Beta-blocker usage and breast cancer survival: a nested case-control study within a UK Clinical Practice Research Datalink cohort

Transcription

1 Published by Oxford University Press on behalf of the International Epidemiological Association ß The Author 2013; all rights reserved. International Journal of Epidemiology 2013;42: doi: /ije/dyt196 MISCELLANEOUS Beta-blocker usage and breast cancer survival: a nested case-control study within a UK Clinical Practice Research Datalink cohort Chris R Cardwell, 1 Helen G Coleman, 1 Liam J Murray, 1 Frank Entschladen 2 and Des G Powe 3 * 1 Cancer Epidemiology & Health Services Research Group, Centre for Public Health, Queen s University Belfast, Northern Ireland, 2 Institute of Immunology, ZBAF, Witten/Herdecke University, Stockumer Str. 10, Witten, Germany and 3 Department of Cellular Pathology, Queen s Medical Centre, Nottingham University Hospitals NHS Trust, UK and John van Geest Cancer Research Centre, Nottingham Trent University, Nottingham, UK *Corresponding author. Department of Cellular Pathology, Queen s Medical Centre, Nottingham University Hospitals NHS Trust, Derby Road, Nottingham, NG7 2UH, UK. des.powe@nottingham.ac.uk Accepted 28 August 2013 Background To investigate the association between post-diagnostic beta-blocker usage and risk of cancer-specific mortality in a large populationbased cohort of female breast cancer patients. Methods A nested case-control study was conducted within a cohort of breast cancer patients identified from cancer registries in England (using the National Cancer Data repository) and diagnosed between 1998 and Patients who had a breast cancer-specific death (ascertained from Office of National Statistics death registration data) were each matched to four alive controls by year and age at diagnosis. Prescription data for these patients were available through the Clinical Practice Research Datalink. Conditional logistic regression models were used to investigate the association between breast cancer-specific death and beta-blocker usage. Results Post-diagnostic use of beta-blockers was identified in 18.9% of 1435 breast cancer-specific deaths and 19.4% of their 5697 matched controls, indicating little evidence of association between beta-blocker use and breast cancer-specific mortality [odds ratio (OR) ¼ 0.97, 95% confidence interval (CI) 0.83, 1.13]. There was also little evidence of an association when analyses were restricted to cardio non-selective beta-blockers (OR ¼ 0.90, 95% CI 0.69, 1.17). Similar results were observed in analyses of drug dosage frequency and duration, and beta-blocker type. Conclusions In this large UK population-based cohort of breast cancer patients, there was little evidence of an association between post-diagnostic beta-blocker usage and breast cancer progression. Further studies which include information on tumour receptor status are warranted to determine whether response to beta-blockers varies by tumour subtypes. Keywords Breast cancer survival, beta-blockers, pharmacoepidemiology, medication, propranolol, targeted therapy 1852

2 BETA-BLOCKERS AND BREAST CANCER SURVIVAL 1853 Introduction Breast cancer is the most common cancer and the leading cause of female cancer mortality worldwide. 1 Current breast cancer treatments are not universally effective and often incur potentially harmful side effects. 2 Recently, it has been proposed that cancer progression may be prevented by beta-blocker medications, 3 5 commonly prescribed for heart disease and hypertension. 6 Metastatic spread involves complex interconnecting processes including blood and lymphatic routes 7 and may be maintained by external or tumour-paracrine norepinephrine production. 8 Numerous in vitro studies have demonstrated that beta-blockers can disrupt migratory activity and inhibit angiogenesis of cancer cells. 9,10 Propranolol, a non-selective beta-2 adrenoceptor antagonist, appears to have particularly potent anti-migratory and antiangiogenic properties in human cancer cell lines and animal models Beta-2 adrenoceptor protein expression has been shown to be predominantly expressed in human breast cancer, whereas beta-1 was not. 15 Despite accumulating evidence from laboratory studies of the potential for beta-blockers to restrict cancer progression, few epidemiological studies have been conducted. These studies have largely favoured the protective hypothesis. In 2010, the first of these reported a 71% reduction in cancer-related mortality and a 57% reduced risk of distant metastasis in 43 breast cancer patients using beta-blockers. 5 This is supported by findings from three subsequent US studies, one observing a significant 70% reduction in recurrence amongst 102 beta-blocker users, 16 one identifying a non-significantly reduced risk of cancer-specific death in 204 beta-blocker users 17 and the most recent observing a 25%, though not statistically significant, reduction in risk of cancer-specific mortality. 18 Similarly, an Irish study observed a highly significant protective association with breast cancerspecific mortality in 70 propranolol users commencing use in the year before diagnosis. 19 In contrast, a recent Danish study observed no evidence of a reduction in cancer recurrence in beta-blocker users but could not investigate cancer-specific mortality. 20 However, due to the relatively small size of most of these studies and differences in design, further investigation of any protective effect of beta-blockers in breast cancer is necessary to inform about their suitability as adjunctive therapy. Therefore, in a large population-based cohort of newly diagnosed female breast cancer patients, we examined whether post-diagnostic betablocker usage was associated with cancer-specific mortality. Methods Study design A cohort study was conducted utilizing recent linkages between the National Cancer Data Repository (NCDR), Clinical Practice Research Datalink (CPRD, formerly the GPRD) and Office of National Statistics death registration data (ONS). The NCDR contains data on all cancer patients identified in English cancer registries, including date and site of primary cancer diagnosis, stage and treatment data. The CPRD is the world s largest database of longitudinal patient records comprising around 8% of the UK population, and includes demographic information, clinical diagnoses and prescription data which are of documented high quality. 21 Ethical approval for all observational research using CPRD data has been obtained from a multicentre research ethics committee. CPRD also contains ONS mortality data providing date and cause of death up to Breast cancer cases were included if they had a CPRD breast cancer diagnosis code which was confirmed in the NCDR with a primary diagnosis of breast cancer [based upon ICD codes C50.0-C50.9 (version 10) or (version 9)] from 1998 to Individuals with a previously recorded cancer diagnosis (with the exception of in situ neoplasms and non-melanoma skin cancers) in the NCDR were excluded. The breast cancer cohort was linked to the CPRD (and ONS) data using a deterministic algorithm based upon the patient s NHS number, gender, date of birth and postcode. After linkage, patients were excluded if prescribed hormone therapy (including tamoxifen and aromatase inhibitors) more than 8 weeks prior to their cancer diagnosis, indicating an incorrect diagnosis date. Exposure data Beta-blockers were determined from GP prescribing data and defined as all drugs classified in the British National Formulary (BNF:58), chapter 2.4. Prescriptions for beta-blockers were converted to daily defined doses (DDD) on the basis of the quantity and strength of beta-blocker prescriptions. A quantity of 28 tablets, based upon the average, was assumed for approximately 5% of prescriptions where quantity was missing or implausible (<7 or 4365 tablets), to allow DDD calculation. Separate analyses were conducted in cardio non-selective beta-blockers, by intrinsic sympathomimetic activity (ISA) and by beta-blocker type, with classifications shown in Supplementary Table 1 (available as Supplementary data at IJE online). Confounders Various potential confounders were recorded. Cancer stage and surgery, chemotherapy and radiotherapy

3 1854 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY were determined in the 6 months after diagnosis on the basis of NCDR treatment data. GP prescribing data were used to determine hormone (BNF chapter , including tamoxifen and aromatase inhibitors), statin (BNF chapter 2.12), angiotensin converting enzyme inhibitor (ACEIs, BNF chapter ), angiotensin 2 receptor blocker (ARBs, BNF chapter ) and non-steroidal anti-inflammatory drug (NSAID) prescriptions (BNF chapter ) in the period after breast cancer diagnosis, and hormone replacement therapy (HRT) for estrogen and progestogens (BNF chapters and ) prior to diagnosis. Smoking, alcohol and body mass index (BMI) were determined from the earliest GP record prior to breast cancer diagnosis (records older than 10 years were recorded as missing). Comorbidities prior to diagnosis were determined from GP diagnosis codes on the basis of the eight most common diagnoses making up a recent adaptation of the Charlson comorbidity index for GPRD, 22 and Read codes identified in this article were used to identify these comorbidities. 22 Data analysis The breast cancer cohort was initially analysed using a nested case-control approach which accounts for immortal time bias. 23,24 Cases were members who had died due to breast cancer (with ICD code for breast cancer as the underlying cause of death) and these were matched on age (in 5-year intervals) and year of cancer diagnosis (in 1-year intervals) to four controls living at least as long after diagnosis as their matched case. The exposure period in cases was the period from their breast cancer diagnosis until 6 months prior to cancer-specific death. The exposure period in the controls was of the same duration as their matched cases starting from the date of their breast cancer diagnosis. Prescriptions in the 6-month period prior to death were removed as these may reflect end-of-life treatment or increased exposure to healthcare professionals. Consequently, analysis was restricted to individuals with over 1 year from diagnosis to cancer-specific death or 1 year from diagnosis to censoring (due to death from other causes or end of ONS coverage). Conditional logistic regression calculated odds ratios (ORs) and 95% confidence intervals (95% CIs). Adjusted analyses were conducted including potential confounders. Various sensitivity analyses were conducted. Analyses were conducted restricting the exposure period to 12, 24 and 36 months prior to cancer-specific death and excluding prescriptions in the first 12 months after diagnosis. Analyses were also conducted stratified by pre-diagnostic use of beta-blockers, stage and hormone therapy. Further analysis was conducted comparing beta-blockers users with only non-beta-blocker users who used other antihypertensive medications in the exposure period, such as diuretics, vasodilator antihypertensive drugs, centrally acting antihypertensive drugs, alphaadrenoceptor blocking drugs, ACEIs, ARBs, renin inhibitors and calcium channel blockers. Further sensitivity analyses were conducted restricted to registries with higher rates of available stage information. All stratified analyses were conducted after re-matching cases to controls within the strata of interest. In the main adjusted analysis, a missing category indicator was used when adjusting for smoking; repeating this analysis using only complete cases produced identical results (data not shown). In another analysis, breast cancer-specific death was based upon a breast cancer ICD code recorded as any cause of death in ONS data, rather than just the underlying cause of death as used in the main analysis. Also, a Cox proportional hazards model was applied to the entire breast cancer cohort, investigating beta-blocker exposure as a time-varying covariate (in which an individual was a non-user until first use and then remained a user until end of follow-up, applying a 6-month lag to mimic the case-control analysis) 23 and also adjusting for the competing risk of deaths from other causes, using competing-risks regression utilizing the proportional subhazards model. 25 The study protocol was registered on the European Network of Centres for Pharmacoepidemiology and Pharmacovigilance (ENCEPP) database ( encepp.eu/encepp/viewresource.htm?id¼2573). The original power calculation estimated 1700 cancer-specific deaths, 8500 controls and beta-blocker usage at around 20%, allowing over 80% power to detect as significant at the 5% level an OR of 0.80 in patients receiving beta-blockers, equivalent to a 20% risk reduction in cancer-specific death. Although observed numbers were slightly lower (1435 cancer-specific deaths and 5697 controls), power remained over 80%. Analyses were conducted using STATA (StataCorp, TX). Results Patient characteristics Overall, there were primary breast cancer cases newly diagnosed between 1998 and 2007 identified in NCDR and linked to CPRD. The final cohort for analysis contained 9817 breast cancer cases (restricted as shown in Figure 1), with a median of 6.0 years of follow-up, in whom there were 1443 cancer-specific deaths and 2274 deaths from any cause. This cohort was converted to nested case-control data with 1435 cancer-specific deaths and 5697 controls. Table 1 shows characteristics of cancer-specific deaths and controls. The average time to death in the cases was 3.9 years and varied from 1 to 12.8 years. Patients in the cancer-specific deaths group had higher stage, less

4 BETA-BLOCKERS AND BREAST CANCER SURVIVAL breast cancer patients with CPRD coverage and no prior history of cancer* Exclusions: 884 breast cancer diagnoses preceded CPRD research quality records 116 death registration data unavailable 308 hormone therapy records preceded breast cancer diagnosis by >8weeks 728 patients had <1 year follow-up 9827 breast cancer patients 1443 breast cancer specific deaths 1435 breast cancer-specific deaths: cases matched 1:4 ratio to 5697 controls (N.B. 8 cases had 3 matched controls, 7 cases had 2 matched controls, 7 cases had 1 matched control) * Excluding in situ neoplasms and non-melanoma skin cancers Figure 1 Patient selection for analysis Exclusion: 8 patients for whom no suitable control could be identified surgery and were more likely to receive chemotherapy and to smoke compared with controls. Other characteristics were generally similar between cases and controls (Table 1). Association between beta-blocker use and survival The association between beta-blocker usage and cancer-specific death is shown in Table 2. Overall, the proportion of beta-blocker users after cancer diagnosis was similar in the cancer-specific deaths compared with controls (18.9% vs 19.4%, respectively) indicating little evidence of association (OR ¼ 0.97, 95% CI 0.83, 1.13). There was little evidence of dose response by prescriptions for beta-blockers or by DDDs. There was also little evidence of association by type or when analyses were restricted to cardio non-selective (beta-1/beta-2 acting) beta-blockers (OR ¼ 0.90, 95% CI 0.69, 1.17), or beta-blockers which exhibit intrinsic sympathomimetic activity (OR ¼ 0.80, 95% CI 0.27, 2.34). Adjusting for potential confounders, including stage, did not markedly alter these estimates. Largely similar associations were observed for all-cause mortality (Supplementary Table 2, available as Supplementary data at IJE online). Sensitivity analyses Sensitivity analyses are shown in Table 3. The findings were little altered when the period of exposure was changed after cancer diagnosis, when stratifying by stage, when stratifying by pre-diagnosis betablocker use, when the analysis was restricted to individuals receiving hormone therapy (a proxy for hormone receptor status), when restricting nonusers to individuals who used other antihypertensive medications in the exposure period, or when breast cancer-specific death was based upon breast cancer recorded as any cause of death (rather than just as the main underlying cause). Additional analyses further investigated the potential for confounding by stage. There was no association (P ¼ 0.19) between stage and beta-blocker prescriptions in the 1st year after cancer diagnosis in the main cohort with beta-blockers received by 14% (283/2010) of stage 1 patients, 14% (338/2357) of stage 2, 14% (52/361) of stage 3, 15% (15/97) of stage 4 and 14% (701/4992) of patients with missing stage. An analysis (Table 3) restricted to cancer registries with higher levels of available stage (available for over 85% of individuals) demonstrated no association with beta-blocker usage after adjustment for stage (OR ¼ 1.19). Beta-blocker usage in the year prior to cancer diagnosis was also investigated (in 1559 cases and 6186 controls with at least 1 year of prescription records prior to diagnosis). Overall, considering prescriptions in the year prior to diagnosis, there was no effect in beta-blocker users (14.2% usage in cases and 14.6% usage in controls, OR ¼ 0.97, 95% CI 0.83, 1.14), cardio non-selective beta-blocker users (OR ¼ 1.18,

5 1856 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY Table 1 Characteristics of breast cancer patients (identified from cancer registry data) who die from breast cancer (cases) compared with controls Characteristics Cancer-specific deaths n (%) Controls n (%) P-value [n ¼ 1435] [n ¼ 5697] Year of cancer diagnosis (34.8) 1979 (34.7) Matched (36.1) 2060 (36.2) (29.1) 1658 (29.1) Age at cancer diagnosis (years) <40 92 (6.4) 352 (6.2) Matched (17.2) 985 (17.3) (19.9) 1144 (20.1) (19.0) 1092 (19.2) (22.2) 1271 (22.3) (12.6) 722 (12.7) (2.7) 131 (2.3) Exposure period (years): mean (SD) 3.9 (2.3) 3.9 (2.3) Matched range Stage 1 72 (11.0) 1133 (41.8) < (61.5) 1379 (50.8) (17.7) 167 (6.2) 4 64 (9.8) 35 (1.3) Missing Treatment within 6 months of cancer diagnosis Surgery 1087 (75.8) 4842 (85.0) <0.001 Chemotherapy 573 (39.9) 1344 (23.6) <0.001 Radiotherapy 2703 (47.5) 700 (48.8) 0.31 Smoking prior to cancer diagnosis Non-smoker 710 (59.9) 3037 (64.0) 0.03 Ex-smoker 226 (19.1) 832 (17.5) Current smoker 249 (21.0) 874 (18.4) Missing Alcohol prior to cancer diagnosis Never consumed alcohol 203 (19.6) 806 (19.2) 0.94 Alcohol consumer 831 (80.4) 3399 (80.8) Missing BMI (kg/m 2 ) prior to cancer diagnosis Mean (SD) 26.6 (5.4) 26.3 (5.1) 0.04 Missing Comorbidity prior to cancer diagnosis Cerebrovascular disease 67 (4.7) 219 (3.8) 0.15 Chronic pulmonary disease 249 (17.4) 906 (15.9) 0.18 Congestive heart disease 38 (2.7) 140 (2.5) 0.72 Diabetes 81 (5.6) 274 (4.8) 0.18 Myocardial infarction 20 (1.4) 79 (1.4) 0.97 Peptic ulcer disease 35 (2.4) 145 (2.6) 0.81 Peripheral vascular disease 35 (2.4) 73 (1.3) Rheumatological disease 62 (4.3) 187 (3.3) 0.07

6 BETA-BLOCKERS AND BREAST CANCER SURVIVAL 1857 Table 2 Exposure to beta-blockers post diagnosis (excluding 6 months prior to death) in breast cancer patients who die from breast cancer (cases) compared with controls Cancer-specific deaths n (%) Controls n (%) Unadjusted OR (95% CI) Model 1 adjusted OR (95% CI) Model 2 adjusted OR (95% CI) Exposure Number of cases/controls 1435/ / /2565 Prescriptions for beta-blockers post diagnosis (81.1) 4590 (80.6) 1.00 (ref. cat.) 1.00 (ref. cat.) 1.00 (ref. cat.) 1 or more 271 (18.9) 1107 (19.4) 0.97 (0.83, 1.13) 1.11 (0.94, 1.32) 1.20 (0.92, 1.57) Prescriptions for beta-blockers post diagnosis (81.1) 4590 (80.6) 1.00 (ref. cat.) 1.00 (ref. cat.) 1.00 (ref. cat.) (8.2) 493 (8.7) 0.94 (0.76, 1.17) 1.08 (0.85, 1.36) 1.19 (0.83, 1.71) 12 or more 154 (10.7) 614 (10.8) 0.99 (0.82, 1.21) 1.15 (0.92, 1.44) 1.21 (0.87, 1.70) Beta-blockers prescriptions post diagnosis in daily defined doses (DDDs) 0 DDDs 1164 (81.1) 4590 (80.6) 1.00 (ref. cat.) 1.00 (ref. cat.) 1.00 (ref. cat.) DDDs 143 (10.0) 612 (10.7) 0.93 (0.76, 1.13) 1.06 (0.86, 1.32) 1.08 (0.76, 1.52) 366 or more DDDs 128 (8.9) 495 (8.7) 1.03 (0.83, 1.27) 1.18 (0.93, 1.50) 1.36 (0.96, 1.93) Cardio non-selective a 70 (4.9) 309 (5.4) 0.90 (0.69, 1.17) 0.96 (0.72, 1.28) 1.13 (0.71, 1.81) ISA activity b 4 (0.3) 20 (0.4) 0.80 (0.27, 2.34) 0.80 (0.24, 2.62) 0.43 (0.05, 4.04) Metoprolol 13 (0.9) 32 (0.6) 1.60 (0.84, 3.04) 1.48 (0.74, 2.97) 1.14 (0.29, 4.55) Propranolol 42 (2.9) 207 (3.6) 0.80 (0.57, 1.13) 0.87 (0.61, 1.24) 0.98 (0.57, 1.71) Atenolol 172 (12.0) 682 (12.0) 1.01 (0.84, 1.21) 1.17 (0.95, 1.44) 1.14 (0.84, 1.56) Carvedilol 7 (0.5) 15 (0.3) 1.87 (0.76, 4.58) 1.91 (0.72, 5.10) 2.66 (0.21, 34.1) Sotalol 17 (1.2) 67 (1.2) 1.00 (0.59, 1.71) 1.02 (0.57, 1.80) 2.02 (0.75, 5.44) Bisoprolol 40 (2.8) 149 (2.6) 1.07 (0.75, 1.53) 1.24 (0.83, 1.85) 1.09 (0.59, 2.02) a Includes propranolol, sotolol, timolol, nadolol, carvidolol, pindolol, oxprenolol and labetolol. b Includes acebutolol, labetolol, pindolol, oxprenolol. c Model 1 includes surgery within 6 months of diagnosis, chemotherapy within 6 months, radiotherapy within 6 months, tamoxifen (post diagnosis), aromatase inhibitors (post diagnosis), NSAID use (post-diagnosis), ACEI use (post-diagnosis), ARB use (post-diagnosis), statin use (post-diagnosis), hormone replacement therapy (pre-diagnosis), comorbidities (pre-diagnosis, including myocardial infarction, cerebrovascular disease, congestive heart disease, chronic pulmonary disease, peripheral vascular disease, peptic ulcer disease and diabetes), and smoking (pre-diagnosis, with missing included as a category). d Model 2 additionally includes stage, restricted to individuals with an available stage record.

7 1858 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY Table 3 Sensitivity analysis for association between beta-blocker usage and breast cancer-specific death in breast cancer patients Cancer-specific Comparison a deaths Controls OR (95% CI) beta-blocker users vs non-users OR (95% CI) 1 11 beta-blocker prescriptions vs none OR (95% CI) 12þ beta-blocker prescriptions vs none Main analysis: diagnosis to 6 months prior to death (0.83, 1.13) 0.94 (0.76, 1.17) 0.99 (0.82, 1.21) Patients with no pre-diagnostic beta-blocker use b (0.66, 1.11) 0.76 (0.55, 1.06) 1.01 (0.68, 1.51) Patients with pre-diagnostic beta-blocker use b (0.81, 1.88) 1.34 (0.80, 2.24) 1.18 (0.75, 1.85) Diagnosis to 1 year prior to death c (0.83, 1.16) 0.97 (0.77, 1.23) 0.99 (0.80, 1.22) Diagnosis to 2 years prior to death d (0.84, 1.22) 0.85 (0.65, 1.11) 1.18 (0.93, 1.50) Diagnosis to 3 years prior to death e (0.89, 1.39) 1.25 (0.94, 1.66) 0.98 (0.72, 1.34) 1 year after diagnosis to 6 months prior to death f (0.83, 1.18) 0.98 (0.76, 1.26) 0.99 (0.80, 1.24) Stage 1 and 2 breast cancer patients only (0.82, 1.36) 1.09 (0.76, 1.55) 1.03 (0.74, 1.44) Stage 3 and 4 breast cancer patients only (0.59, 1.56) 0.76 (0.37, 1.57) 1.15 (0.62, 2.14) Restricted to patients receiving hormone therapy (0.84, 1.16) 0.96 (0.76, 1.22) 1.01 (0.82, 1.24) Patients with a recorded stage from cancer registries (0.90, 1.57) 1.02 (0.68, 1.52) 1.35 (0.94, 1.94) with high rates of stage recording (adjusted for stage) g Restricting analysis to users of other (0.77, 1.11) 0.90 (0.70, 1.16) 0.94 (0.77, 1.18) antihypertensive medications h Including breast cancer-specific deaths where breast (0.87, 1.16) 0.98 (0.81, 1.20) 1.02 (0.85, 1.22) cancer is recorded as any cause of death Time-varying covariate analysis i (0.81, 1.06) 0.91 (0.75, 1,10) 0.95 (0.80, 1.10) Time-varying covariate analysis j (0.83, 1.08) 0.91 (0.75, 1.10) 0.98 (0.82, 1.16) a All sensitivity analyses refer to beta-blocker usage in the time period from breast cancer diagnosis to 6 months before death, and are adjusted for matching criteria only, unless otherwise stated. b Pre-diagnostic beta-blocker use in 2 years prior to breast cancer diagnosis, restricted to individuals with 2 years of medication records prior to diagnosis. c Restricted to individuals with at least 1.5 years of follow-up so relevant exposure period is at least a duration of 6 months. d Restricted to individuals with at least 2.5 years of follow-up so relevant exposure period is at least a duration of 6 months. e Restricted to individuals with at least 3.5 years of follow-up so relevant exposure period is at least a duration of 6 months. f Restricted to individuals with at least 2 years of follow-up so relevant exposure period is at least a duration of 6 months. g Stage missing for 15% of individuals and breast cancer patients included from the Northern and Yorkshire Cancer Registry and Information Service, the Trent Cancer Registry, the Eastern Cancer Registration and Information Centre, the Oxford Cancer Intelligence Unit and the West Midlands Cancer Intelligence. h Antihypertensive medications include diuretics, vasodilator antihypertensive drugs, centrally acting antihypertensive drugs, alpha-adrenoceptor blocking drugs, angiotensin converting enzyme inhibitors, angiotensin 2 receptor antagonists, renin inhibitors and calcium channel blockers. i Reported estimates are hazard ratios and 95% CIs, adjusted for age (as continuous) and year of breast cancer diagnosis (as continuous). j Reported estimates are subdistribution hazards ratios and 95% CIs, adjusted for competing risks of death, age (as continuous) and year of breast cancer diagnosis (as continuous).

8 BETA-BLOCKERS AND BREAST CANCER SURVIVAL % CI 0.85, 1.63), propranolol users (OR ¼ 1.18, 95% CI 0.79, 1.75) or individuals with 12 or more betablockers prescriptions in the previous year (OR ¼ 0.92, 95% CI 0.67, 1.25). Repeating analysis using a Cox proportional hazards model with a time-varying covariate produced similar estimates before [hazards ratio (HR) ¼ 0.93, 95% CI 0.81, 1.06] and after additional adjustment for competing risks of deaths from other causes (subdistribution HR ¼ 0.95, 95% CI 0.83, 1.08). Discussion In this large UK population-based cohort of newly diagnosed breast cancer patients, there was little evidence of an association between post-diagnostic betablocker usage and breast cancer progression. Our findings largely agree with those of a UK study that showed no association between beta-blocker usage prior to diagnosis and all-cause mortality in 984 breast cancer patients 26 and a recent Danish study which showed no evidence of a reduction in cancer recurrence in beta-blocker users, 20, but contrasts with earlier epidemiological findings. A study of 466 consecutive breast cancer patients in Nottingham 5 reported that patients receiving betablockers for hypertension had a 71% reduced risk of cancer-associated mortality (HR ¼ 0.29, 95% CI 0.12, 0.72). An Irish study observed an 81% (HR ¼ 0.19, 95% CI 0.06 to 0.60) reduction in rates of breast cancer-specific death in 70 propranolol users in the year before diagnosis. 19 Similarly, a US study observed improved disease-free survival rates particularly in triple-negative breast cancer patients, although that analysis incorporated only 29 betablocker users. 16 Finally, another US study 17 showed a non-significant reduction in the risk of cancer-specific death in 204 individuals using beta-blockers, but not ACE-inhibitors (HR ¼ 0.76, 95% CI 0.44, 1.33), and more recently another US study 18 observed reductions in beta-blocker users [relative risk (RR) ¼ 0.76, 95% CI 0.54, 1.05). These inconsistent results could have a number of explanations. The timing of exposure could explain some of the variation because many of the earlier studies investigated beta-blocker use prior to diagnosis. 5,19,26.Our main analysis investigated betablocker use after diagnosis, arguably a more clinically relevant exposure period, but no association was observed with pre-diagnostic beta-blocker usage. Alternatively, as suggested recently, 27 heterogeneity due to beta-adrenergic receptor expression, and estrogen receptor (ER), progesterone receptor (PR) and human epidermal growth factor receptor-2 (HER2) status, might be important, particularly as one study observed marked effects in triple-negative breast cancer patients. 16 The types of beta-blocker investigated could also explain the variation. In our study, most beta-blockers used displayed beta-1 adrenoceptor activity. We previously hypothesized that beta-2 specific beta-blockers, which are noncardio-selective, may be of more importance. 15,28 However, previous studies 5,16 have observed marked protective effects for largely cardio-selective betablockers and we observed only a weak inverse association with cardio non-selective beta-blockers, that did not achieve statistical significance. This study has several strengths. Our investigation contains the largest number of cancer-specific deaths studied to date for beta-blocker users in breast cancer patients followed up over a relatively long time period. Importantly, the linkage with NCDR and ONS allowed robust verification of cancer diagnosis and death data. GP prescription data were recorded prior to and without knowledge of cancer-specific mortality, enabling temporal associations to be explored and eliminating the potential for recall bias incurred in questionnaire-based studies. However, the beta-blocker drug data reflect GP prescriptions, not drugs actually consumed. Nevertheless, null associations were also observed for individuals with multiple prescriptions, suggesting that non-compliance is not greatly impacting upon the results. NSAIDs, but importantly not beta-blockers, are available over the counter, potentially causing misclassification of this confounder. We were also unable to stratify analysis by the molecular subtype of the tumours studied although no association was observed in patients receiving hormone therapy (a proxy for estrogen receptor hormone status). A further limitation was our inability to study breast cancer metastasis or recurrence as an outcome, since this information is not routinely collected by English cancer registries. Although it is possible that some deaths have been incorrectly misclassified when determining cancer-specific mortality, a recent methodological study indicated that in comparative studies in which differential misclassification between groups is unlikely, like ours, such misclassification is unlikely to impact on effect estimates. 29 There is also the possibility of residual confounding, although we accounted for many confounders. Although stage data were incomplete, it seems unlikely that stage was confounding the main findings since we found no association between stage and beta-blocker usage and adjustments for stage had little influence in analyses of patients for whom stage was available. Confounding by indication can be problematic but it is not obvious how such confounding could obscure any protective effect of beta-blockers, and findings were similar when beta-blocker users were compared with users of other antihypertensive medications to reduce the possible impact. In conclusion, the findings of this large populationbased study do not support a universal role for betablocker usage in protecting against breast cancer progression. Further studies which include information on tumour receptor status are required.

9 1860 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY Supplementary Data Supplementary data are available at IJE online. Funding This work was supported by a project funding grant from Cancer Research-UK [C19630/A13265]. F.E. was supported by the Fritz Bender Foundation (Munich, Germany). Acknowledgements This study is based in part on data from the General Practice Research Database obtained under licence from the UK Medicines and Healthcare Regulatory Agency. However, the interpretation and conclusions contained in study are those of the authors alone. Conflict of interest: None declared. KEY MESSAGES Numerous in vitro studies have demonstrated that beta-blockers can disrupt migratory activity and inhibit angiogenesis of cancer cells, suggesting that these drugs could have a role in the treatment of patients with breast cancer. In this large UK population-based cohort of breast cancer patients, there was little evidence of an association between beta-blocker usage and breast cancer-specific mortality. References 1 Youlden DR, Cramb SM, Dunn NA, Muller JM, Pyke CM, Baade PD. The descriptive epidemiology of female breast cancer: an international comparison of screening, incidence, survival and mortality. Cancer Epidemiol 2012;36: Ewertz M, Jensen AB. Late effects of breast cancer treatment and potentials for rehabilitation. Acta Oncol 2011;50: Ganz PA, Cole SW. Expanding our therapeutic options: Beta blockers for breast cancer? J Clin Oncol 2011;29: Entschladen F, Drell TL 4th, Lang K, Joseph J, Zaenker KS. Neurotransmitters and chemokines regulate tumor cell migration: potential for a new pharmacological approach to inhibit invasion and metastasis development. Curr Pharm Des 2005;11: Powe DG, Voss MJ, Zanker KS et al. Beta-blocker drug therapy reduces secondary cancer formation in breast cancer and improves cancer specific survival. Oncotarget 2010;1: Aronow WS. Current role of beta-blockers in the treatment of hypertension. Expert Opin Pharmacother 2010;11: Cao Y. Tumor angiogenesis and molecular targets for therapy. Front Biosci 2009;14: Wong HP, Yu L, Lam EK, Tai EK, Wu WK, Cho CH. Nicotine promotes cell proliferation via alpha7- nicotinic acetylcholine receptor and catecholaminesynthesizing enzymes-mediated pathway in human colon adenocarcinoma HT-29 cells. Toxicol Appl Pharmacol 2007;221: Lang K, Drell TL 4th, Lindecke A et al. Induction of a metastatogenic tumor cell type by neurotransmitters and its pharmacological inhibition by established drugs. Int J Cancer 2004;112: Pasquier E, Ciccolini J, Carre M et al. Propranolol potentiates the anti-angiogenic effects and anti-tumor efficacy of chemotherapy agents: implication in breast cancer treatment. Oncotarget 2011;2: Annabi B, Lachambre MP, Plouffe K, Moumdjian R, Beliveau R. Propranolol adrenergic blockade inhibits human brain endothelial cells tubulogenesis and matrix metalloproteinase-9 secretion. Pharmacol Res 2009;60: Benish M, Bartal I, Goldfarb Y et al. Perioperative use of beta-blockers and COX-2 inhibitors may improve immune competence and reduce the risk of tumor metastasis. Ann Surg Oncol 2008;15: Sloan EK, Priceman SJ, Cox BF et al. The sympathetic nervous system induces a metastatic switch in primary breast cancer. Cancer Res 2010;70: Palm D, Lang K, Niggemann B et al. The norepinephrinedriven metastasis development of PC-3 human prostate cancer cells in BALB/c nude mice is inhibited by betablockers. Int J Cancer 2006;118: Powe DG, Voss MJ, Habashy HO et al. Alpha- and betaadrenergic receptor (AR) protein expression is associated with poor clinical outcome in breast cancer: an immunohistochemical study. Breast Cancer Res Treat 2011;130: Melhem-Bertrandt A, Chavez-Macgregor M, Lei X et al. Beta-blocker use is associated with improved relapsefree survival in patients with triple-negative breast cancer. J Clin Oncol 2011;29: Ganz PA, Habel LA, Weltzien EK, Caan BJ, Cole SW. Examining the influence of beta blockers and ACE inhibitors on the risk for breast cancer recurrence: results from the LACE cohort. Breast Cancer Res Treat 2011;129: Holmes MD, Hankinson SE, Feskanich D, Chen WY. Beta blockers and angiotensin-converting enzyme inhibitors purported benefit on breast cancer survival may be explained by aspirin use. Breast Cancer Res Treat 2013;139: Barron TI, Connolly RM, Sharp L, Bennett K, Visvanathan K. Beta blockers and breast cancer mortality: a population- based study. J Clin Oncol 2011;29: Sorensen GV, Ganz PA, Cole SW et al. Use of ß-Blockers, Angiotensin-Converting Enzyme Inhibitors, Angiotensin

10 BETA-BLOCKERS AND BREAST CANCER SURVIVAL 1861 II Receptor Blockers, and Risk of Breast Cancer Recurrence: A Danish Nationwide Prospective Cohort Study. J Clin Oncol 2013;31: Jick H, Jick SS, Derby LE. Validation of information recorded on general practitioner based computerised data resource in the United Kingdom. BMJ 1991;302: Khan NF, Perera R, Harper S, Rose PW. Adaptation and validation of the Charlson Index for Read/OXMIS coded databases. BMC Fam Pract 2010;11:1. 23 Levesque LE, Hanley JA, Kezouh A, Suissa S. Problem of immortal time bias in cohort studies: example using statins for preventing progression of diabetes. BMJ 2010;340: b Etminan M, Samii A. Pharmacoepidemiology I: a review of pharmacoepidemiologic study designs. Pharmacotherapy 2004;24: Fine JP, Gray RJ. A proportional hazards model for the subdistribution of a competing risk. J Am Statist Assoc 1999;94: Shah SM, Carey IM, Owen CG, Harris T, Dewilde S, Cook DG. Does beta-adrenoceptor blocker therapy improve cancer survival? Findings from a population-based retrospective cohort study. Br J Clin Pharmacol 2011;72: Barron TI, Sharp L, Visvanathan K. Beta-adrenergic blocking drugs in breast cancer: a perspective review. Ther Adv Med Oncol 2012;4: Powe DG, Entschladen F. Targeted therapies: Using betablockers to inhibit breast cancer progression. Nat Rev Clin Oncol 2011;8: Sarfati D, Blakely T, Pearce N. Measuring cancer survival in populations: relative survival vs cancer-specific survival. Int J Epidemiol 2010;39: