Dichlorodiphenyl dichloroethylene (DDE) has been shown to reduce the duration of breast feeding in two studies. In addition to duration, we examined whether DDE lowers the initiation of breast feeding. Between 1973 and 1991, the Michigan Department of Community Health conducted three surveys to assess polychlorinated biphenyls (PCBs) and DDE serum concentrations in Michigan anglers. Through telephone interviews with parents, we retrospectively ascertained information on breast feeding. Based on repeated maternal serum measurements between 1973 and 1991, we arrived at the level of exposure at the time of delivery by extrapolating PCB and DDE serum levels. One mother may have contributed more than one child; however, serum concentrations varied between children from the same mother. The maternal DDE and PCB serum concentrations were categorised as follows: 0 to <5 microg/L, 5 to <10 microg/L, >or=10 microg/L. Repeated measurement models and survival analyses were used to determine the relationship between DDE and PCBs and characteristics of breast feeding while controlling for cohort effects, maternal age at delivery, education, and smoking during pregnancy. We focused on 176 pregnancies of 91 mothers who had maternal exposure information and gave birth between 1969 and 1995. Initiation of breast feeding was lowered by 39.5% and duration shortened by 66.4% in children of mothers who smoked during pregnancy. In children of non-smoking mothers, the incidence ratio for breast-feeding initiation was 0.45 [95% CI 0.15, 0.94] and 0.42 [95% CI 0.10, 1.03] when maternal DDE concentrations were 5 to <10 microg/L and >or=10 microg/L respectively, compared with the lowest DDE exposure group. In these offspring (of non-smoking mothers), breast-feeding duration was shorter when DDE concentrations were higher: 13 weeks for >or=10 microg/L DDE, compared with 21.7 weeks for lower DDE. We did not detect any association between PCBs and breast feeding. In the absence of the distorting effects of maternal smoking, DDE exposure may decrease initiation and duration of breast feeding.

Great Lakes sport-caught fish are contaminated with various organochlorines (OCs) such as polychlorinated biphenyls (PCBs). Through consumption of these fish, humans are subject to continuing levels of OC contamination. To assess potential adverse effects of past exposure, we compared three different backward extrapolation models. The data originated from OC determinations in a cohort of anglers and their families. Repeated PCB measurements collected in the 1970s, 1980s, and 1990s were used when testing the backward extrapolations. We applied a simple and a complex decay model based on assumptions used in previous studies; a third was a regression model incorporating markers of OC intake and loss. These techniques provided past exposure estimates. Intraclass correlation coefficients (ICCs) were calculated comparing measured and estimated PCB values. ICC values for the regression model equations were 0.77 and 0.89; ICC values for the simple and complex decay models were significantly lower, with ranges of 0.07-0.45 and -0.14-0.69, respectively. Plots showing trends of OC concentrations in fish and humans indicate comparable increases and decreases of PCB in fish and humans, with fish concentrations peaking approximately 10 years before that in humans. Our findings suggest that one should be cautious when using simple backward extrapolation techniques to estimate OC exposure in situations involving changing environmental exposures. Whenever repeated measurements are available, regression analyses seem to produce more accurate backward estimations of exposure.