Recent Developments
MS-Combo Model
MS-Combo model calculates the
probability of developing any combination
of multiple tumors observed in a bioassay
P(d) = 1 - exp{(/S0 + M + p2d2 + )}
Implemented at the recommendation of
the IMRC
Ł
1
BMDS Wizard
Excel workbook that simplifies BMD
modeling by providing a structured
interface to maintain all inputs, outputs,
and decisions made in the modeling
process
CatReg
Allows for the meta-analysis of toxicity
data from multiple studies, endpoints, and
test species
Estimates the probability that a response
of a severity level (s) or greater occurs,
given a concentration (C) and duration (t):
P(Y > s|C, t) = ll\as + pls * C + p2s * t
Log Duration
Risk Assessment Impacts
Human Health assessments-
implementing scientifically sound
dose-response methods ensures that
EPA (IRIS and PPRTV) assessments
reflect the strongest science possible
Modernizing risk assessment
methods - developing new dose-
response methods advances the
practice of dose-response analysis in
EPA, allowing for the better
characterization of uncertainty and
variability, quantifying incremental
risk, and addressing susceptibility
www. epa. Qov/ncea/bmds
Allen Davis- 513-569-7024
Davis, alien @epa. qov
Jeff Gift- 919-541-4828
Gift. ieff@epa. qov
EPA's Bench
Dose Software
(BMDS)
EPA's Benchmark Dose
Software is the primary dose-
response tool for use in human
health risk assessments within
the EPA and globally.
Currently\ there are over 5,000
registered users across 90
countries. BMDS supports the
use of BMD methods by
Agency partners and a wide
array of stakeholders,
including international', federal
and state regulatory agencies,
industry, scientific
organizations, academia, and
others.
-------
Benchmark Dose Modeling
Accepted as a default dose-response
modeling approach by US EPA
National Center for Environmental
Assessment (NCEA) built and supports
Benchmark Dose Software (BMDS,
current version 2.7.0.4) to facilitate BMD
analyses
Currently, BMDS supports the modeling
of dichotomous, nested dichotomous,
continuous, repeated measure, and
concentration x time data
9
Św - IU-II.I-
* Z I a'n
r
,.._ r :
:
j f.
IBS
-r4^
Terminology
Benchmark Response (BMR) - a change
in response relative to background
response in controls
Benchmark Dose (BMD) - the dose
associated with a selected BMR
Benchmark dose lower confidence level
(BMDL) - a one-sided confidence
interval on the BMD, usually 95%
Model Averaging
Future Directions and User
Support
NCEA and NIOSH developing draft
Bayesian model averaging method to
address model uncertainty
Estimates a weight-averaged
BMD/BMDL from all models being
considered
Uses maximum a posteriori methods and
Laplace approximation-based model
weights
I
'Best" BMD "Lowest" BMD
Ś - -+
r
Ś H n
MA may reduce uncertainty in
BMD estimates
MA BMD
(Shao and Gift, 2014)
MA may enhance accuracy of
BMD estimates
New GUI under development
Allows for assignment of model
parameter priors and model weights,
allowing for incorporation of biological
or other prior information
For example, information of a particular
endpoint's mode of action may support
weighing non-linear models more
heavily than linear ones
Future Directions
Hybrid approach-
characterizes
continuous risk
using the percentage
change of a popu-
lation in the tail of
the distribution
Log-normal distribution - allows for the
user to assume responses are log-
normally distributed.
Model with Nofmality Assumption
Probabilistic dose-response methods
have been proposed (NRC, 2008; 2013)
to assist risk management decisions
Meta-analysis tools using Bayesian
statistics and hierarchical modeling will
be used to support future EPA health
assessments
User Support
BMDS features a graphic user interface
that facilitates efficient modeling,
interpretation of results and reporting
capabilities.
The BMDS website (epa.gov/bmds)
contains a full suite of training
materials, including a Quickstart guide,
numerous training webinars, and links
to currently BMD technical guidance
documents.
------- |