Objectives: To empirically determine thresholds for volumetric assessment of response and progress of liver metastases in line with the unidimensional RECIST thresholds.
Methods: Patients with metastatic colorectal cancer initially enrolled in a multicentre clinical phase-III trial were included. In all CT scans, the longest axial diameters and volumes of hepatic lesions were determined semi-automatically. The sum of diameters and volumes of 1, ≤2 and ≤5 metastases were compared to all previous examinations. Volumetric thresholds corresponding to RECIST 1.1 thresholds were predicted with loess-regression. In sensitivity analysis, the concordances of proposed thresholds, weight-maximizing thresholds and thresholds from loess-regression were compared. Classification concordance for measurements of ≤2 metastases was further analyzed.
Results: For measurements of ≤2 metastases, 348 patients with 629 metastases were included, resulting in 4,773 value pairs. Regression analysis yielded volumetric thresholds of -65.3% for a diameter change of -30%, and +64.6% for a diameter change of +20%. When comparing measurements of unidimensional RECIST assessment with volumetric measurements, there was a concordance of significant progress (≥+20% and ≥+65%) in 88.3% and of significant response (≤-30% and ≤-65%) in 85.0%.
Conclusions: In patients with hepatic metastases, volumetric thresholds of +65% and -65% were yielded corresponding to RECIST thresholds of +20% and -30%.
Key points: • Volumes and diameters of liver metastases from colorectal cancer were determined. • Volumetric thresholds of +65%/-65% corresponding to RECIST 1.1 are proposed. • Comparing both measurements, concordance was 88.3% (significant progress) and 85.0% (significant response).
Keywords: Image interpretation, computer-assisted; Liver; Neoplasm metastasis; Response evaluation criteria in solid tumors; Tomography, X-Ray computed.