Following experience with several clinical trials, the semantic, logical, and statistical properties of the Clinical Global Impressions Scale (CGI-ECDEU, 1976; CIPS, 1986) were investigated. It was concluded that the CGI in its present version (a) may induce inconsistent rating behavior, (b) contains redundant information, (c) consists of items which exhibit extremely abnormal distribution properties. Furthermore (d), some of these items are inappropriately constructed or (e) cast some doubts on their significance as measurements. It is suggested that item #1 (seriousness of disease) be used at the beginning of a study only, and that item #2 (global improvement) be used sparingly and as single measurement to express the doctor's general impression. There would seem to be no basis for the judgment of a pure drug effect (item #3). The instruction to measure improvement by comparing the patient's present condition to that at admission to the study cannot be recommended. Poor distribution properties and a presumably restricted significance of change ratings suggest that parametric statistics should not be referred to.

Doctors’ prescription and dosing behaviour was investigated using data from 9 clinical trials in 550 patients treated with psychotropics. 7 trials were conducted under double- and 2 under single-blind conditions. In 3 of these trials, oral and i.m. preparations were used demanding a double-dummy design. All patients were evaluated on a weekly or 2-week basis using psychopathological rating scales (i.e. Hamilton Anxiety Scale, Hamilton Depression Scale, Clinical Global Impressions, Simpson and Angus EPS). It was found that (a) oral-medication titration was 3- to 4-fold more broad-ranging than i.m. medication titration, (b) oral placebo was titrated to the same extent as the oral investiga-tional drugs, and (c) the titration schedule did not follow protocol requirements. Moreover, the average doses in all drug and placebo groups were the same. Concomitant medication like sleep inducers was found to be more closely related to doctors’ habits than to actual medical need. Independent of trial and investigational drug, 10–33% of all patients received additional sleep inducers.