BANG gel (MGS Research, Inc., Guilford, CT) has been evaluated for measuring intensity-modulated radiation therapy (IMRT) dose distributions. Treatment plans with target doses of 1500 cGy were generated by the Peacock IMRT system (NOMOS Corp., Sewickley, PA) using test target volumes. The gels were enclosed in 13 cm outer diameter cylindrical glass vessels. Dose calibration was conducted using seven smaller (4 cm diameter) cylindrical glass vessels irradiated to 0-1800 cGy in 300 cGy increments. Three-dimensional maps of the proton relaxation rate R2 were obtained using a 1.5 T magnetic resonance imaging (MRI) system (Siemens Medical Systems, Erlangen, Germany) and correlated with dose. A Hahn spin echo sequence was used with TR = 3 s, TE = 20 and 100 ms, NEX = 1, using 1 x 1 x 3 mm3 voxels. The MRI measurements were repeated weekly to identify the gel-aging characteristics. Ionization chamber, thermoluminescent dosimetry (TLD), and film dosimetry measurements of the IMRT dose distributions were obtained to compare against the gel results. The other dosimeters were used in a phantom with the same external cross-section as the gel phantom. The irradiated R2 values of the large vessels did not precisely track the smaller vessels, so the ionization chamber measurements were used to normalize the gel dose distributions. The point-to-point standard deviation of the gel dose measurements was 7.0 cGy. When compared with the ionization chamber measurements averaged over the chamber volume, 1% agreement was obtained. Comparisons against radiographic film dose distribution measurements and the treatment planning dose distribution calculation were used to determine the spatial localization accuracy of the gel and MRI. Spatial localization was better than 2 mm, and the dose was accurately determined by the gel both within and outside the target. The TLD chips were placed throughout the phantom to determine gel measurement precision in high- and low-dose regions. A multidimensional dose comparison tool that simultaneously examines the dose-difference and distance-to-agreement was used to evaluate the gel in both low-and high-dose gradient regions. When 3% and 3 mm criteria were used for the comparisons, more than 90% of the TLD measurements agreed with the gel, with the worst of 309 TLD chip measurements disagreeing by 40% of the criteria. All four MRI measurement session gel-measured dose distributions were compared to evaluate the time behavior of the gel. The low-dose regions were evaluated by comparison with TLD measurements at selected points, while high-dose regions were evaluated by directly comparing measured dose distributions. Tests using the multidimensional comparison tool showed detectable degradation beyond one week postirradiation, but all low-dose measurements passed relative to the test criteria and the dose distributions showed few regions that failed.