Interpreting breast international group (BIG) 1-98: a randomized, double-blind, phase III trial comparing letrozole and tamoxifen as adjuvant endocrine therapy for postmenopausal women with hormone receptor-positive, early breast cancer

The Breast International Group (BIG) 1-98 study is a four-arm trial comparing 5 years of monotherapy with tamoxifen or with letrozole or with sequences of 2 years of one followed by 3 years of the other for postmenopausal women with endocrine-responsive early invasive breast cancer. From 1998 to 2003, BIG -98 enrolled 8,010 women. The enhanced design f the trial enabled two complementary analyses of efficacy and safety. Collection of tumor specimens further enabled treatment comparisons based on tumor biology. Reports of BIG 1-98 should be interpreted in relation to each individual patient as she weighs the costs and benefits of available treatments. Clinicaltrials.gov ID: NCT00004205.


Introduction
Reports of large trials of breast cancer confi rm the value of aromatase inhibitors as adjuvant systemic therapy for postmenopausal women with endocrine-responsive early breast cancer [1][2][3][4][5][6][7][8][9]. Th e inclusion of an aromatase inhibitor in the adjuvant treatment program for this population has been recommended by both the American Society of Clinical Oncology and St. Gallen guidelines [10,11]. Studies have shown that 5 years of adjuvant therapy with an aromatase inhibitor alone improved disease-free survival (DFS) and time to distant recurrence (TDR) in comparison with 5 years of tamoxifen in this population [1][2][3]12], and recently the Breast International Group (BIG) 1-98 trial showed improved overall survival (OS) with the aromatase inhibitor letrozole [13]. Other studies have shown that switching to an aromatase inhibitor after 2 years of tamoxifen improves outcome [4][5][6][7][8]. Results were confi rmed in an overview analysis [9].
Th e BIG 1-98 study is a double-blind, four-arm trial comparing 5 years of monotherapy with tamoxifen or with letrozole or with sequences of 2 years of one of these agents followed by 3 years of the other ( Figure 1). Centers participated in one of two randomization options (twoarm or four-arm). Between 1998 and2003, 8,010 patients were enrolled. Th e trial is designed to answer two questions concerning how best to use endocrine agents for the treatment of early breast cancer in post menopausal women with hormone receptor-positive tumors, the fi rst to compare letrozole monotherapy with tamoxifen monotherapy and the second to determine the benefi t of letrozole in sequence with tamoxifen. Table 1 presents a summary of the study subpopulations contributing to various data analyses of effi cacy questions. In BIG 1-98, the primary endpoint is DFS, defi ned as the time from random assignment to the earliest time of invasive recurrence in local, regional, or distant sites; a new invasive breast cancer in the contralateral breast; any second (non-breast) malignancy; or death from any cause. Secondary outcomes are OS, TDR, and safety [14,15].

Abstract
The Breast International Group (BIG) 1-98 study is a four-arm trial comparing 5 years of monotherapy with tamoxifen or with letrozole or with sequences of 2 years of one followed by 3 years of the other for postmenopausal women with endocrine-responsive early invasive breast cancer. From 1998 to 2003, BIG 1-98 enrolled 8,010 women. The enhanced design of the trial enabled two complementary analyses of effi cacy and safety. Collection of tumor specimens further enabled treatment comparisons based on tumor biology. Reports of BIG 1-98 should be interpreted in relation to each individual patient as she weighs the costs and benefi ts of available treatments. ClinicalTrials.gov ID: NCT00004205. Patients were seen for follow-up at clinic visits every 6 months during treatment to gather general safety data, to document predefi ned toxicity data, and to receive a new supply of study medications. Survival, disease status, and cardiac, bone, and endometrial adverse events (AEs) were reported every 6 months for 5 years from random assignment and were followed by yearly reports after 5 years.
Various aspects of the results of BIG 1-98 have been published in separate papers in the past 5 years. Th e purpose of the present review is to bring the various results together in one place to give an overall inter pretation of their signifi cance for science and patient care.

Letrozole versus tamoxifen
Th e fi rst report of the results of the BIG 1-98 trial was based on the primary core analysis, which was published in 2005 [1]. Th is analysis included all 8,010 patients but did not include any events after the fi rst 2 years (the time of the switch) for patients in the two sequential arms. Th e results of the primary core analysis showed that letrozole improved DFS and TDR in comparison with tamoxifen alone. After a median follow-up of 25.8 months, 5-year DFS estimates were 84.0% and 81.4%, respectively. As compared with tamoxifen, letrozole signifi cantly reduced the risk of a DFS event (hazard ratio (HR) = 0.81, 95% confi dence interval (CI) 0.70 to 0.93; P = 0.003) and the risk of distant recurrence (HR = 0.73, 95% CI 0.60 to 0.88; P = 0.001).
For a report early in follow-up (when most of the observed events occurred within the fi rst 2 years), the primary core analysis was appropriate, but for future updates, which are infl uenced substantially by events beyond 2 years, only those patients randomly assigned to the monotherapy arms (4,922 patients from both twoarm and four-arm options) are included in the head-tohead comparison of 5 years of letrozole versus tamoxifen [2,13,14]. Th e fi rst results of the trial restricted to patients in the two monotherapy groups were published in 2007 [2]. At a median follow-up time of 51 months, results confi rmed the earlier report, showing improved DFS (5-year DFS rates of 84.0% and 81.1% for letrozole and The design allows both head-to-head comparison of letrozole monotherapy versus tamoxifen monotherapy (4,922 patients) and assessment of the role of sequential endocrine treatments (6,182 patients

Number of Population patients Comments
Intention-to-treat population 8,010 Eighteen randomly assigned patients withdrew consent to participate before starting treatment. Two-arm randomization option 1,828 Four-arm randomization option 6,182 Primary core analysis population 8,010 Follow-up for two sequential treatment groups is censored at 2 years (time of treatment switch).

Monotherapy population 4,922
Patients were randomly assigned to receive 5 years of tamoxifen alone or letrozole alone. Two-arm randomization option 1,828 Four-arm randomization option 3,094 Sequential treatment analysis population 6,182 Patients in the intention-to-treat population enrolled in the four-arm randomization option.

Selective crossover on the tamoxifen monotherapy arm
In 2005, the presentation of the fi nding that letrozole signifi cantly reduced both distant recurrences and DFS events compared with tamoxifen led to the decision to inform patients in the tamoxifen monotherapy group of their treatment assignment. 'Selective crossover' in this trial refers to the act of choosing to switch from tamoxifen monotherapy to letrozole prior to completion of 5 years of tamoxifen, after the results were known in 2005. While selective crossover is often in the best interest of patients, it complicates later trial analyses with further patient follow-up because the randomized, blinded trial design is compromised. Th rough a protocol amendment, women who were randomly assigned to tamoxifen alone and who were disease-free could receive letrozole. Of 2,459 patients in the tamoxifen-alone treatment arm, 619 (25.2%) selectively crossed over to letrozole. Th e crossing over of 25% of patients from the less eff ective to the more eff ective regimen called into question whether the intention-to-treat (ITT) analysis, which ignores the crossover, will give the most clinically relevant updated estimate of the relative treatment eff ect. Selective crossover represents a special case of nonadherence to a randomized treatment following the report of positive trial results. Specifi cally, control group patients are off ered and accept the opportunity to cross over to the experimental treatment. Selective crossover is distinct from a protocol-defi ned treatment switch or ad hoc non-adherence to a random assignment because the crossover is motivated by randomized evidence that the experimental treatment is more eff ective. Th us, the ITT control group subject to selective crossover is likely to have a better outcome than if the control group had continued to receive the randomized treatment. Selective crossover disturbs the randomized comparison in updated analyses performed subsequent to fi rst results and therefore particularly impacts the OS results, which require longer follow-up.
With selective crossover and further follow-up, the treatment received by patients in the tamoxifen control arm in BIG 1-98 becomes a mix of the control and experimental treatments. With the potential for selection bias among the patients who decide to cross over, the benefi ts of random assignment erode and the trial becomes a hybrid of a randomized trial and observational study. For a randomized clinical trial, the ITT analysis is the 'gold standard' analytic approach designed to control bias. However, external evidence from a meta-analysis of similar trials [9] shows that the estimated outcomes of the tamoxifen arm in the ITT analysis are likely to be better than if the trial had not been amended and all patients randomly assigned to tamoxifen continued to receive the drug for 5 years. As the selective crossover transforms the randomized trial toward an observational study, established modeling methods to address issues of bias due to treatment selection are needed. For the BIG 1-98 study, the methodology called inverse probability of censoring weighted (IPCW) analysis is used to clarify the clinical benefi t of letrozole compared with tamoxifen [13]. Th e IPCW analysis estimates the clinical benefi t of letrozole versus tamoxifen which would have been observed had there been no selective crossover in the trial. An editorial accompanying the manuscript confi rmed the appropriateness of this method and discussed the interpretation of the IPCW results in the trial specifi cally and cancer trials more broadly [16].

Monotherapy comparison using inverse probability of censoring weighted analysis
IPCW is a well-established approach in randomized and observational studies to overcome selection biases. Th e method is one of several approaches to causal inference and analyzes only the available data under conditions of 'informative' missing data. Specifi cally, we employed IPCW Cox modeling [17]. In the setting of selective cross over, IPCW modeling artifi cially creates a scenario of 'missing' follow-up data by censoring the follow-up of each woman at the time she crossed over ('informative' censoring). However, the truncated follow-up is 'recreated' by applying weighting to the follow-up of women who have similar demographic and disease characteristics and who did not cross over. In this way, the follow-up of women who remain on tamoxifen accounts not only for themselves in the analysis but also for comparable women whose experience remaining on tamoxifen can not be observed, because they selectively changed treatments.
In the IPCW monotherapy analysis, we analyzed the data set of the protocol-specifi ed update at 10 years after trial initiation [12] and took into consideration the selective crossover. After follow-up was truncated, the median follow-up was reduced to 74 months (IPCW) from 76 months (ITT).
Weighted Cox models, using IPCW, estimated a statistically signifi cant, 18% reduction in the hazard of an OS event with letrozole treatment (HR = 0.82, 95% CI 0.70 to 0.95). IPCW estimates of 5-year OS were 91.8% and 90.4% for letrozole and tamoxifen, respectively. Th e IPCW HRs of DFS and TDR events were 0.83 (95% CI 0.74 to 0.94) and 0.80 (95% CI 0.67 to 0.94), respectively ( Figure 2). Th e results show that adjuvant treatment with letrozole, compared with tamoxifen, signifi cantly reduces the risk of death, the risk of recurrent disease, and the risk of recurrence at distant sites in postmenopausal women with hormone receptor-positive breast cancer. Th e OS results were consistent in most subgroups ( Figure 2).

Sequential treatment comparisons
Th e analysis of the sequential treatments was limited to the 6,182 patients entered into the four-arm option ( Figure 1). Th e sequential treatment analysis was pub lished in 2009 [12], with emphasis on comparing the two letrozole-containing sequential regimens with letrozole monotherapy from the time of random assignment; the three letrozole-containing regimens had remained blinded. Th e tamoxifen monotherapy regimen was already shown to be less eff ective than letrozole, had been unblinded, and therefore was not included in the effi cacy comparisons. Tamoxifen was, however, included in AE comparisons with the other three regimens.
At a median follow-up of 71 months from random assign ment, neither tamoxifen followed by letrozole The size of the boxes is inversely proportional to the standard errors of the hazard ratio. The solid vertical line is placed at 0.82, which is the hazard ratio estimate for the overall analysis of the OS endpoint. *Other includes ER + /PgR unknown, ER unknown/ PgR + , ER -/PgR -(ineligible), ER -/PgR unknown, and ER unknown/PgR unknown. CI, confi dence interval; ER, estrogen receptor; L, letrozole; Nx, regional lymph nodes not examined; PgR, progesterone receptor; T, tamoxifen. Reprinted with permission from the Journal of Clinical Oncology [13]. Copyright 2011, American Society of Clinical Oncology.

A End Point Analysis (N=4,922) B Subgroup Analysis (OS)
Favors Letrozole Favors Tamoxifen  Figure 3). Women assigned tamoxifen followed by letrozole had a greater incidence of early relapses than those assigned letrozole monotherapy. Sequential treatment regimens with letrozole and tamoxifen did not improve DFS compared with letrozole monotherapy (Figure 3). Th e estimated 5-year DFS percentages for the three letrozole-containing regimens of the sequential treatment analysis population are 87.9%, 87.6%, and 86.2% for letrozole monotherapy, letrozole followed by tamoxifen, and tamoxifen followed by letrozole groups, respectively. Th ese diff erences are not statistically signifi cant.

Translational research
As early as 2006, the BIG 1-98 database was used to identify prognostic factors for breast cancer relapse.
Using the primary core analysis database [1] and Cox proportional hazards regression, we analyzed all eligible patients treated on BIG 1-98 and evaluated factors predictive of early relapse (within 2 years from random Letrozole followed by tamoxifen versus letrozole monotherapy. Both analyses were stratifi ed for chemotherapy use. The size of the boxes is inversely proportional to the standard error of the hazard ratio. As specifi ed in the protocol to account for multiple comparisons, 99% confi dence intervals are shown. Results are shown for the diseasefree survival, overall survival, and time to distant recurrence endpoints. Interaction tests between treatment and nodal status are not statistically signifi cant. CI, confi dence interval; Let, letrozole; Nx, regional lymph nodes not examined; Tam, tamoxifen. Reprinted with permission from The New England Journal of Medicine [12]. Copyright 2009, Massachusetts Medical Society.  assignment). Locally determined predictive factors for early relapse were node positivity, negative or missing estrogen receptor (ER) or progesterone receptor (PgR), high tumor grade, HER2 overexpression/amplifi cation, large tumor size, treatment with tamoxifen monotherapy, and peritumoral vascular invasion. Up front letrozole resulted in signifi cantly fewer early relapses than tamoxifen, even after adjusting for signifi cant prognostic factors. Th ese early fi ndings [18] helped set the stage for future translational research.
In 2005, the International Breast Cancer Study Group (IBCSG) Central Pathology Laboratory, recognizing the impor tance of accurate assessment of the molecular targets that might predict the better treatment, began collecting formalin-fi xed, paraffi n-embedded tumor blocks or, if these were not available, unstained slides for the assessment of ER, PgR, HER2, and Ki-67 labeling index (LI). Th e material was reviewed for histopathological features and expression of tumor biomarkers without knowledge of patients' treatment assignment or outcome. Th e fi rst reports using these assessments focused on comparing letrozole and tamoxifen monotherapy by using the database reported at median followup of 51 months [2].

Role of accurate assessment of steroid hormone receptor status
Th e fi rst translational analysis focused on the primary target for endocrine treatment, steroid hormone receptors. To be enrolled in BIG 1-98, the primary tumor had to be determined by local pathologists prior to random assignment to be positive for ER or PgR.
Th e impact and importance of central assessment of ER and PgR for patients receiving monotherapy were reported in 2007 [19]. Local assessment for eligibility classifi ed 99.9% of enrolled patients as having hormone receptor-positive disease. Central assessment using immunohistochemistry (IHC) of 6,291 tumors classifi ed 97.0% as positive (ER or PgR levels (or both) of at least 10%). Monotherapy cohort patients whose tumors were classifi ed locally as hormone receptor-positive and reclassifi ed as hormone receptor-negative by central review had worse outcomes, with estimated 65% 3-year DFS compared with 91% among patients whose tumors were classifi ed concordantly as positive (Figure 4).

Development of the Subpopulation Treatment Eff ect Pattern Plot
A key analysis tool for determining the importance of these continuous-value biomarkers (such as level of ER staining) for determining treatment eff ectiveness is the Subpopulation Treatment Eff ect Pattern Plot (STEPP) [20]. STEPP was specifi cally developed for exploring patterns of treatment outcome diff erences across sub popu lations of patients defi ned according to overlapping intervals of the biomarker values. Th e STEPP graphical presentation provides an overview of outcomes according to all values of a biomarker and can be used in addition to comparing treatments within subgroups based on estab lished or arbitrarily defi ned cutpoints ( Figures 5 and 6).

PgR and HER2 as predictive factors for letrozole versus tamoxifen monotherapy
Central assessment of ER, PgR, and HER2 status was evaluable for 3,650 patients (74%) from the monotherapy arms [19,21]. Note that these patients were treated prior to the availability of trastuzumab for HER2-positive tumors. Tumors were considered HER2-positive if amplifi ed by fl uorescence in situ hybridization (FISH) (geneto-chromosome 17 ratio of at least 2.0) or if staining intensity was 3+ by IHC (circumferential and intense membrane staining of greater than 10% invasive tumor cells) in the few tumors with unevaluable/unavailable FISH results. In tumors confi rmed to express ER, 10% had no PgR expression and 7% were HER2-positive. Alhough absence of PgR expression and HER2 positivity were each associated with poorer DFS, there remained a DFS advantage of letrozole for 5 years over tamoxifen for 5 years irrespective of PgR or HER2 status [19,21]. Accordingly, PgR or HER2 alone or together should not be used as discriminators in selecting initial adjuvant endocrine   Regan et al. Breast Cancer Research 2011, 13:209 http://breast-cancer-research.com/content/13/3/209 therapy for postmenopausal women with endocrineresponsive early breast cancer.

Ki-67 labeling index as predictive factor for letrozole versus tamoxifen monotherapy
Among patients in the monotherapy analysis, 2,685 had primary tumor material available for central pathology assessment of Ki-67 LI by IHC and had tumors confi rmed to express ER after central review [22]. Higher values of Ki-67 LI were associated with poorer DFS, and there may be an interaction of Ki-67 LI with treatment as the magni tude of the treatment benefi t for letrozole versus tamoxifen monotherapy was observed to be greater among patients with high-tumor Ki-67 LI than among those with low-tumor Ki-67 LI. In BIG 1-98, Ki-67 LI is a prognostic factor, and high Ki-67 LI levels may identify a patient group that particularly benefi ts from initial letrozole adjuvant therapy.

Composite risk assessment in the sequential treatment population
When the sequential treatment analysis was published [12], the IBCSG had collected and assessed additional tumor material for the four biomarkers. Th e investigation of the role of these biomarkers for the sequential treatments (four-arm option) ( Figure 1) included 5,177 patients (84%) with centrally confi rmed ER expression and other biomarkers available. Th is analysis was based on the database used for the primary outcome of the sequential treatments at median follow-up of 71 months [12]. Th is analysis also accounted for selective crossover.
We assessed whether centrally determined ER, PgR, and HER2 overexpression/amplifi cation and Ki-67 LI, alone or in combination with other prognostic features, predicted the magnitude of letrozole eff ectiveness compared with either sequence or tamoxifen mono therapy [23]. One motivation for this approach was that, despite a lack of data, a majority of the 2009 St. Gallen panelists preferred an initial aromatase inhibitor 'particularly for patients at higher risk' [11]. Second, in some settings, the cost and (for some patients) the side eff ects of aromatase inhibitor therapy make tamoxifen the preferable treatment. We therefore attemp ted to examine whether clinical and pathological features can identify patient groups for whom it is more or less important that a 5-year program include only or some aromatase inhibitor therapy.
Individually, none of the biomarkers signifi cantly predicted diff erential treatment eff ects among the treatment groups ( Figure 5). STEPP analysis of a composite measure of prognostic risk -which was based on an individual's number of involved lymph nodes, tumor grade, tumor size, presence of peritumoral vascular invasion as determined by local pathology, plus age and the four centrally assessed tumor biomarkers -revealed three patterns of treatment eff ects (Figure 6b). Patients at highest risk did best when treated with 5 years of letrozole; any of the three letrozole-containing regimens was acceptable for those patients in an intermediate-risk range; whereas patients at lowest risk did similarly well with letrozole monotherapy, a sequence of letrozole and tamoxifen, or tamoxifen monotherapy. Th us, in BIG 1-98, we confi rmed that the principle of composite assessment of risk, analogous to the clinical practice of integrating multiple risk factors when physicians and patients are deciding on the best adjuvant endocrine treatment for the individual patient, informs treatment selection better Tamoxifen Regan et al. Breast Cancer Research 2011, 13:209 http://breast-cancer-research.com/content/13/3/209 than individual biomarkers and supports the choice of 5 years of letrozole for patients at the highest risk for recurrence.

Future translational research
With the collection of over 5,000 tumor blocks, the IBCSG Central Pathology Offi ce was able to create tissue micoarrays to facilitate future research. Recent improvements in the extraction of DNA and RNA from formalinfi xed, paraffi n-embedded tumor blocks enable the investi gation of molecular features of the tumor and these features may predict a better patient outcome from endocrine therapy strategies. Plans are under way to investigate disease-related features, such as polymorphisms in or gene expression of ER-related genes, and patient-related features, such as polymorphisms in genes related to tamoxifen or aromatase inhibitor metabolism. Th e fi rst of these investigations evaluated the clinical relevance of cytochrome P450 2D6 (CYP2D6). Polymorphisms leading to reduced CYP2D6 enzyme activity may result in lower plasma concentrations of tamoxifen's clinically active metabolite endoxifen, and this may in turn adversely impact the effi cacy of tamoxifen. Because of confl icting results of several small studies that investigated the relation of CYP2D6 polymorphisms with clinical outcomes, whether polymorphisms that result in phenotypes of reduced enzyme activity are in fact associated with poorer disease control among tamoxifentreated patients is not clear. Results based on genotyping CYP2D6 polymorphisms among almost 5,000 BIG 1-98 patients were presented at the San Antonio Breast Cancer Symposium in December 2010 [24] and a manuscript is in preparation.

Side eff ects
For patients undergoing study drug administration, AEs were recorded at each 6-month follow-up visit by using check boxes on the case report forms. Severity was classifi ed according to National Cancer Institute Common Toxicity Criteria Version 2.0. Th e relationship of the AE to study treatment was assessed by the local investigator. Targeted AEs that were explicitly collected were cardiovascular events (that is, myocardial infarction, cerebrovascular accident, transient ischemic attack, angina requiring percutaneous luminal coronary angioplasty or coronary artery bypass graft, thromboembolic event, hypercholesterolemia (cholesterol values reported, mostly non-fasting), and 'other'), bone fractures, vaginal bleeding, endometrial pathology, nausea, vomiting, hot fl ushes, night sweats, and events leading to therapy dis continuation. Other cardiovascular AEs were collected by using an open-text comment fi eld for specifi cation by the investigator. Open-text fi elds were coded according to the Medical Dictionary for Regulatory Activities (MedDRA), and MedDRA preferred term [25] codes were further grouped into categories by IBCSG oncologists.
Senior oncologists at the IBCSG Coordinating Center, who were blinded to treatment assignment, reviewed all grade 3, 4, or 5 cardiovascular AEs; other grade 3 to 5 AEs whose causes were unclear; and all deaths occurring prior to a DFS event. Pre-existing cardiovascular morbidi ties reported at the time of enrollment were also medically reviewed. A report of the cardiovascular side eff ects following this extensive review was published in 2007, documenting a low overall cardiovascular AE incidence that diff ered between monotherapy treatment arms [26]. Th e incidence of hypertension or cerebrovascular events was similar as was the overall incidence of cardiac AEs, although there were more grade 3 to 5 cardiac AEs on letrozole; this excess was only partially attributable to prior hypercholesterolemia. Th ere were more overall and grade 3 to 5 thromboembolic AEs on tamoxifen. Th e incidence of AEs has been updated through a median follow-up of 76 months (Figure 7).
In addition, several focused reports on the safety of letrozole and tamoxifen in BIG 1-98, including bone events [27] in all patients and AEs in elderly patients [28], have been published. Consistent with other reports of aromatase inhibitors, the incidence of bone fractures was higher among patients treated with letrozole versus tamoxifen. Th e wrist was the most common site of fracture in both treatment groups. Risk factors for bone fractures during treatment included age, smoking history, osteoporosis at baseline, previous bone fracture, and previous hormone replacement therapy. Effi cacy results were similar for all age groups, including the 6% of 'elderly' patients (over age 75 years) who were less likely to complete 5 years of either trial treatment. Incidence of bone fractures did not diff er by age. Among elderly patients, letrozole had a signifi cantly higher incidence of any grade 3 to 5 protocol-specifi ed non-fracture AE compared with tamoxifen, but diff erences were not signifi cant for thrombo embolic or cardiac AEs. On the basis of a small number of patients older than 75 years, age per se should not be the only consideration for the choice of adjuvant endocrine therapy.
Detailed AE results comparing the entire 5-year treatment period for all four treatments [12] and for the mono therapy cohort [13] have been published. As previously reported, patients on tamoxifen experienced greater incidences of thromboembolic events, vaginal bleeding, hot fl ushes, and night sweats. Patients on letrozole experienced greater incidences of vaginal dryness, bone fractures, osteoporosis, and arthralgia/ myalgia ( Figure 7a) and higher-grade cardiac events (Figure 7b). It is important to note that these analyses present the incidence of AEs for one regimen (letrozole) compared with the other (tamoxifen), and it is possible that tamoxifen in particular off ers protection from cardiac or bone events [29,30]. Th e incidences of the AEs occurring in the sequential arms generally show results similar to those of the monotherapies during the time the patient was on the individual agents (that is, fi rst 2 years or last 3 years) [12]. Figure 8 shows Kaplan-Meier

Adverse Event
Tamoxifen (N=2447) Natural Log of Relative Risk (95% CI) cumu lative incidence estimates of time to the fi rst occurrence of four targeted AEs. Th e incidence rates of cardiac events, thromboembolic events, and bone fractures or severe osteoporosis (Figure 8a-c) were relatively constant throughout follow-up. In contrast, 24% of women taking letrozole and 29% taking tamoxifen reported hot fl ushes/night sweating within the fi rst year of therapy, with the incidence rates gradually decreasing over time (Figure 8d) [12].

Cognitive function substudy
In this substudy, cognitive function assessments were completed at the end of treatment (5 years) and about 1 year later (6 years). Th e 5-year analysis showed that cognitive function was signifi cantly better among the patients who were receiving letrozole at the end of the 5-year treatment period in comparison with those receiving tamoxifen [31]. A second analysis comparing the 5-year assessments of cognitive function with those collected about 1 year later showed a signifi cant improvement in cognitive function following completion of endocrine therapy for both treatment regimens; the magnitude of improvement was not signifi cantly diff erent between treatments [32].

Discussion
Th e BIG 1-98 trial was designed to compare 5 years of tamoxifen with 5 years of the aromatase inhibitor letrozole and to compare the strategy of the sequential treatments with the monotherapy approach. Twelve years after the fi rst patient was entered, these goals have been met. During these 12 years, evolving data from other trials and the selective crossover of one quarter of the tamoxifen-treated patients to a more eff ective treatment in this trial led to the adaptation of analysis plans to present the most accurate and clinically useful long-term results to the oncology community.
Although all of the patients have completed protocol treatment, BIG 1-98 continues to provide useful information beyond the primary questions of effi cacy. Th e translational reports on known biomarkers and the composite assessment obtained by putting the available tumor characteristics together to assess the risk of recurrence provide a basis on which to make informed patient treatment selection decisions. Future translational research will assess, for example, outcome according to CYP2D6, further clarifying individual patient diff erences that may be used to make treatment decisions. Th e detailed safety data from this study also provide information that can be used to assist in treatment selection. Th ese data are reassuring for the average patient, and certain patient co-morbidities and conditions should be considered in weighing the treatment options.

Conclusions
In all reports of this trial, the authors have attempted to present a clear interpretation of the treatment comparisons for both the average and the individual patient. Reports to date have presented the most clinically meaningful analyses and interpretations. Th e enhanced study design has enabled the presentation of two research questions from a single trial, thus saving time and money. On average, letrozole improves DFS, TDR, and OS in comparison with tamoxifen. Neither sequence of treatments improves results over letrozole monotherapy. However, there may be groups of patients (for example, those at low or intermediate risk) for whom tamoxifen or a sequence of the two agents represents a reasonable choice.
Th ere is no single interpretation of this trial, but for postmenopausal women with endocrine-responsive early breast cancer, BIG 1-98 results can be used to weigh the risks and benefi ts of the various treatment options in terms of effi cacy in tumor-defi ned subgroups, AEs observed among 8,000 women, and tumor biology not yet determined.

Competing interests
Novartis (Basel, Switzerland), as the sponsor of the BIG 1-98 trial, contributed funds to the IBCSG to conduct the trial and retrospective tissue collection. Some of these funds were used to partially fi nance the IBCSG Statistical Center (MMR, AG-H, KNP, and RDG) and participating centers (BT). BT owns stock in Novartis.

Author information
MMR is IBCSG group statistician. KNP is IBCSG director of scientifi c administration. AG-H is BIG 1-98 trial statistician. BT is BIG 1-98 study chair. RDG is IBCSG statistical and data management center director.