A note on the selection of control groups and control variables in comorbidity research

Recently, investigators have become interested in "comorbidity" in eating disorders (EDs). However, it appears that many investigators fail to appreciate that demonstrating that ED patients have more of a given type of concomitant psychopathology than would be expected by chance requires the rejection of the following two null hypotheses: H0-1, Persons with EDs have no more symptomatology of the putatively comorbid condition than do persons without ED; and H0-2, Persons with EDs have no more symptomatology of the putatively comorbid condition than they have other kinds of psychopathology not specific to ED. The choice of proper control group for testing H0-1 and the necessity of testing H0-2 are illustrated with simulated data. These same general issues apply to the assessment of comorbidity in other conditions, but the appropriate comparison group obviously changes according to the exact hypothesis being tested.