I would like to run a multiple linear regression to test if my protein of interest differs among different stages of a disease, while controlling for covariates/ confounders such as age and gender. The disease stage variable has four levels. I would like to compare the "control" stage with the three remaining stages. From what I understand, once I specify the disease stage variable as categorical and specify the reference group ("control"), the multiple regression output will compare the mean of the "control" group with each disease stage. I believe this is similar to running an ANCOVA. However, in ANCOVA analyses, a post-hoc test with an adjustment for multiple comparisons is performed (such as Bonferroni). I am aware that some statistical softwares (such as SPSS) perform post-hoc comparisons on unadjusted group means.
I have noticed that my predictor "disease stage" is significant in my multiple regression analysis. Therefore, I am wondering if it is possible to run a regression each time I would like to separately compare the mean of the control group to another, while controlling for the covariates. Finally, I would apply the Benjamini-Hochberg (FDR) procedure to the p values for the group differences. I understand some softwares (such as JMP) provide the regression output with all the comparisons with the reference group. However, I would think that it would be more appropriate to exclude certain pairwise comparisons from the regression. Such that we only focus on one comparison in the "post-hoc" regression model? I hope my last point makes sense.