Comparison of linear, bi-dimensional, and volumetric measurements in evaluating tumor response of hepatocellular carcinoma lesions in the arterial and portal venous phases on MRI

There are unmet needs in evaluating treatment response of hepatocellular carcinoma in research protocols. Early predictors, such as imaging biomarkers, could allow for earlier judgment of treatment effect. Currently RECIST is the most widely accepted criterion in clinical trials. A modified RECIST (mRECIST) criterion was developed to take into account the unique imaging characteristics of HCC lesions. Much discussion has occurred regarding linear measurements and their appropriateness for evaluating change in tumor burden over time. The simplicity of currently accepted criteria differs with the increasing sophistication of imaging techniques. Tumor volume change on 3D imaging can provide insight into actual action of treatment rather than an estimate of action as shown by linear and bi-dimensional measurements. It was the aim of this study to determine whether linear, bi-dimensional, and volumetric percent changes of HCC lesions, in both the arterial and portal venous phases, are significantly comparable. 27 HCC lesions (identified on 25 subjects) were measured at two timepoints by each method on 3D GRE MRI scans in both phases. Percent change was calculated per lesion for each measurement type in both the arterial and portal venous phases. Signed rank tests, paired t tests, and comparison of change tests were run to evaluate the data. Significant differences between the percent changes of linear measurements versus volumetric measurements were observed using a Wilcoxon signed-rank test which showed p = 0.0000. A simple correlation assessment showed positive correlations for all measurements, with the lowest being correlations 0.8679 for the arterial linear percent change versus the arterial volumetric percent change and 0.8434 for the portal venous linear percent change versus the portal venous volumetric percent change. Differences between percent changes of linear versus bi-dimensional measurements and bi-dimensional versus volumetric measurements were significant as well (Linear versus bi-dimensional p = 0.0001, bi-dimensional versus volumetric p = 0.0004). To conclude, the differences in the percent changes when comparing the measurement types are statistically significant, particularly when comparing linear and volumetric measurements. Establishing a reproducible volumetric criterion could lead to improvements in the implementation of clinical trials