Article type
Year
Abstract
Objective: To systematically study the time aspect of the placebo effect in clinical trials.
Methods: Metaanalysis of placebo-controlled, long-term (duration of study > 3 months) clinical trials in human patients (Medline 1992 & 1993 and hand-search of main journals). Out of 464 abstracts, 26 studies fulfilled our inclusion criteria. We extracted information on disease treated, severity of disease, duration of study, improvement under placebo and under treatment, dropouts, mean age, evaluation according to intention to treat. We regressed study characteristics on the magnitude of the placebo effect, defined as improvement in the placebo group, weighted for study size.
Results: Surprisingly, duration of study was not significantly correlated with improvement under placebo. However, improvement under placebo was significantly correlated with treatment effect (r = .59). In the regression model, (R = .67; R2 = .42; p < 0.0001) 42 % of the variance was explained by two predictors: success rate in the treatment group (beta-weight = .60; p < 0.0001) and evaluation according to intent-to-treat (beta-weight = .37; p < 0.001). Severity of disease and type of disease treated were no significant predictors.
Discussion: Thus, the effect in the placebo group is strongly dependent on the treatment effect and on the type of evaluation. The latter finding points to the fact that part of the placebo effect is artefactual. But there seems also a strong relationship between placebo-effect and treatment-effect. This finding, which incidentally is numerically nearly identical to data reported by Evans (1974), will be discussed and possibly validated against a larger data base, which we are presently building up.
Methods: Metaanalysis of placebo-controlled, long-term (duration of study > 3 months) clinical trials in human patients (Medline 1992 & 1993 and hand-search of main journals). Out of 464 abstracts, 26 studies fulfilled our inclusion criteria. We extracted information on disease treated, severity of disease, duration of study, improvement under placebo and under treatment, dropouts, mean age, evaluation according to intention to treat. We regressed study characteristics on the magnitude of the placebo effect, defined as improvement in the placebo group, weighted for study size.
Results: Surprisingly, duration of study was not significantly correlated with improvement under placebo. However, improvement under placebo was significantly correlated with treatment effect (r = .59). In the regression model, (R = .67; R2 = .42; p < 0.0001) 42 % of the variance was explained by two predictors: success rate in the treatment group (beta-weight = .60; p < 0.0001) and evaluation according to intent-to-treat (beta-weight = .37; p < 0.001). Severity of disease and type of disease treated were no significant predictors.
Discussion: Thus, the effect in the placebo group is strongly dependent on the treatment effect and on the type of evaluation. The latter finding points to the fact that part of the placebo effect is artefactual. But there seems also a strong relationship between placebo-effect and treatment-effect. This finding, which incidentally is numerically nearly identical to data reported by Evans (1974), will be discussed and possibly validated against a larger data base, which we are presently building up.