Predictors for malignant tissue (MT) in residual masses (RM) of non-seminomatous germ cell tumors (NSGCT) treated by chemotherapy (CT) before surgery.

Abstract

405 Background: About half of the patients operated for RM after CT for a NSGCT doesn’t benefit from surgery because RM contain only necrosis. We develop and validate a predictive score for identifying MT (teratoma and/or viable cancer cells) after primary CT for disseminated NSGCT. Methods: We have conducted a retrospective study. The development cohort (cohort 1) consists in 193 pts treated with CT and surgery of RM at the Oscar Lambret Cancer Center between 1997 and 2014. The validation cohort (cohort 2) consists in 98 pts treated in 5 other French sites. Size of RM, modification in size after CT, pre- and post-CT AFP levels, pre- and post-CT HCG levels, pre- and post-CT LDH levels were collected. The predictive factors have been first identified using univariate logistic regression analysis and then selected using step-by-step (forward) logistic regression. Discrimination has been assessed using Akaike Information Criterion (AIC) and ROC curve. Results: The rates of MT in the RM were 52% and 68%, in the cohort 1 and 2, respectively. The 1-, 2- and 5-y OS in the development cohort were 99, 97 and 95%, respectively. The 1-, 2 and 5-y relapse-free survival in the development cohort were 96, 92 and 91%, respectively. In univariate analysis, the predictive factors for MT were: size of RM > 15 mm (OR=3.32 [95%-CI: 1.6-6.2], p=0.001), stable RM versus diminution in size of RM (OR=3.0 [1.5-5.9] p=0.001) or increase in size versus diminution in size of RM (OR=8.6 [1.7-41.9] p=0.008) and pre-CT AFP level > ULN (OR=2.9 [1.4-6.3], p=0.006). The multivariate analysis had retained 2 predictors for MT: size of RM and pre-CT AFP level > ULN. The ROC was 0.71 [0.61-0.80] with an AIC=143 in the development cohort, and was 0.59 [IC95% 0.46-0.72] in the validation cohort. The model showed adequate calibration with a Hosmer-Lemeshow test p=0.69. Conclusions: Surgery of RM is part of standard of care of disseminated NSGCT. However, this surgery is morbid and at the end about 50% of pts are free of MT. We have developed and validated a predictive score for identifying pts benefiting of surgery.