
132 Design & Analysis of Clinical Trials for Economic Evaluation
predicted EQ-5D estimates at 4 months and all subsequent months
should be equal to zero. A model could have been used to incorporate
the censoring feature into QoL data, but this complicates the model con-
siderably. The simple option is to set all EQ-5D values to 0 after death
(which may also have implications for the correlations between EQ-5D
over time).
For this example, the Kaplan–Meier survival times up to 2 years for
Treatment A are as shown in Table5.3 with predicted EQ-5D.
The expected (i.e. mean) QALY (or quality-adjusted survival [QAS]) is
given by
QALY = Σ EQ − 5D
i