Decline in the Incidence of Distant Recurrence of Breast Cancer: A Population-Based Health Record Linkage Study, Australia 2001–2016

Breast cancer is now the most frequently diagnosed cancer worldwide (1). In high-income countries, breast cancer–related mortality rates have declined by 40% over the last three decades, with five-year relative survival now exceeding 90% (1–3). These data demonstrate the substantial impact of breast cancer screening programs and treatment advances on population health. For Stage I–III breast cancer (approximately 95% of new breast cancer diagnoses; ref. 4), an important but understudied measure to assess the effectiveness of cancer care and cancer health disparities is the incidence of distant recurrence (DR).

A reduction in the incidence of DR has been reported in cohort studies covering periods when adjuvant chemotherapy and endocrine therapy were first introduced and mammographic screening became widespread (5, 6). However, there are only very limited descriptions of the population impact of changes in breast cancer care on DR because population-based cancer registries, including the Surveillance, Epidemiology, and End Results program in the United States, do not routinely report recurrence occurring after early-stage (non-metastatic) breast cancer diagnoses. This is an important knowledge gap for the planning and delivery of cancer care.

Over the last two decades newer adjuvant treatment regimens for early-stage breast cancer, such as aromatase inhibitors for hormone receptor–positive breast cancer, HER2-targeted therapies for HER2-positive breast cancer, and taxane-containing chemotherapy regimens have been introduced on the basis of clinical trial evidence of improved survival. Tracking DR rates over time is essential to determine whether these treatment protocols have translated into population health benefits, and to identify and address potential treatment inequalities for those who are socio-economically disadvantaged, for whom lower breast cancer survival is well documented (7, 8). At least one quarter of breast cancer is diagnosed in people 70 years or older. Treatment decisions for older people are more complex and data for this important group are particularly lacking (9). Population-based data about anatomical sites of DR would also help inform health service delivery. Earlier studies have reported a higher proportion of first distant spread to visceral sites than to bone as DR rates have declined (10–12). Anatomical sites vary by tumor receptor type (13), and first site of DR may change further as new adjuvant targeted therapies are introduced. Follow-up beyond five years is required to assess changes in the longer-term incidence of recurrences. Long-term data from clinical trials of adjuvant chemotherapy and up to five years of tamoxifen show reduced risk of “early” recurrence (commonly defined as ≤five years after diagnosis) but little impact on “late” recurrence (>five years; refs. 14, 15).

This population-based descriptive study aimed to quantify the 9-year cumulative incidence, annual hazard, and site of first DR for women diagnosed with non-metastatic breast cancer in 2001 to 2002 and 2006 to 2007; and to describe adjuvant therapy use during these periods. A priori subgroups are defined and examined by extent of disease at initial diagnosis, age, treatment-defined tumor receptor status, and area-level socio-economic status (SES).

We conducted a population-based cohort study using record linkage to access routinely collected administrative health data. We included two cohorts of females ages ≥18 years with a first incident primary invasive non-metastatic breast cancer registered in the statewide New South Wales Cancer Registry (NSWCR) during January 1, 2001–December 31, 2002 (Cohort 1, C1) and January 1, 2006–December 31, 2007 (Cohort 2, C2). Exclusion criteria were: Unknown or distant stage disease at diagnosis; death ≤30 days after breast cancer diagnosis; a non-breast primary cancer before breast cancer because it would not be possible to distinguish the source of a subsequent DR; having no NSW hospital records due to the potential for under-ascertainment of DR if treated outside of NSW. We excluded the small number of males due to the risk of statistical disclosure. We determined sex from the NSWCR record, which records the sex of the person as reported on the cancer case notification forms received (16).

NSW is the most populous state in Australia, comprising nearly one third of Australia's total population (17). Australia has a publicly funded universal healthcare system with free treatment in public hospitals and outpatient services. Non-pharmaceutical treatments, including radiotherapy are subsidized through the Medicare Benefits Schedule (MBS). Prescription medicines are subsidized through the Pharmaceutical Benefits Scheme (PBS). The Commonwealth Government Herceptin Program subsidized trastuzumab for women with HER2-positive metastatic breast cancer from December 2001 to June 2015, after which it was subsidized by the PBS. Adjuvant therapies that became available and PBS-listed during the study period included: Broader access to taxane-based regimens for node-positive breast cancer (December, 2006; ref. 18); aromatase inhibitors for ER-positive tumors (December, 2004; ref. 19); and trastuzumab for HER2-positive tumors (October, 2006; ref. 20; Supplementary Fig. S1). The NSW BreastScreen Program has provided free screening mammograms for women ages ≥40 years since 1991. Screening participation rates for the target age group (50–69 years) have been stable since 1998, but decreased after 2005 for ages 40–49 years and ≥70 years when active recruitment was restricted to the target age group (21–23).

Data sources were the NSWCR, NSW Registry of Births, Deaths and Marriages, NSW Cause of Death Unit Record File, NSW Admitted Patient Data Collection (APDC) for hospital records, PBS claims for anticancer medicines, Herceptin Program records, and MBS claims for radiotherapy services. APDC records were available from July, 2001, PBS records from July, 2002, and other data sources from January, 2001. All data sources provided records to September 30, 2016 or longer. We have reported details of datasets, linkage, and power calculations in our research protocol (doi:10.1136/bmjopen-2018–026414; ref. 24).

We used NSWCR records to categorize age at breast cancer diagnosis (<40, 40–49, 50–69, ≥70 years); country of birth (Australia/New Zealand, other); and area level SES using quintiles of the Index of Relative Socio-economic Disadvantage constructed for the whole of Australia by the Australian Bureau of Statistics from census data for individuals and households (25) and mapped to place of residence for cancer cases by the NSWCR. To avoid small groups, we collapsed area-level SES into two categories for subgroup analyses (more disadvantaged: quintiles 1–3, less disadvantaged: quintiles 4–5). We used the NSWCR classification for extent of disease at diagnosis as localized (T1–3 and no spread to axillary lymph nodes) or regional breast cancer (axillary lymph node spread or locally advanced disease) and ICD-0–3 morphology codes to classify tumor morphology as ductal, lobular and mixed ductal/lobular, or other (16).

We defined receipt of adjuvant systemic therapy as a first dispensing of endocrine, HER2-targeted or cytotoxic therapy ≤12 months after breast cancer diagnosis. We categorized adjuvant chemotherapy dispensed within 12 months as anthracycline-based, taxane-based, anthracycline and taxane combination or sequential, other; and adjuvant endocrine therapy as tamoxifen only, aromatase inhibitor only, tamoxifen and aromatase inhibitor (at any time before DR to incorporate endocrine switches), other. We defined tumors as estrogen or progesterone receptor–positive (abbreviated as ER-positive) if endocrine treatment was dispensed; HER2-positive if HER2-targeted treatment was dispensed; and triple-negative if cytotoxic therapy was dispensed without endocrine or HER2-targeted treatment.

The primary outcome was time to first DR, defined as metastasis in distant organs or non-regional lymph nodes (26), measured in days from the date of breast cancer diagnosis recorded in the NSWCR to the date of the first DR record. The secondary outcome was first recorded site of spread, defined using ICD-10-AM diagnosis codes for secondary malignant neoplasms of bone, liver, lung/pleura, brain, distant lymph nodes, other.

We used seven criteria to classify DR from administrative health records (Supplementary Table S1): First notification of DR recorded in the NSWCR. Notification of new and recurrent cancer cases is a statutory requirement for hospitals, pathology laboratories, radiotherapy services, outpatient departments, day procedure centers and nursing homes; An APDC hospital episode-of-care with an ICD-10-AM diagnosis code for secondary cancer, excluding secondary cancer to locoregional lymph nodes; An MBS code for a radiotherapy service to a secondary cancer site (codes available from May, 2003), or to an unspecified cancer site and classified as palliative (codes discontinued in April, 2003); A PBS code for medicine dispensing restricted to advanced/metastatic disease or a Herceptin Program prescription; A PBS code for dispensing chemotherapy after the adjuvant treatment period (>12 months) and ≥90 days after prior adjuvant therapy, and not classified as treatment for locoregional recurrence, or a second primary breast cancer; or date of death where secondary cancer was recorded as a cause and no other DR records were identified. For people meeting one or more of these criteria, we used the date of the first record as the date of first DR.

We present descriptive statistics for demographic and tumor characteristics of all women in the 2001 to 2002 and 2006 to 2007 cohorts. We compared adjuvant therapy between cohorts in the subset of women with a breast cancer diagnosis from July, 2002 (when complete PBS medicine data were available) who remained DR-free at 12 months, overall and for subgroups (extent of disease at diagnosis, age group). We used Pearson's χ² tests to test the null hypothesis of no difference in the distribution of treatment characteristics between C1 and C2. We also compared systemic adjuvant therapy use by area-level SES in C2.

We used a non-parametric method to calculate the cumulative incidence function for DR in the presence of competing risks (27) and Gray's weighted log-rank test to test the null hypothesis of equality between cohorts (28). We quantified the cumulative incidence of DR at five and nine years for each cohort and subgroups, and present absolute differences between cohorts at these time points with 95% confidence intervals (CI) based on the standard errors of the cumulative incidence in each cohort. We defined competing events as death due to any cause without DR recorded, or a DR record following a NSWCR record of a second (non-breast) primary cancer. To achieve comparable follow-up times for each cohort, we followed C1 to September 30, 2011 and C2 to September 30, 2016. Censoring only occurred for those reaching the end of follow-up without DR or a competing event. We restricted analyses of treatment-defined receptor subgroups to those remaining DR-free at 12 months after breast cancer diagnosis to avoid immortal time bias, and counted time to DR from 12 months after breast cancer diagnosis for these analyses. For consistency, these results are labeled in graphs by the time since breast cancer diagnosis (where 5 and 9 years correspond to 4 and 8 years after the one-year landmark). We restricted analysis of HER2-positive and triple-negative subgroups to C2 because adjuvant trastuzumab was not available in C1 to define HER2 status.

We examined the pattern of recurrence over time for each cohort by plotting the annual hazard rate for DR estimated at the mid-point of each year. We used the absolute differences in 5-year DR risk to compare the risk of early DR (≤5 years) between cohorts. To compare the risk of late DR (>5 years) between cohorts, we also report the cumulative incidence of DR at nine years after breast cancer diagnosis for those remaining DR-free at 5 years (4-year conditional risk) with the absolute difference between cohorts and 95% CI.

As a sensitivity analysis, we examined the cumulative incidence of DR in the C1 subset with breast cancer diagnosed July—December, 2002, who had PBS medicine dispensing records available from the time of diagnosis to assess the potential for underestimating DR risk in the full C1 cohort due to the lack of medicine dispensing records before July, 2002.

Analyses were conducted using SAS V.9.4 statistical software. Figures were created using R. Ethics approval for this study with a waiver of consent was provided by the NSW Population and Health Services Research Ethics Committee (HREC/17/CIPHS/19), the Australian Institute of Health and Welfare Ethics Committee (EO2017/2/255) and the University of Notre Dame Human Research Ethics Committee (0-17–144S). The research was conducted in accordance with the Australian National Health and Medical Research Council, National Statement on Ethical Conduct in Human Research consistent with the Declaration of Helsinki.

The data generated in this study are not publicly available without specific approval from the relevant data custodians and human research ethics committees to protect patient privacy under Australian laws and regulations. Derived aggregate data generated for the purpose of the current study and approved for release by the data custodians are available from the corresponding author on request.

16,521 women had a first primary invasive breast cancer diagnosis recorded in the NSWCR during 2001 to 2002 or 2006 to 2007. Of these, 13,170 women with non-metastatic breast cancer were included (C1 2001–2002, n = 6,338; C2 2006–2007, n = 6,832; Supplementary Fig. S2). Overall, 7,821 (59%) had localized (T1–3N0) disease at breast cancer diagnosis; and 9,819 (75%) were ages ≥50 years (Table 1). Over the 9-year follow-up from diagnosis, we identified DR records for 2,194 (17%) women, including 12 DR only identified in death records, and 1,279 (10%) women had competing events.

Of 8,549 women with breast cancer diagnosed from July 2002 and linked PBS records, 8,013 were alive and DR-free at 12 months (C1 1,576 and C2 6,437). Of these, adjuvant endocrine therapy was dispensed to 66% of C1 and C2 (Table 2). Aromatase inhibitors were PBS-listed for adjuvant therapy in 2004, and thus were available as a switch for C1 and as an initial therapy or switch for C2. Of those dispensed endocrine therapy, 514 (49%) were dispensed an aromatase inhibitor in C1 and 3107 in C2 (73%, P < 0·0001). Adjuvant chemotherapy was dispensed to 474 (30%) of C1 and 2,565 (40%) of C2 (P < 0·0001), with higher use in C2 than C1 for localized (16% C1, 24% C2, P < 0.0001) and regional disease (52% C1, 62% C2, P < 0·0001). Of those dispensed chemotherapy in C1, 59 (12%) received an anthracycline and taxane compared with 1,269 (49%) in C2 (P < 0·0001, Table 2). Adjuvant anti-HER2 therapy was dispensed to 0% in C1 and 569 (9%) in C2. In both cohorts, adjuvant chemotherapy and HER2-targeted therapy were more frequently dispensed to those <50 years than older ages (Table 2).

In C2, adjuvant chemotherapy was more commonly dispensed to those living in less socioeconomically disadvantaged areas than more socioeconomically disadvantaged areas (43%, 38%, P < 0·0001). However, we observed similar chemotherapy use across SES-defined areas for those ages <50 years (Supplementary Table S2). For those ages 50 to 69 years, adjuvant endocrine therapy was also more commonly dispensed to those living in less socioeconomically disadvantaged areas (68%) than more socioeconomically disadvantaged areas (65%, P = 0.04). Similar proportions of adjuvant trastuzumab were dispensed across SES-defined areas, overall and for all age groups (Supplementary Table S2).

The 9-year cumulative incidence of DR was 15.0% in C2 and 18.6% in C1 (9-year risk difference 3.6%; 95% CI, 2.3%–4.9%, Fig. 1A, Table 3). Results for C1 were similar for breast cancer diagnoses from July, 2002 when medicine dispensing records were available from the date of diagnosis (9-year cumulative incidence, 18.7%). The lower cumulative incidence of DR in C2 was seen for both localized and regional disease (Fig. 1B, Table 3). For both cohorts, the annual hazard of DR was lower in each year since diagnosis of localized or regional breast cancer, with the largest absolute difference in annual hazard between C1 and C2 occurring within the first five years after breast cancer diagnosis (Supplementary Fig. S3). Overall, the 5-year absolute risk difference for DR between cohorts was 3.3% (95% CI, 2.2%–4.5%). In comparison, for those remaining DR-free at 5 years, absolute difference in the cumulative incidence of DR within the next four years (to year 9 after breast cancer diagnosis) was 0.6% (95% CI, 0.3%–1.4%).

The risk of DR varied by the age group (Fig. 2A–D, Table 3). The largest decline in DR between C1 and C2 was observed in those ages <40 years (9-year risk difference 14.4%, 95% CI, 8.3%–20.6%, Fig. 2A). No decline in DR risk was observed for those ≥70 years (9-year risk difference −0.9%; 95% CI, −3.6%–1.8%, Fig. 2D). For those remaining DR-free at 12 months, the 8-year cumulative incidence of DR declined for both ER-positive tumors (risk difference 4.1%; 95% CI, 1.6%–6.6%, Fig. 3A) and for ER-negative tumors (risk difference 3.5%; 95% CI, −3.6%–10.5%, Fig. 3B). The corresponding risk difference in cumulative incidence was 2.0% (95% CI, 1.2%–5.3%) for those receiving no adjuvant systemic therapy (Fig. 3C). In C2, the 8-year cumulative incidence of DR risk for those remaining DR-free at 12 months was higher for triple-negative (21.2%) and ER-negative HER2-positive tumors (20.9%) than ER-positive tumors (ER-positive HER-negative 12.6%, ER-positive HER2-positive 13.7%; Supplementary Fig. S4). The risk of DR was lower in C2 than C1 across each quintile of area-level SES disadvantage, but remained lower for those in the least socioeconomically disadvantaged areas (Supplementary Table S3). The risk of metastasis to bone as the first recorded single or multiple DR site was higher than other sites in both cohorts (Table 3). The risk of metastasis to bone, lung/pleura or liver as first recorded site was lower in C2 than C1, whereas the risk of metastasis to brain was similar between cohorts (Table 3).

In this whole-of-population study, the 9-year cumulative incidence of DR for non-metastatic breast cancer diagnosed during 2006 to 2007 was 3.6% (95% CI, 2.3%–4.9%) lower than for those diagnosed in 2001 to 2002. This decline coincided with changes in the use of adjuvant therapies between the two cohorts, and provides evidence for the real-world impact of access to new adjuvant systemic therapies, although this may not be the only contributor to the observed decline.

Meta-analyses of clinical trials of adjuvant therapies conducted in selected patient groups with relatively few people older than 70 years have reported relative risk reductions for DR with absolute risk differences varying according to baseline risk (15, 29–31). Our findings extend this evidence by providing an overall estimate of DR risk for the broader breast cancer population. The observed risk differences reflect the real-world use and impact of newer adjuvant treatments combined with any other (unmeasured) changes in breast cancer screening, staging, treatment and support services over the study period that may alter DR risk; and possibly other factors such as the changes in the prevalence of risk factors for breast cancer associated with tumor type and thereby DR risk. We add to the evidence of a decline in DR from earlier population-level studies (10–12) and more selected breast cancer cohorts (5, 6, 32) that have reported substantial DR decline since the 1980s by comparing two relatively contemporary breast cancer cohorts both of which had access to the population breast screening program for ≥10 years. The 2001–2002 cohort was diagnosed when the benefits of adjuvant chemotherapy for node-positive and high-risk node-negative breast cancer, and tamoxifen for ER-positive breast cancer were well established (33). Adjuvant therapies prescribed to the 2006 to 2007 cohort (aromatase inhibitors, taxane-containing chemotherapy, trastuzumab) continue to be used in current practice. The 9-year DR risk observed for this cohort is similar to the 15% 10-year risk reported in a Netherlands study of women diagnosed with breast cancer in 2005 when adjuvant trastuzumab was introduced and all of whom received surgery (34).

Breast cancer screening participation rates in NSW were relatively stable for the target age group (50–69 years) during the study period, and lower in 2006 to 2007 for eligible age groups outside this range (40–49 years, ≥70 years; refs. 21–23). NSWCR data show that the incidence of localized disease did not increase during the study period, whereas the incidence of regional spread was higher in 2006 to 2007 compared with 2001 to 2002 across all age groups (21). These data may indicate stage shift during the study period, potentially due to improved imaging and staging of primary cancers, and this may contribute to the observed reduction in DR risk. In conjunction with this shift, access to adjuvant taxanes for regional spread expanded during the study period.

The largest absolute difference in annual DR risk between cohorts occurred in the first five years after breast cancer diagnosis when DR risk is highest, with a more modest difference in the risk of late recurrence. This finding is consistent with the effects of adjuvant systemic therapy on early recurrence reported from early trials of chemotherapy, endocrine therapy (duration five years), and trastuzumab (12 months; refs. 15, 29, 30, 35). We have previously reported the sustained risk of late recurrence for ER-positive tumors in C1 (36) consistent with other breast cancer cohorts treated before adjuvant HER2-targeted therapy was introduced (37). In the present study, our findings for C2, who had access to HER2-targeted adjuvant therapy and aromatase inhibitors, indicate the risk of late recurrence persists for ER-positive HER2-negative tumors (the largest tumor receptor subgroup), and is higher than treatment-defined HER2-positive tumors. Follow-up >10 years will be valuable to describe the population-level impact of extended duration adjuvant endocrine therapy regimens on long-term DR risk for ER-positive tumors.

We observed a large variation in DR decline by age. DR risk was highest in people <40 years, possibly reflecting the higher proportions of HER2-positive and triple-negative breast cancer and higher tumor cell proliferation in this age-group than older ages (38, 39). We observed the largest decline in 9-year risk of DR in C2 in this age group, who also had the highest uptake of adjuvant trastuzumab and, together with those ages 40 to 49 years, more use of adjuvant chemotherapy (including taxanes) than older ages. In contrast, for people ≥70 years, DR risk was relatively low in C1 and did not decline in C2. This may largely reflect that ER-positive breast cancer is more common in people ≥70 years and endocrine therapy use was relatively high for this age group in both cohorts. Considering that aromatase inhibitor use was higher in C2 than C1, the population-level benefits of aromatase inhibitors in people ≥70 years appear to be more modest than trial estimates (29). Screening participation rates for women ages 70 to 74 years fell from over 40% to 3% after the screening program restricted active recruitment to women ages 50 to 69 years in 2005 (21), which also may have contributed to these findings. Low use of chemotherapy (10%) and trastuzumab (3%) by people ≥70 years observed in C2 has also been reported by others (40), and is likely due to the need to consider the toxicity and tolerability of these agents with the increased co-morbidities often seen in older people.

The main limitation of our study methods was the potential for under or over-ascertainment of DR using administrative health records (41). Tamoxifen and aromatase inhibitors were not included as DR criteria because it was not possible to distinguish between adjuvant and metastatic use. This may have led to underestimation of DR for ER-positive breast cancer if other DR criteria such as hospital care, chemotherapy or radiotherapy occurred later, or not at all during the study period. Furthermore, any differences in DR ascertainment from study data sources between C1 and C2 could introduce bias in the comparison of DR incidence. Such differences could arise from any changes in data availability or recording over time. In the present study, the lack of medicine dispensing records before July 2002 may have led to underestimation of DR rates in C1. However, sensitivity analysis restricting C1 to breast cancer diagnosed during July–December, 2002 with complete ascertainment of PBS records showed similar 9-year cumulative incidence of DR between the C1 treatment subset and all of C1. Finally, pathology information was not available for tumor size, proliferation (grade, Ki67) or receptor type. However, this limitation does not restrict the overall analysis of changes in DR incidence over time incorporating all tumor types.

In clinical practice, these study findings are helpful when counseling women about the benefits of treatment and may help improve treatment adherence, particularly in women ages <40 years in whom the largest reduction in DR incidence was observed and the side effects of treatment are often most bothersome. Our finding that DR incidence did not decline in people ≥70 years supports the importance of research on effective and tolerable adjuvant treatment regimens for older people at high risk of DR, including: Initiatives to increase enrolment of older people who have been significantly under-represented in breast cancer trials; trials designed to address questions specifically relevant for older people; and decision support tools. For low and middle-income countries where breast cancer mortality remains high, our findings demonstrate the potential population benefit of initiatives to establish comprehensive breast cancer management services such as the Breast Health Global Initiative (42). Our finding of higher DR risk for those living in more socioeconomically disadvantaged areas adds to existing evidence of SES disparities in breast cancer survival (7, 8, 43). For further research, mediation analyses would be valuable to investigate the relative effects of newer treatments alongside other contributors to DR risk, including socioeconomic position.

Finally, our findings support the need for surveillance of DR incidence to better understand cancer health disparities and to assess the effectiveness of new interventions for equitable cancer control. This will require the addition of cancer registry protocols for routine registration of DR. Although we demonstrate that the risk of DR has declined, the number of people living with a diagnosis of metastatic breast cancer is estimated to be rising (44) due to the increasing number of people diagnosed with breast cancer, no decline in the incidence of de novo metastatic breast cancer, and more effective treatments for metastatic breast cancer. Cancer registry protocols to support reporting of DR incidence will also enable reporting of metastatic breast cancer prevalence and post-metastasis survival. Others have raised the importance of these data across all tumor types (45). Breast cancer consumer representative groups (46) are increasingly calling for this information to ensure that their clinical and support service needs are adequately met regardless of age or the socioeconomic group.

S.J. Lord reports grants from The National Health and Medical Research Council during the conduct of the study. B. Daniels reports grants from Cancer Institute NSW outside the submitted work. B.E. Kiely reports personal fees and other support from Novartis; personal fees from Gilead, MSD Oncology, and Eisai; and other support from Pfizer outside the submitted work. N. Houssami reports grants from National Breast Cancer Foundation (NBCF) and National Health & Medical Research Council (NHMRC) during the conduct of the study. No disclosures were reported by the other authors.

S.J. Lord: Conceptualization, formal analysis, funding acquisition, investigation, methodology, writing–original draft, writing–review and editing. B. Daniels: Conceptualization, formal analysis, investigation, methodology, writing–review and editing. D.L. O'Connell: Conceptualization, supervision, investigation, methodology, writing–review and editing. B.E. Kiely: Conceptualization, investigation, methodology, writing–review and editing. J. Beith: Conceptualization, investigation, methodology, writing–review and editing. A.L. Smith: Conceptualization, methodology, writing–review and editing. S.-A. Pearson: Conceptualization, methodology, writing–review and editing. K.-L. Chiew: Conceptualization, investigation, methodology, writing–review and editing. M.K. Bulsara: Conceptualization, supervision, methodology, writing–review and editing. N. Houssami: Conceptualization, supervision, funding acquisition, investigation, methodology, writing–review and editing.

We thank the NSW Cancer Registry, the NSW Ministry of Health, and the Australian Department of Health for providing the data for this study. We thank the Australian Institute of Health and Welfare Data Integration Services Centre and the NSW Centre for Health Record Linkage (CHeReL) for undertaking data linkage. We thank Dr. T. Li for her work on a systematic review of population-based studies of MBC incidence undertaken before this study. We thank Dr. K. Bahlmann for initial development of the timeline of PBS subsidization of targeted therapies included in this article. We thank Chantal Corthals (Breast Cancer Network Australia) for comments on the research questions and an earlier draft of this manuscript. S.J. Lord was funded by a National Health and Medical Research Council (NHMRC) Project Grant ID: 1125433; N. Houssami by the National Breast Cancer Foundation Chair in Breast Cancer Prevention grant (EC-21–001) and a NHMRC Investigator (Leader) grant (194410); B. Daniels by a Cancer Institute NSW Early Career Fellowship (ECF1381); and B. Daniels and S.A. Pearson were supported by the NHMRC Centre of Research Excellence in Medicines Intelligence (1196900).

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.