I agree with Adrian that you need to look at the test for the interaction term.  However, I believe it is customary to include all lower order terms for variables involved in an interaction.  So I would include a first-order term for diabetes in the second model, like this:

p = b0 + b1g + b2d + b3gd + error

In this model, the test on b3 would be the test for effect modification. 

p.s. - Adrian, how did you make subscripts in here?

The usual formal test of effect modification is by test of the interaction term. In other words, if p represents phenotype, g gentotype and d presence of diabetes, then I assume your base model looks like:  

p = b0 + b1g + error, and you would add the term

p = b0 + b1g + b2dg + error

if b2 is statistically significant, then effect modification is present

I agree with Adrian that you need to look at the test for the interaction term.  However, I believe it is customary to include all lower order terms for variables involved in an interaction.  So I would include a first-order term for diabetes in the second model, like this:

p = b0 + b1g + b2d + b3gd + error

In this model, the test on b3 would be the test for effect modification. 

p.s. - Adrian, how did you make subscripts in here?

Hi Ferdinamarie,

I agree with Bruce.

As you are asking how to do it on SPSS, it might be helpful to calculate (transform/compute variable) an interaction variable by simply multiplying g*d. Once you have done this you can incorporate the new variable into the model as Adrian and Bruce have shown.

Thank you All for your insights, it was very helpful.