PROGNOSTIC FACTORS FOR METASTATIC TESTICULAR GERM-CELL TUMORS - THE MEMORIAL-SLOAN-KETTERING CANCER MODEL

Since November 1982, patients treated at Memorial Hospital for advanced germ cell tumours have been allocated to good or poor risk clinical trials by the use of a mathematical model. A probability of complete response (CR) is based upon a patient's actual pre-treatment values of LDH and HCG and the number of metastatic sites. In good risk trials, a regimen consisting of four cycles of etoposide+cisplatin results in a durable proportion of CR of about 90% and causes minimal toxicity. Modifications of standard dose regiments have failed to increase the CR proportion in poor risk patients despite greater toxicity. The data show that about 10% of good risk patients will die despite favourable prognostic features, and about 35% of poor risk patients will achieve a durable CR with standard therapy despite unfavourable clinical characteristics. Therefore better markers of prognosis are needed. Two addition categories of markers are under study at Memorial Hospital. Firstly, the half-life decline of AFP and HCG (a 'post-treatment' marker) measured during the first two months of initial therapy is strongly associated with the likelihood of a CR. A prolonged decline of either AFP or HCG resulted in an inferior survival in both good and poor risk patients. Because few good risk patients had a prolonged marker half-life decline, this observation is being confirmed prospectively in an ongoing randomized trial. However, in poor risk patients, a prolonged half-life marker decline now results in an immediate change to high-dose therapy with autologous stem cell rescue. Secondly, i(12p) copy number and Hst-1/kFGF expression ('tumour-specific' variable) have been associated with stage and/or prognosis. In addition, p53 mutations have been observed. Therefore, the relative contribution of these tumour-specific markers to a patient's prognosis is being defined prospectively by evaluating them in paraffin-embedded primary tissue in patients allocated to good and poor risk trials. The independence (or dependence) of these putative new prognostic variables will be established by multivariate analysis. In this fashion, traditional markers of prognosis will be tested against new markers and, if appropriate, a new allocation algorithm established.