EC Contract: FI6RCT2004508847

1

Risk Characterisation
Institute for Radiological Protection and Nuclear
Safety
For the ERICA project
2
Table of Content

• What is risk characterisation?
• What are the methods to characterise the risk?
• ERICA Tiered Approach and Risk Characterisation
• Definitions of Benchmarcks
• How to derive those  safe levels ?
• Principles and assumptions of SSD
• The two methods adapted for radioactive
  substances
• Methodology Applied To FRED Data
• Application to FRED(ERICA) data
• Illustration Chronic g external exposure
  conditions
• Comparison of the benchmarks values obtained with
  the two methods
• ERICA screening values (SSD method)
• Conclusions for Tiers 1 and 2
• Tier 3 Effect analysis towards more realistic
  estimates to reduce the uncertainty

3
What is risk characterisation?
Exposure analysis (pathways, transfers, dosimetry)
Effect analysis (dose-effect relationships, safe
levels)
PED(R) Predicted Exposure Dose (rate) (Gy or
Gy/time) Or PNEC Predicted Exposure Concentration
(Bq/L or Bq/kg)
PNED(R) Predicted No-Effect Dose (rate) (Gy or
Gy/time)
Risk

• Risk characterisation is the final step of any
  Ecological Risk Assessment.

• It represents the synthesis of information
  obtained during risk
• assessment for use in management decisions.
• It includes an estimation of the probability (or
  incidence) and
• magnitude (or severity) of the adverse effects
  likely to occur in
• an ecosystem or its sub-organisational levels,
  together with
• identification of uncertainties.

4
What are the methods to characterise the risk?
Predicted Exposure Dose (rate)
Predicted No-effect Dose (rate)

• Deterministic method

PNED(R)
PED(R)
The risk index is expressed as the ratio
PED(R)PNED(R)

• Semi-probabilistic method

pdf
Uncertainty is introduced only in the exposure
estimate. The risk index is expressed as a
probability that the exposure estimate exceeds
the PNED(R).
1
0
pdf

• Probabilistic method

1
A statistical distribution is also alsigned to
PNED(R) (e.g. a species sensitivity
distribution). This allows to calculate the
probability of the risk.
0
5
ERICA Tiered Approach and Risk Characterisation

• Tiers 1 and 2 use deterministic method through
  the Risk Quotient.

• PED(R)PNED(R)
• Risk characterisation requires risk assessment
  benchmarchs,
• the so called PNED(R) Tier 2, and the
  corresponding
• environmental media limiting activities Tier
  1.

• Tier 3 uses probabilistic method.

Problem Formulation
Fixed values in Tiers 1 and 2 (PNED(R ) for
acute(chronic) exposure conditions)

risk?


Tier 1 media-specific benchmarks are
backcalculated (the lowest environmental media
limiting activity concentration (water,
sediment, soil) Tier 2 Direct use of PNEDR
corresponds to biota limiting benchmarck
Tier 1 Screening

risk?



Tier 2 Generic Assess.

Methods to derive appropriate risk
characterisation
risk?



Tier 3 Full Assess.
Tier 3 no fixed value - Receptor/site-specific

risk?




6
Definitions of Benchmarcks

• Benchmarks are numerical values used to
• guide risk assessors at various decision
• points in a tiered approach.

• These values need to be provided by a
  transparent and scientific
• reasoning.
• They correspond to screening values when they
  are used in
• screening tiers (e.g. ERICA Tiers 1 and 2).

• These screening values are concentration, dose
  or
• dose rate that are assumed to be safe based
  on.
• exposure response information (e.g.
  ecotoxicity test
• endpoints found in FREDERICA). They
  represent  safe
• levels  for the ecosystem.

• For radioactive substances, concentration are
  expressed in Bq per
• unit of volume or mass, dose in Gy and dose
  rate in Gy per unit
• of time.

7
How to derive those  safe levels ?(1/2)

• Methods recommended by EC for chemicals,
• adapted for radioactive substances

• The EC recommends to develop an assessment of
  effect and to characterise
• risk on ecosystems with the minimum amount of
  empirical information while
• using extrapolation methodology (Technical
  Guidance Document TGD, EC
• 2003).

• All existing approaches are based on available
  ecotoxicity data, typically EC50
• for acute exposure conditions (short-term) and
  EC10 (preferred to NOEC) for
• chronic exposure conditions (long-term).

Exposure-response relationship from ecotoxicity
tests (stressor, species, endpoint)

Effect ()
100
Observed data
Regression model
50
LOEC Lowest observed effect concentration
NOEC No observed effect concentration
10
Concentration (Bq/L or kg) Dose (Gy) Dose Rate
(µGy/h)
EC50 ED50 EDR50
EC10 ED10 EDR10
8
How to derive those  safe levels ?(2/2)

• Methods recommended by EC for chemicals,
• adapted for radioactive substances

• Whatever the method, the ecotoxicity data
  selection is of major importance
• as the benchmark values greatly depend on their
  relevancy, their quality and
• their quantity.

• Two main methods, recommended by EC (TGD,
  revised in 2003) to derive
• PNEC for chemicals.

• Safety Factor Method (SF method)
• based on the application of safety factors to
  ecotoxicity data stringent method as the PNEC
  value is obtained by dividing the lowest critical
  data by an appropriate SF ranging from 10 to 1000.

(2) Species Sensitivity Distribution Method
(SSD) based on a statistical extrapolation model
to address variation between species in their
sensitivity to a stressor.
9
Principles and assumptions of SSD (1/2)

• Safe levels can be calculated with statistical
  extrapolation models. The
• extrapolation is made from tested endpoint(s)
  for a set of tested species to
• the same endpoint(s) in the full set of the
  potentially exposed species.

• Two main assumptions
• The species for wich results are known are
  representative, in terms of sensitivity, of
• the totality of the species in the ecosystem.
• The endpoints measured in laboratory tests are
  indicative of effects on populations in
• field.

PAF ( of Affected Species)
Calculation of a dose(rate) that is assumed to
protect a given of species In the Technical
Guidance Document (2003) the agreed
concentration is the hazardous concentration
affecting 5 of species with 50
confidence. For radioactive substances -gtHD5 or
HDR5 useful to derive PNED or PNEDR
respectively.
100
80
60
40
Dose (Gy) or Dose Rate (µGy/h)
20
5
0
1
10
100
1000
10000
PNED(R) x SF
HD(R)5
10
Principles and assumptions of SSD (2/2)

• This method requires the selection of an
  appropriate level of protection (95
• of species). This cut-off value selected as
  protective level for the entire
• ecosystem can be argued on the basis of
  ecological theories.

• The aim of the cut-off value selection is to
  indirectly protect ecosystem
• structure and processes by protecting the most
  sensitive species.

• The more functionally redundancy that there is
  in a system, the more
• overprotective such an assumption will be. This
  is adequate when changes
• in structure is more sensitive than changes in
  process.
• In the difficult case of keystone species
  (species playing much larger
• functional role than others), the only way to
  deal with a cut-off value as
• protection level is to identify those
  unprotected species and to examine
• whether they correspond to an engineer.

11
The two methods adapted for radioactive
substances (1/2)
SF from 10 to 1000 according to rules given in
the TGD (2003)

• Safety Factor Method

12
The two methods adapted for radioactive
substances (2/2)

• Species Sensitivity Distribution Method

HD(R)5 for acute(chronic) PNED(R) combined with
a SF ranging from 1 to 5 according to rules given
in the TGD (2003)
13
Methodology Applied To FRED Data

• STEP 1 Extracting appropriate data sets
• Data from FRED sorted per ecosystem, per
• exposure condition, per bibliographic reference
  
• and per test . Quality of data describing each
  test
• is assessed.

FRED FASSET Radiation Effect Database

• STEP 2 Building dose(rate)-effect relationships
• Dose-effect relationship was built for each
  accepted
• test. Estimated toxicity values are ED50 for
  acute
• exposure or EDR10 for chronic exposure.
• The quality of the fitted model was judged.

• STEP 3 Deriving PNED(R) values
• The two methods recommended by EC were applied
  on the basis
• of the estimated toxicity values accepted after
  Step 2.
• Comparison of the obtained PNED(R) values.

14
Application to FRED(ERICA) data

• Data allocation per ecosystem and per exposure
  pathway (external or internal
• irradiation) and duration (acute or chronic).

• Only data devoted to effects induced by external
  irradiation pathway are
• quantitatively adequate to be mathematically
  structured in terms of dose-effect
• relationships (Hill model).

15
Illustration Chronic g external exposure
conditions (1/2)

• Set of toxicity values EDR10
• geometric means per species and
• effect category - obtained for
• terrestrial, freshwater and marine
• ecosystems.
• Application of the SF method and the
• SSD method on the dataset.

Results
AF method 3 EDR10 for 3 trophic level and lowest
value (contributing to the geometric mean of
31.3) equal to 6.7 µGy/h -gtSF10 PNEDR 0.6
µGy/h
16
Illustration Chronic g external exposure
conditions (2/2)

• Comparison of species radiosensitivity among
  ecosystems gave no difference
• (bilateral Wilcoxon test, a0.05). This allowed
  the construction of a unique SSD
• for generic ecosystems (SWFWTER) chronically
  exposed to external ?
• irradiation.

Results
SSD method

• HDR581.8
• SF 5
• rounded down and keeping 1 digit
• PNEDR
• 10 µGy/h

17
Comparison of the benchmarks values obtained with
the two methods

• As expected, the SF method gave much more
  stringent screening
• values. ERICA preferred to apply the SSD method
  due the use of the
• whole data set (not the lowest one).

18
ERICA screening values (SSD method)

• The present safe levels were derived to be used
  in the first tiers (and
• generic ecosystem) of the ERICA tiered approach
  that can be applied
• across the range of activities that use
  radioactive substances.

• As they were derived on the basis of ecotoxicity
  data from external g
• irradiation, it is suggested when needed to
  apply a safety factor to take
• account for the higher biological efficiency of
  a and b emitters.

19
Conclusions for Tiers 1 and 2

• The present safe levels were derived to be used
  in the first tiers (and
• generic ecosystem) of the ERICA tiered approach
  that can be applied
• across the range of activities that use
  radioactive substances. Our
• values are consistent with background and with
  former guidelines at
• which no significant effects were expected (see
  Table 16 in ERICA D5).

• As they were derived on the basis of ecotoxicity
  data from external g
• irradiation, it is suggested when needed to
  apply a safety factor to take
• account for the higher biological efficiency of
  a and b emitters.

• Selection of the 95 cut-off level for the
  application of the SSD method is consistent
  with the threshold level used for chemicals
  assessments in the
• TGD.

• To take account for the possible existence of
  keystone species among
• the 5 that are unprotected, we suggest to
  identify the species and
• the effect endpoint present in the lowest
  quartile of the distribution and
• to consider the possible consequences of this.

20
Tier 3 Effect analysis towards more realistic
estimates to reduce the uncertainty

• For Tier 3, the ERICA consortium decided that it
  would not be
• appropriate to make specific recommendations on
  numeric values.
• Rather, guidance on the sorts of approaches
  that may be applied for
• refined effect analysis has been provided.

In ERICA D5, examples are given for several cases
as follows

• To apply the SSD methodology to introduce more
  ecological realism by
• (1) using more conservative levels of protection
  (95 to 99 )
• (2) applying trophic/taxonomic weightings that
  better describe the structure of a specific
  ecosystem
• (3) restricting the statistical analysis to a
  particular endpoint
• (reproduction) and/or a
  particular trophic/taxonomic group (fish).

• To focus on the protection of keystone species
  and/or endangered species
• with a specific search in FREDERICA.

• To refine the effects analysis with additional
  experimental studies and
• modelling. Two extrapolation issues of concern,
  i.e. individual to population
• and external to internal irradiation effects.