- © 2001 by American Society of Clinical Oncology
Adjuvant Tamoxifen: Predictors of Use, Side Effects, and Discontinuation in Older Women
- From the Boston University School of Public Health, Boston University School of Medicine, and Boston Medical Center, Boston, MA.
- Address reprint requests to Rebecca A. Silliman, MD, PhD, Boston Medical Center, 88 E Newton St, F4, Boston, MA 02118; email rsillima{at}bu.edu
Abstract
PURPOSE: To identify predictors of adjuvant tamoxifen use, side effects, and discontinuation in older women.
PATIENTS AND METHODS: We followed a cohort of 303 women ≥ 55 years of age diagnosed with stage I or stage II breast cancer for nearly 3 years. Data were collected from women’s surgical records and from computer-assisted telephone interviews at 5, 21, and 33 months after primary tumor therapy.
RESULTS: Two hundred ninety-two (96%) of 303 patients in the study provided information about tamoxifen use. Tamoxifen use was reported by 189 patients (65%); 26 (15%) discontinued use during the follow-up period. Patients who were 65 to 74 years of age (relative to those 55 to 64 years of age), had stage II disease, were estrogen receptor–positive, saw a greater number of breast cancer physicians, and had better perceptions of their abilities to discuss treatment options with physicians had greater odds of tamoxifen use. Those who had better physical function, had received standard primary tumor therapy, and had obtained helpful breast cancer information from books or magazines had lesser odds of tamoxifen use. Patients ≥ 75 years of age (relative to those 55 to 64 years of age) and patients with better emotional health had significantly lesser odds of reporting side effects. Patients who were estrogen receptor–positive were less likely to stop taking tamoxifen; patients who experienced side effects were more likely to stop taking tamoxifen.
CONCLUSION: Deviations from a prescribed course of adjuvant tamoxifen occur relatively frequently. The clinical consequences of this deviation need to be identified.
AN ESTIMATED 176,300 women were diagnosed with breast cancer in 1999,1 more than one half of whom were 65 years of age or older.2 A substantial literature has documented that older women are less likely to receive standard care for a new diagnosis of breast cancer.3-14 Although less-than-standard therapy has been linked to higher rates of breast cancer recurrence and mortality,15-18 it is not known whether this is due to deficiencies in primary tumor therapy, deficiencies in the prescribing of and adherence to systemic adjuvant therapy, or both.
Over the past decade, the threshold for recommending systemic adjuvant therapy has progressively lowered, and the focus has shifted from reducing mortality rates to improving recurrence-free survival rates.19 For example, at the 1990 National Institutes of Health Consensus Development Conference on the treatment of early-stage breast cancer, it was noted that although “the majority of patients with node-negative breast cancer are cured by breast-conservation treatment or total mastectomy and axillary dissection,” combination chemotherapy or at least 2 years of tamoxifen was recommended.20 By early 1998, systemic adjuvant therapy was recommended for all women, except for node-negative women at low risk of recurrence by virtue of having tumors 1 cm or less in diameter or having grade 1, estrogen receptor–positive tumors with no lymphatic invasion.21,22 Tamoxifen was recommended also for node-negative women older than 70 years of age at high risk of recurrence, regardless of estrogen receptor status.22 The 1998 St. Gallen 6th International Consensus Panel on the Treatment of Primary Breast Cancer further refined these guidelines.23 With the exception of low- risk, node-negative patients (< a 10% risk of relapse at 10 years) and those who are estrogen receptor–negative, the panel recommended a full 5 years of tamoxifen therapy for all elderly women with breast cancer.23 These recommendations were undoubtedly influenced by the overview update of randomized trials of adjuvant tamoxifen therapy that concluded that 5 years of tamoxifen therapy substantially reduces the risk of recurrence, mortality, and contralateral disease among women whose tumors are estrogen receptor–positive and that this benefit is independent of age, nodal status, and receipt of chemotherapy.24
Although 5 years of adjuvant tamoxifen therapy is now recommended,22,23 information about tamoxifen adherence and discontinuance rates is sparse. We know of no published data regarding tamoxifen discontinuance rates in clinical practice. In the context of clinical trials, discontinuance rates have ranged from 23% to 40%. In the tamoxifen chemoprevention trials, for example, the Royal Marsden Hospital trial reported a 40% premature discontinuance rate in the treatment group during a median follow-up of 70 months (5.8 years), compared with 31% of women in the placebo group.25 The National Surgical Adjuvant Breast and Bowel Project Breast Cancer Prevention Trial reported that 24% of women in the tamoxifen group discontinued therapy, compared with 20% of women in the placebo group.26 In the adjuvant tamoxifen setting, Fisher et al27 reported that 23% of patients participating in the B-14 trial discontinued tamoxifen therapy before the occurrence of an event during the first 5 years after randomization, compared with 24% of women in the placebo group.
To better understand patterns of adjuvant tamoxifen use and discontinuation, we followed a cohort of 303 women 55 years of age or older who were diagnosed with stage I or stage II breast cancer for nearly 3 years after primary tumor therapy. Specifically, we sought to identify predictors of adjuvant tamoxifen use, side effects, and discontinuation.
PATIENTS AND METHODS
Study Sample
The study’s methods have been described.14 To be eligible for study participation, women had to be 55 years of age or older, newly diagnosed with stage I or stage II breast cancer, and have no history of a prior breast cancer. Consecutive eligible women from five participating hospitals were sent an introductory letter and a consent form after their most definitive surgical treatment. Our interviewer followed-up by telephone and provided additional information about the study, answered questions, and obtained informed consent. Of the 388 eligible patients, 303 (80%) agreed to participate and were enrolled between January 1993 and April 1996.
Data Collection
Data were collected from women’s surgical records and from computer-assisted telephone interviews at 5, 21, and 33 months after primary tumor therapy. Data collected from medical records included histology, stage, estrogen receptor status, and surgeries performed (modified radical mastectomy or breast-conserving surgery). Medical records were monitored for 6 months after surgery to determine whether radiation therapy and chemotherapy were initiated and completed and whether adjuvant tamoxifen therapy was initiated. The baseline telephone interview included questions about sociodemographic characteristics (including age, education, and marital status); general health-related quality of life (as measured by the Medical Outcomes Study Short Form [SF-36])28; breast cancer-specific quality of life; the presence of physician-diagnosed cardiopulmonary diseases and the frequency of associated symptoms; the number of physicians with whom breast cancer treatment options were discussed and the specific treatments chosen; the perceived helpfulness of various sources of information about breast cancer and its treatment; and perceptions of doctor-patient communication. Follow-up interviews included detailed questions about adjuvant tamoxifen use, side effects, and discontinuance.
Major Analytic Variables
Outcome variables.
Tamoxifen use was defined as taking tamoxifen at any time during the study period (from baseline through the second follow-up interview). At each interview patients were asked, “Are you taking tamoxifen at the present time?” At the second follow-up interview, if the answer to the current tamoxifen use question was no, they were also asked, “Did you ever take tamoxifen?” Responses across the three interviews were then summarized as a dichotomous variable with a yes or no response to ever having taken tamoxifen.
Side effects.
Information on tamoxifen side effects was collected from patients who reported taking it. Patients were asked if they were experiencing hot flashes, vaginitis, phlebitis, depression, nausea, edema, or any other side effects. Two dichotomous side-effect variables with yes/no responses were considered for analysis: hot flashes alone and any side effects. The definition of the latter variable included hot flashes as well as reports of any other side effects. We chose to consider both definitions because we expected the reporting of hot flashes to be highly age-dependent but the reporting of any side effects to be less so.
Discontinuation of tamoxifen therapy.
Whether or not women were continuing to take tamoxifen was evaluated at the second follow-up interview. For patients with missing tamoxifen information, we used the last observation carried forward approach to fill in the missing values.29 We categorized women as either still taking tamoxifen (yes) or no longer taking tamoxifen (no). Patients who had experienced a breast cancer recurrence by the second follow-up interview were excluded from this definition.
Explanatory Variables
We considered variables from five categories. First, we considered sociodemographic characteristics, including age (55 to 64, 65 to 74, and 75+ years), marital status (currently married/not married), and education (< high school/≥ high school). Second, we considered two measures of health status: comorbidity and physical function. Our measure of comorbidity was based on patients’ reports of diagnoses of chronic obstructive pulmonary disease, congestive heart failure, and ischemic heart disease, as well as related disease manifestations.30 We evaluated physical function using the 10-item physical function scale from the SF-36.28 Third, we considered two measures of emotional health: (1) the five-item measure of general emotional health from the SF-36,28 and (2) a four-item measure of breast cancer–specific emotional health that assesses feelings and worries pertaining to potential problems related to the progression of breast cancer.31 Fourth, we considered breast cancer–related variables: breast cancer stage (I/II), estrogen receptor status (positive/negative), and whether patients received standard primary tumor therapy, defined as modified radical mastectomy or breast-conserving surgery and axillary dissection followed by radiation therapy (yes/no). Fifth, we considered aspects of the treatment decision-making process: (1) the number of breast cancer specialists with whom the patient discussed treatment options; (2) patients’ perceptions of their abilities to communicate with their physicians; (3) patients’ perceptions of their physicians’ abilities to give information, discuss treatment options, and tailor treatments; (4) patients’ ratings of their physicians’ technical and interpersonal care; and (5) sources of helpful information about breast cancer and its treatment.31 For the latter, we asked patients how helpful a series of sources had been, including friends and family, breast cancer doctors and their office staff, written materials such as books or magazines, and television specials or news spots. Responses ranged from “very helpful” to “not helpful at all.” For each item we created a dichotomous variable (“very helpful” v all other responses).
Data Analysis
To explore the crude associations between categorical and continuous variables we used two-sample t tests (or analysis of variance procedure), and for associations between categorical variables, we used χ2 tests of proportions (or Fisher’s exact test, when needed). The association between each of the three outcome variables and the explanatory variables was evaluated using multiple logistic regression analysis. In this multivariable analysis, all of the explanatory variables were eligible to enter the final model. We used a stepwise selection procedure to develop parsimonious models. Because of variations in tamoxifen use by estrogen receptor status, we also performed separate analyses for estrogen receptor–positive patients. There was insufficient information to perform subgroup analyses for patients who were estrogen receptor–negative. Results with P values less than .05 were deemed to be statistically significant in this report.
RESULTS
Sample Characteristics
The average age of patients was 67.7 (SD = 8.7) years. Approximately one half were married and 83% had completed at least a high school education. Fifty-nine percent had no cardiopulmonary comorbidity, 63% had stage I breast cancer, the majority (76%) were estrogen receptor–positive, and 78% had received definitive primary tumor therapy. Two hundred ninety-two (96%) of the 303 patients in the study provided information regarding tamoxifen use. Tamoxifen had been taken by 189 patients (65%), but only 166 (88%) provided information about their experience of side effects (23 patients [12%] did not provide any information about side effects). One hundred four (63%) reported at least one side effect while they were taking tamoxifen: hot flashes (45%), vaginitis (16%), fluid retention (13%), depression (15%), nausea (7%), fatigue (5%), thrombophlebitis (2%), vision problems (2%), vaginal bleeding (2%), and other side effects (17%). The remaining 62 patients (37%) reported that they had not experienced any side effects.
Twenty-six patients (15%) had stopped taking tamoxifen by the time of the second follow-up interview for reasons other than recurrence of breast cancer. Of these, 18 (69%) were estrogen receptor–positive and eight (31%) were estrogen receptor–negative. Thirteen patients who had taken tamoxifen and developed a breast cancer recurrence were excluded from the adherence analysis because we were uncertain whether they had stopped taking tamoxifen before or after their recurrences were clinically apparent.
Summary information regarding our explanatory variables by patient age is displayed in Table 1. Younger patients were more likely to be married and were more highly educated. As expected, they were less likely to have comorbidity and their physical function scores were higher. Emotional health status did not vary by age. Younger patients were more likely to have received definitive primary tumor therapy. They were more likely to report that they obtained helpful breast cancer information from books or magazines and television spots, they saw a greater number of breast cancer physicians, they rated their and their physicians’ abilities to communicate more highly, and they rated their physicians’ technical and interpersonal skills more highly.
Study Variables by Age Group
Tamoxifen Use
Table 2 lists the results of our multiple logistic regression analysis with tamoxifen use as the outcome. Patients who were 65 to 74 years of age (compared with those who were 55 to 64 years of age), had stage II disease, were estrogen receptor–positive, saw a greater number of breast cancer physicians, and had better perceptions of their abilities to discuss treatment options with physicians had greater odds of tamoxifen use. Those who had better physical function, had received standard primary tumor therapy, and had obtained helpful breast cancer information from books or magazines had lesser odds of tamoxifen use.
Multiple Logistic Regression Analysis With Tamoxifen Use as the Outcome
Findings from a logistic regression analysis based on the subset of patients who were estrogen receptor–positive were similar to the findings from the full data set for all explanatory variables except for age. Among estrogen receptor–positive patients, there were no statistically significant differences in the odds of tamoxifen use among the three age groups, although the effect sizes were similar to those seen for the full data set.
Side Effects
The multiple logistic regression analysis with any side effects as the outcome (Table 3) indicates that patients ≥ 75 years of age (relative to those 55 to 64 years of age) and patients with better emotional health had significantly lesser odds of reporting side effects. Findings were similar when the outcome was restricted to hot flashes (Table 4). In addition, educational attainment was associated with reporting hot flashes. Patients who completed at least the 12th grade were more than five times more likely to report hot flashes than those who did not complete high school education.
Multiple Logistic Regression Analysis With Side Effects as the Outcome: Any Side Effects
Multiple Logistic Regression Analysis With Side Effects as the Outcome: hot flashes
Tamoxifen Discontinuation
Patients who were estrogen receptor–positive were less likely to stop taking tamoxifen during the follow-up period, and patients who experienced side effects were more likely to stop taking tamoxifen (Table 5). With respect to individual side effects, there were statistically significant differences in the rates of tamoxifen discontinuation between those who did and did not report side effects as follows: depression (36% v 12%), nausea (40% v 14%), vision problems (75% v 14%), and vaginal bleeding (75% v 14%). Age and standard primary tumor therapy were not statistically significantly associated with tamoxifen discontinuance (Table 5).
Multiple Logistic Regression Analysis With Tamoxifen Discontinuation as the Outcome
DISCUSSION
In this study of older women with early-stage breast cancer, we have found that—in addition to clinical factors—age, physical function, standard primary tumor therapy, and aspects of the decision-making process were associated with tamoxifen use. The associations between older age, poorer physical function, and less-than-standard primary tumor therapy and tamoxifen use suggest the substitution of tamoxifen for therapies with greater likelihood of side effects and its use in women with poorer physical capacity. It is of interest, however, that while women 65 to 74 years of age were more likely than women 55 to 64 years of age to take tamoxifen, this was not true for the oldest women (75+ years of age). This oldest group was about as likely to take tamoxifen as was the youngest group.
As we have found in studies of primary tumor therapy14 and of the combination of primary tumor therapy and systemic adjuvant therapy,32 aspects of doctor-patient communication are independently associated with tamoxifen use. In the case of tamoxifen, the number of breast cancer physicians seen and patients’ confidence in their abilities to communicate with their physicians about breast cancer–related issues were both independently associated with its use. The diagnosis of breast cancer is frightening, and it may take several conversations with physicians for women to truly understand their options.
With respect to side effects, the oldest women and those with better emotional health at baseline were less likely to report experiencing side effects. This was also true when the analysis was restricted to hot flashes. The fact that the oldest women were less likely to report side effects is consistent with the fact that these women have the lowest levels of circulating estrogen and are therefore least likely to be affected by the antiestrogenic effects of tamoxifen. In addition, older persons in general are more likely to tolerate cancer treatments than are their younger counterparts.33 Women whose baseline general emotional health scores were lower (worse) were more likely to report side effects. Again, this is consistent with what is well-described in persons with mood disorders. Somatic symptoms are more likely to be reported by older (60+ years) community-dwelling persons with major depression as well as dysthymia than by their healthy counterparts.34
The strong relationship between education and hot flashes is difficult to explain, especially because the relationship persisted after statistical control for age and emotional health. The presence of an unmeasured confounder is a possible explanation. For example, it is possible that women with more education are also in vocational and life roles that make hot flashes less tolerable. It is also possible that the relationship is real. This finding warrants further exploration, particularly given the association between side effects and tamoxifen discontinuance.
Women who were estrogen receptor–negative and those who reported side effects were more likely to have stopped taking tamoxifen by 3 years after diagnosis. The former is not surprising, given the recent data suggesting that tamoxifen is less beneficial in women with estrogen receptor–negative tumors,23 and the latter could be anticipated. The individual side effects significantly associated with stopping tamoxifen were depression, nausea, visual complaints, and vaginal bleeding.
Although our findings represent some of the first to examine tamoxifen use and its sequelae among older women with early-stage breast cancer cared for in the community, they must be accepted with the following caveats. First, other than at baseline, we did not collect tamoxifen use information from medical records. Although we think that it is unlikely that women would report that they had stopped taking tamoxifen when in fact they had not, it is possible that women who reported continuing to take it had indeed stopped taking it. Were this the case, we would have underestimated the number of women who had discontinued tamoxifen therapy. Second, our sample of older women was relatively young, well-educated, and in good health. Variations in tamoxifen prescribing, side effects, and discontinuance might have been greater had our sample been more heterogeneous. Third, losses to follow-up and our reliance on annual telephone interviews precluded the collection of detailed temporal information about tamoxifen discontinuance. However, our strategy of excluding from analysis the 13 women who experienced recurrences should have minimized the impact of this circumstance. Fourth, we studied patterns of tamoxifen use during a period of time when guidelines for use were changing. Fifth, our sample is too small and our follow-up has not been long enough to ascertain whether the variations in primary tumor therapy and adjuvant tamoxifen therapy observed will be reflected in variations in the critically important outcomes of breast cancer recurrence and breast cancer-specific mortality.
Larger and longer-term studies are needed to examine these questions, because it is unlikely that clinical trials of therapies known to be efficacious in younger postmenopausal women will be undertaken to confirm or disprove their efficacy in older postmenopausal women. Furthermore, with broader indications for adjuvant tamoxifen and longer durations of recommended therapy, it is critical that patterns of discontinuance and their consequences be identified and quantified. Although the 15% discontinuation rate that we observed seems favorable in comparison with that observed in the National Surgical Adjuvant Breast and Bowel Project B-14 trial (23%), it still represents a significant degree of discontinuation and is perhaps greater than might be appreciated by many practicing oncologists.