Plasma biomarker study of lenvatinib in gastroenteropancreatic neuroendocrine tumors reveals Ang2 and FGF2 as predictors of treatment response: Results from the international phase II TALENT trial (GETNE 1509)

Abstract

Background: Predictive biomarkers of response to antiangiogenics currently remain elusive, with several molecules discovered but not fully validated nor clinically applied. TALENT trial is a multicenter prospective phase II study of lenvatinib, a VEGFR1-3 and FGFR1-4 multikinase inhibitor (MKI), in advanced G1/G2 neuroendocrine tumors (NETs) from pancreatic (panNETs) and gastrointestinal (giNETs) origins, which has reported the highest overall response rate (ORR) by central radiology assessment with a MKI in this setting. Here we report the plasma biomarker study. Methods: Proangiogenic profiling of plasma samples from patients included in the trial were analyzed by multiplex ELISA (custom made Quantibody Array, RayBiotech). Quantitative determinations of VEGF-A, FGF2, FGF4, Ang2, IL8, PlGF, VEGF-C, VEGF-D and VEGFR2 were obtained from 85 samples of sufficient quality (out of 111 patients included in the study). Association and prediction of response were evaluated for each biomarker and correlated with PFS, OS and ORR in all patients and the different subgroups included in the trial. Results: While none of the factors were able to discriminate PFS or OS in the whole population, a significant association of high-Ang2 and low-FGF2 to ORR was observed. Subgroup analysis confirmed this association in the two cohorts of patients, giNETs (p = 0.003) and panNETs (p = 0.024). In the panNET cohort, prior targeted therapy was mandatory. Prior sunitinib exposure was observed in 8 patients. Of the factors studied, Ang2 and VEGFR2 were significantly decreased in plasma from sunitinib-pretreated patients (-73% p = 0.022 and -62% p = 0.042 respectively). Furthermore, in these sunitinib-pretreated patients Ang-2 and VEGFR2 levels associated to ORR (p = 0.029 and p = 0.029 respectively) and were able to discriminate PFS (log rank p = 0.027 and 0.007 respectively). ROC curve analysis defined Ang2 and VEGFR2 plasma levels with optimal predictive power to be 415pg/ml and 1770pg/ml respectively (p = 0.021). Conclusions: Plasma proangiogenic profiling of TALENT patient samples unraveled high-Ang2 and low-FGF2 as predictive biomarkers of response to lenvatinib in both cohorts. In antiangiogenic-pretreated patients, Ang2 and VEGFR2 levels significantly predict response to treatment in panNETs. Importantly, current studies with MKIs should confirm the value of these markers for advanced NETs not only to predict response, but also to stablish sequential treatment options for these patients.