OBJECTIVES: Clinical trials of oncology products often allow crossover of control patients to the treatment arm following disease progression.  This can underestimate a product’s overall survival (OS) and raises challenges for cost-effectiveness analyses compared.  METHODS: A lifetime model compared the cost-effectiveness of a hypothetical pancreatic cancer therapy (CRG001) to gemcitabine.  Gemcitabine survival data were derived from published studies.  A hazard ratio of 0.55 was assumed for CRG001.  CRG001 cost $2500 every 2 weeks for a maximum of 12 cycles and gemcitabine cost $200 every 1 week for a maximum of 24 cycles.  Analyses were conducted: 1) 0%; 2) 50%; and 3) 85% crossover of gemcitabine patients to CRG001.  Patient crossover occurred at the time of disease progression.  Crossover patients received the CRG001 hazard ratio.  Patients progressing in CRG001 were assumed to receive palliative care.  A secondary analysis allowed 50% crossover but excluded second-line costs of CRG001.   Costs and outcomes were discounted at 5%. RESULTS: The cost per QALY gained for CRG001 compared with current care was $81,352 with no cross-over, $69,292 with 50% crossover and 40,992 with 85% crossover.  In the analysis where the cost of CRG001 was excluded in crossover patients, the ratio was $140,118. CONCLUSIONS: The first three analyses illustrate that CRG001 cost-effectiveness decreases with increasing cross-over of gemcitabine patients, if the costs of CRG001 for crossover patients are included.  In our experience, however, reimbursement agencies often require a primary analysis that excludes second-line costs of the study drug for patients that cross-over.  This analysis yields a high ratio that could lead to a negative reimbursement decision.  In this case, where second-line CRG001 costs are excluded, adjustment of OS for crossover of gemcitabine patients is required. Overall, consideration must be given to the extent and potential impact of crossover when conducting cost-effectiveness analysis of new oncology products.