Adjusting for Age in GWAS where case and control are not age matched.

Question

[Mel]

Hi All,

I'm conducting single variant analysis (i.e. logistic regression) on exome chip data. Normally one would adjust for age, however, my cases are all <65yrs of age and my controls are >65yrs of age. My thought is to not adjust for age in this data set, as any true associations that occur will appear as though they are associated with age, so adjusting for age will hide those associations.

Am I right or wrong in thinking this?

Thanks in advance

Answer 1

[brentp]

It will depend on (among other things) how big the age ranges are within each group,

You could do the analysis as you suggest, without age, then, on variants that meet your cutoff for significance, you could check for a relationship with age within each group separately.

[Mel]

Thanks brentp, good suggestion, I'll be sure to do this.

Regarding the age ranges..

Cases have a mean age of 56.8 plus or minus 5.5 standard deviations

Controls have a mean age of 75.1 plus or minus 7.2 standard deviations