Abstract

The modern debate about schizophrenia began over 100 years ago, with Kraepelin’s description of “dementia praecox”. Despite this, central aspects of the disease remain mysterious and the disease itself is still associated with a high probability of an enduring limitation of the patient’s quality of life. While several conceptions of schizophrenia exist and are still under discussion, at least a provisional consensus regarding a valid measure of schizophrenia seems to have been reached: The Positive and Negative Syndrome Scale (PANSS) quantifies the current state of a person with schizophrenia by combining 30 different schizophrenia-associated symptoms into a single scale value. Even though the scale is widely used and is the measure of choice in many clinical trials, its psychometric properties are still the reason for serious confusion. In many research papers, one important fact about the PANSS is overlooked: it is an interval scale and, therefore, straightforward calculations of proportions are not appropriate. In other words, calculating simple percentage changes is incorrect and a prior scale correction is required. These kinds of calculations often appear in conjunction with responder analyses, as the definition of response is usually based on a predefined cut-off in terms of percent scale change. Two of the presented papers of this thesis are dealing with this urgent problem: using real data as well as simulated data sets, it is shown that ignoring the scale level of the PANSS can, in many cases, even lead to false test decisions concerning an examined treatment effect. Furthermore, an analysis of the problem’s urgency with regard to academic discussions, performed by way of a systematic study of literature in the highest-ranked journals dealing with schizophrenia, showed that incorrect calculations are widespread in the literature and that there is a strong need for a general clarification. As incorrectly calculated percent changes might be a reason for the published low cut-offs of response, as e.g. 20% or 30% cut-offs, the third included article in this thesis analyzes the association of correctly calculated percent changes in the PANSS with a generally measured therapy response. An equipercentile linking of percent PANSS changes and the improvement item of the Clinical Global Impression Scale (CGI) confirmed the choice of a considerably higher response cut-off of 50%. The combined conclusion of the three included articles is the emphasis on the need for a general methodological consensus in schizophrenia research. Valid and replicable research is only possible on the basis of generally accepted methods that rely on the correct application of scale theory in these studies.