Risk of Stroke in Older Women Treated for Early Invasive Breast Cancer. Tamoxifen vs. Aromatase Inhibitors:

Transcription

1 Risk of Stroke in Older Women Treated for Early Invasive Breast Cancer Tamoxifen vs. Aromatase Inhibitors: A Population based Retrospective Cohort Study by Don Thiwanka Dilshan Wijeratne MBBS A thesis submitted in conformity with the requirements for the degree of Masters of Science Department of Health Policy Management and Evaluation University of Toronto Copyright by Don Thiwanka D Wijeratne, 2010

2 ii Risk of Stroke in Older Women Treated for Early Invasive Breast Cancer Tamoxifen vs. Aromatase Inhibitors: A Population based Retrospective Cohort Study Don Thiwanka Dilshan Wijeratne MBBS Submitted for the degree of Masters of Science Department of Health Policy Management and Evaluation University of Toronto 2010 Abstract Tamoxifen and aromatase inhibitors are treatment options for women with breast cancer and evidence on the risk of stroke is important in choosing between these two options. A systematic review of two randomized controlled trials and their nine related trial reports showed different methods for adverse event reporting and inconsistent estimates of stroke risk. In an observational cohort study of 5443 Ontario women, aged 66 years or older with early stage breast cancer, 86 ischemic stroke events (1.6%) occurred during follow-up of 5 years. There was no statistically significant difference in the risk of stroke between the hormone therapy groups [adjusted HR for tamoxifen compared to AI (0.810, 2.179)]. Results were similar across cardiovascular disease risk groups and were robust to different follow up periods and analytic methods. This study suggests that there is no significant difference in stroke between these treatment options.

3 iii Acknowledgments I would like to thank the following people for their invaluable advice and input into the work presented by this thesis Thesis supervisor: Paula. A Rochon, MD MPH FRCPC Thesis Committee: Geoffrey M. Anderson, MD PhD Peter Austin, PhD Lorraine L Lipscombe, MD MSc FRCPC Andrea Gruneir, PhD Others: Dallas Seitz MD FRCPC Hadas Fischer MD MSc Sunila R. Kalkar MBBS MD MEd Wei Wu BSc MSc Onil Bhattacharyya MD PhD Ophira Ginsburg MD FRCPC Moira Kapral MD FRCPC MSc Refik Saskin MSc Staff members: Women s College Research Institute (WCRI) and The Institute of Clinical Evaluative Sciences (ICES) Family: My wife Nishardi and other members of my family

4 iv Table of Contents Abstract...ii Acknowledgments...iii List of Tables...vi List of Figures...vii List of Appendices...viii Introduction and study objectives Introduction Research Question Hypothesis Research Objectives... 3 Among early breast cancer survivors;... 3 Background Hormonal therapy and breast cancer Implications of hormonal therapy Adverse events with hormonal therapy... 7 Systematic review Evidence from RCTs Focus of the systematic review Chapter objectives Introduction Methods Results Trial characteristics Methods to report adverse events Reported adverse events Discussion Retrospective cohort study: Methods Rationale Hypothesis Research Objectives Chapter Objectives Overview of study methodology Data Sources Ontario Cancer Registry Ontario Drug Benefits Program Canadian Institutes for Health Information Ontario Health Insurance Program (OHIP) Registered Persons Database (RPDB) Creating an early invasive breast cancer cohort Definitions of study variables Exposures of interest Outcomes of interest Follow-up Potential confounders... 50

5 4.9.1 Demographic information Co-morbidity (in the year before cohort entry) Healthcare visits in the year before diagnosis (OHIP/IPDB) Medical conditions and procedures Drug therapy Analytic strategies Results Objectives Chapter Abstract Cohort characteristics Healthcare utilization Co-morbidities Cancer related therapies Information on exposure to hormonal therapy Outcomes up to 5 years of follow-up Ischemic stroke All stroke Stroke outcomes with and without preexisting cardiovascular disease Stroke severity Analyses based on a maximum follow-up (31st March 2009) Ischemic stroke All stroke Events Prior cardiovascular disease No prior cardiovascular disease Stroke severity Discussion Objectives Chapter abstract Cohort Characteristics Comorbidities in the cohort Comparison of RCT evidence Information on exposure to hormonal therapy Outcomes Implications for individual care Implications on healthcare delivery Implications for future research Limitations of the study Overall importance of this study Future Directions Appendices Appendix 1 Modeling process on ischemic stroke Appendix 2 Dataset Creation plan Appendix 3 Copy of Ethics approval v

6 vi List of Tables Table 1 Common adverse events reported with tamoxifen and AI Table 2 Description of ATAC and BIG trials and their subsequent reports Table 3 Details on reporting adverse events Table 4 Cardiovascular events in the ATAC and BIG RCTs Table 5 Algorithm used to create the cohort of females with early breast cancer Table 6 CCP codes for beast cancer surgery Table 7 Drugs and their DINs used as exposures of interests Table 8 Codes used to define stroke Table 9 Medical conditions and procedures known to influence risk of stroke Table 10 Medical conditions and procedures captured in our study and their codes. 63 Table 11 Drug Therapy known to influence the risk of stroke Table 12 Drug Therapy captured through the ODB Table 13 Cohort Characteristics Table 14 Comorbidities Table 16 Information on exposure to Hormonal therapy Table 17 Information on exposure to Hormonal therapy Table 18 Information on Follow-up and Outcomes Table 19 Analyses of outcomes Table 20 Analysis of Outcomes up to 31st March

7 vii List of Figures Figure 1 Flow Diagram of Included and Excluded Studies Figure 2 Schema of outcomes captured in this review Figure 3 Study specific days Figure 4 Schema of the relationship of confounders Figure 5 Outline of model building strategy Figure 6 Individuals with breast cancer by year of diagnosis Figure 7 Cumulative incidences of stroke and deathchapter

8 viii List of Appendices Appendix 1 Modeling process on ischemic stroke Appendix 2 Dataset Creation plan Appendix 3 Copy of Ethics approval

9 1 Chapter 1 Introduction and study objectives 1.1 Introduction Over 65% of patients with breast cancer are known to have estrogen or progesterone receptor positive disease (1). Tamoxifen and aromatase inhibitors (AI) are estrogen modulating pharmacological agents that are extensively used to treat receptor positive breast cancers. These therapies contribute to very high survival benefits offering up to 89% survival over five years in early breast cancer patients reducing risk of local and distant cancer recurrence and mortality (2). Treatment guidelines to date suggest both tamoxifen and AI as recommended treatment options. Either of these therapies is commenced following surgery for breast cancer with or without radiotherapy and chemotherapy. Some risk factors associated with these hormonal therapies are well known. Tamoxifen is known to be associated with thromboembolism and endometrial cancer while AIs are more commonly associated with fractures and arthralgias(3). Previous studies suggest that stroke may be an important adverse outcome associated with tamoxifen. However there is conflicting evidence on the risk of stroke from randomized controlled clinical trials (RCTs). There are two major trials that compare monotherapy of tamoxifen to AI. The ATAC trial showed significantly higher risk of stroke with tamoxifen relative to AI(4) while the BIG trial did not find any association(5). Hence there is a need to understand what the RCTs found, what the absolute risks were and how stroke risk fits in with other cardiovascular events. It is important to compare the outcomes between these therapies, more so than with placebo controls, in order to find out the relative risks of these events with these therapeutic options. It is unclear whether strokes with hormonal therapies follow an atherosclerotic or thromboembolism pathophysiology. It is possible that tamoxifen may increase the risk of stroke

10 2 with its prothrombotic properties. By comparing stroke outcomes with other cardiovascular outcomes and thromboembolic events we would be able to delineate the relationship of stroke to either atherosclerotic or thromboembolism pathophysiologies. Also such estimates on the risk of stroke in RCTs seem less generalizable to the real world as these studies include pre-selected individuals. Furthermore little is known about risk of stroke in older individuals and those with pre-existing cardiovascular diseases (CVD). Older individuals have a higher baseline risk of cardiovascular events. Both relative and absolute risk of stroke and other cardiovascular events of these individuals may be different than those represented in clinical trials. Understanding the risk of stroke associated with the use of tamoxifen compared to AI would facilitate informed therapy selection, which would help health providers and patients make decisions about available treatment options. First, this thesis provides a detailed review of the evidence available from RCTs on the risk of cardiovascular adverse events with hormonal therapy in women treated for early invasive breast cancer. This systematic review describes the comorbidities present in study subjects and methods used to report adverse events in those RCTs. Second, we conducted a population level study to explore stroke events within each hormonal therapy group used to treat breast cancer. This is a retrospective cohort study comparing the risk of stroke in older females who have ever used hormonal therapy (tamoxifen versus AI) for treatment of breast cancer. The cohort of this population-based study consists of all adult females, aged 66 years or older, in Ontario, who were diagnosed with early operable breast cancer defined by the Ontario Cancer Registry and who survived for at least 365 days. They were followed for up to 5 years. This study describes the population of individuals who take these medications in the real world. This provides the opportunity to compare them with those represented in clinical trials. This systematic review and population level study provides the opportunity to explore longerterm stroke outcomes with hormonal therapy used in individuals with breast cancer. By understanding the risk associated with tamoxifen versus AI in different subgroups, physicians and patients may have important information to minimize potential harm associated with long term use of these medications. This also sets the forum to answer different research questions on other outcomes relating to these therapies using parallel study designs.

11 3 1.2 Research Question What is the risk of stroke with tamoxifen versus aromatase inhibitors in older women treated for early invasive breast cancer? 1.3 Hypothesis There may be an increased risk of stroke with tamoxifen when compared to aromatase inhibitors in older women treated for early invasive breast cancer. 1.4 Research Objectives Among early breast cancer survivors; 1 to describe the characteristics of older women who are dispensed tamoxifen and aromatase inhibitors (anastrozole, letrozole, examestane) in RCTs and in a real world cohort. 2 to compare the risk of ischemic stroke between users of tamoxifen and aromatase inhibitors. 3 to compare the risk of all stroke between users of tamoxifen and aromatase inhibitors. 4 to determine whether there are differences between women with and without a previous history of cardiovascular disease in the risk of stroke (ischemic or all ) associated with tamoxifen versus AI. 5 to describe the severity of stroke (ICU admission or death within 30 days of stroke) among older women treated with tamoxifen or an AI.

12 4 Chapter 2 Background 2.1 Hormonal therapy and breast cancer Approximately two-thirds of postmenopausal women (PMW) diagnosed with breast cancer have estrogen-receptor positive pathology (1). These women are offered estrogen suppressing systemic adjuvant hormonal therapy in the form of tamoxifen or aromatase inhibitors (anastrozole, letrozole or examestane) either immediately following surgery or after a course of radiotherapy and intravenous chemotherapy. The hormonal therapy significantly reduces risk of local and distant cancer recurrence and mortality and offers good survival benefit of up to 89% five years survival (2). Tamoxifen, a selective estrogen receptor modulator (SERM), has been the mainstay of adjuvant endocrine therapy since 1970 offering significant survival benefit (6). The third generation aromatase inhibitors (AI) have been used as an alternative to tamoxifen for the last decade (7). Anastrozole, letrozole, examestane are considered to have similar anti-tumor activity and toxicity profiles (8). Both tamoxifen and AIs have comparable cancer related outcomes including disease free survival, overall survival and cancer recurrence (9). Both Cancer Care Ontario(10) and American Cancer Society guidelines(2) suggest either tamoxifen or AI as acceptable treatment options in patients with early invasive breast cancer. Tamoxifen is a drug therapy with potent anti-estrogenic properties which compete with estrogen for binding sites in breast and other tissues. It competitively binds to estrogen receptors on tumors and other tissue targets, producing a nuclear complex that decreases DNA synthesis and inhibits estrogen effects. Therefore, tamoxifen is cytostatic rather than cytocidal. Tamoxifen is well absorbed from the gut and appears in high concentrations in uterus, endometrial and breast tissue. It has 99% protein binding. Tamoxifen metabolism is hepatic via CYP2D6 to 4-

13 5 hydroxytamoxifen and via CYP3A4/5 to N-desmethyl-tamoxifen. It is then further metabolized into endoxifen (4-hydroxy-tamoxifen via CYP3A4/5 and N-desmethyl-tamoxifen via CYP2D6). Both 4-hydroxy-tamoxifen and endoxifen are fold more potent than tamoxifen. The half life elimination of tamoxifen is 5-7 days. Excretion is via feces (26% to 51%) and urine (9% to 13%). (11) Aromatase inhibitors (AI) are potent inhibitors of estrogens. Anastrozole and letrozole are nonsteroidal aromatase inhibitors while examestane is an irreversible steroidal aromatase inactivator. However all three drugs are considered to have similar anti-tumor activity and toxicity profiles. Anastrozole is the most commonly used AI. It acts by inhibiting aromatase, the enzyme which catalyzes the conversion of androstenedione to estrone, and testosterone to estradiol. The reduction in circulating estrogen prevents the stimulatory effect of estrogenic stimulation on hormone-receptor positive tumor cells.. Anastrozole causes an 85% decrease in estrone sulfate levels. The onset of estradiol reduction is 70% reduction after 24 hours and 80% after 2 weeks therapy. The duration of action is 6 days. Anastrozole is well absorbed from the gut. It is 40% protein bound. Metabolism of anastrozole is mainly hepatic (around 85%) via N- dealkylation, hydroxylation, and glucuronidation. The primary metabolite (triazole) is inactive. The elimination half-life is hours. Excretion of anastrozole is via feces and urine while urinary excretion accounts for about 11% of total elimination, mostly as metabolites.(12) Although both tamoxifen and AIs suppress the effects of estrogen on hormone-receptor positive breast cancer cells, the differences in pharmacokinetics and pharmacodynamic changes the way in which they achieve this goal. This is also evident by the different adverse event profiles and the mechanisms by which they occur. Tamoxifen has mixed agonist and antagonist activity, depending on the target tissue it acts on. Differences in coactivator/corepressor status among different tissues may help explain the variable response to tamoxifen in target sites. (13-14) As an example tamoxifen provides some protection against menopausal bone loss, presumably due to their partial agonist activity.(15) Also of interest, tamoxifen induces endometrial hyperplasia as an estrogen agonist effectand increases the risk of developing endometrial cancer. On the other hand as an estrogen antagonist effect tamoxifen induce hot flashes.(16)

14 6 Aromatase inhibitors act by inhibiting estrogen synthesis hence producing anti-estrogenic effects by reducing the overall circulating levels of estrogen.(17) This is evident by bone demineralisation and increasing the risk of fracture associated with estrogen deprivation. The estrogen agonist effects like endometrial hyperplasia are not evident with AIs. It is unclear how these differential mechanisms of tamoxifen and AIs translate in to cardiovascular outcomes (See also section 2.3). Currently three different treatment strategies are used for administering adjuvant hormonal therapy. The first strategy involves using either tamoxifen or an AI mono-therapy for 5 years. The second strategy called the early switching strategy uses 2-3 years of tamoxifen followed by a switch to an AI, for a total of up to 5 years. The third extended switching strategy involves a full 5 years of tamoxifen, followed by 5 years of an AI, usually letrozole. (8) The choice of specific adjuvant therapy and how long it should be continued is dependent on the age of the individual, other diagnoses or medical conditions and the balance between the potential benefits and possible side effects. It is well established that tamoxifen use is associated with serious adverse events such as thromboembolism and endometrial cancer. (3, 18) AIs are known to be associated with fractures and arthralgias. (3) As adjuvant hormonal therapy is generally administered for long periods of time, the safety and tolerability of agents used in this setting are of particular importance. In Ontario both tamoxifen and AIs are dispensed though the Ontario Drug Benefit plan (ODB) free of charge to individuals over 65 years of age. AIs however are classified as limited use medication where the physician has to justify its prescription given its higher costs. AIs are much more expensive than tamoxifen. A daily dose of 1 mg AI (anastrozole, brand name Arimidex) costs $4.95 while a daily dose of 2.5mg AI (letrozole, brand name Femara) costs $5.51. In contrast a daily dose of 20 mg tamoxifen costs cents.(19)

15 7 2.2 Implications of hormonal therapy Tamoxifen and AIs are two types of hormonal therapy that are known to have similar benefit on breast cancer outcomes. However they are associated with different adverse events. To facilitate the formulation of a customized therapeutic strategy it is important to consider the different adverse event profiles of these therapies. The majority of individuals with an early breast cancer diagnosis survive the cancer with combinations of surgery, radiotherapy and chemotherapy. Cancer survivors now live longer due to effective longer term therapies like hormonal therapy.(9) With more people surviving breast cancer, this condition is now viewed as a chronic disease. With increased survival these individuals are now more likely to develop conditions unrelated to cancer such as cardiovascular disease, in addition to late or long term effects that can result from their cancer therapies. 2.3 Adverse events with hormonal therapy The adverse event profiles of tamoxifen and AI are largely predictable consequences of estrogen deprivation and partially overlap due to biological properties of these drug classes.(20) Table 1 compares the adverse events between tamoxifen and AI. AIs include anastrozole, letrozole, examestane and are considered to have similar anti-tumor activity and toxicity profiles.(8) AIs reduce circulating endogenous estrogen levels to almost undetectable levels. Tamoxifen acts as an estrogen receptor (ER) antagonist or ER agonist, depending on the target tissue.(20). Because of their differing mechanism of action, their adverse event profiles are different. Tamoxifen has been reported to have a generally favourable effect on serum lipid profiles. (21) Patients with breast cancer and subsequent chemotherapy-induced ovarian failure developed marked increases in total cholesterol and low-density lipoprotein levels; adjuvant tamoxifen was found to decrease these serum lipid concentrations to below baseline concentrations.(22) There is also a suggestion that this may be translated into cardiovascular benefits. A cohort study of adjuvant tamoxifen found that the drug reduced the incidence of myocardial infarction (23), and

16 8 a RCT of the same therapy compared to placebo also showed a trend towards a decrease in mortality from coronary heart disease. (24) The selective estrogenic properties of tamoxifen are known to increase the risk of thromboembolism even though the exact pathophysiology is yet to be determined.(25) Aromatase inhibitors on the other hand are known to increase fracture risks by reducing bone mineral density. This is thought to occur through suppression of estrogen levels induced by AIs that inhibit insulin-like growth factor 1 and vitamin D bioavailalability.(17) The effects of AI on cardiovascular events are less well characterized. Published RCTs and systematic reviews have reported inconsistent information about risk of stroke and other cardiovascular adverse events (i.e. ischemic heart disease events) related to tamoxifen versus AI.(26-27) While tamoxifen is considered to be generally cardioprotective it is unclear why the risk of stroke remains high in studies that compare tamoxifen to AI. There is some suggestion that tamoxifen therapy leads to more stroke events when compared to AI (26-29). This may be partly explained by the prothrombotic effects of tamoxifen which is not evident with AIs. However the incidence of ischemic heart disease is higher with AI than tamoxifen.(26-27). Tamoxifen is generally considered to be protective against atherosclerosis which is a risk factor for cardiovascular disease.(30) It is unclear as to why these inconsistencies exist on cardiovascular events and risks. There are two major trials that compare monotherapy of tamoxifen to monotherapy with AI. The ATAC trial showed significantly higher risk of stroke with tamoxifen relative to AI while the BIG trial did not find this association.(4, 30-31) Both stroke and cardiovascular events are primarily attributed to either atherosclerosis or thrombus formation. Stroke could also occur as a result of thromboembolism.(32) It is unclear whether strokes with hormonal therapies follow an atherosclerotic or thromboembolism pathophysiology. By comparing stroke outcomes along with other cardiovascular outcomes and thromboembolic events, we are able to explore the relationship of stroke to either atherosclerotic or thromboembolic pathophysiologies.

17 9 There is a need to identify gaps in knowledge on the risk of cardiovascular events with hormonal therapy which would inform a population level cohort study. Furthermore little is known about the frequency and impact of such events in the real world where these therapies are used by cancer survivors who include an older population and those with a previous history of cardiovascular disease (CVD). These individuals are not well studied in the large scale RCTs. The older population also has unique features that place them at increased risk of adverse events with these hormonal therapies. Specifically, they may be more prone to osteoporosis which is known to be aggravated by AI.(3) These patients also have higher absolute risk of developing strokes and other cardiovascular events.(33) Hence it is important to know the relative risks versus benefits with hormonal therapies so that an informed therapeutic decision could be made. Chapter Conclusion There is a need to revisit the available evidence on the risk of stroke and other cardiovascular outcomes associated with hormonal therapy used in early breast cancer. Information is required on patients with pre-existing cardiovascular risk and in those who are elderly. This information will help clinicians facilitate a better-informed therapeutic decision while offering the most appropriate treatment options to those who are vulnerable to particular adverse events.

18 10 Table 1 Common adverse events reported with tamoxifen and aromatase inhibitors Tamoxifen Gynecological Hot flushes, vaginal bleeding Vaginal discharge Endometrial cancer Cardiovascular Stroke Thromboembolic Deep venous thrombosis Pulmonary embolism Aromatase inhibitors Musculoskeletal Fractures Arthralgia Osteoporosis Cardiovascular Ischemic heart disease

19 11 Chapter 3 Systematic review Evidence from RCTs 3.1 Focus of the systematic review This systematic review describes the randomized controlled clinical trial evidence on cardiovascular risk and events with hormonal therapy (tamoxifen vs. aromatase inhibitors) used in early breast cancer. This review was conducted in order to identify gaps in knowledge on the risk of cardiovascular events and thromboembolism with hormonal therapy. This information was then used to inform the population level cohort study described in chapters 4-6 of this thesis. 3.2 Chapter objectives Among RCTs of early breast cancer treated with monotherapy of tamoxifen versus aromatase inhibitors: 1 to describe the demographic and co-morbid characteristics of individuals in these RCTs; 2 to discuss differences in study inclusion and exclusion criteria and the methods used to report adverse events; and 3 to compare the risk of cardiovascular events between tamoxifen and AI including the risk of stroke.

20 Chapter Abstract The majority of postmenopausal women (PMW) with estrogen-receptor positive breast cancer receive estrogen suppressing adjuvant hormonal therapy [tamoxifen or AI]. As adjuvant hormonal therapy is generally administered for long periods of time, the safety and tolerability of these agents are of particular importance. We included RCTs of adult females diagnosed with early invasive operable breast cancer in order to measure the risk of cardiovascular events with hormonal therapy. We then selected RCTs that compared monotherapy arms of either tamoxifen or AI. Two RCTs, Arimidex Tamoxifen Alone or in Combination (ATAC) and Breast International Group trial (BIG) were identified. Both RCTs include post menopausal individuals who had surgery for their breast cancers. These two parent trials resulted in nine subsequent published trial reports that describe our outcomes of interest (5 from the ATAC trial and 4 from the BIG trial) As RCTs report their results at different time points, we compared adverse reporting from the RCTs at three defined time periods of less than 3 years, 3 to 5 years and more than 5 years. Based on our review findings the risk of stroke with tamoxifen compared to AI was mixed. While BIG trial results showed no incremental risk of stroke with tamoxifen when compared to AI, ATAC showed persistent elevated risk of stroke with tamoxifen during treatment. This discrepancy may be attributed to the way that information on adverse events was captured. When looking at other cardiovascular events, the AI groups generally seem to have more ischemic heart disease and cardiac failure events during this follow-up period although this difference was not statistically significant. Hypertension was significantly more likely to occur in the AI group. Hypercholesterolemia was also significantly more likely in the AI groups when compared to tamoxifen. Thromboembolic events continue to be significantly more likely with tamoxifen when compared to AI.

21 Introduction Tamoxifen and aromatase inhibitors (AI) are extensively used to treat estrogen / progesterone receptor positive breast cancers. Tamoxifen is a selective estrogen receptor modulator. AIs (i.e. anastrozole, letrozole, examestane) are potent inhibitors of estrogen synthesis where all of the AIs have comparable efficacy and safety profiles.(8) These medications have proven prognostic benefit in reducing breast cancer recurrence, increasing disease free survival and overall survival. (9) However they may be associated with adverse outcomes in different organ systems of the body.(3) Information from randomized controlled trials and observational studies has highlighted stroke as an important adverse event associated with the use of tamoxifen compared to AI. These studies provide somewhat contradictory evidence related to stroke and other cardiovascular events. Many systematic reviews have addressed cardiovascular adverse events with hormonal therapy. They report increased risk of stroke events with tamoxifen when compared to AI. (26-29) However all these reviews include information from RCTs that compare tamoxifen to AI that have a 2-5 year baseline exposure to tamoxifen before individuals are randomized. Four out of the six major trials that compare tamoxifen to AI namely MA-17, IES, ARNO/ABCSG and ITA have such baseline exposure to tamoxifen before individuals are randomized to either tamoxifen or an AI.(10) Furthermore these reviews also include information on adverse events during periods of switching treatment between tamoxifen and AI which were predefined components of these study algorithms. These factors make attribution of adverse events to either therapy problematic. As the current practice issue deals with which drugs to start with (tamoxifen or AI), trials comparing monotherapy with tamoxifen to aromatase inhibitors (i.e. anastrozole, examestane, letrozole) were deemed to be relevant for this review. We were primarily interested in the risk of stroke events [any type ( all stroke ) and, ischemic] with these hormonal therapies as reported stroke events from previous studies are inconsistent with other cardiovascular events reported in the same studies.(26-27) Hence we compared evidence reported in RCTs on the risk of stroke and other cardiovascular adverse events among

22 14 early breast cancer survivors using data from RCTs that look at individuals who are on monotherapy with tamoxifen versus AI to explore this issue. 3.5 Methods We conducted a systematic review of the early invasive breast cancer RCTs that compared tamoxifen or AI monotherapies. We were interested in the risk of stroke, stroke severity including death and other cardiovascular events relating to atherosclerosis and thromboembolism. We included RCTs of adult females diagnosed with early invasive operable breast cancer. Treatment guidelines to date recommend the commencement of either tamoxifen or AI as initial treatment with hormonal therapy (2, 10). As the current clinical decisions deal with which drugs to start with (tamoxifen or AI), trials that compare monotherapy with tamoxifen to aromatase inhibitors have been considered in this review. Those where chemotherapy was provided at the same time as tamoxifen or AI were excluded. The selection of studies and articles for this review was multi-fold. We initially did a MEDLINE (1980 to 2010 March 31st) and EMBASE (1980 to March 31st 2010) search of RCTs that compare tamoxifen to AI for breast cancer as monotherapies. By using the terms [breast cancer and tamoxifen and aromatase inhibitors or AI (anastrozole) or AI (letrozole) or examestane and RCTs] we identified 119 articles. We then selected RCTs that contained monotherapy arms of either tamoxifen or an AI. Two RCTs Armidex Tamoxifen Alone or in Combination (ATAC)(34) and Breast International Group trial (BIG)(35) were identified. Second, the Early breast cancer Trialists' Collaborative Group (EBCTCG) (1) RCTs registry was screened to identify RCTs that met our inclusions. The same two major RCTs (i.e. ATAC and BIG) were identified. This selection was then further validated by the same selection in the recent meta-analysis conducted on breast cancer outcomes.(9) Subsequent published reports of longer-term follow-up of these two RCTs were

23 15 then captured using MEDLINE and EMBASE searches and through citations of these articles. Five publications reporting results from the ATAC trial (4, 34, 36-38) and four publications reporting results from the BIG trial(5, 35, 39-40) that include our outcomes of interest were identified. In addition one publication of the BIG trial described the methods and outcomes of the trial. (31) See Figure 1 for a flow diagram of included and excluded studies. Two independent reviewers (Thiwanka Wijeratne and Wei Wu) extracted information on the cohort (individual number and accrual), demographic information which included the mean (standard deviation) age, menopausal status, comorbidities and pathology information and study exclusions. We then extracted information on interventions including the drugs used, their dosing, frequency and duration, measures relating to the outcomes, duration and completeness of follow-up. Of those individuals who were treated with sequential therapy with tamoxifen and AI, only the information while on the initial therapy was included. The methodological quality of the RCTs was also assessed. Differences were resolved by discussion and when required with the help of a third arbitrator (Sunila Kalkar). Adverse events We extracted information on strokes and other cardiovascular events that were reported in these RCTs using the definition of cardiovascular events by the World Health Organization, Council for International Organizations of Medical Sciences (CIOMS), reporting of adverse drug reactions.(41) Cardiovascular events were defined as myocardial infarction, stable/unstable angina, percutaneous coronary intervention (PCI), coronary artery bypass graft (CABG), congestive heart failure, transient ischemic attack (TIA), stroke and/or peripheral vascular disease. In addition we also extracted information on thromboembolism as a related vascular event. We focused on the risk of stroke events [any type of stroke ( all stroke ), ischemic] with these hormonal therapies. The ATAC reported ischemic stroke events while the BIG captured all stroke events. The definitions of such events were not described.

24 16 Both stroke and cardiovascular events are primarily attributed to either atherosclerosis or thrombus formation although minority of strokes could occur with thromboembolism.(32) It is unclear whether strokes with hormonal therapies follow an atherosclerotic or thromboembolism pathophysiology. Given the plausible dual pathophysiology of ischemic strokes relating to atherosclerosis and hypercoagulability/thromboembolism we looked at other adverse events which share similar mechanisms. On events related to atherosclerosis we compared other cardiovascular events as defined by CIOMS(41) namely ischemic heart disease events, cardiac failure, and peripheral vascular disease. Although specific descriptors on ischemic heart disease events were not provided in the two trials, ATAC reported ischemic cardiovascular events while the BIG reported ischemic heart disease events. For the purpose of this review we present these findings as ischemic heart disease events. Deep venous thrombosis and pulmonary embolism were considered as outcomes relating to hypercoagulability and embolic phenomena. We also looked at incident reports of hypertension and hypercholesterolemia as risk factors of cardiovascular events. See Figure 2 for a schema of outcomes captured in this review. As RCTs report their results at similar time ranges but different specific time points, we compared adverse reporting from the RCTs at time periods of less than 3 years, 3 to 5 years and more than 5 years. Serious Adverse events We collected data on serious adverse events while on treatment and off treatment. Serious adverse events (SAE) were defined by the BIG trial in accordance to the International Conference on Harmonization (ICH) Guidelines(42), and were graded based on the Common Toxicity Criteria of the National Cancer Institute.(43) This system classified grade 3 to 5 events as serious adverse events. In the ATAC trial SAE were defined using the Good Clinical Practice definitions.(44) We have described SAE using the same definitions that were employed by

25 17 individual trials. See Table 3. The ATAC trial reported outcomes based on on and off treatment. The BIG trial only reported SAE based on median follow-up times. In order to compare the BIG and ATAC trial results, the median follow-up time of 71 months in the BIG trial was considered as an on treatment analysis. Analytic strategies We reported the number of adverse events provided at each time point in subsequent reports of the two parent trials (ATAC and BIG). We have also included the number of individuals in each hormonal therapy group at each time point of the two trials. In order to compare the adverse events we used the raw numbers provided in these trials and from this information we calculated the odds ratio (OR) for the adverse events at each time point. These methods are recommended for adverse event reporting by the Cochrane collaboration.(45) For all OR calculations the AI group was considered the reference group. Hence the OR estimates are for the tamoxifen groups relative to AI. When available the published estimates on the risk of adverse events are also presented. These trials however report their risk estimates on AIs relative to tamoxifen. When reporting these published estimates, we have indicated this difference in the reference group in comparison to what we have calculated. Clinical heterogeneity was assessed based on individual populations, the outcome definitions and on how they were collected. Statistical heterogeneity was assessed based on the Cochrane s chi-square test of homogeneity and the I 2 statistic although results of the two RCTs were not combined. Methodological quality assessment The methodological quality of each trial was assessed using the Cochrane risk of bias assessment tool.(46) We used four domains from this tool (i.e. assessment of randomization, masking of intervention, masking of outcome assessment and completeness of follow up). Randomization was assessed based on methods for sequence generation and allocations to interventions. Assessment of masking of intervention and outcome assessment was based on the precautions

26 18 taken to maintain double blinding. Completeness of follow-up was assessed based on the accountability of the cohort that was represented at inception and in subsequent analyses. The fulfillment of each of these domains is presented as a dichotomized variable (yes/no). 3.6 Results Trial characteristics We identified two major randomized controlled clinical trials (ATAC(34) and BIG-1-98(35)) that compared monotherapy with tamoxifen versus AI. These two trials had 9 subsequent published reports (5 related to ATAC(4, 34, 36-38) and 4 related to BIG(5, 35, 39-40)) that described our outcomes of interest. These reports provide follow-up information up to 2009 which includes up to 100 months of median follow-up. The ATAC trial (4), compared the effects of AI (anastrozole) versus tamoxifen alone as well as in combination used sequentially. ATAC enrolled 9366 post menopausal women age (SD) 64.3 (9.1) years with estrogen or progesterone receptor positive early breast cancer. The median follow-up was 100 months (range months).the interventions were AI (anastrozole) 1 mg/day (n=3125) for a mean (SD) duration of 3.97 (1.71) years vs. tamoxifen 20 mg/day (n=3116) mean (SD) duration of 4.11 (1.65) years versus combination (not included in the review as we were interested to account for adverse events with monotherapy). See Table 2 for a description of the RCTs cohort characteristics. The BIG trial (5) compared the effects of AI (letrozole) versus tamoxifen. Participants included 8028 individuals median age of 61 (range 38-89) years who were post menopausal with estrogen or progesterone receptor positive operable breast cancer. Three thousand seventy four individuals were given monotherapy throughout the trial while the remainder switched between treatment options. The maximum median follow-up time published to date was up to 71 months. The interventions were tamoxifen 20 mg/d (n=1540) vs. AI (letrozole) 2.5 mg daily (n= 1534) for 5 years. In addition sequential therapy AI (letrozole) (1540) or tamoxifen (1548) for 2 years followed by the other for 3 years (total up to 5 years) was also considered in this trial.

27 19 Information on adverse events up to treatment switching of these individuals is included in this review. This review reports information on early breast cancer individuals (6241 in ATAC and 8028 in BIG) who were given hormonal mono-therapies, either tamoxifen alone or AI alone. Both studies were well designed RCTs. The quality assessment using the Cochrane bias assessment tool indicates that they were at low risk for introducing bias. See Table 2 for a description of RCTs for quality assessment Methods to report adverse events In the ATAC trial at every 6 month visit, investigators asked individuals whether they experienced any adverse events and they recorded events on the trial case-report forms. Prespecified checklists of potential adverse events were not used. Instead investigators recorded all adverse events occurring while an individual was receiving treatment or up to 14 days after study treatment ended. In the BIG trial for individuals undergoing study drug administration, adverse events (AEs) were recorded also at each 6-month follow-up visit. Check boxes on the case report forms were used to capture information about specific adverse events. These adverse events were recorded during or within 28 days after study treatment. Table 3 summarizes information on reporting adverse outcomes and other trial endpoints Reported adverse events As RCTs report their findings at different time points of follow-up, the adverse events in this systematic review are grouped in to different time frames (first 3 years, between 3-5 years and beyond 5 years). In addition serous adverse events are reported for on and off treatment periods. See Table 4.

28 First 3 years follow-up Within the first 3 years of follow-up the ATAC trial reports adverse events at 33.3 months of median follow-up. The BIG reports adverse events at two time points. The early report was based on information available from a median follow-up at 25.8 months and the late report from 30.1 months of median follow-up. Stroke events - Both ATAC and BIG trials reported stroke events. The ATAC trial reported significantly more ischemic stroke events in the tamoxifen group compared to the AI (anastrozole) group (p=0.0006). We found that individuals on tamoxifen were twice as likely to develop an ischemic stroke. The unadjusted OR that we calculated for tamoxifen relative to AI (anastrozole) was 2.12 [95%CI 1.38, 3.26]. In the BIG trial the unadjusted ORs we calculated both in the early, (OR 1.05 [95%CI 0.67, 1.63]) and late, (OR 1.04 [95%CI 0.70, 1.55]) reports found no difference between the tamoxifen and AI (letrozole) groups on the risk of all stroke events. In the BIG trial late report, the published risk ratio for all stroke events for AI (letrozole) relative to tamoxifen was The published hazard ratio (HR) for stroke events for AI (letrozole) relative to tamoxifen was 0.94 [95% CI 0.63, 1.41]. Our calculated ORs and the published estimates showed inconsistent findings between the ATAC and BIG trials with a trend for more events in the tamoxifen groups relative to the AI groups. The number of events reported in each trial is shown in Table 4. Other cardiovascular events related to atherosclerosis Ischemic heart disease events (IHD) - Both ATAC and BIG trials reported IHD events. In the ATAC trial there was no difference in the risk of IHD events reported between the tamoxifen and AI (anastrozole) groups (p=0.14). The unadjusted OR we calculated for IHD when tamoxifen was compared to AI (anastrozole) was 0.77 [95%CI 0.55, 1.09]. In the BIG trial the unadjusted ORs we calculated both in the early, 0.80 [95%CI 0.54, 1.19] and late, 0.88 [95%CI 0.62, 1.25] reports during this period found no significant difference between the tamoxifen and AI (letrozole) groups on the risk of IHD events. However there was a non significant trend for more events to be reported in the AI groups. In the BIG trial late report, the published risk ratio for

29 21 IHD events when AI (letrozole) was compared to tamoxifen was The published HR for all IHD events for AI (letrozole) relative to tamoxifen was 1.07 [95% CI 0.75, 1.51]. Our calculated ORs and the published estimates showed a non significant trend for fewer events in the tamoxifen groups relative to the AI groups. The number of events reported in each trial is shown in Table 4. Cardiac failure The ATAC trial did not report cardiac failure (CF). The BIG trial early report showed significantly less CF events in the tamoxifen group when compared to the AI (letrozole) group (p=0.01). The unadjusted OR we calculated for CF events when tamoxifen was compared to AI (letrozole) was 0.45 [95%CI 0.24, 0.84]. However the late report on the BIG showed no significant difference between the tamoxifen and AI (letrozole) groups on the risk of CF events (p=0.19). The unadjusted OR we calculated was 0.72 [95%CI 0.45, 1.16]. In the BIG trial late report the published risk ratio for CF events for AI (letrozole) relative to tamoxifen was The published HR for CF events for AI (letrozole) relative to tamoxifen was 1.38 [95% CI 0.85, 2.23]. Our calculated ORs and the published estimates showed inconsistent findings with a trend for fewer events in the tamoxifen groups relative to the AI groups. The number of events reported in each trial is shown in Table 4. Peripheral vascular disease - The ATAC trial and the BIG trial early report did not report peripheral vascular disease (PVD). In the late report of the BIG trial there were significantly less PVD events reported in the tamoxifen group when compared to the AI (letrozole) group (p=0.01). The unadjusted OR we calculated for PVD events when tamoxifen was compared to AI (letrozole) was 0.29 [95%CI 0.11, 0.79]. The published risk ratio from the BIG late report for PVD events for AI (letrozole) relative to tamoxifen was Hence relative to tamoxifen, AI (letrozole) was more than 3 times as likely to result in the development of PVD. Our calculated ORs and the published risk ratio showed significantly fewer events in the tamoxifen groups relative to the AI (letrozole) group. The number of events reported in each trial is shown in Table 4.

30 22 Risk factors for cardiovascular disease Hypertension - The ATAC trial and the BIG trial early report did not report hypertension. In the BIG trial late report, no significant difference was found between the tamoxifen and AI (letrozole) groups on the risk of hypertension (p=0.37). The unadjusted OR we calculated for hypertension when tamoxifen was compared to AI (letrozole) was 0.89 [95%CI 0.71, 1.13]. The published HR for hypertension for AI (letrozole) relative to tamoxifen was 1.10 [95% CI 0.87, 1.39]. Events related to hypercoagulability and embolism Thromboembolic events Both ATAC and BIG trials reported thromboembolic events. The ATAC trial reported significantly more thromboembolic events in the tamoxifen group compared to the AI (anastrozole) group (p=0.02). The unadjusted OR we calculated when tamoxifen was compared to AI (anastrozole) was 1.70 [95%CI 1.09, 2.64]. In the BIG trial the unadjusted ORs we calculated both in the early, OR 2.33 [95%CI 1.72, 3.16] and late, OR 2.31 [95%CI 1.73, 3.08] reports found significantly more thromboembolic events in the tamoxifen group compared to the AI (anastrozole) group. The published risk ratio in the BIG trial late report when AI (letrozole) was compared to tamoxifen was The published HR for thromboembolic events for AI (letrozole) relative to tamoxifen was 0.44 [95% CI 0.33, 0.58]. Hence relative to AI (letrozole), tamoxifen was more than twice as likely to result in the development of thromboembolism. Our calculated ORs and the published estimates showed significantly more events in the tamoxifen groups relative to the AI groups. The number of events reported in each trial is shown in Table years of follow-up Within the 3 to 5 years of follow-up period, the ATAC trial reports adverse events at 36.9 months of median follow-up while the BIG trial reported adverse events at 51 months of median follow-up.

31 23 Stroke events - Both ATAC and BIG trials reported stroke events. The ATAC trial reported significantly more ischemic stroke events in the tamoxifen group compared to the AI (anastrozole) group (p<0.001). The unadjusted OR we calculated for tamoxifen when compared to AI (anastrozole) was 2.08 [95%CI 1.38, 3.15]. The BIG trial showed no difference between the tamoxifen and AI (letrozole) groups on the risk of all stroke events (p=0.90). The unadjusted OR we calculated for all stroke episodes was 1.03 [95%CI 0.64, 1.66] with more events in the tamoxifen group compared to the AI (letrozole) group. Our calculated ORs showed inconsistent findings between the ATAC and BIG trials with a trend for more events in the tamoxifen groups relative to AI groups. The number of events reported in each trial is shown in Table 4. Other cardiovascular events related to atherosclerosis Ischemic heart disease events (IHD) Both ATAC and BIG trials reported IHD events. In the ATAC trial there was no difference found on the risk of IHD events between the tamoxifen and AI (anastrozole) groups (0.121). The unadjusted OR we calculated for IHD for tamoxifen relative to AI (anastrozole) was 0.77 [95%CI 0.56, 1.07]. The BIG trial also showed no difference between the tamoxifen and AI (letrozole) groups on the risk of IHD events (p=0.21). The unadjusted OR we calculated for IHD events was 0.76 [95%CI 0.50, 1.14]. Our calculated ORs showed a non significant trend for fewer events in the tamoxifen groups relative to the AI groups. The number of events reported in each trial is shown in Table 4. Cardiac failure The ATAC trial did not reported CF. The BIG trial showed no difference between the tamoxifen and AI (letrozole) groups on the risk of CF events (p=0.14). The unadjusted OR we calculated for CF events was 0.58 [95%CI 0.30, 1.13] with fewer events in the tamoxifen groups relative to the AI (letrozole) groups. The number of events reported in each trial is shown in Table 4.