The association between breast density and breast cancer pathological response to neoadjuvant chemotherapy

Breast density refers to the proportion of fibroglandular breast tissue relative to adipose tissue in the breast [1]. Wolfe first described the association between mammographically dense breasts and excess breast cancer risk in 1976 and since then numerous other studies have demonstrated a strong association between mammographic density and the risk of breast cancer [2, 3]. Women with the highest categories of breast density have a 4- to 6-fold increased risk of developing breast cancer in comparison to their counterparts with less dense breasts, although this may also be due to a ‘masking effect’ from surrounding breast tissue [4, 5].

The theory underpinning breast density and breast cancer risk is hypothesised to be due to cancer potentiation through incorporating a larger amount of high-risk epithelial cells that may proceed to malignancy within the suitable microenvironment [6]. The extra-cellular matrix (ECM) and collagen organisation within the breast stroma of dense breasts differ considerably from that of women with less dense breasts. Similarly, the presence of small proteoglycans and stromal matrix regulators such as metalloproteinases is overexpressed in women with mammographically dense breasts which may initiate tumourigenesis [5]. In essence, high mammographic breast density is thought to represent a proliferative and pro-inflammatory breast microenvironment [7].

Neoadjuvant chemotherapy (NACT) for breast cancer was once reserved for locally advanced breast cancer to induce downstaging and render tumours operable, however up to 30% of breast cancer patients may now receive NACT [8, 9]. Although the long-term survival outcomes for patients treated with neoadjuvant chemotherapy are comparable to those treated with adjuvant chemotherapy, the proposed benefits associated with neoadjuvant chemotherapy are numerous [10]. The most stringent definition of a pathological Complete Response (pCR) is complete disappearance of invasive and in situ residual tumour disease in the breast and axilla (ypT0) [11]. The Residual Cancer Burden (RCB) index developed from the M.D. Anderson Cancer Centre (MDACC) is a scoring system used to quantify residual disease following NACT. The RCB index combines pathological findings in the primary tumour bed and the regional lymph nodes to calculate a continuous index or score.

Considering the individualised response to NACT, predictive markers are required to select patients who will derive a clinical benefit from NACT. Previous studies examining the value of mammographic density in predicting pathological response to NACT have yielded conflicting results. The aim of this study was to investigate the association between BI-RADS breast composition categories and pCR/RCB index after NACT in a consecutive cohort of 292 Irish women with breast cancer.

Mammographic density was categorised according to the BI-RADS 5th edition at the time of diagnosis and recorded by five radiologists. The BI-RADS score ranged from a to d with category a meaning “the breasts are almost entirely fatty”, category b meaning “scattered fibroglandular density”, category c meaning “heterogeneously dense” and category d meaning “extremely dense”. Six patients did not have their breast density recorded at the time of diagnosis and were categorised retrospectively by a radiologist who was blinded for patient outcomes. Tumour size was calculated using mammography and ultrasound (from the radiology report). In this institution all patients with breast cancer are routinely screened for axillary nodal disease using ultrasonography ± fine needle aspiration (FNA) of indeterminate or suspicious nodes. The Royal College of Pathologists Guidelines are used for classification of fine needle aspirate cytology. C1 is considered inadequate, C2 is considered benign, C3 contains atypical cells, C4 contains cells suspicious for malignancy and C5 contains malignant cells. There were no patients with C3 axilla FNA cytology. Tumour phenotype was recorded from data derived from diagnostic core needle biopsies. Oestrogen Receptor (ER) and Progesterone Receptor (PgR) were considered positive when > 10% of cells showed positive staining. Progesterone receptor staining was not performed routinely in our institute until 2019. HER2 status was assessed by immunohistochemistry (IHC) and Fluorescent in situ Hybridisation (FISH) confirmation in cases with 2+ by IHC. HER2 positivity was defined as either 3+ with IHC and/or amplified with FISH. Patients received NACT according to standard guidelines. Patients with HER2 positivity received HER2 blockade (trastuzumab) concomitantly with NACT. From July 2020 patients received dual-HER2 blockade with pertuzumab and trastuzumab in line with national guidelines. A complete pathological response referred to the absence of any residual invasive or situ disease in the breast and axilla (pCR) following completion of NACT. The residual cancer burden score and class were recorded in the histology report or calculated using the available data. Individual data on menopausal status were not available and therefore arbitrarily the age 52 was used to categorise women into pre- and post-menopausal categories. The average age of menopause in Ireland is 52 and therefore women < 52 were considered pre-menopausal and those aged ≥ 52 were considered post-menopausal [12].

The results of this study demonstrate that breast density did not appear to be an independent predictive marker of pathological response in the neoadjuvant setting. We found that oestrogen receptor positivity was significantly associated with BI-RADS category d, but MD was not an independent predictor of response to NACT.

These findings are concordant with the recent “NeoDense” study conducted by Skarping et al. in Sweden. Their prospective analyses of 200 breast cancer patients observed that whilst MD decreased in many patients during NACT it was not a predictive marker of pathological response to NACT [13]. Previously, the same group performed a retrospective study of 302 patients and concluded that women with BI-RADS d MD were less likely to achieve pCR post-NACT after adjusting for age and pre-treatment tumour characteristics. The results of both study sets were recently collated [14] and demonstrated that pre-menopausal women with category MD d were less likely to achieve pCR after NACT [14]. Conversely, a recent study of 442 patients by Di Cosimo et al. concluded that those with denser breasts showed an increased likelihood of pCR with odds ratio (OR) of 1.70, 2.79 and 1.47 for b, c and d categories. However, none of the MD categories were associated with pCR in the univariable logistic regression analysis and the results only proved marginally significant for MD category C rate using a fully adjusted model (OR c versus a: 2.79, 1.04–7.48) [15]. This distribution of patients in MD category a and d were significantly higher than those represented in the present study. An association between breast density and pCR/RCB score was not demonstrated in the current study.

This study did not chronologically track changes in breast density; however, there is substantial evidence to suggest that age and hormonal events may result in changes in MD. Menopause in particular results in a significant reduction in breast density [16]. Several studies have also investigated the change in breast density during neo- and adjuvant chemotherapy. A significant reduction in breast density post-adjuvant chemotherapy was demonstrated by Knight et al. and Sandberg et al. [17, 18]. Interestingly, a greater than 10% reduction in MD was associated with a significantly reduced risk of contralateral breast cancer compared to women with little or no change in MD. MD reduction is more prominent in younger pre-menopausal women receiving adjuvant chemotherapy compared to post-menopausal women [19]. The underlying biological explanation for reduced MD in pre-menopausal women during chemotherapy may be explained by the induction of amenorrhoea causing lobular atrophy [20].

A pCR was achieved in 21.9% of patients in this study, defined as no invasive or in situ residual disease in the breast and nodes. This strict definition of pCR was adopted as it can best discriminate between favourable and unfavourable patient outcomes [11]. Traditionally, patients with residual in situ disease or foci of invasive disease were considered to have achieved a pCR; however, a pooled analysis of 6377 patients treated with NACT reported an increased risk of relapse and death compared with the group of patients with stage ypT0 N0 breast cancer [11]. The pCR definition restricted to ypT0N0 showed the lowest adjusted HR for DFS and OS compared with the other definitions and should be more accurately adopted. Von Minckwitz et al. pooled analysis of seven NACT randomised controlled trials achieved an overall pCR rate of 15% which is significantly lower than the 21% achieved in this study. This can be explained by their historical patient cohort (1998–2006) and that the findings from this study resulted in a more careful patient selection for NACT. Results from this study suggest that patients with HER2-enriched breast cancer are almost four times more likely to achieve pCR. pCR is a surrogate endpoint for event-free and overall survival. Specifically, pathological complete response is associated with a 52% reduction in the probability of an event and a 64% reduction in the probability of death [21].

To the best of our knowledge this is the first study to examine the predictive value of MD using the RCB score in the setting of NACT. This index is subdivided into four classes with an increasing amount of residual disease: RCB 0 (pCR), RCB-I, RCB-II and RCB-III. An RCB score 0/class O refers to a complete pathological response with no evidence of invasive disease in the breast or axilla, the presence of in situ disease is not considered. 29.8% of patients in this study had Residual Cancer Burden score 0 which includes ypTisypN0 pathological stages. The RCB categories have been shown to correlate with long-term survival outcomes across triple-negative breast cancer (TNBC) and HER2 breast cancer subtypes and several clinical study groups, such as I-SPY [1, 2], ACOSOG (Z11103), CALGB (40601, 40603), NSABP (B-40, B-41) and ABCSG [22]. The RCB assessment is clinically useful as it is highly reproducible, with reproducible long-term prognostic significance [22]. Mammographic density was not associated with RCB assessment in this cohort of 292 patients.

In conclusion, our data suggest that whilst MD assessed at the time of diagnosis is significantly associated with oestrogen-positive breast cancer, MD is not an independent predictive marker of response to breast cancer NACT.