Can you draw any insights from the inconsistencies in results between F344 (diet) and CD (gavage) studies where the dose to targets arguably could have been higher in the negative (CD) study? Is this simply a reflection of strain differences or can we say anything about dose-rate effects and toxicity?

What is the concentration of 1,3 D in Telone II and in DD-92? Are they the same?

I agree that information on possible differences in composition of the test material in the two oral rat studies is crucial

I'm not sure where any further discussion of historical controls should go (here or on some other page), but I continue to be somewhat skeptical of Table 16 in the White Paper. As I wrote under topic 5.5 in the prior round, "Table 16 in the white paper represents extreme cherry-picking. The historical control rate for male F344 rats at Dow in the year of the study is 2% (pooled 5/250). The complete control rate in Table 16, for all years, is 4.6% (40/865). Only by picking the highest possible control rate for males (16% in one year) and adding an irrelevant value for females (8%) was the white paper able to inflate the control rate by an amount large enough to make the 18% incidence at 25 MKD seem POSSIBLY not significant."

So comments above to the effect that the liver tumor incidence in mid-dose males (12%) was within historical control rates is not really fair-- and of course that says nothing about the 18% incidence in the high-dose males. If these tumors are to be ignored for biological reasons, so be it, but I don't think it's appropriate to ignore them as "elevated in the experiment but not elevated if you happen to compare them to either of 2 hand-picked prior experiments out of the 17 that we have data for..."

Somewhat agree with user-37600 [Expert 2] , but it is strange that the same dose, but a much lower Cmax causes tumors in one study. Same dose, but much higher Cmax does not do anything in a 2nd study ? Again emphasize that detailed information on the composition of the materials tested could help explain the complex dataset. It is not known what other impurities except epichlorohydrin were/are present and if variable impurities contribute to the inconsistent results.

I agree with 477751 [Expert 11] immediately above-- just flagging that my comment above was about historical controls GENERALLY, and I'm not sure if any of the pages on this site have a discusssion of the way the White Paper (mis?)uses historical control data to suggest that statistically significant increases are not what they seem. I'm happy to move this comment and my prior one to a different page if asked.

This comment pertains to the comment by user-37600 [Expert 2] above questioning the use of historical control data as described in the White Paper with reference to Table 16. In my experience , historical control tumor data are often used to aid in the interpretation of tumor data from a specific study. The historical control incidences represent a range of incidences that may occur in a control group by chance alone. If the tumor incidence in a specific study falls within the range of historical control incidences, it is fair to conclude that this tumor incidence may have occurred by chance alone and may not be treatment-related. Application of HCD in this way is fairly standard and is not considered to be "cherry picking". At the same time, application of HCD in this way does not "prove" that a specific tumor incidence was not treatment-related. This is true for the liver adenoma incidences in the Scott, 1995 study. We should also note that whereas Table 16 lists HCD on liver tumors over a 21 year interval (1991-2012), the highest HCD incidence in males (16%) occurred in 1992 and the highest incidence in females (8%) occurred in 1997. Since the Scott study was completed in 1995, the highest HCD incidences occurred within a few years of the completion of the Scott study. The time frame of these HCD data in relation to the Scott study increases their relevance in terms of interpreting the tumor data from the Scott study.

In response to user-750802 [Expert 9] : I agree that historical controls are relevant; my problem is with the non-statistical handwaving inherent in Table 16 and the accompanying narrative. The historical control rate is 4.6%. This number has some uncertainty around it, but the right way to portray the uncertainty is not to choose the single highest experiment out of about 20 different experiments and call that the "upper bound" in order to say that 18% (four times the pooled historical control rate) is "within the range."

(PS: the simplest formula for the uncertainty in the pooled historical control rate of 40/865 (4.6%) gives 95% confidence limits that extend from 3.2% to 6.0%, both very far below 18%. This formula assumes that there is not some systematic effect making the individual experiments non-random, so if in fact the control rate was changing from year to year not by chance, these limits would be too narrow. But the White Paper makes no attempt to posit any factor that would make some control rates relevant to the experiment being analyzed and others not, let alone to posit some model for pooling the rates in some non-uniform way).