Here's a good answer to quiz question about thalidomide:
With thalidomide the first "evidence" that animal testing did not apply to humans was the lack of sedative effect (by a human sedative) on the animals tested. Studies found that while an increase in phocomelia occurred in humans, no defects were found in rat and mice populations tested. In rabbits there were some defects but not of the same type. It took testing with primates, such as the marmoset, to replicate the devastating effect the drug had on human fetuses. This shows the limits and dangers of using animal models to determine human effects (teratogenic or otherwise).

***Q. Why is the RfD sometimes based on the LOAEL and sometimes based on the NOAEL?  (I realize it is also sometimes based on the benchmark dose, if there’s a specific effect they’re worried about).  Wouldn’t you always want to be on the safe side and base it on the NOAEL and not the LOAEL?  Also, why would the EPA base the NOAEL on the level below which mottling occurs if they do not consider mottling to be an adverse health effect and consider it to be only cosmetic?
A. There are technical issues with very low doses, and often they do not find a NOAEL.  That is, there were some effects even at the lowest dose tested.  It is often impractical, especially with a two year test, to do it again.  They account for this in the uncertainty factor they use [http://www.raperkins.org/ENVE_652/Module05/Submodule5A/5A_5.htm].  Clearly the effects they see at the LOAEL are small.  If they were great and they had the money, they would repeat using smaller doses.

***Q. How does the use of benchmark dose (BMD) impact the RfC and RfD, and what is its importance compared to using NOAEL and LOAELs when calculating RfC and RfD? The effective dose when using the BMD is lower than that when using noael and loael. http://www.epa.gov/iris/subst/0039.htm
A. First of all , remember how they weigh pigs in Bavaria. Next, for popular chemicals, i.e., ones for which there is money to study them, there have been lots of tests done. Tests use different species, etc. But even in one species, different laboratories will report different numbers. Most often the numbers have a mean and standard deviation. It is not justifiable to just take the lowest number, and decide that is the LOAEL or NOAEL, although, by definition, it is. A better way is to take all the relevant data you have, and treat it statistically. When you do, you can plot an upper and a lower 95% confidence level. I showed how that graph flares out that the far end, i.e., near the LOAEL. So, the “benchmark dose” is the upper 95% confidence level at the LD10 or ED10, where the confidence limits are narrower. That is the point of departure for the extrapolation to zero effect at zero dose. Note this is very similar to the Linearized Multistage model for cancer extrapolation, but that uses the beta distribution to fit the point, because cancer is a multistage process and the beta fits that best. For the benchmark, non-cancer dose, a different curve might be used to fit the data. Probably a Probit.

** Q. From IRIS:
Oral Slope Factor -- 5.7E-3/mg/kg/day
Drinking Water Unit Risk -- 1.6E-7/ug/L
Extrapolation Method -- Linearized multistage procedure, extra risk
Drinking Water Concentrations at Specified Risk Levels:
How is EPA able to extrapolate data to gain inferences of risk levels in drinking water? Do they use the data they gain from studying oral doses in rats? This just seems like a difficult, and not very trustworthy, extrapolation. Can you further explain this to me? I must be missing something.
A. They should compute, that is, if you knew the amount of water people drink, (we do this in ENVE651) the translation between Oral Slope Factor and Drinking water should be straightforward. Do you use the average water consumption, or the 95% upper confidence level of water consumption? I don't know - I'm a toxicologist, not a regulator.

**Q. When referring to multi-generational studies, what does the ‘F' in F1 and F2 actually stand for? I assume the F1 generation are the “children” which the F2s are the “grandchildren.” http://www.epa.gov/iris/subst/0060.htm section 1.A.4
A. P stands for “parental,” F 1 is the first “filial” generation, the children, F 2 the second filial generation, the grandchildren. In Latin, filius is son and filia is daughter.

**Q. According to IRIS, an EPA contractor did a review in Aug 03 and found no critical new studies, and that this information had last been updated in 1989. As a result of earlier research, I have '93 and '96 profiles from EXTOXNET that have very specific information on studies and health affects. The '96 report also says it is in EPA toxicity Class III, slightly toxic. This information had to come from somewhere. Having dealt with EPA contractors in other areas, I find anything they say somewhat suspect. How does EPA classify or determine a "critical new study", and what information is reliable enough to be placed in IRIS?
A. Remember our module on scientific literature. There are top journals and bottom journals, although you need to be in field to know which is which. When scientists get together, they often discount many of the studies. It take money to do the testing, and the tests are not cutting edge research and little “scientific” knowledge is gained by seeing how many rats die. The science is in causes and mechanisms, not counting tumors. So, much of the testing is not reported in the literature. Some of the literature is a regurgitation of older testing. Sometimes things are reported in the lower level journals that are just someone's opinion. In the final analysis, the people who sit down and look at the literature and write the report are the ones who define “critical.”

**Q. No data for an oral Rfd Assessment or inhalation Rfc assessment is available for  1,2-dichloroethane.  Will EPA ever do these assessments or are there not human populations to do such tests?
A. The tests may not have been done, or that it relatively non-toxic compared with its carcinogenic potential – therefore that is all that is studied.

** Q. The IRIS information provided for 1,3 Butadiene describes the RfC being calculated based on the most sensitive effects to chronic exposure in mice, however there are significant fetal effects which result from sub-chronic exposure at what appear to be lower doses, why is the chronic response used to determine the RfC and not the sub-chronic response?
http://www.epa.gov/NCEA/iris/subst/0139.htm
A. Remember how they weigh pigs in Bavaria.  Anyhow, it appears that they had strong confidence in the mouse two-year study and went with that.  The sub-chronic study used 1000 ppm, which is far higher than the RfC they derived from the two-year study.  It is quite common in tox literature of common chemicals that different labs get very different results.  The regulators do not always choose the most conservative results.

** Q. What is the statistical foundation of the uncertainty factors used by the EPA in the IRIS reports? How well have they served as indicators of future finding of clinical pathology?
A. There may not be statistics involved.  http://www.faculty.uaf.edu/ffrap/ENVE_652/Module05/Submodule5A/5A_5.htm
Say they only have one study, but it is a good one.  They take the NOAEL and apply the factors.  If they have several good studies, then they have to think, but not necessarily statisticate.  For example if they have some good studies, but with different species, which do they use?  Answer, usually, the most sensitive species and sex, as long as it is a standard species of lab animal, not an exotic species, or one known to be especially sensitive.  But suppose they have some good studies ( through studies by reputable labs) that are in agreement, pretty much, then some studies by lesser known labs or incomplete or poorly reported studies, but these some much greater sensitivity.  Again, they use judgment, not statistics.  Remember the pages on “extrapolation,” what is found in the animals may be quite different than what will happen in the human – but that extrapolation is the best that we can do.  See topic on PBPK later, which can help close the gap, sometimes.