Time series for modelling counts from a relapsing-remitting disease: application to modelling disease activity in multiple sclerosis.

Many chronic diseases are relapsing-remitting diseases, in which subjects alternate between periods with increasing and decreasing disease activity; relapsing-remitting multiple sclerosis is an example. This paper proposes two classes of models for sequences of counts observed from a relapsing-remitting disease. In the first, the relapsing-remitting nature of the data is modelled by a Poisson time series with a periodic trend in the mean. In this approach, the mean is expressed as a function of a sinusoidal trend and past observations of the time series. An algorithm that uses GLIM is developed, and it results in maximum-likelihood estimation for the amplitude, frequency and autoregressive effects. In the second class of models, the relapsing-remitting behaviour is described by a Poisson time series in which changes in the mean follow a latent Markov chain. An EM algorithm is developed for maximum-likelihood estimation for this model. The two models are illustrated and compared with data from a study evaluating the use of serial magnetic resonance imaging as a measure of disease activity in relapsing-remitting multiple sclerosis.

[1]  H. McFarland,et al.  Clinical worsening in multiple sclerosis is associated with increased frequency and area of gadopentetate dimeglumine–enhancing magnetic resonance imaging lesions , 1993, Annals of neurology.

[2]  N. Patronas,et al.  Serial gadolinium‐enhanced magnetic resonance imaging scans in patients with early, relapsing‐remitting multiple sclerosis: Implications for clinical trials and natural history , 1991, Annals of neurology.

[3]  R. Hamman,et al.  Seasonality comparisons among groups using incidence data. , 1988, Biometrics.

[4]  Models for rates with poisson errors , 1984 .

[5]  P. Albert,et al.  Models for longitudinal data: a generalized estimating equation approach. , 1988, Biometrics.

[6]  S. Zeger,et al.  Markov regression models for time series: a quasi-likelihood approach. , 1988, Biometrics.

[7]  Roland Martin,et al.  Using gadolinium‐enhanced magnetic resonance imaging lesions to monitor disease activity in multiple sclerosis , 1992, Annals of neurology.

[8]  N. Kashiwagi,et al.  Smoothing serial count data through a state-space model , 1992 .

[9]  G. Box,et al.  Transformation of the Independent Variables , 1962 .

[10]  Kung-Yee Liang,et al.  Prediction of Random Effects in the Generalized Linear Model , 1993 .

[11]  A. Singh,et al.  State Space Modelling of Cross-Classified Time Series of Counts , 1992 .

[12]  M. West,et al.  Dynamic Generalized Linear Models and Bayesian Forecasting , 1985 .

[13]  S. Zeger.  A regression model for time series of counts , 1988 .

[14]  B E Kendall,et al.  Breakdown of the blood-brain barrier precedes symptoms and other MRI signs of new lesions in multiple sclerosis. Pathogenetic and clinical implications. , 1990, Brain : a journal of neurology.

[15]  D. Rubin,et al.  Maximum likelihood from incomplete data via the EM - algorithm plus discussions on the paper , 1977 .

[16]  L. Baum,et al.  A Maximization Technique Occurring in the Statistical Analysis of Probabilistic Functions of Markov Chains , 1970 .