Selection of a PNEC
Selecting a PNEC for the combined major estrogens that would be protective of fish reproductive effects is a different approach from selecting a level below which no effects whatsoever are assumed. Although it is accepted that subtle endocrine disruption effects, such as VTG induction, may occur at concentrations below those where reproductive effects could start, it is fish reproductive ability that is of greatest concern.
Caldwell et al. 20, 21 used species sensitivity distributions of available chronic fish reproductive toxicity data to develop PNECs of 0.1 and 2.0 ng/L for EE2 and E2, respectively, protective of reproductive effects in fish following long-term exposures (i.e., exposures encompassing several life stages or multiple generations). These PNECs were used in the present study. Unfortunately, insufficient data were available to employ the same methodology to derive PNECs for E1 and E3 21. Instead, Caldwell et al. 21 employed in vivo VTG induction studies to determine the relative ability to induce VTG by each of the steroid estrogens on the assumption that the estrogens would have the same relative effects on reproductive end points. Thus, Caldwell et al. 21 derived PNECs protective of long-term exposures of 6 and 60 ng/L for E1 and E3, respectively.
Caldwell et al. 21 recognized that E2 and EE2 are unique in that the potential aquatic effects of each has been investigated several times by studies that expose fish to constant levels of these steroid estrogens for multiple life stages and even multiple generations. Such constant long-term exposure seldom, if ever, occurs in natural waters. Concentrations are expected to vary, being lowest during high river flows and highest during low river flows. It is common for aquatic risk assessments to focus on the effects of the highest concentrations that occur during periods of low river flow by comparing low flow PECs to PNECs. By definition, such critical low flows only last for several days, not the many months used in the multi-generation studies to derive the long-term PNECs. Flows closer to the critical low flow than the mean annual flow may be present for several weeks or months during, for example, the late summer and fall in many U.S. rivers; however, those low flows typically do not persist for the entire life cycle of a fish species.
Caldwell et al. 21 found that the no-observed-effect concentrations (NOECs) for E1 and E2 reported by studies investigating exposures of less than 60 d were higher than the NOECs from long-duration, multi-generation studies. The short-term studies developed NOECs for reproductive effects such as changes in sex ratio, intersex, fecundity, and fertilization. In several instances, VTG induction, which is a potential measure of transitory exposure to estrogens, was observed at concentrations that did not elicit the more serious reproductive effects for which NOECs were reported and on which the short-term PNECs are based. On that basis, Caldwell et al. 21 derived PNECs protective of short-term exposures (<60 d) of 20, 5, 200, and 0.5 ng/L for E1, E2, E3, and EE2, respectively.
Two sets of PNECs are used in this risk characterization to evaluate the potential aquatic risks associated with exposure to steroid estrogens. Potential aquatic risks associated with long-duration exposures (represented by mean annual flow PECs) are characterized using PNECs of 6, 2, 60, and 0.1 ng/L for E1, E2, E3, and EE2, respectively. Potential aquatic risks associated with short-term exposures (<60 d) are assessed using PNECs of 20, 5, 200, and 0.5 ng/L for E1, E2, E3, and EE2, respectively. For the risk assessment, these estrogens were assumed to act in an additive way so that after accounting for their relative potencies, their concentrations in surface water could be summed to derive an E2-eq concentration 40, 41.
The PECs of estrogens in U.S. surface waters
Mean and critical low flow (i.e., 7Q10) PECs of estrogens in surface water resulting from humans were estimated using the PhATE model, Ver. 2.1.1 17. This GIS-based model is able to generate PECs for 12 watersheds in the United States. The watersheds included in PhATE were selected to represent a typical range of watersheds in the United States 17. The watersheds (1,488,661 km2) make up about 19% of the land area and were selected to represent the range of geography and climates present in the contiguous 48 states. They include 1,132 POTWs on about 41,508 kilometers of rivers divided into 2,713 segments. Fifty-seven of the POTWs included in PhATE serve over 50,000 people. In total, 692 have secondary treatment, 391 have advanced treatment, 7 have either primary or advanced primary treatment, and 42 do not discharge to a surface water. About 40 million people live within the 12 watersheds. The hydrological data underpinning the model are derived from U.S. Geological Survey river flow measurements, collected over varying durations, depending on watersheds through 2000.
The PhATE model requires several compound-specific inputs including the per capita use for prescribed estrogens or excretion rate for naturally occurring estrogens, metabolism for prescribed estrogens, POTW removal rate, and in-stream removal rate. Key inputs are summarized in Table 1 and are discussed in more detail in the Supplemental Data available online. The total amount of estrogens released into surface water through POTWs as a result of humans was estimated by summing the excreted mass of estrogens sold for therapeutic use in the United States and the excreted mass of naturally produced estrogens (Table 1 and Supplemental Data). Note that nonpoint sources such as releases from animal husbandry operations are not included in the present study. Excreted mass of estrogens from therapeutic use was estimated by adjusting annual sales volume by reported metabolism. Excreted mass of naturally produced estrogens was estimated by identifying age- and gender-specific excretion rates and weighting those rates according to the fraction of the total population that each age and gender category comprises in a method similar to Johnson and Williams 42 (see Supplemental Data). Assumed POTW removal rates are the median of removal rates reported in the literature (Table 1 and Supplemental Data). In-stream decay rates, for which relatively few studies are available, are based on on a review of the literature (Table 1 and Supplemental Data). Note that in the portion of a river reach immediately downstream of a POTW outfall, little degradation occurs because residence times for discharged estrogens are very short.
Table 1. Summary of key inputs used in the Pharmaceutical Assessment and Transport Evaluation Model to estimate predicted environmental concentrationsa
|Compound||Metabolism (%)b||Total mass excreted||Publicly owned treatment works removal||In-stream decay (1/d)|
|Per capita (µg/d)||Total U.S. (kg/y)c||Secondary (%)||Advanced secondary (%)|
Estimating E2-eq concentrations
Within each river segment modeled by PhATE, PECs of each individual estrogen were converted to an E2-eq concentration and then the E2-eq concentrations of each individual estrogen were summed to estimate a total E2-eq concentration for each river segment. The 17β-estradiol equivalents concentrations of each individual estrogen were derived by multiplying the concentration of each estrogen by its relative estrogenic potency to E2. Relative potency of the different estrogens was based on the relative difference of PNECs for each of the estrogens 21. For example, for long-term exposures, EE2 was assumed to have a relative potency of 20 (equal to the E2 PNEC [2 ng/L] divided by the EE2 PNEC [0.1 ng/L]). Similarly, E1 and E3 were assumed to have relative potencies of 0.33 and 0.033, respectively 21, for long-term exposures. For short-term exposures, the relative potencies were assumed to be 0.25, 0.025, and 10 for E1, E3, and EE2, respectively. Thus, a river reach that has mean flow concentrations of 0.1, 0.02, 0.004, and 0.004 ng/L of E1, E2, E3, and EE2, respectively (equal to the ∼90th percentile mean flow PECs; Supplemental Data) was assumed to have a long-term E2-eq concentration of 0.13 ng/L ([0.1 ng/L E1 × 0.33] + 0.02 ng/L E2 + [0.004 ng/L E3 × 0.033] + [0.004 ng/L EE2 × 20]). The short-term E2-eq concentration for those same steroid estrogen concentrations (equal to between the 60th to 90th percentile of critical low flow PECs depending on steroid estrogen; Supplemental Data) would be 0.09 ng/L ([0.1 ng/L E1 × 0.25] + 0.02 ng/L E2 + [0.004 ng/L E3 × 0.025] + [0.004 ng/L EE2 × 10]). The E2-eq concentrations were estimated for all 2,713 reaches in the 12 watersheds modeled by PhATE. Annual average flow PECs were used to develop long-term E2-eq concentrations and critical low flow concentrations were used to estimate short-term E2-eq concentrations. By converting all four steroid estrogens into E2-eq and combining their E2-eq concentrations, this evaluation inherently assumes additivity of the effects of the four estrogens, as has been demonstrated experimentally 40, 41.
Given the recent report of endocrine disruption effects in bass 36, a further analysis was carried out that compared the E2-eq PECs from PhATE and the intersex occurrence and VTG induction data reported by Hinck et al. 36 to investigate whether a relationship existed between PECs and the intersex and VTG observations. Twenty-one of the river reaches for which Hinck et al. 36 reported intersex occurrence in bass overlapped one or more segments in PhATE (Supplemental Data). Sixteen segments for which VTG induction was reported by Hinck et al. 36 overlapped one or more PhATE segments (Supplemental Data). Overlap occurred in the following four watersheds: Mississippi Headwaters, Columbia River, Lower Colorado Basin, Apalachicola River (Supplemental Data). Three reaches for which intersex information were available, and four for which VTG data were available, were represented by more than one PhATE segment. In those cases, the arithmetic average of the E2-eq PECs generated by PhATE for each of the segments was used in the comparison (Supplemental Data).
To examine the effect of uncertainty in the PEC to PNEC comparison, the margin of safety (MOS) was derived and plotted for each PhATE segment in each watershed, as well for all watersheds combined. The MOS is equal to the reciprocal of the PEC:PNEC ratio. Thus, a segment with an annual mean flow E2-eq PEC of 0.4 ng/L has a PEC:PNEC ratio of 0.2, assuming the long-term E2-eq PNEC is 2 ng/L, and a MOS of 5 (PNEC/PEC = 5). The MOS of 5 indicates that the PEC could be up to five times higher than estimated (i.e., it could have been underestimated by as much as fivefold), and still not exceed the PNEC. Alternatively, the PNEC could also be five times lower and still not be exceeded by the PEC. Thus, larger margins of safety have associated with them greater confidence that potential aquatic risks are not present.