A U.S. government scientist poses the following questions on dose-spacing and the inclusion of "human-relevant" dose groups for toxicity testing purposes ($500 for best answer): PAID

The best way to set doses in animal studies is to use the onset of nonlinearity in the systemic dose of a chemical and/or its metabolite(s) (biomarkers). Using this method will take care of the issue of absorption as the systemic doses will be associated with chosen external (nominal) doses. The highest dose should be chosen slightly above the POD from linearity or KMD and less than MTD as the appearance of systemic dose nonlinearity is a biological response that occurs at a much lower dose than MTD - use of MTD is subjecting animals to cruelty. The other two doses can be selected at 10 and 100 or 3 and 10 fold lower than the highest dose. Extrapolating the two lower (or the lowest) dose, where the biology of the animals is not compromised, to human-relevant doses will be more meaningful than using doses that are cause some effect in animals. In case to determine the profound toxicity of a chemical, a dose at MTD can be added to satisfy some of the regulatory agencies still asking to conduct studies at MTD; however, I am not sure what beneficial information would be gleaned from using an MTD dose. Alternatively, when a better titration of dose-response is needed one of more doses can be added at the lower end.

The common testing procedure is to test at a dose equal to the MTD and then to add one or two doses at fractions of the MTD.  Testing at the MTD has the potential advantage that a negative result has about the best assurance available from the bioassay that the health effect is not caused in the test species by this substance and this route of exposure (absent some quirk in the metabolism or distribution of the substance).    

Testing at “human relevant” doses (doses similar in some sense to doses in a human population) seldom provides useful information about the risk at doses to which human are exposed (because each animal tested will be a stand in for perhaps thousands of humans and consequently an increase in a serious health effect such as death by as much as 1 %, although likely considered an extreme risk among humans, would be very unlikely to be detected in a standard bioassay with 50 animals per dose.   Testing at very small doses (e.g., ≤ 1/10 of the MTD), would likely have the same effect in a trend test as testing at zero dose, so results of such testing would be equivalent in a practical sense to increasing the size of the control group. 

Given this, the best use of multiple doses below the MTD are probably in determining the shape of the dose response.  I.e., does the dose response appear to be linear or non-linear?  Perhaps the way to address this question is to have several doses (say, two or three) in addition to the MTD spread out between around 30% of the MTD and the MTD.  

 

The NAS Committee on Risk Assessment Methodology (1993) authored the document, “Issues in Risk Assessment”, which discusses this issue at some length, and suggests several options. 

Testing the estimated human relevant dose (x), 3x, 10x for pharmaceuticals. Adding a 100x for industrial chemicals, pesticides or food chemicals.

The Dosing should be based on TK, MoA, in vitro and read-across data, if possible.

Avoiding MTD or doses causing severe toxicity, specially in DART studies.

See ECETOC Guidance, 2021 https://www.ecetoc.org/wp-content/uploads/2021/03/ECETOC-TR-138-Guidance-on-Dose-Selection.pdf

In my opinion, it is correct to test human-relevant exposure levels. However, the human relevant exposure levels are not always known and may vary according to different factors. In addition, it is always important to have,  in the toxicity study, doses that can cause adverse effects in order to then test lower doses and determine the NOAEL . Therefore, dose range finding studies are important to inform about the doses which cause adverse effects. From the dose range studies, the lower doses can be defined.  

 I don't see how you can get away from some version of the MTD as the presumptive default for the limit dose.  There may be exceptions, e.g.  human pharmaceuticals but even there toxicity often becomes evident at doses not much higher than the therapeutic  dose.   Having said that, there should be room to make that a rebuttable presumption based on ADME, chemical class or other factors which may be specific to the regulatory paradigm.   In short, hypothesis driven generation of experimental data  should always  be considered an acceptable alternative to the generic definition of MTD. 

The low dose in a tox study should ideally deliver an exposure around the expected human exposure. An MTD should be established at some point during product development. Alternatively, the high dose should provide a large (50X) safety margin for exposure. The mid dose should represent an exposure (not dose) in between.

I agree that  MTD is criticized, and it will be better to use "human relevant" exposure levels for toxicity testing. As regards, proper dose spacing it can detect NOAEL or is feasible to calculate MOE with non genotoxic chemicals. With genotoxic chemicals, proper dose spacing should be appropriate to estimate slope factor or unit risk. 

The use of pharmacokinetics has been largely overlooked when selecting doses for a toxicity study. Bioavailability (fraction actually bioavailable for toxicity), clearance half-life, and measures of internal dosimetry (e.g., plasma levels) are critical parameters for dose spacing and the inclusion of the most relevant doses for humans. When actual pharmacokinetic information is not available for dose selection/spacing, various predictive tools are available for parameter estimation or analogue identification for which there may be data available. The most useful toxicity studies are those designed with dose spacing based on pharmacokinetics information in the test species as compared to humans.