To create this model, the authors utilized 2,291 patients treated for metastatic GCT at a single institution and identified 154 patients that had brain metastatic disease. They split the data set in half for training, with the other half for validation, and used logistic regression to identify predictive factors in a step-wise fashion. Pulmonary metastases size was calculated using the long axis diameter. The model will the largest area under the curve was tested in validation.
The characteristics of the cohort are shown below.
The final model ended up utilizing age, pre-chemotherapy beta-HCG level, the presence of bone metastases, choriocarcinoma predominant histology, and pulmonary metastases size, assigning different points to each characteristic as shown below. Based on the total points allocated by each characteristic, the authors demonstrated the probability of presence of brain metastatic disease.
Patients with brain metastatic disease in this cohort were managed with radiation only (42%), surgery only (14%), or both radiation and surgery (9%), with 35% of patients not receiving local therapy. The 2-year progression-free survival for patients with brain metastases was lower than patients without brain metastases (17% versus 65%, p < 0.001), and the 2-year overall survival were similarly lower (62% versus 91%, p < 0.001).
The authors concluded that their predictive model may be useful for helping clinicians identify metastatic germ cell tumor patients at high-risk for brain metastatic disease and worse outcomes.
Presented by: Ryan Ashkar, MD, Indiana University School of Medicine
Written by: Alok Tewari, MD, PhD, Medical Oncologist at the Dana-Farber Cancer Institute, during the 2021 American Society of Clinical Oncology Genitourinary Cancers Symposium (#GU21), February 11th-February 13th, 2021