Comparison of one-, two-, and three-dimensional measurements of childhood brain tumors

Background. End points for assessing drug activity in brain tumors are dete rmined by measuring the change in tumor size by magnetic resonance imaging (MRI) relative to a pretreatment or best-response scan. Traditionally, two- dimensional (2D) tumor measurements have been used, but one-dimensional (1D ) measurements have recently been proposed as an alternative. Because softw are to estimate three-dimensional (3D) tumor volume from digitized MRI imag es is available, we compared all three methods of tumor measurement for chi ldhood brain tumors and clinical outcome. Methods: Tumor size from 130 MRI scans from 32 patients (32 baseline and 98 follow-up scans, for a total of 130 scans; median, three scans per patient; range, two to 18 scans) was mea sured by each method. Tumor-response category (partial response, minor resp onse, stable disease, or progressive disease) was determined from the perce ntage change in tumor size between the baseline or best-response scan and f ollow-up scans. Time to clinical progression was independently determined b y chart review. All statistical tests were two-sided. Results: Concordances between ID and 2D, ID and 3D, and 2D and 3D were 83% (95% confidence inter val [CI] = 67% to 99%), 61% (95 % CI = 47% to 75%), and 66% (95% CI = 52% t o 80%), respectively, on follow-up scans. Concordances for ID and 3D and fo r 2D and 3D were statistically significantly lower than the concordance for ID and 2D (P<.001 and P =.003, respectively). Concordance among 1D, 2D, an d 3D methods in detecting partial response was high; there was less concord ance in classifying tumors in the minor response and progressive-disease ca tegories. Median times to progression measured by the 1D, 2D, and 3D method s were 154, 105, and 112 days, respectively, compared with 114 days based o n neurologic symptoms and signs (P =.09 for overall comparison). Conclusion s: Detection of partial responses was not influenced by the measurement met hod, but estimating time to disease progression may be method dependent for childhood brain tumors.