American Statistical Association New York City Metropolitan Area Chapter Mailman School of Public Health Columbia University Department of Biostatistics Colloquium

The recently revised Response Evaluation Criteria for Solid Tumors (RECIST 1.1) recommends measuring up to a total of five lesions per patient when measuring tumor burden in patients with multiple lesions. This recommendation represents a change from the previous guidelines (RECIST 1.0) which advise measuring up to a total of ten target lesions per patient. There are no existing guidelines which specify the minimum number of lesions that should be measured. In this talk, we will derive an approximation of the distribution for the ratio that quantifies the relative percent change in tumor burden between a baseline and follow-up measurement. We extend this theory to look at the conditional distribution of the relative percent change in tumor burden when a subset of lesions is measured given knowledge of the distribution when all lesions are measured. This conditional distribution can be used to calculate the sensitivity and specificity of the response assessment based on a subset of lesions. These metrics are familiar to radiologists and clinicians and have the potential to help guide them in their choice of the minimum number of lesions that need to be measured in a given application.

Chaya Moskowitz is a biostatistician at Memorial Sloan-Kettering Cancer Center with extensive experience collaborating with its Department of Radiology. She recently contributed to work used by the RECIST working group when revising their guidelines.