Urinary BPA concentrations were measured in 217 samples collected from 82 subjects. On average, each subject contributed 2.6 urine samples, ranging from 1 (n = 34 subjects) to 13 samples (n = 1 subject). Twenty-nine (13%) urine samples had BPA concentrations < LOD. Forty-five women contributed 145 urine samples and 37 men contributed 72 specimens. Among subjects, 24 women (93 specimens) and 24 men (53 specimens) were partners. Of the urine samples from these 24 couples, 42 pairs of specimens were time-matched (collected on the same day). Eleven couples had two or more time-matched urine samples, and 10 couples had a single time-matched urine sample. Three couples had no time-matched urine samples.

The study subjects ranged in age from 28 to 54 years, with a mean ± SD of 35.5 ± 4.4 years (Table 1). Seventy subjects (85%) were Caucasian. Only five subjects (three men and two women) were current smokers. BMI (in kilograms per square meter) ranged from 16.5 to 44.7, with a mean ± SD of 26.8 ± 5.8. Ten of the women became pregnant during the follow-up period, as confirmed by the presence of a fetal heart beat on ultrasound. One pregnancy loss occurred.

Among the 217 urine samples, the median BPA concentration was 1.20 μg/L (25th percentile, 0.60; 75th percentile, 2.70), with a range from < LOD of 0.4 μg/L to 42.6 μg/L. However, because the number of urine samples collected from subjects varied widely, subjects contributed differentially to the median and percentiles of the urinary BPA concentrations. For instance, a subject with four urine samples contributed four times as many data points as a subject with one sample. In addition, multiple urine samples from a given subjects are not independent. To account for both of these factors, we calculated the percentiles of the urinary BPA distribution based on the geometric mean of the urinary BPA concentrations for each subject with two or more urine samples and the single urine sample concentration for subjects with one sample (Table 2). After accounting for these factors, the median was 1.30 μg/L (25th percentile, 0.72; 75th percentile, 2.09). The median was slightly lower among women (1.08 μg/L) than among men (1.40 μg/L). Urinary concentrations unadjusted for SG (Table 2) allow for comparisons with other study populations, because few studies have used SG to adjust for urinary dilution.

Because spot urine samples were collected at different times throughout the day, we explored whether time of day was associated with urinary SG-adjusted BPA concentrations (Table 3). Urine samples collected before 0900 hours (time of collection ranged from 0530 to 0859 hours) had a median of 2.4 μg/L compared with medians of 1.92 μg/L for samples collected from 0900 to 1159 hours, whereas samples collected between noon and 1559 hours had a higher median of 3.24 μg/L. Median BPA then decreased to 1.62 μg/L among samples collected after 1600 hours (range 1600 to 1830 hours). In mixed regression analysis, there was a statistically significant difference in geometric mean SG-adjusted BPA concentration by time of day when urine samples were collected (p = 0.01 for comparison of samples collected between noon and 1559 hours with those collected after 1600 hours).

Although age, BMI, sex, and pregnancy status were not statistically significant predictors of urinary BPA concentrations, we found weak associations after reverse transformation of the regression coefficients. Controlling for SG, BPA concentrations increased by 2% [95% confidence interval (CI), –2 to 6%) for each year age increased. To explore associations of BMI with urinary BPA concentrations, BMI was categorized as underweight (< 20), normal (20–24.9), overweight (25–29.9), or obese (≥ 30). Compared with normal BMI, underweight BMI was associated with a 49% increase (95% CI, –25 to 197%) in BPA concentration, overweight BMI with an 8% increase (95% CI, –27 to 61%), and obese BMI with a 3% decrease (95% CI, –38 to 51%). BPA concentrations for women were 9% lower (95% CI, –35 to 28%) than those for men, and concentrations for urine samples collected from women while pregnant were 26% higher (95% CI, –26 to 115%) than urine samples collected from nonpregnant women. Restricting to the 10 women who became pregnant during follow-up, samples collected while women were pregnant had concentrations 33% higher (95% CI, –33 to 165%) than those collected prepregnancy (n = 49 samples).

To investigate within-couple correlation of urinary BPA concentrations, we first noted that the covariance parameters in the mixed-effects models were estimated to be approximately 0.04 for subject, 0.04 for within-couple collection date, and 0.11 for residual variance. Because the variance component for couples had been estimated to be zero, these numbers suggest that BPA concentrations between partners will correlate substantially only when the measurements are taken on the same day. As a second approach, we used the geometric mean of SG-adjusted BPA concentrations as an average exposure measurement for each individual; using ordinary linear regression, we compared the female’s geometric mean with that of her male partner (n = 24 couples). We observed no significant association between the means (Spearman correlation = 0.04, p = 0.84); similar results were found using arithmetic means or arithmetic means on the log₁₀ scale (p = 0.92 and 0.84).

However, when female BPA concentration (SG-adjusted and log₁₀-transformed) on a particular day was compared with that of her male partner on the same day (n = 42 pairs of day-matched, couple-matched specimens), a moderately strong association was found (Figure 1). Using a linear mixed model accounting for correlation from repeated observations on the same couple by inclusion of random effects, each one unit increase in male log₁₀ SG-adjusted BPA urine concentration was estimated to correspond to a 0.428 increase (95% CI, 0.110 to 0.747) in the female partner’s urine BPA concentration from the same day (Spearman correlation = 0.36; p = 0.02).

Among subjects with at least three urine samples, the calendar time between collection of the first and last urine samples ranged from 14 to 482 days, with a mean (median) of 172 (150) days. The time between collection of first and last urine samples ranged from 150 to 482 days for subjects with at least six samples, with a mean (median) of 304 (329) days.

Sensitivity and specificity analyses to assess the ability of one urine sample to predict a subject’s exposure tertile are shown in Table 4. The proportion of subjects who “truly” were in the highest exposure tertile (top 33%), identified as those with a single urine sample collected anytime during the subject’s participation in the study (i.e., sensitivity) was 0.64. The proportion of subjects who “truly” were in the lowest exposure tertiles (e.g., first and second tertiles) and classified as such (i.e., specificity) was 0.76. The proportion of single urine samples that classified an individual in the highest tertile who was “truly” classified in the highest group (i.e., positive predictive value) was 0.63. We also calculated these measures for the ability of two urine samples to predict an individual’s geometric mean (Table 4). When two BPA samples were used to classify each subject, we found increases in sensitivity (0.67), specificity (0.84), and positive predictive value (0.85).

Apart from a small study in Japan on five subjects (four men and one woman) (Arakawa et al. 2004), we report some of the first data on long-term temporal variability of urinary BPA concentrations among a large sample of both men and women. We found temporal within-subject variability in urinary SG-adjusted BPA concentrations. As suggested by earlier small studies, we also found that men had slightly higher BPA concentrations than women. Japanese researchers reported associations of male sex and androgen levels with higher BPA concentrations (Takeuchi and Tsutsumi 2002; Takeuchi et al. 2004). However, Kim et al. (2003) found that BPA urine concentrations in males and females were similar. Further exploration of a potential sex- or androgen-related difference in urinary BPA concentrations is needed to resolve the inconsistencies across limited studies.

To the best of our knowledge, the present study is the first to compare urinary BPA concentrations in women before and during pregnancy. Although only 10 women contributed data, there was a nonsignificant 33% increase in urinary concentrations of BPA during pregnancy compared with prepregnancy. During pregnancy, there are numerous physiologic changes that may affect BPA distribution, metabolism, and/or clearance. For instance, the glomerular filtration rate is increased during pregnancy, creatinine clearance increases, and in some pregnant women, urine output increases due to increased metabolism of antidiuretic hormone. To explore whether changes in urinary concentrations of BPA are partially due to changes in urine SG during pregnancy, we regressed SG on pregnancy status, clustering by subject. The estimated SG for pregnant women was 0.0029 higher than for nonpregnant women (95% CI, –0.0014 to 0.0071; p = 0.19). Therefore, changes in SG during pregnancy would contribute to a slightly lower SG-adjusted BPA concentration, rather than an increase, as we observed during pregnancy.

Other potential explanations for differences in urinary BPA concentrations during pregnancy include a) changes in exposure, such as an increased use of products or in the consumption of foods that contain BPA; b) changes in volume of distribution and/or altered sequestration of BPA into different body fluid compartments, such as amniotic fluid; and c) changes in BPA metabolism and/or excretion. Application of physiologically based pharmacokinetic approaches is necessary to better understand changes in BPA dynamics during pregnancy, but it is beyond the scope of the present report.

Because exposure to BPA is thought to be primarily via ingestion, we hypothesized that urinary BPA concentrations will be more similar among individuals that are partners than those who are not partners. We found evidence of a moderately strong relationship between urinary BPA concentrations in samples collected on the same day among male and female partners. This suggests that BPA exposure occurs through a common household lifestyle factor, most likely diet. This finding warrants further investigation because sources of BPA among the general population are unclear.

Interestingly, there appeared to be a trend in urinary BPA concentrations by the time of day when the sample was collected. Urinary BPA concentrations were highest in the samples collected between 1200 and 1600 hours, compared with morning or late afternoon/ evening samples. This may reflect dietary BPA exposure during the midday meal, because BPA has a short half-life (Volkel et al. 2002). We did not collect data on the time of day meals or foods were consumed throughout the day, so we are unable to confirm this directly.

We also conducted sensitivity and specificity analyses to determine the ability of one urine sample or a pair of urine samples to predict long-term urinary BPA concentrations over several months. Despite the presence of temporal variability in urinary BPA concentrations, our sensitivity and specificity analyses suggested that a single urine sample correctly classified subjects into the highest BPA exposure tertile based on the geometric mean of repeat samples for that subject. However, there were several limitations to our approach. First, because there was no gold standard to use as the “true” urinary BPA concentration, we relied on using the geometric mean of repeat urine sample concentrations. How well these geometric means represent an individual’s true exposure is unclear. Furthermore, there is structural dependency between the “true” urinary BPA concentrations and the single or pair of urine samples that will over-inflate the sensitivity and specificity. Second, the timing of the collection of the repeat urine samples was not balanced. For instance, the duration between the collection of consecutive repeat urine samples ranged from 3 days to a year or more, but all samples were treated equally without respect to the time between collection. Thus, although we found that two urine samples improved the ability to predict BPA exposure tertiles, the preferred temporal spacing of the two samples is currently unclear. However, based on previous work on other nonpersistent chemicals, such as phthalates, at least 1 month between samples is preferable when the exposure duration of interest is over several months (Hauser et al. 2004).

Despite a much longer span of time between the collection of the first and last urine samples among subjects in the two-sample sensitivity analysis, the predictive ability was increased with the collection of a second sample. For the subset included in the two-sample sensitivity analysis (subjects with at least six samples), the average time interval between samples was 304 days compared with 175 days among the group of subjects included in the one-sample sensitivity analysis (subjects with at least three samples). Because samples from the two-sample sensitivity analysis represent a much wider exposure period, we would not have been surprised by a lack of improvement in the two-sample analysis compared to the one-sample analysis. However, our finding of increased predictive ability in the two-sample analysis may suggest that, because BPA exposure is ubiquitous and likely occurs daily, urinary concentrations exist in a relatively pseudosteady state over the course of months or years (National Research Council 2006).

The urinary BPA concentrations found in the present study were within the range reported from studies in several countries, including Japan, Korea, and the United States. In a nonrepresentative subset of NHANES III that included men and women between 20 and 59 years of age, the geometric mean urinary BPA concentration was 1.33 μg/L (Calafat et al. 2005). Among girls 6–8 years of age, Wolff et al. (2007) reported a geometric mean of 2.0 μg/L. Yang et al. (2003) reported a geometric mean urinary BPA concentration of 9.54 μg/L among 73 Koreans (34 men and 39 women) In another Korean study on 15 men and 15 women, Kim et al. (2003) reported a mean (SE) urinary BPA concentration of 2.82 μg/L (0.73) and 2.76 μg/L (0.54), respectively, with no difference by sex. Among 48 women in Japan, Ouchi and Watanabe (2002) reported a median urinary BPA concentration of 1.2 μg/L, with a range of 0.2–19.1 μg/L. Comparisons across studies are difficult because they may be confounded by differences in timing and methods of urine collection, containers used, analytical methods to quantify BPA, different limits of detection, and methods for correction of urinary dilution. Additionally, country-specific regulations on the use of BPA and occupational exposure to BPA may affect urinary concentrations. Countries with bans on the use of BPA products for packaging foods may have populations with lower urinary and serum BPA concentrations. However, we are not aware of data showing population differences in exposure before and after bans.

Despite within-person variability in urinary BPA concentrations, a single sample is predictive of long-term exposure (over weeks to months) and provides good sensitivity to classify individuals into tertiles in epidemiologic studies. The addition of a second urine sample improved this classification ability. Although we did not find strong relationships of urinary BPA concentrations with age, sex, BMI, and pregnancy status, these associations were suggestive and are worthy of further follow-up. Discovery of whether urinary concentrations increase during pregnancy, reflecting either increased exposure or alterations in metabolism during pregnancy, is especially important. The correlation between urinary BPA concentrations among male and female partners suggests that BPA exposure is shared through diet or common residential source(s). Finally, replication of this study in other populations, including the general population, should be conducted to determine whether predictors vary across populations. In addition, further work needs to be done to determine the utility of urine samples to define an individual’s exposure level during pregnancy.