Purpose: We sought to categorize longitudinal radiation-induced rectal toxicity data obtained from men participating in a randomised controlled trial for locally advanced prostate cancer. Materials and methods: Data from self-assessed questionnaires of rectal symptoms and clinician recorded remedial interventions were collected during the TROG 96.01 trial. In this trial, volunteers were randomised to radiation with or without neoadjuvant androgen deprivation. Characterization of longitudinal variations in symptom intensity was achieved using prevalence data. An integrated visualization and clustering approach based on memetic algorithms was used to define the compositions of symptom clusters occurring before, during and after radiation. The utility of the CTC grading system as a means of identifying specific injury profiles was evaluated using concordance analyses. Results: Seven well-defined clusters of rectal symptoms were present prior to treatment, 25 were seen immediately following radiation and 7 at years 1, 2 and 3 following radiation. CTC grading did not concord with the degree of rectal ‘distress’ and ‘problems’ at all time points. Concordance was not improved by adding urgency to the CTC scale. Conclusions: The CTC scale has serious shortcomings. A powerful new technique for non-hierarchical clustering may contribute to the categorization of rectal toxicity data for genomic profiling studies and detailed patho-physiological studies.
Radiotherapy and Oncology Vol. 90, Issue 3, p. 400-407