2ª Reunión Científica Dako. II-TRADITIONAL MORPHOLOGICAL FACTORS

Until relatively recently the main role of the diagnostic histopathologist lay in the establishment of a diagnosis of breast cancer from excision biopsy or frozen section. Apart from the examination of loco-regional lymph nodes for the presence or absence of metastases it was unusual for any other prognostic information to be supplied, or indeed requested. The treatment of breast cancer was standardised, predominantly surgical and there was little attempt to stratify patients for appropriate therapy on an individual basis. However, as mentioned previously, in the last two decades the treatment of breast cancer has undergone dramatic changes and a much wider range of both local and systemic therapeutic options is now available. Early diagnosis, especially since the advent of mammographic breast screening, is detecting tumours which are likely to have a favourable outcome and it has become extremely important to assess prognosis for each patient before a therapeutic plan is agreed.

A considerable amount of useful prognostic information is available from the careful histopathological examination of routine breast carcinoma specimens.

The following factors, all relatively simple to assess, have been shown to provide clinically relevant prognostic information, provided that careful attention is paid to diagnostic guidelines and protocols.

For correlation with prognosis the size of tumours should only be assessed on pathological specimens, as clinical measurement is notoriously inaccurate. If an estimate of clinical tumour size is required for therapeutic planning purposes it should be checked by an ultrasonic measurement.

We recommend that the tumour diameter is measured in three planes to the nearest millimetre, initially in the fresh state. These measurements are then re-checked after fixation and the greatest diameter is taken as the tumour size. If there is any doubt about the size measurement tumour extent should be checked on histological sections using the stage micrometer. This is particularly appropriate for small tumours, especially those measuring 1 cm or less, and for tumours with a large in-situ component. Examples of the measurement of tumour size in a variety of circumstances are given in Fig 2.1.

As a time dependent factor tumour size has been shown consistently in many studies to influence prognosis,

patients with smaller tumours having a better long term survival than those with larger tumours. The significant correlations found by Elston et al,

Fisher et al

and Neville et al

are of particular interest since the initial pathological assessments in these large multicentre trials, including measurement of tumours size, were all carried out by local participating pathologists rather than central pathologists. This emphasises the inherent strength of tumour size as a prognostic factor.

The estimation of tumour size has assumed particular importance in breast screening. The term 'minimal breast cancer' (MBC) was originally introduced to delineate certain forms of breast cancer which carried an exceedingly good prognosis.

MBC included all cases of in situ carcinoma (ductal and lobular) and invasive carcinomas measuring 5 mm or less. Subsequently, for no clearly defined reason the invasive component has been re-defined by various different groups. The Breast Cancer Detection Demonstration Projects

and the American Cancer Society

have used 9 mm or less as the maximum diameter, whilst the American College of Surgeons

favour up to and including 10 mm.

This lack of uniformity in definition causes problems in the interpretation of data from different studies, but there is little doubt that minimal invasive carcinomas (MIC) are of an earlier stage than tumours which measure more than 10 mm in diameter. In most series the frequency of axillary lymph node metastasis in MIC is 15-20%,

compared with over 40% in tumours measuring 15 mm or more.

Even more favourable results are obtained in women with breast cancer detected during the prevalent round of breast screening, axillary node metastasis ranging from 0-15%.

Surprisingly it is difficult to obtain accurate data on the relationship between MIC and prognosis but, as expected, survival appears to be better than for patients with larger tumours. For example, in the long term study from the Memorial-Sloan Kettering Cancer Centre

the projected relapse-free survival rates 20 years after initial treatment were as follows:

However, our own studies from the Nottingham Tenovus Primary Breast Cancer Study (NTPBCS) suggest that the cut of point of 10 mm is not necessarily the best discriminator for MIC (Fig 2.2). Life table analysis shows that the more significant size is 15 mm. The majority of the patients in this study are from symptomatic practice and it will be important to ascertain whether screening-detected tumours have the same pattern of prognosis.

It is clear, therefore, that tumour size is a valuable prognostic factor, and it has also become an important quality assurance measure for breast screening programmes,

used in part to judge the ability of radiologists to detect impalpable invasive carcinomas on mammography. For example, in the United Kingdom National Health Service Breast Screening Programme (NHSBSP) it is required of radiologists that 50% of the invasive cancers detected must measure less than 15 mm.

It is, therefore, incumbent upon pathologists to measure tumour diameter as accurately as possible. As size decreases so the risk of errors in measurement increases, and marked inconsistencies have been reported.

Figure 2.2 Relationship between pathological tumour size and survival in 1763 patients with primary operable breast cancer.

Until comparatively recently, the histological classification of carcinoma of the breast was restricted to a main sub division into in situ and invasive carcinoma. The majority of invasive carcinomas were designated simply as scirrhous adenocarcinomas. It is now recognised that invasive carcinomas may be further sub divided morphologically according to their degree of differentiation. This is achieved in two ways, by assessing histological type and histological grade.

It is now accepted that a wide range of morphological patterns may be recognised in invasive carcinoma of the breast

and that histological type provides useful prognostic information.

The diagnostic criteria have been described in some detail previously

and will not be repeated here. However, it must be appreciated that a considerable subjective element remains and there is not yet universal agreement for all types. This is reflected in the relative proportions of different types in published series (Table 2.1). Furthermore, in the NHSBSP pathology quality assurance scheme consistency of diagnosis of histological type was disappointingly low,17 implying that pathologists need to work to agreed diagnostic protocols.

The favourable prognosis of certain histological types of invasive carcinoma of the breast is well established. Thus tubular carcinoma,

mucinous carcinoma,

invasive cribriform carcinoma,

medullary carcinoma,

infiltrating lobular carcinoma

and tubulolobular carcinoma

have all been reported to have a more favourable prognosis than invasive carcinomas of no specific type (ductal NST). Very few comprehensive long term follow up studies, relating histological type to survival, have been carried out. Dawson and colleagues

found a relative excess of tubular, mucinous, medullary and infiltrating lobular carcinomas in patients who had survived at least 25 years after mastectomy, compared with those having a survival of less than 10 years. These findings were confirmed in a similar study from Edinburgh

with the addition of papillary and invasive cribriform carcinomas amongst the long term survivors. The Edinburgh group have also found a relative excess of these 'special type' or 'specific type' tumours in cases of invasive carcinoma detected in the prevalent round of mammographic breast screening,

and this has been confirmed in our own studies.

We have obtained further objective evidence that histological type can provide powerful prognostic information from the NTPBCS.

In a series comprising over 1500 patients with primary operable invasive carcinoma followed up for a minimum of 10 years the excellent prognosis of the special types, pure tubular, invasive cribriform and mucinous was confirmed (Fig 2.3). The study also showed that the categories of tubular mixed carcinoma and mixed ductal NST and special type are worth recording, since they carry a good prognosis, considerably better than ductal NST carcinoma (Fig 2.4). In previous studies such mixed types were rarely recognised and the tumours included within the general category of ductal NST. Since they account for at least 15 per cent of cases in a symptomatic series, valuable prognostic information is lost if they are overlooked.

It has become accepted dogma that medullary carcinoma carries an excellent or good prognosis.

It is interesting that this view has persisted despite the fact the other studies have been unable to confirm a survival advantage for medullary carcinoma compared with ductal NST carcinomas.

However, some of the latter studies have shown that medullary carcinoma does have a better prognosis than ductal NST carcinomas of grade 3.

Ellis and colleagues

have therefore concluded that medullary carcinoma should be regarded as having a moderate rather than a good prognosis.

Overall, infiltrating lobular carcinoma conveys a slightly better prognosis than ductal NST carcinoma,

although the 10 year survival of 54% in the latter study clearly implies no more than a moderate prognostic category. However, Dixon and colleagues

found significant differences in survival between the morphological subtypes of lobular carcinoma, and this has been confirmed in our own studies.

Thus the classical type carries a good prognosis (60% 10 year survival), mixed lobular type an average prognosis (55% at 10 years) and the solid lobular type has a poor prognosis (40% at 10 years). Tubulolobular carcinoma which has an excellent prognosis (over 9% 10 year survival) is currently considered as a separate and distinct type because of the lack of agreement concerning its assignment as a tubular or lobular variant.

Patients with invasive carcinoma of the breast can therefore be stratified into broad prognostic groups according to their histological type. The excellent prognosis group comprises the special types (tubular, cribriform, mucinous) and tubulolobular carcinoma, the good group tubular mixed, mixed ductal NST/special type and classical lobular carcinoma, the average group mixed lobular, medullary and atypical medullary carcinoma and the poor group is composed of ductal NST, mixed ductal and lobular and solid lobular carcinoma.

In addition histological typing of breast cancer adds to our understanding of the biology of breast cancer. For example, infiltrating lobular carcinomas show estrogen receptor (ER) expression more frequently than ductal NST carcinomas

and they also have a different pattern of metastatic spread with a predeliction for unusual sites such as the retroperitoneum and serosal surfaces.

There is now very interesting evidence emerging that there is a correlation between histological type and expression of the BRCA1 and BRCA2 genes. For example, in two important studies there appears to be an excess of high grade ductal NST carcinomas with medullary features in BRCA1 related cases in comparison with BRCA2 cases.

It is to be hoped that further clarification of these genetic associations will be provided by more comprehensive large scale studies.

Table. 2.1 Comparison of relative percentage of main morphological types of invasive breast cancer in different published series.

Figure 2.3 Suvirval curves for patients with 'special type' invasive carcinomas of the breast compared with that for patients with ductal/NST carcinoma.

Figure 2.4 Survival curves for patients with tubular mixed carcinomas compared with that for patients with ductal NST carcinoma.

One of the most fundamental aspects of oncological pathology, which has undoubtedly stood the test of time, has been the recognition that the detailed morphological structure of tumours can be correlated with their degree of malignancy. Nearly 70 years ago Greenhough,

in Boston, USA undertook the first formal study of the grading of histological differentiation in breast cancer. He assessed 8 morphological features in a rather subjective way, but showed a good correlation with so-called 'cure', although the latter was not defined properly. Much credit should also go to Scarff and his colleagues at the Middlesex Hospital in London who re-examined Greenhough's method and decided that only three factors, tubule formation, nuclear pleomorphism and hyperchromatism were of importance.

Scarff's method has formed the basis of all subsequent grading systems, whether they use multiple cellular factors

or nuclear factors.

It is remarkable that given the diversity of methods employed very many studies have demonstrated a significant association between grade and survival indicating the powerful prognostic information provided; this data has been reviewed in some detail previously.

Despite this evidence, and the adoption of one of the methods, the so-called Scarff-Bloom-Richardson method by the World Health Organization,

acceptance of histological grade into routine diagnostic histopathological practice has been slow until relatively recently. In the past this was due in part to lack of clinical demand, as mentioned previously. A further reason for the reluctance to rely on histological grading is the subjective nature of previously published methods and a perceived poor reproducibility and consistency

despite the fact that a substantial number of studies have reported acceptable levels of inter and intra observer variability.

These conflicting views have highlighted the need for grading to be carried out by trained histopathologists who work to an agreed protocol. Given the nature of the methods, assessment of histological differentiation will always carry an underlying subjective element, but one of the fundamental problems with many of the systems used in previous studies has been the lack of strictly defined written criteria. Bloom and Richardson

made a useful contribution by adding numerical scoring to the method described by Patey and Scarff

but did not provided clear criteria for their cut off points. In Nottingham we have employed further modifications to the above methods in order to introduce greater objectivity.

Three characteristics of the tumour are evaluated, tubule formation as an expression of glandular differentiation, nuclear pleomorphism and mitotic counts (Table 2.2).

Table 2.2 Summary of semiquantitative method for assessing histological grade in breast carcinoma.

Feature

Score

Tubule formation

Majority of tumour (>75%)

1

Moderate degree (10-75%)

2

Little or none (<10%)

3

Nuclear pleomorphism

Small, regular uniform cells

1

Moderate increase in size and variability

2

Marked variation

3

Mitotic counts

Dependent on microscope field area

1-3

(see Table 2.3 and Fig. 2.5)

A numerical scoring system on a scale of 1-3 is used to ensure that each factor is assessed individually. In evaluating tubules only structures exhibiting clear central lumina are counted; cut off points of 75% and 10% of tumour area are used to allocate the points. In mucinous carcinoma lumina with tumour cell islands are used as a surrogate for tubules and this also applies to the cribriform structures in invasive cribriform carcinoma. Nuclear pleomorphism is assessed by reference to the regularity of nuclear size and shape of normal epithelial cells in adjacent breast tissue. Increasing irregularity of nuclear outlines and the number and size of nucleoli are useful additional features in allocating points for pleomorphism. The most important modification concerns the evaluation of mitotic figures. Qualitatively care must be taken to count only clearly defined mitotic figures; hyperchromatic and pyknotic nuclei are ignored since they are more likely to represent apoptosis than proliferation. Quantitatively a more accurate assessment is required than designations such as 'about 2 or 3 mitoses per high power field' (HPF)

since the area of a single HPF may vary by as much as sixfold from one microscope to another.

However, although estimation of mitotic index

is the most accurate way of counting mitoses it is laborious and unlikely to be of practical use in the routine laboratory. We have, therefore, compromised and standardised mitotic counts to a fixed field area (Table 2.3).

Table 2.3 Assignment of points for mitotic counts according to the field area, using several microscopes.

Microscope

Leitz

Ortholux

Nikon

Labophot

Leitz

Diaplan

Objective

x25

x40

x40

Field diameter (mm)

0.59

0.44

0.63

Field area (mm2)

0.274

0.152

0.312

Mitotic count*

1 point

0-9

0-5

0-11

2 points

10-19

6-10

12-22

3 points

>20

>11

>23

Using this system any microscope can be calibrated to obtain reproducible and comparable data. Alternatively, mitoses may be assessed using a grid system.

The total number of mitoses per 10 high power fields is counted and the point allocation plotted according to the measured field diameter (Fig 2.5). To our knowledge, in only one other method, that of Contesso et al

is reference made to the expression of mitotic counts per defined field area. Unfortunately, their method for counting mitoses is somewhat idiosyncratic and no attempt is made to quantify tubular structures or increase precision in the estimation of pleomorphism.

The Nottingham method described above has been fully evaluated in the NTPBCS. The results, based on life table analysis of over 1800 patients followed-up for between 3 and 18 years, confirm conclusively the highly significant relationship between histological grade and prognosis; survival worsens with increasing grade (Fig 2.6).

The method has now been shown to have good reproducibility in other centres

and it has been adopted for use in the pathological data set of the United Kingdom NHSBSP,

by the European Breast Screening Pathology Group

and the Association of Directors of Anatomic and Surgical Pathology in the United States.

In Nottingham histological grading is carried out in all cases of invasive carcinoma of breast, regardless of morphological type and this view has been accepted by the United Kingdom NHSBSP.

This practice has been criticised by some pathologists who feel that grading is not appropriate for the special histological types such as pure tubular, invasive cribriform, mucinous, medullary and infiltrating lobular carcinomas. However, we have found that when grade is analysed in tumours of particular histological types it is usually found to be appropriate.

For example, most infiltrating lobular carcinomas, especially those of classical subtype, are designated as grade 2 and the overall survival curve of lobular carcinoma overlies that of all other types of grade 2 carcinoma. A minority fall into the grade 1 or grade 3 category and for these cases the survival curves show an appropriate and significant separation (Fig 2.7). Furthermore, we have shown that for some special types such as mucinous and tubular mixed carcinoma, grading provides a more appropriate estimation of prognosis than type alone.

Histological grade is therefore a very powerful prognostic factor in breast cancer.

Figure 2.6 Relatioship between histological grade and survival in 1830 patients with primary operable carcinoma of breast.

It has long been recognised that involvement of loco-regional lymph nodes in breast cancer is one of the most important prognostic factors. It is now generally accepted that clinical assessment of lymph node status is not sufficiently accurate for therapeutic use, and that evaluation of lymph node stage should be based on histological examination of excised nodes.

Numerous studies have shown that patients who have histologically confirmed loco-regional lymph node involvement have a significantly poorer prognosis than those without nodal involvement.

Overall 10 year survival is reduced from 75% for node negative patients to 25-30% for node positive patients. Prognosis also related to the number and level of loco-regional lymph nodes involved; the greater the number of nodes involved the poorer the patient survival.

In the United States the NSABP divides patients into two groups for therapeutic purposes, those with 1-3 positive nodes and those with 4 or more positive. Similarly, involvement ofnodes in the 'higher' levels of the axilla, and specifically the apex, carry a worse prognosis

as does involvement of the internal mammary nodes.

In the NTPBCS we have demonstrated that highly significant prognostic information can be obtained by a lymph node sampling method with examination of a node from the low axilla, apex of axilla and second intercostal space (Fig 2.8).

In recent years there has been considerable debate regarding the extent of axillary lymph node dissection with arguments in favour of both axillary sampling and axillary clearance.

It has been argued that a minimum of 10 nodes should be obtained before deeming a patient node negative

but this is disputed by others.

A greater number of nodes can be obtained at clearance compared with sampling, but the price for the additional prognostic information is the greater post operative morbidity, including reduced shoulder mobility and chronic lymphoedema. In practical terms we believe that a sensible compromise between the two methods, clearance and sampling, should be employed. Internal mammary sampling only provides useful information in medially sited tumours and need not be performed for lateral tumours. Low axillary clearance, carried out below the level of the intercosto-brachial nerve, produces enough lymph nodes (usually between four and 15) for accurate prognostication, with minimal morbidity. Apical node biopsy is additive, but to reduce the need for an additional incision, should only be performed when the primary operation is mastectomy.

A further refinement of lymph node sampling may be provided by the recently developed technique of sentinel node biopsy. The sentinel node is the first lymph node encountered by lymphatics draining from tissues around a tumour. The concept was first introduced in 1977 by Cabanas

in relation to the lymphatic drainage of penile cancer. The principle of sentinel node biopsy rests on the theory that if metastatic spread has occurred it will first involve the sentinel node and biopsy of this node will be an accurate determinant of stage.

Figure 2.7 Correlation between histological grade and overall survival in 341 patients with infiltrating lobular carcinoma. x² - 21.5, 2 d.f.: p<0.001

Sentinel node biopsy was first introduced clinically in 1992 in cases of malignant melanoma.

Blue dye was injected around the melanoma pre-operatively. At operation the dye was traced to the regional lymph nodes and a 'blue' sentinel node excised. In a large series of patients the sentinel node was identified in 82% and accurate staging was provided in 99% of those in whom it was identified. Lymphoscintigraphic techniques have since been introduced to improve the sentinel node detection rate. This involves the injection of 99m-Tc labelled colloid around the tumour instead of blue dye; portable gamma detecting probes can then be used intra-operatively to detect a specific 'hot' (sentinel) node. This radio-isotope technique was also applied successfully to patients with malignant melanoma.

In breast cancer the first reported study used the blue dye method; sentinel nodes were identified in 65% of patients which staged the axilla accurately in 96% of that cohort.

In a subsequent report by the same group it was demonstrated that a learning curve exists for the technique; the detection rate and sensitivity improved in this second series to 93% and 100% respectively.

Sentinel node biopsy has now been assessed in numerous studies. Detection rates average 95% and axillary nodal status is predicted accurately in over 95% of these. It appears that detection rates are improved if a combination of blue dye and radiolabelled colloid is used.

These data suggest that sentinel node biopsy can be used to stage the axilla in primary breast carcinoma. The value of the technique in clinical practice will depend on the false negative rate and the incidence of regional recurrence after sentinel node biopsy alone. Since patients with positive sentinel nodes can be selected for full axillary clearance the clinical utility will also depend on the ability of the technique to determine histological involvement of the node pre, or intra-operatively.

A number of different techniques have been advocated including frozen section, imprint cytology and immunohistochemistry; all are time consuming and labour intensive and their value has yet to be established conclusively

Prospective clinical trials, which are required to address these issues, are currently being planned. For a fuller account of the pathological aspects of sentinel node biopsy the reader is referred to recent review article by Anderson (1999) and Lee et al (2000).

The significance of the presence of metastatic carcinoma in the adipose tissue surrounding axillary lymph nodes, the so-called extranodal spread or extracapsular metastasis (ECM), is uncertain with conflicting data. It was suggested by Mambo and Gallager,

in a retrospective analysis, that this feature conveyed a poor prognosis in patients with up to three nodes involved, but not in those with four or more nodes involved. Similar results were obtained by Cascinelli et al

in a study of mastectomy without adjuvant radiotherapy or chemotherapy except

Figure 2.8 Relationship between lymph node stage (A - no node involvement, B - low node involvement, C - high node involvement) and survival in 693 patients with primary operable carcinoma of the breast.

that the effect of extranodal spread on recurrence rate was seen in patients with two or more nodes affected, irrespective of the number of nodes involved. Fisher et al

have demonstrated in a prospective study that extranodal spread occurs significantly more frequently in patients who have four or more nodes involved by metastatic tumour, and although such patients were more likely to suffer short term relapse they could not demonstrate that this effect was independent of nodal status. This view supports that of Hartveit

who found that extranodal spread had no intrinsic prognostic significance and concluded that the presence of tumour cells in efferent vessels was the only indicator of poorer prognosis in patients with involved nodes. More recently Donegan and colleagues

have confirmed the association between ECM and the number of nodes involved, but found that there was no significant influence on prognosis. They concluded that ECM was not an indicator per se for irradiation after complete axillary clearance.

There has also been debate on the actual size of lymph node metastases, and in particular the question of so-called occult metastases. Several studies have shown that the presence of 'micrometastases' measuring 2 mm or less does not affect survival adversely, compared with that of node negative patients.

The data concerning the detection of tiny deposits, often single cells, by serial sectioning or immunostaining is conflicting. Some groups have found a significantly worse prognosis for such patients,

whilst others could detect no difference in survival.

Hartveit and Lilleng

have suggested that the site of the deposits is the most important factor; in their study survival of patients with subcapsular deposits was the same as that of node negative patients, but deposits within the lymphoid tissue conveyed the sample prognosis as node positive patients. Further studies are clearly required.

In routine practice it is inappropriate to examine serial sections in every case because of the workload implications, and a sensible compromise is necessary. If nodes are obviously involved on gross examination a single confirmatory section is sufficient. Nodes measuring less than 5 mm in length may be processed in groups and need only be cut at two levels. Nodes greater than 5 mm in length should be sliced at intervals at right angles to the long axis, multiple blocks up to 4 in number being taken according to the overall node size. Immunostaining should be reserved for the small number of cases in which the morphological appearances are suspicious, but not diagnostic of metastatic carcinoma.

The prognostic value of the estimation of vascular invasion in breast cancer is disputed.

Some studies have found no significant correlation with clinical outcome

whilst others have shown that the presence of vascular invasion predicts for both recurrence

and long term survival.

One explanation for such discrepancies may be the wide variation in the reported frequency of vascular invasion (20-54%) and the related problem of the distinction of true vessels, especially lymphatics, from artefactual soft tissue spaces.

Although muscular blood vessels are occasionally involved, tumour emboli are usually identified within thin walled vascular channels. It is impossible to determine whether such spaces are lymphatics, capillaries or venules and for this reason we believe that vascular permeation should be left unspecified and the broad term 'vascular invasion' used. In order to avoid overdiagnosis care must be taken to avoid misinterpretation of both ductal carcinoma in situ and shrinkage artefact associated with cords of invasive tumour as vessel invasion.

These problems are greatly reduced by good fixation, as emphasised in the section on specimen preparation earlier, and by working to a simple but strict protocol. Vascular invasion should only be assessed in the breast tissue surrounding tumour and not within it. Tumour emboli must be present within spaces having a complete lining of endothelial cells; these spaces are often in close proximity to small muscular blood vessels and may by separated from the main tumour by normal lobular units, topographical patterns emphasised by Orbo et al (Fig 2.10).

Immunostaining for endothelial markers such as laminin, type IV collagen, Factor VIII related antigen and Ulex Europeus agglutinin I is not helpful in distinguishing vessels from duct structures, but may be useful in excluding shrinkage artefact.

In routine practice their use should be confined to the resolution of equivocal cases. Reproducibility of the evaluation of vascular invasion has been shown to be satisfactory,

and even when this was questioned complete agreement was obtained in over 85% of cases.

Vascular invasion correlates very closely with loco-regional lymph node involvement,

and possibly because of this association it has been claimed that it can provide prognostic information as powerful as lymph node stage.

There is certainly a correlation between the presence of vascular invasion and early recurrence in lymph node negative patients

and Roses et al

have shown that the adverse prognostic effects are also independent of occult axillary node involvement. In the NTPBCS we have confirmed the prognostic value of vascular invasion in relation to long term survival (Fig 2.11) but, like Roses et al,

we have also demonstrated that this effect is independent of lymph node stage, using multivariate analysis.

However, at the present time it appears that the most important application for the assessment of vascular invasion lies in its power as a predictor of local recurrence following conservation therapy.

VI also predicts significantly for the risk of flap recurrence after mastectomy.

Figure 2.11 Relationship between vascular invasion and survival in 1623 patients with primary operable carcinoma of the breast.

A number of other morphological features of breast carcinoma have been proposed as prognostic factors, but are of relatively less importance than those discussed above.

There is no doubt that angiogenesis is an important factor in the growth and metastatic potential of carcinomas.

It has therefore been suggested that tumours showing a high level of new vessel formation would have a poorer prognosis than those with relatively little angiogenesis. This association appears to have been demonstrated for a number of tumours including breast carcinoma.

However, other studies have failed to identify such an association

and our own experience is similar to this second group of investigators.

We have assessed vascular density in breast carcinomas using both random field selection and pre-selection of the perceived area of highest vascularity, the so-called 'hot spot' using immunostaining for expression of two endothelial markers, CD 34 and CD 31. Weidner

has commented that others have not used his precise technique which involves use of the antibody to factor VIII and analysis of the 'hot spot' regions. These techniques were used by Costello and colleagues,

but they were unable to demonstrate any relationship between vessel density and clinical outcome. From these latter studies it appears that there are major difficulties in reproducibility of the measurement of new vessel formation in breast carcinomas and it is our view that the performance of vascular counts on routinely selected blocks of tumour tissue offers no advantage over the traditional pathological prognostic factors.

Tumour necrosis is a relatively common phenomenon, occasionally visible macroscopically as a sharply demarcated area of dullness, usually in a central position. Microscopically necrotic tumour is characterised, as in any tissue, by the nuclear changes of karyorrhexis, pyknosis and karyolysis, with a change to granular eosinophilic cytoplasmic degeneration. When tumour necrosis has been present for a sufficient length of time it may be accompanied by replacement fibrosis. Necrosis is almost entirely confined to ductal NST carcinomas and appears to occur most frequently in those of high grade.

The prognostic value of tumour necrosis has been evaluated in several studies and its presence has been equated with decreased survival and early treatment failure.

Unfortunately, in none of these studies is a precise definition given of terms used, such as 'extensive' necrosis, which limits their value. Parham and colleagues have proposed a new 'simplified' method for grading breast cancer by combining tumour necrosis and mitotic counts.

In summary there is some published evidence which suggests that the presence of tumour necrosis may be a poor prognostic feature. However, before tumour necrosis is accepted as a useful prognostic factor reproducible criteria for the definition of necrosis and evaluation of its extent must be devised. Tumour necrosis must also be tested in multivariate analysis against other prognostic factors, especially histological grade with which it appears to be closely associated.

Stromal fibrosis is found frequently in invasive carcinoma of the breast, but in varying amounts.

The prognostic significance is uncertain, and stromal fibrosis has been associated with a favourable prognosis,

poorer survival

and to have no effect on survival.

In this respect tumour type is a confounding factor, since extensive fibrosis may be found both in low grade tumours such as tubular carcinomas and in high grade ductal NST carcinomas. For this reason it is unlikely that stromal fibrosis will provide useful prognostic information.

Stromal elastosis is also a feature of many breast carcinomas, distributed either in a periductal or a diffuse pattern.

As with stromal fibrosis there are conflicting data on the prognostic significance of elastosis. Some studies have suggested that its presence is associated with a good prognosis

but this has not been confirmed by others.

Giri et al

found that central elastosis had greater clinical significance, but this was based on a small study with only short-term follow-up. Elastosis is particularly associated with tumours having a relatively good prognosis (eg tubular, tubular mixed, invasive cribriform);

this suggests that stromal elastosis is not an independent prognostic factor.

The amount of ductal carcinoma in situ (DCIS) associated with invasive carcinomas of the breast is extremely variable, and the assessment of its extent is highly subjective. Some groups have suggested that the presence of prominent DCIS within invasive carcinomas conveys a better prognosis and a decreased frequency of nodal metastases.

However, it has been suggested that the DCIS component in invasive carcinomas may be of greater importance in the management of patients considered for conservation therapy. At an EORTC meeting in 1989 it was concluded that 'the principal risk factor for breast relapse after breast conserving treatment is large residual burden and the main source of this burden is an extensive in situ component (EIC)'.

This statement was based on data from a number of studies

but a subsequent publication from the Boston group has cast some doubt on its validity.

They found that assessment of excision margins was by far the most powerful factor influencing local recurrence rates and that EIC was not a predictive factor if complete excision was obtained. EIC did, however, predict for the likelihood that margins would be involved and it therefore has some value in this respect.