Recent clinical trials have shown the efficacy of the selective

Transcription

1 Annals of Internal Medicine Review Systematic Review: Comparative Effectiveness of Medications to Reduce Risk for Primary Breast Cancer Heidi D. Nelson, MD, MPH; Rongwei Fu, PhD; Jessica C. Griffin, MS; Peggy Nygren, MA; M.E. Beth Smith, DO; and Linda Humphrey, MD, MPH Background: Trials demonstrate the efficacy of medications to reduce the risk for invasive breast cancer. Purpose: To summarize benefits and harms of tamoxifen citrate, raloxifene, and tibolone to reduce the risk for primary breast cancer. Data Sources: MEDLINE and Cochrane databases from inception to January 2009, Web of Science, trial registries, and manufacturer information. Study Selection: Predefined eligibility criteria were used to select articles. English-language reports of randomized, controlled trials (RCTs) for benefits and RCTs and observational studies for harms were included. Data Extraction: Two reviewers assessed study data, quality, and applicability. Data Synthesis: Seven placebo-controlled RCTs and 1 head-tohead trial provide results for main outcomes. Tamoxifen (risk ratio, 0.70 [95% CI, 0.59 to 0.82]; 4 trials), raloxifene (risk ratio, 0.44 [CI, 0.27 to 0.71]; 2 trials), and tibolone (risk ratio, 0.32 [CI, 0.13 to 0.80]; 1 trial) reduce risk for invasive breast cancer compared with placebo by 7 to 10 per 1000 women per year. Tamoxifen and raloxifene reduce estrogen receptor positive breast cancer but not estrogen receptor negative breast cancer, noninvasive breast cancer, or mortality. All medications reduce fractures. Tamoxifen (risk ratio, 1.93 [CI, 1.41 to 2.64]; 4 trials) and raloxifene (risk ratio, 1.60 [CI, 1.15 to 2.23]; 2 trials) increase thromboembolic events by 4 to 7 per 1000 women per year; raloxifene causes fewer events than tamoxifen. Tamoxifen increases risk for endometrial cancer (risk ratio, 2.13 [CI, 1.36 to 3.32]; 3 trials) compared with placebo by 4 per 1000 women per year and causes cataracts compared with raloxifene. Tibolone causes strokes in older women. Limitations: Bias, trial heterogeneity, and a dearth of head-to-head trials limit this review. Data are lacking on doses, duration, and timing of the medications; long-term effects; and nonwhite and premenopausal women. Conclusion: Three medications reduce risk for primary breast cancer but increase risk for thromboembolic events (tamoxifen, raloxifene), endometrial cancer (tamoxifen), or stroke (tibolone). Primary Funding Source: Agency for Healthcare Research and Quality. Ann Intern Med. 2009;151: For author affiliations, see end of text. This article was published at on 15 September Recent clinical trials have shown the efficacy of the selective estrogen receptor modulators tamoxifen citrate and raloxifene and of the selective tissue estrogenic activity regulator tibolone to reduce the risk for invasive breast cancer in women without preexisting cancer. Tamoxifen and raloxifene are approved by the U.S. Food and Drug Administration (FDA) for this purpose among women at high risk for breast cancer. The FDA indications define high risk as having a breast biopsy showing lobular carcinoma in situ or atypical hyperplasia, 1 or more firstdegree relatives with breast cancer, or a 5-year predicted risk for breast cancer of 1.66% or more calculated by the modified Gail model (1 3). These are similar to entry criteria of the major U.S. clinical trials (4 8). Tamoxifen is also approved for treatment of early and advanced estrogen receptor positive breast cancer in pre- and postmenopausal women. Raloxifene is most often used to prevent and treat osteoporosis in postmenopausal women. Tibolone is not currently approved for use in the United States, but it is approved to treat menopausal symptoms in 90 countries and to prevent osteoporosis in 45 (9, 10). In the United States, use of these medications for reducing breast cancer risk is low (4). This comparative effectiveness review summarizes the available evidence for the effectiveness and safety of tamoxifen citrate, raloxifene, and tibolone for reducing risk for primary breast cancer in women in general and among population subgroups of women. This review highlights outcomes that are most commonly associated with the medications and are clinically most important. METHODS For all steps of the review, we followed a standard protocol that is consistent with the Agency for Healthcare Research and Quality Effective Healthcare Program Methods Reference Guide for Effectiveness and Comparative Effec- See also: Print Editors Notes Editorial comment Related articles , 727, 738 Web-Only Appendix Appendix Tables Appendix Figures CME quiz Conversion of graphics into slides 17 November 2009 Annals of Internal Medicine Volume 151 Number

2 Review Comparative Effectiveness of Medications to Reduce Breast Cancer Risk Context Medications that reduce breast cancer risk may have other potential benefits and harms. Contribution This review of 8 trials found that tamoxifen, raloxifene, and tibolone each reduced risk for invasive breast cancer in women more than placebo. Tamoxifen and raloxifene reduced risk for estrogen receptor positive but not estrogen receptor negative breast cancer or death. All drugs reduced fracture risks. They also increased risk for thromboembolic events (tamoxifen and raloxifene), endometrial cancer (tamoxifen), and strokes (tibolone). Implication Although tamoxifen, raloxifene, and tibolone sometimes reduce risk for primary breast cancer, they also may increase a woman s risk for particular harmful events. The Editors tiveness Reviews (11). Additional information about methods for this review, including detailed search strategies, inclusion criteria, extraction, and rating processes, is available in the appendices to this article ( and the full technical report ( (12). Key questions guiding this review were developed through the Effective Healthcare Program. Investigators created an analytic framework that incorporated the key questions and outlined the patient population, interventions, and clinical outcomes (Appendix Figure 1, available at Additional outcomes and evidence related to key questions about risk stratification and adherence to medications are available in the full technical report (12). The target population includes women who did not have preexisting invasive or noninvasive breast cancer or precursor conditions and were not known carriers of breast cancer susceptibility mutations (BRCA1, BRCA2, or others). Data Sources and Searches In conjunction with a research librarian, the investigators searched MEDLINE, the Cochrane Central Register of Controlled Trials, and the Cochrane Database of Systematic Reviews from inception to January 2009 for relevant English-language reports of studies, systematic reviews, and meta-analyses. We also manually reviewed reference lists, citations for major trials in Web of Science, and clinical trial registries. We requested scientific information packets from manufacturers of the 3 medications, although the only packet provided was for raloxifene. We contacted investigators of the major trials for additional unpublished data specifically addressing population subgroups but received none. Study Selection Before reviewing abstracts and articles, we developed inclusion and exclusion criteria for studies based on the key questions and target population. We included studies with treatment durations of 3 months or more that enrolled 100 or more participants to ensure adequate drug exposure and power to support results. For comparative benefits, we included only doubleblind, placebo-controlled, or head-to-head randomized, controlled trials (RCTs) of tamoxifen, raloxifene, or tibolone to reduce risk for breast cancer that enrolled women without preexisting breast cancer. We included trials that were designed and powered to demonstrate invasive breast cancer incidence as a primary or secondary outcome. The technical expert panel for this project advised including only RCTs because of the lack of observational studies of tamoxifen and raloxifene with breast cancer outcomes in women without preexisting cancer and concerns for bias in observational studies in which women were using these medications for other indications. We defined our inclusion criteria for comparative harms more broadly. We included RCTs and observational studies of tamoxifen, raloxifene, or tibolone in women without breast cancer that were designed for multiple types of outcomes. However, studies must have had a nonuser comparison group or direct comparisons between tamoxifen, raloxifene, or tibolone to be included. We considered all adverse outcomes at all reported follow-up times to capture potential short- and long-term adverse effects. However, because the National Surgical Adjuvant Breast and Bowel Project P-1 (NSABP P-1) trial was unblinded after investigators reported initial results in 1998, we focused on data from the earlier 1998 publication (8) and then compared these results with data from the subsequent 2005 publication (7). Appendix Figure 2, available at provides detailed inclusion and exclusion criteria for benefits and harms. After an initial review of abstracts, we retrieved fulltext articles of potentially relevant material and conducted a second review to determine inclusion. A second reviewer confirmed results of the initial reviewer, and discrepancies were resolved by group consensus. Data Abstraction and Quality Assessment From the included studies, investigators abstracted study design, setting, participant characteristics, enrollment criteria, interventions, numbers enrolled and lost to follow-up, methods of outcome ascertainment, and results for each outcome. A second investigator confirmed the accuracy of the abstracted information. We used predefined criteria developed by the U.S. Preventive Services Task Force to assess the quality of individual studies (good, fair, or poor) (13). We assessed study applicability by following the population, intervention, comparator, outcomes, timing of outcomes measurement, and setting format (good, fair, or poor) (11). Two November 2009 Annals of Internal Medicine Volume 151 Number 10

3 Comparative Effectiveness of Medications to Reduce Breast Cancer Risk Review investigators independently rated quality and applicability of each study, and final ratings were determined by consensus. Investigators determined the overall strength of the body of evidence through group consensus by using the Evidence-based Practice Center GRADE (Grading of Recommendations Assessment, Development, and Evaluation) approach (high, moderate, low, or insufficient) (11). Data Synthesis and Analysis We combined results of eligible trials to obtain more precise estimates of the major health outcomes. To determine the appropriateness of meta-analysis, we considered clinical and methodological diversity and assessed statistical heterogeneity. We abstracted or calculated estimates of risk ratios (rate ratio, hazard ratio, or relative risk) and their SEs from each trial and used them as the effect measures (Appendix, available at We assessed the presence of statistical heterogeneity among the studies by using standard chi-square tests and evaluated the magnitude of heterogeneity by using the I 2 statistic (14). We used a random-effects model to account for variation among studies (15). When there is no variation among studies, the random-effects model yields the same results as a fixedeffects model. For all meta-analyses, we combined results separately for tamoxifen and raloxifene and provided 95% CIs. To explore whether combined estimates differ among subpopulations, we performed subgroup analysis by age ( 50 vs. 50 years), family history of breast cancer (yes vs. no), use of menopausal hormone therapy (yes vs. no), menopausal status (premenopausal vs. postmenopausal), and body mass index ( 25 vs. 25 mg/kg 2 ), when at least 2 studies reported results. We also performed subgroup analysis for tamoxifen trials stratified by active versus posttreatment periods when studies reported these data. We used Stata software, version 9.1, for all these analyses (Stata, College Station, Texas). To facilitate the evaluation of benefits and harms across trials, we abstracted or calculated event rates per 1000 woman-years for both treatment and placebo groups using steps similar to those used to obtain risk ratios. When the event rates were not reported or calculable, we indicate them as such. To obtain the combined event rates, we conducted a meta-analysis of the placebo event rates by using a random-effects Poisson model and raw data of number of events and woman-years of follow-up. We used PROC NLMIXED in SAS 91.3 software for this analysis (SAS Institute, Cary, North Carolina). Role of the Funding Source The Agency for Healthcare Research and Quality provided the initial key questions and copyright release for this manuscript but did not participate in the literature search, data analysis, or interpretation of the results. RESULTS From electronic database searches, secondary references, and the scientific information packet, we identified 4230 abstracts and included 58 full-text articles. Fourteen articles from 6 trials provided data for the meta-analysis of major health outcomes (Appendix Figure 3, available at Characteristics of Primary Prevention Trials Eight large RCTs that enrolled women without preexisting breast cancer and reported breast cancer outcomes provided most of the data for this review and all of the data for the meta-analysis (Appendix Tables 1 and 2, available at Additional outcomes from these trials include mortality, fractures, thromboembolic events, cardiovascular disease events, uterine abnormalities, cataracts, and other adverse effects. Trials include a head-to-head trial of tamoxifen and raloxifene the Study of Tamoxifen and Raloxifene (STAR) (4, 16); 4 placebo-controlled trials of tamoxifen the International Breast Cancer Intervention Study (IBIS-I) (17, 18), the NSABP P-1 (5 8), the Royal Marsden Hospital trial (19, 20), and the Italian Tamoxifen Prevention Trial (21 24); 2 placebo-controlled trials of raloxifene the Multiple Outcomes of Raloxifene Evaluation (MORE) study with long-term follow-up in the Continuing Outcomes Relevant to Evista (CORE) study (25 39) and the Raloxifene Use for the Heart (RUTH) trial (40, 41); and 1 placebo-controlled trial of tibolone the Long-Term Intervention on Fractures with Tibolone (LIFT) (9). The tamoxifen trials, including STAR, were designed to determine breast cancer incidence as the primary outcome (4, 7, 8, 17 24). Inclusion criteria considered breast cancer risk in all these trials except the Italian trial (22). Two trials, STAR and NSABP P-1, used the modified Gail model to select participants (1, 2). In STAR, women were eligible if they were postmenopausal and had a Gail model predicted 5-year risk for breast cancer of 1.66% or more (4). The NSABP P-1 used this same threshold as well as additional criteria, such as age 60 years or older or a history of lobular carcinoma in situ (8). Most women aged 60 years or older have a Gail model risk of at least 1.66% without additional risk factors because age is a dominant predictor in the model. The IBIS-I and Royal Marsden trials defined eligibility criteria on the basis of numbers of relatives with breast cancer as well as personal history of benign findings on breast biopsy (17, 19). Breast cancer incidence was 1 of 2 primary outcomes in RUTH and a secondary outcome in MORE and LIFT. The latter 2 studies enrolled women with osteoporosis to determine the efficacy of raloxifene or tibolone in preventing fractures (9, 32). Eligibility criteria for both trials included a bone mineral density T-score of 2.5 or less at the femoral neck or lumbar spine or low bone mineral density with preexisting vertebral fractures at baseline. The RUTH trial was designed to determine the efficacy of raloxifene in preventing coronary events and enrolled 17 November 2009 Annals of Internal Medicine Volume 151 Number

4 Review Comparative Effectiveness of Medications to Reduce Breast Cancer Risk women with coronary heart disease or multiple risk factors for coronary heart disease (40). Differences in trial designs led to the enrollment of dissimilar groups of women into the trials. The mean age of participants at entry ranged from 47 years (19) to 51 years (22, 42) in the tamoxifen trials and from 67 years (28) to 68 years (9, 40) in the raloxifene and tibolone trials. Mean age in the STAR trial was 59 years (4). Risks for most outcomes measured in these trials increase with age, including risks for such adverse events as thromboembolic events and strokes. The 15- to 20-year age difference between participants in different trials would be expected to influence results and limit comparisons between medications. Trials also varied by follow-up times. For placebocontrolled trials of tamoxifen, median treatment duration was approximately 4 years (8, 23). In the MORE trial, results were reported after 3 and 4 years of treatment (25 33, 35, 38). Results of CORE, a continuation study of MORE (43), were reported for 4-year and combined 8-year outcomes (MORE and CORE) (36, 37, 39). Median treatment durations were 2.8 years in LIFT (9), 3.1 to 3.2 years in STAR (44), and 5.1 years in RUTH (40). Although most trials reported similar main outcomes, the ascertainment of outcomes varied by trial. The diagnostic criteria for several outcomes were not well described in the trials, and differences in results between trials for some of these outcomes may be due, at least in part, to how the outcomes were determined and measured. Quality and Applicability of Primary Prevention Trials All the primary prevention trials met criteria for fair or good quality for major outcomes. The most important methodological limitation was the inclusion of women using estrogen in the Italian (14% of women), Royal Marsden (15% to 27%), and IBIS-I (40%) tamoxifen trials. Estrogen use could modify or confound breast cancer risk and other outcomes, such as thromboembolic events (45, 46). Trials generally met criteria for good applicability, except for the Italian trial, which exclusively enrolled women who had previously undergone hysterectomy (22). Women with oophorectomies may be at lower-than-average risk for breast cancer (47, 48). In addition, most participants in all trials were white, and none of the trials provided outcomes specific to racial or ethnic groups. The trials were multicenter, relevant to primary care, and large ranging from the Royal Marsden trial (19), which enrolled 2471 women, to STAR (4), which enrolled Participants were recruited from clinics and communities located in many countries, with North America, Europe, and the United Kingdom most represented. Comparative Effectiveness of Tamoxifen, Raloxifene, and Tibolone to Reduce Risk for Breast Cancer, Death, and Fractures In placebo-controlled trials, tamoxifen (risk ratio, 0.70 [95% CI, 0.59 to 0.82]; 4 trials) (7, 18, 20, 23), raloxifene (risk ratio, 0.44 [CI, 0.27 to 0.71]; 2 trials) (40, 43), and Table. Results of Primary Prevention Trials Major Health Outcome Head-to-Head Trial* Placebo-Controlled Trials Raloxifene vs. Tamoxifen Tamoxifen vs. Placebo Risk Ratio (95% CI) Events Reduced or Increased (95% CI), n Risk Ratio (95% CI) [Trials, n ] Events Reduced or Increased (95% CI), n Strength of Evidence Benefits Invasive breast cancer 1.02 ( ) 0.70 ( ) [4] 7 (4 12) High Estrogen receptor positive 0.93 ( ) 0.58 ( ) [4] 8 (3 13) High invasive breast cancer Estrogen receptor negative 1.15 ( ) 1.19 ( ) [4] Moderate invasive breast cancer Noninvasive breast cancer 1.40 ( ) 0.85 ( ) [4] Low All-cause mortality 0.94 ( ) 1.07 ( ) [4] High Vertebral fracture 0.98 ( ) 0.75 ( ) Low Nonvertebral fracture Insufficient data 0.66 ( ) 3 (0.2 5) Moderate Harms Thromboembolic events 0.70 ( ) 6 (2 10); more with tamoxifen 1.93 ( ) [4] 4 (2 9) High Coronary heart disease events 1.10 ( ) 1.00 ( ) [4] High Stroke 0.96 ( ) 1.36 ( ) [4] Moderate Endometrial cancer 0.62 ( ) 2.13 ( ) [3] 4 (1 10) High Cataracts 0.79 ( ) 13 (5 21); more with tamoxifen 1.25 ( ) [3] Low * Study of Tamoxifen and Raloxifene (STAR), 2006 (4). Long-Term Intervention on Fractures with Tibolone (LIFT), 2008 (9). If meta-analysis. Number of events reduced for benefits or increased for harms compared with placebo or other comparator per 1000 women-years, assuming 5 years of use. High consistent results from many ( 5) or large definitive trials; moderate some evidence (3 5 trials) suggests an effect, but results could be altered by future research; low few ( 2) trials exist or existing trials have inconsistent results; insufficient data inadequate or not reported. National Surgical Adjuvant Breast and Bowel Project P-1 (NSABP P-1), 2005 (7) November 2009 Annals of Internal Medicine Volume 151 Number 10

5 Comparative Effectiveness of Medications to Reduce Breast Cancer Risk Review tibolone (risk ratio, 0.32 [CI, 0.13 to 0.80]; 1 trial) (9) reduced the incidence of invasive breast cancer in middleaged and older women by approximately 30% to 68% (Table and Figure 1). Tamoxifen and raloxifene had similar effects in the STAR head-to-head trial (4). Reduction of invasive breast cancer continued at least 3 to 5 years after discontinuation of tamoxifen therapy in the 2 trials that provided posttreatment follow-up data (Figure 1) (18, 20). Rates of invasive cancer were low and risk was not reduced in the Italian trial compared with the other trials, possibly reflecting the underlying lower risk among the women enrolled in this trial (23). In placebo-controlled trials, tamoxifen (7, 18, 20, 23) and raloxifene (40, 43) reduced estrogen receptor positive but not estrogen receptor negative invasive breast cancer and had similar effects in the STAR head-to-head trial (4) (Table). Tamoxifen (7, 18, 20, 23) and raloxifene (40, 43) did not reduce noninvasive breast cancer, including ductal carcinoma in situ, in combined estimates (Figure 2). However, tamoxifen reduced noninvasive breast cancer in the NSABP P-1 trial (risk ratio, 0.63 [CI, 0.45 to 0.89]) (7). The STAR head-to-head trial suggested a higher incidence of noninvasive breast cancer for raloxifene compared with tamoxifen (risk ratio, 1.40 [CI, 0.98 to 2.00]) (4). All-cause mortality was similar for women using raloxifene compared with those using tamoxifen in the STAR trial (4) and for those receiving tamoxifen (7, 18, 20, 23), raloxifene (40, 43), and tibolone (9) compared with those in the placebo group (Table). Tamoxifen did not reduce breast cancer mortality compared with placebo (risk ratio, 1.07 [CI, 0.66 to 1.74]; 4 trials) (7, 18, 20, 23). Trials of raloxifene and tibolone did not report breast cancer mortality. In placebo-controlled trials, raloxifene (29, 40) and tibolone (9) reduced vertebral fractures (Table), tamoxifen (7) and tibolone (9) reduced nonvertebral fractures (Table), and tibolone reduced wrist but not hip fractures (9). Tamoxifen and raloxifene had similar effects on fractures at multiple sites in the STAR head-to-head trial (4). In LIFT, tibolone reduced risk for colon cancer (risk ratio, 0.31 [CI, 0.10 to 0.96]) (9). Adverse Effects of Tamoxifen, Raloxifene, and Tibolone Tamoxifen (risk ratio, 1.93 [CI, 1.41 to 2.64]; 4 trials) (8, 18, 20, 21) and raloxifene (risk ratio, 1.60 [CI, 1.15 to 2.23]; 2 trials) (33, 40) caused more thromboembolic events than placebo (Table and Figure 3). Risk returned to normal after discontinuation of tamoxifen therapy in the 2 trials providing posttreatment data (Figure 3) (18, 20). Raloxifene caused fewer thromboembolic events than tamoxifen in the STAR head-to-head trial (4). Tibolone did not increase risk for thromboembolic events in the LIFT trial, although data are limited (9). Tamoxifen (7, 18, 20, 23), raloxifene (26, 40), and tibolone (9) did not increase risk for coronary heart disease events in placebo-controlled trials (Table), although data for tibolone are limited. Most trials used composite measures that included myocardial infarction, the acute coronary syndrome, and severe angina (12). Tibolone caused Table Continued Placebo-Controlled Trials (continued) Raloxifene vs. Placebo Tibolone vs. Placebo Risk Ratio (95% CI) [Trials, n ] Events Reduced or Increased (95% CI), n Strength of Evidence Risk Ratio (95% CI) Events Reduced or Increased (95% CI), n Strength of Evidence 0.44 ( ) [2] 9 (4 14) High 0.32 ( ) 10 (3 17) Moderate 0.33 ( ) [2] 8 (4 12) High Insufficient 1.25 ( ) [2] Moderate Insufficient 1.47 ( ) [2] Moderate Insufficient 0.91 ( ) [2] High Insufficient 0.61 ( ) [2] 7 (5 9) High 0.55 ( ) 44 (25 61) Moderate 0.97 ( ) [2] High 0.74 ( ) 34 (8 56) Moderate 1.60 ( ) [2] 7 (2 15) High 0.57 ( ) Low 0.95 ( ) [2] High 1.37 ( ) Low 0.96 ( ) [2] Moderate 2.19 ( ) 11 (1 36) Moderate 1.14 ( ) [2] Moderate Insufficient 0.93 ( ) [2] High Insufficient 17 November 2009 Annals of Internal Medicine Volume 151 Number

6 Review Comparative Effectiveness of Medications to Reduce Breast Cancer Risk Figure 1. Meta-analysis of risk ratios for incidence of invasive breast cancer. Study, Year (Reference) Participants, n Treatment Placebo Mean or Median Duration, y Intended Treatment Total Follow-up Treatment Patients, n Rate* Placebo Patients, n Rate* Risk Ratio (95% CI) Risk Ratio (95% CI) Tamoxifen NSABP P-1, 2005 (7) ( ) IBIS-I, 2007 (18) ( ) Royal Marsden trial, 2007 (20) ( ) Italian Tamoxifen Prevention Trial, 2007 (23) ( ) Combined (test of heterogeneity: Q = 4.8, I 2 = 37.6%; P = 0.186) 0.70 ( ) Raloxifene MORE and CORE, 2004 (43) or ( ) RUTH, 2006 (40) ( ) Combined (test of heterogeneity: Q = 3.0, I 2 = 66.7%; P = 0.084) 0.44 ( ) Tibolone LIFT, 2008 (9) ( ) By Treatment Period Tamoxifen: active treatment IBIS-I, 2007 (18) ( ) Royal Marsden Trial, 2007 (20) ( ) Combined (test of heterogeneity: Q = 0.7, I 2 = 0.0%; P = 0.416) 0.80 ( ) Tamoxifen: after treatment IBIS-I, 2007 (18) ( ) Royal Marsden trial, 2007 (20) ( ) Combined (test of heterogeneity: Q = 0.1, I 2 = 0.0%; P = 0.795) 0.70 ( ) Favors Treatment Favors Control Error bars represent 95% CIs. CORE Continuing Outcomes Relevant to Evista; IBIS-I International Breast Cancer Intervention Study; LIFT Long-Term Intervention on Fractures with Tibolone; MORE Multiple Outcomes of Raloxifene Evaluation; not reported; NSABP P-1 National Surgical Adjuvant Breast and Bowel Project P-1; RUTH Raloxifene Use for the Heart. * Per 1000 woman-years. Italian Tamoxifen Prevention Trial and RUTH reported mean or median duration of actual treatment period. Analysis included data from both MORE and CORE. Participants from MORE had 4-year treatment, and those who continued in CORE had 4 additional years of treatment. Total follow-up time is averaged over both MORE and CORE for 7705 participants. more strokes than placebo in the LIFT trial, and these occurred more often in women older than 70 years (9). Tamoxifen (7, 18, 20, 23) and raloxifene (26, 40) did not increase stroke incidence in placebo-controlled trials (Figure 4). However, in the RUTH trial, women randomly assigned to raloxifene had higher stroke mortality than those assigned to placebo (risk ratio, 1.49 [CI, 1.00 to 2.24]) (40). In placebo-controlled trials, tamoxifen caused more cases of endometrial cancer (risk ratio, 2.13 [CI, 1.36 to 3.32]; 3 trials) (7, 18, 20) (Figure 5). Women using tamoxifen also had more benign gynecologic conditions (18, 49); surgical procedures, including hysterectomy (18, 20, 49); and uterine bleeding (18, 49) than did women using placebo. Raloxifene did not increase risk for endometrial cancer (33, 40) (Figure 5) or uterine bleeding (27, 40, 50 58). In the STAR head-to-head trial, raloxifene caused fewer cases of endometrial hyperplasia (risk ratio, 0.16 [CI, 0.09 to 0.29]) and was associated with fewer hysterecto November 2009 Annals of Internal Medicine Volume 151 Number 10

7 Comparative Effectiveness of Medications to Reduce Breast Cancer Risk Review mies (risk ratio, 0.44 [CI, 0.35 to 0.56]) compared with tamoxifen but did not cause fewer cases of endometrial cancer (risk ratio, 0.62 [CI, 0.35 to 1.08]) (4). In the NSABP P-1 trial (8), women using tamoxifen had more cataract surgeries than those using placebo (risk ratio, 1.57 [CI, 1.16 to 2.14]), although risk for cataracts was not increased in combined estimates (8, 18, 20) (Figure 5). Raloxifene did not increase risk for cataracts (Figure 5) or cataract surgery in placebo-controlled trials (33, 40). Raloxifene caused fewer cataracts (risk ratio, 0.79 [CI, 0.68 to 0.92]) and cataract surgeries (risk ratio, 0.82 [CI, 0.68 to 0.99]) than tamoxifen in the STAR head-to-head trial (4). Most common side effects for tamoxifen are hot flashes and other vasomotor symptoms (8, 18, 20, 23) and vaginal discharge, itching, or dryness (8, 18, 20, 23). For raloxifene, vasomotor symptoms (27, 40, 52, 53, 55) and leg cramps (27, 40, 55) are most common. In STAR, raloxifene users reported more musculoskeletal problems, dyspareunia, and weight gain, whereas tamoxifen users had more gynecologic problems, vasomotor symptoms, leg cramps, and bladder control symptoms (16). Tibolone increases vaginal bleeding (59 61), but in contrast to tamoxifen and raloxifene, it reduces the number and severity of hot flashes (60, 62, 63). Variability of Outcomes in Population Subgroups Tamoxifen and raloxifene had similar effects on breast cancer outcomes regardless of age and family history of breast cancer in the STAR head-to-head trial (4). Tamoxifen reduced breast cancer incidence in several subgroups of women evaluated in placebo-controlled primary prevention trials (12). Results are similar for women aged 50 years or younger (risk ratio, 0.68 [CI, 0.55 to 0.85]; 3 trials) versus those older than 50 years (risk ratio, 0.68 [CI, 0.51 to 0.90]; 3 trials) (7, 18, 23); premenopausal (risk ratio, 0.63 [CI, 0.46 to 0.85]; 2 trials) versus postmenopausal (risk ratio, 0.68 [CI, 0.44 to 1.05]; 2 trials) women (18, 20); estrogen users (risk ratio, 0.75 [CI, 0.47 to 1.20]; 2 trials) versus nonusers (risk ratio, 0.68 [CI, 0.54 to 0.86]; 3 trials) (7, 18, 23); and those with a family history of breast cancer (risk ratio, 0.58 [CI, 0.46 to 0.73]; 1 trial) versus those with no family history (risk ratio, 0.54 [CI, 0.34 to 0.83]; 1 trial) (7). In the NSABP P-1 trial, cancer rates were highest and risk reduction was greatest among women in the highest modified Gail model risk category (5-year risk 5%) and among women with previous atypical hyperplasia (7). Some harms, including thromboembolic events, strokes, and endometrial cancer, were more common in older (aged 50 years) than younger women in the NSABP P-1 trial (7). Figure 2. Meta-analysis of risk ratios for incidence of noninvasive breast cancer. Study, Year (Reference) Participants, n Treatment Placebo Mean or Median Duration, y Intended Treatment Total Follow-up Treatment Patients, n Rate* Placebo Patients, n Rate* Risk Ratio (95% CI) Risk Ratio (95% CI) Tamoxifen NSABP P-1, 2005 (7) ( ) IBIS-I, 2007 (18) ( ) Royal Marsden trial, 2007 (20) ( ) Italian Tamoxifen Prevention Trial, 2007 (23) ( ) Combined (test of heterogeneity: Q = 5.9, I 2 = 48.9%; P = 0.118) 0.85 ( ) Raloxifene MORE and CORE, 2004 (43) or ( ) RUTH, 2006 (40) ( ) Combined (test of heterogeneity: Q = 0.9, I 2 = 0.0%; P = 0.349) 1.47 ( ) Favors Treatment Favors Control Error bars represent 95% CIs. CORE Continuing Outcomes Relevant to Evista; IBIS-I International Breast Cancer Intervention Study; MORE Multiple Outcomes of Raloxifene Evaluation; NSABP P-1 National Surgical Adjuvant Breast and Bowel Project P-1; RUTH Raloxifene Use for the Heart. * Per 1000 woman-years. Italian Tamoxifen Prevention Trial and RUTH reported mean or median duration of actual treatment period. Analysis included data from both MORE and CORE. Participants from MORE had 4-year treatment, and those who continued in CORE had 4 additional years of treatment. Total follow-up time is averaged over both MORE and CORE for 7705 participants November 2009 Annals of Internal Medicine Volume 151 Number

8 Review Comparative Effectiveness of Medications to Reduce Breast Cancer Risk Figure 3. Meta-analysis of risk ratios for incidence of venous thromboembolism. Study, Year (Reference) Participants, n Treatment Placebo Mean or Median Duration, y Treatment Patients, n Rate* Placebo Patients, n Rate* Risk Ratio (95% CI) Risk Ratio (95% CI) Tamoxifen: active treatment NSABP P-1, 1998 (8) ( ) Italian Tamoxifen Prevention Trial, 2005 (21) ( ) IBIS-I, 2007 (18) ( ) Royal Marsden trial, 2007 (20) ( ) Combined (test of heterogeneity: Q = 2.0, I 2 = 0.0%; P = 0.565) 1.93 ( ) Tamoxifen: after treatment IBIS-I, 2007 (18) ( ) Royal Marsden trial, 2007 (20) ( ) Combined (test of heterogeneity: Q = 0.2, I 2 = 0.0%; P = 0.688) 1.02 ( ) Raloxifene MORE, 2004 (33) ( ) RUTH, 2006 (40) ( ) Combined (test of heterogeneity: Q = 1.3, I 2 = 22.3%; P = 0.257) 1.60 ( ) Tibolone LIFT, 2008 (9) ( ) Favors Treatment Favors Control Error bars represent 95% CIs. IBIS-I International Breast Cancer Intervention Study; LIFT Long-Term Intervention on Fractures with Tibolone; MORE Multiple Outcomes of Raloxifene Evaluation; not reported; NSABP P-1 National Surgical Adjuvant Breast and Bowel Project P-1; RUTH Raloxifene Use for the Heart. * Per 1000 woman-years. For tamoxifen trials, venous thromboembolic events include deep venous thrombosis and pulmonary embolism only. For other trials, additional thromboembolic events may be included. Events were reported from at least 3 months after treatment was stopped until the end of follow-up. Results of breast cancer incidence for raloxifene are similar among women who had previously used estrogen (risk ratio, 0.45 [CI, 0.19 to 1.07]; 2 trials) versus those who had not (risk ratio, 0.44 [CI, 0.30 to 0.65]; 2 trials) (36, 41) and among those with a body mass index of 25 mg/kg 2 or less (risk ratio, 0.47 [CI, 0.17 to 1.33]; 2 trials) versus those with an index greater than 25 mg/kg 2 (risk ratio, 0.43 [CI, 0.30 to 0.63]; 2 trials) (36, 41). Results also do not vary by age (36, 41), age at menarche (41), parity (41), and age at first live birth (41). Estimates from subgroups based on family history of breast cancer and previous hysterectomy or oophorectomy are limited by smaller numbers of participants (36, 41). DISCUSSION In middle-aged and older women without preexisting breast cancer, tamoxifen, raloxifene, and tibolone reduce the risk for invasive breast cancer by 30% to 68%. Effects were similar for tamoxifen and raloxifene in the STAR head-tohead trial. Reduction of invasive breast cancer continued after discontinuation of tamoxifen therapy in the 2 trials that provided follow-up data. Tamoxifen and raloxifene reduced estrogen receptor positive but not estrogen receptor negative invasive breast cancer and had similar effects on these subtypes when directly compared. Tamoxifen reduced noninvasive breast cancer in the NSABP P-1 trial but not in the other tamoxifen trials. Raloxifene did not decrease noninvasive cancer, and the STAR trial suggested that more women using raloxifene had noninvasive breast cancer than those using tamoxifen. Tamoxifen and raloxifene reduced the incidence of invasive breast cancer for all population subgroups evaluated; these subgroups have not been studied for tibolone. On the basis of our meta-analysis of placebocontrolled primary prevention trials, the number of cases of November 2009 Annals of Internal Medicine Volume 151 Number 10

9 Comparative Effectiveness of Medications to Reduce Breast Cancer Risk Review Figure 4. Meta-analysis of risk ratios for incidence of stroke. Study, Y ear (Reference) Participants, n T reatment Placebo Mean or Median Duration, y T reatment Patients, n Rate* Placebo Patients, n Rate* Risk Ratio (95% CI ) Risk Ratio (95% CI ) T amoxifen: active treatment NSABP P-1, 1998 (8 ) ( ) IBIS-I, 2007 (18) ( ) Royal Marsden trial, 2007 (20) ( ) Italian T amoxifen Prevention T rial, 2007 (23) ( ) Combined (test of heterogeneity: Q = 2.9, I 2 = 0.0%; P = 0.400) 1.36 ( ) T amoxifen: after treatment IBIS-I, 2007 (18) ( ) Royal Marsden trial, 2007 (20) ( ) Combined (test of heterogeneity: Q = 1.9, I 2 = 46.8%; P = 0.170) 0.83 ( ) Raloxifene MORE, 2002 (26) ( ) RUTH, 2006 (40) ( ) Combined (test of heterogeneity: Q = 2.5, I 2 = 59.3%; P = 0.117) 0.96 ( ) T ibolone LIFT, 2008 (9) ( ) Favor s Tr eatment Favor s Contro l Error bars represent 95% CIs. IBIS-I International Breast Cancer Intervention Study; LIFT Long-Term Intervention on Fractures with Tibolone; MORE Multiple Outcomes of Raloxifene Evaluation; not reported; NSABP P-1 National Surgical Adjuvant Breast and Bowel Project P-1; RUTH Raloxifene Use for the Heart. * Per 1000 woman-years. Events were reported from at least 3 months after treatment was stopped until the end of follow-up. invasive breast cancer reduced is approximately 7 to 10 per 1000 women per year, assuming 5 years of use. Conclusions about the comparative effectiveness of medications are limited by the heterogeneity of participants in the trials and the existence of only 1 head-to-head trial. Trials demonstrate risk reduction not only for breast cancer but also for fractures, providing additional benefits for women. In the United States, the current choices of medications are raloxifene for postmenopausal women and tamoxifen for premenopausal and postmenopausal women. These medications also increase risks for serious and potentially life-threatening adverse events. Thromboembolic events are the most common serious complications, more so with tamoxifen than with raloxifene in the STAR trial. Risk was increased by 60% to 90% in the placebocontrolled primary prevention trials that enrolled women with no history of thromboembolic events (4 to 7 per 1000 women per year). Clinicians considering these medications need to assess history and risk factors for thromboembolic events in treatment candidates. The effects of tamoxifen on endometrial cancer (4 per 1000 women per year), endometrial hyperplasia, and hysterectomy are also important. These problems could be avoided if use of the drug were limited to women who have had a hysterectomy. However, because tamoxifen is the only medication approved for use in premenopausal women, close monitoring of adverse uterine effects would be required for some users. Raloxifene and tamoxifen also cause adverse effects that could affect quality of life, such as hot flashes, vaginal symptoms, and musculoskeletal symptoms. Although tibolone is not currently available in the United States, its effect on stroke raises concern. This review is limited by potential biases and evidence gaps. These include publication bias and biases resulting from our selection criteria, such as using English-only reports. Trials may not have been truly blinded because side effects of active medications may have led to differential ascertainment of outcomes. Active surveillance ended with completion of therapy in most trials, and important long November 2009 Annals of Internal Medicine Volume 151 Number

10 Review Comparative Effectiveness of Medications to Reduce Breast Cancer Risk Figure 5. Meta-analysis of risk ratios for incidence of endometrial cancer and cataracts. Study, Year (Reference) Participants, n Treatment Placebo Mean or Median Duration, y Treatment Patients, n Rate* Placebo Patients, n Rate* Risk Ratio (95% CI) Risk Ratio (95% CI) Endometrial Cancer Tamoxifen NSABP P-1, 1998 (8) ( ) IBIS-I, 2007 (18) ( ) Royal Marsden trial, 2007 (20) ( ) Combined (test of heterogeneity: Q = 1.2, I 2 = 0.0%; P = 0.551) 2.13 ( ) Raloxifene MORE, 2004 (33) ( ) RUTH, 2006 (40) ( ) Combined (test of heterogeneity: Q = 0.2, I 2 = 0.0%; P = 0.630) 1.14 ( ) Tibolone LIFT, 2008 (9) Cataracts Tamoxifen NSABP P-1, 1998 (8) ( ) IBIS-I, 2007 (18) ( ) Royal Marsden trial, 2007 (20) ( ) Combined (test of heterogeneity: Q = 3.9, I 2 = 48.5%; P = 0.144) 1.25 ( ) Raloxifene MORE, 2004 (33) ( ) RUTH, 2006 (40) ( ) Combined (test of heterogeneity: Q = 0.3, I 2 = 0.0%; P = 0.557) 0.93 ( ) Favors Treatment Favors Control Error bars represent 95% CIs. IBIS-I International Breast Cancer Intervention Study; LIFT Long-Term Intervention on Fractures with Tibolone; MORE Multiple Outcomes of Raloxifene Evaluation; not reported; NSABP P-1 National Surgical Adjuvant Breast and Bowel Project P-1. * Per 1000 woman-years. Rates were based on number of women with an intact uterus. The rate and risk ratio were recalculated on the basis of the number of women at risk (having an intact uterus). The values reported by the study were based on all randomly assigned participants. The number of women at risk (having an intact uterus) was not reported, and the risk ratio is calculated on the basis of the number of randomly assigned participants at baseline. term outcomes may have been underreported. For some tamoxifen trials, participants randomly assigned to placebo switched to active medications after closure of the trial, compromising long-term tracking of outcomes. All efficacy trials were powered to detect statistical differences in breast cancer incidence, not adverse outcomes. Risks for some adverse outcomes may be underestimated because of lack of statistical power. Underestimation of adverse outcomes may also relate to inadequate ascertainment. For example, rates of cataracts and cataract surgery in the NSABP P-1 trial are substantially higher than rates in the other trials, most likely because the trial enlisted a more aggressive detection method (8). Evidence gaps include determination of optimal doses, duration and timing of use, persistence of effects after treatment, and outcomes in population subgroups. Data are lacking for nonwhite women, premenopausal women, and women who have comorbid conditions or are taking additional medications for other indications. Follow-up of women enrolled in existing trials would provide needed data on long-term outcomes. Additional evaluation of tibolone is necessary to determine important clinical out November 2009 Annals of Internal Medicine Volume 151 Number 10

11 Comparative Effectiveness of Medications to Reduce Breast Cancer Risk Review comes, particularly regarding its safety. Trials of other emerging medications to reduce breast cancer risk, such as aromatase inhibitors and retinoids, will be needed as these are developed. Controlled trials of lifestyle modification interventions to reduce risk for breast cancer, such as weight loss and exercise, should also be explored. Clinical applications require caution. Women enrolled in clinical trials tend to be healthier than the general population and probably have fewer adverse outcomes. The heterogeneity of study populations must be considered in interpreting results of the trials. Results may best apply to patients with characteristics similar to those of the study participants. In general, tamoxifen results apply to younger pre- and postmenopausal women meeting breast cancer risk criteria; tibolone results, to older postmenopausal women with osteoporosis; and raloxifene results, to postmenopausal women meeting breast cancer risk criteria and to older postmenopausal women with osteoporosis, coronary heart disease, or risk factors for coronary heart disease. Before applying these findings to practice, clinicians must ensure that women understand their individual risks for breast cancer and can favorably balance these with the unwanted effects of risk-reducing medications. From Oregon Health & Science University, Veterans Affairs Medical Center, and the Women and Children s Health Research Center, Providence Health & Services, Portland, Oregon. Disclaimer: The findings and conclusions in this document are those of the authors, who are responsible for its content, and do not necessarily represent the views of the Agency for Healthcare Research and Quality. No statement in this report should be construed as an official position of the Agency or of the U.S. Department of Health and Human Services. Acknowledgment: The authors thank Andrew Hamilton, MLS, MS, and Rose Campbell, MLIS, MS, for literature searches and Jennifer Nguyen for administrative assistance at the Oregon Evidence-based Practice Center at the Oregon Health & Science University. They also acknowledge the contributions of Agency for Healthcare Research and Quality Officers Shilpa Amin, MD, MBsc, and Kenneth Lin, MD, and members of the Technical Expert Panel and expert reviewers. Grant Support: This manuscript is based on research conducted by the Oregon Evidence-based Practice Center under contract to the Agency for Healthcare Research and Quality, Rockville, Maryland (contract ). Potential Conflicts of Interest: None disclosed. Requests for Single Reprints: Heidi Nelson, MD, MPH, Oregon Evidence-based Practice Center, Oregon Health & Science University, Mailcode BICC, 3181 SW Sam Jackson Park Road, Portland, OR ; , nelsonh@ohsu.edu. Current author addresses and author contributions are available at References 1. Gail MH. The estimation and use of absolute risk for weighing the risks and benefits of selective estrogen receptor modulators for preventing breast cancer. Ann N Y Acad Sci. 2001;949: [PMID: ] 2. Gail MH, Brinton LA, Byar DP, Corle DK, Green SB, Schairer C, et al. Projecting individualized probabilities of developing breast cancer for white females who are being examined annually. J Natl Cancer Inst. 1989;81: [PMID: ] 3. U.S. Food and Drug Administration Center for Drug Evaluation and Research. FDA approved drug products. Accessed at /cder/drugsatfda/index.cfm?fuseaction Search.Label_ApprovalHistory on 18 August Vogel VG, Costantino JP, Wickerham DL, Cronin WM, Cecchini RS, Atkins JN, et al; National Surgical Adjuvant Breast and Bowel Project (NSABP). 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. JAMA. 2006;295: [PMID: ] 5. Day R, Ganz PA, Costantino JP. Tamoxifen and depression: more evidence from the National Surgical Adjuvant Breast and Bowel Project s Breast Cancer Prevention (P-1) Randomized Study. J Natl Cancer Inst. 2001;93: [PMID: ] 6. Day R; National Surgical Adjuvant Breast and Bowel Project P-1 study (NSABP P-1). Quality of life and tamoxifen in a breast cancer prevention trial: a summary of findings from the NSABP P-1 study. National Surgical Adjuvant Breast and Bowel Project. Ann N Y Acad Sci. 2001;949: [PMID: ] 7. Fisher B, Costantino JP, Wickerham DL, Cecchini RS, Cronin WM, Robidoux A, et al. Tamoxifen for the prevention of breast cancer: current status of the National Surgical Adjuvant Breast and Bowel Project P-1 study. J Natl Cancer Inst. 2005;97(22): Fisher B, Costantino JP, Wickerham DL, Redmond CK, Kavanah M, Cronin WM, et al. Tamoxifen for prevention of breast cancer: report of the National Surgical Adjuvant Breast and Bowel Project P-1 Study. J Natl Cancer Inst. 1998; 90: [PMID: ] 9. Cummings SR, Ettinger B, Delmas PD, Kenemans P, Stathopoulos V, Verweij P, et al; LIFT Trial Investigators. The effects of tibolone in older postmenopausal women. N Engl J Med. 2008;359: [PMID: ] 10. Medicines and Healthcare Products Regulatory Agency. UK Public Assessment Report Tibolone (Livial): benefit-risk evaluation Accessed at on 18 August Agency for Healthcare Research and Quality. Methods Reference Guide for Effectiveness and Comparative Effectiveness Reviews, version 1.0. Rockville, MD; Agency for Healthcare Research and Quality. Draft posted October Accessed at /repfiles/2007_10draftmethodsguide.pdf on 18 August Nelson HD, Fu R, Humphrey L, Smith ME, Griffin J, Nygren P. Comparative Effectiveness of Medications to Reduce Risk of Primary Breast Cancer in Women. Comparative Effectiveness Review. (Prepared by the Oregon Health & Science University Evidence-based Practice Center under contract no ). Rockville, MD: Agency for Healthcare Research and Quality; Harris RP, Helfand M, Woolf SH, Lohr KN, Mulrow CD, Teutsch SM, et al; Methods Work Group, Third US Preventive Services Task Force. Current methods of the US Preventive Services Task Force: a review of the process. Am J Prev Med. 2001;20: [PMID: ] 14. Higgins JP, Thompson SG. Quantifying heterogeneity in a meta-analysis. Stat Med. 2002;21: [PMID: ] 15. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials. 1986;7: [PMID: ] 16. Land SR, Wickerham DL, Costantino JP, Ritter MW, Vogel VG, Lee M, et al. Patient-reported symptoms and quality of life during treatment with tamoxifen or raloxifene for breast cancer prevention: the NSABP Study of Tamoxifen and Raloxifene (STAR) P-2 trial. JAMA. 2006;295: [PMID: ] 17. Cuzick J, Forbes J, Edwards R, Baum M, Cawthorn S, Coates A, et al; IBIS investigators. First results from the International Breast Cancer Intervention Study (IBIS-I): a randomised prevention trial. Lancet. 2002;360: [PMID: ] 18. Cuzick J, Forbes JF, Sestak I, Cawthorn S, Hamed H, Holli K, et al; International Breast Cancer Intervention Study I Investigators. Long-term re November 2009 Annals of Internal Medicine Volume 151 Number