Treatment and survival from breast cancer: the experience of patients at South Australian teaching hospitals between 1977 and 2003

Transcription

1 Blackwell Publishing LtdOxford, UKJEPJournal of Evaluation in Clinical Practice Blackwell Publishing Ltd Original Article Treatment and survival from breast cancerc. Luke et al. Journal of Evaluation in Clinical Practice ISSN Treatment and survival from breast cancer: the experience of patients at South Australian teaching hospitals between 1977 and 2003 Colin Luke MB BS MPH FAFPHM, 1 Grantley Gill MD FRACS, 2 Stephen Birrell BM BS FRACS PhD, 3 Vlad Humeniuk MB BS FRACS, 4 Martin Borg MD FRANZCR, 5 Christos Karapetis MB BS MMedSc (ClinEpi) FRACP, 6 Bogda Koczwara BM BS FRACP Mbioethics, 7 Ian Olver MD PhD CMin FRACP FAChPM MRACMA, 8 Michael Penniment MB BS FRANZCR, 9 Ken Pittman MB BS MD FRACP, 10 Tim Price MB BS FRACP, 11 David Walsh MB BS FRACS, 12 Eng Kiat (Eric) Yeoh MD FRCP FRCR 13 and David Roder BDS DDSc MPH 14 1 Senior Medical Consultant, Clinical Epidemiologist, Epidemiology Branch, Department of Health, South Australian Government, Adelaide, SA, Australia 2 Surgeon, Head, Breast Endocrine & Surgical Oncology Unit, Royal Adelaide Hospital Cancer Centre and Department of Surgery, University of Adelaide, Adelaide, SA, Australia 3 Surgeon, Director, Flinders Medical Centre Breast Cancer Unit, Adelaide, SA, Australia 4 Surgeon, Senior Visiting Surgeon, Breast Endocrine Unit, Department of Surgery, Queen Elizabeth Hospital, Adelaide, SA, Australia 5 Radiation Oncologist, Principal Consultant, Department of Radiation Oncology, Royal Adelaide Hospital Cancer Centre, Adelaide, SA, Australia 6 Medical Oncologist, Senior Consultant Medical Oncologist, Department of Medical Oncology, Flinders Medical Centre, Adelaide, SA, Australia 7 Medical Oncologist, Head, Department of Medical Oncology, Flinders Medical Centre, Adelaide, SA, Australia 8 Medical Oncologist, The Cancer Council Professor of Cancer Care, University of Adelaide, and Clinical Director, Royal Adelaide Hospital Cancer Centre, Adelaide, SA, Australia 9 Radiation Oncologist, Staff specialist, Department of Radiation Oncology, Royal Adelaide Hospital Cancer Centre, Adelaide, SA, Australia 10 Medical Oncologist, Director of Medical Oncology, Queen Elizabeth Hospital, Adelaide, SA, Australia 11 Medical Oncologist, Head, Oncology Research and Cancer Registry, Queen Elizabeth Hospital, Adelaide, SA, Australia 12 Surgeon, Head, Breast Endocrine Surgical Unit, Queen Elizabeth Hospital, Adelaide, SA, Australia 13 Radiation Oncologist, Director, Department of Radiation Oncology, Royal Adelaide Hospital Cancer Centre, Adelaide, SA, Australia 14 Epidemiologist, Consultant Epidemiologist, The Cancer Council South Australia (TCCSA), Adelaide, SA, Australia Keywords breast cancer, registries, survival trends, treatment patterns Correspondence David Roder The Cancer Council South Australia (TCCSA) PO Box 929 Unley SA 5061 Australia droder@cancersa.org.au Accepted for publication: 30 August 2005 doi: /j x Abstract Rationale Treatment guidelines recommend a more conservative surgical approach than mastectomy for early stage breast cancer and a stronger emphasis on adjuvant therapy. Registry data at South Australian teaching hospitals have been used to monitor survivals and treatment in relation to these guidelines. Aims and objectives To use registry data to: (1) investigate trends in survival and treatment; and (2) compare treatment with guidelines. Methods Registry data from three teaching hospitals were used to analyse trends in primary courses of treatment of breast cancers during (n = 4671), using univariate analyses and multiple logistic regression. Disease-specific survivals were analysed using Kaplan Meier product limit estimates and multivariable Cox proportional hazards regression. Results The 5-year survival was 79.9%, but with a secular increase, reaching 83.6% in The relative risk of death (95% confidence limits) was 0.74 (0.62, 0.88) for , compared with previous diagnoses, after adjusting for tumour node metastasis stage, grade, age and place of residence. Treatment changes included an increase in conservative surgery (as opposed to mastectomy) from 51.7% in to 76.8% in for stage I (P < 0.001) and from 31.1% to 52.2% across these periods for stage II (P < 0.001). Adjuvant radiotherapy also became more common (P < 0.001), with 20.6% of patients receiving this treatment in compared with 60.7% in Radiotherapy generally was more prevalent when conservative surgery was provided, Blackwell Publishing Ltd, Journal of Evaluation in Clinical Practice 13 (2007)

2 C. Luke et al. Treatment and survival from breast cancer although also relatively common in mastectomy patients when tumour diameters exceeded 50 mm or when there were four or more involved nodes. The proportion of patients receiving chemotherapy increased (P < 0.001), from 19.6% in to 36.9% in , and the proportion having hormone therapy also increased (P < 0.001), from 34.3% to 59.4% between these periods. Conclusions Survivals appear to be increasing and treatment trends are broadly consistent with guideline directions, and the earlier research on which these recommendations were based. Introduction Breast cancer has been the most commonly notified cancer and leading cause of cancer death in Australian women since cancer registration began in the 1970s [1,2]. Attempts to improve treatment outcomes have included earlier detection initiatives, primarily mammography screening [3], and promoting evidence-based treatment guidelines [4,5]. Population-based screening mammography was introduced in the early 1990s and has accounted for an increasing proportion of cancer diagnoses [6]. By 2001, 30% of invasive breast cancers in South Australia (SA) were being found through screening mammography [7]. The effect of promoting early detection appears to have been pronounced, in that the proportion of cancers found when small (diameter <15 mm) has increased threefold since the early 1980s [8]. An annual reduction in breast cancer mortality of 2.2% occurred in Australian women between 1991 and 2001 [1]. Although this reduction appeared to strengthen in the late 1990s, its commencement was too early to be attributed solely to screening mammography [1,9]. Similarly, a corresponding mortality decline in Britain from 1989 was regarded as starting too early to be due entirely to screening, with treatment gains being cited as a likely contributor [10]. Other researchers, reporting a fall in breast cancer mortality in Australia in the early 1990s, expressed the view that improvements in both stage at diagnosis and treatment could have been responsible [11]. A treatment contribution to mortality reductions seems plausible, given the attention directed at quality of care. In 1995, the National Health and Medical Research Council and National Breast Cancer Centre released treatment guidelines, and further guidelines and updates were released in 2001 [4,5]. These guidelines were a culmination of earlier research results, which had been formally reviewed by a parliamentary committee and led to a range of clinical recommendations [12]. They included adoption of a more conservative approach to surgical management than mastectomy for early stage disease and a stronger emphasis on adjuvant therapy [4,5]. Quality assurance initiatives also have included the introduction of a National Breast Surgery Audit for cancer treatment, which has been endorsed by the Royal Australian College of Surgeons for ongoing monitoring of surgical practice in Australia and New Zealand [13]. The Audit has extended progressively and would now be covering about 30% of newly diagnosed breast cancers. Because there has not been an ongoing collection of treatment data throughout Australia, we are now presenting data from South Australian hospital-based cancer registries to indicate secular trends in treatment and survival at these centres [14]. While not selected to be nationally representative, the data show trends at teaching hospitals in one of the country s eight states and territories that may be indicative of broader trends. These registries were introduced in 1987 and have been used to monitor survivals and treatment by stage and other clinical characteristics, and for research and hospital administration. Earlier results showed that 35% of breast cancers diagnosed in SA in the 1990s were recorded in these registries [14]. Although the teaching hospitals associated with these registries were tertiary referral centres, the tumour node metastasis (TNM) stage profiles of their patients have been similar to those reported for Australia as a whole. For example, surgical patients at these hospitals in had similar stage profiles to surgical patients in a 1995 national survey, in that respective percentages were for: stage I 44% and 40%; stage II 44% and 48%; stage III 8% and 8%; and stage IV 4% and 4% [14,15]. Survivals from breast cancers treated at these hospitals also have been similar to survivals for all South Australian patients [16]. For example, data indicated a 5-year survival of 80% for patients at these hospitals, compared with 78% for the State as a whole in There is also evidence that South Australian survivals are similar to national figures, in that the 5- year survival in SA was 82% in and 84% in , compared with the 84% for Australia overall in [14,17]. Data collected prospectively by our hospital registries since 1987, and retrospectively for , are used now to show survival and treatment trends, both for general benchmarking and to indicate extent of concordance with treatment guidelines and with the earlier evidence on which these guidelines were based [4,5]. As the coverage of retrospective data was less comprehensive than for prospective data, they may be less reliable and should be interpreted more cautiously. Methods Data sources and coding The operational procedures of these hospital registries have been described previously [14]. The registries were established with approval of hospital ethics committees and authorized under section 64D of the South Australian Health Commission Act Blackwell Publishing Ltd 213

3 Treatment and survival from breast cancer C. Luke et al. The registries seek initial core data on age, sex, residential address and other person descriptors, organ site of cancer, histological type, date of diagnosis, and where relevant, date and cause of death, from the State Cancer Registry, multidisciplinary tumour meetings, pathology reporting systems and other hospital sources [14]. Postcode data are used to classify patients as metropolitan (State capital) or non-metropolitan (country) residents [14]. The registries then add data on, grade, diameter, nodal status, primary course of treatment, and immediate treatment outcomes, obtained from tumour meetings, case notes, and individual clinicians [14]. Clinical data items were classified in the present study period according to conventional international coding practices and the American Commission on Cancer Registry Operations and Data Standards [18,19]. Where available, pathological stage was used in preference to clinical stage, according to the best stage practice used by North American registries [20]. Where available in medical records (76% of breast cases), oestrogen and progesterone receptor expression were recorded as described in earlier reports as the percentages of nuclei that were positive for the relevant receptor [14]. These percentages were categorized as: negative up to 10%; positive (low) higher but not exceeding 75%; and positive (high) over 75%. Death data provided to hospital registries by the State Registry were obtained through electronic linkage with State death records. Additional information on deaths occurring outside SA was obtained from interstate registries and the National Death Index at the Australian Institute of Health and Welfare [14]. The extent of loss to follow-up has been checked through active tracing and found to be small and to have little effect on calculated survival [14,21]. Statistical analyses Analyses were undertaken with STATA 8.2 software [22]. Kaplan Meier product limit estimates were calculated to show diseasespecific survivals, with a censoring of live cases on 31 December 2003 [23]. This method has been shown to provide similar survival estimates to the relative-survival method in South Australian population-based studies [24]. For example, the 5-year survival from female breast cancers diagnosed in was 78% according to the Kaplan Meier estimate and 79% for the relativesurvival method. The Kaplan Meier estimate was preferred in the present study because patients risks of death from competing causes could not be assumed to equate with population norms (an underlying assumption in relative survival [25]). Cox proportional hazards regression was used to show differences in disease-specific fatality by tumour characteristic and to indicate levels of statistical significance [23]. Censoring criteria were the same as for the Kaplan Meier analyses. All variables were entered into the regression analyses, with backwards elimination where this did not reduce model fit (P > 0.05). Assumptions underlying these tests, such as proportionality and an absence of colinearity, were tested and found to be satisfied. Patterns of primary care were assessed by tumour characteristic, initially using the Pearson chi-square or Mann Whitney U-test for univariate analyses, depending on whether characteristics were distributed on a binary or nominal scale, or in an ordinal manner [22,23]. Fisher s exact test was substituted for the chi-square test when expected values were less than five. Logistic regression was used for multivariable analyses of associations of sociodemographic and tumour variables with treatment modality, again using backwards elimination [23]. Because the data came from different hospitals, the presence of effect modification by hospital was investigated in the regression analyses by examining interaction terms, and clustering was assessed using the robust variance estimator [22], but neither of these effects was detected. Regression analyses were stratified by hospital to avoid confounding from any inter-hospital differences. Results Survivals Case survivals for the 4671 breast cancers recorded on these registries ranged from 79.9% at 5 years to 68.9% at 10 years, 62.1% at 15 years, and 52.5% at 20 years. Survivals varied by age, with the lowest survivals applying in the youngest and (more so) the oldest age groups (Table 1). Survivals were predictably lower for higher s, with the 5-year figure ranging from 95.8% for stage I to 26.8% for stage IV. Survivals also were lower for less differentiated lesions, larger tumours and those with nodal environment (Table 1). Survivals varied by histology type (P < 0.001), with a lower 5- year survival of 79.8% for ductal as opposed to 89.6% for other specified types. Lower oestrogen and progesterone receptor expression was associated with a lower survival (P < for each receptor). For example, oestrogen receptor negative cases had a 5-year survival of 70.6%, compared with a corresponding 83.3% survival for receptor positive (low) and 90.2% for receptor positive (high) tumours. A secular increase in survival was observed during the 1990s, with the 5-year survival increasing from 78.8% for and 79.7% for to 83.6% for (P < 0.001). This increase occurred, despite an increase from 2.9% in to 8.3% in in the proportion of patients aged 80 years or more (P < 0.001), who had relatively low survivals (Table 1). The increase in survival was accompanied by lower stage distributions, with the proportion of cancers classified as stage I varying from 22.0% in to 35.9% in (P < 0.001). Meanwhile the small tumour proportion (diameter < 15 mm) increased from 14.1% in to 30.7% in (P < 0.001). Multivariable analysis confirmed that predictors of death from breast cancer included higher stages and less differentiated lesions. After adjusting for these characteristics, risk of death was higher in patients over 80 years of age at diagnosis, and potentially in non-metropolitan residents, although the difference by residential area achieved only a marginal statistical significance (P = 0.047) (Table 2). Patients diagnosed in had a lower risk of death than those diagnosed in previous years, with a relative risk (95% confidence limits) of 0.74 (0.62, 0.88). When the model was re-run, replacing with nodal involvement and diameter (mm) (<15 as the reference category and 15 19, 20 29, and 40 as indicator variables), a similar reduction in risk was observed, with a relative risk of 0.74 (0.61, Blackwell Publishing Ltd

4 C. Luke et al. Treatment and survival from breast cancer Table 1 Percentage of case survivals from female breast cancer; South Australian hospital-based registries, * Period from diagnosis (year) Characteristic 5 (%) 10 (%) 15 (%) 20 (%) P-value All cases (n = 4671) Age at diagnosis (year) Under 40 (n = 383) (n = 833) (n = 1117) P < (n = 1238) (n = 820) (n = 280) By I (n = 1494) IIA (n = 1354) II (other) (n = 943) IIIA (n = 225) P < III (other) (n = 172) IV (n = 388) By differentiation Well (n = 854) Moderate (n = 1449) P < Poor/undifferentiated (n = 1416) By diameter (mm) Under 10 (n = 399) (n = 672) (n = 698) (n = 1125) P < (n = 647) (n = 796) By nodal status Negative (n = 2615) P < Positive (n = 1820) *Survival from the primary cancer (see text); P-values derived from proportional hazards regression; Date of censoring: 31 December ) for Also, when the model was re-run for cases with data on hormone receptor expression, negative oestrogen receptor status was retained as a predictor of death (P < 0.001), and as before, a lower risk presented for than previous years, the relative risk being 0.77 (0.63, 0.95). Treatment patterns (primary course) Any treatment Overall, 99.4% of patients received surgery, radiotherapy, chemotherapy and/or hormone therapy as part of the primary course of care. The probability of not receiving one or more of these treatments was higher for patients aged 80 years or more (P = 0.004), those with stage IV disease, and those diagnosed prior to 1997 (Table 3). Multiple logistic regression analysis confirmed that predictors of not receiving any of these treatments comprised: being aged 80 years or more (P = 0.001); and having a more advanced stage (P = 0.008). After adjusting for stage, the relative odds (95% confidence limits) of not receiving any of these treatments were 4.76 (1.85, 12.22) for patients aged 80 years or more when compared with younger patients. Surgery The proportion receiving surgery as part of the primary treatment was 92.4%, with a lower proportion applying to older patients and those with advanced disease (Table 3). Multiple logistic regression analysis confirmed that the relative odds of surgery decreased with age and higher stage. Compared with patients less than 70 years of age, the relative odds of surgery were 0.26 (0.19, 0.36) for years old and 0.06 (0.04, 0.09) for patients aged 80 years or more, after adjusting for stage. Using stage I as the reference category, the relative odds of surgery reduced progressively with higher stage to 0.02 (0.01, 0.03) for stage IV. The proportion of surgical patients treated conservatively (i.e. not receiving a mastectomy) did not vary by calendar year during (P = 0.183), but it increased progressively from 36.6% in to 53.3% in and 58.7% in (Table 4). The trend varied by stage, with a secular increase applying for stage I and stage II, but with a decrease occurring for higher stages (Table 4). While secular trends were similar, irrespective of place of residence, conservative surgery was more common among surgical patients residing in metropolitan than country areas (52.3% compared with 47.7%; P = 0.027). This difference 2006 Blackwell Publishing Ltd 215

5 Treatment and survival from breast cancer C. Luke et al. Table 2 Relative risk (95% confidence limits) of case fatality from female breast cancer by diagnostic period and patient and cancer characteristic; South Australian hospital-based registries, * Multivariable proportional hazards regression Characteristic persisted after stage stratification in a Mantel Haenszel analysis (P = 0.028). Radiotherapy Relative risk Age at diagnosis (year) Under 80 (reference) (n = 3521) (n = 162) 1.48 (1.04, 2.11) Place of residence Adelaide (reference) (n = 2961) 1.00 Country in SA (n = 722) 1.20 (1.00, 1.43) I (reference) (n = 1269) 1.00 IIA (n = 1136) 2.22 (1.71, 2.89) II (other) (n = 791) 4.10 (3.17, 5.30) IIIA (n = 178) 5.12 (3.67, 7.15) III (other) (n = 114) 8.05 (5.73, 11.30) IV (n = 195) (15.42, 27.10) Differentiation Well (reference) (n = 848) 1.00 Moderate (n = 1435) 1.56 (1.19, 2.06) Poor/undifferentiated (n = 1400) 2.56 (1.96, 3.35) (reference) (n = 1963) (n = 1720) 0.74 (0.62, 0.88) *Includes cases with data on these characteristics. Date of censoring: 31 December Analysis stratified by hospital. The proportion of patients receiving radiotherapy as part of their primary course of care was 40.9%. This figure was lower for older age groups, and advanced stage, and increased from 12.6% in to 60.7% in (Table 3). A secular increase in proportion receiving radiotherapy was evident for each stage category (P < 0.001). By , 43.6% of patients treated by mastectomy were receiving adjuvant radiotherapy, which compared with 76.8% of patients receiving radiotherapy among those having conservative surgery (P < 0.001). For patients under 70 years of age treated by conservative surgery, the proportion receiving adjuvant radiotherapy was 84.0% in Although radiotherapy was generally less common when a mastectomy was provided, this was not so for all patients, in that the proportion of mastectomy patients receiving radiotherapy in was 74.1% when tumour diameters exceeded 50 mm and 78.9% when there were four or more involved nodes. Multiple regression analysis confirmed the trends observed in analyses of univariate predictors, with confirmatory results also applying when the regression analysis was restricted to surgical patients. In addition, a strong association between conservative surgery and radiotherapy was evident, the relative odds of radiotherapy being 6.27 (5.30, 7.41) when conservative surgery was provided rather than a mastectomy (Table 5). Chemotherapy Overall, 28.3% of patients had chemotherapy as part of their primary care, with this proportion reducing with age (Table 6). The proportion was larger in the higher stages, although reaching a peak for stage IIIA, before declining for the more advanced cases. There was also an increase in the chemotherapy proportion from 14.0% in to 36.9% in (Table 3), with secular increases applying for each stage category (P < 0.001). These trends were confirmed in a multiple logistic regression analysis (Table 6). Cyclophosphamide, methotrexate, fluorouracil (CMF) was the chemotherapy of choice for 85.5% of chemotherapy treatments in , but this proportion reduced to 74.8% in and 51.6% in Meanwhile, an increasing proportion of treatments included CAF/CEF, A C, or (to a lesser extent) other agents such as docetaxel, paclitaxel, gemcitabine, vinorelbine or capecitabine. Hormone therapy Approximately half the patients (49.9%) received hormone therapy as part of their primary treatment, with higher proportions applying for older ages, higher stage, and more recent diagnostic period (Table 3). Between and , the proportion receiving hormone therapy increased from 18.0% to 59.4%. About two-thirds (68.4%) of oestrogen-receptor positive cases diagnosed in received this therapy. A secular increase in proportion was evident for each stage category, although the evidence was stronger for stages I to III (P < 0.001) than stage IV (P = 0.077). These trends were confirmed in the multiple logistic regression analysis, where the relative odds of hormone therapy were found: (1) to increase with age to (11.96, 32.96) for patients aged 80 years or more when compared with those under 40 years as the reference category; (2) to increase with stage to 2.49 (1.75, 3.55) for stage IIIA, compared with stage I, and then decrease to 2.02 (1.75, 3.55) for stage IV; (3) to be lower at 0.53 (0.46, 0.62) for poorly or undifferentiated lesions compared with better differentiated ones; and (IV) to increase progressively over time to 6.31 (3.83, 10.41) for compared with When the analysis was repeated for patients with a recorded oestrogen-receptor status, the same predictors were retained, plus oestrogen receptor status, with the relative odds of hormone therapy being 2.67 (2.09, 3.40) for receptor positive (low), and 3.51 (2.81, 4.40) for receptor positive (high), when compared with receptor negative cases as the reference category. The hormone drug of choice was tamoxifen, which comprised 97.3% of hormone treatments in This proportion reduced to 87.6% in , due to an increased use of other agents such as anastrozole, letrozole, and LHRH analogues. Discussion Hospital registry data appeared broadly representative, in that survivals calculated from them equated with population-based Blackwell Publishing Ltd

6 C. Luke et al. Treatment and survival from breast cancer Table 3 Percentage of female breast cancer cases by treatment mode, as part of the primary course of care; South Australian hospital-based registries, * % receiving Characteristic No treatment (%) Surgery Radiotherapy Chemotherapy Hormone therapy All (n = 4593) Age (year) <40 (n = 381) (n = 823) (n = 1105) (n = 1218) (n = 799) (n = 267) P-value <0.001 <0.001 <0.001 <0.001 I (n = 1484) IIA (n = 1344) II (other) (n = 937) IIIA (n = 223) III (other) (n = 167) IV (n = 367) UK (n = 71) P-value <0.001 <0.001 <0.041 <0.001 < (n = 222) (n = 954) (n = 1483) (n = 1934) P-value <0.001 <0.001 <0.001 *Treatment details complete for 98% of cases. Derivation of P-values (see text) (data for unknown TNM values excluded). Table 4 Percentage of surgical female breast cancer cases by and surgical treatment mode, as part of the primary course of care; South Australian hospital-based registries, * Surgical mode (n = 1040) (%) (n = 1345) (%) (n = 1746) (%) P-value All conservative (n = 2123) P < mastectomy (n = 2008) I conservative (n = 1018) P < mastectomy (n = 423) II conservative (n = 930) P < mastectomy (n = 1232) III conservative (n = 70) P = mastectomy (n = 256) IV conservative (n = 84) P = mastectomy (n = 66) Unknown conservative (n = 21) P = mastectomy (n = 31) *Surgical details complete for 97% of surgical cases. Derivation of P-values (see text) Blackwell Publishing Ltd 217

7 Treatment and survival from breast cancer C. Luke et al. Table 5 Relative odds (95% confidence limits) of surgical cases of female breast cancer receiving radiotherapy, as part of the primary course of treatment; South Australian hospital-based registries, * Multiple logistic regression Characteristics Relative odds of radiotherapy Age at diagnosis (year) Under 60 (reference) (n = 2128) (n = 1116) 0.66 (0.55, 0.79) (n = 668) 0.26 (0.21, 0.33) 80+ (n = 164) 0.04 (0.02, 0.07) I or II (reference) (n = 3600) 1.00 IIIA (n = 209) 3.76 (2.68, 5.26) III (other) (n = 117) 2.88 (1.86, 4.45) IV (n = 150) 1.18 (0.79, 1.75) Surgery type mastectomy (reference) (n = 1977) 1.00 conservative (n = 2099) 6.27 (5.30, 7.41) (reference) (n = 192) (n = 828) 1.76 (1.08, 2.88) (n = 1336) 2.35 (1.46, 3.78) (n = 1720) (7.17, 18.42) *Analysis stratified by hospital. Table 6 Relative odds (95% confidence limits) of female breast cancer cases receiving chemotherapy, as part of the primary course of treatment; South Australian hospital-based registries, * Multiple logistic regression Characteristic Relative odds of chemotherapy Age at diagnosis (year) Under 40 (reference) (n = 373) (n = 808) 0.74 (0.55, 1.00) (n = 1093) 0.25 (0.18, 0.33) (n = 1199) 0.07 (0.05, 0.10) (n = 788) 0.01 (0.01, 0.02) 80+ (n = 261) 0.00 (0.00, 0.01) I (reference) (n = 1484) 1.00 IIA (n = 1344) 7.32 (5.69, 9.42) II (other) (n = 937) (12.51, 21.62) IIIA (n = 223) (24.02, 55.04) III (other) (n = 167) (20.54, 53.47) IV (n = 367) (11.04, 22.43) (reference) (n = 215) (n = 938) 2.13 (1.33, 3.43) (n = 1466) 3.94 (2.47, 6.28) (n = 1903) (7.78, 19.70) *Analysis stratified by hospital. breast cancer survivals [14,17]. The stage distribution for surgical cases at these hospitals was equivalent to the stage distribution reported in a national study [15]. Gains in case survivals have been apparent from Australian population-based data, but without data on stage, it has not been possible to assess respective contributions of earlier detection and treatment advances [17]. Our hospital registry data indicate that recently diagnosed patients had better survivals than previous patients, after adjusting for stage and grade. While residual confounding from imperfect measures (e.g. as may occur from stage shift), or from lead-time and related biases, may have contributed, the lower population-based breast cancer mortality in Australian women in than in the prior 10-year period suggests that real gains in case survival have occurred [9]. Although a decrease in national incidence has not been apparent [1], the mean annual breast cancer mortality rate, age standardized to the 2001 Australian population, was 16.1% lower in than that in [9]. Hospital data show changes in treatment, including an increased use of conservative surgery (as opposed to mastectomy) for stage I and stage II. This may reflect increased patient and surgeon preference for a conservative approach, plus greater opportunities to remove smaller than larger tumours through conservative surgery. Researchers have estimated that in North America, around 57% of stage I, and 52% of stage II breast cancers would be eligible for conservative surgery [26], which compares with the corresponding figures of 76.8% and 52.2%, respectively, of patients receiving conservative surgery in in this study. Since adjuvant radiotherapy is considered to be a standard of care for patients receiving conservative surgery [4], it is not surprising that rates of adjuvant radiotherapy have increased and were higher in patients treated by conservative surgery. By , 76.8% of patients receiving conservative surgery (84.0% of those under 70 years of age) were getting adjuvant radiotherapy. It is possible that some older patients would not have received radiotherapy due to preferences for alternative protocols, such as conservative therapy and adjuvant hormone therapy. Although the proportion of mastectomy patients receiving adjuvant radiotherapy in was lower at 43.6%, the proportion was comparatively high for mastectomy patients at 74.1% when tumour diameters exceeded 50 mm, and at 78.9% when four or more nodes were involved. This accords with guideline recommendations that adjuvant radiotherapy be considered for mastectomy patients with large tumours and extensive regional spread [4]. The trends observed in surgery type and adjuvant radiotherapy, and the secular increases in chemotherapy and hormone therapy, are broadly consistent with guideline directions and the earlier research on which they were based [4,5]. Advances in systemic treatment over the past 20 years have included recognition that premenopausal women benefit from hormonal chemotherapy, a greater use of anthracyclines and taxanes [27 29], and more recently, adjuvant and neo-adjuvant use of aromatase inhibitors [30,31]. Our multivariable analyses pointed to poorer outcomes in country than metropolitan residents, but only a marginal level of statistical significance was achieved (i.e. P = 0.047). If real, this Blackwell Publishing Ltd

8 C. Luke et al. Treatment and survival from breast cancer difference could reflect the referral to these teaching hospitals of the more difficult country cases, including those with greater frailty and co-morbidity. Women residing in rural areas have been found in other studies to be more inclined to receive a mastectomy as opposed to more conservative surgery [32], which would reduce the need for travel to metropolitan areas for radiotherapy and other adjuvant therapy. A small difference also was observed in this study, with 52.3% of country compared with 47.7% of metropolitan surgical patients having a mastectomy. Nonetheless, country residents were not found to have had less exposure to radiotherapy or to adjuvant treatments in general (P > 0.750). Survivals were lower in patients aged 80 years, after adjusting for stage, grade, place of residence, and diagnostic period. This may reflect decisions to restrict the range of treatment exposures, due to greater co-morbidity. Although few patients survived stage IV disease in the long term (i.e. 7.7% at 20 years from diagnosis), the 26.8% survival at 5 years is a comparatively high figure, equating with the corresponding 26.1% indicated by SEER data for for all races combined and the 27.7% for white women [25]. It was evident from our hospital data that although the proportion of stage IV patients receiving hormone therapy as part of their primary care did not vary over time between and (P = 0.546), an increasing proportion received radiotherapy over this period (an increase from 23.4% to 53.3%) and chemotherapy (an increase from 23.4% to 44.0%). Conclusions Trends towards conservative surgical management of early stage disease, and more general increases in the use of adjuvant radiotherapy, chemotherapy and hormone therapy, are consistent in broad terms with guideline recommendations and the earlier research results on which they were based. Reassuring gains in survivals from breast cancer are evident, after adjusting for stage, grade and other prognostic indicators, which may reflect treatment advances in addition to artificial influences. Old patients have lower survivals than younger patients of comparable stage and other prognostic indicators, which may reflect compromising effects of greater co-morbidity and frailty on treatment planning. Country residents are more likely than metropolitan residents to have a mastectomy, as opposed to more conservative surgery, despite treatment at the same clinical centres. The present data provide useful benchmarks for monitoring survival and treatment, which can be used by individual hospitals when evaluating their own experience. Similar analyses are advocated for other states and territories, in order to test the representativeness of results from this study. Acknowledgements Staff members of hospital registries are acknowledged for their careful attention to detail in the collection of these data over so many years. In particular, our gratitude is extended to: Ms Dianne Buranyi-Trevarton and Ms Carol Beeke, Royal Adelaide Hospital Registry; Ms Margaret Colbeck and Ms Vendra Severin, Queen Elizabeth Hospital Registry; and Ms Maureen McMellon, Flinders Medical Centre Registry. The Epidemiology Branch of the SA Department of Health also is acknowledged for providing core data and backup support services, particularly Ms Lesley Milliken, Mr Kevin Priest and staff members of the South Australian Cancer Registry. References 1. Australian Institute of Health and Welfare (AIHW) & Australasian Association of Cancer Registries (AACR) (2004) Cancer in Australia Cancer Series no. 28. Canberra: AIHW. 2. New South Wales Central Cancer Registry (NSWCCR) (2005) Cancer in New South Wales. Incidence and Mortality Sydney: The Cancer Institute New South Wales. 3. Australian Institute of Health and Welfare (AIHW) (2000) Breast- Screen Australian Achievement Report Cancer Series no. 13. Canberra: AIHW. 4. National Breast Cancer Centre (NBCC) (2001) Clinical Practice Guidelines for the Management of Early Breast Cancer, 2nd edn. Canberra: National Health and Medical Research Council. 5. National Health and Medical Research Council (NHMRC) (1995) The Management of Early Breast Cancer. Clinical Practice Guidelines. Canberra: Australian Government Printing Service. 6. Australian Institute of Health and Welfare (AIHW) (2005) Breast- Screen Australia monitoring report Cancer Series no. 29. Canberra: AIHW. 7. South Australian Cancer Registry (SACR) (2003) Epidemiology of cancer in South Australia. Incidence, Mortality and Survival 1977 to Incidence and Mortality Adelaide: Department of Human Services. 8. Luke, C., Nguyen, A. M., Priest, K. & Roder, D. (2004) Female breast cancers are getting smaller, but socio-demographic differences remain. Australian and New Zealand Journal of Public Health, 4, Australian Institute of Health and Welfare (AIHW) (2004) GRIM (General Record of Incidence of Mortality) Books. Canberra: AIHW. 10. Beral, V., Hermon, C., Reeves, G. & Peto, R. (1995) Sudden fall in breast cancer deaths in England and Wales (letter). Lancet, 345, Smith, C. L., Kricker, A. & Armstrong, B. K. (1998) Breast cancer mortality trends in Australia: 1921 to Medical Journal of Australia, 164, House of Representatives Standing Committee on Community Affairs (HRSCCA) (1995) Report of the Management and Treatment of Breast Cancer in Australia. Canberra: Australian Government Printing Service. 13. Malycha, P. & Tyson, S. (2000) National breast surgery audit. Australian and New Zealand Journal of Surgery, 70, South Australian Cancer Registry (SACR) (2000) Epidemiology of cancer in South Australia. Incidence, Mortality and Survival, 1977 to Incidence and Mortality, Adelaide: Openbook Publishers. 15. Hill, D., Jamrozik, K., White, V., Collins, J., Boyages, J., Shugg, D., Pruden, M., Giles, G. & Byrne, M. (1999) Surgical Management of Breast Cancer in Australia in Sydney: National Health & Medical Research Council, National Breast Cancer Centre. 16. The Cancer Council South Australia (TCCSA) (2002) Cancers of the Female Breast and Gynaecological Organs. Adelaide: The Cancer Council South Australia. 17. Australian Institute of Health and Welfare (AIHW) & Australasian Association of Cancer Registries (AACR) (2001) Cancer Survival in Australia. Part 2: Statistical Tables. Cat no. CAN14. Canberra: AIHW Blackwell Publishing Ltd 219

9 Treatment and survival from breast cancer C. Luke et al. 18. World Health Organization (WHO) (1977) Manual of the International Statistical Classification of Diseases, Injuries and Causes of Death (Based on Recommendations of the Ninth Revision Conference, 1975). Geneva: World Health Organization. 19. Hahn Johnson, C. & Richards, L. (eds) (1998) Standards of the Commission on Cancer, Volume 2: Registry Operations and Data Standards (ROADS). Chicago: Commission on Cancer. 20. Osteen, R. T., Winchester, D. P. & Cunningham, M. P. (1995) Breast Cancer. In National Cancer Data Base Annual Review of Patient Care, 1995 (eds G. D. Steele, J. M. Jessup, D. P. Winchester, H. R. Menck & G. P. Murphy), pp Atlanta: The American Cancer Society. 21. Bonett, A., Roder, D. & Esterman, A. (1988) Cancer case-survival rates for South Australia: a comparison with US rates and a preliminary investigation of time trends. Medical Journal of Australia, 148, StataCorp. (2004) STATA Statistical Software. Release 8.2. Texas: STATA Corporation, College Station. 23. Armitage, P. & Berry, G. (1987) Statistical Methods in Medical Research. Oxford: Blackwell Scientific Publications. 24. South Australian Cancer Registry (SACR) (1997) Epidemiology of cancer in South Australia. Incidence, Mortality and Survival, 1977 to Incidence and Mortality, Adelaide: Openbook Publishers. 25. Ries, L. A. G., Eisner, M. P., Kosary, C. L., Hankey, B. F., Miller, B. A., Clegg, L., Mariotta, A., Feuer, E. J. & Edwards, B. K. (eds) (2005) Cancer Statistics Review, Bethesda, MD: National Cancer Institute. 26. Tyldesley, S., Foroudi, F., Barbera, L., Boyd, C., Schulze, K., Walker, H. & MacKillop, W. J. (2003) The appropriate rate of breast conserving surgery: an evidence-based estimate. Clinical Oncology, 15, Aebi, S., Gelber, S., Castiglione-Gertsch, M., et al. (2000) Is chemotherapy alone adequate for young women with oestrogen-receptorpositive breast cancer? Lancet, 355, Di Leo, A., Ciarlo, A., Panella, M., Pozzessere, D., Santani, S., Vinci, E. & Biganzoli, L. (2004) Controversies in the adjuvant treatment of breast cancer: the role of taxanes. Annals of Oncology, 15, iv17 iv Tack, D. K., Palmieri, F. M. & Perez, E. A. (2004) Anthracycline vs nonanthracycline adjuvant therapy for breast cancer. Oncology (Huntingt), 18, Ellis, M. J., Coop, A., Singh, B., et al. (2001) Letrozole is more effective neoadjuvant endocrine therapy than tamoxifen for Erb-1- and/or ErbB-2-positive, estrogen receptor-positive primary breast cancer: evidence from a phase III randomized trial. Journal of Clinical Oncology, 19, Baum, M., Budzar, A. U., Cuzick, J., Forbes, J., Houghton, J. H., Klijn, J. G., Sahmoud, T. (ATAC Trialists Group) (2002) Anastrozole alone or in combination with tamoxifen versus tamoxifen alone for adjuvant treatment of postmenopausal women with early breast cancer: first results of the ATAC randomised trial. Lancet, 359, Craft, P. S., Primrose, J. G., Lindner, J. A. & McManus, P. R. (1993) Surgical management of breast cancer in Australian women in 1993: analysis of Medicare statistics. Medical Journal of Australia, 166, Blackwell Publishing Ltd