The effect of smoking on breastfeeding was studied in 252 mothers with insulin-dependent diabetes mellitus (IDDM) giving birth in the period 1985-1995 and of whom 31% were classified as White group B, 16% group C, 37% group D and 16% group F or R. As part of a centralized regimen a detailed smoking history was taken during pregnancy. Forty-six percent were smokers. In 1997, a questionnaire was sent out asking for information on the duration of breastfeeding. The response rate was 87%. The duration of breastfeeding was significantly longer in the non-smoking group (p = 0.004). Fifty-five percent of non-smokers versus 33% of smokers were still breastfeeding 4 mo after birth. A strong dose-response relationship was established (p = 0.009). Cox multiple regression analysis showed a significantly negative influence of smoking (p = 0.01) and of hospitalization of the infants during the first year of life (p = 0.02) and a significantly positive influence of maternal age (p = 0.02) and birthweight (p = 0.03) on the length of breastfeeding. The breastfeeding curves for non-smokers and smokers were mainly divergent in the first month after birth, suggesting that the effect of smoking on breastfeeding is mainly exerted during that period. [ABSTRACT FROM AUTHOR]