Progresive diseases study using Markov′s multiple stage models

Risk factors and their degree of association with a progressive disease,such as Alzheimerís disease or liver cancer, can be identifi edby using epidemiological models; some examples of these modelsinclude logistic and Poisson regression, log-linear, linear regression,and mixed models. Using models that take into account not onlythe different health status that a person could experience betweenvisits but also his/her characteristics (i.e. age, gender, genetic traits,etc.) seems to be reasonable and justifi ed. In this paper we discussa methodology to estimate the effect of covariates that could beassociated with a disease when its progression or regression canbe idealized by means of a multi-state model that incorporates thelongitudinal nature of data. This method is based on the Markovproperty and it is illustrated using simulated data about Alzheimerísdisease. Finally, the merits and limitations of this method are discussed.