Low Prevalence of HER2-Positive Breast Carcinomas among Screening Detected Breast Cancers

Conflicting results have been reported regarding the prevalence of screen-detected human epidermal growth factor receptor 2 (HER2)-positive breast carcinomas and non-screen detected HER2-positive breast carcinomas. To address this issue, we evaluated the prevalence of HER2-positive breast carcinomas in two independent regional screening programs in Spain. The clinicopathologic and immunohistochemical characteristics of 479 (306 and 173) screen-detected breast carcinomas and 819 (479 and 340) non-screen-detected breast carcinomas diagnosed in women between 50 and 69-year-olds were compared. The prevalence of HER2-positive breast carcinomas was 8.8% and 6.4% in the two series of screen-detected tumors, compared with 16.4% and 13% in non-screen-detected carcinomas. These differences were statistically significant. This lower prevalence of HER2-positive in-screen-detected breast carcinomas was observed in both hormone receptor positive (luminal HER2) and hormone-receptor-negative (HER2 enriched) tumors. In addition, a lower prevalence of triple-negative and a higher prevalence of luminal-A breast carcinomas was observed in screen-detected tumors. Moreover, a literature review pointed out important differences in subrogate molecular types in screen-detected breast carcinomas among reported series, mainly due to study design, technical issues and racial differences.


Introduction
Since the introduction of screening programs for breast cancer (BC) detection, several studies have demonstrated that screen detected breast cancers (SDBC) are diagnosed in early stages, since they are smaller and have less lymph node metastases than those detected outside of screening [1][2][3][4][5]. In addition, SDBC have intrinsic good prognosis features, given that low-grade carcinomas and favorable types such as tubular carcinoma are more frequently diagnosed [3,6,7]. Partially due to these features, screening is associated with a relative mortality reduction of 20% [8].
During the last two decades, the introduction of the molecular classification of BC has prompted researchers to analyze the differences of the intrinsic molecular subtypes between SDBC and nonscreen-detected breast cancers (NSDBC). Most studies have reported higher percentage of estrogen receptor (ER)-positive/progesterone receptor (PR)-positive tumors among SDBC leading to a higher prevalence of luminal A and a lower prevalence of triple negative (TN) breast carcinomas in this group of tumors [4,[9][10][11]. However, there are conflicting results regarding the frequency of detected human epidermal growth factor receptor 2 (HER2)-positive BC among SDBC, since most series have not found differences when compared with NSDBC [7,[10][11][12][13][14][15], while other studies have reported a lower prevalence of HER2-positive tumors within SDBC group [3,4,13,16]. These discrepancies can be attributed to different factors. Reported series are heterogeneous regarding to variables such as case and controls numbers, year and patient age at diagnosis, screening-program inclusion criteria or subrogate intrinsic molecular classification used.
Knowing the true frequency of HER2-positive BC among SDBC has an important clinical and epidemiological implication. To identify the correct proportion of HER2-positive tumors after benchmarking to standard references is an important measure to assurance quality in HER2 testing. As the detection mode seems to impact on HER2 status, the proportion of HER2-positive BC may differ in each institution depending on the relative number of tumors coming from screening programs. In addition, as currently available anti-HER2 drugs have improved HER2-positive tumors prognosis and their diagnosis is mostly outside the screening programs, there is probably an impact on the screening prognosis.
In this study, we establish the prevalence of the different subrogate intrinsic molecular subtypes in two independent hospital-based series of SDBC from two different regional screening programs in Spain, one from Andalusia (Programa de Detección Precoz del Cáncer de Mama (PDP)) and other from Madrid (Detección Precoz del Cáncer de Mama (DEPRECAM)). In both series, pathologic and immunohistochemical features were centrally reviewed and the control groups consisted of tumors from patients in the same age range, in order to avoid bias related to pathologic characteristics and/or age. Thereby, it is well established that the biologic characteristics of BC differ depending on the menopausal status of the patients. After an extensive literature review, our results in both series support the latest data indicating a low prevalence of HER2-positive among SDBC.

Clinicopathologic Differences of NSDBC According to Age: Selection of the Control Group
First, in the hospital-based series from Andalusia, we analyzed possible differences among NSDBC according to age. We compared three age groups: ≤49 years, 50-69 years and ≥70 years. We observed statistically significant differences in all variables analyzed, except in PR expression (Table 1). Accordingly, we selected only NSDBC from women with 50-69 years for further comparisons with SDBC.

Clinicopathologic Differences between SDBC and NSDBC Groups
We observed statistically significant differences in all variables analyzed when compared SDBC with NSDBC (Table 2). Thus, near two thirds of SDBC were pT1 and N0, in contrast to only 50% in NSDBC. Regarding histological type, fewer infiltrating lobular carcinomas and poor-prognosis subtypes were found among SDBC. In addition, a two-fold increase in histological grade 1 was observed in SDBC with respect to NSDBC. In addition, 14.4% of SDBC showed lymphovascular invasion (LVI) in contrast with 33% of NSDBC.

Immunohistochemical and Subrogate Molecular Classification Differences between SDBC and NSDBC Groups
SDBC showed an increased proportion of ER-positive and PR-positive cases. The most important difference was observed regarding HER2 status. Whereas 16.4% of NSDBC were HER2positive by immunohistochemistry (IHC) and/or fluorescence in situ hybridization (FISH), only 8.8% of SDBC were HER2-positive (p = 0003) ( Table 3). According to biomarker results, we observed differences in the distribution of subrogate molecular types. The greater difference was observed in the frequency of HER2-positive/hormone receptor (HR)-negative BC, which dropped from 10% in NSDBC to 3.9% in SDBC. In addition, an increase in the frequency of luminal A and a decrease in the frequency of HER2-positive/HR-positive and TN BC was observed among SDBC. Regarding TN BC, the decrease was more prominent among those tumors that expressed basal markers. The proportion of luminal B BC was similar among SDBC and NSDBC (Table 4).

Prognosis
With a median follow-up of 120 months, we observed a total of 29 (9.4%) relapses and 16 deaths (5.22%) among SDBC (PDP) and 85 relapses (18%) and 75 deaths (15.9%) among NSDBC. However, deaths attributable to BC were 9 (2.9%) among SDBC and 48 (10.2%) among NSDBC. Kaplan-Meir plot is provided in Figure 1. Table 5; Table 6 show the results of the univariate and multivariate analysis BC specific death. Analysis on time to relapse is provided in Table S1. In the univariate analysis, SDBC had a better prognosis than NSDBC. However, in the multivariate analysis, screening did not retain statistical significance as a prognostic factor.

Literature Review
This review included 13 series (Tables 7 and 8) in which immunohistochemical data and/or the subrogate molecular classification were evaluated.
The series reported by Joensuu et al. [14] and Lehtimäki et al. [15] included a similar cohort of patients, although the method of evaluation of HR differed between both studies. Series were heterogeneous with respect to different variables, as follows:   [7,11,19], one series from United States of America [17] and one series from Oceania [20]. No series from Africa or South America were found.

Age
Seven series [7,[10][11][12]17,19,20] did not define an age-limit to include cases, whereas the remaining series limited the study to those included in the age range of the screening programs, which was variable among countries.

Type of Cohort
Seven cohorts were composed of patients included in hospital-based databases, whereas the remaining series obtained the cases from population-based registries [3,[13][14][15]18,20].

Centralized Review of Clinicopathologic Data
Only in one series, conventional clinicopathologic data were centrally reviewed [18]. In the remaining series, data were collected from medical records.

Molecular Classification Results
The results are summarized in Table 9. In control groups, the frequency of luminal (ER-positive and/or PR-positive, HER2-negative) tumors ranged from 54.3% to 72%; the frequency of luminal-HER2 (ER-positive and/or PR-positive, HER2-positive) tumors from 6% to 20.2%; the frequency of HER2-positive or HER2-enriched (ER-negative, PR-negative and HER2-positive) tumors from 3.5% to 13.9%; and the frequency of TN (ER/PR-negative and HER2-negative) tumors from 7.5% to 22.6%.

Discussion
In this study, we confirmed that the frequency of subrogate molecular subtypes of BC differed between SDBC and NSDBC. Specially, we demonstrated that HER2-positive BC were underrepresented in SDBC, especially those tumors that were HR-negative. In addition-and according to most of the previously published series-we observed that SDBC showed statistically significant differences in most conventional clinicopathologic features analyzed when compared with NSDBC. Thus, SDBC presented at earlier stage, since they were smaller and had less axillary lymph node metastases. Regarding histological types, less infiltrating lobular carcinomas and poorprognosis subtypes were found among SDBC. Moreover, the proportion of G1 carcinomas was higher in SDBC.
Regarding patient selection, one major strength of our study is that we used a control group composed by tumors affecting women in the same age range that those attending screening programs, since we demonstrated that clinicopathologic features differed among women with different ages. Thus, in both women younger than 49 years and older than 69 years, a lower proportion of stage I BC was observed due to larger tumors and more axillary involvement. In women younger than 49 years, a higher frequency of TN BC and a lower frequency of luminal A BC was observed. On the other hand, the group of women older than 69 years showed a lower proportion of HER2-positive BC.
Only three out of thirteen series reviewed used our same age range (50-69 years) [4,14,15] as a criterion for the control group and only in one of them the authors observed differences in HER2 expression regarding SDBC [4]. The reported lower prevalence of HER2-positive tumors within SDBC group in comparison with NSDBC group is probably related to a lead time bias due to early detection. Whereas the mean tumor size in HER2-positive SDBC in PDP series was 1,4 cm, a statistically significant higher mean size (2,65 cm) was observed in the non-screen detected HER2positive BC group (p < 0012).
The definition of case was also variable among the series. Thus, whereas we considered cases as only those detected during women's participation in the population screening programs, other studies included tumors detected by mammography, without clinical symptoms, in patients from and outside screening programs [10,11]. Hence, a proportion of control tumors in our two cohorts could have been detected by mammography, without clinical symptoms. Whether or not these tumors have biologic characteristics more similar to SDBC or NSDBC remains to be established. In this sense, Iwamoto et al. [21] reported that the proportion of HER2-positive BC was 17% among "self-detected", 15% among "screening-detected (asymptomatic)" and 15% among ''screeningdetected (symptomatic)''.
Technical differences in the determination and evaluation of biomarkers were also present among series. Thus, different antibodies and thresholds criteria were used for the evaluation of ER, PR and HER2. Regarding ER and PR, some studied considered positive those tumors with at least 1% of positive cells, whereas others used a 5% [17], 10% [4] or 20% [3] threshold. For the evaluation of HER2, all but one [18] of the studies used the internationally accepted evaluation criteria. Probably these technical differences may partially explain the wide range of positivity of different biomarkers among series in both SDBC and NSDBC, as presented in Table 8. Regarding HER2, the reviewed series reported 12% to 30.1% of positivity in the control group and 8% to 25% of positivity in SDBC. Since we observed some differences in the incidence of HER2-positive BC between PDP and DEPRECAM series in our study, we carried out a concordance analysis in a small group of tumors with an overall kappa value for immunohistochemistry of 0,86 and 1 for in situ hybridization, corresponding to an almost perfect agreement. This concordance analysis suggested that the differences between both series were not due to technical issues (Table S2). Our results in the NSDBC group are in accordance with national data in Spain, where a central data base [22] including 135,173 cases, reported a 16.4% of HER2 positivity. In addition, in a database from UK and Ireland, the frequency of HER2 positivity among 1.537 tumors was 9% in SDBC and 13.3% in symptomatic patients [23].
Regarding TN BC, most previous series demonstrated a reduction of this molecular type among SDBC [3,4,10,11,13,17,19,20]. However, there were marked differences in the relative frequency of this molecular type in SDBC, ranging from 1.8% to 18% in the different series. In addition to the previously mentioned factors, race can partially explain these differences, since a higher percentage was observed in that series including a large proportion of Black-American women, a population group in which TN BC is more frequent [24]. Only two previous series analyzed differences in the expression of basal markers among TN SDBC [10,13]. Our results, as the study of Crispo et al. [10], suggested a lower percentage of basal-like TN BC in this group. In contrast, Dawson et al. [13] did not observed this difference. Further studies are needed in order to establish whether or not different subtypes of TN BC are differentially represented between SDBC and NSDBC.
Most previous series have classified luminal tumors according to HER2 expression, and only three series have also classified luminal tumors according to Ki67 expression [3,17,19]. In our study, and the one from Brewster et al. [17], we defined luminal A carcinomas as those ER and/or PR positive BC that had Ki67 expression in ≤15% of neoplastic cells. Falck et al. [3] defined luminal tumors according to both Ki67 (20% threshold) and PR (20% threshold) expression. Despite differences in definition criteria, the three series observed a similar frequency of luminal B BC in SDBC and NSDBC, but a significant higher percentage of luminal A BC in the SDBC group. These data, together with the lower prevalence of HER2-positive and TN BC indicated that screening increased the proportion of low proliferative BC. Accordingly, we observed differences in Ki67 proliferative index among subrogate molecular groups (Table S3).
Some studies have reported that screening is a favorable prognostic factor in BC. Other authors, however, have suggested that prognosis in this group of patients is related with associated clinicopathologic good prognostic factors. In this sense, in our series, patients with SDBC had better prognosis that patients with NSDBC in the univariate analysis. However, screening status did not show prognostic significance in the multivariate analysis (Table 6). On the other hand, tumor size, stage and. Higher tumor size and more advanced stage were associated with worse outcomes. These variables were also important when evaluating HER2-positive BC only. In this subgroup of tumors, the method of detection was not associated with the prognosis, whereas tumor size and stage were (Table S4).
In spite of HER2 status did not was associated with prognosis, the molecular subtype was associated. luminal A molecular subtype had a better disease-free survival and BC-specific survival, whereas basal BC had a worse disease-free survival and BC-specific survival. These results highlight the importance of using molecular classification on BC, as it can provide key prognosis information.
Our review pointed out important differences in subrogate molecular types in SDBC among reported series, partially attributable to differences in study designs, technical issues and racial differences. To know the true incidence of intrinsic molecular subtypes of SDBC is of great importance to increase epidemiological knowledge and to stablish benchmarks for quality controls, which could ultimately impact on patients' treatment.

Immunohistochemistry and Tumor Classification
IHC in Cohort 1 was analyzed on TMA. TMA construction was performed as previously reported [25] including two cores for each tumor. Analysis of TMA IHC and fluorescent in situ hybridization was performed by two pathologists (MAL-G and JP) and the final results were obtained by consensus.
HER2 in Cohort 2 was re-evaluated on core biopsies by two pathologists (BP-M and JP) and the final results were obtained by consensus. On HER2-positive cases, ER and PR were also evaluated. Details of the antibody clones, probes, suppliers, dilutions and scoring criteria following international guidelines [26] used are provided in Tables S5 and S6. Since we used different reagents in PDP and DEPRECAM series to analyze HER2 status, we performed a concordance analysis between the two immunohistochemical methods using 2 TMA sections, including a total of 58 cases, displaying different levels of HER2 expression that were double-blinded evaluated.
For molecular classification, BC were grouped according to immunohistochemical criteria. As a subrogate definition of the intrinsic subtypes of BC the following criteria were applied [27]: tumors showing ER or PR expression and no HER2 expression were regarded as luminal A or B depending on their Ki67 index (≤15 or >15, respectively); tumors with ER or PR positivity and HER2 overexpression and/or amplification were classified as ER-positive/PR-positive/HER2-positive (luminal HER2); tumors showing overexpression and/or amplification of HER2 and absence of expression of ER and PR were regarded as HER2-positive and TN tumors with CK5/6, CK14, CK17 and/or EGFR expression were ascribed to the basal phenotype, whereas all other TN breast carcinomas were termed TN not otherwise specified (NOS).

Statistical Analysis
The χ2 test was used to evaluate differences among clinicopathologic characteristics (summarized with percentages). Differences in median size (in cm) among the different groups of study were analyzed with the Kruskal-Wallis and Mann-Whitney U tests. Breast cancer-specific survival was defined as the time from surgery to the time of death from BC, with deaths from other causes being censored; whereas in the time to relapse analysis, the endpoint was BC recurrence, either local or distant. Survival curves were estimated using the Kaplan-Meier method the differences in survival were evaluated using the log-rank test. Cox's proportional hazards modeling of parameters potentially related to survival was conducted to calculate hazard ratios in both univariate and multivariate analyses.
Kappa values were used to assess HER2 status concordance between the results obtained by the two immunohistochemistry methods applied in this study.
The present study was performed in accordance with the standard ethical procedures dictated by Spanish law (Ley de Investigación Orgánica Biomédica, 14 July 2007). The procedure was approved by the local ethics committee (Code 0292-N-15, approval date 10 April 2015, by Comité de Ética de la Investigación-HHUU Virgen Macarena-Virgen del Rocío). Written informed consent was obtained from all the patients and all the clinical analyses were conducted in accordance with the principles of the Helsinki Declaration.

Literature Review
A literature review was conducted using the MEDLINE and PubMed. The following keywords were used to perform flexible searches within these databases: "screening" AND "breast cancer", "pathology", "immunohistochemistry", "HER2", "estrogen receptor", "progesterone receptor", "Ki67", "molecular classification". Only papers published in English were included.

Conclusions
In our series, HER2-positive BC are underrepresented in SDBC compared to NSDBC, particularly HR-negative tumors. In addition, within SDBC there is a higher prevalence of luminal A BC and a lower prevalence of TN BC. The literature review pointed out important differences in biomarker expression results as well as in subrogate molecular types, partially attributable to differences in study designs, technical issues and racial differences.

Supplementary Materials:
The following are available online at www.mdpi.com/2072-6694/12/6/1578/s1, Table  S1: Disease-free survival in the age range 50-69-year-old, Table S2: Concordance analysis between HER2 status in both hospitals, Table S3: Ki67 differences between the subrogate phenotypes in the series (NSDBC and SDBC) (PDP), Table S4: Cox model for histopathologic features for disease-free survival and cancer-specific free survival within the HER2-positive breast cancer tumors, Table S5: Antibodies, Table S6: HER2 fluorescence in situ hybridization (FISH) used in PDP and DEPRECAM. Funding: Supported by grants from the Instituto de Salud Carlos III (ISCIII) (PI 19/01331) and CIBERONC (CB16/12/00316), cofinanced by the European Development Regional Fund 'A way to achieve Europe' (FEDER) and by the Spanish Association Against Cancer Scientific Foundation (ECC) (Grupos Coordinados Traslacionales ECC 2018).

Conflicts of Interest:
The authors declare no conflicts of interest.