The imaging criteria mRECIST, which was initially introduced to assess the efficacy of HCC vascular bed destruction induced by TACE, is now often used to assess subtler vascular effects induced by anti-VEGFR therapy. However, variability of mRECIST assessment has been only partially investigated. In this study, we evaluated the inter-reader variability and its sources in the mRECIST assessment of treatment response for HCC treated by anti-VEGFR drugs.
A subset of 41 advanced HCC tumors in 24 patients treated by anti-VEGFR drugs were selected from a phase I/II nalysedre study (the original study). These data were retrospectively reviewed according to mRECIST criteria by 3 mRECIST non-expert radiologists each having different levels of experience. Each liver lesion measurement was hand drawn using an electronic caliper. Results from these 3 radiologists were then compared to those extracted by an mRECIST expert at the time of the original study. The precision of measurements among the 3 non-experts and between the expert and the 3 non-experts were nalysed by assessing bias and standard deviation (SD) using the Bland-Altman method. The agreement of readers’ responses was assessed using the Kappa coefficient statistic. The causes of discrepancies were nalysed.
Among the 3 non-experts, SD of measurements ranged [24.9%; 36.3%] and the Kappa coefficients were moderate 0.41 [0.28; 0.55]. SD in measurements of expert versus non-experts ranged [33.2%; 41.1%] and Kappa coefficients were poor 0.20 [0.06; 0.35]. Pooling the four readers together, the rate of discrepancy at declaring either Progressive Disease (PD) or Partial Response (PR) per patient was identical at 41.7% (10/24). The main cause of discrepancy at declaring PD came from the complexity of HCC enhancement patterns and the poor definition of tumors boundaries. Discrepancies at detecting PR came from the reader variability at selecting only the viable part or the entire liver tumor.
When used by mRECIST non-experts to assess the subtle vascular effect induced on HCC by anti-VEGFR therapy, mRECIST appears to lack reproducibility. It is therefore important when using mRECIST to require specific training to reduce readers variability.
Clinical trial identification
Legal entity responsible for the study
Has not received any funding.
H. Beaumont, N. Faye, C. Klifa: Employee: Median Technologies. All other authors have declared no conflicts of interest.