A few thoughts on these discussions:
-
The TD50s used for limit setting are typically the rodent TD50s, firstly since that is the standardised carcinogenicity assay (thus that exceptionally potent primate result for NDEA is ignored from the dataset), and secondly since rodent TD50s were used for the TTC and CoC initial development by Kroes, Cheeseman and co-workers. Additionally, a primate assay is less likely to be lifetime, due to longer lifespans, would use fewer animals for cost and ethics reasons, and is thus more typically used for mechanistic exploration rather than performed to make a formal determination of potency. In this case all non-rodent results seem to come from a single study looking at organ- and species- specificity in carcinogenesis (Adamson 1982).
-
Lhasa TD50s, for all that, as a Lhasa employee, I would argue are better, are based ultimately on the same data that the Gold TD50s were based on. The Lhasa TD50 calculation requires a certain threshold of data, such as having multiple dose groups, so there are many compounds (e.g. single-dose studies) for which no Lhasa TD50 exists; however, where both exist the values are typically very close (Thresher et al 2019 - we were aiming to replicate the original work of the Gold TD50s published in Peto et al 1984 but could not do so exactly), so it appears that the following process has been used:
- determine whether a Lhasa TD50 exists for the compound/study, and use that as an initial ‘quality gate’ for the carc data.
- there’s a caveat in this: For some reliable studies, such as the Peto et al 1991, we were unable to replicate the ‘lifetable’ method used by Gold, so there is no Lhasa TD50 currently published for this extremely-reliable study - we can calculate one, but it would be comparing apples with oranges, so the simple presence/absence of a Lhasa TD50 is not a complete quality gate.
- use the Gold TD50 for that compound/study, for consistency with e.g. the TTC and CoC data. -
There is a final subtlety, which is that the summary harmonic mean TD50s such as 26.5 ug/kg/day for NDEA are not the TD50s from a single most reliable study; a case could be made for the use of the TD50 derived from the 4000-rat Peto et al 1991 study, which would lead to 49.8 ug/kg/day (most sensitive organ/sex/species in most reliable study, as per M7). As I’ve mentioned in a couple of presentations ‘it’s hard to argue with 4000 rats’! The harmonic mean is a conservative way of determining a mean, and thus can be swayed lower by a single, lower-reliability, study that has an overestimation of potency w.r.t. the robust study.