Inter-observer variability of clinical target volume delineation in radiotherapy treatment of pancreatic cancer: a multi-institutional contouring experience

BACKGROUND An observational multi-institutional study has been conducted aimed to evaluate the inter-observer variability in clinical target volume (CTV) delineation among different radiation oncologists in radiotherapy treatment of pancreatic cancer. METHODS A multi-institutional contouring dummy-run of two different cases of pancreatic cancer treated by postoperative and preoperative radiotherapy (RT) was performed. Clinical history, diagnostics, and planning CT imaging were available on AIRO website (http://www.radioterapiaitalia.it). Participants were requested to delineate CTVs according to their skills and knowledge. Aiming to quantify interobserver variability of CTVs delineations, the total volume, craniocaudal, laterolateral, and anteroposterior diameters were calculated. Descriptive statistic was calculated. The 95% Confidence Interval (95% CI) for coefficient of variation (CV) was estimated. The Dice Similarity Index (DSI) was used to evaluate the spatial overlap accuracy of the different CTVs compared with the CTVs of a national reference Centre considered as a benchmark. The mean DSI (mDSI) was calculated and reported. RESULTS A total of 18 radiation oncologists from different Institutes submitted the targets. Less variability was observed for the Elective CTV rather than the Boost CTV, in both cases. The estimated CV were 28.8% (95% CI: 21.2-45.0%) and 20.0% (95% CI: 14.9-30.6%) for the Elective CTV, in adjuvant (Case 1) and neoadjuvant (Case 2) case, respectively. The mDSI value was 0.68 for the Elective CTVs in both cases (range 0.19-0.79 in postoperative vs range 0.35-0.79 in preoperative case). The mDSI was increased to 0.71 (Case 1) and 0.72 (Case 2) if the observers with a worse agreement have been excluded. On the other hand, a CV of 42.4% (95% CI: 30.1-72.4%) and 63.8% (95% CI: 43.9-119.2%) with a mDSI value of 0.44 and 0.52, were calculated for the Boost CTV in Case 1 and Case 2, respectively. CONCLUSIONS The CV and mDSI obtained values for Elective CTVs showed an acceptable agreement among participants either in postoperative as well in preoperative setting. Additional strategies to reduce the variability in Boost CTV delineation need to be found and promoted.