Method: The OPTIMA study (an ongoing RCT of treatment strategies for patients with advanced HIV disease for whom standard therapies have failed) measures the Health Utilities Index (HUI2 and HUI3) and the EQ5D at 3 monthly intervals. While both instruments were administered at the same time, the HUI used a recall period of the past week, whereas the EQ5D asked about health status on the day of administration. In order to identify whether an acute adverse event was captured by the utility instruments, we classified assessments into three groups; (i) when a non-AIDS related serious adverse event (SAE) was ongoing and unresolved (so the recall period was applicable for both the HUI and EQ5D), (ii) when an SAE had resolved within the past 7 days (recall period applicable only for HUI), and (iii) when no SAE had occurred within the past 7 days. We fitted a Generalized Linear Model (using GEE) including the instrument type, the assessment type and their interaction terms. The effect of recall time was estimated by the examining the difference between the HUI and EQ5D in assessing events resolved in the past 7 days, after negating the effect of instrument difference estimated by assessing events which were unresolved.

Results: Among 329 patients enrolled as of March-2005, there were 295 assessments identified as type (i) and 62 as type (ii). The adjusted effect of recall time was found to be statistically significant (HUI2vsEQ5D=-0.152[-0.257,-0.048] and HUI3vsEQ5D=-0.215[-0.343,-0.088]). As anticipated, the comparison between HUI2 and HUI3, where recall time is identical, showed no statistically significant recall effect (-0.063[-0.022,0.148]).

Conclusions: This study demonstrates that patients accurately adhered to the length of recall requested. The instrument with the shorter recall (EQ5D) did not capture the impact of all adverse events. Whether a longer recall period than 1 week would be more appropriate in this disease group needs further research since patients might reflect on their recovery rather than the event. Validation of these findings is warranted in a new study which should use the same instrument but include multiple recall times of different lengths.