Fax: (312) 926-1722
Acceptance of tamoxifen chemoprevention by physicians and women at risk
Article first published online: 22 MAR 2004
Copyright © 2004 American Cancer Society
Volume 100, Issue 9, pages 1800–1806, 1 May 2004
How to Cite
Tchou, J., Hou, N., Rademaker, A., Jordan, V. C. and Morrow, M. (2004), Acceptance of tamoxifen chemoprevention by physicians and women at risk. Cancer, 100: 1800–1806. doi: 10.1002/cncr.20205
- Issue published online: 19 APR 2004
- Article first published online: 22 MAR 2004
- Manuscript Accepted: 12 FEB 2004
- Manuscript Revised: 10 FEB 2004
- Manuscript Received: 8 JAN 2004
- SPORE. Grant Number: P50-CA89018
- Avon Foundation
- tamoxifen chemoprevention;
- atypical hyperplasia (AH);
- lobular carcinoma in situ (LCIS);
- breast carcinoma risk
In the National Surgical Adjuvant Breast and Bowel Project (NSABP) P-1 trial, tamoxifen was shown to reduce breast carcinoma risk by 49% in high-risk women. The purpose of the current study was to identify factors associated with being offered, and accepting, tamoxifen chemoprevention.
The records of 219 women who sought risk evaluation after the publication of the NSABP P-1 trial between September 1998 and October 2002 were reviewed. Risk was calculated using the model of either Gail et al. or Claus et al. The impact of individual risk factors on the offering and acceptance of tamoxifen was compared using the Fisher exact test and logistic regression analysis.
Tamoxifen was offered to 137 women (63%) in the current study. The magnitude of Gail risk, age, menopausal status, hysterectomy, and history of lobular carcinoma in situ (LCIS) or atypical hyperplasia (AH) were all found to be significant predictors of a patient being offered tamoxifen. On multivariate analysis, only a history of AH or LCIS and hysterectomy were found to be significant, with odds ratios of 20.3 and 3.4, respectively. Fifty-seven of the women who were offered tamoxifen (42%) took the drug. Only a history of LCIS or AH and older age were found to be predictive of tamoxifen acceptance.
In the current study, risk due to AH or LCIS was found to be the main predictor of being offered and accepting tamoxifen chemoprevention. Cancer 2004. © 2004 American Cancer Society.
Laboratory evidence that tamoxifen reduced the incidence of mammary carcinoma in rats and mice,1, 2 coupled with the observation that women who received tamoxifen as adjuvant endocrine therapy for early-stage breast carcinoma had a significantly reduced rate of contralateral breast carcinoma,3, 4 led to studies of tamoxifen as a chemopreventive for women at increased risk of developing breast carcinoma. To our knowledge, four prospective randomized trials have been performed to date to test the efficacy of tamoxifen for this purpose.5–8 The reported benefits of tamoxifen varied from no reduction to a 49% reduction in risk, largely due to differences with regard to study design and the level of risk of the populations studied. A recent overview analysis of these trials reported a 38% reduction in the incidence of breast carcinoma (95% confidence interval [95% CI], 28–48%).9
On September 2, 1998, the Food and Drug Administration (FDA) Oncology Drugs Advisory Committee recommended that the FDA approve tamoxifen for breast carcinoma risk reduction. Since the approval of tamoxifen as a chemopreventive, there have been only limited data published regarding its use in risk-eligible women in the clinical setting. Exactly how many women in the U.S. are eligible and, more important, will benefit from tamoxifen for risk reduction is not clear. Extrapolating from data from the year 2000 National Health Interview Survey, Freedman et al.10 estimated that approximately 10.2 million women, or 15.5% of all women in the U.S., were eligible for tamoxifen using the National Surgical Adjuvant Breast and Bowel Project (NSABP) P-1 trial criteria.5 By applying a risk/benefit ratio analysis11 to the estimates derived from the National Health Interview Survey, approximately 2.4 million women were estimated to derive a net benefit from tamoxifen. Although the study by Freedman et al.10 provided a better identification of the population with the potential to benefit from tamoxifen, the acceptance of tamoxifen chemoprevention by physicians and women at risk in the clinical setting remains largely unknown.
To identify factors that are important to both physicians and women in making a decision regarding the use of tamoxifen as a chemopreventive agent, we identified a group of 219 women seeking risk evaluation at our breast center after the publication of the NSABP P-1 trial.5 The rates of tamoxifen being offered and accepted as well as factors affecting these rates were analyzed.
MATERIALS AND METHODS
Between the publication of the NSABP P-1 study in September 19985 and October 2002, 219 women seeking risk evaluation were seen at the Lynn Sage Breast Center. Information concerning breast carcinoma risk factors was collected. Women at risk on the basis of a family history of breast carcinoma that was suggestive of genetic mutation were offered genetic counseling. The model developed by Gail et al.12 was used to numerically estimate risk, except in cases of suspected genetic mutation or in those patients at risk on the basis of lobular carcinoma in situ (LCIS). Diagnoses of atypical hyperplasia (AH) or LCIS made at other institutions were confirmed by a review of pathology slides prior to the estimation of risk. For this report, the model developed by Claus et al.13 was used to estimate the risk of developing breast carcinoma in women who had a strong family history of breast carcinoma and had a 5-year Gail risk of < 1.7%, but was not part of the clinical counseling process.
Neutral risk counseling regarding the risks and benefits of tamoxifen was provided to all women. Based on the data from the NSABP P-1 trial,5 women were advised that 5 years of tamoxifen therapy reduced the risk of breast carcinoma by approximately 50%. Women at risk on the basis of AH were advised that the risk reduction was approximately 85%. Premenopausal women were counseled regarding side effects such as hot flashes, menstrual irregularity, vaginal discharge, vaginal dryness, and the need to use barrier contraception to prevent pregnancy. Postmenopausal women, in addition to being advised about symptomatic side effects, were informed of the increased risk of endometrial carcinoma, thromboembolic problems, and cataract surgery. Information concerning both absolute benefits and risks was provided using data from the NSABP P-1 trial.5
After obtaining institutional review board approval, high-risk women were identified by chart review. All office notes made between September 1998 and October 2002 were reviewed to determine whether women were offered or accepted tamoxifen at any time during the study period. The percentage of patients who were offered or accepted tamoxifen was compared between risk factor categories using the Fisher exact test. Age was analyzed using the Wilcoxon rank sum test. Multivariate analysis was performed using logistic regression analysis.
The characteristics and risk factor profiles of the 219 women studied are summarized in Table 1. The mean age of the patients was 47 years (range, 22–75 years); 47% of the group was age ≤ 45 years, and 93% were white. There were 73 women (33%) with a diagnosis of AH or LCIS. The median 5-year Gail risk of breast carcinoma in the 189 women who did not have a diagnosis of LCIS was 2.3% (range, 0.3–9.1%). Gail risk was not calculated for the remaining 30 women with LCIS because the model is not validated in LCIS. There were 103 women (54%) with a Gail risk of 1.7–4.9%, whereas 69 women (37%) had a 5-year Gail risk of < 1.7%. Of the 69 women with a 5-year Gail risk < 1.7%, 61 reported a family history of at least 1 affected first-degree or second-degree relative with breast carcinoma. Two women, including one with no family history of breast carcinoma, had a history of radiation exposure during childhood. Another woman, who had no family history of breast carcinoma, had a personal history of ovarian carcinoma and was found to be a BRCA2 gene mutation carrier. There were a total of 5 known breast carcinoma susceptibility gene carriers in our cohort of 219 women (2.3%): 3 with BRCA1 mutations and 2 with BRCA2 mutations. All had a 5-year Gail risk of < 0.6%.
|Mean age (range)||47||(22–75)|
|5-year Gail risk (%) (n = 189)|
|Family history of breast carcinomaa||170||78|
|History of AH||43||20|
|History of LCIS||30||14|
|History of hysterectomy|
|Parity (among premenopausal women)|
Characteristics of Women Offered Tamoxifen
Of the 219 women studied, 137 were offered tamoxifen (63%). The pattern of offering tamoxifen was significantly influenced by the magnitude of the 5-year Gail risk, as shown in Table 2. Twenty of the 69 women with a 5-year Gail risk < 1.7% were offered tamoxifen. The cumulative risk of breast carcinoma (at ages 49 years, 59 years, or 69 years, depending on the patient age at the time of consultation), calculated using the model of Claus et al.,13 was 6.4% (range, 2.2–12.8%) for those women with a Gail risk < 1.7% who were offered tamoxifen compared with 3.9% (range, 0.5–11.0%) for those women who were not offered tamoxifen (P = 0.003). Approximately 60% of the women who were offered tamoxifen had at least 3 affected family members compared with 11% of the women who were not offered tamoxifen.
|Risk factors||Overall no.||Tamoxifen offered no. (%)||Tamoxifen not offered no. (%)||P valuea|
|All patients||219||137 (63)||82 (37)|
|Mean age (yrs) (SEM) (range)||48 (0.8) (22–75)||44 (1.3) (24–67)||0.003|
|5-year Gail risk||189||107 (57)||82 (43)|
|< 1.7||69||20 (29)||49 (71)||< 0.001|
|1.7–2.9||57||38 (67)||19 (33)|
|3.0–4.9||46||32 (70)||14 (30)|
|≥ 5.0||17||17 (100)||0 (0)|
|Prior breast biopsy|
|Yes||131||93 (71)||38 (29)||0.003|
|No||88||44 (50)||44 (50)|
|History of LCIS or AH|
|Yes||73||68 (93)||5 (7)||< 0.0001|
|No||146||69 (47)||77 (53)|
|Family History of breast carcinoma|
|Yes||170||97 (57)||73 (43)||0.002|
|No||49||40 (82)||9 (18)|
|Pre||120||66 (55)||54 (45)||0.012|
|Post||99||71 (72)||28 (28)|
|History of hysterectomy|
|Yes||42||34 (81)||8 (19)||0.007|
|No||174||101 (58)||73 (42)|
|Parity (among premenopausal women)|
|Nulliparous||51||25 (49)||26 (51)||0.27|
|Parous||69||41 (59)||28 (41)|
Women with Gail risk scores > 1.7% were offered tamoxifen in 67–100% of cases (Table 2), with an increasing proportion of women being offered tamoxifen as the Gail score increased. Of the 137 women who were offered tamoxifen, 68 (49.6%) had a diagnosis of LCIS or AH, which was the individual risk factor found to be most strongly predictive of being offered tamoxifen (P < 0.0001) (Table 2). All 30 women with LCIS and 39 of the 44 patients with AH were offered tamoxifen. Three of the 5 women with AH who were not offered tamoxifen were age < 40 years (age 30 years, 36 years, and 39 years, respectively) and had 5-year Gail risks of 0.9%, 1.1%, and 1.5%, respectively. Of the other two women with atypia who were not offered tamoxifen, one had a diagnosis of AH in a single duct and one woman failed to return for follow-up surveillance.
Age and menopausal status also were found to be significantly correlated with being offered tamoxifen, with older women (mean age, 48 years) more likely to be offered tamoxifen than their younger counterparts (mean age, 44 years) (P = 0.003). Approximately 72% of postmenopausal women were offered tamoxifen compared with 55% of premenopausal women (P = 0.012), although this may in part reflect the increased tendency to offer tamoxifen to women who had undergone a hysterectomy (Table 2). The effect of individual risk factors on the likelihood of being offered tamoxifen is summarized in Table 2. In a multivariate analysis of the significant variables in Table 2 (excluding the overall Gail risk score), only 2 variables a history of LCIS or AH and hysterectomy were found to be significant, with odds ratios of 20.3 (95% CI, 7.0–59.0; P = 0.0001) and 3.4 (95% CI, 1.4–8.3; P = 0.007), respectively.
Of note, there were 33 women with a 5-year Gail risk of > 1.7 and < 5.0 who were not offered tamoxifen. A review of their records demonstrated no documentation of medial conditions (e.g., thromboembolic events, endometrial carcinoma, or stokes) that would contraindicate the use of the drug.
Characteristics of Women Accepting Tamoxifen
Fifty-seven of the 137 women offered tamoxifen (42%) opted to take the drug. As shown in Table 3, there was no correlation found between the Gail risk score and the acceptance of tamoxifen in women with Gail risks > 1.7%. The only factors found to be predictive of accepting tamoxifen were a history of AH or LCIS and older age. Women accepting tamoxifen had a mean age of 51 years compared with 46 years in those women who refused the drug (P = 0.03). Thirty-eight of 57 women (67%) who accepted tamoxifen had a history of AH or LCIS, and women with AH or LCIS were much more likely to accept tamoxifen than those at risk on the basis of other factors (56% vs. 28%; P < 0.0001) (Table 3).
|Risk factors||Tamoxifen offered No.||Tamoxifen accepted No. (%)||Tamoxifen refused No. (%)||P valuea|
|All patients||137||57 (42)||80 (58)|
|Mean age (yrs) (SEM), (range)||51 (0.9) (39–67)||46 (1.2) (22–75)||0.03|
|5-year Gail risk|
|< 1.7||20||0 (0)||20 (100)||< 0.0001|
|1.7–2.9||38||17 (45)||21 (55)|
|3.0–4.9||32||16 (50)||16 (50)|
|≥ 5.0||17||8 (47)||9 (53)|
|Prior breast biopsy|
|Yes||93||44 (47)||49 (53)||0.06|
|No||44||13 (30)||31 (70)|
|History of LCIS or AH|
|Yes||68||38 (56)||30 (44)||< 0.0001|
|No||69||19 (28)||50 (72)|
|Family History of breast carcinoma|
|Yes||97||38 (39)||59 (61)||0.45|
|No||40||19 (48)||21 (53)|
|Pre||66||24 (36)||42 (64)||0.30|
|Post||71||33 (46)||38 (54)|
|History of hysterectomy|
|Yes||34||18 (53)||16 (47)||0.16|
|No||101||38 (38)||63 (62)|
|Parity (among premenopausal women)|
|Nulliparous||25 (38)||8 (32)||17 (68)||0.61|
|Parous||41 (62)||16 (39)||25 (61)|
A comparison of the 38 women with LCIS or AH who accepted tamoxifen and the 30 with these diagnoses who refused the drug demonstrated no significant differences with regard to mean age, family history of breast carcinoma, menopausal status, or history of hysterectomy.
Overall, a family history of breast carcinoma, a history of prior breast biopsy, menopausal status, and having undergone a hysterectomy were not found to be predictive of the acceptance of tamoxifen. Four of the 57 women (7%) who accepted tamoxifen stopped the drug during the study period: 2 because of mood changes and 1 because of night sweats, and 1 patient opted to undergo bilateral prophylactic mastectomy for greater risk reduction.
Tamoxifen was approved for the reduction of breast carcinoma risk by the FDA in 1998, but little is known regarding its use in the clinical setting. Port et al.14 reported that only 2 of 43 risk-eligible women (4.7%) seen at a risk clinic in New York City consented to receive tamoxifen. Limited information regarding the risk characteristics of their population was provided, but of the two women who elected to receive tamoxifen, one had AH and one had LCIS. Our study of 219 women, including 73 with LCIS or AH, clearly demonstrates that the presence of these histologic risk factors is a highly significant predictor of being offered, and accepting, tamoxifen chemoprevention. One potential explanation for this finding is the greater risk reduction achieved in women at risk on the basis of having LCIS or AH compared with those at risk from other causes. The NSABP P-1 study demonstrated an overall risk reduction in breast carcinoma of 49%.5 The risk reduction was 86% for the 579 women with AH and 56% for women with LCIS.5 However, it is unlikely that the increased efficacy of tamoxifen in this subpopulation is the entire explanation for the greater acceptance of the drug in practice because women at risk on the basis of AH or LCIS consented to participate in the NSABP P-1 trial more frequently than those with an equal level of risk due to other factors,15 even before the added benefit of tamoxifen therapy in this subpopulation was recognized. In the NSABP P-1 trial, 23% of risk-eligible women without atypia joined the study compared with 32% of those with atypia, a 39% increase in enrollment.15 In the ongoing Study of Tamoxifen and Raloxifene (STAR) trial, 36% of risk-eligible women with atypia had joined the trial at the time of last follow-up compared with 21% of all eligible subjects, a 71% relative increase.16 This suggests that women and their physicians may regard histologic lesions as more valid markers of personal risk than other risk factors, a finding that is consistent with the literature indicating that women undergoing prophylactic mastectomy are more likely to have undergone benign breast biopsies than those who chose not to have the procedure.17, 18
Overall, 26% of the 219 women in the current study received tamoxifen for chemoprevention, a figure that is substantially higher than the 4.7% rate reported by Port et al.,14 and very similar to the 32% acceptance rate of women with atypia who were offered enrollment in the Breast Cancer Prevention Trial.15 The similarity of these figures is not encouraging. Participation in the NSABP P-1 trial involved consenting to take a drug of unproven benefit, as well as the possibility of being randomized to the placebo arm of the study. This is associated with considerably more uncertainty than the clinical use of a drug proven to reduce breast carcinoma incidence by > 50%. The findings of the current study indicate that both physician practice and the attitudes of at-risk women are responsible for the low rates of tamoxifen usage. Only 63% of the current study population was offered tamoxifen, and the likelihood of being offered tamoxifen was related to the level of risk as predicted by the Gail model.12 In general, the pattern of offering tamoxifen reflects the selection of women at increased risk who are likely to have a greater net benefit from the drug: those with AH or LCIS and those who have undergone a hysterectomy. However, the older mean age of patients offered tamoxifen and their greater likelihood of being postmenopausal does not fit this pattern because the most favorable risk benefit ratio is seen in premenopausal women.11 Women with a 5-year Gail risk < 1.7% were rarely offered tamoxifen, although many of these women had significant elevations of risk as estimated by the model of Claus et al.,13 reflecting the greater availability of the Gail model calculator during the time of the current study.
The current study data concerning factors predicting the acceptance of tamoxifen chemoprevention demonstrated no relation between the level of breast carcinoma risk or factors that would influence the net benefit of tamoxifen and tamoxifen acceptance for women with a 5-year Gail risk of > 1.7%. Women who had undergone a hysterectomy were no more likely than their counterparts who had not undergone hysterectomy to accept tamoxifen, despite the observation that a fear of side effects is a significant factor in tamoxifen refusal.14 We previously reported that premenopausal women with T1a or T1b lymph node-negative breast carcinoma who are nulliparous are significantly less likely than their parous counterparts to accept chemoprevention with tamoxifen.19 In the current study, parity appeared to have no effect on the physician's decision to offer tamoxifen or a woman's decision to accept the drug.
The importance of a diagnosis of AH or LCIS for the use of tamoxifen chemoprevention is interesting in light of recent studies suggesting that atypia may be identified in asymptomatic high-risk women by epithelial sampling, either with random fine-needle aspiration20 or ductal lavage.21. Histologic atypia was present in only 3.6%, of 10,366 breast biopsies performed for clinical abnormalities in a report by Dupont and Page.22 In a review of 2983 biopsies that were performed for mammographic abnormalities, a 7% incidence of atypia was noted.23 However, Fabian et al.20 and Dooley et al.21 reported atypia in 21% and 24%, respectively, of high-risk women when random epithelial sampling was performed. Studies of prophylactic mastectomy specimens from high-risk women also suggest that clinically silent histologic atypia is present in a significant number of high-risk women.24 The current study findings regarding the importance of atypia in the decision to undertake tamoxifen chemoprevention provide a strong rationale for epithelial sampling in the high-risk woman who is uncertain regarding tamoxifen use.
A potential weakness of the current study is that it is a reflection of practice at a single institution, which may not be generalizable to other settings. This is particularly true of the data concerning physician patterns of offering tamoxifen. However, the pattern observed of offering tamoxifen to women with greater levels of risk or those at lower risk of serious side effects (i.e., women who have undergone a hysterectomy) is consistent with guidelines for selecting women for tamoxifen chemoprevention,11, 25, 26 and provides a real-world clinical perspective on the relevance of the estimates of the number of women predicted to derive a net benefit from tamoxifen based on the National Health Interview Survey data.10 There is less reason to believe that the data regarding factors influencing the acceptance of tamoxifen are not generally applicable. Although the population in the current study was nearly all white, so was the population studied in the NSABP P-1 Trial5 and the population from which the model of Gail et al. was derived and validated,12 and estimates of tamoxifen benefit and harm are felt to be more valid in these women than in women of other races.27 The pattern of a greater rate of tamoxifen acceptance in women at risk on the basis of histologic lesions is consistent with the findings regarding recruitment to two large, prospective, randomized chemoprevention trials involving a broad spectrum of women.15, 16 A recent review of the data regarding chemoprevention of breast carcinoma concluded that the number of eligible women who would choose to receive 5 years of chemoprevention was uncertain and that more research regarding the counseling of at-risk women concerning the effects of chemoprevention was needed.27 This study provides initial data characterizing women who accept and those who refuse chemoprevention. Further investigation of the differences in the perception of women at risk on the basis of histologic factors compared with those at risk because of other causes may provide the basis for improved educational strategies for chemoprevention counseling in the future.
- 19Impact of the breast cancer prevention trial (BCPT) on tamoxifen (tam) utilization in patients with T1a/b N0 breast cancer [abstract 96]. Proc Am Soc Clin Oncol. 2003; 22: 24., , , et al.
- 26American Society of Clinical Oncology technology assessment on breast cancer risk reduction strategies: tamoxifen and raloxifene. J Clin Oncol. 1999; 17: 1929–1955., , , .