Risk factors for recurrence in patients with hormone receptor-positive/human epidermal growth factor receptor 2-negative early breast cancer in Japan: a systematic literature review and meta-analysis

Abstract

Background

The clinicopathological factors indicating risk of recurrence are used to guide the choice of perioperative therapy in patients with breast cancer. Although several risk factors for recurrence have been reported in patients with hormone receptor-positive (HR+)/human epidermal growth factor receptor 2-negative (HER2−) early breast cancer in Japan, there has been no systematic review quantifying potential risk factors.

Methods

We performed a systematic literature review and meta-analysis using the MEDLINE, Embase, Cochrane CENTRAL, and Japan Medical Abstract Society databases to identify risk factors for recurrence in HR+/HER2− early breast cancer in Japan. The primary outcome was relapse-free or disease-free survival (RFS/DFS), and the secondary outcomes were overall survival and breast cancer-specific survival (BCSS).

Results

Searches identified 42 eligible publications. Meta-analyses identified lymph node metastasis (hazard ratio: 2.76 [95% confidence interval: 1.97–3.88]), large tumor size (1.67 [1.24–2.23]), high histological grade (1.50 [1.04–2.16]), and high nuclear grade (2.02 [1.61–2.54]) as risk factors for RFS/DFS. Lymph node metastasis (2.43 [1.28–4.63]), large tumor size (1.80 [1.24–2.62]), and high histological grade (2.02 [1.44–2.84]) were also risk factors for overall survival, and high progesterone status was a possible favorable prognostic factor for BCSS (0.20 [0.10–0.42]).

Conclusions

Identified risk factors were consistent with the previous reports, and this study provides quantitative summary of risk factors for HR+/HER2– early breast cancer recurrence in Japan. (PROSPERO Registration ID, CRD42022338391.)

Keywords:

• Risk factors
• breast neoplasms
• systematic review
• meta-analysis
• receptor
• ErbB-2

Introduction

Breast cancer is the most common cancer among women in the US and Japan^1,2. Approximately 92 000 new cases of breast cancer were reported in Japan in 2020², and the average annual incidence of breast cancer is estimated to increase until 2025³. Breast cancer mortality has also been increasing since the 1970s⁴.

Nearly 90% of patients have either localized or regional breast cancer at diagnosis, and are commonly treated by surgery in combination with other modalities such as radiation, chemotherapy, hormone therapy, or targeted immunotherapy with curative intent^5,6. However, breast cancer is a heterogeneous disease with several molecular/clinical subtypes based on factors such as hormone receptor status and human epidermal growth factor receptor 2 (HER2) protein expression, and the optimal treatment is guided by the expression of biomarkers including estrogen receptor, progesterone receptor (PgR), and HER2^7,8. Hormone receptor-positive (HR+), HER2-negative (HER2−) breast cancer is the most prevalent subtype, and is recommended the treatment included hormonal therapies. However, around 30% of HR+/HER2− patients with certain features in terms of lymph node metastasis, tumor size, tumor grade, or other proliferation markers (e.g. Ki-67) are at higher risk of relapsing with incurable metastatic disease^9–13.

Asian patients with breast cancer exhibit distinct clinicopathological characteristics compared to non-Asian patients^14,15. In the monarchE study that examined the effects of adding abemaciclib to adjuvant endocrine therapy in patients with high-risk HR+/HER2− early breast cancer (EBC), patients from Asia were generally younger, more frequently premenopausal and had a higher incidence of ≥4 positive nodes than those from non-Asian countries¹⁶. As noted above, several factors are known to be associated with the risk of cancer recurrence. However, it is unclear whether the risk factors for recurrence based on Western data contribute to the risk of recurrence to a similar extent in Asians.

In this study, we systematically summarized published evidence regarding the risk factors for breast cancer recurrence and conducted meta-analyses to evaluate their impact on recurrence in patients with HR+/HER2− EBC in Japan.

Methods

This systematic literature review was registered in PROSPERO (International Prospective Register of Systematic Reviews; Registration ID: CRD42022338391; website: http://www.crd.york.ac.uk/PROSPERO/) and conducted based on the preferred reporting items for systematic reviews and meta-analyses guidelines.¹⁷

Search strategy

We searched the MEDLINE, Embase, Cochrane Central Register of Controlled Trials (CENTRAL), and the Japan Medical Abstracts Society (JAMAS) databases for relevant publications between January 2010 and June 2022. The search was restricted to studies published in English or Japanese and studies in humans. The detailed search strategies are presented in Supplementary Table 1.

Eligibility criteria

Publications that met the following criteria were included in this systematic review and meta-analyses: studies including early-stage (stage I, II, and III [operable only]) and HR+/HER2− breast cancer patients in Japan; no limitations on interventions or comparators; outcome factors associated with the risk of cancer recurrence or other relevant outcomes included relapse-free survival (RFS) or disease-free survival (DFS), overall survival (OS), and breast cancer-specific survival (BCSS), with RFS or DFS (RFS/DFS) as the outcome of primary interest; observational studies and interventional clinical trials with sample sizes ≥25 patients; and English/Japanese reports published between 1 January 2010 and 22 June 2022. Studies were excluded if <50% of the patient population was either HR + or HER2−. Letters and commentaries were also excluded.

Study selection

Three authors (N. M., T. I., and K. N.) conducted a comprehensive literature search, eliminated duplicate publications, and independently reviewed the titles and abstracts of the studies for possible relevance. The same authors subsequently assessed the full texts of the articles for inclusion. Disagreements among the authors were resolved by discussion among the same three authors.

Risk of bias assessment

The risk of bias (RoB) was assessed using the quality in prognosis studies tool¹⁸, which evaluates the validity and bias in studies of prognostic factors using six domains: study participation, study attrition, prognostic factor measurement, outcome measurement, study confounding, and statistical analysis and reporting. The assessment of RoB in each domain was based on multiple low/moderate/high questions. The overall RoB for each article was concluded based on the results of the six domains as follows: low overall RoB, four to six domains classified as low RoB and the rest as moderate RoB; moderate overall RoB, one domain classified as high RoB and the rest as moderate or low RoB, two domains classified as high RoB and the rest as low RoB, or three domains classified as moderate RoB and low RoB; and high overall RoB, all other cases. Two authors (N. M. and K. N.) conducted the RoB evaluation, and any disagreements between these two authors were resolved by discussion among N. M., T. I., and K. N.

Statistical analysis

Meta-analyses were conducted when more than three reports evaluated the relevant risk factor for each endpoint. RFS and DFS were unified as an endpoint RFS/DFS, since RFS and DFS are similar endpoints and are sometimes used interchangeably. Hazard ratios (HRs) were used as effect size measures. Multivariable-adjusted HRs by covariates for risk factors were used for primary analyses, and univariable HRs for risk factors were used for sensitivity analyses. HRs were combined using a random-effects model with the Sidik–Jonkman estimator¹⁹ and Hartung–Knapp’s method²⁰. Visual inspection of the forest plots was performed, and heterogeneity parameters (I² and τ²) and their 95% confidence intervals (CIs) were evaluated to investigate the possibility of statistical heterogeneity. An I² value ≥75% indicated considerable heterogeneity. There are no clear criteria for τ², and the decision of considerable heterogeneity was therefore based on the estimated average effect. If heterogeneity existed, subgroup analysis of the study characteristics was performed, and the factors used for subgroup analysis were determined by checking forest plots. Influence analyses were performed using Viechtbauer and Cheung’s leave-one-out method²¹. If only one or very few studies were responsible for the heterogeneity, a sensitivity analysis was performed, excluding the study or group of studies. Funnel plots and Egger’s test were used to assess publication bias. If standard errors were not reported, they were calculated from the standard deviation, P-values, or CIs and if CIs were not reported, they were calculated from standard errors or P-values. If the data to be synthesized were missing, the study was excluded. For the overall analyses, sensitivity analyses were conducted including only the studies with 100% HR+/HER2− patients. Statistical analyses were carried out using R version 4.2.0 (R Foundation for Statistical Computing; Vienna, Austria).

Results

Study selection

We conducted a systematic literature search of the MEDLINE, Embase, Cochrane CENTRAL, and JAMAS databases up to June 2022. A total of 2397 reports were identified (MEDLINE: n = 1166, Embase: n = 1102, Cochrane CENTRAL: n = 34, JAMAS: n = 95). After discarding duplicates, the titles and abstracts of 2060 reports were screened, of which 158 were selected for full-text review. Forty-two reports were finally eligible for inclusion (Figure 1). No reports were added by manual searching.

Figure 1. PRISMA diagram illustrating the study selection process. PRISMA: Preferred Reporting Items for Systematic Reviews and Meta-analyses.

Study characteristics

The characteristics of the 42 reports are presented in Table 1. Three reports were randomized controlled trials and 39 were observational studies, including 33 retrospective and six prospective studies. The evaluated outcomes were RFS/DFS (32 reports), OS (9 reports), and BCSS (11 reports). Twenty-five reports included only HR+/HER2− patients. The potential risk factors evaluated in each study are also summarized in Table 1. The RoB was high in 19 reports, moderate in 21, and low in two (Table 1, Supplementary Table 2).

Table 1. Characteristics of the 42 articles identified in the systematic literature review.

Author, year [reference]	Study design	Sample size	Stage	HR+, %	HER2−, %	Outcome	Evaluated risk factors	RoB
Shibuta et al. (2011)^22	Observational: retrospective	4266	I-III	78.0%	83.0%	DFS	subtype	High
Tokiniwa et al. (2012)^23	Observational: retrospective	71	I-III	87.3%	100.0%	RFS	tumor size, Ki-67, topoisomerase II alpha expression	High
Hisamatsu et al. (2012)^24	Observational: prospective	122	I-III	100.0%	100.0%	RFS	pT, LN metastasis, nuclear grade, Ki-67, FOXA1	High
Ishitobi et al. (2012)^25	Observational: retrospective	375	I-III	50.5%	81.7%	OS, RFS, IBTR	LN status, tumor size, histologic grade, ER, IBTR, age, postoperative chemotherapy, postoperative hormone therapy, tumor morphology, margin width, lymphovascular invasion, HER2, therapeutic effect (neoadjuvant)	Moderate
Masuda et al. (2012)^26	Interventional: RCT Phase III	185	IIA	100.0%	100.0%	Response	regimen (anastrozole + goserelin vs tamoxifen + goserelin)	Low
Tokunaga et al. (2012)^27	Observational: retrospective	246	I-III	68.7%	78.6%	RFS	tumor size, nodal involvement, nuclear grade, ER, PgR, HER2, LOH	Moderate
Kikuchi et al. (2013)^28	Observational: retrospective	1987	I-III	100.0%	100.0%	DFS	p53	High
Satoh et al. (2013)^29	Observational: retrospective	471	I-III	100.0%	100.0%	DFS	menopausal status, lymphatic invasion, nuclear grade, number of LN metastases, comedo-type cancer, family history of breast cancer	High
Kurebayashi et al. (2014)^30	Observational: retrospective	261	I-III	100.0%	100.0%	DDFS, BCSS	lymphovascular invasion, PgR, Ki-67, grade	Moderate
Niikura et al. (2014)^31	Observational: retrospective	1331	I-III	80.0%	84.0%	RFS	T stage, N stage, Ki-67, adjuvant endocrine therapy	Moderate
Nishimiya et al. (2014)^32	Observational: retrospective	253	I-III	71.0%	80.0%	RFS, OS	age, preoperative CEA, preoperative TPA, preoperative CA15-3, pT, pN, histology, HR, HER2, Ki-67, method (total/partial), adjuvant hormone therapy, adjuvant chemotherapy	High
Hisamatsu et al. (2015)^33	Observational: prospective	141	I-III	100.0%	100.0%	DFS	pT, LN metastasis, nuclear grade, Ki-67, GATA-3, FOXA1	Moderate
Honma et al. (2015)^34	Observational: retrospective	1111	I-III	86.5%	96.1%	DFS, OS	menopause, T, LN, grade, ER, PgR, HER2, Bcl-2	High
Kawase et al. (2015)^35	Observational: retrospective	42	II-III	100.0%	100.0%	DDFS	tumor size, LN status, tumor grade, pre ER, post ER, pre PgR, post PgR, pre Ki-67, post Ki-67, pre Bcl-2, post Bcl-2, pre FOXA1, post FOXA1, pre MAPT, post MAPT	Moderate
Nishimukai et al. (2015)^36	Observational: prospective	327	I-III	100.0%	100.0%	DDFS	tumor size, LN metastasis, Ki-67, PgR, menopausal status, nuclear grade, ER, adjuvant chemotherapy	High
Ono et al. (2015)^37	Observational: retrospective	369	I-III	100.0%	100.0%	DFS, BCSS	age, menopause status, tumor size, histology, histological grade, nuclear grade, lymphovascular invasion, Ki-67	High
Higuchi et al. (2016)^38	Observational: retrospective	387	I-III	77.5%	80.6%	RFS	age, menopausal status, T size, LN metastasis, nuclear grade, ER, PgR, HER2, subtype, Ki-67, SUVmax, adjuvant chemotherapy	High
Kawai et al. (2016)^39	Observational: retrospective	9850	I-III	100.0%	100.0%	RFS, BCSS	BMI	Moderate
Miyoshi et al. (2016)^40	Observational: retrospective	639	I-III	100.0%	100.0%	OS	ALDH1, age, bilateral breast cancer, tumor size, nodal status, PgR, Ki-67, tumor grade	High
Okuma et al. (2016)^41	Observational: retrospective	122	II-III	56.6%	61.5%	pCR, DFS	age, stage, subtype, histological grade, response (neoadjuvant), number of axillary LNs, gene expression risk score	Moderate
Sato et al. (2016)^42	Observational: retrospective	118	I-III	100.0%	100.0%	BCSS	PgR, histological grade, Ki-67, tumor size, adjuvant hormonal therapy	Moderate
Kurozumi et al. (2017)^43	Observational: retrospective	177	I-III	100.0%	100.0%	RFS, BCSS	Ki-67, PgR, menopausal status, pT, pN, histological grade, adjuvant chemotherapy, adjuvant endocrine therapy	Moderate
Ono et al. (2017)^44	Observational: retrospective	368	I-III	100.0%	100.0%	DFS, BCSS	age, menopause, pT, grade, nuclear grade, histology, PgR	Moderate
Hayashi et al. (2018)^45	Observational: retrospective	719	0-III	68.5%	76.9%	DFS, OS	pCR status (neoadjuvant)	Moderate
Imamura et al. (2018)^46	Observational: retrospective	1076	I-III	79.3%	83.4%	DFS	tumor size, LN metastasis, tumor grade, ER, PgR, chemotherapy, CEA, CA15-3	Moderate
Kurozumi et al. (2018)^47	Observational: prospective	107	I-III	100.0%	100.0%	RFS, BCSS	residual tumor size, residual node status, residual Ki-67 level, PgR, preoperative endocrine prognostic index	High
Yamada et al. (2018)^48	Observational: retrospective	1806	I-III	100.0%	100.0%	RFS	ER/PgR, tumor size, number of involved LNs, histological grade, postoperative radiation therapy, neoadjuvant/adjuvant chemotherapy, adjuvant hormonal therapy	Moderate
Arima et al. (2019)^49	Observational: retrospective	1866	I-III	100.0%	100.0%	DFS	regimen (endocrine vs endocrine + chemo), PgR/Ki-67	Moderate
Fujimoto et al. (2019)^50	Observational: retrospective	717	I-III	81.3%	85.2%	DFS, OS	menopausal status, tumor size, LN metastasis, nuclear grade, ER, HER2, adjuvant chemotherapy, TILs	High
Naoi et al. (2019)^51	Observational: retrospective	257	II-III	100.0%	80.1%	distant recurrence rate, pCR	menopause, tumor size, histological grade, PgR, Ki-67, 95GC score	Moderate
Ohara et al. (2019)^52	Observational: retrospective	124	I-III	100.0%	75.0%	DDFS, pCR	RCB, menstrual status, T, N, histological grade, PgR, HER2, Ki-67, TILs, PAM50	High
Takuwa et al. (2019)^53	Observational: retrospective	418	0-III	100.0%	100.0%	DFS, OS, BCSS	radiotherapy	Moderate
Horimoto et al. (2020)^54	Observational: prospective	117	I-III	100.0%	100.0%	DFS	age, tumor size, LN metastasis, tumor grade, FOXA1	High
Ishiguro et al. (2020)^55	Interventional: RCT Phase II	193	I-III	100.0%	100.0%	pCR	NAC	Moderate
Sugano et al. (2020)^56	Observational: retrospective	336	I-II	100.0%	100.0%	DFS, BCSS	histological grade, nuclear grade, ACTN4, age, menopausal status, area of invasive tumor, histology, lymphovascular invasion, Ki-67	Moderate
Fujihara et al. (2021)^57	Observational: retrospective	120	0-III	85.3%	81.6%	local RFS	lymphatic vessel invasion, surgical margin lateral, surgical margin vertical, age, MRI findings, histology, tumor size, nodal status, stage, ER, PgR, HER2, subtype, EIC, nuclear grade, operation, adjuvant hormonal therapy, chemotherapy	High
Takahashi et al. (2021)^58	Observational: prospective	189	I-II	73.0%	90.5%	RFS	pathological stage, pathological LN, TP53 status, tumor size, histological grade, ER, PgR, HER2, Ki-67, adjuvant chemotherapy, adjuvant endocrine therapy	Moderate
Tanaka et al. (2021)^59	Observational: retrospective	169	I-III	100.0%	100.0%	DDFS, DFS	Ki-67, podoplanin, age, tumor size, lymphatic invasion, histological grade, Ki-67, PgR, TILs, adjuvant chemotherapy	Moderate
Toi et al. (2021)^60	Interventional: RCT Phase III	1930	I-III	100.0%	100.0%	invasive DFS, OS	regimen (endocrine vs endocrine + S-1)	Low
Kurozumi et al. (2022)^61	Observational: retrospective	142	I-III	100.0%	100.0%	BCSS	LAT1 expression, Ki-67, PgR, tumor size, nodal status, histological grade	High
Sueta et al. (2022)^62	Observational: retrospective	95	I-III	66.3%	100.0%	pCR, RFS, BCSS	tumor size, LN metastasis, histological grade, Ki-67, tumor subtype, response (NAC), BRCAness	High
Yamamoto et al. 2022^63	Observational: retrospective	57	I-III	100.0%	100.0%	RFS, OS	clinical stage, pathological stage, biological feature prior to NAC, biological feature following NAC	High

Abbreviations. HR: hormone receptor; RoB: risk of bias; RCT: randomized controlled trial; DFS: disease-free survival; RFS: relapse-free survival; OS: overall survival; IBTR: ipsilateral breast tumor recurrence; DDFS: distant disease-free survival; BCSS: breast cancer-specific survival; pCR: pathological complete response; pT: pathological tumor; LN: lymph node; FOXA1: Forkhead box protein A1; ER: estrogen receptor; HER2: human epidermal growth factor receptor 2; PgR: progesterone receptor; LOH: loss of heterozygosity; CEA: carcinoembryonic antigen; TPA: tissue polypeptide antigen; CA15-3: carbohydrate antigen 15-3; pN: pathological node; GATA-3: GATA-binding protein 3; bcl-2: B cell lymphoma 2; MAPT: microtubule-associated protein tau; SUVmax: maximum standardized uptake value; BMI: body mass index; ALDH1: aldehyde dehydrogenase 1; TILs: tumor-infiltrating lymphocytes; 95GC: 95‑gene classifier; RCB: residual cancer burden; PAM50: prediction analysis of microarray 50; NAC: neoadjuvant chemotherapy; ACTN4: actinin-4; MRI: magnetic resonance imaging; EIC: extensive intraductal component; LAT1: L-type amino acid transporter 1; BRCA: breast cancer susceptibility.

Meta-analyses

Risk factors for RFS/DFS

Nine factors were eligible for meta-analysis in relation to RFS/DFS (Table 2, Figure 2): lymph node metastasis, tumor size, Ki-67, adjuvant chemotherapy, PgR status, histological grade, nuclear grade, menopausal status, and age were reported in relation to RFS/DFS in more than three reports using either multivariable or univariable results. Lymph node metastasis, tumor size, histological grade, and nuclear grade were significant risk factors in the multivariable analyses (HR: 2.76, 95% CI: [1.97–3.88], 1.67 [1.24–2.23], 1.50 [1.04–2.16], 2.02 [1.61–2.54], respectively), and the results of the univariable analyses were similar. Ki-67 tended to be a high-risk factor and was a significant risk factor in univariable analysis (2.65 [1.76–4.00]), but not in multivariable analysis (1.50 [0.94–2.41]). Adjuvant chemotherapy was not identified as a risk factor in either multivariable or univariable analyses (0.81 [0.47–1.38], 1.19 [0.54–2.61], respectively). High PgR status was a significant favorable prognostic factor in univariable analysis (0.51 [0.27–0.96]) and tended to be associated with an improved prognosis, but without statistical significance, in multivariable analysis (0.60 [0.29–1.22]). Menopausal status and age were not risk factors in univariable analyses and were not examined in multivariable analyses because there were only three and two reports, respectively.

Figure 2. Forest plots for relationships between various factors and RFS/DFS in multivariable analyses. HRs represented by gray boxes; size of boxes correlated with weight; length of lines represents 95% CIs; diamonds indicate pooled HRs. (A) Lymph node metastasis; yes vs no (reference). (B) Tumor size; large vs small (reference). (C) Ki-67; high vs low (reference). (D) Adjuvant chemotherapy; yes vs no (reference). (E) PgR status; high vs low (reference). (F) Histological grade; high vs low (reference). (G) Nuclear grade; high vs low (reference). PgR: progesterone receptor; RFS/DFS; relapse-free survival or disease-free survival; HR: hazard ratio; CI: confidence interval; HK: Hartung-Knapp.

Table 2. Results of meta-analysis by outcomes and factors.

Factor	RFS/DFS		OS		BCSS
	Multivariable	Univariable	Multivariable	Univariable	Multivariable	Univariable
Lymph node metastasis yes vs no (reference)	K = 15, 2.76 [1.97; 3.88]	K = 10, 3.15 [1.99; 4.99]	K = 6, 2.43 [1.28; 4.63]	K = 4, 3.89 [2.76; 5.47]	K = 2	K = 3
Tumor size large vs small (reference)	K = 12, 1.67 [1.24; 2.23]	K = 14, 2.27 [1.60; 3.21]	K = 6, 1.80 [1.24; 2.62]	K = 4, 2.48 [1.34; 4.57]	K = 3	K = 6, 2.39 [1.17; 4.87]
Ki-67 high vs low (reference)	K = 7, 1.50 [0.94; 2.41]	K = 10, 2.65 [1.76; 4.00]			K = 5, 2.12 [0.53; 8.49]	K = 6, 3.76 [1.79; 7.93]
Adjuvant chemotherapy yes vs no (reference)	K = 6, 0.81 [0.47; 1.38]	K = 6, 1.19 [0.54; 2.61]
Adjuvant endocrine therapy yes vs no (reference)	K = 3	K = 3
PgR status high vs low (reference)	K = 4, 0.60 [0.29; 1.22]	K = 9, 0.51 [0.27; 0.96]			K = 4, 0.20 [0.10; 0.42]	K = 4, 0.14 [0.03; 0.73]
Histological grade high vs low (reference)	K = 9, 1.50 [1.04; 2.16]	K = 9, 1.90 [1.49; 2.41]	K = 5, 2.02 [1.44; 2.84]	K = 3	K = 5, 2.28 [0.60; 8.69]	K = 5, 5.63 [2.24; 14.13]
Nuclear grade high vs low (reference)	K = 5, 2.02 [1.61; 2.54]	K = 6, 2.06 [1.31; 3.24]			K = 2	K = 3
Menopausal status post vs pre (reference)	K = 3	K = 7, 1.06 [0.84; 1.35]	K = 3	K = 2	K = 1	K = 3
Age high vs low (reference)	K = 2	K = 5, 0.95 [0.78; 1.14]				K = 3
Histology IDC vs other (reference)		K = 3				K = 3
Lymphovascular invasion yes vs no (reference)		K = 3			K = 1	K = 2

Quantitative synthesis results expressed as hazard ratio [95% confidence interval]. RFS/DFS: relapse-free survival or disease-free survival; OS: overall survival; BCSS: breast cancer-specific survival; K: number of studies; PgR: progesterone receptor; IDC: invasive ductal carcinoma.

In the primary analyses using multivariable results, only lymph node metastasis showed high heterogeneity (I² = 79%) (Figure 2A) and therefore we underwent subgroup analysis to explore the cause of the heterogeneity (Supplementary Figure 1). We could only perform a subgroup analysis for perioperative chemotherapy (Yes/No), because other factors were not consistently reported. The Yes group included 13 reports, with a pooled HR of 2.83 [2.08–3.85], and the No group only included two reports, resulting in a very wide CI for the pooled HR, which precluded further evaluation. In the influence analyses for the primary analyses using multivariable results, all leave-one-out results were included in the original pooled effect size of 95% CIs, and no reports caused heterogeneity, and sensitivity analyses based on the influence analysis were therefore not performed (Supplementary Figure 2). The meta-analysis results for publishing bias showed no significant publication biases (Supplementary Figure 3). The sensitivity analyses using only studies with 100% HR+/HER2− patients were consistent with the primary analyses (Supplementary Table 3).

Risk factors for OS

Three factors were eligible for meta-analyses in relation to OS (Table 2, Figure 3): lymph node metastasis, tumor size, and histological grade were reported in relation to OS in more than three reports with either multivariable or univariable results. Lymph node metastasis and tumor size were significant risk factors in both multivariable (2.43 [1.28–4.63], 1.80 [1.24–2.62], respectively) and univariable analyses (3.89 [2.76–5.47], 2.48 [1.34–4.57], respectively). Histological grade was a significant risk factor in multivariable analysis (2.02 [1.44–2.84]), but only three reports included univariable results.

Figure 3. Forest plots for relationships between various factors and OS in multivariable analyses. HRs represented by gray boxes; size of boxes correlated with weight; length of lines represents 95% CIs; diamonds indicate pooled HRs. (A) Lymph node metastasis; yes vs no (reference). (B) Tumor size; large vs small (reference). (C) Histological grade; high vs low (reference). OS: overall survival; HR: hazard ratio; CI: confidence interval; HK: Hartung-Knapp.

Risk factors for BCSS

Four factors were eligible for meta-analyses in relation to BCSS (Table 2, Figure 4): tumor size, Ki-67, PgR status, and histological grade were reported in relation to BCSS in more than three reports with either multivariable or univariable results. Ki-67 and histological grade were strong risk factors in univariable analyses (3.76 [1.79–7.93], 5.63 [2.24–14.13], respectively) and tended to be high-risk factors in multivariable analyses, but the results were not significant (2.12 [0.53–8.49], 2.28 [0.60–8.69], respectively). High PgR status was a significant favorable prognostic factor in both multivariable and univariable analyses (0.20 [0.10–0.42], 0.14 [0.03–0.73], respectively). Tumor size was a significant risk factor in univariable analysis (2.39 [1.17–4.87]), but only three reports included multivariable results.

Figure 4. Forest plots for relationships between various factors and BCSS in multivariable analyses. HRs represented by gray boxes; sizes of boxes correlated with weight; lengths of lines represent 95% CIs; diamonds indicate pooled HRs. (A) Ki-67; high vs low (reference). (B) PgR status; high vs low (reference). (C) Histological grade; high vs low (reference). PgR: progesterone receptor; BCSS: breast cancer-specific survival; HR: hazard ratio; CI: confidence interval; HK: Hartung-Knapp.

Discussion

We conducted a systematic literature review and meta-analyses of recurrence risk factors in patients with HR+/HER2− EBC in Japan. The identified risk factors were consistent with those used in the molecular classification of HR+/HER2− EBC into luminal A/B subtypes, and with the risk factors indicated in the existing breast cancer guidelines.⁶ The results of this study provide a quantitative summary, which can be used to assess the relative importance of each risk factor in patients with breast cancer in Japan. Lymph node metastasis, tumor size, and histological grade were consistently identified as risk factors for RFS/DFS and OS. Regarding BCSS, histological grade was a significant risk factor in univariable but not in multivariable analysis, but showed a consistent trend toward high risk. Lymph node metastasis is a well-known risk factor⁶, and accordingly had the highest HR among these three factors in the present study.

Ki-67 was a significant risk factor for RFS/DFS and BCSS in univariable analyses but not in multivariable analyses. This apparent discrepancy may be attributable to different methods of evaluation among studies, given that a consensus Ki-67 threshold for predicting risk of recurrence is not well established and testing and assessment methods are not standardized⁶⁴; indeed, most identified in this review had thresholds that varied from 10%–30%. Another potential explanation is that Ki-67 is a marker of cell proliferation, which is a fundamental biological process, and it may therefore be confounded by other relevant risk factors.

Abemaciclib has recently been approved for postoperative adjuvant therapy in high-risk patients based on the monarchE trial⁶⁵, with high risk defined as having at least four positive pathologic axillary lymph nodes or up to positive three axillary lymph nodes plus tumor size ≥5 cm, histologic grade 3 disease, or centrally assessed Ki-67 ≥ 20%. This review also shows that it is reasonable to define high risk according to lymph node metastasis, tumor size, and histological grade.

Overall, PgR status was suggested to be a favorable prognostic factor; it was a significant favorable prognostic factor for RFS/DFS in univariable analyses but not in multivariate analyses, and a significant favorable prognostic factor for BCSS in both analyses. The Japanese Breast Cancer Society Clinical Practice Guidelines and the American Society of Clinical Oncology/College of American Pathologists guidelines also recommend PgR testing to evaluate the risk of future breast cancer, and the results were consistent with the global consensus of clinicians^6,66.

Adjuvant chemotherapy was not associated with a low recurrence risk. The use of adjuvant chemotherapy is at physician’s discretion⁶, and perioperative chemotherapy regimens have evolved with time. Study periods were 2000–2001 in Kurozumi (2017)⁴³, 1992–2010 in Yamada (2018)⁴⁸, 1995–1999 in Nishimiya (2014)³², 2008–2014 in Higuchi (2016)³⁸, 2004–2014 in Imamura (2018)⁴⁶, and 2008–2017 in Fujimoto (2019)⁵⁰, which indicates variability across studies. Due to the variability in perioperative chemotherapy regimens over time, the limited number of studies, and variations in the factors considered in multivariable analyses across different studies, no definitive difference could be obtained. This factor is, in addition, a posteriori in terms of recurrence risk assessment, and the results of each report should be interpreted with caution.

Breast cancer in Asians is characterized by early tumor onset², i.e. young age and pre-menopausal status, but menopausal status and age were only reported in limited numbers of studies in this review, and were not identified as significant risk factors for RFS/DFS. Although young-onset breast cancer has been reported to be more aggressive and have a poorer prognosis, the lack of a clear definition of risk classification by age⁶⁷ and relatively older age thresholds (50 or 57 years) in the studies included in this review preclude an adequate discussion of these factors.

Domestic guidelines list lymphovascular invasion as a risk factor for breast cancer recurrence⁶; however, few reports in this review assessed it as a potential risk factor, and it was therefore not included in the current meta-analyses.

This systematic review and meta-analysis had some limitations. Some of the risk factors may be missed since most of the original studies are retrospective observational studies. For example, variables like the age of onset and BMI are known to differ in Japanese from Western people like Americans, so can affect outcomes differently in Japanese population. However, most of the 42 studies included in this systematic review did not report the age at diagnosis and BMI, thus it was not possible to perform analysis on these factors. The population characteristics differed between studies, the list of assessed risk factors varied among studies, and the adjustments in multivariable analyses also differed. Meta-analyses using the results of univariable analyses were therefore performed as sensitivity analyses, which confirmed and supplemented the primary analyses. In addition, this study included reports in which >50% of the patients were HR+/HER2−, and 17 reports used in the primary analysis included some non-HR+/HER2− patients. However, we checked the robustness of the results by performing a sensitivity analysis using only the 25 reports in which all patients were HR+/HER2−, and these results were consistent with the primary analyses. Although ER and HER2 were also assessed as risk factors in some of the mixed population studies, these were not relevant to the current study and were therefore omitted from the analyses. Early and late recurrence could not be compared because only two reports included this information. Adjuvant endocrine therapy and radiation therapy also could not be evaluated either because the number of reports included was three and one, respectively. The cause of the high heterogeneity in lymph node metastasis was unclear because we could not perform detailed subgroup analyses due to a lack of information. Different thresholds were used for some risk factors, or groupings were made in each report; for example, some reports treated risk factors as positive vs. negative, while others classified them as grades 0–4. In this study however, we classified and analyzed all risk factors as single binary variables, e.g. high vs low. In addition, most studies included in this systematic review were observational studies, and most ROB scores were moderate or high.

Conclusion

In conclusion, the risk factors identified in the systematic review and meta-analyses were consistent with body of knowledge, including the Japanese Breast Cancer Society Clinical Practice Guidelines, but the current results provide novel information quantifying each risk factor in Japanese patients with HR+/HER2− EBC. Notably, the choice of postoperative adjuvant treatment is based on an assessment of recurrence risk. Accurate assessment of recurrence risk followed by decision of treatment strategies based on it are necessary to improve the clinical outcome of the patients. The results of this study will support these aims and help in the development of better treatment strategies.

Transparency

Declaration of funding

This study was funded by Eli Lilly Japan and outsourced from Eli Lilly Japan to the Biostatistics Unit of Keio University Hospital.

Author contributions

Y. T., Z. C., and T. K. contributed to the study conceptualization and design. N. M., T. I., and K. N. performed a literature search and contributed to data acquisition. K. N. performed statistical analyses. All authors contributed to the interpretation of data. N. M. contributed to manuscript drafting. All authors critically reviewed and revised the draft manuscript and approved the final version for submission.

Acknowledgements

None.

Declaration of financial/other relationships

H. S. has received speaker’s honorarium and medical writing support from Daiichi Sankyo as well as research funding from Eisai. Y. T, Z. C., and T.K. are employees of Eli Lilly Japan, and minor stockholders of Eli Lilly and Company. J. T.’s institution received study grants from Eli Lilly, Eisai, Daiichi Sankyo, Kyowa Kirin, Taiho, MSD, and Nihon Kayaku, Ono, the West Japan Oncology Group and he was paid personal fee for lecture by the Eli Lilly, Eisai, Daiichi Sankyo, Kyowa Kirin, Taiho, MSD, and Nihon Kayaku Japan.

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Data availability statement

The datasets used and/or analyzed during this study are available from the corresponding author upon reasonable request.