Effects of Tamoxifen vs Raloxifene on the Risk of Developing Invasive Breast Cancer and Other Disease Outcomes: The NSABP Study of Tamoxifen and Raloxifene (STAR) P-2 Trial

Context Tamoxifen is approved for the reduction of breast cancer risk, and raloxifene has demonstrated a reduced risk of breast cancer in trials of older women with osteoporosis.

Objective To compare the relative effects and safety of raloxifene and tamoxifen on the risk of developing invasive breast cancer and other disease outcomes.

Design, Setting, and Patients The National Surgical Adjuvant Breast and Bowel Project Study of Tamoxifen and Raloxifene trial, a prospective, double-blind, randomized clinical trial conducted beginning July 1, 1999, in nearly 200 clinical centers throughout North America, with final analysis initiated after at least 327 incident invasive breast cancers were diagnosed. Patients were 19 747 postmenopausal women of mean age 58.5 years with increased 5-year breast cancer risk (mean risk, 4.03% [SD, 2.17%]). Data reported are based on a cutoff date of December 31, 2005.

Intervention Oral tamoxifen (20 mg/d) or raloxifene (60 mg/d) over 5 years.

Main Outcome Measures Incidence of invasive breast cancer, uterine cancer, noninvasive breast cancer, bone fractures, thromboembolic events.

Results There were 163 cases of invasive breast cancer in women assigned to tamoxifen and 168 in those assigned to raloxifene (incidence, 4.30 per 1000 vs 4.41 per 1000; risk ratio [RR], 1.02; 95% confidence interval [CI], 0.82-1.28). There were fewer cases of noninvasive breast cancer in the tamoxifen group (57 cases) than in the raloxifene group (80 cases) (incidence, 1.51 vs 2.11 per 1000; RR, 1.40; 95% CI, 0.98-2.00). There were 36 cases of uterine cancer with tamoxifen and 23 with raloxifene (RR, 0.62; 95% CI, 0.35-1.08). No differences were found for other invasive cancer sites, for ischemic heart disease events, or for stroke. Thromboembolic events occurred less often in the raloxifene group (RR, 0.70; 95% CI, 0.54-0.91). The number of osteoporotic fractures in the groups was similar. There were fewer cataracts (RR, 0.79; 95% CI, 0.68-0.92) and cataract surgeries (RR, 0.82; 95% CI, 0.68-0.99) in the women taking raloxifene. There was no difference in the total number of deaths (101 vs 96 for tamoxifen vs raloxifene) or in causes of death.

Conclusions Raloxifene is as effective as tamoxifen in reducing the risk of invasive breast cancer and has a lower risk of thromboembolic events and cataracts but a nonstatistically significant higher risk of noninvasive breast cancer. The risk of other cancers, fractures, ischemic heart disease, and stroke is similar for both drugs.

Trial Registration clinicaltrials.gov Identifier: NCT00003906

Trial Registration Published online June 5, 2006 (doi:10.1001/jama.295.23.joc60074).

Tamoxifen, a selective estrogen receptor modulator (SERM)¹ that has been used to treat both early and advanced breast cancer for more than 3 decades,² has been thoroughly evaluated for the reduction of the risk of both invasive and noninvasive breast cancer in women at increased risk.^3-9 Raloxifene, a second-generation SERM,¹ has been shown to reduce the incidence of mammary malignancy in preclinical models,^10-12 and several clinical trials evaluating it for the prevention and treatment of osteoporosis have suggested that it may also have a role in reducing the risk of invasive breast cancer in postmenopausal women.^13-20

The Multiple Outcomes Raloxifene Evaluation (MORE) study, a multicenter, randomized, placebo-controlled, double-blind clinical trial completed in 1999, was designed to test whether raloxifene, at a daily dose of either 60 mg or 120 mg, reduced the risk of fracture in postmenopausal women with osteoporosis.¹⁸ The primary end point was the development of fracture. Eligible women had a history of at least 1 vertebral body fracture. At 36 months of follow-up in 6828 women, the risk of vertebral fracture was reduced by 30% in the women who received raloxifene. These women had increased risk of venous thromboembolus when compared with those assigned to placebo (relative risk, 3.1; 95% confidence interval [CI], 1.5-6.2), but raloxifene did not cause an increase in vaginal bleeding.

A secondary end point in the MORE trial was invasive breast cancer.¹⁴ After 4 years of follow-up, there were 22 cases among 5129 postmenopausal women randomly assigned to raloxifene, compared with 39 cases among 2576 postmenopausal women assigned to placebo. The MORE trial concluded that among older postmenopausal women with osteoporosis, the risk of estrogen receptor–positive invasive breast cancer was decreased by 72% during 4 years of raloxifene treatment, with no apparent decrease in the incidence of estrogen receptor–negative tumors. Like tamoxifen, raloxifene increased the risk of thromboembolic disease but did not appear to increase the risk of endometrial cancer.

The Continuing Outcomes Relevant to Evista (CORE) trial examined the effect of 4 additional years of raloxifene therapy on the incidence of invasive breast cancer in women in the MORE trial who agreed to continue therapy.²⁰ After 4 years of participation in the CORE trial by 5213 participants, the risk of invasive breast cancer was reduced by 69% (hazard ratio, 0.31; 95% CI, 0.24-0.71) in the raloxifene group compared with the placebo group. During the 8 years of both the MORE and CORE trials, the incidence of invasive breast cancer and estrogen receptor–positive invasive breast cancer were reduced by 66% (hazard ratio, 0.34; 95% CI, 0.22-0.50) and 76% (hazard ratio, 0.24; 95% CI, 0.15-0.40), respectively, in the raloxifene group compared with the placebo group. During the CORE trial, the relative risk of thromboembolism in the raloxifene group compared with that in the placebo group was 2.17 (95% CI, 0.83-5.70). No increase in the risk of endometrial cancer was observed with raloxifene. Raloxifene has previously been shown to be less active than tamoxifen in stimulating endometrial cancer growth under laboratory conditions.^21-23

Based on our findings from the National Surgical Adjuvant Breast and Bowel Project (NSABP) Breast Cancer Prevention Trial (BCPT) (P-1), tamoxifen was approved by the US Food and Drug Administration.³ The NSABP Study of Tamoxifen and Raloxifene (STAR) trial was launched to directly compare tamoxifen with raloxifene in a population of women at increased risk for breast cancer. The objective of STAR is to compare raloxifene with tamoxifen in terms of their relative effects on the risk of invasive breast cancer and numerous other diseases influenced by tamoxifen in the BCPT.

The participants in this study were given a detailed description of the trial and provided written informed consent. The protocol and consent form were approved by the National Cancer Institute and the institutional review boards of all participating institutions. This trial was a double-blinded study with neither the participants nor their clinicians aware of which of the 2 treatments they were receiving. In addition, the central review of all protocol events and toxicities was also performed in a blinded fashion. The data reported are based on a cutoff date of December 31, 2005.

To be eligible for participation in the STAR trial, a woman had to have at least a 5-year predicted breast cancer risk of 1.66% based on the Gail model²⁴; be at least 35 years of age and postmenopausal; not be taking tamoxifen, raloxifene, hormone therapy, oral contraceptives, or androgens for at least the previous 3 months; not currently be taking either warfarin or cholestyramine; have no history of stroke, pulmonary embolism, or deep vein thrombosis (DVT) and no history of any malignancy diagnosed less than 5 years before randomization except basal or squamous cell carcinoma of the skin or carcinoma in situ of the cervix; have no uncontrolled atrial fibrillation, uncontrolled diabetes, or uncontrolled hypertension; and have no psychiatric condition that would interfere with adherence or a performance status that would restrict normal activity for a significant portion of each day. Postmenopausal women aged 35 years and older could enter the trial if they had a history of lobular carcinoma in situ (LCIS) treated by local excision alone. For the purposes of this trial, menopause was defined as (1) a history of at least 12 months without spontaneous menstrual bleeding or (2) a documented hysterectomy and bilateral salpingo-oophorectomy or (3) age 55 years or older with a hysterectomy with or without oophorectomy; or (4) age younger than 55 years, either with a hysterectomy without oophorectomy or with unknown ovary status, and with a documented level of follicle-stimulating hormone confirming elevation in the postmenopausal range.

A total of 184 460 women were screened using the modified Gail model^24,25 to determine their breast cancer risk. Of these, 96 368 had a 5-year risk of at least 1.66% (Figure 1). From this group, 20 616 agreed to be screened to determine full eligibility for the trial based on the medical criteria defined below; 20 168 were found to meet all eligibility criteria of the study. Of this latter group, 19 747 women expressed a desire to go forward with participation in the trial, signed a consent form, and were randomized to receive either tamoxifen or raloxifene. Participants were screened and enrolled through nearly 200 clinical centers throughout North America.

The mean age of participants at the time of randomization was 58.5 years. Nine percent were younger than 50 years, 49.8% were between 50 and 59 years, and 41.2% were 60 years or older (Table 1). More than 93% of participants were white, 2.5% were African American, 2.0% were Hispanic, and the remainder were of other racial/ethnic populations. Race information was collected because it is a risk factor for breast cancer and is one of the factors used in the Gail model to determine the predicted risk of breast cancer. It was collected as self-reported by the participants from options supplied by the investigators. More than half of the participants reported having undergone a hysterectomy prior to randomization. Almost 19% reported a family history of breast cancer in 2 or more first-degree relatives, and more than 71% reported a history of invasive breast cancer in 1 or more. More than 9% reported a personal history of LCIS prior to enrollment in the trial, and 22.7% had breast biopsy results prior to trial enrollment that showed either atypical ductal or lobular hyperplasia. The mean predicted 5-year risk of developing breast cancer among the study population was 4.03% (SD, 2.17%). The trial opened for participant entry on July 1, 1999, completing accrual on November 4, 2004.^26,27

Eligible women were randomly assigned to receive either tamoxifen (20 mg/d) or raloxifene (60 mg/d) for a maximum of 5 years.

The mean time of follow-up was 3.9 (SD, 1.6) years. Participant adherence to protocol therapy was within the levels expected and planned for when designing the trial. At the time of the cut-off for this analysis, the percentage of women persistent with the protocol regimen was 68.3% for those in the tamoxifen group and 71.5% for those in the raloxifene group. In designing the trial we planned for a nonadherence rate of 9.2% per year as measured by study dropouts and women who permanently discontinued therapy. The observed rate is below the planned level. The mean duration of treatment was 3.1 (SD, 1.7) and 3.2 (SD, 1.6) years for the tamoxifen and raloxifene groups, respectively; this differs from the mean duration of follow-up due to participants' discontinuing their study drug before the end of the 5-year period. Information on all individuals was included up to the time they underwent follow-up, regardless of whether or not they were adherent. During the course of the study 605 women in the tamoxifen group and 532 in the raloxifene group were lost to follow-up. In designing the trial we planned for a loss to follow-up rate of 2% per year of follow-up. The observed rate for this parameter is well below the planned rate. Those lost to follow-up contributed an average of 24 months of information before becoming lost to follow-up.

The primary end point was invasive breast cancer. Secondary end points included endometrial cancer, in situ breast cancer, cardiovascular disease, stroke, pulmonary embolism, DVT, transient ischemic attack, osteoporotic fracture, cataracts, death, and quality of life; data on all other invasive cancers also were collected prospectively. The cardiac disease end points included fatal and nonfatal myocardial infarction, severe angina, and acute ischemic syndrome. Severe angina was defined as angina requiring revascularization by percutaneous coronary intervention (angioplasty or stent) or coronary artery bypass graft surgery. Acute ischemic syndrome was defined as the presence of a new Q wave on electrocardiogram or angina requiring hospitalization without surgery. Vascular-related events included stroke and transient ischemic attack, as well as DVT and pulmonary embolism. To judge impact on osteoporosis, we selected 3 fracture sites known to be indicative of this disease: hip, spine, and Colles fractures of the wrist.

Randomization was accomplished using a biased-coin minimization algorithm.²⁸ The women were stratified by age (35-49, 50-59, ≥60 years), race/ethnicity (white, African American, Hispanic, other), history of LCIS (yes, no), and 5-year predicted risk of breast cancer (<2.5%, 2.5%-3.9%, and ≥4.0%). Follow-up occurred at 6 months after treatment initiation and every 6 months thereafter through 5 years. After 5 years, follow-up occurred annually. Clinical breast examination was to be performed every 6 months, and bilateral mammograms were to be performed annually. Gynecologic examinations, complete blood cell counts, and routine serum chemistry tests were to be obtained annually. Information regarding the occurrence of all protocol-defined end points was ascertained at each follow-up visit and verified by the collection of pathology reports, mammographic reports, surgical reports, discharge summaries, and other medical record documents. Self-reported symptoms were collected at each contact, and in-depth quality-of-life assessments were performed at selected clinical centers on a subset of 1983 participants using the Medical Outcomes Study Short-Form 36,^29-32 the Center for Epidemiologic Studies-Depression Scale,³³ and the Medical Outcomes Study Sexual Functioning Scale.³⁴ Symptoms were self-reported on a 5-point qualitative severity scale ranging from not having the symptom at all to having the symptom with severity that was reported as “extremely bothersome.” The 3 intermediate symptom severity levels were “slightly bothersome,” “moderately bothersome,” and “bothered quite a bit.” The findings from the in-depth quality-of-life and symptom assessments are reported in a companion article.³⁵

The STAR trial was monitored by an independent data monitoring committee composed of individuals with expertise in research ethics, oncology, clinical trial methodology, gynecology, epidemiology, and biostatistics. A consumer representative was also included as a member of the committee. The trial was monitored using a stopping rule based on that proposed by Fleming et al³⁶ using a 2-tailed log-rank test. The monitoring plan was based on detecting a statistically significant difference between treatment groups in the incidence of invasive breast cancer—the primary end point of the trial—and included 6 interim analyses and a final analysis initiated when at least 327 cases of invasive breast cancer were diagnosed in the entire study cohort. With this number of events, the study design provided a 95% probability that we would correctly conclude that the 2 treatments were equivalent, if they really were so.

All analyses were based on the treatment assignment made at the time of randomization, regardless of the treatment status at the time of analysis. The analyses included all randomized participants with follow-up data who were at risk at baseline for the diagnosis of an incident case of breast cancer. Comparison between treatment groups of the study end points was based on the determination of rates per 1000 person-years, the risk ratio (RR) contrasting the rate in the raloxifene group to the rate in the tamoxifen group, and the 95% CIs for the RR. The rate was defined as the number of observed events divided by the total number of observed event-specific person-years at risk. The CI for each RR was determined assuming that the event followed a Poisson distribution, conditioning on the total number of events and person-years at risk. Plots comparing the treatment groups in terms of the cumulative incidence over time were also developed for several end points. The cumulative incidence was determined accounting for the competing risk due to death.³⁷P values to assess statistically significant differences between cumulative incidence curves were determined by the log-rank test. Differences between treatment groups in the distributions of demographic and pathological characteristics were assessed by the χ² test. All P values reported are 2-sided; P<.05 was used to determine statistical significance. Analyses were performed using SAS version 8.2 (SAS Institute Inc, Cary, NC).

There was no difference between the effect of tamoxifen and the effect of raloxifene on the incidence of invasive breast cancer (Table 2 and Figure 2); there were 163 cases of invasive breast cancer in the women assigned to tamoxifen and 168 cases in those assigned to raloxifene. The rate per 1000 was 4.30 in the tamoxifen group and 4.41 in the raloxifene group (RR, 1.02; 95% CI, 0.82-1.28). The P value testing the difference between treatment groups in invasive breast cancer incidence determined from the log-rank test and including the stratification factors as covariates was .96. The cumulative incidence through 72 months for the 2 treatment groups was 25.1 and 24.8 per 1000 for the tamoxifen and raloxifene groups, respectively (P = .83, Figure 2). When the treatment groups are compared by baseline categories of age, history of LCIS, history of atypical hyperplasia, Gail model–derived 5-year predicted risk of breast cancer, and the number of relatives with a history of breast cancer, the pattern of no differential effect by treatment assignment is consistent, and none of the RRs in these subsets are statistically significant (Table 2). Histological characteristics, tumor size, and nodal status were derived from submitted pathology reports; there was no central review of pathology slides. When we assessed the pathological characteristics of the tumors in these patients, there were no differences between the treatment groups in regard to distributions by tumor size, nodal status, or estrogen receptor level (Table 2).

In contrast to the findings for invasive breast cancer, there were fewer noninvasive breast cancers in the tamoxifen group than in the raloxifene group (Table 3, Figure 2), although this difference did not reach statistical significance. There were 57 incident cases of noninvasive breast cancer among the women who took tamoxifen and 80 among the women who took raloxifene. (Rate for noninvasive breast cancer, 1.51 per 1000 women assigned to tamoxifen and 2.11 per 1000 women assigned to raloxifene [RR, 1.40; 95% CI, 0.98-2.00].) Cumulative incidence through 6 years was 8.1 per 1000 in the tamoxifen group and 11.6 in the raloxifene group (P = .052, Figure 2). About 36% of the cases were LCIS and 54% were ductal carcinoma in situ (DCIS), with the balance being mixed types. The pattern of fewer cases among the tamoxifen group was evident for both LCIS and DCIS.

There was a trend toward a decreased incidence of uterine cancer in the raloxifene group, but the difference was not statistically significant—36 cases (tamoxifen) vs 23 (raloxifene) (Table 3, Figure 3). Annual incidence rates were 2.00 per 1000 (tamoxifen) and 1.25 per 1000 women (raloxifene) (RR, 0.62; 95% CI, 0.35-1.08). Cumulative incidence rates through 7 years were 14.7 per 1000 (tamoxifen) and 8.1 per 1000 (raloxifene) (P = .07, Figure 3). Only 1 case of uterine cancer occurred among women younger than 50 years, in a participant in the tamoxifen group. At the time of analysis, clinicopathological stage was unknown for 3 cases (1 in the tamoxifen group, 2 in the raloxifene group). The majority of the others who developed uterine cancer (56 [91%]) were diagnosed with stage I disease. Of the remaining cases, there was 1 case of stage II disease in each of the treatment groups, 2 with stage III disease in the raloxifene group, and 1 with stage IV disease in the raloxifene group. Two of these cases were mixed Mullerian cell type; both were in the tamoxifen group.

While there were no significant differences with respect to risk of uterine cancer, there were differences between the treatment groups indicating that the effect of raloxifene on the uterus is less than that of tamoxifen. Among those who did not have a diagnosis of uterine cancer, there was a statistically significant difference between the groups in the incidence of uterine hyperplasia. The rates were 84% less in the raloxifene-treated group (14 cases) than in the tamoxifen-treated group (84 cases) (RR, 0.16; 95% CI, 0.09-0.29). This magnitude of difference between treatment groups was evident for hyperplasia both with and without atypia. For the tamoxifen and raloxifene groups, respectively, there were 12 cases and 1 case with atypia (RR, 0.08; 95% CI, 0.00-0.55) and 72 and 13 cases without atypia (RR, 0.18; 95% CI, 0.09-0.32). There also was a statistically significant difference between the treatment groups in the number of hysterectomies performed during the course of follow-up. Among women who were not diagnosed with endometrial cancer, there were 244 hysterectomies performed in those assigned to tamoxifen compared with 111 in those assigned to raloxifene (RR, 0.44; 95% CI, 0.35-0.56).

Table 4 shows the site-specific incidence of invasive cancers other than breast and uterine malignancy. There were no statistically significant differences between the treatment groups in regard to the number of women who developed any other cancer, in total or by specific site of diagnosis. Colorectal, lung, and leukemia/hematopoietic cancers were the most frequently diagnosed sites of other primary tumors. Differences between treatment groups were small for colorectal cancer (31 tamoxifen, 30 raloxifene) and for leukemia/hematopoietic cancers (32 tamoxifen, 28 raloxifene). The difference between the treatment groups was larger for lung cancer but not statistically significant; there were 28 cases of lung cancer in the tamoxifen group and 39 in the raloxifene group (RR, 1.38; 95% CI, 0.83-2.34). The next most frequently diagnosed sites of cancer included skin, ovary, and thyroid. There were 14 cases of skin cancer in the tamoxifen group and 12 in the raloxifene group (RR, 0.85; 95% CI, 0.36-1.98); 12 cases of ovarian cancer in the tamoxifen group and 18 in the raloxifene group (RR, 1.48; 95% CI, 0.68-3.38); and 8 cases of thyroid cancer in the tamoxifen group and 18 in the raloxifene group (RR, 2.24; 95% CI, 0.93-5.95).

The findings for the 3 types of ischemic heart disease events in the study are shown in Table 5. Overall, there were 114 events in those assigned to tamoxifen and 126 in those assigned to raloxifene. This difference was not statistically significant (RR, 1.10; 95% CI, 0.85-1.43). In addition, we found no significant differences between the treatment groups when we inspected the separate types of events. In women assigned to raloxifene, there were 11 fewer myocardial infarctions (RR, 0.77; 95% CI, 0.48-1.20), 12 more cases of severe angina (RR, 1.23; 95% CI, 0.84-1.81), and 11 more cases of acute ischemic syndrome (RR, 1.72; 95% CI, 0.88-3.50).

The findings for strokes and thromboembolic events are shown in Table 5. The difference in total number of strokes was small, with the number of events occurring in women assigned to tamoxifen being only 2 more than in those assigned to raloxifene (53 vs 51). There was no statistically significant difference between tamoxifen and raloxifene in the number of transient ischemic attacks that occurred (41 in the tamoxifen group vs 50 in the raloxifene group; RR, 1.21; 95% CI, 0.79-1.88). However, there was a statistically significant difference between the treatment groups for the incidence of thromboembolic events, with the raloxifene group experiencing fewer cases of pulmonary embolism and DVT. Overall, there were 141 events with tamoxifen and 100 with raloxifene, indicating that the risk was 30% less in the raloxifene group (RR, 0.70; 95% CI, 0.54-0.91). The cumulative incidence at 6 years was 21.0 per 1000 and 16.0 per 1000 for the tamoxifen and raloxifene groups, respectively (P = .01; Figure 3). Pulmonary embolism and DVT occurred in 54 vs 35 women (RR, 0.64; 95% CI, 0.41-1.00) and in 87 vs 65 women (RR, 0.74; 95% CI, 0.53-1.03) assigned to tamoxifen and raloxifene, respectively.

Although all fracture sites were recorded, the prespecified fracture sites of interest in the trial were the hip, spine, and Colles fractures of the wrist. No screening was performed for vertebral fractures, and only data for clinically apparent vertebral fractures were captured. There was no difference between treatment groups in the total number of these fractures or in the number for any of the specific types of fracture (Table 5). One hundred four women in the tamoxifen group and 96 in the raloxifene group experienced 1 of these fractures (RR, 0.92; 95% CI, 0.69-1.22). With regard to the specific types of fracture for the tamoxifen and raloxifene groups, respectively, there were 26 and 23 hip fractures (RR, 0.88; 95% CI, 0.48-1.60), 53 and 52 spine fractures (RR, 0.98; 95% CI, 0.65-1.46), and 27 and 23 Colles fractures (RR, 0.85; 95% CI, 0.46-1.53).

At the time of randomization, 2808 participants reported a history of cataracts (Table 5, Figure 4). Among those who were cataract-free at baseline, 707 developed cataracts during the course of follow-up. The differences between treatment groups for the incidence of cataracts and cataract surgery were statistically significant, with occurrence for both being less in the raloxifene group. Of those assigned to tamoxifen, 394 were diagnosed with cataracts and of those assigned to raloxifene, 313. The RR for cataract incidence was 0.79 (95% CI, 0.68-0.92). Cumulative incidence at 6 years for tamoxifen and raloxifene was 77.9 and 56.3 per 1000, respectively (P = .002). Of these women, 260 in the tamoxifen group and 215 in the raloxifene group had cataract surgery. The RR for cataract surgery was 0.82 (95% CI, 0.68-0.99).

A total of 197 deaths occurred among the study participants; the distribution of these fatalities by cause of death is shown in Table 6. Mortality in the treatment groups was similar. There were 101 deaths in those assigned to tamoxifen and 96 in those assigned to raloxifene, resulting in a rate per 1000 of 2.64 and 2.49, respectively (RR, 0.94; 95% CI, 0.71-1.26). Distribution by cause of death did not differ by treatment group. Forty-six and 47 women in the tamoxifen and raloxifene groups, respectively, died of cancer; 22 and 20, respectively, died as a result of cardiovascular disease; and 33 and 29, respectively, died of other causes. There were only 10 deaths from stroke, with 2 more occurring in women assigned to tamoxifen than in those assigned to raloxifene. The most frequent cause of death was lung cancer. There were 30 lung cancer deaths, 11 in the tamoxifen group and 19 in the raloxifene group; this difference was not statistically significant (RR, 1.72; 95% CI, 0.78-3.99). The proportion of smokers was the same in the tamoxifen and raloxifene treatment groups.

In this initial report of the STAR trial, raloxifene and tamoxifen were equivalent in efficacy for lowering the risk of invasive breast cancer. The cumulative incidence rates were 25.1 per 1000 women (raloxifene) vs 24.8 per 1000 (tamoxifen) (P = .83). Consistent with preclinical findings and with results from other large-scale studies showing that, compared with placebo, raloxifene does not increase endometrial cancer risk,^17,20 the rate of endometrial cancer in the STAR trial, although not statistically significant, was 38% lower in the raloxifene group than in the tamoxifen group. In contrast to tamoxifen, raloxifene does not reduce the risk of noninvasive breast cancer. Raloxifene also was associated with significantly less risk of thromboembolic events and cataracts. Combined, these results demonstrate that raloxifene is an alternative for lowering risk of invasive breast cancer in postmenopausal women with higher Gail risk scores and in those with LCIS for whom the Gail model does not apply.

The mean baseline 5-year breast cancer risk of STAR participants was 4%. More than 70% of participants had a history of invasive breast cancer in a first-degree maternal relative, an important component of a woman's perceived risk of breast cancer.³⁸ More than 20% of participants reported a history of atypical lobular or ductal hyperplasia on breast biopsy prior to entry into the trial, and more than 9% reported a history of LCIS. Atypical hyperplasia or LCIS and a family history of breast cancer in a first-degree relative increase the likelihood of developing breast cancer and account for the overrepresentation of women with these characteristics in the study population. The large benefit of tamoxifen that was demonstrated in the BCPT in women with atypical hyperplasia or LCIS may explain the large numbers of women with atypical hyperplasia or LCIS willing to participate in the STAR trial. The large proportion (51.6%) of women in STAR who had already undergone hysterectomy is likely attributable to the fact that such women had no risk for uterine malignancy, which was associated with tamoxifen in the BCPT.

In the STAR trial, tamoxifen and raloxifene had equivalent effects in reducing risk of invasive breast cancer in all examined subgroups, including women with a history of atypical hyperplasia or LCIS, who had the highest annual rates of invasive breast cancer (Table 2). The LCIS subgroup rates of 9.83 (tamoxifen) and 9.61 (raloxifene) per 1000 women are about 2.5 times higher than those for women who participated in the study and had no history of LCIS (3.76 [tamoxifen] and 3.86 [raloxifene] per 1000). The annual rates of invasive breast cancer among women aged 50 years or older (baseline) with no history of LCIS were similar in our 2 prevention studies (3.30 per 1000 in the BCPT). However, the rates in women with an LCIS history were about 1.7 times higher in the STAR trial than in the tamoxifen group of the BCPT. Similar patterns emerged when we compared the STAR trial and the BCPT in terms of breast cancer results in women with or without a baseline history of atypical hyperplasia. The greater mean age at entry and the higher 5-year risk of breast cancer in STAR participants may be the explanation for these differences. Pathological size of incident invasive breast cancers was similar between study groups; the status of the axillary lymph nodes or presence of estrogen receptor protein was not significantly different between the groups.

We found no statistically significant difference between raloxifene and tamoxifen in the risk of noninvasive disease (LCIS and DCIS) (incidence, 1.51 vs 2.11 per 1000 per year; RR, 1.40; 95% CI, 0.98-2.00). However, this study may have been underpowered to detect such a difference. Therefore, the clinical impact of this finding remains to be seen. The mechanism that would allow for a decrease in invasive breast cancers but a lesser impact on noninvasive disease is unknown. However, similar results were seen in the MORE and CORE studies,^8,20 in which raloxifene did not reduce the risk of noninvasive breast cancer, although the number of events in those studies was very small. All of these results taken together suggest that different SERMs have unique and specific mixes of benefits and risks and that neither a benefit nor a risk seen with one SERM can be generalized across the entire class.^39,40

We are in the process of obtaining further data on all of the cases of in situ breast cancer, but, to date, the noninvasive cancers occurring in these individuals in either treatment group appear to be similar to noninvasive cancers occurring in the general population. Most of the STAR cases were diagnosed as a result of mammograms demonstrating increasing calcifications. The individuals were undergoing careful follow-up and as a result their cancers were small, and most were treated surgically with lumpectomy. Approximately 36% of the cases were LCIS and 64% were DCIS or mixed LCIS and DCIS. The difference between the tamoxifen- and the raloxifene-treated individuals with DCIS was quite small (0.4 per 1000 per year). In the NSABP DCIS trial B-24, the 10-year rate of ipsilateral invasive breast tumors was only 6%.⁴¹ The CORE results through 8 years of follow-up show that raloxifene continues to offer a significant reduction in invasive disease, suggesting that raloxifene has a durable benefit despite this lesser impact on noninvasive disease.

Previous studies have shown that raloxifene does not increase the risk of uterine malignancy when compared with placebo.^14,17,20 In the STAR trial, only 59 invasive uterine cancers were diagnosed in both study groups during more than 76 000 woman-years of follow-up. Approximately 25% fewer cases of uterine cancer were diagnosed in the raloxifene than in the tamoxifen group (Table 3). Although uterine cancer of the mixed Mullerian type occurred in only 2 cases in the tamoxifen group of the STAR trial, there have been isolated case reports of this tumor associated with raloxifene.⁴² The rates of uterine cancer were 2.00 per 1000 (tamoxifen) and 1.25 per 1000 (raloxifene). This difference did not reach statistical significance (RR, 0.62; 95% CI, 0.35-1.08). However, endometrial hyperplasia, a risk factor for endometrial cancer, was far more common in the tamoxifen-treated group than in the raloxifene group (RR, 0.16; 95% CI, 0.09-0.29). The number of participants undergoing a hysterectomy for non–cancer-related reasons was significantly reduced in the raloxifene group (RR, 0.44; 95% CI, 0.35-0.56). It is important to note that the difference between the treatment groups in non–cancer-related hysterectomies has likely caused an underestimate of the true magnitude of endometrial cancer risk associated with tamoxifen and an underestimate of the true magnitude of difference between the 2 treatment groups for this end point.

Several nonsignificant differences in rates of invasive cancers other than breast or uterine cancer occurred between the 2 groups in the STAR trial. More cancers of the lung, kidney, ovary, and thyroid gland were seen with raloxifene, but the RRs were small and CIs were large (Table 4), suggesting that raloxifene is not associated with increased risks of these diseases. More data are needed to confirm this observation, and continuing follow-up of the women who participated in the STAR trial plus data from other studies involving raloxifene will shed more light on whether raloxifene is associated with other cancer risks.

There were no differences in the rates of myocardial infarction, severe angina, or acute ischemic syndrome between the tamoxifen and raloxifene groups in our study. Anecdotal reports of an association between raloxifene and elevated levels of serum cholesterol, triglycerides, or both appeared in the literature during the STAR trial,⁴³ but no such association was causally linked to an increased risk of cardiovascular events. We did see important differences, however, when we compared the rates of thromboembolic vascular events reported with the 2 agents. Although incidence of stroke or transient ischemic attack did not differ statistically significantly between groups, there was a statistically significant 30% reduction in the risk of thromboembolic events in the raloxifene group vs the tamoxifen group (RR, 0.70; 95% CI, 0.54-0.91). For pulmonary embolism, the reduction in risk was 36% and for DVT, 26%. Compared with placebo in the MORE trial,⁴⁴ raloxifene demonstrated a 3-fold increase in the risk of pulmonary embolism (RR, 3.0; 95% CI, 1.2-9.3) and a 60% increased risk of DVT (RR, 1.6; 95% CI, 0.91-2.86). These data indicate that both tamoxifen and raloxifene increase the risk of thromboembolic events but raloxifene less so.^45,46

The Women's Health Initiative trial indicated that estrogen plus progestin hormone therapy had hazard ratios of 1.41 (95% CI, 1.07-1.85) for stroke, 2.07 (95% CI, 1.49-2.87) for DVT, and 2.13 (95% CI, 1.39-3.25) for pulmonary embolism compared with placebo in postmenopausal women.^47,48 Venous thromboembolic events occurred at similar rates among the postmenopausal women who took hormone therapy in the Women's Health Initiative study and the postmenopausal women who took raloxifene in the STAR trial.

SERMs generally are known to reduce the risk of fracture in postmenopausal women. This was confirmed in the STAR trial. Rates of fracture were virtually identical in the raloxifene and tamoxifen groups and were similar to previously reported rates for both agents.^3,13,18

Tamoxifen is known to increase rates of both cataracts and cataract surgery.^3,49 In the STAR trial, compared with tamoxifen, raloxifene was associated with a 21% lower rate of cataract development and an 18% lower rate of cataract surgery. Although women may not perceive cataracts to be as serious as certain other STAR end points (such as uterine malignancy or venous thromboembolic events), the lower frequency of cataract development/cataract surgery in the women who took raloxifene is an important consideration in weighing its relative merits (vs tamoxifen) for reducing the risk of breast cancer. The rate of cataract development in the women who took tamoxifen was lower in the STAR trial than in the BCPT (12.3 vs 24.8 per 1000), but the rate of cataract surgery was higher (8.0 vs 4.7 per 1000). The rate of cataract development among women in the STAR trial who took raloxifene (9.7 per 1000) was substantially lower than this rate in the women in the BCPT who took placebo (21.7 per 1000).

Detailed data about symptoms and quality of life outcomes are presented in the article by Land and her colleagues in this issue of JAMA.³⁵ In summary, there were no significant differences between tamoxifen and raloxifene in participant-reported outcomes for physical and mental health.

A single small (fewer than 30 participants) trial examining short-term raloxifene therapy in premenopausal women found an increased incidence of ovarian cysts.⁵⁰ Without larger studies to further examine the efficacy and safety of raloxifene for breast cancer risk reduction in premenopausal women, raloxifene should not be prescribed for these women. Its risk-benefit advantages over tamoxifen in postmenopausal women notwithstanding, raloxifene is not a substitute for tamoxifen in premenopausal women.

We made special efforts in the STAR trial to recruit women from racial and ethnic groups in proportion to their numbers in the North American population. The trial enrolled more than 6% racial/ethnic minority women, almost double the number of minority participants enrolled in the BCPT, but we did not achieve our goal of fully proportional minority participation relative to the North American population of women at increased risk. Additional educational and recruitment efforts will be required in future clinical trials of risk reduction to achieve this goal. Raloxifene demonstrated a positive risk-benefit ratio among STAR minority participants, who likely are representative of North American minority women at increased risk for breast cancer.

The question of how long to use raloxifene to achieve optimal benefit remains unanswered by the STAR trial and other relevant published data. Continued follow-up is both required and ongoing among STAR participants, who agreed to undergo follow-up indefinitely after unblinding; this follow-up should help answer questions about the duration of raloxifene treatment for breast cancer risk reduction. Uncertainty over duration, however, should not be a barrier to use of the drug. Similar questions about the duration of tamoxifen therapy for breast cancer persists 25 years after that drug's initial approval for treating the disease and have not impeded either its widespread use or its benefit.

The initial NSABP P-1 trial demonstrated that tamoxifen could reduce the risk of invasive breast cancer by 49% and established proof of principle that the chemoprevention of breast cancer is possible. Nevertheless, primary care physicians have not broadly accepted the idea of tamoxifen use for breast cancer chemoprevention, in part because the drug has been characterized as too toxic. There is an important distinction to be made between tamoxifen and raloxifene relative to the latter's potential use for breast cancer chemoprevention. Tamoxifen was well known to oncologists who had used it extensively in the treatment of receptor-positive breast cancer, but the drug was not commonly prescribed by primary care physicians, who are the most involved in preventive care. Tamoxifen was viewed as a cancer drug, and the news reports highlighting its toxicity may have hampered primary care physicians in exploring its use as a preventive agent. In contrast, raloxifene is well known to the primary care community and is widely prescribed for the prevention and treatment of osteoporosis in postmenopausal women. More than 500 000 women in the United States are currently taking raloxifene (John Mershon, MD, Eli Lilly and Co, oral communication, April 18, 2006), the majority of whom are older and at lower risk of breast cancer than are the women in the STAR trial.

This trial confirms the previously reported benefit of raloxifene in reducing the risk of invasive breast cancer and indicates that raloxifene is as active as tamoxifen in this regard.⁵¹ If raloxifene is approved by the Food and Drug Administration for the prevention of breast cancer, primary care physicians may be more willing, given their experience with raloxifene, to prescribe it for breast cancer chemoprevention than they have been to prescribe tamoxifen.

Corresponding Author: Victor G. Vogel, MD, MHS, Magee-Womens Hospital, University of Pittsburgh School of Medicine, 300 Halket St, Room 3524, Pittsburgh, PA 15213-3221 (vvogel@magee.edu).

Published Online: June 5, 2006 (doi:10.1001/jama.295.23.joc60074).

Author Contributions: Dr Costantino had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Vogel, Costantino, Wickerham, Cronin, Robidoux, Lippman, Runowicz, Ford, Jordan, Wolmark.

Acquisition of data: Vogel, Costantino, Cronin, Atkins, Bevers, Fehrenbacher, Pajon, Wade, Robidoux, Margolese, James, Ganz, McCaskill-Stevens.

Analysis and interpretation of data: Vogel, Costantino, Cronin, Cecchini, Robidoux, Lippman, Ganz, Reis, Ford, Wolmark.

Drafting of the manuscript: Vogel, Costantino, Cronin, Cecchini, Robidoux.

Critical revision of the manuscript for important intellectual content: Vogel, Costantino, Wickerham, Cronin, Atkins, Bevers, Fehrenbacher, Pajon, Wade, Robidoux, Margolese, James, Lippman, Runowicz, Ganz, Reis, McCaskill-Stevens, Ford, Jordan, Wolmark.

Statistical analysis: Costantino, Cecchini.

Obtained funding: Costantino, Cronin, McCaskill-Stevens, Jordan.

Administrative, technical, or material support: Vogel, Costantino, Wickerham, Cronin, Atkins, Bevers, Wade, Robidoux, Margolese, James, Ganz, Reis, Ford, Wolmark.

Study supervision: Vogel, Costantino, Cronin, Bevers, Fehrenbacher, Pajon, Lippman, Ford, Wolmark.

Active NSABP STAR P-2 Clinical Centers as of May 3, 2006: United States:Albert Einstein Comprehensive Cancer Center of Montefiore Medical Center, Bronx, NY: Joseph Sparano (Principal Investigator [PI]), Una Hopkins (Project Coordinator [PC]); Albert Einstein Healthcare Network, Philadelphia, Pa: Ajit M. Desai (PI), Mary P. Murphy (PC); Allegheny Cancer Center Protocol Office, Pittsburgh, Pa: Norman Wolmark (PI), Deborah S. Davison (PC); Alvin J. Siteman Cancer Center at Barnes-Jewish Hospital & Washington University School of Medicine, St Louis, Mo: Antonella L. Rastelli (PI), Gwendolyn Randall (PC); Arizona Cancer Center, Tucson: David S. Alberts (PI), Heidi Fritz (PC); Aultman Hospital, Canton, Ohio: Eleftherios Paul Mamounas (PI), Nancy E. Hansen (PC); Baltimore-Washington Regional STAR Center, Baltimore, Md: John L. Zapas (PI), Sally Brown (PC); Baptist Cancer Institute, Jacksonville, Fla: Neil Abramson (PI), Andrea B. Canto (PC); Baptist Clinical Research Center, Memphis, Tenn: Lee S. Schwartzberg (PI), Patricia M. Gish (PC); Baptist Hospital East/CBC, Louisville, Ky: Daniel C. Scullin, Jr (PI), Judy E. Sisk (PC); Bassett Healthcare, Cooperstown, NY: Anne N. Nafziger (PI), Jennifer M. Victory (PC); Baylor College of Medicine, Houston, Tex: Powel H. Brown (PI), Toni E. Sinclair (PC); Baylor-Sammons Cancer Center, Dallas, Tex: Michael D. Grant (PI), Elaine Lagow (PC); Berkshire Hematology Oncology, PC, Pittsfield, Mass: Harvey Zimbler (PI), Gloria F. Gero (PC); Beth Israel Deaconess Medical Center, Boston, Mass: Nadine Tung (PI), Rena Leib (PC); Boca Raton Community Hospital, Boca Raton, Fla: Louise E. Morrell (PI), Mary Klaus-Clark (PC); CAMC Health Education and Research Institute Center for Cancer Research, Charleston, WVa: Steven J. Jubelirer (PI), Augusta Kosowicz (PC); Cancer Research Network Inc, Plantation, Fla: Charles L. Vogel (PI), Filiz Gokce (PC); CCOP Cancer Research for the Ozarks, Springfield, Mo: J. Wendall Goodwin (PI), Deborah A. Cane (PC); CCOP Cedar Rapids Oncology Project-Mercy Medical Center, Cedar Rapids, Iowa: Abby R. Thrower (PI), Kathy V. Fleming (PC); CCOP Marshfield Clinic, Marshfield, Wis: James L. Hoehn (PI), Kelly J. Noreen (PC); CCOP Alton Ochsner Medical Foundation, New Orleans, La: Carl G. Kardinal (PI), Kate Rodger (PC); CCOP Atlanta Regional, Atlanta, Ga: Janice Galleshaw (PI), Karen E. Packard (PC); CCOP Benaroya Research Institute at Virginia Mason, Seattle, Wash: Andrew D. Jacobs (PI), Cathleen Goetsch (PC); CCOP Carle Cancer Center, Urbana, Ill: Kendrith M. Rowland (PI), Shelly B. McCaskill (PC); CCOP Central Illinois, Decatur, Ill: James Lloyd Wade III (PI), Peggy S. Verrill (PC); CCOP Christiana Care Health Services, Newark, Del: Timothy Francis Wozniak (PI), Pamela G. Eppes (PC); CCOP Colorado Cancer Research Program, Denver: Eduardo R. Pajon, Jr (PI), Robin Szekely (PC); CCOP Columbia River Oncology Program, Portland, Ore: Keith S. Lanier (PI), Jenny L. DeKorne (PC); CCOP Columbus, Columbus, Ohio: John Philip Kuebler (PI), Julie O’Brian (PC); CCOP Dayton, Dayton, Ohio: Howard M. Gross (PI), Bernadette L. Bensman (PC); CCOP Duluth Cancer Center, Duluth, Minn: Steven A. Kuross (PI), Becky Buchanan (PC); CCOP Evanston/Highland Park Hospital, Highland Park, Ill: Douglas E. Merkel (PI), Nancy J. Haefling (PC); CCOP Grand Rapids Clinical Oncology Program, Grand Rapids, Mich: Marianne K. Lange (PI), Connie M. Szczepanek (PC); CCOP Greenville, Greenville, SC: Jeffrey K. Giguere (PI), Claudette Phinney (PC); CCOP Heartland Cancer Research: St Louis, Mo: Alan P. Lyss (PI), Rebecca L. Tackett (PC); CCOP Hematology-Oncology Associates of CNY, Syracuse, NY: Jeffrey J. Kirshner (PI), Sue Averson (PC); CCOP Iowa Oncology Research Association, Des Moines: Roscoe F. Morton (PI), Mary L. Loots (PC); CCOP Kalamazoo, Kalamazoo, Mich: Raymond S. Lord III (PI), Joan Westendorp (PC); CCOP Kansas City Clinical Oncology Program, Kansas City, Mo: Rakesh Gaur (PI), Mabel Goodpaster (PC); CCOP Main Line Health, Bryn Mawr, Pa: Thomas G. Frazier (PI), Rosemary Wiggins (PC); CCOP Mayo Clinic Scottsdale, Scottsdale, Ariz: Tom R. Fitch (PI), Kortni R. Jones (PC); CCOP MeritCare Hospital, Fargo, ND: Ralph Levitt (PI), Valorie Steichen (PC); CCOP Metro-Minnesota, St Louis Park, Minn: Patrick James Flynn (PI), Ann Deshler (PC); CCOP Michigan Cancer Research Consortium Community Clinical Onc Program, Ann Arbor: Philip J. Stella (PI), Kathy A. Flynn (PC); CCOP Missouri Valley Cancer Consortium, Omaha, Neb: Gamini Soori (PI), Mary Beth Wilwerding (PC); CCOP Montana Cancer Consortium, Billings: David B. Myers (PI), Shirley Hall (PC); CCOP Nevada Cancer Research Foundation, Las Vegas: John A. Ellerton (PI), Kathleen Van Wagenen (PC); CCOP North Shore University Hospital, Manhasset, NY: Lora R. Weiselberg (PI), Doreen Mayberry (PC); CCOP Northern Indiana Research Consortium, South Bend: Robin T. Zon (PI), Maria Schwieter (PC); CCOP Northern New Jersey/Hackensack University Medical Center, Hackensack: Deborah M. Capko (PI), Kathleen Laico (PC); CCOP Northwest, Tacoma, Wash: Lauren K. Colman (PI), Karyn Hart (PC); CCOP Oklahoma, Tulsa: James B. Lockhart (PI), Kayleen R. Joyce (PC); CCOP Santa Rosa Memorial Hospital, Santa Rosa, Calif: Ian C. Anderson (PI), Marianne Dias (PC); CCOP Scott and White Hospital, Temple, Tex: Kathy Kimmey (PI), Jan P. Pickett (PC); CCOP Sioux Community Cancer Consortium/SDHRF, Sioux Falls, SD: Loren K. Tschetter (PI), Judy Norman (PC); CCOP South Florida STAR Group, Miami Beach: Rogerio C. Lilenbaum (PI), Francine Mineau (PC); CCOP Southeast Cancer Control Consortium Inc, Winston-Salem, NC: James N. Atkins (PI), Robin Burgess (PC); CCOP St Louis-Cape Girardeau, St Louis, Mo: Bethany Graham Sleckman (PI), Carol J. Antinora (PC); CCOP Toledo Community Hospital Oncology Program, Toledo, Ohio: Paul L. Schaefer (PI), Jeanette M. Ceglio (PC); CCOP Upstate Carolina, Spartanburg, SC: Steven W. Corso (PI), Donna Hewitt (PC); CCOP Western Regional, Phoenix, Ariz: David K. King (PI), Susan Colvin (PC); CCOP Wichita, Wichita, Kan: Shaker R. Dakhil (PI), Marge J. Good (PC); CCOP William Beaumont Hospital, Royal Oak, Mich: David A. Decker (PI), Jeanne Archer (PC); City of Hope National Medical Center, Duarte, Calif: Lawrence D. Wagman (PI), Annette S. Brown (PC); Clarian Health Partners Inc, Indianapolis, Ind: Anna Maria V. Storniolo (PI), Jane V. Berby-Todd (PC); Cleveland Clinic Health System Star BCPT, Warrensville Heights, Ohio: Roger F. Classen (PI), Alice Somich (PC); Coborn Cancer Center, St Cloud, Minn: Harold E. Windschitl (PI), Linda Meehl (PC); Columbia University Medical Center, New York, NY: Victor R. Grann (PI), Rossy Sandoval (PC); Comprehensive Cancer Institute, Huntsville, Ala: Jeremy K. Hon (PI), Susan C. Meyer (PC); Connecticut BCPT Task Force, Hartford: Patricia A. DeFusco (PI), Camille Alyce Servodidio (PC); Dana-Farber Cancer Institute, Boston, Mass: Judy E. Garber (PI), Jennifer Thibodeau (PC); Desert Regional Medical Center, Palm Springs, Calif: Janet K. Ihde (PI), Gregory S. Jackson (PC); Don and Sybil Harrington Cancer Center, Amarillo, Tex: Brian Pruitt (PI), Kathy Morris (PC); Duke University Medical Center, Durham, NC: Paul K. Marcom (PI), Donna C. Harper (PC); East Carolina University, Greenville, NC: Rosa Cuenca (PI), Melinda F. Friday (PC); Eastern Cooperative Onc Group/National Medical Association, Philadelphia, Pa: Edith P. Mitchell (PI), Charmaine Green (PC); Eastern Maine Medical Center, Bangor: A. Merrill Garrett (PI), Laurie Lewis (PC); Ellis Fischel Cancer Center, Columbia, Mo: Edward R. Sauter (PI), Marta L. Fuemmeler (PC); Fox Chase Cancer Center, Philadelphia, Pa: Mary B. Daly (PI), Joan James (PC); Fred Hutchinson Cancer Research Center, Seattle, Wash: Richard B. Clarfeld (PI), Joelle Machia (PC); Geisinger Breast Clinic, Danville, Pa: James F. Evans (PI), Marie Lamey (PC); Glendale Memorial Hospital Comprehensive Cancer Center, Glendale, Calif: Deanna Attai (PI), Mary M. Eichenhofer (PC); Greenebaum Cancer Center, Baltimore, Md: Barry R. Meisenberg (PI), Nancy S. Tait (PC); H. Lee Moffitt Cancer Center and Research Institute, Tampa, Fla: Susan Minton (PI), Pamela Dawson (PC); Harris Methodist Fort Worth, Fort Worth, Tex: Mark W. Redrow (PI), Tamara D. Plant (PC); Hendrick Cancer Center, Abilene, Tex: Victor J. Hirsch (PI), Jo Ann Jameson (PC); Hennepin County Medical Center, Minneapolis, Minn: Richard T. Zera (PI), Carol A. Schmidt (PC); Huntsman Cancer Institute at the University of Utah, Salt Lake City: John Harris Ward (PI), Chaunda L. Biggs (PC); Indiana Community Cancer Care, Indianapolis: William M. Dugan, Jr (PI), Carolyn Rhoton (PC); Ireland Cancer Center at Case Western Reserve University, Cleveland, Ohio: Rosemary A. Leeming (PI), Cheryl B. Henkin (PC); Joe Arrington Cancer Research and Treatment Center, Lubbock, Tex: Paul J. Anderson (PI), Vickie Reynolds (PC); Kaiser Permanente San Diego, San Diego, Calif: Jonathan Polikoff (PI), Carol P. Elliott (PC); Kaplan Comprehensive Cancer Center at NYU Medical Center, New York, NY: Anna Pavlick (PI), Karen Carapetyan (PC); Kimmel Cancer Center at Thomas Jefferson University, Philadelphia, Pa: Bruce M. Boman (PI), Colleen M. Cohick (PC); Lehigh Valley Hospital, John and Dorothy Morgan Cancer Center, Allentown, Pa: Gregory R. Harper (PI), Deborah R. Kane (PC); Loma Linda University Cancer Institute, Loma Linda, Calif: Carlos A Garberoglio (PI), Pedro Bernal (PC); Lombardi Cancer Center, Washington, DC: Claudine Isaacs (PI), Joy Dritschilo (PC); Loyola-Cardinal Bernardin Cancer Center, Maywood, Ill: Kathy Albain (PI), Kathy O’Day (PC); Mayo Clinic Rochester, Rochester, Minn: Sandhya Pruthi (PI), Jean M. Jensen (PC); MBCCOP Gulf Coast, Mobile, Ala: Paul O. Schwarzenberger (PI), Carolyn A. Robinson (PC); MBCCOP Howard University Cancer Center, Washington, DC: Lucile L. Adams-Campbell (PI), Joan Pearson (PC); MBCCOP Louisiana State University Health Sciences Center, New Orleans: Julia A. Lawrence (PI), Tasha N. Moore (PC); MBCCOP Our Lady of Mercy Medical Center, Bronx, NY: Janice Dutcher (PI), Maria Serrano (PC); MBCCOP San Juan, San Juan, Puerto Rico: William Caceres (PI), Doris Cuadrado (PC); MBCCOP University of Hawaii, Honolulu: Nancy L. Furumoto (PI), Ann M. Kelminski (PC); MBCCOP University of Illinois at Chicago (UIC): Divyesh G. Mehta (PI), Judith Murray (PC); MBCCOP University of New Mexico Cancer Research and Treatment Center, Albuquerque: Aroop Mangalik (PI), Ann W. Parsons (PC); MBCCOP Virginia Commonwealth University, Richmond: Mary Helen Hackney (PI), Gwendolyn G. Parker (PC); Memorial Sloan-Kettering Cancer Center, New York, NY: Leslie Montgomery (PI), Yong Anna Park (PC); Mercy Cancer Center, Scranton, Pa: Martin Hyzinski (PI), Vee A. Pauli (PC); Methodist Cancer Center Omaha, Omaha, Neb: Stephen J. Lemon (PI), Kathy Christiansen (PC); Michigan State University, East Lansing: Carol A. Slomski (PI), Cheryl Robins (PC); Mid-Delta Family Practice Clinic II Inc, Cleveland, Miss: Nathaniel Brown (PI), Sarah L. Cooks (PC); Milwaukee Prevention Trial Sinai Samaritan Medical Center MIL, Waukesha, Wis: Paul J. LeMarbre (PI), Julie Fries (PC); Naval Medical Center San Diego, San Diego, Calif: Preston S. Gable (PI), Debbie A. Vacek (PC); Norris Cotton Cancer Center at the Dartmouth-Hitchcock Medical Center, Lebanon, NH: Bradley A. Arrick (PI), Jo A. Strohbehn (PC); North Idaho Cancer Risk Assessment Clinic–NICRAC Coeur D, Alene: Haluk Tezcan (PI), Tolli J. Willhite (PC); North Mississippi Medical Center, Tupelo: Charles W. Montgomery (PI), Linda M. Roof (PC); Northern California Kaiser Permanente Divison of Research, Oakland: Louis Fehrenbacher (PI), Roberta Gross (PC); Northwestern Memorial Hospital, Chicago, Ill: Seema A. Khan (PI), Rachel Cohn (PC); NortonHealthcare Inc, Louisville, Ky: John T. Hamm (PI), Belma Kantardzic (PC); Ohio State University/James CHRI, Columbus: William B. Farrar (PI), Beth A. Putz (PC); Oklahoma NSABP Consortium, Oklahoma City: Karl K. Boatman (PI), Glenda J. Bell (PC); Oregon Health and Science University, Portland: Rodney F. Pommier (PI), Paula Fry (PC); Providence Hospital Cancer Center, Southfield, Mich: Judie R. Goodman (PI), Jaswinder Grewal (PC); Radiant Research, Stuart, Fla: Darrell N. Fiske (PI), Diane Andersen (PC); Rapid City Regional Oncology Group, Rapid City, SD: Richard Charles Tenglin (PI), Beth A. Bement-Stump (PC); Resurrection Medical Center, Chicago, Ill: Adrian Bianco (PI), Mary Ann J. Kosic (PC); Roswell Park/WNY STAR Consortium, Buffalo, NY: Stephen B. Edge (PI), Susan G. Wood (PC); Rush University Medical Center, Chicago, Ill: Janet M. Wolter (PI), Yelizaveta Yanovskaya (PC); Saint Barnabas Health Care System Consortium, Newark, NJ: Frederick B. Cohen (PI), Luzminda B. Costin (PC); San Gabriel Clinical Oncology Research Program, Pasadena, Calif: Benjamin T. Stafford (PI), Maria Francisco-Arriola (PC); Scripps-Stevens Cancer Center, San Diego, Calif: Joel I. Bernstein (PI), Melba Nelson (PC); St Luke's Breast Care Center, Chesterfield, Mo: David P. Krajcovic (PI), Margaret A. Klass (PC); St Vincent's Hospital-Manhattan, New York, NY: Stephanie F. Bernik (PI), Bethany S. Gourley (PC); St Mary Medical Center/Pacific Shores Medical Group, Long Beach, Calif: Nerses Simon Tchekmedyian (PI), Jann E. Buaiz (PC); St Vincent Hospitals and Health Services, Indianapolis, Ind: Ruemu E. Birhiray (PI), Mary Amy Waddell (PC); Stanford University, Palo Alto, Calif: Robert W. Carlson (PI), Christine Schurman (PC); Summa Health System, Akron, Ohio: Ralph Douglas Trochelman (PI), Nancy J. Rinella (PC); SUNY Upstate Medical University, Syracuse, NY: Jayne R. Charlamb (PI), Linda Ellinwood (PC); Sutter Health Cancer Research Group-Eastern Division, Sacramento, Calif: Vincent A. Caggiano (PI), Isabel Rodrigues-Fong (PC); Texas Tech University Health Sciences Center Southwest Cancer Center, Lubbock: Everardo Cobos (PI), Wendy A. Carr (PC); The Cancer Center at Glens Falls Hospital, Glens Falls, NY: Robert W. Sponzo (PI), Barbara A. Sponzo (PC); The Cancer Institute of New Jersey, New Brunswick: Deborah L. Toppmeyer (PI), Angelina G. Owens (PC); The Johns Hopkins Oncology Center, Baltimore, Md: Kala Visvanathan (PI), Sarah Ord (PC); The Regional Cancer Center, Erie, Pa: Roy E. Smith (PI), Lisa Quirk (PC); The Western Pennsylvania Hospital, Pittsburgh: John A. Lech (PI), Laura Getty Gibson (PC); Thompson Cancer Survival Center, Knoxville, Tenn: Daniel Mark Ibach (PI), Josie B. Stanga (PC); UCLA-Center for Health Sciences, Los Angeles, Calif: Patricia A. Ganz (PI), Barbara Kahn (PC); University Hospital & Medical Center at Stony Brook, Stony Brook, NY: Stefan Madajewicz (PI), Gerty Fortune (PC); University of Arkansas for Medical Sciences/Arkansas Cancer Res Center: Little Rock: V. Suzanne Klimberg (PI), Maureen A. Smith (PC); University of California San Diego Cancer Center: Anne M. Wallace (PI), Thuy Pham (PC); University of California, Irvine Medical Center, Orange : Randall F. Holcombe (PI), Brandy Fitzhenry (PC); University of Chicago, Chicago, Ill: Nora T. Jaskowiak (PI), Yolanda M. Davis (PC); University of Cincinnati, Cincinnati, Ohio: Elizabeth A. Shaughnessy (PI), Ruth L. Steele (PC); University of Iowa Hospitals and Clinics, Iowa City: Neal Walter Wilkinson (PI), Michelle Arnold (PC); University of Kansas Medical Center, Kansas City: William R. Jewell (PI), Holly J. Smith (PC); University of Kentucky Medical Center, Lexington: Edward H. Romond (PI), Valorie A. Gray (PC); University of Louisville/Brown Cancer Center, Louisville, Ky: Jeffrey B. Hargis (PI), Jackie H. Sheffield (PC); University of Michigan University Hospital, Ann Arbor: Sofia Merajver (PI), Janet H. Tarolli (PC); University of North Carolina at Chapel Hill, Chapel Hill: Stephen A. Bernard (PI), Nancy Pope (PC); University of Pittsburgh Medical Center Health Systems, Pittsburgh, Pa: Victor G. Vogel (PI), Marsha MacIntyre (PC); University of Tennessee Medical Center, Knoxville: Daniel M. Green (PI), Mary E. Donovan (PC); University of Texas Health Science Center at San Antonio: Morton S. Kahlenberg (PI), Betty Razvillas (PC); University of Texas M. D. Anderson Cancer Center, Houston: Therese B. Bevers (PI), Valerie Sepeda (PC); University of Texas Southwestern Medical Center at Dallas: Ann Marilyn Leitch (PI), Mary E. De Haas (PC); University of Wisconsin Comprehensive Cancer Center, Madison: James A. Stewart (PI), Mary Hamielec (PC); URCC Cancer Prevention Network, Rochester, NY: Gary R. Morrow (PI), Kelly A. Kita (PC); USC/Norris Cancer Center, Los Angeles, Calif: Darcy V. Spicer (PI), Elizabeth M. Sales (PC); Valley Hospital, Ridgewood, NJ: Harold Bruck (PI), Melissa F. Cooper (PC); Vanderbilt Cancer Center, Nashville, Tenn: Mark Kelley (PI), Elizabeth Broome (PC); Vassar Brothers Hospital, Poughkeepsie, NY: Carol S. Woo (PI), Camille Finkle (PC); Vermont Cancer Center University of Vermont, Burlington: Seth P. Harlow (PI), Karen M. Wilson (PC); Wake Forest University School of Medicine, Winston Salem, NC: Edward A. Levine (PI), Lisa B. Pruitt (PC); Wayne State University Karmanos Cancer Institute, Detroit, Mich: Michael Simon (PI), Mary Akagi (PC); West Florida Cancer Institute, Pensacola: Mark S. Boatright (PI), Laura J. Malek (PC); Wilford Hall Medical Center, Lackland AFB, Tex: Allyson Lynn Harroff (PI), Blanca N. Velazquez (PC); Yakima Valley Memorial Hospital/North Star Lodge Cancer Center, Yakima, Wash: Thomas E. Boyd (PI), Cheryl Hopkins (PC); Yale University School of Medicine-Yale Cancer Center, New Haven, Conn: Donald R. Lannin (PI), Michele D. Alguard (PC); York Cancer Center, York, Pa: Ronald E. Hempling (PI), Debi Oxenberg (PC). Canada:CancerCare Manitoba, Winnipeg: Andrew L. Cooke (PI), Jeannette Kostenuik (PC); Centre Hospitalier Affilie Universitaire de Quebec, Hopital du St-Sacrement, Quebec: Louise Provencher (PI), Guylaine Julien (PC); Cross Cancer Institute, Edmonton, Alberta: Barbara E. Krause (PI), Peggy Blize (PC); Jewish General Hospital, Montréal, Quebec: Richard Gordon Margolese (PI), Linda Robitaille (PC); Le Groupe Hospitalier de l’Universite de Montréal, Montréal, Quebec: Andre Robidoux (PI), Nicole Tremblay (PC); Montreal General Hospital, Montréal, Quebec: Michael P. Thirlwell (PI), Catherine Meaney (PC); Ottawa Hospital Regional Cancer Centre, Ottawa, Ontario: Shailendra Verma (PI), Peggy Florack (PC); Royal Victoria Hospital, Montréal, Quebec: Henry Ryusuke Shibata (PI), Rosalia Santos (PC); St Michael's Hospital Breast Centre, Toronto, Ontario: Jarley Koo (PI), Daisy J. Dastur (PC); Thunder Bay Regional Health Sciences Centre, Thunder Bay, Ontario: Margaret Lyn Anthes (PI), Cindy Sinnott (PC); Tom Baker Cancer Centre, Calgary, Alberta: Alexander H. G. Paterson (PI), Barb Gore-Hickman (PC); UBC-Vancouver Hospital & Health Science Center, Vancouver, British Columbia: Urve Kuusk (PI), Lynn Fearn (PC); Women's College Campus, Toronto, Ontario: H. Lavina A. Lickley (PI), Maria Oldfield (PC).

Funding/Support: This study was supported by Public Health Service grants U10-CA-37377, U10-CA-69974, U10CA-12027, and U10CA-69651 from the National Cancer Institute, National Institutes of Health, Department of Health and Human Services; and by AstraZeneca Pharmaceuticals and Eli Lilly and Co.

Role of the Sponsor: The study sponsors had no role in any aspect of study design; in the collection, analysis, and interpretation of data; or in the development of the manuscript. Per contractual arrangement, the manuscript was submitted to AstraZeneca and Eli Lilly before submission.

Acknowledgment: We thank Barbara C. Good, PhD, Director of Scientific Publications for the NSABP, for editorial assistance and Lynne Anderson, Biostatistical Center, NSABP, for her valuable assistance in the conduct of the study.