STATISTICS IN MEDICINE Statist. Med. 2001; 20:951–963 Repeated-measures models to evaluate a hepatitis B vaccination programme Didier Renard 1;*;† , Liesbeth Bruckers 1 , Geert Molenberghs 1 , Akke Vellinga 2 and Pierre Van Damme 2 1 Limburgs Universitair Centrum; Center for Statistics; Universitaire Campus; B-3590 Diepenbeek; Belgium 2 Center for the Evaluation of Vaccination; Epidemiology and Community Medicine; University of Antwerp; Universiteitsplein 1; B-2610 Antwerp; Belgium SUMMARY In 1985–1986, a hepatitis B vaccination programme was conducted in a Belgian institution for the mentally handicapped. A group of 97 residents was followed up for 11 years in order to characterize the long-term persistence of hepatitis B antibodies after vaccination. This paper proposes the use of linear mixed-eects models to account for serial correlation and between-individual heterogeneity in the data, while adjusting the analysis for various individual characteristics and important risk factors in the response to vaccination. We propose several model building strategies and focus on the prediction of future levels of antibodies. Copyright ? 2001 John Wiley & Sons, Ltd. 1. INTRODUCTION Mentally handicapped individuals residing in institutions are at high risk for hepatitis B virus (HBV) acquisition and subsequent carrier state. The higher risk of non-parenteral transmission in this population is due to the typical behaviour of mentally retarded patients, the type of mental retardation and the closed setting of the institutions which all enhance spreading of the virus. Hepatitis B vaccination of residents and sta is a general recommendation and has become part of today’s hepatitis B prevention programme. Data on long-term persistence of antibodies against HBV are scarce, especially in this population. Data available from other high-risk populations showed that 67 to 85 per cent of the vaccinated individuals still had antibody levels higher than 10 International Units= litre (IU= l), 9 to 12 years after the rst vaccine dose [1–5]. * Correspondence to: Didier Renard, Limburgs Universitair Centrum, Center for Statistics, Universitaire Campus, B-3590 Diepenbeek, Belgium † E-mail: didier.renard@luc.ac.be Contract=grant sponsor: LUC Bijzonder Onderzoekfonds Received September 1998 Copyright ? 2001 John Wiley & Sons, Ltd. Accepted April 2000