High levels of uPA and PAI-1 predict a good response to anthracyclines

Simona Borstnar (1) , Aleksander Sadikov (2) , Barbara Mozina (1) , Tanja Cufer (3)

1 Institute of Oncology
2 Faculty of Computer and Information Science, Artificial Intelligence Laboratory
3 University Clinic Golnik

Urokinase-type plasminogen activator (uPA) and its main inhibitor (PAI-1) were shown with level 1 evidence to be prognostic factors for primary breast cancer. Our preliminary retrospective study on a cohort of 1,220 consecutive patients hinted that uPA and PAI-1 could also serve as predictive factors for systemic therapy, namely that patients with high levels of the two markers benefit much more from anthracycline-based chemotherapy than patients with low levels of the two markers. The latter could equally well be treated with less toxic CMF-based chemotherapy (cyclophosphamide, methotrexate, and fluorouracil). The retrospective study, however, suffered from severely uneven patient and tumor characteristics as the patients were treated per institutional guidelines valid at the time and were not randomized between the anthracycline and CMF arms. In the present paper, we attempted to remedy this shortcoming and recheck our previous observations on more balanced data. To this end we employed a custom-made computer algorithm that selected 180 patients out of a total of 1,220 patients such that we obtained very well balanced anthracycline and CMF arms according to patient and tumor characteristics. Moreover, the low and high uPA/PAI-1 subgroups within both arms were also completely balanced. The algorithm in a way created a similar setting to that of a randomized study at the expense of greatly reducing the number of patients included into the study. In this setting, we observed the 3-year disease-free survival (DFS) in all four subgroups (according to treatment and levels of markers: both uPA and PAI-1 low versus one or both high). We report that the 3-year DFS in the CMF arm differed significantly: 87.1% for patients with low levels of markers versus 77.0% for patients with high levels of markers ( = 0.044, HR = 2.81, 95% CI = 0.98-8.04). On the other hand, the 3-year DFS in the anthracycline arm did not differ much between the two marker level subgroups: 85.2% for patients with low levels of markers versus 81.8% for patients with high levels of markers. Our observation points out that worse prognosis correlated to high uPA and PAI-1 levels can be reversed by treatment efficacy achieved through anthracycline-based chemotherapy. Based on this observation, we hypothesize that uPA/PAI-1 combination could be predictive for response to systemic therapy.

Mots clés: uPA; PAI-1; Predictive value; Breast cancer; Anthracycline-based chemotherapy; Balancing patients' and tumor characteristics

Simona Borstnar, Aleksander Sadikov, Barbara Mozina, Tanja Cufer. High levels of uPA and PAI-1 predict a good response to anthracyclines. Breast Cancer Research and Treatment, 2009, 121 (3), pp.615-624. ⟨10.1007/s10549-009-0691-8⟩. ⟨hal-00612980⟩