To determine correlation of inter reader variability in sum of diameters using RECIST 1.1 with end point assessment in lung cancer.

Abstract

e13557 Background: Lung cancer is the leading cause of cancer death in the world including more than 160,000 deaths in the US. The purpose of the study was to determine whether inter reader variability in Sum of Diameters (SOD) of tumor burden has any correlation with variability in end point assessment in lung cancer progression. RECIST 1.1 is based on the SOD of target lesions seen on imaging studies. Response criteria for evaluation of target lesions include - Complete response (CR), Partial response (PR), Progressive disease (PD) and Stable disease (SD). The key determinant of patient response is based on Target Lesion response which in turn is determined by SOD. Inter reader variability study plays an important role in the development of reliable diagnostic tools and understanding of imaging outcomes given the confounding factors like effusion, atelectasis and consolidation in lung cancer that affect Target Lesion selection. Methods: Retrospective analysis of 470 patients was carried out using RECIST 1.1. Double read with adjudication is the preferred read model for submission studies where images are read by two independent reviewers blinded to treatment allocation. As per RECIST 1.1, lesions were measured in the longest diameter for non-nodal and short axis for nodal lesions. This was followed by the calculation of SOD for total tumor burden. If these two primary reviewers disagree, then a third radiologist, the “adjudicator”, reviews the assessments performed by the first two radiologists and selects between the more accurate one. For further analysis, patients were divided into 2 groups, the one with no adjudication i.e. agreement between both readers and the second group with adjudication i.e. disagreement between both readers and ANOVA was used to perform analysis of Variance. Results: Of 470 patients, 332 patients with disagreement were adjudicated, while there was agreement on 138 patients assessments between both readers. SOD of baseline visits for all patients was assessed using ANOVA - single factor with following results: F ratio of 4.76 for Disagreement group was more than F crit (3.86) with P-value 0.03, while for Agreement group F value was less than F crit. Conclusions: There is a direct relationship of variability in SOD at baseline between two readers to the possibility of disagreement in their end point assessment. Additional rules around selection and measurement of Target Lesions should be proposed in protocol to reduce variability and improve endpoint assessment outcomes.[Table: see text]

Can we predict discordant RECIST 1.1 evaluations in double read clinical trials?

“…Our study confirmed findings by Sharma et al (31) who concluded that there was an association of the variability of SOD at baseline with the variability of the study endpoint. However, our percentage of specific sum of diameters (SPropSOD) analysis did not confirm or contradict other works about dissociated responses (28,32), probably because this phenomenon is reportedly observed in only around 10% of cases (28).…”

Breaking down the RECIST 1.1 double read variability in lung trials: What do baseline assessments tell us?

“…We know from Sharma et al. ( 16 ) that the variability in SOD at baseline is a risk for discrepant responses but, while threshold values for detecting significant longitudinal SOD changes (in follow up) are proposed ( 17 ), none are proposed for critical values for differences in SOD with double reads at baseline. Darkeh et al.…”