[DSM IV and research. The problem of clinical trials]

Like its predecessors, DSM IV allows the prospective study population for clinical trials to be defined as accurately possible. The inclusion and exclusion criteria allow clear definition of the diagnostic category to which patients belong, and the simplest way of allowing the clinician to find his way through the criteria under consideration is to place the latter in a "checklist". The clinician then checks whether or not a given criterion is present, and after completing the checklist calculates whether the checked criteria are sufficient to warrant inclusion of a given patient. This approach, though methodical nevertheless, contains a significant drawback. The diagnostic criteria of DSM IV are at once "positive" (symptoms which a patient must present to qualify for inclusion in a diagnostic category) and "negative" (symptoms which the patient must not present in order to qualify for inclusion). In some cases it is relatively simple to establish the existence of such negative criteria (e.g. cases in which it must be ascertained that symptoms are due neither to ingestion of a toxic substance nor to recent bereavement). However, very often these negative symptoms are not presented one by one in the list of criteria, but are instead grouped together in an allusive phrase such as: "the symptoms must not be due to any associated disorder", or, "the symptoms must not occur exclusively during episodes of an associated disorder", or again, "the criteria of such and such a disorder are not met". Thus, in theory, each time a negative criterion of this type is encountered, the investigator must refer to the disorder for which diagnosis constitutes an exclusion criterion and verify that the conditions of this disorder are not met. For example, criterion B for major depressive episodes states that "the symptoms do not meet criteria for a mixed episode". It is therefore necessary to verify that the exclusion criteria are indeed absent. This type of "Chinese box" procedure thus implies that checklists should not only contain the diagnostic criteria for the particular disorder being sought, but that they should include a complete list of diagnostic criteria for all other disorders requiring exclusion. Since such an awkward procedure would make it impossible to draw up reasonable case report forms, a selection of exclusion criteria is generally made, often on a somewhat arbitrary basis. It might in fact prove more useful to adopt diagnostic algorithms.