OBJECTIVE: To compare model estimates of the cost-effectiveness of latanoprost and travoprost versus timolol. METHODS: The probability of developing a visual field defect (VFD) was estimated using data from a double-masked double-dummy Phase III multi-centre clinical trial comparing travoprost 0.004% od, latanoprost 0.005% od and timolol 0.5% bid, and the 2 discriminant functions published by Stewart (1993). A Markov model was constructed to estimate the cost and effectiveness (time to disease progression) of patients treated over five years for two states: those who do and those who do not develop a VFD. Resource utilisation was estimated from a 5-year retrospective patient chart analysis in France, from the UK General Practitioner Data Base in the UK, and from expert interviews in Germany. Both costs and outcomes were discounted at a 5% rate. The economic perspective was that of the Heath Care Sick Fund. RESULTS: The average time to disease progression was estimated to be 2.81 years with timolol, 3.28 with latanoprost and 3.42 with travoprost. In France, over 5 years, a latanoprost treated-patient is predicted to cost 277.6 more than a timolol-treated patient while the figure for a travoprost-treated patient is EUR 403.7. In Germany, these figures were €858.60 and €714.80, for latanoprost and travoprost respectively. The corresponding amounts in the UK were €288.9 and €251.1. In the UK and in Germany, travoprost was predicted to be more effective and less costly than latanoprost. Comparing the prostaglandin analogues in France, travoprost's incremental additional 5 year cost is modelled to be €951.4 per additional disease free year. CONCLUSION: According to our model, travoprost is a cost-effective alternative to timolol and latanoprost.