High chromosomal instability is associated with higher 10‐year risks of recurrence for hormone receptor‐positive, human epidermal growth factor receptor 2‐negative breast cancer patients: clinical evidence from a large‐scale, multiple‐site, retrospective study

Abstract

Long‐term survival varies among hormone receptor‐positive (HR+) and human epidermal growth factor receptor 2‐negative (HER2−) breast cancer patients and is seriously impaired by metastasis. Chromosomal instability (CIN) was one of the key drivers of breast cancer metastasis. Here we evaluate CIN and 10‐year invasive disease‐free survival (iDFS) and overall survival (OS) in HR+/HER2−– breast cancer. In this large‐scale, multiple‐site, retrospective study, 354 HR+/HER2− breast cancer patients were recruited. Of these, 204 patients were used for internal training, 70 for external validation, and 80 for cross‐validation. All medical records were carefully reviewed to obtain the disease recurrence information. Formalin‐fixed paraffin‐embedded tissue samples were collected, followed by low‐pass whole‐genome sequencing with a median genome coverage of 1.86X using minimal 1 ng DNA input. CIN was then assessed using a customized bioinformatics workflow. Three or more instances of CIN per sample was defined as high CIN and the frequency was 42.2% (86/204) in the internal cohort. High CIN correlated significantly with increased lymph node metastasis, vascular invasion, progesterone receptor negative status, HER2 low, worse pathological type, and performed as an independent prognostic factor for HR+/− breast cancer. Patients with high CIN had shorter iDFS and OS than those with low CIN [10‐year iDFS 11.1% versus 82.2%, hazard ratio (HR) = 11.12, p < 0.01; 10‐year OS 45.7% versus 94.3%, HR = 14.17, p < 0.01]. These findings were validated in two external cohorts with 70 breast cancer patients. Moreover, high CIN could predict the prognosis more accurately than Adjuvant! Online score (10‐year iDFS 11.1% versus 48.6%, HR = 2.71, p < 0.01). Cross‐validation analysis found that high consistency (83.8%) was observed between CIN and MammaPrint score, while only 45% between CIN and Adjuvant! Online score. In conclusion, high CIN is an independent prognostic indicator for HR+/HER2− breast cancer with shorter iDFS and OS and holds promise for predicting recurrence and metastasis.

Introduction

Breast cancer is a heterogeneous disease, with an estimated 2.26 million new cases worldwide in 2020 [1]. Although there have been recent advances based on molecular subtyping, for patients with late recurrences and metastases the 10‐year breast cancer‐specific mortality approaches 50% [2]. In PR China, nearly 50–60% of patients are hormone receptor‐positive (HR+) and human epidermal growth factor receptor 2‐negative (HER2−) [3]. Compared with other subtypes, HR+/HER2− breast cancer has lower recurrence risk in the first 5 years post‐surgery, but has high recurrence risk in the second 5 years; moreover, the recurrence risk remains stable for a long time [4, 5, 6, 7]. Therefore, for HR+/HER2− breast‐cancer patients with high recurrence risk, the National Comprehensive Cancer Network (NCCN) guidelines recommend 10 years of adjuvant‐intensive endocrine therapy or adjuvant chemotherapy [8]. Patients with better prognoses do not require adjuvant chemotherapy, avoiding overtreatment‐related toxic side effects and financial costs [9]. Thus, accurate recurrence risk classification of HR+/HER2− breast‐cancer cases is crucial for more precise clinical decisions.

Several risk scores and prognostic models have been developed to predict breast‐cancer recurrence risk. The clinically used Adjuvant! Online (AO) has focused strongly on clinicopathological features [10, 11]. Despite being inexpensive to assay and ubiquitous, features such as age, tumor size, lymph node status, tumor grade, estrogen receptor (ER) status, and comorbidity can be influenced by subjectivity and pathological discordance level [8, 12]. Risk assessments incorporating genomic features may improve objectivity, accuracy, and robustness.

Prognostic models based on multigenic parameters are expected to accurately predict breast cancer recurrence, including Oncotype DX (21 genes), MammaPrint (MMP) (70 genes), EndoPredict (12 genes), PAM50 (50 genes), Breast Cancer Index (BCI), and RecurIndex (28 genes) [13, 14]. Based on clinical studies, indicators for different populations have identified and play complementary roles. However, there are differences between these models in the types and numbers of genes and in suitable patients; therefore, none can completely reflect the overall clinical situation.

Approximately 89% invasive breast cancers exhibit aneuploidy. Such chromosomal instability (CIN) is an important feature of cancer that plays an important role in its occurrence, development, and drug resistance [15, 16, 17, 18]. For example, mutations in hBUB1, MAD2, BRCA1, BRCA2, hCDC4, and TP53 are associated with CIN, which promotes tumor development [19, 20]. Moreover, CIN regulates gene copy number by increasing copy numbers of chromosomes with oncogenes and reducing the copy number of chromosomes with tumor suppressor genes [21]. CIN not only drives tumor heterogeneity, mediating tumor biological behavior, but also promotes tumor metastasis by causing chronic inflammation through the cyclic GMP‐AMP synthase‐stimulator of interferon genes pathway [22, 23]. Reducing CIN in cancer cells without affecting other genetic abnormalities can disable metastasis [22]. With the advancement of low‐pass whole‐genome sequencing (LPWGS), CIN assessments can be easily performed on tumor samples, even those that are formalin‐fixed paraffin‐embedded. Herein, we analyzed CIN in HR+/HER2− breast‐cancer patients with LPWGS, investigated its relationship with breast‐cancer clinicopathological features, and further assessed its utility as a prognostic predictor.

Materials and Methods

Patients

Data were collected from patients admitted to Zhejiang Cancer Hospital from 2002 to 2019, Jinhua Municipal Central Hospital from 2014 to 2020, and The First Hospital of Jiaxing from 2010 to 2019. Patients diagnosed with HR+/HER2− breast cancer without other malignant tumors were enrolled. In total, 354 patients were enrolled from three hospitals (Figure 1). This study was approved by the Clinical Research Ethics Committee of each institution. Written informed consent was obtained from all subjects.

Figure 1.

Study design. Patients with HR+/HER2− breast cancer with complete clinicopathological and prognostic information provided informed consent for us to obtain archival primary tumor tissue and perform LPWGS. BC, breast cancer; HER2, human epidermal growth factor receptor 2; HR, hormone receptor; LPWGS, low‐pass whole‐genome sequencing.

Data collection

Patient clinicopathological characteristics were collected, including age at onset, tumor size, lymph node status, pathological type, pathological grade, vascular invasion, and ER, progesterone receptor (PR), and HER2 status. ER+ and/or PR+ ≥10% nuclear staining were scored as HR+. C‐erbB‐2 0 or 1+ assessed by immunohistochemistry was scored as HER2−, while a score of 2+ corresponding to HER2 status was further confirmed through fluorescence in situ hybridization. Patients were followed up at 3‐month intervals during the first 2 years postsurgery, at 6‐month intervals at 3–5 years postsurgery, and once annually thereafter. If any signs or symptoms were detected at any time during this period, a review was required. Most patients were followed up regularly until September 2021. Invasive disease‐free survival (iDFS) was defined as the time interval between the date of surgery and the first day of diagnosis of ipsilateral invasive breast tumor, locoregional invasive disease, distant recurrence, contralateral invasive breast cancer, second primary invasive cancer, or death from any cause [24]. Overall survival (OS) was defined as the time from the date of surgery to death from any cause [24].

LPWGS

Sequencing was performed using a customized workflow described previously [25, 26]. Generally, prepared libraries were sequenced using the HiSeq Xten platform (Illumina, San Diego, CA, USA) and the average coverage for each 200 kbp genomic bin was evaluated by SAMtools mpileup; the coverage of each bin was normalized using Formula (1) [25, 26]:

$$\text{coverag}{\mathrm{e}}_{\text{normalized}}=\frac{\text{coverag}{\mathrm{e}}_{\mathrm{raw}}-\text{mean}\left(\text{coverag}{\mathrm{e}}_{\text{controls},\mathrm{raw}}\right)}{\mathrm{SD}\left(\text{coverag}{\mathrm{e}}_{\text{controls},\mathrm{raw}}\right)}$$ (1)

Chromosomal copy number change $\text{coverag}{\mathrm{e}}_{\mathrm{chr},\mathrm{raw}}$ is summarized as the average of all bins of this chromosome. Z _chr is the Z‐score of the chromosome calculated using Formula (2):

$$Z_{\mathrm{chr}}=\frac{\text{coverag}{\mathrm{e}}_{\mathrm{chr},\mathrm{raw}}-\text{mean}\left(\text{coverag}{\mathrm{e}}_{\mathrm{chr},\text{controls},\mathrm{raw}}\right)}{\mathrm{SD}\left(\text{coverag}{\mathrm{e}}_{\mathrm{chr},\text{controls},\mathrm{raw}}\right)}$$ (2)

A cutoff value Z _chr ≥ 3 indicates copy gain, whereas Z _chr ≤ −3 indicates copy loss. A chromosome with |Z _chr| ≥ 3 was defined as an incidence of CIN. Generally, the higher the CIN, the higher the recurrence rate. The step‐forward method was used to determine the optimal cutoff value. For each cutoff analysis, patients were divided into two groups. The R package ‘survival’ was used to calculate hazard ratios (HRs). We started with chromosomal copy number variation (CNV) of 1 as the cutoff. HRs were then calculated using a Cox proportional hazards model. Then 2, 3, and more were used as the cutoff. The best CIN count cutoff was determined once the maximum HR was identified. Finally, at least three chromosomes were scored as high CIN.

Statistical analysis

All analyses were performed using SPSS v26 (IBM, Armonk, New York, USA) and R v4.3.0 (R Foundation for Statistical Computing, Vienna, Austria). The p values <0.05 were considered statistically significant. Pearson's chi‐square test and Fisher's exact test were used to compare clinicopathological features between CIN groups. A multivariate Cox proportional risk model was established; potential correlations between CIN, iDFS, and OS were analyzed, and the HR was estimated. Survival was calculated from the date of surgery. Both iDFS and OS were estimated using the Kaplan–Meier method.

Results

CIN and CNVs

CIN in HR+/HER2− breast‐cancer tissue samples from Zhejiang Cancer Hospital was assayed using LPWGS with a median genome coverage of ×1.86 (range, ×1.03–×3.17). On average, seven chromosomal CNVs were present per sample (range, 0–51). Using the step‐forward method, we scored ≥3 CINs/sample as high CIN and <3 as low; 118 patients had low and 86 had high CIN, with median follow‐up times of 124 and 83.5 months, respectively. Patients with recurrence had more CNVs; the most common CNVs were chromosomes 1q, 8q, 12q, 17q, 17p, 8p, 22q, 11q, 14q, 9p, 6q, and 13q (supplementary material, Figure S1).

Patients and clinicopathological features

We enrolled 204 patients with HR+/HER2− breast cancer admitted to Zhejiang Cancer Hospital between 2002 and 2019. Clinicopathological data of all patients were collected; baseline characteristics are summarized in Table 1. The median age of the 204 patients at initial diagnosis whose tumor DNA was sequenced was 47 (range, 28–84) years. Lymph node metastasis was observed in 68.1% (139/204) patients. Invasive ductal carcinoma was diagnosed in 88.7% (181/204) patients. Pathological grade II or higher was diagnosed in 94.6% (193/204) patients. All patients underwent surgery, including 100 with postoperative recurrence or metastasis and 104 without recurrence or metastasis. Median follow‐up times were 96 and 118 months for patients with recurrent and nonrecurrent disease, respectively. Furthermore, 88.7% (181/204) patients received (neo)adjuvant chemotherapy; 38.2% (78/204) received adjuvant radiotherapy; and 80.9% (165/204) received adjuvant endocrine therapy.

Table 1.

Comparison of the difference in clinicopathological characteristics between low CIN and high CIN groups

Clinical characteristic	All patients (n = 204)	Low CIN (n = 118)	High CIN (n = 86)	p
Age at onset				0.06
Median (range)	47 (28–84)	49 (29–84)	46 (28–81)
BMI				0.72
<24	118 (57.8)	67 (56.8)	51 (59.3)
≥24	86 (42.2)	51 (43.2)	35 (40.7)
Age at menarche				0.94
<15	66 (32.4)	38 (32.2)	28 (32.6)
≥15	135 (66.2)	77 (65.3)	58 (67.4)
Unknown	3 (1.5)	3 (2.5)	0 (0.0)
Menopause at diagnosis				0.98
Yes	75 (36.8)	43 (36.4)	32 (37.2)
No	126 (61.8)	72 (61.0)	54 (62.8)
Unknown	3 (1.5)	3 (2.5)	0 (0.0)
Pregnancy frequency				0.64
0	5 (2.5)	2 (1.7)	3 (3.5)
1–2	169 (82.8)	96 (81.4)	73 (84.9)
≥3	27 (13.2)	17 (14.4)	10 (11.6)
Unknown	3 (1.5)	3 (2.5)	0 (0.0)
Tumor size (cm)				0.20
≤2	50 (24.5)	33 (28.0)	17 (19.8)
>2	153 (75.0)	85 (72.0)	68 (79.1)
Unknown	1 (0.5)	0 (0.0)	1 (1.2)
Lymph node status				<0.01
0	65 (31.9)	45 (38.1)	20 (23.3)
1–3	90 (44.1)	58 (49.2)	32 (37.2)
≥4	49 (24.0)	15 (12.7)	34 (39.5)
Pathological type				0.01
Invasive special carcinoma with good prognosis*	7 (3.4)	7 (5.9)	0 (0.0)
IDC	181 (88.7)	105 (89.0)	76 (88.4)
ILC	9 (4.4)	5 (4.2)	4 (4.7)
Other types with poor prognosis †	7 (3.4)	1 (0.8)	6 (7.0)
Grade				0.15
I	11 (5.4)	7 (5.9)	4 (4.7)
II	143 (70.1)	88 (74.6)	55 (64.0)
III	50 (24.5)	23 (19.5)	27 (31.4)
Vascular invasion				<0.01
Yes	64 (31.4)	25 (21.2)	39 (45.3)
No	140 (68.6)	93 (78.8)	47 (54.7)
ER status				0.41
Positive	199 (97.5)	116 (98.3)	83 (96.5)
Negative	5 (2.5)	2 (1.7)	4 (3.5)
PR status				0.01
Positive	162 (79.4)	101 (85.6)	61 (70.9)
Negative	42 (20.6)	17 (14.4)	25 (29.1)
HER2 expression				<0.01
0	116 (56.9)	80 (67.8)	36 (41.9)
Low	88 (43.1)	38 (32.2)	50 (58.1)
Endocrine therapy				0.40
TAM	95 (46.6)	54 (45.8)	41 (47.7)
AI ± TAM	70 (34.3)	35 (29.7)	35 (30.7)
None	15 (7.4)	6 (5.1)	9 (10.5)
Unknown	24 (11.8)	23 (19.5)	1 (1.2)
Adjuvant radiotherapy				<0.01
Yes	78 (38.2)	33 (28.0)	45 (52.3)
No	126 (61.8)	85 (72.0)	41 (47.7)
(Neo)Adjuvant chemotherapy				0.01
Taxanes or anthracyclines	77 (37.7)	53 (44.9)	24 (27.9)
Taxanes + anthracyclines or taxanes + platinum	104 (51.0)	49 (41.5)	55 (64.0)
None	22 (10.8)	15 (12.7)	7 (8.1)
Unknown	1 (0.5)	1 (0.8)	0 (0.0)
Surgery				0.61
Breast‐conserving surgery	14 (6.9)	9 (7.6)	5 (5.8)
Mastectomy	190 (93.1)	109 (92.4)	81 (94.2)

Data are represented as n (%). Bold text indicates p values that showed significance.

AI, aromatase inhibitor; CIN, chromosomal instability; ER, estrogen receptor; IDC, invasive ductal carcinoma; ILC, invasive lobular carcinoma; PR, progesterone receptor; TAM, tamoxifen.

^*Invasive special carcinoma with good prognosis included tubular, mucinous, papillary, and cribriform carcinoma.

^† Other types with poor prognosis included micropapillary carcinoma and carcinoma simplex.

Relationships between CIN and clinicopathological characteristics

Associations between CIN levels and clinicopathological variables are summarized in Table 1. High CIN levels were significantly correlated with more lymph node metastases (p < 0.01), vascular invasion (p < 0.01), PR negativity (p = 0.01), HER2 low expression (p < 0.01), and worse pathological type (p = 0.01), but not with age at onset, body mass index, age at menarche, pregnancy frequency, or tumor size. Compared with patients with low CIN, those with high CIN received more intensive adjuvant therapy, such as radiotherapy and chemotherapy with taxanes and anthracyclines.

High CIN is associated with poor survival

Using Kaplan–Meier analyses, we observed a significant difference in 5‐year and 10‐year iDFS between CIN‐low and CIN‐high groups. Patients with high CIN had shorter iDFS than those with low CIN (5‐year iDFS, 40.5% versus 94.6%; 10‐year iDFS, 11.1% versus 82.2%; HR = 11.12, p < 0.01; Figure 2A; supplementary material, Table S1). Patients with high CIN had shorter OS than those with low CIN (5‐year OS, 72.4% versus 98.0%; 10‐year OS, 45.7% versus 94.3%; HR = 14.17, p < 0.01; Figure 2B; supplementary material, Table S1). Patients with HR+/HER2− breast cancer with high CIN had a greater risk of recurrence and death.

Figure 2.

High CIN is associated with poor prognosis. Survival (Kaplan–Meier) as a function of CIN status of patients recruited in Zhejiang Cancer Hospital for (A) iDFS and (B) OS of all patients, and (C) iDFS and (D) OS of lymph node metastasis‐negative patients. CIN, chromosomal instability; iDFS, invasive disease‐free survival; OS, overall survival.

Of 65 patients without lymph node metastases, 45 and 20 had low and high CIN, respectively. Furthermore, 6.2% (4/65) patients did not receive endocrine therapy, and 81.5% (53/65) did not receive postoperative radiotherapy; 16.9% (11/65) patients did not receive (neo)adjuvant chemotherapy, but 81.5% (53/65) did (supplementary material, Table S2). We found no significant differences in treatment patterns between patients with low and high CIN; however, those with high CIN had more frequent recurrences (HR = 5.71, p < 0.01; Figure 2C) and poorer survival (HR = 7.23, p < 0.01; Figure 2D).

CIN is an independent prognostic indicator

Univariate Cox regression showed the following were associated with shorter iDFS: high CIN (HR =10.79, p < 0.01), >3 lymph node metastases (HR = 3.71, p < 0.01), age < 35 years at onset (HR = 2.65, p < 0.01), vascular invasion (HR = 2.12, p < 0.01), HER2 expression (HR = 2.24, p < 0.01), and PR negativity (HR = 1.87, p = 0.01). High CIN (HR = 10.26, p < 0.01), >3 lymph node metastases (HR = 2.52, p = 0.01), and vascular invasion (HR = 2.28, p < 0.01) were significantly associated with shorter OS, whereas other factors, including tumor size, pathological type, and grade, were not significantly associated with OS (Table 2).

Table 2.

Univariate parameter Cox regression analysis

Parameters	iDFS		OS
	HR (95% CI)	p	HR (95% CI)	p
CIN
Low	1		1
High	10.79 (6.63–17.55)	<0.01	10.26 (5.01–21.02)	<0.01
Age at onset
≥35	1		1
<35	2.65 (1.47–4.75)	<0.01	1.36 (0.58–3.21)	0.48
Menopause at diagnosis
Yes	1		1
No	0.96 (0.64–1.44)	0.82	0.87 (0.50–1.50)	0.61
Tumor size (cm)
≤2	1		1
>2	1.36 (0.84–2.21)	0.21	1.29 (0.68–2.45)	0.43
Lymph node status
0	1		1
1–3	1.32 (0.80–2.22)	0.30	1.02 (0.52–2.01)	0.95
≥4	3.71 (2.18–6.31)	<0.01	2.52 (1.27–4.99)	0.01
Pathological type
IDC	1		1
ILC	0.84 (0.31–2.29)	0.73	0.75 (0.18–3.08)	0.69
Other types with poor prognosis*	2.11 (0.98–4.57)	0.06	0.65 (0.15–2.76)	0.56
Invasive special carcinoma with good prognosis †	0.00 (0.00–5.733E+170)	0.95	0.00 (0.00–2.065E+251)	0.97
Grade
I	1		1
II	0.96 (0.39–2.73)	0.92	1.17 (0.28–4.85)	0.83
III	1.05 (0.40–2.74)	0.92	1.66 (0.38–7.19)	0.50
Vascular invasion
No	1		1
Yes	2.12 (1.41–3.19)	<0.01	2.28 (1.34–3.89)	<0.01
PR status
Positive	1		1
Negative	1.87 (1.20–2.91)	0.01	1.35 (0.74–2.48)	0.33
HER2 expression
0	1		1
Low	2.24 (1.50–3.35)	<0.01	1.65 (0.97–2.86)	0.06

Bold text indicates HRs that showed significance.

CI, confidence interval; CIN, chromosomal instability; HR, hazard ratio; IDC, invasive ductal carcinoma; iDFS, invasive disease‐free survival; ILC, invasive lobular carcinoma; OS, overall survival; PR, progesterone receptor.

^*Other types with poor prognosis included micropapillary carcinoma and carcinoma simplex.

^† Invasive special carcinoma with good prognosis included tubular, mucinous, papillary, and cribriform carcinoma.

We chose factors associated with iDFS and OS from the univariate analysis for the multivariate Cox regression analysis (Figure 3; supplementary material, Table S3). CIN was an independent prognostic factor for both iDFS and OS (HR =8.55, p < 0.01; HR = 9.37, p < 0.01, respectively). Other independent risk factors associated with breast‐cancer recurrence and death were age at onset (HR = 2.25, p = 0.02) and vascular invasion (HR = 2.07, p = 0.03).

Figure 3.

CIN is an independent prognostic indicator for HR+/HER2− breast cancer patients. Evaluation of CIN and clinicopathological characteristics using multiple‐parameter Cox regression forest plots with HR, 95% CI and p‐values for (A) iDFS and (B) OS. AI, aromatase inhibitor; CI, confidence interval; CIN, chromosomal instability; ER, estrogen receptor; HR, hazard ratio; iDFS, invasive disease‐free survival; OS, overall survival; PR, progesterone receptor; TAM, tamoxifen. Bold text indicates statistically significant HRs.

CIN predicts the prognosis more accurately than AO

Among the patients assessed, 91.2% (186/204) and 42.2% (86/204) were classified as being at high risk based on AO and CIN, respectively. However, patients with high risk based on the AO prediction had longer iDFS and OS than those with high risk predicted based on CIN (10‐year iDFS 48.6% versus 11.1%, HR = 2.71, p < 0.01; 10‐year OS 72.6% versus 45.7%, HR = 2.34, p < 0.01; supplementary material, Figure S2).

Validation using independent cohorts

We next chose two independent cohorts of 70 patients with HR+/HER2− breast cancer. Their clinicopathological characteristics are presented in the supplementary material, Table S4. Survival analysis in the two independently validated external samples showed that patients with high CIN at The First Hospital of Jiaxing and Jinhua Municipal Central Hospital had shorter iDFS (HR = 4.51, p = 0.02; HR = 23.82, p < 0.01, respectively; supplementary material, Figure S3A,B) and those at Jinhua Municipal Central Hospital had shorter OS (HR = 7.55, p < 0.01; supplementary material, Figure S3D). Although OS of patients at The First Hospital of Jiaxing did not reach statistical significance, curves remained separate (supplementary material, Figure S3C).

Cross‐validation with other risk evaluation assays

CIN, AO, and MMP were used to assess 80 HR+/HER2− operable breast cancer cases from Zhejiang Cancer Hospital. Frequencies of high CIN and risk determined using MMP and AO were 38.7% (31/80), 35.0% (28/80), and 88.7% (71/80), respectively (supplementary material, Figure S4A). The consistency between CIN testing and MMP was 83.8%, but only 45.0% compared with AO scores (supplementary material, Figure S4B). Eight cases with high CIN were assessed as being low risk based on MMP whereas, in five cases, the inverse was true (supplementary material, Table S5). Finally, in patients defined as high risk based on AO, CIN testing reclassified 42 individuals as having low risk.

Discussion

Using LPWGS, we found that high CIN was significantly associated with more aggressive clinicopathological characteristics, including lymph node metastasis, pathological type, vascular invasion, and PR negativity. The association between high CIN and aggressive clinicopathological features of breast cancer is consistent that reported previously [27]. In this study, although patients with high CIN received stronger adjuvant therapy, such as radiotherapy and chemotherapy with taxanes and anthracyclines (Table 1), their prognoses were worse than those of patients with low CIN. Of 65 patients without lymph node metastasis, 20 had high CIN, and 80.0% (16/20) had recurrence, with a 10‐year iDFS rate of 28.1%. The remaining 45 patients had low CIN with a 10‐year iDFS rate of 79.7% (Figure 2C). Twenty patients with high CIN showed better clinicopathological features; 65.0% (13/20) tumors were of pathological grade II and 85.0% (17/20) had no vascular invasion (supplementary material, Table S3). Therefore, CIN has significant promise as a predictor of recurrence risk and as an aid to identify groups of patients whose traditional clinicopathological features appear to be good but who actually have a poor prognosis.

Amplification of chromosomes 1q, 8q, 12q, and 17q and deletions in chromosomes 17p, 8p, 22q, 11q, 14q, 9p, 6q, and 13q were frequently detected in tumors with recurrence (supplementary material, Figure S1). Chromosomal gains in 18q11 and 19p13 are associated with poor survival in patients with metastatic triple‐negative breast cancer (TNBC), probably caused by a change in loci at the copy number altered region [28]. Furthermore, 1q21.3 amplification is associated with recurrence and poor prognosis in breast cancer [29]; S100 calcium‐binding protein family members, primarily S100A7, S100A8, S100A9, and IRAK1, whose genes are in this chromosomal region, establish a feedback loop to drive tumor growth, which can be disrupted by pacritinib [29]. Amplification of 17q23 has been associated with adverse clinical outcomes in HER2+ breast cancer. Co‐amplification of WIP1 and MIR21 on 17q23 has a synergistic effect on tumor progression by inhibiting oncogene‐induced senescence in breast epithelial cells; inhibitors of products of both genes kill trastuzumab‐resistant HER2+ breast tumor cells [30]. Thus, CNVs play an important role in breast cancer progression and provide a basis for new therapies. Further studies should focus on specific chromosomal variations.

Lee et al [27] used fluorescent immuno‐hybridization to detect CIN in multiple cancer subtypes and found that DFS in patients with high CIN was significantly shortened. Stover et al [28] assayed genomic copy‐number variation (GCNV) in cell‐free DNA using ultra‐LPWGS and reported that GCNV was an independent prognostic factor for metastatic TNBC. Our results showed that high CIN, evaluated using LPWGS, was an independent prognostic factor for HR+/HER2− breast cancer. However, nine fatal tumors were scored as low CIN (supplementary material, Table S6): all cases (100%; 9/9) were invasive ductal carcinoma, 77.8% (7/9) had ≤3 lymph node metastases, 77.8% (7/9) were without vascular invasion, and 77.8% (7/9) were PR positive. Despite good clinicopathology, they still had poor prognoses. Underlying biological behavior and treatment regimen may be a primary cause of mortality, and further research may be needed to explore the underlying biological mechanisms and whether there are specific genetic variants or molecular markers that could guide more personalized and precise treatment strategies.

Genomic prediction models, including Oncotype DX, MMP, EndoPredict, BCI, and PAM50, which are recommended in the NCCN guidelines, are targeted for different patient populations based on clinical studies and play a complementary role. Oncotype DX is used to predict endocrine therapy, chemotherapy efficacy, and prognosis in patients with no lymph node metastasis or those with one to three lymph‐node metastases [31, 32, 33, 34]. MMP is applicable to exemption from chemotherapy in clinical high‐risk patients aged ≥50 years [35, 36, 37]. BCI predicts a 5‐year or longer risk of distant recurrence in HR+ and node‐negative cases [38, 39, 40]. PAM50 can be used for prognostic prediction and molecular classification [41, 42, 43]. However, these multigene assessment methods, which use expression profiles of only 10–70 genes, cannot fully reflect the real proliferative and invasive abilities of tumors. These genes were primarily identified in screens of public mRNA expression databases, and case data used for model construction were retrospective. Moreover, there is high heterogeneity among these models, including types and numbers of genes included and parameter settings. In this study, we used LPWGS to detect CIN, which better reflects changes in whole chromosomes at high efficiency and low cost. Studies have established that MMP is more accurate than AO in predicting recurrence risk [44, 45, 46]. In the present study, AO tended to classify more patients with better prognosis into the high‐risk cohort, with less accuracy than CIN. Correlation analysis revealed a significant negative correlation between MMP and CIN scores (supplementary material, Figure S5), wherein the distribution of populations was highly consistent, indicating that both methods accurately assessed the risk of recurrence. The high consistency of CIN and MMP may be attributable to the fact that they both divide risk at the genome level. As MMP has been recommended in the NCCN guidelines [8], CIN determined based on the LPWGS assay may be a promising method for clinical application. When analyzing 14 patients whose LPWGS and MMP parameters differed (supplementary material, Table S5), we found that those with high CIN but low risk according to MMP had increased vascular invasion, which is associated with metastasis and poor prognosis [47, 48, 49, 50]. CIN was significantly correlated with vascular invasion (p < 0.01); therefore, it may be a better predictor of metastasis. Evaluating combinations of two or more multigenic models will likely improve prediction accuracy.

This study has some limitations. First, as surgical specimens were retrospectively analyzed, sample preservation and quality may have been substandard, affecting our results. Second, the sample size was relatively small, possibly producing information bias in the data owing to the long follow‐up period. Third, given that this was a retrospective study, it was difficult to obtain stratified and unified treatment methods which need to be further confirmed through prospective studies.

In conclusion, to the best of our knowledge, this is the first study to evaluate whole CINs in HR+/HER2− breast cancer using the LPWGS assay. Our results suggest that high CIN (three or more chromosomes with instability) is an independent prognostic factor for HR+/HER2− breast cancer. CIN performed better than clinical characteristic‐based assays and was nearly equivalent to the MMP assay for distinguishing between high and low recurrence risk. Thus, CIN is a promising new prognostic method for HR+/HER2− breast cancer.

Author contributions statement

W‐MC, Y‐YL, JFF and XL conceived and designed the study; Y‐YL, JFF and XL coordinated the study; W‐MC, X‐JW and WYS gave financial support; JFF, XL, YY, LZ, XLL, WYS, X‐JW and W‐MC provided study materials or patients. All authors contributed to data collection and assembly. Y‐YL, ZLQ and W‐MC performed data analysis and interpretation. Y‐YL prepared the first draft of the manuscript. W‐MC revised the paper and helped to write the final draft of the manuscript. All authors approved the final version and agreed to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Supporting information

Figure S1. Patients with recurrence have increased copy number variation

Figure S2. Kaplan–Meier analysis of the risk stratification of 204 patients (high‐ and low‐risk) based on the use of CIN and AO

Figure S3. Survival (Kaplan–Meier) as a function of CIN status in two independent validation cohorts

Figure S4. Cross‐validation with using two risk‐evaluation assays

Figure S5. Correlation analysis of CIN and MMP scores

Table S1. Kaplan–Meier estimates of the 10‐year iDFS and 10‐year OS according to CIN status

Table S2. Clinicopathological characteristics between low CIN and high CIN groups in HR+/HER2− breast cancer patients with negative lymph node metastases

Table S3. Multiple parameter Cox regression analysis

Table S4. Clinicopathological characteristics in two independent cohorts

Table S5. Pathologic characteristics of breast cancers with inconsistent results between CIN and MMP

Table S6. Clinicopathologic characteristics of nine patients with low CIN who died from HR+/HER2− breast cancer

Data availability statement

The raw sequence data reported in this paper have been deposited in the Genome Sequence Archive in the National Genomics Data Center, China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences (GSA‐Human: HRA005613) that are publicly accessible at https://ngdc.cncb.ac.cn/gsa-human.