OBJECTIVE: To investigate the extent to which claims of efficacy in applications for requesting the addition of new drugs to the formulary of the Australian Pharmaceutical Benefits Scheme have been based on surrogate outcome data. METHODS: Analyses were drawn from a database of all submissions for new drug/indication pairs considered between 1993 and 1998. Applications were classified according to whether the data presented were based on clincal, intermediate or surrogate endpoints. Drugs utilisation and expenditure data were examined for those drugs subsequently listed. RESULTS: Of more than 400 submissions considered, 216 were initial submissions for drug/indication pairs and 150/216 [69%] were subsequently recommended for addition to the formulary. Of those recommended for listing, 36/150 [24%] applications presented claims of efficacy data that were wholly dependent on surrogate endpoint data, and 68/150 [45%] submissions relied on a combination of surrogate with intermediate and/or clinical endpoint data. Seventy-four (49%) submissions presented some clinical endpoint data, but only 30/150 (20%) submissions were based on clinical endpoint data alone. In calendar 1999 more than AUD 420 million was spent on drugs whose efficacy had only been demonstrated against surrogate endpoints, approximately one sixth of annual expenditure. DISCUSSION: Decision-makers must at times consider the adoption of therapies for which only surrogate outcome data are available. Uncertainties inherent in estimates of comparative efficacy and cost-effectiveness are magnified when data are drawn from clinical trials that report only surrogate outcome data. Comparative cost-effectiveness analysis cannot give rise to technical or allocative efficiencies if the criteria on which assessments are based are unreliable.