Aim To determine the strengths and limitations of current approaches used in economic analyses of advanced cancer care.

Methods We conducted a systematic review of EMBASE, AMED, PsycINFO, CINAHL, and MEDLINE from inception to June 2010 using search terms advanced or terminal illness, cancer, and cost or economic studies. We included studies that reported economic evaluations using decision analytical models, within patients with advanced cancer, with any intervention and published in English. We extracted data on health system context, model type, model time horizon and model outcomes.

Results 28 papers were eligible for inclusion, the majority of these were undertaken in Europe or USA. 17 Markov models and 10 decision trees were identified. Most (z) models used lifetime horizons but (x) used curtailed horizons without any clear rationale between studies for time horizon selected. Model outcomes varied; most used quality of life (QOL), but some used survival, and one estimated the likelihood of effective pain management. Measurement of QOL was the most commonly cited limitation in all models. (Z) studies relied on healthcare professionals' valuations of QOL rather than patients' own valuations.

Conclusion We found no consistent model type or time horizon used in studies assessing cost effectiveness of interventions in advanced cancer. In models with shorter durations, additional future costs such should be considered such as those related to further operative procedures, or palliative treatments. The validity of these models may be weakened by using utility values that are not derived directly from patients with advanced cancer.