use of epidemiological studies for setting a health- based ... › sites › default › files ›...
TRANSCRIPT
![Page 1: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/1.jpg)
Use of epidemiological studies for setting a health-
based guidance value
Þórhallur Ingi Halldórsson
Faculty of Food Science and Nutrition
University of Iceland ([email protected])
![Page 2: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/2.jpg)
Setting health based guidance values (HBGV)
Hazard characterization:
• Identifying critical outcomes and studies.
• Determine point of departure …
• ..preferably by using bench mark dose analyses
• Can the same methodology be used for human
observational studies?
![Page 3: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/3.jpg)
Controlled animal experiments why use something else?
• The controlled setting is major strength that cannot be matched by
human observational studies (when studying potential adverse effects)
• More time and cost effective than large scale epidemiological studies
• Subject to far fewer ethical constraints……but
![Page 4: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/4.jpg)
Controlled animal experiments why use something else?
• The controlled setting is major strength that cannot be matched by
human observational studies (when studying potential adverse effects)
• More time and cost effective than large scale epidemiological studies
• Subject to far fewer ethical constraints……but
≠
![Page 5: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/5.jpg)
HBGV: critical outcomes and critical studies
• The luxury of working with controlled studies often
results in large weights being given to one study
– ….most critical outcome
– ….most critical study
• The study population and the environment partly
determines the dose response
![Page 6: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/6.jpg)
Determining point of departure
GUIDANCE?
≠
![Page 7: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/7.jpg)
Bench Mark Dose (BMD)
• More precise in determining
POD than use of NOAEL
• Takes biological relevance
into considerations
• The model fit provides
some information on
– study quality
– uncertainty
– and biological plausibility
![Page 8: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/8.jpg)
EFSA guidance
![Page 9: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/9.jpg)
EFSA guidance
• Maximum likelihood
• Model averaging for quantile
data
• Restricted set of models
• The continuous models are not
conventionally used in human
epidemiology
![Page 10: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/10.jpg)
Existing guidance designed for controlled animal experiments
• The analysis of human dose–response data
can be more complicated than that of
typical data from animal studies, due to
confounders and imprecision in the
exposure estimates.
• In principle, the BMD approach would
also be applicable to human data.
• Opportunities for modeling human are
more limited. Studies are less standardized
and the modeling often involves additional
considerations, such as adjusting for
covariates.
• Sometimes human toxicological data are
reported in ways that are similar to the
reporting of toxicological .... in these
cases, this guidance document would be
applicable
![Page 11: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/11.jpg)
Existing guidance designed for controlled animal experiments
• The analysis of human dose–response data
can be more complicated than that of
typical data from animal studies, due to
confounders and imprecision in the
exposure estimates.
• In principle, the BMD approach would
also be applicable to human data.
• Opportunities for modelling human data
are more limited. Studies are less
standardized and the modeling often
involves additional considerations, such as
adjusting for covariates.
• Sometimes human toxicological data are
reported in ways that are similar to
animal data .... in these cases, this
guidance document would be applicable
![Page 12: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/12.jpg)
Existing guidance designed for controlled animal experiments
• The analysis of human dose–response data
can be more complicated than that of
typical data from animal studies, due to
confounders and imprecision in the
exposure estimates.
• In principle, the BMD approach would
also be applicable to human data.
• Opportunities for modelling human data
are more limited. Studies are less
standardized and the modeling often
involves additional considerations, such as
adjusting for covariates.
• Sometimes human toxicological data are
reported in ways that are similar to animal
data .... in these cases, this guidance
document would be applicable
No further guidance on human studies is then given
![Page 13: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/13.jpg)
Access to data
• Previously there have been few reasons to report data in a BMD usable format.
• Humans ≠ animals: Data protection rules apply
• Public health authorities need to do their homework as well to facilitate transfer of data
“The EPA recently proposed excluding from
consideration in setting environmental
standards any studies whose raw, individual-
level data are not publicly available”
![Page 14: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/14.jpg)
What characterises use of BMD
modelling
![Page 15: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/15.jpg)
…for animal studies
What characterises use of BMD
modelling
![Page 16: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/16.jpg)
In animal studies you have
1. Controls (zero dose)
2. Large exposure range
3. Controlled conditions →
low risk of bias
4. Variability is small
5. Detailed clinical
assessment of health
outcomes
George, JD et al 1992: The developmental toxicity of
ethylene glycol diethyl ether in mice
![Page 17: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/17.jpg)
In animal studies you have
George, JD et al 1992: The developmental toxicity of
ethylene glycol diethyl ether in mice
1. Controls (zero dose)
2. Large exposure range
1000-fold here
1. Controlled conditions →
low risk of bias
2. Variability is small
3. Detailed clinical
assessment of health
outcomes
![Page 18: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/18.jpg)
In animal studies you have
1. Controls (zero dose)
2. Large exposure range
3. Large spacing between
doses (not always)
4. Variability is small
5. Controlled conditions (in
theory no bias)
NOAEL
LOAEL
D = 350mg/kg/bw
George, JD et al 1992: The developmental toxicity of
ethylene glycol diethyl ether in mice
![Page 19: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/19.jpg)
In animal studies you have
1. Controls (zero dose)
2. Large exposure range
3. Large spacing between
doses
4. Variability is small
5. Controlled conditions (in
theory no bias)
George, JD et al 1992: The developmental toxicity of
ethylene glycol diethyl ether in mice
![Page 20: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/20.jpg)
In animal studies you have
1. Controls (zero dose)
2. Large exposure range
3. Large spacing between
doses
4. Variability is small
5. Controlled conditions
George, JD et al 1992: The developmental toxicity of
ethylene glycol diethyl ether in mice
![Page 21: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/21.jpg)
… and after that: uncertainty factors
![Page 22: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/22.jpg)
…for human studies
What characterises use of BMD
modelling
![Page 23: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/23.jpg)
In human studies you have
1. No zero dose: just subjects with relatively low exposure
2. Often small exposure range
~10-20
~100-300
PCB-153
![Page 24: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/24.jpg)
In human studies you have
1. No zero dose: just subjects with relatively low exposure
2. Often narrow exposure range
PCB-153
![Page 25: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/25.jpg)
In human studies you have
• And this determines where your observations are
placed on the theoretical dose-response curve:
– NULL
– Non-Linear
– or ~linear response
![Page 26: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/26.jpg)
In human studies you have
• And this determines where your observations are
placed on the theoretical dose-response curve:
– NULL
– Non-Linear
– or ~linear response
– or ~linear response
![Page 27: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/27.jpg)
In human studies you have
• And this determines where your observations are
placed on the theoretical dose-response curve:
– NULL
– Non-Linear
– or ~linear response
![Page 28: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/28.jpg)
In human studies you have
1. No zero dose: just subjects with relatively low exposure
2. Often narrow exposure range
3. Spacing between doses is not a problem: exposure is determined by how free
living subjects behave individually
The NOAEL/LOAEL issue is often not relevant
The benefit of using BMD approach relates to
estimating POD taking biological relevance
into consideration NOAEL
LOAEL
![Page 29: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/29.jpg)
In human studies you have
1. No zero dose: just subjects with relatively low exposure
2. Often narrow exposure range
3. Spacing between doses are not a problem
4. LARGE variability (it should be )
Raises some questions on the use of BMDLs
![Page 30: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/30.jpg)
In human studies you have
1. No zero dose: just subjects with relatively low exposure
2. Often narrow exposure range
3. Spacing between doses are not a problem
4. Variability is LARGE
5. And you cannot have controlled conditions
![Page 31: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/31.jpg)
Usually no uncertainty factors
- What you see is what you get -
![Page 32: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/32.jpg)
Conclusion
• There are no major obstacles for using human data to derive HBGV
• BMD analyses can easily be performed but existing conventions may not
be directly applicable and more work is needed (as has been done for
animal data)
• Lack of individual participant data is unlikely to be a key issue
• Due to high variability and varying sample size the use of BMDLs for
human data needs some careful considerations
• Modelling should not be done for the sake of modelling.
![Page 33: Use of epidemiological studies for setting a health- based ... › sites › default › files › event › 180918-co… · Use of epidemiological studies for setting a health-based](https://reader033.vdocuments.mx/reader033/viewer/2022060316/5f0c03527e708231d4335328/html5/thumbnails/33.jpg)
Thank you