Inter-observer variation in rectal cancer gross tumor volume delineation by radiologist and oncologist

BackgroundThe PRORECT trial compares proton and photon radiotherapy in rectal cancer.PurposeTo evaluate inter-observer agreement in gross tumor volume (GTV) delineation between a radiation oncologist and a radiologist in the PRORECT trial. In addition, we evaluated baseline tumor volume and length as predictors of clinical complete response (cCR) and compared tumor length and tumor and nodal staging (TN-stage) between diagnostic magnetic resonance imaging (dMRI) and treatment planning MRI (tpMRI).Material and MethodsA total of 50 consecutive patients underwent dMRI and tpMRI. GTV on tpMRI were delineated by a radiation oncologist (GTVo) and a radiologist (GTVr). Tumor lengths and TN staging were determined by the radiologist on both dMRI and tpMRI.ResultsPatients underwent dMRI and tpMRI with a median interval of 23 days (range=10-51 days). Median GTVo and GTVr volumes were 41.3 cm³ (range=3.7-149.9 cm³) and 29.6 cm³ (range=1.1-137 cm³), respectively. The median Dice score was 0.75 (range=0.3-0.86). GTVo and GTVr were highly positively correlated (r(48) = 0.919; P < 0.001) and the median volume difference was significant (W, Z = -4.5214; P < 0.001). GTVr volume, but not tumor length, was significantly correlated with cCR (r_pb = -0.395, n = 50; P = 0.005). Increasing GTVr was associated with reduced likelihood of achieving cCR. No changes in TN stage between dMRI and tpMRI were observed.ConclusionThe overall inter-observer agreement in GTV delineation between radiologist and radiation oncologist was good (median Dice score = 0.75 [range=0.3-0.86]). There was a small difference in measured tumor length between dMRI and tpMRI, but no change in TN stage between dMRI and tpMRI. GTVr was a significant predictor of achieving cCR.