OBJECTIVES: Due to wide variations in prescribing practice, a key methodological challenge for an economic analysis in oncology is describing current practice. Robust methods of eliciting the full range of expert opinion are necessary to ensure clinical relevance and validity of economic models. Insufficient information about variations in management can lead to model underestimation of uncertainty. This study aims to determine the appropriate model structure for an economic evaluation of a pharmacogenetic test to inform irinotecan prescribing, by identifying current practice. METHODS: Expert opinion was elicited to inform model structure. Through semi-structured postal survey (May-August 2008), clinical experts were asked to describe: the general management of advanced CRC patients on chemotherapy, place of irinotecan-based regimens within the clinical pathway and management of patients on these regimens. Since data on the frequency of key adverse events (neutropaenia and diarrhoea) from irinotecan-based regimens are not readily available, experts were asked to provide frequency estimates and resources associated with management. Descriptive statistics were used to summarise the data. RESULTS: The final expert panel comprised 44 oncologists representing all English NHS regions, from a sample of 98 potential experts. Wide variations in prescribing practice were identified. There were ten first-line, ten second-line and twelve third-line chemotherapy regimens reported, reflecting variation in practice. However, dominant treatment pathways emerged for each stage of treatment, with: first-line treatment with oxaliplatin-based regimens, second-line treatment with irinotecan-based regimens, and third-line treatment with mitomycin-based regimens. Experts estimated the frequency of: febrile neutropaenia 8.4% (95% CI: 6.7 – 10.0), septic neutropaenia 4.7% (95% CI: 3.4 – 6.0), and severe diarrhoea 13.1% (95% CI: 10.8 - 15.5). CONCLUSIONS: This study captured the heterogeneity in chemotherapy prescribing practice and patient management within CRC specialties and obtained necessary data to inform the appropriate model structure for base case and subsequent structural and parameter sensitivity analyses.