Environ. Sci. Technol. 2006, 40, 6498-6505
First Derivation of Predicted-No-Effect Values for Freshwater and Terrestrial Ecosystems Exposed to Radioactive Substances JACQUELINE GARNIER-LAPLACE,* CLAIRE DELLA-VEDOVA, RODOLPHE GILBIN, DAVID COPPLESTONE,† JOANNE HINGSTON,† AND PHILIPPE CIFFROY‡ Laboratoire de Radioe´cologie et d’Ecotoxicologie, Institut de Radioprotection et de Su ˆ rete´ Nucle´aire, IRSN/DEI/SECRE/LRE, Cadarache, Baˆt. 186, BP 3, 13115 Saint-Paul-Lez-Durance, Cedex, France
The FASSET Radiation Effects Database (FRED) constitutes a unique structured resource of the biological effects of ionizing radiation on non-human species mainly from temperate ecosystems, encompassing 26,000 primary data entries. Quality-assessed data were extracted from FRED and dose-effect relationships were constructed to provide estimates of ED50 and EDR10. These estimates are Doses (or Dose Rates) related to the percent change in the average level of the endpoint for a particular effect (50% or 10% for acute or chronic exposure regimes, respectively). Acute and chronic Species Sensitivity Distributions (SSDs) were built on the basis of these data sets, and the Assessment Factor Method (AFM) was applied when data were too scarce. The Hazardous Dose corresponding to 5% of species acutely affected at the 50% effect level varied from 1 to 5.5 Gy according to the ecosystem. For chronic γ external irradiation exposure, no-effect values varied from 10 µGy/h for freshwaters through application of the AFM to 67 µGy/h for terrestrial ecosystems, corresponding to the 5th percentile of the non-weighted SSD (vs 229 µGy/h when trophic weights are applied). These values are higher by ca. ×50 to ×100 than the upper bound of natural background, and lower than dose rates triggering effects at individual levels on contaminated sites.
Introduction At the present time, no internationally agreed criteria or standards directly apply to the protection of the environment under chronic low-level exposure to radioactive substances. Only guideline values, below which biological effects from * Corresponding author phone: +33-4-42-19-95-34; fax: +33-442-19-91-51; e-mail:
[email protected]. † Environment Agency, Ecosystems & Radioactive Substances, P.O. Box 12, Richard Fairclough House, Knutsford Road, Latchford Warrington Cheshire, WA4 1HG, U.K. ‡ Electricite ´ de France, Division Recherche et De´veloppement, De´partement Laboratoire National d’Hydraulique et Environnement, 6 quai Watier, 78401 Chatou, France. 6498
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 20, 2006
ionizing radiation on nonhuman biota are unlikely to have a significant impact on a population, have been published following critical reviews of former radiobiological literature (1-3). However, these values have never been proposed as ecological risk assessment benchmarks or protection criteria. In addition, there is still a lack of criteria or guidance to help in assessing the ecological impact of acute exposure to radioactive substances in the case of accidents. The challenge for today’s scientists, underlined by the European Commission’s position (4), is to fill this gap in environmental radioprotection by rapidly developing a capability for prospective or retrospective quantification of the ecological risk associated with either acute or chronic exposure to radioactive substances. A small number of national bodies have already developed assessment methodologies inter alia (the U.S. Department of Energy (5), the Canadian Environment Agency (6), and the U.K. Environment Agency (7)) but in these cases the assessments rely upon these existing guideline values and not radiation specific risk assessment benchmarks. To address this, the FASSET project (8) produced a framework for the assessment of environmental impact of ionizing contaminants. It generated an extensive database, the FASSET Radiation Effects Database (FRED), which collated the primary data from irradiation experiments, conducted on different wildlife species, mainly representative of temperate ecosystems, reported in scientific peer-reviewed literature in order to summarize the radiation effects to nonhuman species (9). The current ERICA project is continuing the development of an integrated approach to assess and manage environmental risk from radioactive substances (10, 11) and is looking to highlight methods to derive radiation specific risk assessment benchmarks that could be used within future assessments. Simultaneously, the international radioprotection bodies have highlighted the need for consistency between the approaches to be applied in regulating radioactive substances and those applied for chemicals (12-14). This is because, although there are a number of key differences between chemical and radioactive stressorss e.g., for radioactive substances, the effects analysis is dependent on the amount of radiation energy absorbed by the body of the living organism rather than the concentration to which it is exposedsthere is no compelling argument for radioactive substances to be considered in a way different from that used for conventional chemicals. Consequently, if a similar approach is to be adopted, an Ecological Risk Assessment (ERA) could be used as a prelude to risk management (15, 16) for both categories of stressors. Within this context, this paper examines whether the current knowledge on the biological effects of radioactive substances on nonhuman species is adequate and robust enough to allow the derivation of risk assessment benchmarks. Traditionally, two methods are proposed for chemicals depending upon the composition of the ecotoxicity data sets: (1) the use of fixed assessment factors when few chronic ecotoxicity data are available, or (2) the use of the Species Sensitivity Distribution (SSD) approach associated with an arbitrary cutoff value, which is usually set at a protection level of 95% of the species when the available data are more robust. In Europe for instance, the Technical Guidance Document (TGD) (17) suggests the use of these methods to derive the so-called Predicted No-Effect Concentration (PNEC) with the primary object of protection designated as the maintenance of the structure and function of an ecosystem. Until 10.1021/es0606531 CCC: $33.50
2006 American Chemical Society Published on Web 09/14/2006
now, these techniques have not been applied to radioactive substances. Using the approach outlined in the TGD, this paper therefore presents the first derivation of predicted noeffect exposure levels for radioactive substances for terrestrial and freshwater ecosystems. These Predicted No-Effect Dose (rate) values and their associated uncertainties could be useful as risk assessment benchmarks. Once having presented the methodology applied for their derivation, the strengths and weaknesses of these values are critically discussed by comparing them to natural background levels and to exposure levels leading to ecological effects observed in radioactively contaminated sites.
Materials and Methods The FASSET Radionuclide Effect Database (FRED). FRED contains ca. 26,000 data entries from more than a thousand literature references covering the period 1934-2004 (9). These data correspond to pairs of points (exposure level, biological effect) along with information on the conditions in which these data were obtained, e.g., the tested species and its life stage, the exposure regime defined by the exposure duration and the irradiation pathway, and the biological effect endpoint. The following points should be noted. (1) The exposure level is assessed by using a specific unit, that is the absorbed dose or dose rate expressed in Gy or in µGy/h (basic conversion as follows: 1 Gy ) 1 J per kg of tissue; 1 MeV per disintegration corresponds to 5.77 × 10-4 µGy per hour); (2) The absorbed dose alone cannot predict the biological effect as other parameters have to be considered. These parameters may include the exposure pathway for both internal and external irradiation. External irradiation is dependent on where the organism is in the environment, how long it spends in different parts of the environment, and the type and location of the radionuclides in the surrounding air, soil/sediment, or water (dose contribution will be mainly from γ-rays, but also β-radiation for organisms with dimensions < ∼1 cm). Internal irradiation comes from R-, β-, and γ-emitting radionuclides taken up into biological tissues. Here the radiation quality is also important as it is known that an exposure from R-radiationswith high Linear Energy Transfer (LET)sis more effective than, for example, γ-rays and most β-radiation (low LET), per unit of absorbed dose, in producing biological damage (9). As for chemicals, experimental studies of the effects of ionizing radiation on living organisms were broken down into those that employ either acute exposures, i.e., in periods of time that are short, usually minutes but less than an hour, in comparison with the time taken for an effect to become apparent, or chronic exposures, i.e., over all, or a large part, of the life stage of interest. Furthermore, the biological endpoints tested have been grouped into four categories of effects: mortality, reproductive capacity, morbidity, mutation (9). The available data in FRED have approximately a 2:1 bias in favor of acute data with the majority of papers describing the consequences of external γ irradiation exposure. Chronic effects data are very limited and are also largely dominated by external γ irradiation exposure conditions. Quality Assurance and Data Extraction from FRED. No standardized ecotoxicity test exists for radioactive substances. The consequence is that there are major differences within the experiments that have been conducted, e.g., in test species exposure conditions, or the observed effects, or the range of doses or dose rates used. In order to address some of these issues, all FRED references were subjected to a grading for three criteria, which was then aggregated into a total score, with 80 as the maximum value: dosimetry, experimental design, and statistical details (18). Only data from papers
with medium to high scores for these criteria (i.e., papers with total scores ranging from 35 to 62) were extracted and used in the analysis described in this paper. Dose-Effects Reconstruction. A mathematical treatment was systematically applied to each extracted data set to reconstruct the dose(rates)-effect relationships and hence to estimate the critical toxicity endpoints. These critical data are: the estimated ED50 (in Gy), defined as the dose giving 50% change in observed effect for acute exposure, and the estimated EDR10 (in µGy/h), defined as the dose rate giving 10% change in observed effect corresponding to the EC10 (the NOEC for chemicals) for chronic exposure (19, 20). Monotonous dose(rate)-response curves were reconstructed using the commonly used model based on the Hill equation. The common form of the dose (or x) response (or y) curve is as follows:
y(x) ) (y(∞) - y(0)) × f(x) + y(0)
(1)
where y(0) and y(∞)are the boundaries of the effect zone:, i.e., the known response at zero dose (the control group) and the effect expected for a dose tending toward infinity, respectively; and f(x) is a probability function of the dose varying from 0 to 1 with the dose. Two parameters: the Hill number nH and the Dose(Rate) giving 50% effect ED(R)50 are characteristics of the probability function in a Hill model as follows:
f(x) )
x
nH
xnH + ED(R)50nH
(2)
The curve fitting approach applied was based on the Levenberg-Marquardt algorithm and was visually checked, as replicate values were not available in FRED. This fit enables the ED(R)x to be calculated. The ED(R)x is defined as the dose (or dose rate) that corresponds to x% of the effect with respect to the control. More precisely, the ED(R)x is the concentration for which x% of the maximum possible variation in response is observed. The extreme effect values, i.e., those obtained for the control group exposed only to the dose (or dose rate) corresponding to the natural background, y(0), and the group subject to the maximum dose (or dose rate) in the experiment, y(∞), need to be determined in a systematic and robust way as their values greatly influence the resulting curve fit. A rule to initiate the fitting process was thus defined as follows: if the control effect value y(0) is 0 (continuous data), 0% or 100% (percentage data), this value is imposed on the model. Otherwise, the control value can be adjusted. The value for the maximum effect y(∞) used is always imposed on the model to avoid poor estimation (>100% or
