We measured fluoride concentrations in urine, serum, and amniotic fluid from 47 second trimester pregnant women primarily living in Northern California between 2014 and 2016. To our knowledge, this is the first time these types of data have been reported in the US. In our study, which used archived biological samples, community water fluoride samples from the time of collection were not available. Therefore, we used public records to determine the water fluoride concentrations for each community in the year that the samples were collected.
Fluoride measurements using a fluoride ion specific electrode are highly specific when the sample is buffered to a pH below 7 to prevent the interference of hydroxyl ions. However, fluoride concentrations in serum or plasma are close to the limit of detection (LOD) of the electrode of 0.02 mg/L (mg/L) and therefore the hexamethyldisiloxane (HMDS) facilitated diffusion method, originally derived by Taves , and further modified by Martinez-Mier et al., (Martinez-Mier et al., 2011) quantitatively transfers fluoride from the sample into an alkaline trapping solution of smaller volume. This process results in fluoride concentrations in the solution that are above the LOD and on the linear portion of the standard curve. Furthermore, this method is preferred for samples that contain protein as it also releases additional fluoride ions that may be bound to proteins through binding to cations or other positively charged molecular groups.
Urine from healthy individuals, which has a relatively low protein content, can be measured directly, without diffusion. However, because we did not know the medical history of our sample population, we used the diffusion method to measure the fluoride concentration in urine, as well in serum and amniotic fluid. Urine fluoride concentrations were measured in spot samples rather than 24-hour urine samples. Urine spot samples have been shown to be an accurate assessment of fluoride ingestion on a population basis . We found similar MUFSG concentrations for pregnant women in Northern California relative to their community water fluoride concentrations, as reported by Till et al. in a Canadian population . Community water fluoride concentration and MUFSG were associated in both this study, and the study by Till et al., although in this study the confidence intervals for both the adjusted and unadjusted associations crossed the null.
The formula used to correct for specific gravity was originally generated in 1945 by Levine and Jahy , to adjust for urinary lead concentrations. However, this formula has recently been questioned; in particular, there are concerns about whether it overcompensates for the confounding effect of specific gravity in the absence of an appropriately weighted exponential adjustment factor for the substance of interest [13, 21, 22]. There is no such factor yet defined for fluoride; therefore, we also present associations with maternal urine unadjusted for specific gravity (Table 2). We found that unadjusted maternal urine fluoride was significantly positively associated with water fluoride concentrations.
The similarity in urine fluoride measures in Till’s study of a larger cohort of Canadian women  and ours, supports the validity of our relatively small sample size, and underlines the usefulness of MUF as a biomarker to compare study outcomes relative to fluoride intake. However, we did not have access to information on additional possible fluoride exposure through dental products, or the use of tea, bottled drinks or water; which is a limitation to this study. With the inclusion of this data, we may have identified more differences between our US derived samples and those from the Canadian study.
Similarly to the findings by Smith et al.  and Zipkin et al.  in the US in the 1950s, we found that the mean concentration of fluoride in urine of women from fluoridated communities (MUF) was similar (0.74 mg/L) to mean concentrations of fluoride in drinking water (0.8 mg/L) (see Table 2). However, in our study, mean urine fluoride concentrations from communities with lower water fluoride concentrations (0.52 mg/L) were more than twice that of community water fluoride concentrations (0.2 mg/L). It is possible that the relative increase in urine fluoride concentrations of women from non-fluoridated communities, was due to the overnight stay in fluoridated San Francisco. However, these values for non-fluoridated communities were similar to those reported by Till et al. , suggesting that in both Canada and the US, there is increased exposure to other sources of fluoride outside of the community water supply. For example, consumption of bottled drinks made in areas with higher fluoride concentrations, which are then consumed in the lower fluoride areas, may increase fluoride exposure. This so-called “halo” effect of fluoride exposure [24, 25] would not have been present in previous times when most community water, including that used for the manufacturing of food and beverages, was at concentrations less than 1.0 mg/L fluoride.
Smith et al. previously reported that the fold increase of urine fluoride as compared to blood fluoride concentration (4.3 fold), was lower in communities with low fluoride in drinking water as compared to communities with higher water fluoride (28 fold) . This reported difference between lower and higher fluoride exposure suggests an increase in glomerular filtration rate with increasing fluoride exposure. However, Malin et al.  showed evidence of reduced glomerular filtration rates associated with increased water fluoride, which would result in a decrease, rather than an increase in the fold difference between urine and blood fluoride . Instead, the relatively high blood fluoride concentrations as compared to urine fluoride in the low fluoride group reported by Smith et al., may have been due to sampling techniques as suggested by Taves , or more likely, because they had reached the limits of detection in their method for measuring low concentrations of fluoride in blood. Our results show a consistent and significant association between water fluoride, urine fluoride and serum fluoride, and support the use of urine fluoride as a biomarker for systemic fluoride exposure.
We found fluoride concentrations in amniotic fluid to be similar to maternal serum fluoride concentrations, and both were positively correlated to community water fluoride concentrations. These concentrations of amniotic fluid fluoride are similar to those reported by Ron et al.  drawn during mid trimester amniocentesis. However, in that study, maternal plasma concentrations were higher than what we measured for maternal serum. Fluoride concentrations in plasma and serum are comparable, and therefore a possible reason for the differences between ours and Ron’s study, may be related to their use of direct fluoride measurements, without prior diffusion, which would therefore be at the limits of fluoride detection by the electrode.
Our finding of similar fluoride concentrations between maternal serum and amniotic fluid supports direct diffusion of fluoride from maternal serum, without a placental barrier. This is supported by data from Amstrong  who found that ashed sera from maternal and umbilical blood obtained after cesarean section, contained similar concentrations of fluoride. Shen and Taves subsequently measured fluoride in maternal and cord blood at birth in 5 subjects using the fluoride diffusion method and found the concentration of fluoride in cord blood to be approximately 75% of maternal serum. They concluded that the high positive correlation between maternal and cord blood (0.86) showed that fluoride passively diffuses across the placenta . It is likely that the difference between these studies may have been related to the time of sampling, as circulating fetal fluoride concentrations are reduced later in gestation as fluoride is taken up into the rapidly growing skeleton. Though Gedalia is frequently quoted as providing evidence that the placenta creates a barrier to fluoride  he reversed this in a later publication  confirming the free passage of fluoride between mother and fetus. The difference in his findings over time has been attributed to methodological differences in his fluoride analysis.
Gupta et al. reported that when maternal plasma fluoride concentrations were greater than 0.4 mg/L, that fetal cord blood fluoride concentrations were relatively reduced . This finding has been interpreted as evidence that a placental barrier occurs at high fluoride concentrations. However, 0.4 mg/L is an extremely high plasma fluoride concentration, and is over 10 fold the highest maternal serum concentration measured in our study. Reports of these high plasma fluoride concentrations suggest the possibility that there were measurement errors. If they are correct, then is likely that other systemic toxic effects that would occur at these concentrations , would account for the differences in fluoride concentrations, rather than a specific placental barrier to fluoride ion diffusion.
The early formed amniotic fluid is formed by diffusion of maternal plasma through the placenta, and is replaced by fetal urine at about 20 weeks. This early formed amniotic fluid is contained in the neural tube when it closes at approximately the 4th week of pregnancy, , and forms the nascent cerebrospinal fluid (CSF) , where it interacts with the developing brain. The importance of this early cerebral spinal fluid in brain development is demonstrated in Xenopus embryos, where exposure to a mixture of ubiquitous chemicals at concentrations found in human amniotic fluid affect thyroid hormone-dependent transcription, gene expression, brain development and behaviour in early embryogenesis . While our analyses were done on midtrimester amniotic fluid, given recent population based correlations between systemic fluoride exposure and neurotoxicity , further studies of how fluoride may also affect early brain development in concert with other environmental stressors, are warranted.