Following Anika's post, this is what they say at pages 19-20 of the provisional pdf at:
(In relevant part it saysIf case control matching was truly maintained those 22 persons with CFS from the Georgia dial-up survey will have exactly the same proportion of cortisol levels as the larger group in Reeves' prior CFS cortisol study in Reference 2. This should be easy to prove or dis-prove, and it would be fun to see if they are lying about this. After all, the cortisol testing had already been done prior to this new paper, and is part of the patient's medical history.
(In relevant part it says
and 22 of 32 persons with CFS and 30 of 51 healthy controls from Georgia....Clinical and demographic characteristics of subjects with specimens available for this study did not differ from those persons who did not have ample specimen volumes and case-control matching was maintained.