2014 - Alicante - Spain

PAGE 2014: Drug/Disease modeling - CNS
Mats Karlsson

Item response theory for analyzing placebo and drug treatment in Phase 3 studies of schizophrenia

Elke H.J. Krekels (1), Ana Kalezic (1), Lena E. Friberg (1), An M. Vermeulen (2), Mats O. Karlsson (1)

(1) Department of Pharmaceutical Biosciences, Uppsala University, Uppsala, Sweden, (2) Janssen Research & Development, a division of Janssen Pharmaceutica NV, Beerse, Belgium

Objectives: In schizophrenia, the effects of placebo and drug treatment are often quantified using total PANSS scores. Treatment effects on the composite positive, negative, and general PANSS subscales have been quantified as well. In the current analysis, item response theory (IRT) is used to quantify treatment effects for each subscale based on PANSS item-level data.

Methods: Longitudinal data were available from 3 Phase 3 studies with a scheduled duration of 43 days, including 344 patients on placebo treatment and 948 patients on paliperidone with daily doses between 3 and 15 mg. Pre-dose data from 358 patients in comparator treatment arms were also available. Based on all baseline data, item characteristics curves (ICC) were constructed describing the probability of each score for each item as a function of an (unobserved) disease state that was modeled as a random effect, normalized to zero for the typical patient. For each patient different disease states were estimated for the positive, negative and general subscales. Changes in disease states over time were modeled using traditional placebo and drug effect models.

Results: On all 3 subscales, the placebo effect was best described by an asymptotic time-course model. The maximum placebo effect in the typical individual was highest on the positive subscale, for the general subscale the typical placebo effect was slightly lower, while on the negative subscale this maximum placebo effect was half that on the positive subscale. The drug effect was best described using a linear dose-effect relationship, in which the slope increased linearly in time from 0 to an estimated maximum at 43 days. For the typical patient, this maximum drug effect was about twice as high on the positive subscale, compared to the negative and general subscale.

Conclusions: In IRT analyses, changes in unobserved disease states rather than observed scores are modeled, complicating a direct comparison with previous findings. However, our results are in line with previous findings that both placebo and drug treatment most strongly improve positive schizophrenia symptoms.

Acknowledgement: This work was supported by the DDMoRe (www.ddmore.eu) project.

Reference: PAGE 23 (2014) Abstr 3145 [www.page-meeting.org/?abstract=3145]
Poster: Drug/Disease modeling - CNS