Evaluation of interobserver and interscale agreement in assessing late bowel toxicity after pelvic radiation in patients with carcinoma of the cervix.

Background: Lack of agreement and inconsistency in capturing late bowel toxicity may be a source of error while reporting trials with toxicity endpoints. Documenting baseline inconsistency while scoring toxicity questionnaires (RTOG/EORTC and CTCAE) may be worthwhile. The present study was conducted as a quality assurance measure prior to initiating a randomized trial (PARCER; NCT01279135) that evaluates the impact of image-guided radiation on bowel toxicity.

Methods: From August 2010 to July 2011, patients with cervical cancer who underwent pelvic chemoradiation >6 months ago, with controlled disease and any bowel symptom at follow-up, were included. RTOG and CTCAE questionnaires were filled by two blinded observers. Interscale (RTOG vs CTCAE) and interobserver (observer A and B) agreement were evaluated with Spearman's correlation and kappa statistic.

Results: Fifty-five patients were included. Twelve patients with symptoms could not be graded by the RTOG scale. Of those graded as asymptomatic on RTOG, distension, vomiting, pain and nausea were identified as the most common symptoms. Amongst these, grade 1, 2 and 3 toxicity was observed in 6, 5 and 1 patient, respectively. The interscale correlation was moderate (Spearman's correlation ρ = 0.56; P = 0.001). High interobserver agreement (92%) was observed within the RTOG scale [kappa (κ) -0.94; 95% CI 0.77-1]. All disagreements were observed while scoring grade 1-2 toxicity. Among CTCAE, agreement was lower with modules such as distension, anorexia, nausea and constipation.

Conclusions: High interobserver agreement was observed for both RTOG and most CTCAE subscales; most disagreements were for grade 1-2. Interscale agreement (RTOG and CTCAE) was moderate. Detailed patient interrogation or use of patient-reported-outcome scores while documenting the aforesaid subscales may be worthwhile.