OBJECTIVES: 1) To quantify and characterize the cost-utility analyses (CUAs) published in the scientific literature through 2003; and 2) to examine methodological practices used in these CUAs. This paper builds upon our previous analyses that evaluated CUAs from 1976 through 2001. METHODS: We systematically searched Medline and the Health Economic Evaluations Database (HEED) for original CUAs written in English and published during 2002-2003 to update a comprehensive registry of CUAs (www.tufts-nemc.org/cearegistry). Two trained readers independently extracted data on the characteristics and methodology of each study. We compared recently published (2002-2003) with previously published (1976-2001) data. RESULTS: Our previous search identified 533 original CUAs published during 1976-2001; a finding of 262 studies published during 2002-2003 increased the sample by almost 50%. In the 1976-2001 and 2002-2003 data, most studies were based in the US (61%, 53%) and written by academics (90%, 92%). Studies examined treatment (63%, 62%) more than primary (14%, 16%) and secondary (22%, 21%) prevention. The most frequent condition studied was cardiovascular disease (21%, 18%). Percentage of funding by the pharmaceutical and device industry did not change substantially (19%, 23%). Studies improved with time across the two datasets: disclosing funding source (65% vs. 71%), stating year of currency (76% vs. 82%), reporting incremental ratios (59% vs. 79%), and using probabilistic sensitivity analyses (9% vs. 28%). CONCLUSIONS: The publication of CUAs has accelerated in recent years. Use of recommended methods has increased, although a sizeable number of analyses still do not adhere to best practice. Continued monitoring of these trends is crucial to enhancing the quality of the field.