Prediction of Placebo Responses by Multiple Regression Methods and CFA. Re-Analysis of an Experiment Conducted by W. Janke between 1963 and 1966
Janke, Wilhelm, Ising, Marcus, Psychology Science
This paper is based on experimental studies on the psychological effects of placebo and their predictability by personality characteristics, which were conducted by the first author in prepration of his habilitation thesis between 1963 and 1966 in Marburg and Gie[beta]en. The first author describes his interactions with Gustav A. Lienert, which preceded the reported studies. He also demonstrates that the original analysis of the data was already inspired by Lienert's idea of a configural approach, although Lienert's first presentation of the Configural Frequency Analysis (CFA) was still forthcoming at that time. A re-analysis of the data confirms that a configurai approach like the CFA is of high value for the identification of personality patterns, which are predictive for placebo responders (PR) and non-responders (PNR). A bi-prediction analysis shows that PR and PNR are characterized by complementary patterns of specific personality traits.
Key words: Configurai Frequency Analysis (CFA), bi-prediction analysis, placebo response
A) Remarks on the background of the studies and Lienert s influence on them between 1962 to 1968
Between 1963 and 1966, when the author was post-doctoral assistant in Marburg and Gie[beta]en, he performed a series of experiments on the effects of placebo and their determinants. These studies formed the basis of his habilitation thesis, which was submitted to the "Mathematisch-Naturwissenschaftliche Fakultat" of the university of Gie[beta]en in autumn 1966 and approved by the faculty in May 1967 (referees were the psychologist Karl-Hermann Wewetzer and the biochemist Hans-Jurgen Staudinger.)3. A part of the work was published in 1986(Janke, 1986).
The studies were based on the assumption that responses to placebo are predictable only by measures from different areas, e.g. factors of drug administration, the situational context and personality characteristics (traits and states).
The study, which is reported in (his paper, included as predictor-variables broad personality characteristics like neuroticism, cxtraversion and narrow traits like suggestibility, aggressiveness after frustration, attitudes towards drugs, expectation of drug effects, and experience with the use of drugs. Criterion-variables were subjective reports of mood and performance-measures. The subjective placebo-responses were assessed directly, from reports of the observed changes from baseline, as well as indirectly, from the reported state before and after the administration of placebo. The placebo responses were defined by dichotomous (Responder - Non-Responder) and by continuous measures.
Gustav Lienert (GL) influenced these experiments in several ways. His paper "On the role of suggestion in pharmacopsychological studies" (Lienert, 1955) had a great impact for the author's interest in placebo-effects. Their study in relation to personality characteristics can also be considered as a continuation of Jankers dissertation "Zur Abhangigkeit der Wirkung psychotropcr Substanzen von Pcrsonlichkeitsmerkmalen: Ein Beitrag zur Begrundung der Differcntiellcn Pharmakopsychologie" (On the dependency of the effects of psychotropic drugs on personality characteristics. A contribution to the foundation of a differential pharmacopsychology) (Janke, 1964). GL was Janke's mentor during the dissertation period from 1957 to 1961. he also proposed the subtitle of the doctoral thesis.
GL also played a considerable role during the planning of the experiments. he gave advice and made proposals. In 1963, when GL left Marburg and went to Hamburg and I (W. Janke) changed to Gie[beta]en, the topic "placebo" had been discussed with him with respect to the habilitation thesis several times.
A major role of GL on my thinking becomes evident from another point: I was convinced that the prediction of placebo responses by means of (raits would not be successful by a linear combination of trait-measures, only, and therefore searched for other models. …