Prognostic value of Helix pomatia in breast cancer. International (Ludwig) Breast Cancer Study Group.

Six hundred and eighty-four primary breast cancers from the International (Ludwig) Breast Cancer Study Group (IBCSG) were studied for Helix pomatia lectin (HPA) binding. There was a weak correlation between lymph node-positive and HPA positive (P = 0.04). In our series there was a large advantage in disease-free survival (DFS) and overall survival (OS) for node-negative patients (P < 0.0001 DFS and OS). However, there was no such advantage for HPA-negative patients (P = 0.23 DFS and P = 0.32 OS). We conclude that in this randomised patient group HPA is of no clinical predictive value.

Reports on the prognostic importance of Helix pomatia lectin binding in human breast cancer have been conflicting. In 1983 Leathem published that 14/14 normal breast specimens and 24/26 breast cancers stained positively with Helix pomatia lectin (Leathem et al., 1983). This was followed by two meetings abstracts (Leathem et al., 1984;Leathem et al., 1985) from the same group that demonstrated a strong relationship between HPA binding and axillary lymph node metastases. In 1987 Fenlon et al., in a series of 100 tumours, reported a significant correlation between HPA binding, tumour stage, local recurrence and survival (Fenlon et al., 1987). The authors did not comment on menopausal status. In 1987 Leathem and Brooks reported that HPA positivity correlated with the time to first recurrence and with survival, but that this was only true for premenopausal women . This is in contrast to the results of Fukutomi et al. (1989) who found HPA positivity to be strongly correlated with poor survival, irrespective of menopausal status (Fukutomi et al., 1989). In a recent report of 153 breast carcinomas using the same biotinylated lectin the significance had dropped to P = 0.05 (Fukutomi et al., 1991). All of these earlier studies were carried out on small numbers of patients and there was a need for a careful analysis of large numbers of cases in different centres. Last year, Brooks and Leathem published a larger series of 373 cases . They found a strong correlation between HPA positivity and the presence of lymph node metastases. In this study there is no comment on whether this correlates with menopausal status and whether this is a completely new data set or an extension of their 179 cases previously reported . The study of Galea et al. (1991) on 459 cases could not confirm these data or their original observation (Fenlon et al., 1987). On the basis of these conflicting results we carried out a pilot study on 363 cases randomised from the IBCSG (Ludwig) Trial V database. We found no correlation between HPA positivity and survival (Taylor et al., 1991). In view of these findings we have enlarged the study to 684 cases from this same data set to re-analyse our HPA positivity in relation to all recorded clinical and pathological parameters. We report here a summary of our findings and some of the conflicting views in this area.

Materials and methods
Cases were taken at random from 13 of the participating centres in the IBCSG Trial V. Details of this trial have been reported elsewhere (Ludwig Breast Cancer Study Group, 1988;Ludwig Breast Cancer Study Group, 1989). We have studied the association between HPA positivity and other prognostic factors and the effect of HPA binding on outcome. In view of the differences in the literature with staining techniques a pilot study was carried out by both the direct peroxidase conjugated method used by Galea (Galea et al., 1991) and the indirect avidin-biotin techniques described by Fukutomi (Fukutomi et al., 1989(Fukutomi et al., , 1991. Similar staining results were obtained for each and the positivity was inhibited by the appropriate sugar (N-acetyl-galactosamine), indicating that both methods produced specific staining. Figures 1 and 2 show similar fields of the same tumours stained by both methods. Both the tumour and the vascular endothelium show similar patterns of reactivity. Owing to the simplicity of the method all subsequent cases were stained using the direct method. All cases were scored according to the method described by , with all positive cases having either greater than 5% of the cells strongly positive to greater than 50% of the cells weakly positive. Samples were coded according to the patient randomisation number and all slides were reviewed independently by two pathologists (RA and BAG). In 10% of cases difficulty was encountered in using the scoring scheme and for these a consensus view was taken. All results were sent to the Biostatistics Center, Dana Farber Cancer Institute, where clinical correlations were studied.

Results
It was noted that in a large proportion of cases there was staining of normal breast vascular endothelium and associated erythrocytes. On a representative sample of randomly selected cases (Table I) there was no clear correlation between positivity of tumour and normal tissues. Table IIA shows that there is a very weak correlation between Helix pomatia positivity and lymph node status and no correlation between HPA binding and either DFS or OS (Table IIB). In a detailed analysis of HPA binding and other patient characteristics no correlations were found with menopausal status, ER, PR, tumour grade, vessel invasion, histological type pathological tumour size or treatment groups (Tables III and  IV). In addition, HPA does not predict a poorer prognosis in a Figure I a, Photomicrograph of a breast carcinoma stained with the direct method for HPA immunoreactivity. (x 360). Note staining on vascular endothelium (arrows). b, The same tumour stained by the indirect method. Note the similar staining pattern of the tumour and the blood vessels. (x 360). Note staining on vascular endothelium (arrows).     Fukutomi et al. (1989) in the paper quoted by Brooks in support of their own findings. In the flow cytometry paper of Alam et al. (1990), that also found a positive correlation between HPA positivity and lymph node involvement, the authors used a direct technique. They, however, state that they selected a cut off when related to grade and lymph node involvement of 20% of cells positive as this was the most informative (Alam et al., 1990).
It is difficult therefore to draw any conclusions from this publication until it is repeated by other groups with a larger data set.
We do not rule out the possibility that two staining methods with a lectin may give a different result dependent upon the cut off used for positivity. The difficulty is that with any study there is a statistical chance finding of a significant correlation between any given parameter and any selected cut 'unknown' categories. off point. We must conclude that there is no consistent data to strongly support the view that the localisation of HPA epitopes is of clinical significance. This issue will not be resolved on the basis of staining procedures. As discussed by , the binding of lectins is very complex and poorly understood. It is therefore essential that the reactive epitopes are properly characterised and more suitable antibodies with defined specificity made available. We await with interest the biochemical evidence for, and characterisation of, the glycoconjugate that Leathem et al., first identified immunocytochemically in 1983.
The lack of correlation between staining of tumour and normal tissues is very important because, although addressed in part by correspondence in the Lancet between Leathem  and Grundbacher (Grundbacher et al., 1987), the possible confounding effects of any analyses by secretor status and the binding ability of Helix pomatia  for Blood group A and AB should not be ignored. It could be argued that in order to analyse these data properly it is necessary to work with material from patients of blood groups B and 0 and to know in all cases the secretor status. Using the indirect technique Fukutomi et al. (1991) found that red cells in patients of blood group A were positively stained. Our incidence of 46% positive staining on the endothelium would correlate with binding to blood group A and AB determinants and the positivity on the luminal surface of the normal mammary glandular elements could in part reflect secretor status. Thus positivity of the tumours could be due to a number of factors. Until the HPA-binding ligands in the tumours are biochemically defined and demonstrated to be different from the determinants on the normal epithelium, there is no reason to assume that the 40% of cases that stain positively in both the tumour and the adja-cent normal tissue carry different epitopes in these two sites. Until that is proven, any hypotheses concerning the gaining of HPA binding sites in the progression to metastasis is premature.
As stated recently by Baum (Baum, 1991), even if it were possible to demonstrate a group of patients at higher risk of lymph node involvement, the false negative and false positive rates of these methods is such that this would not be a significant advance in the management of breast cancer.
We would like to thank Professor G. Westbury and the Lady Joseph Fund for their generous support for this work.
We thank the patients, the doctors (especially the pathologists), the nurses and the data managers who made this research possible. We also acknowledge the Ludwig Institute for Cancer Research We would like to thank Dr R. Millis, Dr D. Barnes, Dr I. Ellis, Dr R. Stoddart and Professor N. Wright for useful discussions during the preparation of this manuscript.