Commentary on Cox CL, Lensing S, Rai SN et al. Proxy assessment of quality of life in pediatric clinical trials: application of the Health Utilities Index 3. Qual Life Res 2005; 14: 1045-1056.

The paper by Cox et al. offers an important potential for informing designs of future studies. The rates of missing measurements in the study are much greater than those reported by other investigators. This is not surprising because the nurse assessors were restricted explicitly by the study protocol from collecting required information, had little contact with many of the patients, and some had a poor understanding of the measurement system. Standard multi-attribute system questionnaires, such as the one used in the Cox et al. study, have been designed carefully and require assessments of aspects of health that are not always readily observable (e.g. emotional status). There is an abundance of published evidence, from studies conducted around the world, of acceptable measurement properties for these instruments. However, the published evidence also indicates that measurement properties are compromised with the use of proxy assessors who have little day-to-day knowledge of patients. We conclude that the Cox et al. study results provide evidence that multi-attribute system assessments should be elicited from patients themselves or, failing that, from proxy assessors known to be well-informed about the patients' capacity on all relevant attributes.