Fax +41 61 306 12 34 E-Mail karger@karger.ch www.karger.com Methods in Neuroepidemiology Neuroepidemiology 2007;28:56–64 DOI: 10.1159/000098518 Disability Evolution in Multiple Sclerosis: How to Deal with Missing Transition Times in the Markov Model? V. Petiot a C. Quantin a G. Le Teuff a M. Chavance c C. Binquet a M. Abrahamowicz d, e T. Moreau b a Service de Biostatistique et d’Informatique Médicale and b Service de Neurology, CHRU Dijon, Dijon, and c Inserm, U472, Paris, France; d Department of Epidemiology and Biostatistics, McGill University and e Division of Clinical Epidemiology, McGill University Health Centre, Montreal, Canada Introduction Multiple sclerosis (MS) is the most frequent chronic disabling disease of the central nervous system in adults in western countries. This disease mainly affects young people and particularly women aged from 20 to 35 years. MS induces a full spectrum of disease ranging from be- nign, and even asymptomatic, to more severe cases in- cluding wheelchair dependence [1]. Consequently, the MS course is usually assessed through the progression of disability measured by means of disability scales such as the Expanded Disability Status Scale (EDSS) [2] or the European Database for Multiple Sclerosis (EDMUS) Grading Scale (EGS), which are concordant [3]. MS is characterized by two types of clinical events: relapse and progression. Relapses are defined as the ap- pearance, the reappearance, or the worsening of symp- toms of neurological dysfunction lasting more than 24 h. Between these relapses, the neurological status of the pa- tient can return to normal or present sequelae [1]. Pro- gression is characterized by a continuous neurological deterioration for a minimum of 6 months [4] . For 85% of patients, at its onset MS is diagnosed as relapsing-remit- Key Words Multiple sclerosis  Markov model  Missing data  Multiple imputation Abstract Markov modeling of disability progression in multiple scle- rosis requires knowledge of all times of transitions from a given level of disability to the next level, but such data are often missing. We address methodological challenges due to partly missing transition times. To estimate the effects of prognostic factors on the risk of transitions between three consecutive disability levels, two methods were used to deal with missing data. Listwise deletion limited the analysis to subjects with complete data. Multiple imputation of missing data revealed that data were missing at random (MAR mech- anism) and imputed the missing transition times from the Weibull model. The results were then compared with the full data set with the actual times established through chart re- view. Multiple imputation estimates were systematically closer to those from the full data set than the listwise dele- tion estimates. Copyright © 2007 S. Karger AG, Basel Published online: January 11, 2007 Catherine Quantin, MD, PhD Service de Biostatistique et Informatique Médicale, CHRU Dijon 1, boulevard Jeanne d’Arc, BP 77908 FR–21079 Dijon Cedex (France) Tel. +33 3 80 29 36 29, Fax +33 3 80 29 39 73, E-Mail catherine.quantin@chu-dijon.fr © 2007 S. Karger AG, Basel 0251–5350/07/0281–0056$23.50/0 Accessible online at: www.karger.com/ned