British Journal of Cancer (1999) 80(3/4), 419–426 © 1999 Cancer Research Campaign Article no. bjoc.1998.0373 Risk-group discrimination in node-negative breast cancer using invasion and proliferation markers: 6-year median follow-up N Harbeck1, P Dettmar2, C Thomssen4, U Berger3, K Ulm3, R Kates3, H Höfler2, F Jänicke4, H Graeff1 and M Schmitt1 1Frauenklinik, Klinikum rechts der Isar, 2Institut für Allgemeine Pathologie und Pathologische Anatomie and 3Institut für Medizinische Statistik und Epidemiologie, Technische Universität München, Ismaningerstr. 22, D-81675 Munich, Germany; 4Universitätsfrauenklinik Eppendorf, Hamburg, Germany Summary Factors reflecting two major aspects of tumour biology, invasion (urokinase-type plasminogen activator (uPA), plasminogen activator inhibiter (PAI-1), cathepsin D) and proliferation (S-phase fraction (SPF), Ki-67, p53, HER-2/neu), were assessed in 125 node- negative breast cancer patients without adjuvant systemic therapy. Median follow-up time was 76 months. Antigen levels of uPA, PAI-1 and cathepsin D were immunoenzymatically determined in tumour tissue extracts. SPF and ploidy were determined flow-cytometrically, Ki--67, p53, and HER-2/neu immunohistochemically in adjacent paraffin sections. Their prognostic impact on disease-free (DFS) and overall survival (OS) was compared to that of traditional factors (tumour size, grading, hormone receptor status). Univariate analysis determined PAI-1 (P , 0.001), uPA (P 5 0.008), cathepsin D (P 5 0.004) and SPF (P 5 0.023) as significant for DFS. All other factors failed to be of significant prognostic value. In a Cox model, only PAI-1 was significant for DFS (P , 0.001, relative risk (RR) 6.2). In CART analysis for DFS, the combination of PAI-1 and uPA gave the best risk group discrimination. For OS, PAI-1, cathepsin D, tumour size and ploidy were statistically significant in univariate, but PAI-1 was the only independently significant factor in Cox analysis (P , 0.001, RR 8.9). In particular, this analysis shows that PAI-1 is still a strong and independent prognostic factor in node-negative breast cancer after extended 6-year median follow-up. Keywords: breast cancer; plasminogen activator inhibitor type 1 (PAI-1); prognosis; S-phase fraction; tumour biology; urokinase-type plasminogen activator (uPA) Node-negative breast cancer patients, in contrast to breast cancer gations have demonstrated that the invasion markers uPA (serine patients whose lymph nodes show tumour cell involvement at time protease urokinase-type plasminogen activator) and PAI-1 of primary therapy, have a low risk of suffering disease recur- (inhibitor of uPA) are statistically independent prognostic factors rences. About 70% of the node-negative patients are cured by to predict disease recurrence and death in node-negative breast surgery alone and will therefore not need any adjuvant systemic cancer. Elevated antigen levels of uPA and PAI-1 are associated therapy. Nevertheless, even within this low-risk breast cancer with poor prognosis (Duffy et al, 1990; Grøhndahl-Hansen et al, group, up to 30% of the patients may relapse within 10 years after 1993; Jänicke et al, 1993; Foekens et al, 1994; Fernö et al, 1996). surgery and eventually die of metastasis (Clark and McGuire, Another protease, cathepsin D, has also been associated with poor 1988). Traditional histomorphological and clinical factors such as patient outcome (Rochefort, 1992). tumour size, tumour grade, steroid hormone receptor status, age, Prior to the study of tumour-associated proteolytic factors, flow or menopausal status, have been used to identify the high-risk cytometric DNA analysis was reported to yield valuable prog- node-negative patients who may benefit from adjuvant systemic nostic information in breast cancer patients. Ploidy and, in partic- therapy (McGuire and Clark, 1992). However, by applying these ular, S-phase fraction have been addressed as rather powerful traditional prognostic factors, more than 75% of all node-negative prognostic factors in node-negative breast cancer (Osborne, 1989). breast cancer patients will receive adjuvant systemic therapy In recent years, immunohistochemical detection of the prolifera- (McGuire and Clark, 1992), even though only about 30% of all tion-associated antigen Ki-67 has also been used for proliferation node-negative patients will eventually develop systemic disease. assessment. The ability to detect Ki-67 in formalin-fixed paraffin This obvious discrepancy has stimulated the search for new prog- sections by means of a monoclonal antibody, MIB1, made this nostic factors, and numerous factors have been suggested so far for technique readily available for determination of tumour cell prolif- the assessment of breast cancer prognosis (Harris et al, 1992). eration (Dettmar et al, 1997). Tumour-biological factors such as those reflecting invasion and Molecular markers, such as the HER-2/neu gene, which codes metastasis or proliferation have strongly been put forward in the for an analogue of the epidermal growth factor receptor, as well as literature as new prognostic markers. Several independent investi- the tumour suppressor gene p53, have also been associated with patient prognosis. Yet their prognostic impact is still quite contro- versial. In addition, their unique tumour-biological role has not yet Received 28 May 1998 been fully determined (Clark, 1996). Revised 21 October 1998 Even though there is abundant literature on so-called new prog- Accepted 19 November 1998 nostic factors in primary breast cancer, most publications merely Correspondence to: N Harbeck compare one or two new factors to the traditional prognostic 419 420 N Harbeck et al factors. In addition, subgroup analyses of clinically relevant Table 1 Node-negative breast cancer patients without adjuvant systemic patient collectives such as node-negative patients are often not therapy. Prospective analyses were performed in all 125 patients, retrospective analyses only in cases where sufficient tumour tissue was left performed due to small patient numbers. Therefore, after a long- for analysis term median follow-up of more than 6 years, we have now evalu- ated the prognostic impact of eight tumour-biological factors Factors n (%) (uPA, PAI-1, cathepsin D, S-phase, ploidy, Ki-67, p53, HER- 2/neu) in a homogeneous, clinically important cohort of node- Tumour size (cm) 125 ²2 65 (52) negative patients whose follow-up data were not altered by effects . 2 and ² 5 56 (44.8) of any adjuvant systemic therapy. In order to ensure comparability . 5 4 (3.2) of the results, we performed analysis of five factors (S-phase, Steroid hormone receptor status 125 ploidy, Ki-67, HER-2/neu, p53) on adjacent paraffin sections of Positive 99 (79.2) the same tissue block. The prognostic impact of the tumour- Negative 26 (20.8) biological factors on disease-free, as well as overall, survival was Grading 125 compared to that of the traditional prognostic factors tumour size, G 1/2 93 (74.4) grading and steroid hormone receptor status. G 3/4 32 (25.6) PAI-1 125 Low (² 14 ng mg21 protein) 99 (79.2) MATERIALS AND METHODS High (. 14 ng mg21 protein) 26 (20.8) uPA 125 Patients Low (² 3 ng mg21 protein) 83 (66.4) High (. 3 ng mg21 protein) 42 (33.6) Traditional factors (pathological tumour size, steroid hormone Cathepsin D 121 receptor status, grading) and tumour-biological factors (uPA, PAI- Low (² 41 pmol mg21 protein) 60 (49.6) 1, cathepsin D, S-phase, ploidy, Ki-67, p53, HER-2/neu) were High (. 41 pmol mg21 protein) 61 (50.4) assessed in tumour tissues obtained from 125 patients with node- S-phase fraction 101 negative breast cancer. Histological grade was scored according to Low (² 6%) 55 (54.5) Elston and Ellis (1991); completely undifferentiated tumours in High (. 6%) 46 (45.5) which a histological subtype could not be determined were classi- Ploidy 101 fied as G4. Diploid (near diploid, diploid) 51 (50.5) Patients either had a modified radical mastectomy (n 5 83) or Aneuploid (an-, multi-, tetraploid) 50 (49.5) underwent breast-conserving surgery with subsequent breast irra- MIB1 116 diation (n 5 42) at the Department of Obstetrics and Gynecology Low (² 25%) 96 (82.8) . of the Technische Universität München, Germany, between 1987 High ( 25%) 20 (17.2) and 1991. In accordance with the standard treatment at the time, p53 111 none of the patients received any adjuvant systemic therapy. Negative 102 (91.9) Positive 9 (8.1) Median age of all patients at primary therapy was 56 years (range 35–82 years). Further patient characteristics are displayed in Table HER-2/neu 101 Negative 55 (54.5) 1. At time of primary therapy no patient had clinical or X-ray Positive 46 (45.5) evidence of distant metastases. Follow-up data was obtained every 3–6 months. Median follow-up of patients still alive at time of analysis was 76 months (range 47–108 months). Twenty-three patients (18.4%) relapsed. Fifteen patients (12%) died of breast according to Harbeck et al (1994) and then calculated using the cancer and eight patients died of causes not related to breast cancer computer program ModFit (Verity Software House, ME, USA) within the follow-up period. (Dettmar et al, 1997). Immunostaining for p53, HER-2/neu (c-erbB-2) and Ki-67 (MIB1 antibody) was performed as described (Dettmar et al, 1997; Methods Harbeck et al, 1998) in adjacent 4-µm-thick paraffin sections using As described earlier (Jänicke et al, 1990, 1994a), uPA and PAI-1 the alkaline phosphatase anti-alkaline phosphatase (APAAP) have been measured in tumour tissue extracts in a prospective method. fashion since 1987 for all breast cancer patients who had their primary surgery performed at our department. uPA and PAI-1 Statistical analysis antigen were determined by commercially available enzyme- linked immunosorbent assays (ELISAs) in detergent extracts of Correlations between continuous variables were analysed using breast cancer tissue specimens and expressed as ng of antigen per the Spearman rank test.