This study, done during the 2002--2003 season among children <6 years of age to investigate influenza vaccine effectiveness and confounding factors, involved 2913 children (1512 vaccinees, 1401 non-vaccinees) recruited from 54 paediatric clinics. Between December 2002 and April 2003, parents reported their children's maximum body temperatures weekly. Influenza-like illness (ILI) was defined as an acute febrile illness (> or =38.0 degrees C) during the peak epidemic period. Adjusted odds ratios (ORs) for ILI were obtained using a logistic regression model. In analysis for total subjects, the ORs were significantly decreased for vaccinees (OR: 0.76, 95% CI: 0.66-0.88) and significantly increased for younger age groups, including children aged 2.0-3.9 years (1.42, 1.18-1.72) and those < 2.0 years (2.02,1.61-2.54), compared to those between 4.0 and 5.9 years. ORs were significantly increased for children who visited a physician within the last 6 months for a cold (1.27, 1.08-1.50), attended preschool (1.72, 1.45-2.04), and had > or =3 siblings (1.42, 1.15-1.74). These confounding factors are suggested to be considered in estimating vaccine effectiveness among young children. In subgroup analysis by age groups, significantly decreased ORs were seen in 2.0-3.9-year-old (0.59, 0.47-0.74) and 4.0-5.9-year-old (0.75, 0.58-0.98) vaccinees; no significant vaccine effectiveness was detected for those < 2.0 years (1.07, 0.80-1.44). Thus, among very young children vaccine effectiveness could not be demonstrated.