Comparison of 1- and 2-dimensional measurements with volumetric measurements for evaluation of change in tumor size

Abstract

3563 Background: Change in tumor size is commonly used as a surrogate endpoint for clinical outcome. Measurement of tumor volume, the gold standard for determining tumor size, is not clinically practical. The 2 most commonly used measurements of tumor size ie, longest diameter in axial plane (LD, used in RECIST) and cross product (CP) of LD and the longest perpendicular dimension in axial plane (used in WHO), are surrogates for tumor volume and assume simple tumor geometry. Since most tumors grow and regress irregularly, we determined which of these measurements most accurately reflects change in tumor volume by comparing change in tumor volume calculated using LD or CP with change in measured tumor volume. Methods: Computed tomography (CT) scans of the chest, abdomen, and pelvis were acquired at baseline and 8-week intervals during treatment in 22 patients (pts) enrolled in 2 first-in-human (FIH) clinical trials. CT scans were analyzed by a central reader to obtain LD, CP, and tumor volume. LD (assuming spherical tumor geometry) or CP (assuming spherical or elliptical tumor geometry) was used to calculate volume. The mean bias (μ: mean difference between measured and calculated percent change in tumor volume from baseline [%ΔVol]) and limits of agreement (μ ± 2s [s = standard deviation of difference]) from Bland-Altman analyses (used for assessing agreement between 2 measurement methods) were used for comparisons between measured and calculated %ΔVol. A nested effects model (tumor lesions considered as nested within pts) was used to test significance of differences between measured and calculated %ΔVol. Results: Bland-Altman analyses showed that μ ± 2s was -5% ± 70% for LD, 5% ± 38% for CP (spherical), and 7% ± 33% for CP (elliptical). Nested analyses indicated consistent results with an observed significant difference (p-value = 0.02) between measured and calculated %ΔVol using LD and non-significant differences (p-values = 0.28 and 0.05) using CP (assuming elliptical and spherical geometry respectively). Conclusions: In this diverse FIH population, 2-dimensional tumor measurements (using CP) are better than 1-dimensional tumor measurements (using LD) in representing changes in tumor volume and should decrease noise in the measurement of treatment response. No significant financial relationships to disclose.