Nieves Velez de Mendizabal1,2, Iñaki F. Troconiz3, Matthew M. Hutmacher 4, Robert R. Bies1, 2
(1) Indiana University School of Medicine; Indianapolis, IN, USA. (2) Indiana Clinical and Translational Sciences Institute (CTSI), Indianapolis, IN, USA. (3) Department of Pharmacy and Pharmaceutical Technology; School of Pharmacy; University of Navarra; Pamplona, Spain. (4) Ann Arbor Pharmacometrics Group (A2PG), Ann Arbor, MI, USA
Objectives: Multiple sclerosis (MS) is a prototypic autoimmune disease which affects to the central nervous system (CNS) with a relapsing-remitting symptomatology (RRMS) [1]. The focal inflammatory events of the CNS are evident on MRI as contrast enhancing lesions (CELs). The natural history of a CEL is highly unpredictable. For the appropriate design of future longitudinal studies and clinical trials, it would be relevant to know the distribution of new CELs. Here we analyzed the best statistical model fitting the distribution of CELs developed by nine RRMS patients whom underwent monthly MRI for 48 months.
Methods: Subjects. Data for this study were obtained from the NINDS-NIH [2]. Nine patients, not receiving immunomodulatory treatment with RRMS who underwent monthly MRI assessments with gadolinium, were studied for 48 months. The number of CELs was recorded for each consecutive month. Data analysis. Fifteen models based on seven different probability distributions were explored: Poisson model [PS], Poisson model with Markov elements, PMAK2, nested PMAK2, nested nested PMAK2 , Poisson model with mixture distribution [PMIX], Zero-Inflated Poisson model [ZIP], Generalized Poisson model [GP, GP PMAK2, GP nested PMAK2], Negative Binomial model [NB, NB PMAK2 , NB nested PMAK2 , NB nested nested PMAK2] and Zero-Inflated Negative Binomial model [ZINB]. Analyses were performed using NONMEM VII v2 (LAPLACIAN). Model evaluation was based on the comparison of several dynamic descriptors calculated for both raw and simulated data.
Results: Based mainly on the -2xLog(Likelihood), and the goodness of a developed Visual Numerical Predictive Check, the selected model was the negative binomial with lambda affected each time t by the observations of the 2 previous time points, t-1 and t-2.
Conclusions: In this study we analyzed the best statistical model fitting the distribution of CELs. Significant improvements were observed in the probability distribution models when the information about what happened in the two previous months was incorporated, although the importance of these previous observations seems to be diluted along the disease course. In the future, mechanistic elements, such as the balance between effector and regulatory T cell, will be incorporated [3] in order to identify latent variables that explain variations in the parameter lambda.
References:
[1] Compston A, Coles A, Multiple sclerosis. Lancet 2008; 372, 1502.
[2] Bagnato F, Jeffries N, Richert ND, Stone RD, Ohayon JM, McFarland HF, Frank JA: Evolution of T1 black holes in patients with multiple sclerosis imaged monthly for 4 years. Brain 2003, 126(Pt 8):1782-1789.
[3] Velez de Mendizabal N, Carneiro J, Sole RV, Goni J, Bragard J, Martinez-Forero I, Martinez-Pasamar S, Sepulcre J, Torrealdea J, Bagnato F, Garcia-Ojalvo J, Villoslada P. Modeling the effector – regulatory T cell cross-regulation reveals the intrinsic character of relapses in Multiple Sclerosis BMC Sys Biol (2011) Jul 15;5(1):114
Reference: PAGE 21 (2012) Abstr 2414 [www.page-meeting.org/?abstract=2414]
Poster: CNS