Chlorinated hydrocarbons may increase breast cancer risk. Most epidemiological studies addressing this possibility have used one biological sample to measure a subject's cumulative exposure to these compounds. Little is known about short-term temporal variation in organochlorines, particularly in individuals with low levels. Thus, the reliability of using one sample to assess blood levels of chlorinated hydrocarbons in an epidemiological study is unknown. To better understand the temporal changes in blood measures among women with nonoccupational exposures to these compounds, we collected two 5-ml blood samples, an average of 2 months apart, from each of 31 nonfasting healthy women, ages 45-81 years. Samples were assayed for 1,1-dichloro-2,2-bis(p-chlorophenyl)ethylene (DDE), polychlorinated biphenyls (PCBs), and trans-nonachlor in blinded, matched pairs. Results were adjusted for estimated total plasma lipids. The correlations between the two blood samples were high for DDE and PCBs (lipid-adjusted, r = 0.96 and r = 0.89, respectively). For trans-nonachlor, the correlation was relatively poor (lipid-adjusted r = 0.57); however, with the removal of one outlier, the correlation improved substantially (lipid-adjusted, r = 0.90). The mean difference between the two blood samples in unadjusted [-0.36 ng/ml, 95% confidence interval (CI), -0.97, 0.24 ng/ml, P = 0.23] and lipid-adjusted (-0.035 microgram/g lipid; 95% CI, -0.124, 0.055; P = 0.44) DDE levels was small. Similarly, there was little change in the mean difference for unadjusted (-0.14 ng/ml; 95% CI, -0.53, 0.25 ng/ml; P = 0.47) and lipid-adjusted (0.006 microgram/g lipid; 95% CI, -0.050, 0.062; P = 0.82) PCB levels. The mean differences in trans-nonachlor levels between the two blood draws were also small: unadjusted (-0.03 ng/ml; 95% CI, -0.07, 0.02 ng/ml; P = 0.20) and lipid-adjusted (-0.003 microgram/g lipid; 95% CI, -0.010, 0.004; P = 0.33). These data suggest that temporal changes in organochlorine levels within a 1 to 3-month period are minimal for noncancer patients and that a single measure for estimating exposure is highly reliable for DDE and PCB. For trans-nonachlor, however, where the correlation between blood draws was lower, three samples would be needed for estimating exposure; if an outlier is removed from our data, however, then we can conclude that only a single measure is sufficient. These data, therefore, offer no clear conclusion for the use of a single measurement for trans-nonachlor.