Tumor Growth Rate to Predict the Outcome of Patients with Neuroendocrine Tumors: Performance and Sources of Variability

Abstract

Introduction: Tumor growth rate (TGR), percentage of change in tumor volume/month, has been previously identified as an early radiological biomarker for treatment monitoring in neuroendocrine tumor (NET) patients. We assessed the performance and reproducibility of TGR at 3 months (TGR_3m) as a predictor factor of progression-free survival (PFS), including the impact of imaging method and reader variability. Methods: Baseline and 3-month (±1 month) CT/MRI images from patients with advanced, grade 1–2 NETs were retrospectively reviewed by 2 readers. Influence of number of targets, tumor burden, and location of lesion on the performance of TGR_3m to predict PFS was assessed by uni/multivariable Cox regression analysis. Agreement between readers was assessed by Lin’s concordance coefficient (LCC) and kappa coefficient (KC). Results: A total of 790 lesions were measured in 222 patients. Median PFS was 22.9 months. On univariable analysis, number of lesions (</≥4), tumor burden, and presence of liver metastases were significantly correlated with PFS. On multivariate analysis, ≥4 lesions (HR: 1.89 [95% CI: 1.01–3.57]), TGR_3m ≥0.8%/month (HR: 4.01 [95% CI: 2.31–6.97]), and watch and wait correlated with shorter PFS. No correlation was found between TGR_3m and number of lesions (rho: −0.2; p value: 0.1930). No difference in mean TGR_3m across organs was shown (p value: 0.6). Concordance between readers was acceptable (LCC: 0.52 [95% CI: 0.38–0.65]; KC: 0.57, agreement: 81.55%). TGR_3m remained a significant prognostic factor when data from the second reader were employed (HR: 4.35 [95% CI: 2.44–7.79]; p value <0.001) regardless his expertise (HR: 1.21 [95% CI: 0.70–2.09]; p value: 0.493). Discussion/Conclusion: TGR_3m is a robust and early radiological biomarker able to predict PFS. It may be used to identify patients with advanced NETs who require closer radiological follow-up.