Prediction interval generalized estimating equations

[mgx]

Because I have time dependent covariates and the full covariate conditional mean assumption (Pepe and Anderson, 1994) does not hold, I have fitted a continuous longitudinal response with generalized estimation equations and independent working correlations in PROC MIXED. My code looks like this

proc mixed data=dataset EMPIRICAl ;
class id timeclss;
model responde = time lag1 lag2 lag3 /solution COVB;
repeated  /type=simple subject=ID group=timeclss;
run;

Is it possible to construct prediction intervals for this type of marginal model?

Kind regards, 

Margaux