Table 5.

Mean EQ-5D-5L utility index by patients’ self-assessed performance status at follow-up

	Performance status				P-value for trend a
Model	0	1	2	3 or 4
Observed data	0.8719	0.7273	0.5941	0.5941
Ordinary least square
Model 3	0.8518	0.7584**	0.6499	0.5886	< 0.001
Model 4	0.8537	0.7581**	0.6518	0.5883	< 0.001
Model 5	0.8560	0.7571**	0.6495	0.5842	< 0.001
Tobit model
Model 3	0.8890	0.7784**	0.6498	0.5772	< 0.001
Model 4	0.8913	0.7780**	0.6520	0.5768	< 0.001
Model 5	0.8944	0.7766**	0.6490	0.5710	< 0.001
Censored least absolute deviation
Model 3	0.8478*	0.7590**	0.6558	0.6004	< 0.001
Model 4	0.8512	0.7580**	0.6545	0.5948	< 0.001
Model 5	0.8505	0.7541**	0.6487	0.5829	< 0.001
Quantile regression
Model 3	0.8478*	0.7590**	0.6558	0.6004	< 0.001
Model 4	0.8512	0.7580**	0.6545	0.5948	< 0.001
Model 5	0.8505	0.7541**	0.6487	0.5829	< 0.001
Logistic quantile regression
Model 3	0.8341**	0.7548**	0.6526	0.5920	< 0.001
Model 4	0.8349**	0.7532**	0.6507	0.5886	< 0.001
Model 5	0.8387**	0.7508**	0.6422	0.5733	< 0.001

^aCuzick’s nonparametric test for trend [29].

^**P-value < 0.01; ^*P-value < 0.05 for signed-rank test between observed and predicted utility index within the same performance status level.