The objective of this article was to review the methods used to obtain quality-of-life (utility) weights reported in assessments carried out for the National Institute for Health and Clinical Excellence (NICE).The design of the review was a cross-sectional survey. Health technology assessment (HTA) reports published on the NICE website up to May 2003 were reviewed. Data were extracted on the following: the approach to utility estimation (direct or indirect), how health states were described for indirect estimation, valuation techniques used (standard gamble [SG], time trade-off [TTO], visual analogue scale [VAS], etc.), whether uncertainty in utility estimates was explored in cost-utility analyses, and whether utility values were identified as a priority for further research by assessment authors.Fifty-six assessments were reviewed, of which 28 reported 45 cost-utility analyses. There was striking variation in the values used to describe different health states. Data from patients were used in 15 (33%) analyses, from the general public in 10 (22%) and from clinicians in 4 (9%). In 16 (36%) cases, the source for utility estimates was unclear. Health states were described using a range of generic and disease-specific measures, although the EQ-5D was used most frequently. In 25 analyses (56%), the valuation technique used was not reported. TTO was used in 11 (24%), SG in 3 (7%), magnitude estimation in 5 (11%) and VAS in 1 (2%). Sensitivity analyses based on utility values were reported in 25 cases (56%), more commonly in reports of analyses carried out by independent teams than technology sponsors although this may be subject to reporting bias. Further research into quality of life was recommended in 17 (61%) of the 28 assessment reports that contained at least one cost-utility analysis. Greater transparency and consistency are required in reporting the methods used to obtain quality-of-life weights in cost-utility analyses, and better sources of data are required. Methodological variation results in important differences in values. Therefore, caution must be exercised when comparing the results of different cost-utility analyses.