Molecular Subtypes of Breast Cancer Are Not Associated with the Clinical Under- or Overstaging of Breast Cancer

Abstract Purpose to evaluate the agreement between the clinical and pathological stagings of breast cancer based on clinical and molecular features. Methods this was a cross-sectional study, in which clinical, epidemiological and pathological data were collected from 226 patients who underwent surgery at the Prof. Dr. José Aristodemo Pinotti Women's Hospital (CAISM/Unicamp) from January 2008 to September 2010. Patients were staged clinically and pathologically, and were classified as: understaged, when the clinical staging was lower than the pathological staging; correctly staged, when the clinical staging was the same as the pathological one; and overstaged, when the clinical staging was greater than the pathological staging. Results understaged patients were younger (52.2 years; p < 0.01) and more symptomatic at diagnosis (p = 0.04) when compared with correctly or overstaged patients. Clinicopathological surrogate subtype, menopausal status, parity, hormone replace therapy and histology were not associated with differences in staging. Women under 57 years of age were clinically understaged mainly due to underestimation of T (tumor staging) (p < 0.001), as were the premenopausal women (p < 0.01). Patients whose diagnosis was made due to clinical complaints, and not by screening, were clinically understaged due to underestimation of N (lymph nodes staging) (p < 0.001). Conclusion the study shows that the clinicopathological surrogate subtype is not associated with differences in staging, while younger women diagnosed because of clinical complaints tend to have their breast tumors understaged during clinical evaluation.


Introduction
The staging of breast cancer involves an adequate assessment of the local and distant extensions of the disease. In the age of precision medicine, clinical tumor staging must be very accurate in order to avoid inadequate treatment. For instance, data from the American College of Surgeons Oncology Group (ACOSOG) and the National Surgical Adjuvant Breast and Bowel Project (NSABP) most recent trials, as well as the European Organization for Research and Treatment of Cancer (EORTC) AMAROS (After Mapping of the Axilla: Radiotherapy or Surgery) and a few other trials defined very conservative approaches to women with early stage breast cancer 1-3 ; it seems sensible, however, that these advantages may not be enjoyed by women whose tumors have been understaged during pretreatment clinical evaluation.
The currently available methods for breast cancer clinical staging rely on physical examination and radiologic assessment of the breast. 4 It is well known that several factors may interfere with clinical breast examination (CBE), such as breast density and volume, and women's menstrual status. 5 The same applies to mammography, since breast density may compromise exam accuracy, for example. 6 Moreover, recent studies have suggested that low-grade luminal A tumors are generally detected with screening, whereas rapidly growing tumors may be detected only when they become symptomatic, 7 which implies that the molecular subtype of breast cancer may be associated with clinical presentation and thence with the sensitivity of the clinical and radiological assessments. Current staging guidelines, however, do not suggest staging procedures tailored according to molecular subtype.
It remains unknown to what extent the characteristics of patients and their tumors may hamper a correct presurgical evaluation of the breast cancer stage. In this study, we examined whether factors such as clinicopathological surrogate subtypes, age, menopause status, and diagnosis by screening or symptoms are associated with the discrepancy between the clinical and pathological stagings of breast tumors in women unaware of the existence of treatment.

Study Design
We performed a secondary analysis of data collected from other studies linked to our study group. [8][9][10][11][12] The principal investigator (Juliana Pinho Espinola -JPE) conducted a further review of 226 medical records, specifically for the present report, in order to obtain data regarding clinical staging, exams, and pathological staging. Clinical and epidemiological data are from January 2008 to September 2010. Data were collected at Professor Dr. José Aristodemo Pinotti Women's Hospital (CAISM/Unicamp). In this study, data pertaining to the molecular and pathological characteristics of the tumors had already been obtained for the studies cited above.
Conclusão o estudo nos mostra que o subtipo clinico-patológico não está associado a diferenças de estadiamento, enquanto mulheres mais jovens, e que tiveram seu diagnóstico por queixa clínica, tendem a ter seus tumores mais frequentemente subestadiados. Inclusion criteria consisted of: women undergoing primary surgical treatment for breast cancer; invasive ductal carcinoma or invasive lobular carcinoma; histological samples that could provide material for the safe and accurate diagnosis of cancer, and the information about its stage for this study. Exclusion criteria consisted of: lack of information about the staging or the molecular features of the tumor (technical inability to perform any of the laboratory procedures); patients undergoing neoadjuvant chemotherapy; patients with metastatic breast cancer; patients already treated for breast cancer before their entry into the study.

Data Collection
Information about the epidemiology, clinical and pathological staging, and surgical treatment performed were reviewed and annotated in a specific form. The review of the medical records was performed only by the main investigator, reducing the potential for bias in the interpretation of the information contained in the medical reports. The data collection form was initially used to evaluate 20 medical records, and the information was entered into the pre-existing database, allowing the adjustment of both the collection record and the database, which was formatted to contain all information relevant to this study.
Information pertaining to the clinical staging was obtained from the patient's initial appointment after the histological diagnosis had been confirmed. All patients were evaluated by a breast surgeon, and the staging of the disease was done according to well-established protocols (such as the American Joint Commission on Cancer -AJCCprotocol). 13 Women underwent two-incidence digital mammographies, craniocaudal (CC) and mediolateral oblique (MLO). Breast ultrasound was used when needed. Clinical lymph nodes status was determined with palpation of the axilla. All pathological reports were reviewed by JPE. Evaluation of the expression of estrogen/progesterone receptors HER2 and Ki67 was performed using immunohistochemistry and fluorescent in situ hybridization (FISH) according to standard protocols. All pathological assessments were performed at the Experimental Pathology Laboratory of CAISM/ Unicamp. The surrogate molecular subtypes of breast cancer were determined according to the 13 th St. Gallen Consensus. 14

Classification of Staging Status
Patients were classified as: understaged, when the clinical staging was lower than the pathological staging; correctly staged, when the clinical staging was the same as the pathological staging; and overstaged, when the clinical staging was greater than the pathological staging.

Data Analysis
Data were analyzed using the R statistical package for microcomputers (R Development Core Team, 2015, Aalborg, Denmark). The significance level was 5% (p 0.05 and confidence interval of 95%). Bivariate analysis consisted of calculating the kappa value for agreement between the clinical and pathological stagings. Cases were classified as: understaged, if the clinical staging was less than pathological; correctly staged, if both matched; and overstaged, if the clinical staging was higher than pathological (we call these conditions "staging statuses"). A logistic regression model was adjusted to determine possible associations between the congruence of staging and the clinical and pathological features of the tumors. Finally, we separately analyzed whether the molecular subtype of breast cancer was associated with staging status using the Chi-square ( 2 ) test.

Results
Eighty-six (38%) patients were understaged, 130 (57.5%) patients were correctly staged, and 10 (4.5%) patients were overstaged. In ►Table 1, we compare clinical and pathological characteristics according to staging status (understaged, correctly staged/overstaged). For the statistical analysis, patients who were correctly staged and overstaged were grouped. Patients understaged were younger (mean age at diagnosis ¼ 52.2 years; p < 0.0001) and more symptomatic at diagnosis (p ¼ 0.04) than their counterparts (mean age at diagnosis ¼ 60.7 years). Menopausal status, parity, family history of breast cancer, post-menopausal hormonal replacement, histology (whether ductal or non-ductal) and body mass index (BMI) were not associated with the staging status. The affected breast quadrant was not related to the discordance of the staging status (p ¼ 0.18, data not shown). The mean time between the first appointment at the institution when clinical staging was performed and the surgery was of 65 days for correctly and overstaged patients, and of 62 days for understaged patients (p ¼ 0.25; data not shown in table).
►Table 2 cross-tabulates the clinical and pathological staging of the women according to the 2010 AJCC consensus 13 , and further dissects T (tumor) and N (lymph nodes) stagings. Out of all women clinically considered as stage I, 32 were later considered as stage II (24 women) or stage III (8 women). From the group of women in stage II, 54 were later considered as stage III, whereas 8 were downstaged. From the stage III group, 2 were downstaged to stage II pathological. Because all metastatic patients were excluded from the analysis, there are no stage IV patients in this study. The intraclass correlation coefficient (ICC) for AJCC staging agreement was 0.54. While analyzing only T staging, ICC was slightly superior (ICC ¼ 0.62). It is worth noting that the agreement was negatively affected by 27 women who were clinically T1, T2 or T3 and were later found to be pT4, due to dermal involvement not seen in the clinical evaluation. The mean pathological size of the tumor was also relevant in understaged patients (mean 32 mm, range 12 to 80 mm) in comparison to correctly staged and overstaged patients (mean 25 mm, range 1 to 75 mm) (p < 0.001).
Restricting the analysis to N staging (ICC ¼ 0.33), agreement was low mainly because of understaging of axillary lymph nodes: out of the 137 N0 patients, 48 were later found to harbor lymph node compromise. Also, patients with a clinically positive axilla were often understaged: 41out of 85 N1 patients were actually pN2 or pN3. It was found that the understaged patients had a mean of 6.9 lymph node involvement (range 0 to 41), and the correctly staged and overstaged patients had a mean of 2.3 (range 0 to 29) (p < 0.001) compromised nodes.
In ►Table 3 we further analyze the significant associations described in ►Table 1, between staging status and age, menopausal status, and whether diagnosis was made during breast cancer screening or triggered by symptoms. We arbitrarily established the cutoff point for this analysis as the median age (57 years) of the patients. We detected that women < 57 years of age were clinically understaged mainly due to underestimation of tumor volume or dermal involvement (T) (p < 0.001). The same was true for premenopausal women (p < 0.01), although these associations were not significant after multivariate adjustment. On the other hand, patients whose diagnosis was triggered by symptoms were clinically understaged due to underestimation of their N status (p < 0.001).
We also analyzed whether the type of surgery performed (conservative or radical) was associated with the staging status (p ¼ 0.33) and the rate of re-operation due to com-promised surgical margins (p ¼ 0.45) (data not shown in tables), but no relation was found.

Discussion
Our study suggests that the surrogate clinicopathological subtype of breast tumor is not associated with the discrepancies between clinical and pathological stagings, whereas younger women and/or those with palpable tumors are at a significantly higher risk of having their disease clinically understaged. This may result in significant detrimental effects for treatment planning in younger women, since suboptimal treatments may be prescribed. Interestingly, younger, premenopausal women had their tumors understaged mainly due to T underestimation, whereas women with symptomatic tumors had their disease understaged mainly due to N underestimation. It is important to notice that CBE workup may vary according to practitioner experience, whereas mammography is performed according to strict technical parameters dictated by BI-RADS® (American College of Radiology, NY School of Medicine, New York, 2014). 15 One possible explanation for our findings is that younger women have faster growing tumors 16 compared with older women, and, in theory, these rapidly growing tumors may be smaller during pretreatment evaluation than they are at the moment surgery is performed.
Our data also revealed that clinical T underestimation was also more common in women with larger tumors, which in turn suggests that the semitechnical strategy for T estimation has sensitivity flaws. It is important to notice that we excluded, as per study protocol, patients with clinical advanced tumors, since these were prescribed neoadjuvant treatments, although 18 patients with cT4 were included, probably because they had small tumors with skin involvement, and they were not sent to neoadjuvant chemotherapy.   We found that out of 93 patients with clinical T1 tumors, 25 were further diagnosed with pT2 tumors, and another 4, with pT4 tumors. However, out of the 106 women with clinical T2 tumors, 21 were later diagnosed with pT4 disease. These figures suggest that the clinical evaluation of the breast, especially in women with larger tumors, may be flawed by the underestimation of skin and chest wall involvement, which may pose serious implications for the success of the primary surgical treatment. The failure to clinically detect lymph node involvement was the main responsible reason for the clinical underestimation of the local extension of the disease. The implication of this must be analyzed with caution, since women with clinical N0 tumors who were later diagnosed with pN1 or pN2 do not represent a technical problem, since it is expected that a percentage of patients, depending on tumor characteristics, will be in this situation, and sentinel lymph node biopsy can be used to overcome this situation. The clinical challenge resides in women who are clinically deemed N1 and are later found to be N2 or N3. For these patients, suboptimal dissection of the axillary nodes may occur due to unexpected technical difficulties during surgery. In many cases, prognosis may be affected, since neoadjuvant treatment might have been a better option, although studies have shown that treatment order does not seem to be associated with mortality due to breast cancer. 17,18 Recent evidence from the ACOSOG Z011 1 and EORTC AMAROS 3 studies have shown that women with early stage disease may benefit from less aggressive treatment protocols; however, the benefits of these less aggressive treatment protocols have not been demonstrated in women who were clinically understaged. 1,3 For instance, in the ACOSOG Z011 trial, women who were found to have tumors ! 5 cm on histology were excluded from analysis.
Overstaging is characterized by the overestimation of tumor size and spread, and may result in overtreatment. In our study, 8 women (3.5%) were thought to be clinically stage II but were pathologically stage I, and 2 women (0.9%) were clinically stage III and were pathologically stage II. We were unable to determine whether patients who were overstaged were also overtreated.
Eighty-six (38%) out of 226 women had their breast cancer understaged. However, the number of women who had to undergo re-operation did not differ between women correctly staged or overstaged and those understaged. This finding reveals that clinical understaging of breast tumors was not associated with suboptimal first-line surgical treatment, and there was no association of clinical understaging with the rate of re-operations or compromised surgical margins. It is worth mentioning that there was no significant difference in the time elapsed from diagnosis to treatment between patients clinically understaged and those correctly staged or overstaged. This is an important finding, because significantly higher intervals between diagnosis and treatment in women understaged could have been interpreted as delayed treatment.
It was our hypothesis that triple-negative, luminal B and HER2 tumors would be more frequently understaged than luminal A tumors due to the rapid growth of the former. However, this hypothesis was not confirmed, and we were surprised by the fact that younger women, and those who had their diagnosis triggered by clinical complaints and not by screening tended to have their tumors understaged more often. Unfortunately, we do not have annotated follow-up data for this cohort of patients, and thus it was impossible for us to evaluate the survival implications of clinical understaging, since there were very significant imbalances in tumor volume and lymph node compromise between groups.