The axillary lymph node (AX) region is the principal site for metastatic spread from carcinoma of the breast. The second major site of regional metastases is the internal mammary lymph node (IM) chain. It is well known that histologic presence or absence of AX and/or IM is a crucial diagnostic parameter for the prognosis of the breast cancer patients (Veronesi et al., 1985; Noguchi et al., 1991 (Leathem et al., 1984;Fenlon et al., 1987;Alam et al., 1990 Postoperatively, lymph nodes were removed from the resected specimens. The average total number of dissected axillary lymph nodes per patients was 26 nodes, with a range of 12-105. At least three sections were made of each lymph node for histological examination with Hematoxylin and Eosin staining. There were 53 patients with negative AX, 20 with one to three positive AX, and 25 with four or more positive AX. IM were positive in 17 patients and negative in 81 patients.The method for preparation of the paraffin sections for HPA staining has been described elsewhere (Fukutomi et al., 1989). Most cases were clearly either intensively positive or completely negative. In a few cases we used the scoring defined by Brooks et al. (1991). Finally cases were classified as positive or negative staining.Statistically, a comparison was made using the chi-square test. In the univariate study, overall or disease-free survival was studied by the Kaplan-Meier method (Kaplan & Meier, 1958), and a log-rank test was used to assess statistical significance. In the multivariate study, Cox's regression test was used to examine several parameters simultaneously, followed by multiple regression analysis to determine which variables are important for predicting AX or IM metastases. From the coefficients of variables selected by a stepwise forward selection method, we were able to construct the following discriminant function. Discriminant score (Z) = a. + a,X, +. ..... + apXp, (a.; constant; al, a2.ap: discriminant coefficient; XI, X2. Xp: explanatory variables). The probability (P) of positive AX or IM was calculated by the following logistic function: P = ez/(1 + ez) = 1/(1 + e-z), e = 2.718.-). For predicting the AX or IM metastases, Z>0 (P>0.5) was regarded as positive AX or IM, and Z.0 (P<0.5) as negative AX or IM.Results Correspondence: M. Noguchi, Operation Center, Kanazawa University Hospital, School of Medicine, Kanazawa University, Takaramachi, 13-1, Kanazawa, 920, Japan. * Recipient of a fellowship from the Japanese-Germany Center, Berlin, Germany. Received 30 June 1992; and in revised form 6 January 1993.Forty-four (45%) of the 98 breast cancers were HPA-positive, and the remaining 54 (55%), HPA-negative. The positive rates of HPA staining were significantly higher as the disease progressed, with 7% for Stage 1, 42% for Stage 2, and 69% for Stage 3 (P<0.01). Moreover, HPA staining was found to be significantly associated with tumour size Br.