Baseline demographic and clinical characteristics
The study population was drawn from a prevalent population of 4,246 female patients alive with a diagnosis of BC between 1st January 2012 and 31st March 2018. Over this 75-month period, 464 patients were identified with a new diagnosis of locally advanced BC or MBC, 318 of whom (68.5%) had HR+/HER2- disease. In all, 196 women with MBC were eligible for the analysis; 47 (38.5%) of those excluded had locally advanced BC, 35 (28.7%) had participated in a clinical trial, 26 (21.3%) had incomplete records and the remainder had another malignancy at index (Additional file1).
Patient characteristics for the study cohort are summarized (Table 1). Median age was 67 years and the majority of patients were post-menopausal (n=167, 85.2%). Median follow-up time was 34 months (range 0.3-77). Approximately two thirds of patients (63.3%, n=124) had recurrent MBC with a median time between primary diagnosis and metastasis of 5.4 years (range 0.5-28); the remainder (36.7%, n=72) had de novo MBC. Metastatic disease at index date was exclusively non-visceral for 73 patients (37.2%) and exclusively visceral (including CNS sites) for 31 (15.8%); the remaining 88 patients (44.9%) had metastases at both visceral and non-visceral sites. Bone was the most commonly recorded site of metastasis (n=133, 67.9%).
Treatment for metastatic disease
Almost all patients (n=192, 98.0%) received SACT during the study period; the median number of distinct treatments with SACT following index date was 2 (range 1-9). At some point, almost all patients received endocrine therapy (n =182, 94.8%) whereas 91 (47.4 %) received chemotherapy. Seven patients were omitted from analysis by LoT because initiation of treatment pre-dated confirmation of metastatic disease. Subsequent analyses are based on the remaining 185 patients receiving SACT post-diagnosis of MBC, for whom the distributions of site of metastasis, and de novo and recurrent MBC were similar to the overall cohort (Table 1).
Patients with recurrent MBC were more likely to have visceral metastases (n=75, 62.5%), than those with de novo MBC (n=28, 38.9%; p<0.01). Although visceral metastases were more frequent amongst patients under 55 years than in older age groups (75.8% [n=25] and 60.8% [n=90], respectively), this difference was not statistically significant (p=0.11).
Endocrine therapy
Endocrine therapy was the modality most frequently used as 1st LoT (n=127, 68.6%), with a further 12 (6.5%) receiving endocrine/targeted therapy. Median treatment duration of 1st LoT endocrine (with/without targeted) therapy was 382 days (range 13-1708). Patients treated with 1st LoT endocrine most commonly received an AI (114 patients, 82.0%); tamoxifen (8.6%) or exemestane/everolimus in combination (8.6%) were used less frequently (Table 2).
Chemotherapy
Forty-six patients (24.9%) received 1st LoT chemotherapy and median treatment duration was 129 days (range 24-278). The proportion of patients receiving 1st LoT chemotherapy appeared to decrease between 2012 and 2017 (33.3% and 15.4%, respectively), although this trend was not statistically significant (R2=0.38, p=0.19).
No single class of chemotherapy predominated as 1st LoT; 26 patients (56.5%) received single agent chemotherapy (paclitaxel, capecitabine or docetaxel); the remaining 20 received combination chemotherapy (Table 2). Half those patients on combination therapy had recurrent MBC and carboplatin with paclitaxel (n=6) was only used for patients with recurrent MBC who had been treated with (neo)adjuvant chemotherapy. Epirubicin and cyclophosphamide (EC; n=14) was only used for patients with visceral metastasis, whether with recurrent MBC following (neo)adjuvant treatment with endocrine or with de novo MBC.
Factors influencing use of chemotherapy as 1st line SACT
Age and site of metastasis
Patients receiving 1st LoT chemotherapy were younger (median age 59, range 33-84; p<0.01), more likely to be pre/peri-menopausal (45.0%, n=9; p=0.02) and more likely to have visceral metastasis (89.1%, n=41; p<0.01) than those receiving endocrine or endocrine/targeted treatment (Table 3). Very few patients with exclusively non-visceral metastasis at index date (<10%) were treated with 1st LoT chemotherapy.
Of the 115 patients with visceral metastasis, 17 (68.0%) of those aged under 55 years were treated with chemotherapy compared with fewer than 6.0% of those aged 75 years and over (p<0.01) (Table 3). The 66 patients with exclusively non-visceral metastasis were, however, almost always treated with 1st LoT endocrine (with/without targeted) therapy, regardless of their age (Table 3).
Mode of presentation with MBC
Patients with recurrent MBC were somewhat more likely to be treated with 1st LoT chemotherapy (27.4%, n=31) than those with de novo MBC (20.8%, n=15); this difference in treatment modality was not, however, statistically significant (p=0.31).
Influence of prior (neo)adjuvant treatment and MFI
All 124 patients with recurrent metastatic disease had received (neo)adjuvant endocrine therapy, with or without (neo)adjuvant chemotherapy. Sixteen patients (12.9%) developed metastatic disease within 24 months of primary diagnosis and 29 patients (23.4%) at least 120 months after that diagnosis; the median MFI was 64.9 months (range 5.6-339.2). This implies that around half of those developing recurrent MBC will have done so whilst on (neo)adjuvant endocrine therapy. Those with MFI longer than average will have been progressively less likely to have relapsed whilst receiving (neo)adjuvant endocrine therapy.
The length of MFI was associated with young age at index date, patients aged under 55 years having the shortest MFI (p=0.01); duration of MFI was not, however, associated with the likelihood of developing visceral metastasis (p=0.93). 1st LoT chemotherapy was significantly more common than 1st LoT endocrine therapy in patients with an MFI less than 60 months (61.3% and 40.2%, respectively; p=0.03), indicating increased use of chemotherapy in those relapsing whilst on (neo)adjuvant endocrine therapy.
Second LoT and subsequent SACT
Of the 185 patients receiving 1st LoT, 123 (66.5%) received subsequent SACT. The mean proportion changing to further treatment following 1st to 3rd LoT was 64.1%, falling to 51.8% following 4th to 6th LoT (Figure 1). Of those receiving 1st LoT chemotherapy, 39 (84.8%) had subsequent SACT, compared with 84 (60.4%) of those receiving 1st LoT endocrine (with/without targeted) therapy (p<0.01). The extent to which this subsequent treatment represents “maintenance” endocrine therapy without disease progression is, however, not known.
Following 1st LoT endocrine (with/without targeted) therapy or chemotherapy, 62 patients did not receive subsequent SACT; 36 of these patients (58.0%) died during or following 1st LoT, the remainder either having completed or continuing 1st LoT. Following 2nd LoT SACT, 48 patients did not receive further SACT, 29 (60.4%) having died during or following 2nd LoT. By the end of the study period, 61.1% treated patients had died, with the remainder being alive either on or off SACT (Figure 1).
The sequence of chemotherapy and endocrine (with/without targeted) therapy received by patients from 1st to 3rd LoT was diverse (Figure 2a). Of the 185 patients who received 1st LoT, 171 (92.4%) received endocrine (with/without targeted) therapy and 85 (45.9%) received chemotherapy at some time during the study period. When given, 2nd LoT was more often endocrine (with/without targeted) therapy (77.4%, n=96) than chemotherapy (22.6%, n=28) (Figure 2a). When given, 3rd LoT was again more often endocrine (with/without targeted) therapy (65.3%, n=49) than chemotherapy (34.7%; n=26).
There was substantial diversity in the sequence of SACT classes and regimens used (Additional file2). One line of endocrine therapy with a single agent AI, was the treatment sequence used most frequently (27.6%, n=51 patients). The next most common were three successive lines (10.3%, n=19) and two successive lines (9.2%, n=17) of endocrine therapy, each without targeted therapy. The fourth most common treatment sequence was the first to include chemotherapy, (3rd LoT capecitabine following endocrine therapy; 4.3%, n=8).
Patients receiving 1st LoT chemotherapy were more likely than those receiving endocrine (with/without targeted) therapy to receive subsequent chemotherapy (26.1% and 11.5%, respectively; p=0.02: Figure 2b). Regardless of 1st LoT modality, all patients receiving 2nd LoT chemotherapy either subsequently received further treatment or died. In contrast, 18 patients (18.8%) receiving 2nd LoT endocrine (with/without targeted) therapy remained alive with no further change of treatment until study end.
Overall survival
Median OS for the study cohort (n=192) following diagnosis of metastatic disease was 29.5 months (95% CI: 23.3-34.4). For patients receiving 1st LoT chemotherapy, median OS was 22.5 months compared to 31.7 months for those receiving endocrine (with/without targeted) therapy (p=0.11) (Figure 3). Median OS was 31.8 months in those with de novo MBC compared to 24.2 months in those with recurrent MBC (p=0.54); the apparent early survival benefit in patients with de novo MBC disappeared after 36 months (Figure 3). By contrast, OS was significantly better for patients with exclusively non-visceral metastasis than those with visceral metastasis (median 36.9 months and 22.8 months, respectively; p<0.01) and this benefit appeared to be regardless of whether patients had de novo or recurrent MBC (p=0.02) and age group (p=0.03) (Figure 3). OS was particularly poor for patients with recurrent MBC who had visceral metastasis (20.1 months).
Given the observed associations of young age, visceral metastasis, presentation with recurrent disease and 1st LoT chemotherapy, a Cox PH analysis was undertaken to further investigate overall survival. Site of metastasis and modality of 1st LoT were highly correlated (p<0.01) and therefore prone to collinearity in the regression model; modality of 1st LoT was subsequently dropped from the final model. Presentation with recurrent disease was not a significant explanatory variable for survival and was also dropped, leaving site of metastasis and age. There were 181 patients available for analysis and 72 patients (39.8%) were censored. Age (p=0.04) and site of metastasis (p<0.01) were both significant determinants of OS in the combined model; when adjusted for age, the HR for death for patients with visceral metastasis was 1.91 (95% CI: 1.27-2.89) relative to patients with non-visceral metastasis only.