> I also wanted to remove age and gender effects at the same time. It seems 3dttest++ doesn't work well with that.

Typically I'd not treat a subject-grouping variable such as gender as a nuisance factor that is simply added in the model but won't be looked at in the results or discussed when writing up. If there is gender effect, that fact warrants some discussion in science per se. In addition, from modeling perspective, an explanatory variable in the model accounts for some data variability no matter how you treat or name it (of interest or not).

With that in mind, sure you can still use 3dttest++ to analyze your situation. Basically you have two groups of subjects (one discrete variable, using option -setA and -setB in 3dtttest++); in addition, you have two continuous variables (using option -covariates in 3dttest++). The subtle issues for you are:

1) Do the two genders have the same or different average blood test score?
2) Do the two genders have the same or different average age?

If the answer for the above two questions is YES, you have a simple case and proper centering should be easy; otherwise the analysis would be a little trickier. The word "remove" in covariate modeling, although very popularly used in literature, is simply a misnomer, which leads to confusion and mistakes in modeling. You can't just "remove" some confounding effect, and then forget about it.

Gang