Although multiple sclerosis is a chronic condition clinical trials and economic evaluations of beta interferon in MS tend to be short term. Thus the majority of analyses suggest that the Cost per QALY for beta interferon lie between £0.5 and £1 million. The PRISM study demonstrated sustained effectiveness of interferon beta-1a for relapse and disease progression.OBJECTIVE: To establish the cost effectiveness of interferon beta-1a 22mcg and 44 mcg tiw vs. no treatment in MS. METHODS: Our markov model defines states using the Kurtzke EDSS. States are specified by utility; maintenance therapy cost; relapse probability; relapse cost; relapse utility; relapse duration; progression probability; and cost of interferon beta-1a. The model has an annual cycle and a 15 year time horizon. Therapy stops at EDSS 6.5. Relapse and progression probabilities, and relapse duration were taken from the PRISM study. Utilities, maintenance therapy and cost of relapse in each state are taken from Parkin et al. The relapse utility was assumed to be utility in remission minus 0.5. The interferon beta-1a is the UK list price. Sensitivity analyses were undertaken on all parameters. Monte carlo simulation was used to explore the uncertainty in the sampled parameters.RESULTS: The central estimates of the cost per QALY for beta interferon therapy are £45,100 for 22mcg and £49,600 for 44mcg. If patients who develop neutralising anti-bodies stop therapy the costs per QALY become £42,400 and £46,400 respectively. These results are sensitive to the cost and duration of therapy. CONCLUSION: The cost per QALY for long-term therapy with IFN beta-1a is much better than published estimates suggest; between £30,000 and £65,000. The significant reduction reflects the evidence of sustained effect provided by the PRISM study and the appropriate modelling of long term treatment for a chronic condition. Decision makers and pharmaceutical companies need to be aware that in chronic conditions short term estimates are likely to be a poor basis for decision making.