Mortality and recurrence rates among systemically untreated high risk breast cancer patients included in the DBCG 77 trials

Abstract

Background: Following loco-regional treatment for early breast cancer accurate prognostication is essential for communicating benefits of systemic treatment. The aim of this study was to determine time to recurrence and long-term mortality rates in high risk patients according to patient characteristics and subtypes as assigned by immunohistochemistry panels.

Patients and methods: In November 1977 through January 1983, 2862 patients with tumors larger than 5 cm or positive axillary nodes were included in the DBCG 77 trials. Archival tumor tissue from patients randomly assigned to no systemic treatment was analyzed for ER, PR, Ki67, EGFR and HER2. Intrinsic subtypes were defined as follows: Luminal A, ER or PR >0%, HER2-negative, PR >10% and Ki67 < 14%; Luminal B, ER or PR >0%, (PR ≤10% or HER2-positive or Ki67 ≥ 14%); HER2E, ER 0%, PR 0%, HER2 positive; Core basal, ER 0%, PR 0%, HER2 negative and EGFR positive. Multivariate categorical and fractional polynomials (MFP) models were used to construct prognostic subsets by clinicopathologic characteristics.

Results: In a multivariate model, mortality rate was significantly associated with age, tumor size, nodal status, invasion, histological type and grade, as well as subtype classification.

Conclusions: With 35 years of follow-up, in this population of high-risk patients with no systemic therapy, no subgroup based on a composite prognostic score and/or molecular subtypes could be identified without excess mortality as compared to the background population.

Introduction

Early breast cancer without recognizable distant metastases is potentially curable but long-term survival is extremely rare in entirely untreated patients [1]. Today, the general treatment principle is to obtain complete loco-regional eradication of cancer tissue with the addition of systemic treatment dependent on tumor biology and the risk of recurrence for the individual patient. Adjuvant tamoxifen and chemotherapy were widely introduced in the early eighties to those who at the time were considered high risk, i.e., those with large tumors, lymph node metastases or invasion of the skin or deep fascia. The evidence was provided by the overviews of Early Breast Cancer Trialists’ Collaborative Group (EBCTCG) [2,3]. More recently the EBCTCG have demonstrated that modern chemotherapy reduces 10-year breast cancer mortality by about a third and importantly, the proportional benefits were similar in older and younger women and independent of age, nodal status, estrogen receptor status and type of chemotherapy regimen [4–6]. Systemic treatment has gradually been extended and will (depending on tumor characteristics) in most patients include endocrine therapy, HER2 targeting and/or chemotherapy [7].

Even in the absence of systemic treatment, about a third of high-risk breast cancer patients will, with efficient local treatment, be free of recurrence at 10 years [8,9]. A wide variation in the risk of recurrence has been demonstrated within different pathological stages [10]. Combining pathological stage with breast cancer subtypes may more accurately determine prognosis even in high risk patients. At least four clinically relevant subtypes – luminal A, luminal B, HER2-Enriched (HER2E) and basal-like – are now considered useful. Initially discovered on gene expression microarrays, these intrinsic molecular subtypes can be determined on clinical specimens using multigene assays such as PAM50 (Prosigna), but can also with reasonable accuracy be identified by inexpensive, widely-accessible immunohistochemistry panels [11,12].

In this study, we evaluated in systemically untreated high-risk breast cancer patients the ability of subtypes as designated by immunohistochemistry panels to predict long-term survival and recurrence. Standard of care for high-risk patients has for a long time included adjuvant systemic treatment, and therefore this patient cohort with long-term, high quality detailed follow-up enables an exceptional study.

Patients and methods

In 1977, the Danish Breast Cancer Cooperative Group (DBCG) provided its first standard diagnostic and treatment protocols for early invasive breast cancer [13]. At the same time, the clinical DBCG database was established and has since prospectively accumulated diagnostic, therapeutic and follow-up data by the use of standardized forms nationwide [14]. The aim of the current study was among high-risk breast cancer patients to compare time to recurrence and long-term mortality according to intrinsic subtype.

Patients

Eligible for registration in the 77 protocol program was patients who were without previous or synchronous malignant disease, who were without evidence of advanced disease by physical examination, radiography of the chest and bone or bone scintigraphy, and who achieved a complete resection of a unilateral invasive adenocarcinoma of the breast by mastectomy with axillary sampling or clearance (level I and part of level II). Patients with axillary lymph node metastasis, tumors >5 cm, or invasion of the deep fascia without distant metastasis were, if premenopausal, eligible for the 77B trial, and if postmenopausal for the 77C trial.

Pathology

Surgical specimens were classified in a predetermined manner including tumor size, examination of tumor margins, invasion into skin or deep fascia, measurement of gross tumor size, total number of lymph nodes identified and number of metastatic nodes. Classification of histological type was performed according to WHO and grade (ductal carcinomas) according to Elston and Ellis [15,16].

Centralized collection of specimens and immunohistochemical classification have been described earlier [9,12]. In brief, from patients enrolled in 77B and 77C, formalin-fixed, paraffin-embedded blocks from primary excisional surgery specimens were retrospectively collected from the corresponding pathology departments. These were used to construct duplicate 0.6 mm core tissue microarrays, from which sections for immunohistochemistry were cut. The definition of subtypes used in prior publications [9,12] was maintained: luminal A = hormone receptor positive (i.e., ER or PR >0%), HER2-negative, PR >10% and Ki67 < 14%; luminal B = hormone receptor positive but (PR ≤10% or HER2-positive or Ki67 ≥ 14%); HER2E = ER negative, PR negative and HER2 positive; core basal = ER negative, PR negative, HER2 negative and EGFR positive.

Adjuvant treatment

Postoperatively, all patients received radiotherapy to the chest wall and regional nodes (40.92 Gy in 22 fractions, five fractions per week; or 36.60 Gy in 12 fractions, two fractions per week). In the 77B trial, patients were assigned randomly to one of four options: no systemic therapy, levamisole, oral cyclophosphamide, or oral cyclophosphamide plus methotrexate and fluorouracil [8]. Patients in the 77C trial were randomly assigned to no systemic therapy or tamoxifen [9].

None of the patients in the cohort analyzed in this current study received any adjuvant systemic treatment. Only patients randomly assigned to radiotherapy alone from the 77B trial and the 77C trial were included.

Follow-up

Treatment related adverse events and findings on clinical examination were recorded until a first event every third month during the first and second year, every six months during the third to fifth year, and thereafter annually to a total of 10 years. Biochemical tests and imaging were done when indicated by existing symptoms or signs. In addition, long-term follow-up was acquired on survival through linkage to the Danish Civil Registration System (CRS) [17].

Endpoints

The primary endpoint was standardized mortality ratio (SMR). Secondary endpoints included time to recurrence, and overall survival (OS). For SMR and OS, complete follow-up was achieved until 1 June 2017 by linkage to the CRS on individual level and retrieving mortality figures of the general Danish female population.

Statistical analysis

The DBCG Data Center undertook central review, query and analysis of data. Follow-up time was quantified in terms of a Kaplan–Meier estimate of potential follow-up [18]. Overall survival was calculated as the time elapsed from the date of definitive surgery until death from any cause, and was estimated using the Kaplan–Meier method. Time to recurrence was defined as the time from surgery to invasive loco-regional recurrence or distant metastases. New contralateral breast cancer, another malignancy or death without prior recurrence were counted as competing events. Cumulative incidence of recurrence in the presence of competing risk was estimated. Univariate and multivariate regression analyses were performed, using the Cox proportional hazards model for OS and Fine-Gray proportional hazards subdistribution model for recurrence. The number of deaths observed was compared with the number of deaths expected, calculated by applying age and calendar year specific female mortality figures of the general Danish population and the corresponding person-years of the respective cohort members. Time at risk was defined as time from definitive surgery until date of death from any cause, emigration or end of follow-up (1 June 2017). The SMR, computed as the ratio of the observed to the expected number of deaths, served as an estimate of relative risk of death, and 95% confidence intervals (CIs) were computed based on the assumption that the observed number of deaths followed a Poisson distribution. The SMR was analyzed using univariate and multivariate Poisson regression models.

Multivariable fractional polynomials (MFPs) were used to assess the functional form of continuous prognostic variables building the multivariate regression models [19]. Factors included in the multivariable analyses were age, tumor size, number of positive lymph nodes, fraction of positive vs retrieved lymph nodes, histological type (ductal, lobular, other histological types), grade (1, 2, 3), and invasion of the tumor into deep fascia. Only ductal carcinomas were graded, and for multivariate models, grade was set to two for non-ductal carcinomas with separate estimates for histologic type. The assumption of proportional hazards was assessed by Schoenfeld residuals and by including time-dependent covariates in the model. To comply with the assumption of proportional hazards, time-dependent components were included in the models. All p values are two-sided. Statistical analyses were done using SAS v9.4 (SAS Institute, Inc., Cary, NC, USA), and STATA IC 14 (StataCorp, College Station, TX, USA).

Results

The DBCG 77B and 77C trials enrolled 2862 patients between November 1977 and January 1982. Following mastectomy for early high-risk breast cancer 1100 (38%) patients were followed without adjuvant systemic treatment after postoperative radiotherapy. Biomarker analyses were performed on archived FFPE tumor blocks in 820 (75%) and subtype classification was available in 745 (68%). Mean age at mastectomy was 61 years. Table 1 summarizes patient and tumor characteristics and shows a high proportion of patients with large tumors and heavy nodal involvement. Among the patients with subtype classification, 199 (27%) were assigned as luminal A, 419 (56%) luminal B, 84 (11%) HER2E and 43 (6%) core basal. The estimated potential median follow-up was 10.0 years for first event and 37.6 years for OS.

Table 1. Patient and tumor characteristics by subtype.

			Subtype
	Total		Luminal A		Luminal B		HER2E		Core basal		NA
Characteristics	N	%	N	%	N	%	N	%	N	%	N	%
No. of patients	1100		199		419		84		43		355
Age
<50	187	17	13	7	58	14	9	11	8	19	99	28
50–59	272	25	51	26	84	20	34	40	9	21	94	26
60–69	370	34	66	33	162	39	29	35	14	33	99	28
≥70	271	25	69	35	115	27	12	14	12	28	63	18
Lymph node status
Negative	152	14	25	13	49	12	7	8	9	21	62	17
1–3 positive	591	54	118	59	220	53	44	52	19	44	190	54
4–9 positive	264	24	40	20	117	28	22	26	12	28	73	21
≥10 positive	39	4	6	3	17	4	8	10	1	2	7	2
Unknown	54	5	10	5	16	4	3	4	2	5	23	6
Total lymph nodes
0 retrieved	54	5	10	5	16	4	3	4	2	5	23	6
1–3 retrieved	259	24	48	24	92	22	17	20	7	16	95	27
4–9 retrieved	610	55	113	57	240	57	44	52	25	58	188	53
≥10 retrieved	177	16	28	14	71	17	20	24	9	21	49	14
Tumor size
0–20 mm	289	26	60	30	104	25	18	21	8	19	99	28
21–50 mm	587	53	106	53	233	56	47	56	25	58	176	50
>50 mm	213	19	33	17	78	19	19	23	10	23	73	21
Unknown	11	1	0	0	4	1	0	0	0	0	7	2
Deep fascia invasion
Absent	818	74	168	84	307	73	60	71	30	70	253	71
Present	275	25	31	16	111	26	24	29	12	28	97	27
Unknown	7	1	0	0	1	0	0	0	1	2	5	1
Histologic type
Ductal carcinoma	975	89	186	93	364	87	75	89	35	81	315	89
Lobular carcinoma	48	4	9	5	21	5	3	4	0	0	15	4
Other	77	7	4	2	34	8	6	7	8	19	25	7
Malignancy grade^a
Grade I	275	28	82	44	103	28	9	12	2	6	79	25
Grade II	534	55	94	51	205	56	47	63	16	46	172	55
Grade III	162	17	10	5	56	15	18	24	15	24	63	20
Unknown	4	0	0	0	0	0	1	1	2	6	1	0

HER2E: Her2-Enriched.

^a Ductal carcinoma only.

Analysis of recurrence

Within 2½ years, 20% of luminal A patients have a breast cancer recurrence increasing to 30% of luminal B and 40% or more in HER2E and core basal breast cancer patients. At 10 years, 526 patients (47.8%) had a breast cancer recurrence, including 450 with distant metastases, and 203 patients (18.5%) had a competing event. Factors significantly associated with recurrence were age, tumor size, nodal status, invasion and grade (all p < .01). Figure 1(A) shows the cumulative incidence of recurrence up to 10 years according to subtype. The corresponding hazard ratios are shown in Table 2, where estimates are split according to time after surgery due to non-proportional hazards for HER2E and core basal as compared to luminal B. There is a highly significant effect of subtype (p < .0001), which is maintained in the multivariate analysis. Considering all first events, including recurrence, contralateral breast cancer, other malignancy, and death as first event, the pattern looks similar (Figure 1(C)).

Figure 1. (Panel A) Cumulative Incidence estimates for recurrence (Rec) of systemically untreated patients with luminal A (Lum A), luminal B (Lum B), HER2-Enriched (HER2) and core basal (CB) breast cancer. (Panel B) Kaplan–Meier estimates of overall survival for systemically untreated patients with luminal A (Lum A), luminal B (Lum B), HER2-Enriched (HER2) and core basal (CB) breast cancer. Estimates for OS at 5, 10 and 20 years after surgery; Lum A 67% (60;73), 46% (39;53), 20% (14;25), Lum B 56% (51;61), 28% (24;33), 13% (10;16), HER2E 41% (30;51), 31% (21;41), 21% (13;31), core basal 42% (27;56), 35% (21;49), 16% (7;29). (Panel C) Kaplan–Meier estimates of disease-free survival (considering events of recurrence, contralateral breast cancer, other malignancy, and death as first event) for systemically untreated patients with luminal A (Lum A), luminal B (Lum B), HER2-Enriched (HER2) and core basal (CB) breast cancer. Patients at risk according to time after surgery listed below x-axis.

Table 2. Proportional hazards models for recurrence and overall survival according to subtype.

		Years after surgery
Subtype	p	HR	95% CI	HR	95% CI	HR	95% CI	Pheterogeneity
Recurrence		0–2½		2½–10
Univariate	<.0001
Luminal A		0.63	(0.43;0.90)	0.67	(0.47;0.96)			.78
Luminal B		Reference
HER2E		1.97	(1.35;2.88)	0.39	(0.18;0.85)			.0002
Core basal		1.55	(0.93;1.60)	0.36	(0.13;1.02)			.01
Multivariate	.001
Luminal A		0.79	(0.55;1.13)	0.69	(0.47;1.00)			.61
Luminal B		Reference
HER2E		1.66	(1.11;2.49)	0.35	(0.16;0.76)			.0004
Core Basal		1.44	(0.84;2.47)	0.40	(0.14;1.15)			.03
OS		0–5		5–10		>10
Univariate	<.0001
Luminal A		0.67	(0.51;0.87)	0.57	(0.41;0.81)	1.09	(.84;1.42)	.03
Luminal B		Reference
HER2E		1.75	(1.29;2.37)	0.44	(0.22;0.89)	0.61	(.39;.95)	<.0001
Core Basal		1.60	(1.06;2.42)	0.30	(0.10;0.94)	0.60	(.32;1.10)	.002
Multivariate	p < .0001
Luminal A		0.79	(0.60;1.03)	0.60	(0.43;0.85)	0.97	(.75;1.26)	.08
Luminal B		Reference
HER2E		1.72	(1.26;2.34)	0.46	(0.22;0.93)	0.48	(.31;.75)	<.0001
Core Basal		1.55	(1.01;2.37)	0.36	(0.11;1.11)	0.63	(.34;1.16)	.008

HER2E: Her2-Enriched; HR: hazard ratio; CI: confidence interval; OS: overall survival.

Hazard ratios for unadjusted and adjusted estimates. p Values for overall significance and for test of heterogeneity in relation to years after surgery (non-proportional hazard).

Analysis of OS and SMR

Of the 1100 women in the study cohort, 1050 died (35-year OS 5.5%; 95% CI 4.3–7.0) while 372 deaths were expected (SMR 2.82; 95% CI 2.65–2.99, p < .0001).

Factors significantly associated with mortality (OS and SMR) were age, tumor size, nodal status and grade. Figure 1(B) shows the OS up to 35 years after surgery. Corresponding figures for OS at 5, 10 and 20 years after surgery are listed in the figure legend. Median survival was distinct across subtypes, with 8.6 years for patients in the luminal A group as opposed to 3.1 years in the HER2E group of patients. The relative risk according to follow-up time and subtype is shown in Table 2, both for univariate and multivariate analyses, with an overall statistically significant effect (p < .0001).

According to subtype, the number of observed vs expected deaths was 407 and 126.2 for luminal B, 194 and 89.1 for luminal A, 80 and 32.0 for HER2E and 39 vs 15 for core basal. The empirical SMR overall and divided by time of follow-up for each subtype are shown in Table 3.

Table 3. Standardized mortality ratio according to subtype, overall and in relation to years after surgery.

	Years after surgery
	0–5		5–10		>10		Overall
Subtype	SMR	95% CI	SMR	95% CI	SMR	95% CI	SMR	95% CI
Lum A	3.20	(2.52;4.08)	2.14	(1.58;2.91)	1.76	(1.43;2.17)	2.18	(1.89;2.51)
Lum B	5.13	(4.43;5.92)	4.46	(3.73;5.35)	1.65	(1.37;2.00)	3.23	(2.93;3.55)
HER2E	13.0	(9.87;17.2)	2.53	(1.26;5.05)	0.88	(0.58;1.33)	2.50	(2.01;3.11)
Core Basal	10.1	(6.81;14.9)	1.35	(0.44;4.19)	1.07	(0.59;1.93)	2.60	(1.90;3.56)

Lum A: luminal A; Lum B: luminal B; HER2E: Her2-Enriched; CI: confidence interval; SMR: standardized mortality ratio.

Identifying patients free of excess mortality

By summing the estimated regression coefficients (β) of age, tumor size, nodal status, deep fascia invasion, and histological type and malignancy grade in the multivariate model for SMR (Supplementary Table 1), we constructed a clinical prognostic score index (cPSI). When patients were divided according to subtypes and cPSI, even those with the lowest decile cPSI combined with a luminal A subtype had an excess mortality as compared to the background population (data not shown), and with excess mortality progressively increasing with increasing score.

Discussion

The result from this historical follow-up clearly shows the appalling consequences of omitting systemic treatment in high-risk breast cancer patients and recalls the medical need of high-risk breast cancer patients not receiving adjuvant systemic treatment. The retrospective determination of breast cancer subtypes revealed a distinct pattern separating luminal A breast cancer from luminal B and both of them from HER2E and core-basal breast cancers. Two and a half years after primary surgery, recurrence rate was more than twice as high in patients with HER2E and core-basal breast cancers as compared to luminal A, while patients with luminal B breast cancers were in between. The diversity in mortality was even more separated among the four subtypes after 5 years. In years 5–10, hazards reversed and mortality approximated to women in background population after 5 years in patients with core-basal cancers and after 10 years in patients with HER2E breast cancer. Patients with luminal breast cancer persistently had an excess mortality as described previously and only 20% were alive at 20 years [20–22]. We were, within this group of high-risk patients, unable to identify a subgroup without major excess mortality, even when combining subtypes with other known prognostic factors. Overall, our data reinforce that all clinically high-risk patients should be recommended some form of systemic treatment even with a luminal A subtype.

The strengths of our study include a prospectively defined and identified study cohort; use of standardized breast cancer management including diagnostic procedures, standardized loco-regional treatment with mastectomy and radiotherapy; and long-term, high quality detailed follow-up. In contrast, treatment information in the SEER registry has been inaccurate [23,24]. The access to complementing registries furthermore allowed us to calculate SMR, which compared to OS is less sensitive to competing mortality occurring over time from deaths from other causes, e.g., ischemic heart disease, stroke and non-breast cancer [25]. Our ability to analyze SMR according to subtype was however hampered by the considerable excess mortality in the first 10 years after diagnosis of breast cancer.

The group of long-term survivors in our study were very small, and clearly a limitation that we were not able to characterize this subset. One possible explanation was that we use immunohistochemistry to determine intrinsic subtype, as opposed to a gene based set-up as, for example the PAM50 or the OncotypeDx. In years 2½–10, patients with luminal subtypes had a higher rate of recurrence as compared to patients with core basal and HER2E subtypes, but unfortunately patients were followed for 10 years only and we are unable to describe the pattern of recurrence beyond 10 years. Immunohistochemical panels provide less prognostic information and do not contain the same level of analytical reproducibility as genomic-based nucleic acid tests [26,27]. In particular, immunohistochemical assessment of Ki67 and PR has analytic variability issues [28,29], and the predefined cutoff used for Ki-67 may have been suboptimal [30]. Other limitations include lack of availability of some tumor blocks and of preanalytical handling guidelines for older specimens. While our study suggests that adjuvant systemic treatment is required in high-risk breast cancer patients, it does not elucidate the possible benefits [12]. Furthermore, our data do not in any way contradict the growing evidence supporting de-escalation of systemic treatment to many low- or moderate-risk breast cancers [31].

In conclusion, if adjuvant systemic treatment is omitted the vast majority of high-risk breast cancer patients are affected by excess mortality and it is not possible at the time of diagnosis to identify the relatively few good prognosis patients.

Disclosure statement

No potential conflict of interest was reported by the authors.