2ª Reunión Científica Dako. V - PROGNOSTIC FACTORS IN PATIENT MANAGEMENT

All the factors described in the previous section have been shown to convey prognostic information to a greater or lesser degree, and several provide a powerful indication of the likely outcome for particular groups of patients. However, no consensus has been reached on which factors should be used routinely in clinical practice, and Hawkins

in a review of a number of recent articles found a staggering range of findings and conclusions. He reiterated the proposal made by McGuire and Clark

that the following guidelines should be utilised in deciding what constitutes a useful prognostic factor:

Missing from this list, perhaps because it was felt to be too obvious to need re-stating, is the most important point of all, namely clinical relevance.

The majority of the factors which we have discussed can be assessed in a routine diagnostic histopathology laboratory and are therefore readily available for clinical management. Histopathologists are used to supplying such information in their routine reports (eg Dukes's staging of-rectal carcinoma, measurement of Breslow thickness and Clark's levels in malignant melanoma), but to avoid wasted effort they need to agree with their clinical colleagues which prognostic factors should be reported. There is no point in histopathologists writing long and elegant reports full of data on the latest prognostic factor if the clinicians have no intention of using the information to plan or stratify therapy for an individual patient.

In breast cancer, until comparatively recently, there has been a depressing lack of interest in the use of prognostic factors in patient management. Indeed, the only factor used consistently in most centres as a guide for therapy has been loco-regional lymph node status and this has also been the case for patient stratification in clinical trials. Lymph node status is a time-dependent prognostic factor - the longer the tumour has been growing the more likely it is that spread to lymph nodes will have occurred. Taken alone, lymph node stage, although a powerful factor, is incapable of defining either a 'cured' group of patients or a group with a close to 100% mortality from breast cancer.

Prognosis in breast cancer depends not only upon the presence of distant metastases, but also on the aggressiveness or virulence of the tumour. The virulence of a tumour depends on a number of intrinsic biological characteristics, some of which can already be evaluated, such as morphological features, growth rate, hormone responsiveness and some which are yet to be measurable such as invasiveness or power of tissue destruction.

If accurate prognostication is required on an individual patient basis then a Prognostic Index is required which uses both Time-dependent Factors and Biological Factors. This is not a new idea, but, as indicated above, one that has been neglected by clinicians until recently. Indeed, when Greenhough

first introduced histological grading 70 years ago he noted that, even in his small series, the combination of high grade malignancy and nodal involvement gave an exceedingly poor prognosis. When Bloom

revived interest in the grading method devised by Patey and Scarff

he too stressed that prediction of survival was improved by combining grade with lymph node stage. Thus the 5-year survival of 94% for patients with grade 1 tumours and uninvolved axillary nodes fell to 65% for those with involved nodes and from 55% to 16% in grade 3 tumours. Similar findings were later reported in the multicentre Cancer Research Campaign trial.

Observations such as these were largely derived from studies of sub-groups using univariate analyses. The Nottingham Tenovus Primary Breast Cancer Study (NTPBCS) was established in 1973 specifically to investigate a wide range of potential prognostic factors. All patients with primary operable breast cancer (clinical size <5 cm) presenting to a single surgical team (Professor R W Blamey) are entered into the study. Initial surgical treatment includes simple or subcutaneous mastectomy or wide local excision and post-operative radiotherapy, together with node sampling. To date over 3000 patients have been entered into the study which has been used both to derive and to test a prognostic index based on multiple factors, the Nottingham Prognostic Index.

From the start of the study prognostic data has been accumulated both prospectively and retrospectively. Basic prospective data has included age at diagnosis, menopausal status (based on menstrual history and checked by FSH levels), tumour size (measured pathologically as described previously), histological grade (assessed by the method described by Elston and Ellis

), oestrogen receptor status (initially using the dextran coated charcoal method (DCC) and latterly using the ELISA and ERICA monoclonal antibody methods) and lymph node stage. The latter is divided into 3 groups, based on histological examination, as follows:

Stage A

No node involvement

Stage B

Involvement of up to 3 low axillary nodes or internal mammary node (for medial tumours)

Stage C

Involvement of four or more low axillary nodes and/or the apical node or low axillary and internal mammary node simultaneously.

In a preliminary study based on this data an initial group of patients with a particularly poor prognosis was identified.

It was found that 85% of patients with tumours >2 cm, grades 2 or 3 and of lymph node stage C had suffered a major recurrence or died within 18 months of diagnosis. However, this index lacked sensitivity since it only identified 50% of patients having a very poor prognosis. Accordingly, in 1982, a retrospective multivariate analysis was carried out of 9 separate factors studied in 387 patients.

Although a number of factors were related to survival in univariate analysis only three remained significant in multivariate analysis, pathological tumour size, histological grade and lymph node stage (Table 5.1).

Table 5.1 Cox's multivariate analysis: Data from 387 patients with primary operable breast cancer 1976 - 1981. Results used to derive the Nottingham Prognostic Index.

Factors	B coefficient	Z value
Menopausal status	0.5	1.5
Tumour size	0.17	2.92*
Histological grade	0.82	4.56*
Lymph node stage	0.76	5.29*
ER content	-0.34	-1.72

The B coefficients in the multivariate analysis show the contribution of each factor to the estimation of survival. Thus, using the B values for weighting an index predicting survival was calculated:

Nottingham Prognostic Index (NPI) = Size (cm) x 0.2 + Stage (lymph node, 1-3 by level) + Grade (1-3: well, moderate or poor differentiation).

Curves of survival by life table analysis methods showed excellent separation of patient groups, depending on the index level, but since the index had been derived from these patients this was a self-fulfilling prophecy. The index was therefore tested prospectively in a further 320 patients and this study confirmed that the data derived from one group of patients, could be applied successfully to another entirely separate group.

Figure 5.1 shows the analysis based on all of the first 1989 patients entered into the NTPBCS, with up to 15 year follow-up. We have identified 3 groups of patients by employing (arbitrary) cut off points of <3.4 for the Good group, 3.41 - 5.4 for the Moderate group and >5.41 for the Poor group; the percentages of patients falling into each group in symptomatic practice, and their predicted 15 year survival are shown in Table 5.2.

Table 5.2 NPI groups for 1989 patients showing the numbers and percentages in each group, the expected 15 year survival and for comparison the expected survival for age-matched females without breast cancer.

NPI	n	%	15 year survival
Aged-matchedfemales			83%
GPG (<3.4)	635	32	80%
MPG (3.41-5.4)	1040	52	42%
PPG (>5.41)	316	16	13%

Whilst it is clear that the NPI provides extremely powerful prognostic information within the NTPBCS it is important to demonstrate its utility and reproducibility in studies from other centres, especially in view of the relatively subjective nature of one of its components, histological grade. In this respect Henson et al

have carried out a retrospective analysis of prognostic data in over 22,000 women as part of the SEER (Surveillance, Epidemiology and End Results) Programme of the National Cancer Institute in the United States. Despite the fact that the data was collected from a large number of institutes and there was no standardisation of methods, especially grading, they confirmed that a combination of stage and grade improved prediction of outcome. Furthermore, they argue powerfully that observer variation in grade assignment has not been proven to interfere with the estimation of prognosis in patients with breast cancer. In an analysis of 379 patients Chevallier and colleagues

identified young age, tumour size and histological grade as factors which added to lymph node stage in the prediction of recurrence. These factors were combined to divide lymph node negative patients into three prognostic groups.

As discussed previously, 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 now been provided by its validation in two large multi-centre studies.

In the Yorkshire Breast Cancer Group the NPI was applied to 1186 patients and in the Danish Breast Cancer Co-operative Group over 9000 patients were studied; both obtained excellent separation into three prognostic groups.

Such studies confirms the inherent power of the pathological factors used in the NPI, and indicate that the index can be used for patient management in any centre provided that histopathologists are prepared to record this relatively simple data.

Figure 5.1 Survival curves for 1989 patients in the Nottingham Tenovus Primary Breast Cancer Study

Although the NPI has the considerable advantage of simplicity, as noted above, there is still a reluctance on behalf of many clinicians to use a system based on routine morphological histopathology, and this had led to a continuing search for more objective factors which may reflect the biology of individual breast cancers more accurately. Several commercial organisations are marketing prognostic indices based on such factors as hormone receptor status, DNA ploidy, S-phase fraction (SPF), Epidermal Growth Factor (EGFR) and C-erbB-2 expression. The Guys group have devised one such index for node-positive patients using C-erbB-2 expression and S-phase fraction. $17-O'Reilly SM, Barnes DM, Camplejohn RS, et al. The relationship between C-erbB-2 expression, S-phase fraction and prognosis in breast cancer. Br J Cancer 1991; 63: 444-446.$ It is interesting that the study from which this index was derived did not include an evaluation of histological grade.

Over the years many biological prognostic factors have been assessed in the NTPBCS. These include oestrogen receptor status (ER),

binding of epithelial mucin antibodies,

DNA index and SPF,

Epidermal growth factor EGFR,

C-erbB-2 expression,

c-myc expression,

Helix pomatia lectin binding

and p53 expression.

Each relates to prognosis, but also to histological grade and in multivariate analysis grade emerges as the more powerful (and, after its inclusion) the only significant

Many of these factors have been the subject of encouraging reports in the world literature. It is worth pointing out that although some show excellent prognostic separation at perhaps two years, any significant differences may be eliminated by 5-10 years. Ploidy, ER and probably EGFR are good examples of this: at 18 months patients with ER positive tumours show a 15% mortality compared with 30% for ER negative tumours - a 100% difference in mortality, but only a 5% difference in case survival. By 10 years the mortality is the same. Analysis too early has lead to many misleading publications on the value of individual prognostic factors. A further point is important; some factors have prognostic importance not strictly related to survival. ER, for example, predicts hormone responsiveness after primary treatment failure

and also in the adjuvant situation.

It has been stated previously that there is little point in pathologists attempting to provide accurate and reproducible prognostic information if it is not going to be use by their clinical colleagues in clinical practice. There are a number of specific applications for prognostic factors in breast cancer, and two main areas will be used as examples.

Table 5.3 Distribution of 134 invasive cancers detected in the prevalent round of mammographic screening in the Nottingham Breast Screening Service. Comparison is made with the distribution of 1629 cancers from an unscreened population.

NPI Group		Screened	Unscreened
	n	%	%
(EPG	59	44%	13%)
GPG	102	76%	29%
MPG	27	20%	54%
PPG	5	4%	17%

There is a marked shift of cases towards the more favourable end of the prognostic range, with more than 3 times as many tumours in the EPG, and a very small percentage in the PPG. In addition to the importance of these observations to breast screening theory there are also obvious

Within breast screening pathological prognostic factors have an important role in the quality assurance of the programme. Individual screening centres can monitor their performance in the detection of the favourable prognostic cases described above, and indeed, radiologists in the NHS BSP are set a target of 15 cancers measuring 1 cm or less per 10,000 women screened in the prevalent round.

It is clear from the data presented above that it is now possible to place individual women with breast cancer into separate prognostic groups. It follows logically from this that the therapy used for patients in the GPG, who have an 80% chance of surviving 15 years (little different from that of an age-matched population without breast cancer), should be different from that for those in the PPG in which only 40% will be alive after 3 years and less than 15% at 15 years, a survival comparable to that of advanced breast cancer (stage III or tumours >5 cm clinically). In the NTPBCS we are now using the NPI for individual patient management. This is discussed in more detail by Blamey

but a brief outline is pertinent here. Together with a consideration of other factors such as tumour type, vascular invasion (VI) and oestrogen receptor status (ER), and clinical features such as patient age, menopausal status and physical and mental health the NPI is used to make decisions on appropriate local and systemic therapy, as follows:

In Nottingham patients take part in the therapeutic decision making process. Once a diagnosis of malignancy is established the patient is counselled, and the first decision to be made is the type of primary local surgery, wide local excision (WLE) or mastectomy. In our initial series of patients treated by wide local excision there were no exclusions based on prognostic factors (all patients received WLE at their request) and an unacceptably high local recurrence rate of 20% ensued.

We found that the factors which were the most significant in predicting the likelihood of local recurrence were young age, tumour size and VI. Accordingly, in the preoperative assessment, patients with tumours >3 cm clinically and/or on imaging, and with evidence on mammography of multifocality are advised against WLE, especially in the under 40 age group. Postoperatively, to avoid conversion to mastectomy we require clear excision margins (>5 mm histologically) and absence of VI (minimal VI is allowed if tumour size is <2 cm). Conversion to mastectomy is strongly advised if tumours are >2 cm on pathological measurement, grade 3, node positive and with definite VI. Using such criteria 12% of patients initially offered wide local excision are converted to mastectomy, the rest proceeding to radiotherapy to the intact breast. This policy has led to a local recurrence rate of only 2.2% at a median five-year follow-up.

Selection for adjuvant radiotherapy to lymph nodes and flaps is also required after mastectomy. Overall, if neither irradiation nor axillary clearance are carried out some 40% of patients will develop loco-regional recurrence.

If prophylactic irradiation were carried out in all these patients 60% would receive unnecessary treatment, with its consequent morbidity. We have shown that loco-regional recurrence is largely confined to patients in the PPG

and have subsequently demonstrated in a randomised trial that adjuvant irradiation of this group significantly reduces loco-regional recurrence.

An index for local recurrence risk has been constructed, based on histological grade, lymph node stage and vascular invasion; adjuvant irradiation is recommended for patients with a high index score who would otherwise have a 40% chance of developing a local recurrence in the next 5 years.

There is still considerable debate concerning the use of adjuvant systemic therapy in primary operable breast cancer, which usually revolves around two main questions.

The NPI is the only prognostic index established to date which can provide satisfactory stratification for the application of systemic therapy. In the NTPBCS it is used as follows:

Patients in this group have a very good survival at 15 years compared with age matched women in the population at large. However, the mortality in the GPG is slightly higher and some 10% will die from breast cancer. Tamoxifen has been shown to give a small survival advantage in these women

but the gain is small and only 1 in 10 women stand to benefit. Although Tamoxifen is largely without serious side effects, problems are increasingly being recognised, the most worrying being an increased risk of endometrial carcinoma. Furthermore, tumours in the GPG are almost always ER positive, and patients are therefore likely to respond to Tamoxifen if metastatic disease presents.

It therefore seems reasonable to withhold adjuvant hormonal therapy in this group. The very small potential benefit for the group as a whole means that cytotoxic therapy is contraindicated in this group.

In these groups two further factors are considered in the allocation of systemic therapy, menopausal status and tumour ER status. Premenopausal patients whose tumours are ER positive, node positive are currently entered into the ICI 2802 clinical trial and are randomly allocated to receive either Zoladex (LH/RH antagonist) or CMF (cytotoxic drug regime). All other premenopausal patients are treated with adjuvant CMF.

Postmenopausal patients whose tumours are ER positive receive adjuvant Tamoxifen. Those whose tumours are ER negative receive cytotoxic therapy if they are deemed physically and mentally capable of tolerating the drug regime. If these patients are considered to be too frail for chemotherapy a trial of Tamoxifen may be given, on the basis that a small percentage of patients with ER negative tumours respond to hormone therapy.

The diagnosis of metastatic disease is a turning point for any breast cancer patient. In such a group of patients, where virtually all will die of their disease within a relatively short time, accurate prediction of prognosis and response to therapies has a considerable impact on management. From a retrospective multivariate analysis of 191 patients with metastatic breast cancer we identified four factors which were independently significant for survival from diagnosis of metastatic disease - histological grade and ER status of the primary tumour, site of initial metastatic disease (SIMD) and disease-free interval (DFI).

An index score was calculated as follows: Score = 4 x Grade (1-3) - 6 x ER (0 or 1, negative or positive) + 4 x SIMD (1-4, bone only, lung only, bone and lung, visceral) - 0.1 x DFI (months). Patients were divided into 3 groups based on the index score:

Respective survival after 18 months (primary endocrine therapy) was 67%, 37%, 0%.

This index has now been validated in a prospective series of patients.

The percentage of patients in each group was similar in both studies and the survival differences also matched. We now use the index to direct therapy in patients with metastases. Patients in Group A continue to receive endocrine therapy as initial treatment. Patients in Group C, who had a very poor prognosis on endocrine therapy now receive chemotherapy. Patients in Group B, whilst showing a very low response rate to endocrine therapy, have a survival rate between those of Groups A and C. Currently we continue to treat these patients with endocrine therapy. However, if cytotoxic therapy is shown to improve prognosis in Group C patients, then a similar therapeutic regimen will have to be considered for Group B patients.

In properly run breast units the psychological and social aspects of breast cancer form an exceedingly important part of patient management. The NPI has a useful role in this respect. For example, a young woman with breast cancer may well ask whether she can have another pregnancy. Her NPI shows her chances of survival: a woman in the PPG is clearly ill-advised to have further children without fully appreciating the consequences, since they will almost certainly be motherless in a few years; on the other hand, women in the GPG (and certainly in the EPG) should be encouraged to look upon themselves as cured and to live a normal life.

In this course we have tried to approach the topic of prognostic factors in breast cancer very much from the perspective of the routine diagnostic histopathology service and to indicate which factors are of most value in determining appropriate therapy for individual patients. In the Nottingham Tenovus Primary Breast Cancer Study a wide range of biological factors has been studied. It is our view that at the present time the Nottingham Prognostic Index, based on careful histopathological evaluation of tumour size, histological grade and lymph node stage, together with ER status, offers the most powerful and reproducible method of assessing prognosis. In the future more objective methods of estimating tumour differentiation and invasiveness may become available, but current techniques do not achieve significance in multivariate analysis when compared with histological grade. The NPI fully satisfies the criteria suggested by McGuire and Clark

and endorsed by Hawkins.

In an issue of Breast Cancer Research and Treatment entirely devoted to prognostic factors in breast cancer Clark stated that both the NPI and the Nottingham index for patients with metastatic breast cancer treated by endocrine therapy 'provide an excellent basis for the evaluation of newer factors that have been more recently proposed'.

The time has never been better for histopathologists to demonstrate the importance of their contribution to the management of patients with breast cancer. If we do not provide prognostic information to our clinical colleagues others, including commercial laboratories, will fill the void.