screening mammogram in the prior 26 months. We chose 26 months (2 years and 2 months for timing flexibility) because Group Health recommends risk-based screening every 1 to 2 years, depending on a woman's risk; in addition, 24 months is consistent with definition used for screening compliance by Healthcare Effectiveness Data and Information Set (23). Western Washington SEER data were used to identify all incident cancers diagnosed including ductal carcinoma in situ (DCIS) and invasive cancers.

We summarized characteristics of the study cohort including total women and person-months of follow-up, the age and mammography distribution over time, and the temporal distribution of HT exposure. We also counted the number of women and characterized their age distribution within each of the HT exposure groups per year.

Women had monthly indicators for whether they were screening compliant (screened within the previous 26 months). Using these monthly data, we estimated the percentage of screening compliant women for each year and HT exposure group, adjusted to the age distribution of the entire study population in the year 2000. Annual compliance rates were calculated based on an age-adjusted Poisson regression model estimated using generalized estimating equations to account for within-woman correlation and using the method of marginal standardization (24, 25).

Age and time since prior screening mammography adjusted invasive and DCIS cancer rates (per 1,000 person-years), corresponding 95% CIs, and tests for cancer trends were also computed using marginal standardization and Poisson regression models. These adjusted rates were estimated for each year and HT exposure group and were standardized to the age and mammography distribution of the study population in the year 2000. Two-sided Wald tests were conducted (based on the regression models) to test contrasts for linear trends in the log risk of cancer across the study period. All analyses were conducted with SAS software, version 9.2 (SAS Institute, Inc.).

We studied 163,490 unique women (8,900,171 person-months). Our study population averaged 70,907 women per year with an annual average of 741,681 person-months and 25,721 screening mammograms. The number of women in our study population rose each year from 1998 (59,284) to 2009 (79,667). There were moderate fluctuations in the overall distribution of women by age and by time since last mammogram, and we saw important changes in HT exposure (Fig. 1). We noted a large decrease in current HT users over time: approximately 41% of the study population in 1998 to approximately 11% in 2009. The distribution of time since last screening mammography suggests the proportion of women with annual screening, and 2-year screening intervals remained relatively unchanged between 1998 and 2009 (1998: 41% annual, 23% biennial; 2009: 40% annual, 22% biennial), but there was a decrease in the proportion of women with >48 month screening intervals (35% in 1998 to 28% in 2009).

In 1998, 46% of women were aged 50 to 59 years, 28% were aged 60 to 69 years, and 26% were aged 70 to 79 years. There was a slow shift in the study population toward a greater proportion of younger women; by 2009, 52% were 50 to 59 years and only 16% were aged 70 to 79 years (Fig. 1). The age distribution also differed by HT group overall and across time (Fig. 2). The absolute number in each HT group can be seen over time with the largest increase in never users from 30,009 in 1998 to 60,688 in 2009. The majority of former and continued EPT users were aged 50 to 59 years; the proportion of this age group increased in never users over the study period.

Before 2002, screening compliance differed significantly by HT exposure with current EPT users having the highest rates (83%) followed by former EPT (77%) and current ET (77%), former ET (72%) and never users hovering at the lowest around 56% (Fig. 3). With the exception of current EPT users, screening increased among the different HT exposure groups in the first few years following release of the WHI findings. By 2006, screening compliance was similar among current and former users of both EPT and ET (∼81%) and showed significant increases among never users (∼62%). The proportion of women in compliance in each group appeared to decrease after 2007 in all groups. Furthermore, while screening increased among most groups between 2002 and 2006, the overall rate of screening compliance remained relatively constant and ultimately declined in the full population over the 12-year study period.

We observed 3,426 cancers over the study period (610 DCIS, 2,816 invasive), with an annual average of 51 DCIS and 235 invasive. The majority of invasive cancers were ER+ (2,180; annual average 182). After adjusting for time since prior screening mammogram and age, invasive breast cancer rates significantly decreased overall (all HT groups combined) across the study period (P_trend < 0.01). When stratifying by HT exposure status, declines were significant (P_trend = 0.01) or of borderline significance (P_trend = 0.05, 0.06) for 4 of the exposure groups. Current EPT users, however, experienced no significant change (P_trend = 0.68; Fig. 4A). DCIS rates did not change over time across any HT group (Fig. 4B).

This is the first study conducted to date that reports on patterns of mammography use for HT users by linking HT pharmacy dispensing and cancer rates within the same cohort of women over time. Consistent with other U.S. reports, we observed clinically significant declines in HT use between 2002 and 2006 with the largest among women aged 50 to 69 years. The change in HT use translated into important changes in mammography screening patterns. Before 2002, former HT users had lower screening rates than current users; this is no longer true—former users have the same or even slightly higher screening rates than current users. During the same period of HT and mammography screening changes, we observed no changes in DCIS rates (P = 0.66), but did observe a significant −3.5% average annual percentage reduction in invasive cancers.

Screening mammography rates have decreased slightly over the past decade in our population, as has been seen in others (17). Prior literature has suggested former HT users have lower mammography screening compared with current users. This held true in our population only during the years before the EP WHI findings were released, after which former users in our study had the same or slightly higher rates than current users; similar findings to another cohort (26). Higher screening rates among women who discontinued use may reflect a healthy user screening predisposition or increased concern over breast cancer risk. Even though we show important increases in screening among non-HT users over time, the group's screening compliance rate remains significantly lower than other HT exposure groups, which may help explain why mammography rates are decreasing in the United States. Because of the timing of this study, we do not think rates of screening were affected by the 2009 change in mammography screening recommendations of the U.S. Preventive Services Task Force, released in October, 2009 (27); furthermore, women in this population were mailed reminders when they were due or overdue for a screening mammogram.

We compared our observed trends in cancer rates with the underlying western Washington SEER registry rates over the same time period to provide a regional comparison for rates (28, 29). Age-adjusted DCIS rates (per 100,000) did not vary meaningfully with rates ranging between 33.4 (2000) and 37.7 (2004), similar to our population (28, 29). Invasive cancer rates (per 100,000) fluctuated a bit more with rates over 150 between 1998 and 2002, which decreased into the 130′s between 2003 and 2008 (28, 29).

An important limitation of these results is the precision of our cancer incidence estimates after stratification by HT status and year due to small sample sizes; particularly for DCIS rates. Although the uncertainty around these incidence rates is high because of the small sample sizes in the HT subgroups, breast cancer incidence declines were observed among all HT groups except current EPT users and were limited to invasive cancers. These findings suggest cancer declines are not due to less screening, but are more likely due to EPT cessation reducing the promotion of established tumors. Unlike longer term WHI EPT findings, we did not observe higher breast cancer rates among former EPT users (30). Consistent with our findings were the recently reported long-term follow-up from the ET WHI trial with persistent lowered invasive breast cancer rates among former ET users (31). However, neither WHI trial included in situ cancers; these account for 20% of all breast cancer diagnoses in the United States (32).

Our study also has several strengths. We were able to collect accurate, automated HT data through Group Health Cooperative's administrative pharmacy database, which has been shown to be 97% complete (3, 21). These data enabled us to directly link exposure, screening, and outcomes within an integrated delivery system in which disenrollment is uncommon (20), and prescription drug coverage is not a barrier for most women. Managed care organizations such as Group Health are also strongly represented in the United States among insured individuals; as of 2008, 25% of the U.S. population was enrolled in an health maintenance organization (33). The Group Health racial and ethnic composition accurately mirrors western Washington state. Compared with the rest of the country, Group Health members are more likely to be Asian or Pacific Islanders (9% versus 5%) but less likely to be Black (4% versus 13%) or report Hispanic ethnicity (4% versus 16%). (34)

Continued surveillance in the United States and other countries should focus on age-specific incidence, histologic subtypes, tumor characteristics (e.g., stage, size, and severity), and outcomes following diagnosis (e.g., mortality) and take into account screening methods and frequency. Doing this will help further explain the decreased breast cancer incidence that has been observed in the United States and other countries. Importantly, when studies are examining the influence of a changing exposure, HT in this example, it is important to examine and account for varying exposures (HT type and dose), exposure prevalence, and characteristics of individuals exposed.

Screening compliance within groups defined by HT use increased in the years immediately following WHI. However, because never-users of HT comprised a significantly greater proportion of women in the population and because this group has consistently exhibited much lower screening utilization than HT users, the population, as a whole, did not show increased mammography compliance. Differential screening mammography rates by HT use (current, former, and never use) do not explain recent declines in invasive breast cancer incidence. Invasive breast cancer rates declined among all HT user groups, except EPT continued users. These data suggest that discontinuing HT—including EPT—has an immediate effect on breast cancer rates, lending additional support (4, 5) to the mechanism that cessation leads to tumor regression.

The authors had full responsibility in the design of the study, the collection of the data, the analysis and interpretation of the data, the decision to submit the manuscript for publication, and the writing of the manuscript.

The data collection was supported by a grant from the National Cancer Institute (CA63731) as part of the Breast Cancer Surveillance Consortium. Investigators time was covered by grant funding from the National Cancer Institute (U01CA63731, U01CA63740, U01CA86076, and U01CA86082). The collection of cancer incidence data used in this study was supported by the Cancer Surveillance System of the Fred Hutchinson Cancer Research Center, which is funded by contract no. N01-CN-67009 and N01-PC-35142 from the SEER Program of the National Cancer Institute with additional support from the Fred Hutchinson Cancer Research Center and the State of Washington. A list of the BCSC investigators and procedures for requesting BCSC data for research purposes are available at: http://breastscreening.cancer.gov/.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.