The clinical assessment of nodal status (as in TNM classification) is unreliable.^2,³ Palpable nodes may be enlarged because they show benign reactive changes whilst nodes bearing tumour deposits can be impalpable. Although axillary sonography is a moderately sensitive and fairly specific technique in the diagnosis of axillary metastatic involvement,⁴ histological examination of the lymph nodes from the axilla should be carried out in all patients with primary operable invasive breast cancer. It is well known that patients who have histologically confirmed lymph node involvement have a significantly poorer prognosis than those without nodal metastases. The 10-year survival is reduced from 85% for patients with no nodal involvement to 40% for those with involvement of multiple nodes. Prognosis worsens the greater the number of nodes involved, and the level of nodal involvement can provide useful information; metastasis to the higher level nodes (level II or III) in the axilla and particularly those at the apex (level III) carries a worse prognosis (Fig. 2.1).

Figure 2.1 Overall survival for patients in the NTPBCS according to lymph node stage (P < 0.0001).

Studies show that overall survival is decreased as the number of nodes that are involved increases.5–7

Optimal management of the axilla is discussed in Chapter 9.

Sentinel node biopsy or blue dye directed axillary sampling has been shown to provide accurate prognostic information and has an extremely low rate of lymphoedema.² Of 1275 patients undergoing axillary sampling, only 0.04% suffered symptomatic complaints of arm swelling and lymphoedema.

A refinement of lymph node sampling is provided by the technique of sentinel lymph node (SLN) biopsy. There have been several studies to examine the validity of this technique. The ALMANAC trial provides level I evidence for UK-based practice.³

From a pathological perspective, there are unresolved questions about how to optimally assess the sentinel node. Studies have shown that the mean number of sentinel nodes is close to 3. As the number of nodes removed during SLN biopsy is fewer, pathologists are now challenged to process these few nodes optimally. Several methods have been studied: routine paraffin histology; intraoperative frozen section; immunohistochemistry and more intensive methods such as serial sectioning and molecular methods including polymerase chain reaction (PCR) and reverse transcription–PCR (RT–PCR).

At present there is no consensus for the optimal handling of the SLN in the laboratory. Most centres have developed their own ‘in-house’ protocols, which are invariably tied to the institution’s research ambitions. This ad hoc approach can influence interpretation of results, as the amount of material assessed between centres is not uniform. In a number of large studies, using different histopathological techniques, the SLN false-negative rate (defined as how often the SLN is negative for malignancy when cancer is present in the rest of the axilla) varies between 0% and 11%.

In Nottingham the MRC ALMANAC protocol is employed to assess all sentinel node biopsies. These nodes are cut into slices about 3 mm thick, taken perpendicular to the long axis of the node to maximise the assessment of the marginal sinus, with one node per cassette. The majority of nodes can be completely embedded in one cassette. Larger nodes have alternate slices embedded and may require more than one cassette. Large, obviously involved nodes have one section taken.

Intraoperative frozen sections or imprint cytology of axillary lymph nodes have been advocated by some authors. Conventional frozen sections have a high false-negative rate of between 10% and 30%. More intensive intraoperative assessment with serial sections and immunohistochemistry has been described,⁸ but is time consuming and labour intensive. Frozen section is best applied to selected cases; for example, if the node is macroscopically abnormal and this is confirmed histologically to be metastatic carcinoma, further axillary surgery can be performed immediately. Some studies have found low false-negative rates of 2–3% with intraoperative imprint cytology,⁹ but not all have been able to achieve this level of accuracy. Lymph node status can be assessed peroperatively using ultrasound combined with FNA or core biopsy, and this reduces the need for preoperative frozen section and imprint cytology to assess axillary nodes intraoperatively.

RT–PCR is more sensitive than immunohistochemistry at detecting metastatic tumour in axillary nodes. Two types of methods have been used to detect tumour cells with molecular techniques. First, a genetic defect such as chromosomal rearrangement or mutation can be used. The problem with this is that no single genetic defect is seen in all breast carcinomas. The second method is to use a molecular marker that is present in tumour cells but not in the adjacent tissue. To identify a marker that has this specificity and is expressed in the majority of tumours is difficult. It may be necessary to use a panel of markers. A major problem with PCR is the potentially high false-positive rate due to the sensitivity of the method and potential for contamination. In addition, it is not possible with PCR to determine whether the DNA comes from cells that are viable. An advantage of haematoxylin/eosin sections and immunohistochemistry is that the morphology of the cells can be examined and malignancy confirmed. A major unresolved question is whether carcinoma detected only by PCR or RT–PCR has the same prognostic significance as haematoxylin/eosin positivity.

The detection rate of micrometastases in axillary lymph nodes has been reported to range from 9% to 46%.¹⁰ Studies have used serial sectioning with or without immunohistochemical stains to detect micrometastatic foci and these methods have increased detection rates.

The European Working Group for Breast Screening Pathology ¹¹ has formulated working guidelines for the assessment and pathological work-up of SLNs in breast cancer. From a literature review available at that time, the committee concluded that it was not possible to determine the significance of micrometastasis or isolated tumour cells. They noted that approximately 18% of cases are associated with other nodal (non-sentinel node) metastases. False-negative rates are most often determined via immunohistochemistry (IHC). However, at present it is recommended that an intensive work-up is not justified on a population level. The committee did suggest multilevel assessment and, where resources permit, intraoperative assessment, although the results of Z0011 (see Chapter 7) suggest that some node-positive patients may not require axillary dissection and this brings the whole issue of intraoperative assessment into question. The current view is that IHC should not be used routinely to assess axillary lymph nodes.

Tumour size

Tumour size is one of the most powerful predictors of prognostic factor in breast cancer.^12,¹³ The frequency of nodal metastases in patients with tumours < 10 mm is 10–20%,¹³ and node-negative patients with tumours < 10 mm have a 10-year disease-free survival rate of greater than 90%.¹⁴ Tumour size is a time-dependent prognostic factor that depends on the period between tumour development and detection, and on the balance between tumour cell proliferation and death (tumour growth rate). It is well known that the association between increasing tumour size and increasing number of positive lymph nodes and worse outcome is highly statistically significant.¹³ Therefore, the main aim of population screening with mammography is to detect smaller tumours that are likely to have a better outcome than those that present symptomatically (of a larger size). Patients with smaller tumours have a better long-term survival than those with larger tumours (Fig. 2.2). Estimation of tumour size has assumed particular importance since the introduction of population screening. In most studies the frequency of axillary lymph node metastasis in small < 15 mm (so-called minimally invasive carcinoma (MIC)) invasive carcinomas is 15–20%,^15–17 compared with over 40% in tumours measuring 15 mm or more.¹⁸ During the prevalent round of breast screening even more favourable results are obtained, with the frequency of axillary lymph node metastasis ranging from 0% to 15%.^19–22 The Nottingham Tenovus Primary Breast Cancer Study (NTPBCS) has generated data that suggest the cut-off point of 10 mm is not the best discriminator for MIC. Life-table analysis of survival curves found no difference between tumours measuring up to 9 mm and those measuring 10–14 mm. This indicates that < 15 mm may be a more realistic watershed in defining small, invasive, carcinomas of good prognosis. It is clear that pathological tumour size is a valuable prognostic factor and it has become an important quality assurance measure for breast screening.^21,^23–26 It is also used in part to judge the ability of radiologists to detect small impalpable invasive carcinomas.

Figure 2.2 Overall survival according to size.

Differentiation

Modern pathologists have recognised that invasive carcinomas can be divided according to their degree of differentiation. There are two ways to achieve this: (i) by allocating a histological type according to the architectural pattern of the tumour; (ii) by assigning a grade of differentiation based on semiquantitative evaluation of structural characteristics.

Certain histological types of invasive carcinoma carry a favourable prognosis. Tubular, mucinous, invasive cribriform, medullary and tubulolobular types, together with rare tumour types such as adenoid cystic carcinoma, adenomyoepithelioma and low-grade and squamous carcinoma, have all been reported to have a more favourable outcome than invasive carcinoma of no special type (ductal NST). Undoubtedly, assessment of histological type provides prognostic information in breast cancer. However, this effect is relatively small in multivariate analysis ²⁷ when compared with the prognostic value of histological grade; histological type may prove to be more useful in increasing our understanding of the biology of breast cancers.²⁸

Invasive lobular carcinoma (ILC) is of particular importance. It comprises approximately 5–15% of breast cancers and appears to have a distinct biology. It is less common than invasive carcinoma of no special type, also called invasive ductal carcinoma (IDC).

Recently, Rakha et al.²⁹ examined a large group of 5680 breast tumours (415 patients (8%) with pure ILC and 2901 (55.7%) with IDC (no special type)) and demonstrated that, compared to IDC, patients with ILC tended to be older and have tumours that were more frequently of lower grade (typically grade 2; 84%), hormone-receptor positive (86% compared to 61% in IDC), of larger size and with absence of vascular invasion. More patients with ILC compared with IDC were placed in the good Nottingham Prognostic Index group (40% compared to 21% in IDC). ILC showed indolent but progressive behavioural characteristics with nearly linear survival curves that crossed those of IDC after approximately 10 years of follow-up, thus eventually exhibiting a worse long-term outcome. Interestingly, ILC showed a better response to adjuvant hormonal therapy (HT), with improvement in survival in patients who received HT compared with matched patients with IDC.

Tumour grade

The Nottingham method (Elston and Ellis) is a modification and improvement of previously existing morphological grading systems and is able to provide greater objectivity of grade.³⁰

In brief, assessment of grade considers three tumour characteristics: tubule formation, nuclear pleomorphism and mitotic counts. A numerical scoring system of 1–3 is used for each factor individually. The three scores are added together to produce scores of 3–9 on which an overall tumour grade is assigned:

• Grade 1 – well differentiated = score 3–5 points.

• Grade 2 – moderately differentiated = score 6–7 points.

• Grade 3 – poorly differentiated = score 8–9 points.

It should be noted that grade is valuable irrespective of histological type.

There is a highly significant correlation between tumour grade with long-term prognosis (Fig. 2.3); patients with grade 1 tumours have a 93% chance of surviving 10 years after diagnosis, whereas in patients with grade 3 tumours this is reduced to 70%. It has now been shown conclusively that the Nottingham method, with its more objective criteria, has excellent reproducibility when used by experienced pathologists.

Figure 2.3 Overall survival according to grade.

Interestingly, grade is not included in the recent revision of the TNM staging system of breast cancer as its value is questioned in certain settings such as lobular type.

The value of grade was addressed by Rakha et al.³¹ in a large study of grade and overall survival in a large and well-characterised consecutive series of operable breast cancer (n = 2219 cases), with a long-term follow-up (median 111 months) using the Nottingham histological grading system. Histological grade was strongly associated with both breast cancer-specific survival (BCSS) and disease-free survival (DFS) in the whole series as well as in different subgroups based on tumour size (pT1a, pT1b, pT1c and pT2) and lymph node (LN) stages (pN0 and pN1 and pN2). The authors were able to demonstrate differences in survival between different individual grades (1, 2 and 3). In multivariate analyses histological grade was an independent predictor of both BCSS and DFS.

Additionally, the usefulness of routinely assessing grade in invasive lobular cancer was examined by this group in 4987 patients,³² of whom 517 were pure ILC cases. The majority of ILC was of classical type or mixed lobular variants (89%). Most ILC cases were moderately differentiated (grade 2) tumours (76%), while a small proportion of tumours were either grade 1 or 3 tumours (12% each). There were significant associations between histological grade and other clinicopathological variables of poor prognosis such as larger tumour size, positive lymph node status, vascular invasion, oestrogen receptor and androgen receptor negativity, and p53 positivity. Multivariate analyses showed that histological grade was an independent predictor of BCSS and disease-free interval.

Vascular invasion

The presence of tumour emboli in vascular and lymphatic spaces has emerged as an important prognostic factor. Several studies have now shown that the presence of vascular invasion correlates closely with local and regional lymph node involvement.^33,³⁴ It has been suggested that it can provide prognostic information as powerful as lymph node stage. Reproducibility is the limiting factor in its widespread adoption and routine clinical assessment. In a Nottingham study of assessing vascular invasion, the issue of reproducibility was addressed specifically. Of 1704 cases, a subset of 400 cases was examined by two or more pathologists. There was a 77% overall inter-observer agreement on histological features and an 85.8% overall agreement in the classification of vascular invasion. Other studies have reported a similarly high concurrence between pathologists.^35,³⁶ The assessment of vascular invasion is subjective but there is good evidence that a high rate of concurrence can be obtained as long as strict criteria are used. Lymphatic and vascular invasion is considered to be a valuable surrogate for lymph node stage in cases where nodes have not been removed for examination. In patients whose axillary nodes are tumour free on histological examination there is a correlation between the presence of lymphovascular invasion (LVI) and early recurrence.

Recently, Lee et al.³⁷ assessed the prognostic value of LVI in a group of 2760 patients with node-negative breast cancer with long-term follow-up (median 13 years). This study demonstrated a strong association between poor histological grade and younger age with LVI-positive cancers. LVI was prognostically significant and was independent of grade, size and type for overall survival (see Fig. 2.4).

Figure 2.4 LVI and breast cancer-specific survival.

Molecular/predictive factors

Despite the overall association between molecular markers with prognosis and outcome, they are limited in their ability to capture the nuances of the complex cascade of events that drive the clinical behaviour of breast cancer. Morphological factors are unable to predict response to systemic treatments, and tumours of apparently homogeneous morphological characteristics still vary in response to therapy and have distinct outcomes. The use of strategies ranging from hormone therapy to chemotherapy and recently to novel receptor-directed therapies and vaccines relies on the expression of predictive factors by the cancer, either individually (e.g. oestrogen receptors (ERs), progesterone receptors (PRs), human epidermal growth factor receptor (HER-2) and epidermal growth factor receptor (EGFR)) or globally (e.g. high-throughput gene expression assays). These molecular markers are used not only to guide treatments but are also being used to monitor response and detect relapse. It is envisaged that molecular predictive markers may form the basis of tailoring an individual’s adjuvant therapy based on the genetic fingerprint of their cancer.

Recently, Tsutsui et al.⁴⁷ reported on a large series of 1029 patients (with adjuvant therapy) and found EGFR to be of independent prognostic value irrespective of nodal status. In contrast, studies by Rampaul et al.⁴⁸ and Ferrero et al.⁴⁹ (without adjuvant therapy) showed EGFR to be of no prognostic value (these latter series also incorporated grade and had longer follow-up).

The predictive value of this marker for hormone resistance or responsiveness is better defined. EGFR tumours are considered more resistant to endocrine therapy, and there exists an inverse relationship with EGFR-negative cancers (they are more likely ER positive), being more often sensitive to endocrine manipulation. Though some studies have been able to demonstrate that EGFR status is a predictor of tamoxifen failure and even response rates,⁵⁰ there are conflicting data from well-designed level II studies⁵¹ that have shown no value of EGFR in predicting the efficacy of tamoxifen in high-risk postmenopausal women.

Perhaps the most exciting use of EGFR status and indeed the driving impetus for clinicians and scientist to measure it accurately is its putative role in defining those who should respond to novel EGFR-directed therapies.

HER-2

HER-2 is an important target in the development of a variety of new cancer therapies, which include monoclonal antibody (mAb)-based therapy, small-molecule drugs directed at the internal tyrosine kinase portion of the HER-2 oncoprotein, and vaccines. The most widely known HER-2-directed therapy is trastuzumab (Herceptin; Genentech, South San Francisco, CA, USA). Trastuzumab is a humanised recombinant mAb that specifically targets the HER-2 extracellular domain. There are a variety of techniques available to determine HER-2 status in breast cancer, some of which are employed for research purposes only. In diagnostic pathology laboratories HER-2 status is assessed routinely either by IHC, which assesses expression of the HER-2 oncoprotein, and fluorescence in situ hybridisation (FISH), which measures the number of HER2 gene copies per chromosome 17 or gene amplification. Modifications of ISH using colorimetric detection are being developed, including chromogenic in situ hybridisation (CISH) and silver-enhanced in situ hybridisation (SISH).

Methods of HER-2 testing: Current guidelines for HER-2 testing ⁵² specify the methods that are suitable to detect either the HER-2 protein (by IHC) or gene amplification (using FISH or other in situ methods). Guidelines stress the need for stringent, reproducible and consistent criteria for testing.

Immunohistochemistry for HER-2 testing: Among the methods in use for determining HER-2 status, IHC is the most widely used. In studies employing various commercially available antibodies, a wide variety of sensitivity and specificity in fixed paraffin-embedded tissues is seen.^53,⁵⁴ Antigen retrieval techniques are currently not standardised and they introduce the potential for false-positive staining. Nonetheless, IHC possesses many advantages to support its widespread adoption: (i) it allows for the preservation of tissue architecture and so can be used to identify local areas of overexpression within a heterogeneous sample, and can distinguish between HER-2 positivity in in-situ and invasive cancer; (ii) it is applicable to routine patient samples, facilitating use as a diagnostic test, and this allows prospective and retrospective research studies of HER-2 status to be undertaken.

Two Food and Drug Administration (FDA)-approved IHC tests for determining HER-2 status are available: HercepTest (DAKO, Carpeteria, CA, USA), based upon a polyclonal antibody; and CB11 (Pathway, Ventana Medical Systems, Tucson, AZ, USA), based upon a monoclonal antibody. The National Comprehensive Cancer Network guidelines ⁵⁵ classify an IHC score of 0 or 1 + as representing HER-2-negative status, 3 + as positive, while 2 + is equivocal. Positive staining is defined as strong, continuous membranous expression of HER-2 in at least 10% of tumour cells. However, a joint report from the American Society of Clinical Oncology (ASCO) and the College of American Pathologists (CAP)⁵² specified a threshold of > 30% strong circumferential membrane staining for a positive result. If both uniformity and a homogeneous, dark circumferential pattern are seen, resultant cases are likely to be amplified by FISH as well as positive for HER-2 protein expression. The equivocal range for IHC (score 2 +), which may include up to 23% of samples, is defined as complete membrane staining that is either non-uniform or weak in intensity but with obvious circumferential distribution in at least 10% of cells. Equivocal or inconclusive results should be tested by FISH. Consistent with previous guidelines, a negative HER-2 test is defined as either an IHC result of 0 or 1 + for cellular membrane protein expression (no staining or weak, incomplete membrane staining in any proportion of tumour cells), though up to 5% of 1 + on IHC may be FISH positive.

In situ hybridisation for HER-2 testing: FISH and CISH measure directly the number of HER2 genes per chromosome 17, and when there is a chromosome centromeric enumeration probe (CEP) included, the copy number of chromosome 17 gene amplification is defined as an increase in HER-2/CEP17 ratio above 2.0. ISH results are semiquantitative, counting the number of signals in non-overlapping interphase nuclei of the lesion using either single-colour (HER-2 probe only, e.g. Ventana Inform) or dual-colour hybridisation (using HER-2 and chromosome 17 centromere probes simultaneously (e.g. Abbott, Chicago, USA, DAKO, Copenhagen, Denmark, etc.), the latter making it easier to distinguish true HER-2 amplification from chromosomal aneuploidy. ISH allows simultaneous morphological assessment, where evaluation of gene amplification can be restricted to invasive carcinoma cells. Many studies have compared FISH and IHC in the evaluation of HER-2 and have demonstrated concordance between the two techniques of up to 91%. Two studies have shown that FISH predicts more accurately HER-2 positivity than IHC when applied to molecularly characterised breast cancers.^56,⁵⁷

Three FISH tests are FDA approved for selecting patients for treatment with trastuzumab. The Path Vysion (Vysis Inc., Downers Grove, IL, USA) and PharmDx (Dako) tests require a ratio (HER-2 to CEP17) of 2.2 or greater for the sample to be considered amplified and both include an HER2 gene probe and a chromosome 17 probe. The INFORM test (Ventana Medical Systems) requires that at least 5.0 gene copies of HER2 be present if a sample is to be considered amplified as this kit uses a single HER2 gene probe without a chromosome 17 probe.

In brief, a two-phase testing algorithm based on IHC assay as the primary screen with FISH being reserved for equivocal cases is currently recommended. This is based on evidence showing very good concordance between IHC and FISH results on breast carcinomas from 37 laboratories when tested in experienced reference centres.⁵⁹

Chromogenic in situ hybridisation (CISH/SISH): CISH and SISH are colorimetric methods to detect gene amplification that can be viewed using a standard light microscope. Concordance between FISH and SISH, for the validation of HER2 gene status, is about 96% (kappa = 0.754).

Prognostic significance and association with other prognostic factors: The seminal work by Slamon et al.⁶⁰ in 1987 showed that HER2 gene amplification independently predicted overall survival (OS) and disease-free survival (DFS) in a multivariate analysis in node-positive patients. Since then most large studies have confirmed this relationship in multivariate analysis. It is now well established that there is a significant correlation between HER-2 overexpression/amplification and poor prognosis for patients with nodal metastasis. There was until recently no consensus on the prognostic value of HER-2 in node-negative breast cancer patients, a group most often diagnosed through screening and representing a subgroup that could potentially benefit from appropriate adjuvant therapy. More recent studies have, however, shown a consistently poorer prognosis when comparing HER-2-positive versus -negative small node-negative breast cancers.

Rilke et al.⁶¹ reported on the prognostic significance of HER-2 expression and its relationship with other prognostic factors. Using specimens from 1210 consecutive patients treated between 1968 and 1971 at a single institution (National Cancer Institute of Milan), with no systemic adjuvant therapy and 20-year follow-up, they found overexpression of HER-2 in 23% and showed a negative impact on survival of node-positive but not node-negative patients. Analysis of HER-2 in relation to the presence of lymphoplasmacytic infiltrate (LPI; favourable prognosis) and nodal status demonstrated that in node-negative, LPI-negative patients, HER-2 overexpression showed the same level of correlation with poor prognosis as with those patients with nodal metastasis. However, in the patients with node-negative disease and LPI positivity, HER-2 overexpression correlated with good prognosis. Some studies have reported a prognostic value for HER-2 in node-negative patients in selected subgroups,^62–64 whereas others have shown no correlation.^65,⁶⁶

Mirza et al.⁶⁶ have published a systematic review of prognostic factors in node-negative disease. Data for HER-2 showed a lack of standardisation of assays and no association with survival. However, HER-2 status was shown to be of independent prognostic significance in several large studies with long-term follow-up, with worstened prognosis for those node-negative patients with HER-2-positive signalling.^67,^68,⁶⁹

There is at present no agreement on the association between HER-2 and other prognostic factors. Several studies have shown a lack of association between HER-2 status and tumour size,^60,^70,⁷¹ although some do report a correlation.^61,^66,^72–76

Prediction of response to therapy: hormonal therapy: Transfection of normal breast cancer cells with the HER2 gene has been shown to result in acquisition of oestrogen-independent growth that is insensitive to tamoxifen.^77,⁷⁸

A number of clinical studies, using various end-points such as time to relapse, more rapid spread to other sites, and DFS or OS, have reported an association between HER-2 positivity and resistance to hormonal therapy.79–83 Some reports have described specific resistance to tamoxifen in HER-2-overexpressing tumours.^81,⁸² The 20-year update of the Naples GUN Trial⁸¹ found that HER-2 overexpression not only predicted resistance to tamoxifen, but that HER-2-positive patients had a worse outcome on tamoxifen therapy compared with those who were untreated.

Several studies have also shown a reduction in response rates to hormonal therapy. Metastatic breast cancer that overexpressed HER-2, measured by high plasma levels of extracellular domain, demonstrated a substantial reduction in response rate to hormonal therapy.⁸⁴ Other studies have failed to find an association or even a trend between HER-2 status and response to hormonal therapy.^85–87 Elledge et al.⁸⁷ examined the response to tamoxifen in 205 tumours with ER-positive disease. In HER-2-positive compared to HER-2-negative patients, they found no significant evidence for a poorer response, time to treatment failure or survival. In a more recent study, the relationship between HER-2 overexpression and response to tamoxifen was examined in the adjuvant setting in 741 (650 ER positive, 91 ER negative/PR positive) of the total of 1572 patients in the CALGB 8541 Trial who had HER-2 measured.⁸⁸ Tamoxifen significantly improved response, DFS and OS, irrespective of HER-2 status. However, tamoxifen was not randomised within this trial and all patients received one of three regimens of doxorubicin. Thus these data on tamoxifen resistance have limitations to their interpretation.

With regards to HER-2 and prognosis, not only is there clinical value in its positive expression but also in its absence – in so-called ‘triple-negative phenotype’ cancers (TNP), i.e. cancers that are HER-2, ER and PR negative. It also denotes a biologically different subgroup. The recognition of basal phenotype in these triple-negative cancers is of growing importance, with several lines of evidence supporting the view that triple-negative tumours and basal phenotype are not interchangeable but rather distinct entities. Ninety-one per cent of TNP tumours display a significant association with the basal-like centroid and about 20% of non-TNP tumours cluster together with TNP tumours in the ‘basal-like’ cluster; these data support the view that the majority of, but not all, TNP tumours have a basal-like phenotype and that the majority of, but not all, basal-like tumours have a TNP phenotype.⁸⁹ Triple-negative breast cancers have a relapse pattern that is very different from hormone-positive breast cancers: the risk of relapse is much higher for the first 3–5 years but drops sharply and substantially below that of hormone-positive breast cancers after that. This relapse pattern has been recognised for all types of triple-negative cancers for which sufficient data exist, although the absolute relapse and survival rates differ across subtypes.

Previous studies have shown that the expression of ‘basal markers’ (i.e. CK5/6, CK14, CK17 and/or EGFR) is associated with a poor prognosis, regardless of hormone receptor expression. The expression of basal markers (basal cytokeratins and EGFR) in TNP tumours (core basal phenotype) also correlates with a worse prognosis and identifies a clinically distinct subgroup within the TNP group. Moreover, it should be noted that identification of a subgroup of tumours solely based on the lack of expression of immunohistochemistry (for example, TNP) risks misassignment based on technical artefacts.⁸⁹

An in-depth discussion of these issues is beyond the scope of this work and the reader is directed to detailed reviews by Rakha et al.⁸⁹

Histopathology of patients with BRCA1 and BRCA2 mutations

The cancers that develop in patients with a genetic predisposition to breast cancer as a result of mutations in the breast cancer susceptibility genes BRCA1 and BRCA2 are of great clinical interest. Identification of histological features that could indicate a genetic predisposition would be useful in providing an insight into the function of these genes and may aid in identifying those in whom screening for genetic mutations would be useful. There is a general agreement that BRCA1-related cancers are more frequently ‘medullary-like’ carcinomas and high grade compared to those in patients without this genetic alteration. Cancers associated with BRCA1 mutations have a significantly higher mitotic rate, a larger proportion of tumour with a continuous pushing margin and more lymphocytic infiltration than sporadic breast cancers.^90,⁹¹ These are also likely to be less positive for ERs and PRs and are more often aneuploid, have a high S-phase fraction, show greater accumulation of p52 protein and expression of basal cytokeratins (CK5/6, CK14 and CK17) than do sporadic breast cancers.^92–97 However, none of these features alone or in combination can be used to identify a cancer as being from a BRCA1 gene mutation carrier. Pathological features reported in BRCA2-related breast cancer have been less consistent. Marcus et al.⁹² noted a significantly higher proportion of tubulolobular cancers in BRCA2 mutation carriers than in other patients. However, it has been reported in another study that tubular carcinomas are less common in BRCA2 mutation carriers.⁹⁴ Armes et al.⁹⁷ investigated the histological phenotypes of breast carcinoma in women under 40 years of age with and without BRCA1 and BRCA2 germ-line mutations. It was found that the pleomorphic variant of invasive lobular carcinoma was more common in patients with BRCA2 mutations. Others have reported that BRCA2-related cancers tend to be of high histological grade,^94,⁹⁸ whereas others have not noted any significant difference in grade between BRCA2-related cancers and controls.⁹⁹

High-throughput molecular techniques as prognostic and predictive tools in breast cancer

Genomics

Mechanisms of genetic aberrations

The sequencing of the entire human genome has permitted the introduction of high-throughput technologies that allow survey of thousands of genes and their products in a single assay. Together with powerful analytical tools this has opened new avenues for classifying breast cancer into biologically and clinically distinct groups based on DNA copy number alterations and gene expression patterns.

Cancer development is driven by the accumulation of DNA changes within the genome. DNA repair defects lead to a genome-wide genetic instability and this can drive further cancer progression. Genomics (the study of the human genome), transcriptomics (the study of the transcriptome (mRNAs)) and proteomics (the analysis of the protein complement of the genome) are the three branches of molecular biology that are being explored in an effort to improve understanding, diagnosis, prognostication and provision of new therapeutic targets for the treatment of breast cancer.

The genetic fingerprint provides information on normal cellular processes and morphological/phenotypic expression. When this message is altered, it forms the nidus for the development and progression of cancer. These alterations can be caused by DNA mutations, chromosomal aberrations, epigenetic modification and protein interactions. DNA mutations can lead to a change in gene function. These mutations are often due to base substitutions that may directly cause a stop codon. As a result the gene is only partly transcribed and any functional protein production terminated. Chromosomal instability can be due to DNA modification, mutation or viral genome integration. Changes in chromosome numbers are also seen and cancers can be aneuploid (60–90 chromosomes) or near diploid (46 chromosomes). In solid tumours virtually all chromosomal rearrangements are unbalanced, the net result being loss or gain in certain parts of the chromosome. By screening the genome, current losses and gains can be identified. This information can be used to investigate regions of overlap to reveal the genes that are involved in malignant transformation and cancer progression. In some situations, the DNA copy number and code remain intact but the accessibility of this DNA for transcription is affected by epigenetic modifications. This process can occur through DNA methylation. It directly silences genes and interferes with the binding of transcription factors by changing the chromatin structure around the gene. Global hypomethylation is a characteristic feature of the genome of a cancer cell. However, some sequences can be hypermethylated, such as CpG islands, which tend to lie around the transcription start sites of approximately half of all human genes. DNA hypomethylation has been shown to correlate with chromosome instability. Chromatin architecture remodelling is essential for gene transcription; thus the prognostic significance of quantitative chromatin changes.^100,¹⁰¹

Assessment of genomic status in breast cancer

Several high-throughput molecular techniques have been developed to assess the status of the genome in a given cell population, in terms of copy number, DNA sequencing and structure. Large-scale DNA or gene copy number alterations can be assessed using loss of heterozygosity (LOH) and comparative genomic hybridisation (CGH) studies as well as other techniques. LOH can be used to identify chromosomal regions where allelic losses are more frequent. In one study,¹⁰² a panel of 150 polymorphic microsatellite markers from throughout the whole genome was used to identify such regions. These data were correlated with clinicopathological features and showed four specific loci that correlated with lymph node metastasis (11q23–24, 13q12, 17p13.3 and 22q13). Microsatellites can also be used to check the ability of cells to repair DNA replication errors. Microsatellite instability (MSI) can be employed to identify both genetic and epigenetic modification. It has been shown that both genetic and epigenetic alterations of hMSH2 and hMLH1 contribute to genomic instability and tumour progression in sporadic breast cancer.

Chromosome-based CGH can be used to identify the loss of one or both copies of a given gene, as well as regions of amplification. In addition, the technique provides information on the number of copies of any part of each chromosome throughout the whole genome. There have been several studies using CGH in breast carcinoma,^103,¹⁰⁴ and those with clinical follow-up have led to the identification of genetic changes related to prognosis. However, the resolution of every chromosome-based CGH technique was limited to approximately 10 million base pairs (Mbp). This limitation makes it difficult to link copy number changes to the genes involved. Another limitation to this technique is the need to perform karyotyping for target identification in every experiment.

Therefore, array CGH has been developed to overcome the problems of chromosome-based CGH. Array CGH does not require a normal metaphase spread but rather an array of DNA fragments (100 bp to 100 kb) and their precise chromosomal locus. This approach can provide resolutions up to 1 Mb. Albertson et al. utilised this technique to map the recurrent breast cancer amplicon at chromosome 20q12.3. They were able to demonstrate that what was previously described as a single amplicon is in fact two distinct amplicons (ZNF217 and CYP24). A novel amplicon at 17q21.3 that is implicated in the amplification and overexpression of the HOXB7 gene in breast cancer has also been recently characterised using array CGH.¹⁰⁵

Assessment of gene expression in breast cancer

In gene expression profiling, each gene is usually represented by a single element (created with a complementary DNA (cDNA) or oligonucleotide for the gene studied) and with high-density oligonucleotide cDNA microarray technology many thousands of gene-specific mRNAs can be measured in parallel in a single tissue sample. The principle of gene expression cDNA array is comparable to array CGH, other than that it uses cDNA generated from mRNA. Similar to the two main types of expression microarrays (cDNA microarrays and oligonucleotide microarrays), serial analysis of gene expression (SAGE) can provide profiling of the transcriptome by taking a raw count of sequence tags, each representing a transcript in an RNA population with each one representing one gene.

Broadly speaking, molecular profiling of breast cancers by gene expression microarrays can be performed in one of two ways: unsupervised or supervised analysis. Unsupervised analysis is used mainly for partitioning tumour samples into groups or classes on the basis of gene expression profiles regardless of other features. The main objective of this approach is to determine whether discrete subsets of cancers can be defined on the basis of gene expression profiles, and whether new classes can be identified (class discovery) that may have clinical significance and to develop a new molecular taxonomy. Supervised analysis, on the other hand, is used to allocate tumours to specific groups based on clinical or pathological features (e.g. clinical outcome or response to therapy). There are two main types of supervised analysis: class comparison and class prediction. The former aims to identify the transcriptomic differences between two classes of tumours, whereas the development of a ‘gene signature’ is the ultimate goal of the latter.

In breast cancer, the class discovery studies were pioneered by the Stanford group,¹⁰⁶ which proved the principle that breast cancer could be classified into molecularly distinct groups based upon gene expression profiles and their similarity to normal cell counterparts. Multiple independent studies have confirmed and expanded the original results. According to these studies, breast cancer was classified into two main classes (ER-positive and ER-negative tumours) and each one can be classified into multiple molecularly distinct subclasses. For example, the ER-negative tumours encompass three subgroups, one overexpressing HER-2, one with tumours expressing genes characteristic of basal/myoepithelial cells (basal-like cancer), and one with a gene expression profile similar to normal breast tissue. cDNAs have also been used to distinguish cancers with BRCA1/BRCA2 mutations^107,¹⁰⁸ and to determine ER status,¹⁰⁹ lymph nodes status^110,¹¹¹ and prognostic subgroups in node-negative breast cancer.¹¹² These recent studies demonstrate the ability of cDNA to have a direct translational use in clinical practice.

Proteomics

The genetic code does not inform on which proteins are expressed by a cancer, or whether they are functional, if expressed. Post-translational modifications such as glycosylations or phosphorylation of proteins affect their expression and function at the protein level, which cannot be detected by assessing the status of the genome or transcriptome in a given cell population. Therefore, proteomics aims to assess changes at the protein level in which the proteome is displayed. Proteomics is quickly evolving to provide supportive and critical information to data generated through genomic approaches in high-throughput formats. This can be achieved by using different techniques such as two-dimensional polyacrylamide gel electrophoresis, isotope-coded affinity tagging, surface-enhanced laser desorption ionisation and matrix-associated laser desorption ionisation–time of flight.

These global expression profiling approaches yield candidate proteins and related genes that require verification through application of other techniques. Tissue microarrays (TMAs) provide a method for high-throughput protein expression analysis of large cohorts of archival samples that can be readily linked to clinicopathological and long-term follow-up databases. TMAs have expedited the validation of the prognostic and predictive significance of several candidate biomarkers. The technique allows for a composite slide of up to 1000 cores of tissue to be constructed into one paraffin block. The advantages are obvious in screening for novel protein expression. There are concerns on the validity of the cores as being representative samples and indeed this aspect has not been well studied, with only one study examining a large sample size with correlation to corresponding larger histology sections.¹⁰³ This technology and its application have been extensively reviewed.¹¹³

Clinical use of the high-throughput molecular techniques

It is envisaged that gene expression profiles may be able to guide decisions on the choice of hormonal, chemotherapeutic or targeted agents for each individual in the future. Presently, one of the best examples of the use of genomics and proteomics in clinical practice is the use of HER-2 expression/ amplification to select patients for trastuzumab. Such profiles are useful to identify prognostically significant genes, which studies have demonstrated can perform better than traditional markers.^104,¹¹⁴ In metastatic disease, expression patterns can be used to establish the origin of the metastatic lesion in patients with more than one primary. Transcription profiling has also been shown to differentiate accurately breast cancers with germ-line mutations in BRCA1/BRCA2 from those without such mutations. Such findings, if validated in other studies, open the way for molecular phenotyping in high-risk families. In addition, gene expression profiling has been used to develop expression predictors (class prediction studies) that can be used for many types of clinical management decisions, including risk assessment, diagnostic testing, prognostic stratification and treatment selection. There are two types of class prediction studies in breast cancer: prognostic and predictive class prediction. Prognostic class prediction includes (i) poor-prognosis gene signatures that can discriminate between a good and a poor outcome by comparison between highly aggressive and less aggressive primary tumours, and (ii) recurrence score gene signatures that define tumours based on the risk of disease relapse. The predictive class prediction includes predictors of response to therapy.^115,¹¹⁶

An example of a prognostic class prediction study is that reported by van’t Veer et al. Expression profiles of 117 primary breast cancers were compared with known prognostic factors and matched with 5 years of follow-up data. Twenty-five thousand genes were used to generate expression profiles, which separated the tumours into two groups. Group 1 developed distant metastasis in 34% and in group 2 70% developed distant metastases. From the 25 000-gene set, 70 genes had great accuracy in predicting recurrent disease. Multivariate analysis with grade, size, LVI, age and ER showed the poor-prognosis microarray profile to be an independent predictor of recurrent disease. This approach was further tested in 295 patients and again the use of gene profiling was able to accurately identify a poor-prognosis group.^101,¹⁰⁹ MammaPrint® assay, which is based on the Amsterdam 70-gene gene signature, is currently used as a molecular diagnostic test for breast cancer prognostication and prediction. Another predictor calculates a recurrence score on the basis of the expression of 21 known genes, with the use of RT–PCR in formalin-fixed, paraffin-embedded tissue. The ‘Oncotype DX assay’ has been developed by researchers at Genomic Health. Oncotype DX is currently used as a diagnostic test to quantify the likelihood of disease recurrence in women with early-stage breast cancer and assesses the likely benefit from tamoxifen and certain types of chemotherapy. Oncotype DX employs a mathematical algorithm called the recurrence score (range 1–100) to calculate continuous risk for relapse and death for patients receiving adjuvant tamoxifen, and has recently been shown in a large population-based case–control study to be an effective predictive test for ER-positive, node-negative breast cancer patients treated or untreated with tamoxifen and no chemotherapy.¹¹⁷ In addition, a number of microarray studies have been published which identified other prognostic signatures with clinical significance. For example, Change et al.¹¹⁸ have used a ‘wound-response gene expression signature’ to stratify breast cancer patients based on the hypothesis that features of the molecular programme of normal wound healing might play an important role in cancer metastasis. They found that this gene signature can improve the risk stratification of early breast cancer over that provided by standard clinicopathological features. A gene expression signature of hypoxia response, derived from studies of cultured mammary epithelial cells’ ‘hypoxia gene signature’, showed a strong predictor of clinical outcomes in breast cancer.¹¹⁹

Although proteomics has a long history that predates profiling at the RNA level, difficulties in protein identification and the lack of reproducibility of several assay platforms have limited the utility of such profiling systems. In breast cancer research, however, proteomics has begun to take on a role in the monitoring of response, prediction of resistance and relapse in patients treated with novel bio-directed therapies.¹²⁰

Clinical use of prognostic factors in patient management

Prognosis in breast cancer depends on the presence of spread of disease and on the inherent aggressiveness of the tumour. The latter depends on a number of intrinsic biological characteristics, some of which have already been evaluated, such as morphological features, growth rate and hormone responsiveness. Accurate prognostication is now required on an individual patient basis and this can only be achieved by using a Prognostic Index that includes both time-dependent and biological factors. The best way to obtain such an index is to take potential factors that have been shown to have some value in univariate analysis and submit them to multivariate analysis. This has been the approach in deriving the Nottingham Prognostic Index (NPI), which is based on three factors using the following formula:

Arbitrary cut-off points of 3.4 and 5.4 are used to divide patients into six prognostic groups: excellent (EPG), good (GPG), moderate (MPG) I and II, poor (PPG) and very poor (VPG) (Tables 2.1–2.3 , Figs 2.5 and 2.6).

Table 2.1

NPI showing percentage in each prognostic group

Prognostic groups: EPG, excellent prognosis; GPG, good prognosis; MPG, moderate prognosis; PPG, poor prognosis; VPG, very poor prognosis.

Table 2.2

Differences in 10-year survival for each prognostic group

Prognostic groups: EPG, excellent prognosis; GPG, good prognosis; MPG, moderate prognosis; PPG, poor prognosis; VPG, very poor prognosis.

CL, confidence limit.

Table 2.3

Relative risk reduction by NPI

Prognostic groups: EPG, excellent prognosis; GPG, good prognosis; MPG, moderate prognosis; PPG, poor prognosis; VPG, very poor prognosis.

ARR, absolute risk reduction; CL, confidence limit; RRR, relative risk reduction.

Figure 2.5 Survival by NPI (1980–86).

Figure 2.6 Survival by NPI (1990–99).

The NPI provides extremely powerful prognostic information within the NTPBCS and was able to demonstrate its utility and reproducibility in studies from other centres. In this respect Henson et al.,¹⁸ in a retrospective analysis of prognostic data in over 22 000 women from the SEER Programme of the National Cancer Institute in the USA, confirmed that a combination of lymph node stage and histological grade improved prediction of prognosis. In a similar way to the NTPBCS, Chevallier et al.¹²¹ have identified ‘young’ age, tumour size and histological grade as factors to be added to lymph node stage in the prediction of recurrence; these factors were used to divide lymph node-negative patients into three prognostic groups.

One of the strengths of the NPI is the fact that it has been verified prospectively in the NTPBCS. Further confirmation of its value has been provided by its validation in two large multicentre studies involving nearly 11 000 patients in total.^122,¹²³ Such studies demonstrate the inherent power of the pathological factors used in the NPI, which has become the most widely used index for the management of patients with breast cancer, certainly in the UK.

Other pathology-based prognostic indices used in breast cancer include Adjuvant!Online (AO)¹²⁴ and the St Gallen criteria,¹²⁵ and more recently Predict (a new UK prognostic model that predicts survival following surgery for invasive breast cancer).¹²⁶ This tool is a population-based validation of the prognostic model PREDICT for early breast cancer.

Adjuvant!Online (http://www.adjuvantonline.com) was developed in the USA and published in 2001; users can input information on a patient’s age, ER status, tumour grade, tumour size and number of positive nodes, and obtain predictions of 10-year OS (the likelihood of being alive 10 years after the diagnosis of breast cancer was first carried out), BCSS (the likelihood of not dying of breast cancer within 10 years of diagnosis) and event-free survival (EFS; the likelihood of surviving 10 years without recurrence (local, regional or distant), a second primary breast cancer, or death from breast cancer), both with and without any proposed adjuvant therapy. The performance of AO has been evaluated in cohorts of patients in Germany, Canada and the UK. It has an advantage over the NPI in that it integrates treatment, and for this reason AO is the programme most commonly used by oncologists.

The current St Gallen-derived algorithm for selection of adjuvant systemic therapy for early breast cancer patients includes tumour size and grade, nodal status, menopausal status, peritumoral vessel invasion, endocrine status and HER-2 status. Predict, which is similar in concept to AO but is based on UK patients, has been developed at the University of Cambridge. This tool is an alternative updated method of assessing probability of survival for UK patients.

All of the above discussion applies to primary operable breast cancer. Prognostic factors have also been explored in locally advanced and metastatic breast cancer, albeit not to the same extent. In a small study of locally advanced pancreatic cancer patients (n = 60) treated with either tamoxifen or radiation therapy and with crossover upon failure, response to therapy correlated significantly with histological grade (P = 0.02), ER (P = 0.02), PR status (P = 0.02), mitotic index (P = 0.01) and tumour ploidy (P = 0.04). Survival from initial therapy correlated significantly with ER (P = 0.01) and PR status (P = 0.04).¹²⁷ In these patients, histological grade, mitotic index, tumour ploidy, and ER and PR status of the primary tumour may predict response and prognosis.

These authors also explored the prognostic value of certain factors as a composite index to guide decision-making in metastatic disease.¹²⁸ The advanced breast cancer (ABC) index comprises the grade, ER status, site of initial metastasis and disease-free interval. This index was tested both retrospectively and prospectively and shown to separate groups where the survival between each was highly significant (P < 0.001).