Utilizing the items from the MDS-UPDRS score to increase drug effect detection power in de novo idiopathic Parkinson's disease patients
Simon Buatois (1,2), Sylvie Retout (1), Nicolas Frey (1), Sebastian Ueckert (2,3)
(1) Pharma Research and Early Development, Clinical Pharmacology, Roche Innovation Center Basel, Switzerland; (2) IAME, UMR 1137, INSERM, F-75018 Paris, France, Univ Paris Diderot, Sorbonne Paris Cité, F-75018 Paris, France; (3) Pharmetheus AB, Uppsala, Sweden
Objectives: In Parkinson’s disease (PD) clinical trials, total score of clinical rating scales, such as MDS-UPDRS [1], is traditionally used to assess the treatment efficacy [2]. However, the high number of failed clinical trials has led to challenge its use as a clinical end point [3-5]. As a result, pharmaceutical companies are increasingly moving to an analysis of subsets of the scale. These subsets are often selected per items family (e.g. motor items), and do not take the sensitivity of each item to a drug effect into account, leading to a potential loss of power. The objectives of this work was first, to construct a subset of the MDS-UPDRS which maximizes the power to detect a drug effect acting on the disease progression (DEDP) in de novo PD patients, and secondly, to compare those results with the ones obtained with an IRT-based analysis.
Methods: An updated version of the IRT model [6], integrating three different pathophysiological processes with different progression rate, was built to analyze the data from the Parkinson’s Progression Marker Initiative database [7]. Based on this model, clinical trial simulations were performed in de novo PD patients with a balanced, placebo versus disease modifying agent, parallel-arm study design. The power to detect a DEDP was computed under several scenarios (with different study durations and drug effects), through the use of three methods: a summary score based analysis i) of the total number of simulated items ii) of the optimal set of items (determined using a greedy algorithm [8]) and iii) an IRT based analysis.
Results: A three latent variables IRT based modeling allowed an adequate description of the data at both item and total score level taking into account the difficulty and the power of discrimination of each item. Compare to the classical analysis, the power to detect a DEDP was increased when using an optimal set of items (e.g. up to 20% increase for a 50% DEDP). However, it requires an accurate approximation of drug effect prior to the analysis, as the optimal set varied between scenarios. The IRT based analysis increased further the power without the need of items selection.
Conclusions: Selection of the most sensitive items of the MDS-UPDRS score can be used to increase the power of a summary score analysis. Nevertheless, an IRT analysis based on all collected data items increase further the power to detect a DEDP without any need of an apriori assumption on its magnitude and is recommended.
References:
[1] C G Goetz et al, Movement Disorder Society-Sponsored Revision of the Unified Parkinson’s Disease Rating Scale (MDS-UPDRS): Scale, Movement Disorders, 2007 Jan, 22(1):41-7
[2] Ramaker, C., Marinus, J., Stiggelbout, A. M., & Hilten, B. J. (2002). Systematic evaluation of rating scales for impairment and disability in Parkinson's disease. Movement Disorders Mov Disord., 17(5), 867-876.
[3] Kalia L, Lang A. Parkinson's disease. The Lancet. 2015;386(9996):896-912.
[4] Louis E, Tang M, Cote L, Alfaro B, Mejia H, Marder K. Progression of Parkinsonian Signs in Parkinson Disease. Arch Neurol. 1999;56(3):334.
[5] Vu T, Nutt J, Holford N. Progression of motor and nonmotor features of Parkinson's disease and their response to treatment. British Journal of Clinical Pharmacology. 2012;74(2):267-283.
[6] Buatois S, Retout S, Gieschke R, Ueckert S, Frey N. Using Item response theory to yield information from the MDS-UPDRS items in Parkinson’s disease clinical trials. http://www.page-meeting.org/?abstract=3417
[7] PPMI: Parkinson’s Progression marker initiative www.ppmi-info.org/data
[8] Cormen T. Introduction to algorithms. Cambridge, Mass.: MIT Press; 2009.
