Can Administrative Data Improve the Performance of Cancer Clinical Trial Economic Analyses?

Abstract

PURPOSE\nTrial economic analyses, such as cost-effectiveness analysis, often rely on trial-collected data, which are burdensome and expensive to collect and may be incomplete. In contrast, administrative databases systematically collect health system encounters. We investigated whether administrative data could improve the performance of cancer trial economic analysis.\n\n\nMETHODS\nHealth administrative data were probabilistically linked to Ontario patient data from the Canadian Cancer Trials Group CO.17 trial (n = 572), which evaluated cetuximab plus best supportive care (75 linked Ontario patients) versus best supportive care alone (73 patients) in previously treated metastatic colorectal cancer. Trial-collected resource utilization data and vital status were compared with administrative data. Cost effectiveness in 2007 Canadian dollars was determined with bootstrap incremental cost-effectiveness ratio (ICER) CIs.\n\n\nRESULTS\nUp to trial date of last contact, administrative data vital status was concordant in more than 96%. Twenty-nine subsequent deaths occurred. Up to trial last contact, there were 50 net additional hospitalizations in administrative data and 33 net additional emergency department visits. Total costs were $3,023,034 for the cetuximab group and $1,191,118 for the control group up to trial last contact. The ICER was $211,128 per life-year gained (90% CI, $101,396 to $694,950) up to trial last contact and $164,378 (90% CI, -$138,260 to $644,555) up to administrative data last contact. ICER estimates were similar to the analysis using trial-collected data.\n\n\nCONCLUSION\nAdministrative data were more complete than trial data for hospital encounters, a key cost driver in economic analysis. There was a longer follow-up. This demonstrates the potential of administrative data to relieve the burden of collecting key data in cancer trials, which represents a considerable effort and expense.