Study objective - The study aimed to assess through a sentinel practice network the validity of data on levels of organochlorine residues in human milk along with personal, lifestyle, and exposure variables of breastfeeding women; to compare the results of this new approach with those of the Lower Saxony breast milk surveillance programme; and to test hypotheses on potential determinants of contamination levels. Design - Eligible women were enrolled into this cross sectional study by a network of 51 paediatric practices when bringing their babies for a U3 infant screening examination (4th to 6th week after delivery). Lifestyle and exposure factors were obtained by questionnaire. All milk samples were analysed for hexachlorocyclohexane, hexachlorbenzole, DDT, dieldrin, poly-chlorinated biphenyls (PCB) and heptachlor; half the samples were also analysed for dioxin. Analytic statistics were computed using polychotomous logistic regression (PLR). Setting - The study was conducted in Lower Saxony, Germany, from summer 1992 to summer 1993. Participants-Altogether 156 primiparous, breast feeding German women, aged 25-35 years, who had been born and had grown up in West Germany, were studied. Main results - Compared with the regular programme, participants in this study had their milk analysed sooner after delivery and were more likely to have grown up in rural areas, less likely to have been exposed to hazardous substances, less likely to have a diet of health food, and slightly less likely to be a smoker at the time of the study. Breast milk contamination levels were comparable in both studies, and in all but two cases well below the tolerable concentrations established by the Deutsche Forschungsgemeinschaft (German Research Fellowship). After adjustment for potential confounders using polychotomous logistic regression, there were statistically significant positive associations between breast milk contamination and age (PCB, test for trend: p = 0.006), average dietary fat intake per week (dioxin, p = 0.01), and proximity of residence to hazardous sites (dioxin, p<0-05), and negative associations between residue levels and relative body weight at the time of the study (PCB; p<0-0001) and difference in body weight (weight minus weight before the pregnancy; PCB, p = 0-0002), respectively. Conclusions - Sentinel practice networks are a feasible and low-biased approach to population based breast milk studies. The contamination levels and associations found are biologically plausible and comparable with the results of other studies. To reduce organochlorine residue levels in human milk in the short term, breastfeeding women should be advised not to try to reduce their weight until after lactation. Public promotion of a lower dietary fat intake may reduce the lifetime accumulation of organochlorine compounds in the human body fat tissue in the long term, resulting in lower concentrations in breast milk as well.