2013 - Glasgow - Scotland

PAGE 2013: Other Drug/Disease Modelling
Ana Novakovic

Application of Item Response Theory to EDSS Modeling in Multiple Sclerosis

Ana Kalezic (1), Radojka Savic (2), Alain Munafo (3), Sebastian Ueckert (1), Mats O Karlsson (1)

(1) Dept of Pharmaceutical Biosciences, Uppsala University, Uppsala, Sweden

Objectives: Traditional approaches to measurement scales generally disregard the underlying nature of the subcomponent data. In contrast, item response theory (IRT) refers to a set of mathematical models that describe, in probabilistic terms, the relationship between a person's response to a survey question and its level of the "latent variable" being measured by the scale [1]. In the area of clinical pharmacology, IRT modeling has previously been applied to ADAS-cog assessments [2].
The objective of this analysis was to apply IRT methodology to the Expanded Disability Status Score (EDSS) [3], a widely used measure of disease disability in multiple sclerosis (MS).

Methods: Data were collected from a 96-week Phase III clinical study with relapsing-remitting MS. For this analysis, 41664 EDSS observations from 1319 patients at baseline or treated with placebo were used.
The assumption is that the outcome of each item constituting EDSS depends on an unobserved variable "disability."
Unlike most measurement scales, EDSS total score does not result from simple addition of individual items, but instead results from individual components via a decision tree.
For each EDSS item, a model was developed in accordance with the nature of data, describing the probability of a given score as a function of disability variable. Sets of parameters characterizing each item were modeled as fixed effects, while the MS disability was modeled as subject-specific random effect with or without time components. All models were fitted using NONMEM 7.2; simulation-based diagnostics for model evaluation also used PsN and R/Xpose4 software.

Results: The final model contained 8 ordered categorical submodels for a total of 54 parameters. Simulations from the IRT model were in good agreement with the observed EDSS and item-level data. The disability variable showed a significant increase (p<0.01) over time in the typical individual, but with considerable variability across patients.

Conclusions: This is the first time that the IRT methodology has been applied to the MS area and to a score that is not a summation of items. For the latter type, IRT models have been shown to increase precision in predictions and power to predict drug effects and linkage to biomarkers [4, 5]. The developed model can be used to explore potential benefits of the IRT methodology for characterizing MS disability.

Acknowledgement: This work was supported by the DDMoRe project.

References:
[1] Reeve BB, Fayers P. Applying item response theory modeling for evaluating questionnaire item and scale properties. In: Fayers P, Hays RD, editors. Assessing quality of life in clinical trials: methods of practice. 2nd ed. Oxford University Press, 2005. p. 55-73.
[2] Ueckert S, Plan E, Ito K, Karlsson M, Corrigan B, Hooker A. Application of Item Response Theory to ADAS-cog Scores Modelling in Alzheimer's Disease . PAGE 21 (2012) Abstr 2318 [www.page-meeting.org/?abstract=2318]
[3] Kurtzke JF, Rating neurologic impairment in multiple sclerosis: an expanded disability status scale (EDSS), Neurology 33 (11):1444-52, 1983
[4] Ueckert S. ACOP 2013 Abstract
[5] Balsis S, Unger A, Benge J, Geraci L, Doody R. Gaining precision on the Alzheimer's disease assessment scale-cognitive: A comparison of item response theory-based scores and total scores. Alzheimers Dement 2012; 8:288-294




Reference: PAGE 22 (2013) Abstr 2903 [www.page-meeting.org/?abstract=2903]
Poster: Other Drug/Disease Modelling
Click to open PDF poster/presentation (click to open)
Top