Early Discontinuation of Endocrine Therapy and Recurrence of Breast Cancer among Premenopausal Women

More than two-thirds of premenopausal women diagnosed with breast cancer have estrogen receptor (ER)-positive disease and are therefore candidates for endocrine therapy (1). Current guidelines recommend that premenopausal women diagnosed with ER-positive tumors receive adjuvant tamoxifen therapy for at least 5 years postdiagnosis to prevent or delay a recurrence (2). Tamoxifen is a selective ER modulator with metabolites that compete with estrogen at the ER-binding site (3). Tamoxifen was first introduced in the adjuvant setting in 1989 with treatment guidelines recommending 1 year of adjuvant therapy (4). As evidence from clinical trials accumulated, treatment guidelines have extended the intended duration of adjuvant tamoxifen therapy from 1 year, to 2 years, to 5 years (5). A pooled analysis of tamoxifen clinical trials among patients with breast cancer diagnosed with ER-positive disease reported risk ratios (RRs) for the effect of 1, 2, and 5 years of tamoxifen therapy versus placebo of 0.79, 0.72, and 0.50, respectively (6). Since 2013, a duration of up to 10 years of tamoxifen is recommended for premenopausal women diagnosed with nonmetastatic breast cancer (7). These changes in guidelines reflect accumulating evidence of the benefit of longer duration tamoxifen therapy.

Barriers to tamoxifen treatment adherence and completion include sociodemographic characteristics suggestive of poor access to resources, concurrent medication use, disease severity, and overall health status of the patient (8–13). Lower treatment completion has been observed in younger patients (ages less than 40 years), unmarried patients, and those with a low level of social support (14). Switching hormonal therapies can also reduce adherence (15, 16). Endocrine therapies sometimes cause adverse side effects (e.g., hot flashes, nausea, fatigue, arthralgia, vaginal bleeding, and fractures) that can affect a patient's choice to continue therapy (17, 18).

Early discontinuation and poor adherence to prescribed therapies may contribute to poor patient outcomes. In the United States, approximately 50% of patients with breast cancer never complete the intended duration of endocrine therapy (19). In Denmark, the proportion of patients completing adjuvant therapy is higher. A population-based study using the Danish Breast Cancer Group (DBCG) clinical database reported that approximately 84% of postmenopausal patients with breast cancer in Denmark completed endocrine therapy, which was defined as medication use for at least 4.5 years (20). No study has quantified the effect of early discontinuation of adjuvant endocrine therapy on breast cancer recurrence in an exclusively premenopausal cohort. In this study, we quantify the association between early discontinuation of endocrine therapy and breast cancer recurrence among premenopausal women with a stage I–III first primary breast cancer diagnosis in Denmark.

Denmark provides universal tax-funded health care with residency-based entitlement and availability of government-maintained nationwide registries, providing longitudinal sources of routinely collected administrative, health, and clinical quality data (21). The Predictors of Breast Cancer Recurrence (ProBe CaRe) cohort is a study of premenopausal women diagnosed with breast cancer in Denmark and reported to the DBCG (22). The cohort includes 4,600 women diagnosed with stage I–III ER-positive primary breast cancer between January 1, 2002 and December 31, 2011. Women were excluded if they had a previous cancer diagnosis. As breast cancer recurrences that occur within the first 1.5 years following breast cancer surgery are unlikely to be related to endocrine therapy, we also excluded 97 women in the cohort (68 of whom were diagnosed with a breast cancer recurrence) who had less than 1.5 years of follow-up. The final study population included 4,503 premenopausal women diagnosed with ER-positive breast cancer. DBCG defines menopausal status based on menostasia, oophorectomy, hysterectomy, menstruation on hormonal therapy, and age (5). Women are considered menopausal if menostasia for 12 months or more, or bilateral oophorectomy or prior hysterectomy, or menstruation on cyclic hormonal therapy and over 55 years of age. Women are classified as perimenopausal if menostasia for 2–12 months, or prior hysterectomy or mensuration on hormonal replacement therapy and 50–54 years of age. All other women are defined as premenopausal.

Members of the cohort had follow-up information available until August 31, 2019, with recurrence follow-up from 1.5 years after surgery until the first of (i) recurrence, (ii) death, (iii) 10 years of follow-up, (iv) loss to follow-up due to emigration, (v) another incident malignancy, or (vi) the end of available follow-up. Mortality and emigration were identified using the Danish Civil Registration System, which is updated daily (23). Emigration or withdrawal from clinical follow-up were the only sources of loss of follow-up, which have had an impact on less than 5% of the study population. This study was approved by the Danish Data Protection Agency (record number KEA-2017-4), the Danish Medicines Agency, the DBCG (DBCG-2014–15), and the local Institutional Review Boards of the U.S. coinvestigators. This study is based on routinely collected registry data, so Danish law does not require an ethical board's approval in Denmark, nor does it require patient consent to participate.

According to the DBCG guidelines, women diagnosed with breast cancer are recommended to undergo semiannual examinations during the first 5 years after diagnosis and annual examinations in years 6–10 (24). Women undergoing treatment for breast cancer receive endocrine therapy at the treating hospital at each follow-up visit. Treatments prescribed during these follow-up visits are registered by the DBCG, which documents whether a treatment was received and the type of therapy (tamoxifen or aromatase inhibitors). At each visit, a 6-month supply of endocrine therapy is dispensed.

Early discontinuation of endocrine therapy was defined as receipt of endocrine therapy for less than 4.5 years. If a recurrence, emigration, or death occurred before 5 years elapsed after treatment initiation, then discontinuation of therapy was deemed early if therapy was discontinued more than half a year before the event. Initiation of endocrine therapy was defined as the first date of receipt of endocrine therapy according to the DBCG registry. Discontinuation of endocrine therapy was based on the last date of receipt of endocrine therapy, with consideration of the 6-month supply of endocrine therapy prescribed. Figure 1 illustrates the cohort entry, induction period, and follow-up considerations.

Using the DBCG registry, the ProBe CaRe premenopausal cohort collected information about the first primary breast cancer diagnosis, including demographic, tumor, and treatment characteristics. This information included the date of surgery, type of surgery (mastectomy or breast conserving surgery), age at diagnosis, tumor size, tumor grade, lymph node status, HER2 expression (available after 2007), stage—derived from tumor size (T), lymph node status (N), and metastases (M)—radiotherapy, anti-HER2 therapy, and adjuvant chemotherapy. Among baseline patient characteristics, potential confounders include comorbidities (Charlson comorbidity index), socioeconomic status (median household income in tertiles), marital status (single, married), and education (high school, college, or graduate school). Medication persistence—defined as the continued per protocol use—can be influenced by time-varying covariates. As such, incident comorbid diagnoses during follow-up, initiation of concomitant medication, and switching from tamoxifen to aromatase inhibitors were treated as potential time-varying confounders.

In the descriptive analysis, we calculated the frequencies and proportions of patient demographic factors, tumor characteristics, treatments received, concomitant medications, and comorbidities within strata of tamoxifen early discontinuation status. We then evaluated the association between early discontinuation and recurrence using two complementary approaches.

First, we used pooled logistic regression models to estimate the time-dependent HRs and 95% confidence intervals (CIs) for the association between early discontinuation of endocrine therapy and breast cancer recurrence (25). The pooled logistic regression models allow for a flexible baseline hazard and produce estimates that approximate the average rate ratio, similar to a Cox proportional hazards model with a time-dependent exposure (26). The exposure is time varying to allow for treatment discontinuation at different time points over the 5 or more years of prescribed endocrine therapy. We updated the exposure annually after endocrine treatment initiation. We present multivariable-adjusted HRs and 95% CI accounting for potential baseline confounders.

Realizing that the experience of women who discontinue adjuvant endocrine therapy is not likely exchangeable with the experience of women who continue on therapy; for the second approach, we used G-computation to estimate the effect of early discontinuation of tamoxifen therapy on breast cancer recurrence adjusted for baseline tumor, treatment, and patient-level characteristics as well as time-varying covariates (27). To perform this analysis, we first calculated stabilized time-varying weights for each year of treatment. The weights for each patient were calculated from multivariable logistic regression models to estimate the probability of early discontinuation in that time interval based on baseline characteristics and time-varying factors, namely, concomitant medication use and comorbidity diagnoses during follow-up. We used stabilized weights, computed by dividing the probability of discontinuing therapy given baseline characteristics by the probability of discontinuing therapy given baseline characteristics and time-varying covariates (27). These stabilized weights were then used as time-varying weights in a pooled logistic regression model to estimate the hazard ratio (HR) for the association between early treatment discontinuation and breast cancer recurrence. We calculated the accompanying 95% CI using nonparametric bootstrapping with 50 resamples (28).

In a sensitivity analysis, we used a longer period to define early discontinuation. We allowed for 1.5 years between last date of prescription and end of follow-up, still considering a 6-month supply of endocrine therapy. This sensitivity analysis assumes that any event that occurred within 1 year of stopping endocrine therapy would not be due to early discontinuation of therapy.

Of the 4,503 premenopausal women included in the cohort, 1,001 (22%) discontinued endocrine therapy over the course of follow-up (Table 1). Among the 1,001 women who discontinued endocrine therapy, 11% discontinued in year 1, 16% discontinued in year 2, 15% discontinued in year 3, 27% discontinued in year 4, and 31% discontinued in year 5 of the course of treatment. Among those who discontinued endocrine therapy, 202 (20%) had a recurrence; among those who completed therapy, 388 (11%) had a recurrence (Fig. 2). Compared with women who completed endocrine therapy, those who discontinued were less likely to have received chemotherapy (86% vs. 92%) or radiotherapy (49% vs. 55%). Women who discontinued were also more likely to have had a comorbid condition at diagnosis (11% vs. 8.4%), have had a lower household income (42% vs. 31%), have been single during follow-up (36% vs. 32%), and have had a high school level of education (22% vs. 19%).

In multivariable-adjusted models, early discontinuation of endocrine therapy was associated with an increase in the estimated rate of recurrence (HR, 1.63; 95% CI, 1.30–2.04; Table 2). Similarly, results from the G-computation approach (using time-varying weights) indicated that women who discontinued endocrine therapy had 1.67 times the estimated rate of recurrence compared with women who continued endocrine therapy (95% CI, 1.25–2.14). The crude results were comparable with the above approaches, suggesting little confounding by measured baseline or time-varying confounders (HR, 1.59; 95% CI, 1.27–2.00).

The sensitivity analyses, which defined completion of adjuvant endocrine therapy as 4 years of therapy or continuation of therapy for up to 1 year before a recurrence, yielded similar, although attenuated estimates (Table 2).

This study is the first to quantify the effect of early discontinuation of adjuvant endocrine therapy on breast cancer recurrence in a population-based cohort of premenopausal women diagnosed with breast cancer. We report that 22% of premenopausal women discontinued endocrine therapy before completing 5 years of recommended treatment. The estimated rate of breast cancer recurrence in this study was about 1.7-fold higher among women who discontinued endocrine therapy compared with those who completed their course of treatment. Among women who did not complete adjuvant endocrine therapy, most (57%) discontinued taking medication in the last 2 years of treatment.

Trial results have reported that 5 years of tamoxifen therapy, compared with placebo, reduced the risk of recurrence by nearly one-half among premenopausal women diagnosed with ER-positive breast cancer (29). This result suggests that lack of tamoxifen therapy among premenopausal women diagnosed with ER-positive breast cancer would nearly double the risk of recurrence compared with tamoxifen therapy of appropriate duration. A recent systematic review reported that 15%–20% of women discontinued tamoxifen therapy in the first year of therapy and 31%–60% discontinued by the end of 5 years (30–34). Although the review primarily included studies of U.S. populations—in which a greater proportion of women than in Denmark discontinue endocrine therapy—the pattern of discontinuation over the first 5 years of therapy is comparable with the pattern observed in our study. A meta-analysis of trials comparing the benefit of 5 years versus 1–2 years of tamoxifen therapy reported 0.82 times the rate of recurrence (29). Therefore, the expected increase in the rate of recurrence comparing 2 versus 5 years of tamoxifen therapy would be approximately 1.22 (1/0.82; ref. 29). The estimated effect from our study is further from the null than this expectation, likely due to a combination of factors. First, there is a difference in target populations, as the trials were conducted among patients with both premenopausal and postmenopausal breast cancer, whereas our study is restricted to premenopausal women. Second, most women in our study who completed 5 years of tamoxifen therapy continued on therapy, given the change in guidelines from 5 to 10 years of therapy in 2013 (35). Third, our study estimated the rate ratio, which will be further from the null than the RR reported from trials. Finally, we excluded recurrences in the first 1.5 years after diagnosis, which were included in the trial results. This exclusion reduces the number of events, and ratios of rarer events are often larger than ratios of more common events (36). Even though these women would have received some benefit from initiating therapy, evidence clearly supports the need for completion of the 5 years of tamoxifen therapy—and even extending to 10 years—to acquire the full preventive benefit.

Before the end of 5 years of adjuvant endocrine therapy, 22% of women in our cohort discontinued treatment. This estimate is substantially lower than those from previous studies conducted in the United States (19); however, the proportion is slightly larger than the corresponding 16% of postmenopausal women who discontinued treatment reported in an earlier Danish study (20). This difference may be expected because some previous studies have shown that younger women are less likely to complete adjuvant endocrine therapy than older women (10). Similar to U.S.-based studies, we observed that women who discontinued therapy were more likely to report a lower household income, be unmarried, and have high school rather than third-level education, compared with women who completed therapy (37). Income, education, and health care access gaps are more pronounced in the United States than in Denmark, which is also correlated with health care access, and may partially explain the differences observed between our study and the review (38, 39). The BIG 1-98 study—a double-blind trial that randomized postmenopausal women to tamoxifen or aromatase inhibitors—reported tamoxifen discontinuation rates that were consistent with those observed in this study (19%; ref. 40). In BIG 1-98, among postmenopausal women who were randomized to the tamoxifen arm, adverse events such as thromboembolic complications were the most common causes for treatment discontinuation (40). In the clinical setting, women who are at risk for or who experience thromboembolic complications are eligible to switch to aromatase inhibitors, which carry about half the risk of thrombotic events compared with tamoxifen (41, 42). Still, adverse side effects are often reported as a reason that patients stop taking tamoxifen (12, 43).

Our results are comparable with those observed in other studies that were not restricted to premenopausal women. A study using a Kaiser Permanente patient database reported that early discontinuation was associated with higher all-cause mortality (HR, 1.26; 95% CI, 1.09–1.46; ref. 44). In a Danish population-based cohort of postmenopausal women diagnosed with breast cancer between 1996 and 2004, those who discontinued therapy had an increased hazard of recurrence (HR = 1.45; 95% CI, 1.14–1.85; ref. 20). The associations observed in our study are more pronounced, underscoring the importance of endocrine therapy completion among premenopausal women to prevent or delay a recurrence. Most premenopausal patients with breast cancer are ineligible for screening. Danish regional and national mammography screening programs have invited only women older than 50 years to be screened, so most women in our cohort were diagnosed before they would have been invited to participate (45). Therefore, optimizing treatment regimens for these patients is essential to improve outcomes.

There are several limitations to consider. First, as with all population-based studies of medication use, we had information on prescriptions dispensed at each follow-up visit but could not be certain that endocrine therapy was taken as prescribed. It is likely that women who discontinued therapy before the recommended 5 years of treatment were also less likely to take the medication as indicated, even when they filled a prescription. Second, since 2013, the Danish guidelines recommend that tamoxifen be prescribed to premenopausal women for up to 10 years (7, 35). We observed that, among women who completed at least 5 years of adjuvant endocrine therapy, 11% (n = 405) completed 10 years of treatment, and 71% (n = 2,565) continued taking medication for more than 5 years. The proportion of women receiving more than 5 years of adjuvant endocrine therapy increased as a function of year of diagnosis between 2002 and 2011. Given that our study population included women diagnosed with breast cancer through 2011, and with follow-up available until August 31, 2019, we were unable to examine the effect of discontinuation of tamoxifen therapy at 5 years compared with continuation for up to 10 years. Third, we were unable to capture menopausal transition. We did observe that 40% of those who completed therapy and 30% of those who did not complete therapy switched from tamoxifen to an aromatase inhibitor. This suggests that a large proportion of women transitioned to menopause, but women may also switch type of therapy due to side effects. Fourth, we did not have information on reasons for discontinuation, which would provide additional insight into clinical interventions to ensure treatment completion. Finally, women who discontinued therapy were more likely to have had a recurrence within the first 5 years following their diagnosis compared with those who completed therapy (70% vs. 52%). This may reflect an innate resistance to adjuvant endocrine therapy (46–48). We excluded women with less than 1.5 years of follow-up or a recurrent event within the first 1.5 years following diagnosis to help mitigate this potential bias. Still, the hazard of recurrence is highest in the first 2–3 years after diagnosis, and the prevalence of discontinuation increases over time (49). The combination of eliminating the first 1.5 years of follow-up and a discontinuation proportion that increases over time suggests that the association observed in this study pertains to recurrences that occur after the initial high hazard period. If we assume that recurrences in the high hazard period reflect an innate resistance of the cancer rather than persistence of endocrine therapy, then our results are unbiased. However, future analyses might consider analyzing the results stratified on the basis of year of discontinuation to look specifically at the influence of endocrine therapy on this early period of high recurrence hazard.

In conclusion, our results highlight the importance of clinical follow-up and behavioral interventions that support persistence of adjuvant endocrine therapy to prevent a recurrent event.

L.J. Collin reports grants from NCI during the conduct of the study. L.A. Waller reports grants from NCI during the conduct of the study; grants from NIH/NIA, NIH/NIAID, NIH/NHLBI, NIH/NIEHS, NIH/NICHD, and CDC outside the submitted work. B. Ejlertsen reports grants from NanoString Technologies, AstraZeneca, Novartis, Oncology-Venture, Roche, and Samsung outside the submitted work. M.-B. Jensen reports grants from Samsung Bioepis, Nanostring Technologies, and Oncology Venture outside the submitted work. H.T. Sørensen reports The Department of Clinical Epidemiology is involved in studies with funding from various companies as research grants to and administered by Aarhus University. None of these studies are related to the current study. T.L. Lash reports grants from multiple public sector and foundation sources of financial support during the conduct of the study; personal fees from Amgen Inc. outside the submitted work. No disclosures were reported by the other authors.

L.J. Collin: Conceptualization, formal analysis, funding acquisition, validation, investigation, visualization, methodology, writing-original draft, writing-review and editing. D.P. Cronin-Fenton: Conceptualization, resources, supervision, funding acquisition, methodology, writing-review and editing. T.P. Ahern: Formal analysis, supervision, methodology, writing-original draft, writing-review and editing. M. Goodman: Supervision, methodology, writing-original draft, writing-review and editing. L.E. McCullough: Conceptualization, formal analysis, supervision, methodology, writing-original draft, writing-review and editing. L.A. Waller: Conceptualization, formal analysis, supervision, methodology, writing-original draft, writing-review and editing. A. Kjærsgaard: Resources, data curation, software, visualization, writing-review and editing. P. Damkier: Investigation, methodology, writing-review and editing. P.M. Christiansen: Resources, supervision, writing-review and editing. B. Ejlertsen: Conceptualization, resources, supervision, funding acquisition, methodology, project administration, writing-review and editing. M.-B. Jensen: Resources, data curation, software, formal analysis, methodology, writing-review and editing. H.T. Sørensen: conceptualization, resources, supervision, funding acquisition, methodology, writing-review and editing. T.L. Lash: Conceptualization, resources, formal analysis, supervision, funding acquisition, validation, methodology, writing-original draft, writing-review and editing.

This work was supported primarily by the U.S. NCI awarded to T.L. Lash (R01CA166825) and the Lundbeck Foundation awarded to D.P. Cronin-Fenton (R167-2013-15861). This work was also supported by the Novo Nordisk Foundation awarded to D.P. Cronin-Fenton (NNF19OC0058710). L.J. Collin was supported by the U.S. NCI (F31CA239566). T.P. Ahern was supported in part by funding from the National Institute of General Medical Sciences (P20GM103644). The content of this article is solely the responsibility of the authors and does not necessarily represent the official views of the U.S. NIH or other sources of funding support.

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.