Single and multidimensional measurements underestimate neuroblastoma response to therapy.

Abstract

Changes in three-dimensional (3D) measurements of neuroblastoma are used to assess response. Linear measurements may not accurately characterize tumor size due to the infiltrative character of these tumors. The purpose of this study was to assess the accuracy of one-dimensional (1D), two-dimensional (2D), and 3D measurements in characterizing neuroblastoma response compared to a reference standard of tumor volume. We retrospectively reviewed imaging for 34 patients with stage 3 or 4 neuroblastoma. Blinded readers contoured or made linear measurements of tumors. Correlation coefficients were used to compare linear measurements to volumetric and 3D measurements. Bland-Altman analyses were used to assess bias between measurements. Sensitivity and specificity for patient events and survival were calculated for each measurement technique. Mean patient age was 2.9 ± 3.0 years (range 0-15 years). There was strong correlation between volumetric and 1D (r = 0.78, P < 0.0001), 2D (r = 0.86, P < 0.0001), and 3D (r = 0.88, P < 0.0001) measurements. Mean bias between volumetric measurements and 1D, 2D, and 3D measurements was 37.1% (95% limits: 6.2-67.9%), 16.1% (95% limits: -11.7-43.8%), and 7.7% (95% limits: -19.7-35.1%), respectively. 1D and 2D measurements undercategorized response versus volumetric change in 88.2% (30/34) and 29.4% (10/34) of cases. 3D measurements incorrectly characterized response in 16.7% (4/24) of cases versus volumetric change. 3D measurements were highly sensitive for patient events and survival, but all measurement techniques had poor specificity. 3D measurements most accurately quantify neuroblastoma size response versus volumetric change in patients with stage 3 and 4 neuroblastoma. 1D and 2D measurements underrepresent tumor response.