|
VOC Cancer Risk Simulation Model
A new
model that integrates pharmacokinetic, cytotoxicity, cell proliferation,
and stochastic carcinogenicity components into a unified framework is
available to download here.
The
Cyclophosphamide Hematotoxicity BBRA Model
To experiment with this model, there are
three items to download:
1. The free run time
version of ITHINK 5.0 (download here)
from High Performance Systems. This is a continuous simulation modeling
environment that will enable you to run and use simulation models created
in ITHINK 5.0. You will not be able to save changes that you make in the
models, however, unless you buy a full version of ITHINK 5.0 from High
Performance Systems.
2. The Cyclophosphamide
Hematotoxicity BBRA Model (download here ~
140KB) developed for the American Petroleum Institute by Cox Associates.
This is a specialized tool that simulates the effects of the drug cyclophosphamide
on blood cells, especially CFU-GM stem cells and both earlier and later
blood cells in the granulocyte- macrophage (GM) lineage. The model takes
a biologically based risk assessment (BBRA) approach,meaning that it simulates
the flows of cells among physiologically important compartments in response
to user-specified dosing scenarios. The model is intended primarily to
simulate responses to low to moderate doses that are small enough so that
full recovery can occur after cessation of dosing. Long-term irreversible
poisoning of the stromal microenvironment is not modeled, but cell kinetics
are described in enough detail to make a variety of testable predictions.
3. A document (download
here ~1MB) describing the Cyclophosphamide
Hematotoxicity BBRA Model and how to use it. This is an electronic document
written in MS Word for Office 97. The current version contains some hyperlinks
to key references and web resources; others will be added over time.
The detailed technical
background for the model and document, including a tutorial review of
normal and perturbed hematopoiesis, is contained in reports to the American
Petroleum Institute prepared by Cox Associates since 1993. Click here
for a summary overview of the model and its validation to date using experimental
and clinical data.
Bayesian methods for assessing uncertain exposures
Click here to download presentation
Marketers and biostatisticians
must often try to estimate a statistical relation between "exposure"
and the probability of a "response" when the true exposure values
are unknown and only rough estimates are available. Applying standard
text-book methods that ignore errors in the estimated values of the independent
variables may yield misleading or meaningless results. Moreover, the conventional
wisdom that ignoring such errors attenuates estimated relations between
exposure and response does not hold when more than one independent variable
is involved. This presentation summarizes recent progress in numerical
Bayesian methods for estimating correct exposure- response relations even
when the correct values of exposure are highly uncertain.
MOLECULAR BIOLOGY OF CHEMICAL LEUKEMOGENESIS
FOR BENZENE
In February, 1999,
Cox Associates delivered to Exxon Biomedical Sciences, Inc., a review
of the detailed biological mechanisms by which benzene can cause hematotoxic
damage and induce secondary Acute Myeloid Leukemia in humans. A key finding
from this study is that many of the mechanisms of benzene-induced health
effects are sub-linear at low doses, helping to explain why recent epidemiological
studies do not find the excess leukemias and health damage at low exposure
concentrations (e.g., 1 ppm or less) predicted from older benzene health
risk assessment models.
Top
of Page
CLINICAL APPLICATIONS OF CYCLOPHOSPHAMIDE
MODEL?
On December 15th,
1998 a Hematoxicity Modeling Workshop was held at the University of Ottawa
Institute of Population Health (Professor Dan Krewski, host), in conjunction
with the Benzene State of the Science Workshop, 1998. The topic of the
one-day Hematotoxicity Modeling Workshop was a review of the simulation
model of cyclophosphamide hematotoxicity developed by Cox Associates
for the American Petroleum Institute. Reviewers included government scientists
from EPA and NIEHS, biomathematical modelers from the Fred Hutchison Cancer
Research Center and the University of Ottawa, and experts in clinical
hematology and research from the University of Ottawa and the University
of Colorado. Workshop papers will be submitted for publication in 1999.
The main finding, reported to the Benzene State of the Science Workshop,
was that the model appeared to offer useful, apparently realistic predictions.
An unexpected finding was that several of the expert hematologists felt
that the model might be useful in clinical practice, in refining dose
regimens for chemotherapy patients.
Top
of Page
Cox Associates' ongoing
applied research in data mining technologies has recently led to a promising
new application: using data mining algorithms to combine predictions from
different expert systems for predicting which chemicals are carcinogenic
without undertaking expensive experiments to find out. We discovered that
the new prediction-combination technique, which is based on classification
trees, can produce "hybrid" predictions than are more accurate
than any individual expert system's prediction and also than the predictions
from previous combination methods. Click here
to download a technical paper describing the new approach and results.
Top
of Page
In September of 1998,
Cox Associates completed a new approach to finding the probable locations
of greatest soil concentrations of contaminants based on soil samples.
Traditional geostatistical and sampling and estimation methods such as
kriging do not fully address the problem of efficient search for buried
hazards when the spatial distribution of yield is highly uncertain. Our
new approach adaptively optimizes the search process so as to continually
maximize its expected yield. This can dramatically cut the sampling costs
needed to show that spatial hazards have been adequately reduced by sampling
and remediation efforts. Details may be found in the attached technical
paper. Click here to download the paper.
Top
of Page
Since 1986, Cox Associates
has developed biologically-based risk assessment (BBRA) simulation models
and methods for understanding the human health risks from chemical exposures.
A program developed for the American Petroleum Institute (API) to predict
the effects on blood cells of dosing with the immunosuppressive drug cyclophosphamide
is now available to other researchers from our web site. Click here
to download the required modeling software and documentation.
Top
of Page
CAUSAL ANALYSIS AND PUBLIC POLICY: THE CASE
OF DIESEL EXHAUST
In early June 1998,
Cox Associates delivered written comments on behalf of the Engine Manufacturer's
Association to the Clean Air Science Advisory Committee (CASAC) on EPA's
recent (2-98) Draft Risk Assessment for diesel exhaust health risks. The
comments address causation (as opposed to statistical associations) in
epidemiological literature on health risks of diesel exhaust. It concludes
that no causal link between diesel exhaust and human lung cancer has been
found.
METHODS: The document
provides references and web links to statistical resources that help untangle
causation from statistical association. This is a novel application of
insights and methods from our data mining and causal forecasting practice
areas.
IMPACT: CASAC has
directed EPA to rework its draft risk assessment to address issues identified
in public comments by Cox Associates and others. The comments document
identifies current methods for overcoming modeling challenges in this
area.
For information,
please see the attached report.
CLIENT CONTACT: Glenn
Keller, Engine Manufacturer's Association, 312-644-6610
Top
of Page
|
