Are DEBtox models applicable for acute exposure data?

[Annika M.-D.]

Hi colleagues,

I realized that DEBtox models seem to use only datasets of chronic exposures. This makes sense, as the response to a stressor develops over time and when the time of observation is shorter than the time to the (measurable) effect, it will be difficult to fit a DEBtox model. But I guess if the endpoint is sensitive already during short periods, it should be fine, right?

So, I was wondering if there are some examples of DEBtox models parameterized or applied to acute toxicity tests (let's define "acute" to less than a week of exposure).

Cheers,
Annika

[Tjalling]

So, you're using "acute" strictly as a measure of test duration, and not to mean "short term survival", I guess (there is often confusion about this in ecotox). In that case, I would define "acute" not as an absolute time period but relative to the life cycle of the species (at that temperature). There are many animals that complete their whole life cycle in a week, easily .

To fit a DEBtox model, without using any a priori information, you would need growth and reproduction data over a good part of the life cycle. To short and you won't be able to fit all basic parameters (and the tox data you probably cannot fit either). If you do use a priori information (e.g., from add-my-pet, from other experimental work, or "educated guesses"), you would be able to parameterise the basic model. However, there would be no way to tell if it is actually working as it should, so it would not be very convincing IMO. Second problem would be whether you can establish the mode of action in a short test, and the third problem whether there is enough information to parameterise the TKTD module. As you said: you need to have substantial effects to fit the model, and even fast and strong toxicity will take time to have a clear effect on growth and reproduction, relative to the control (for survival that would be much easier, but we have GUTS for that).

My guess is that these problems will rapidly make the whole exercise rather futile on "too short" experiments, but I must admit that I haven't tried. I will have a look in the archive to see if there are examples for very short tests, but I think there's good reasons why they are so scarce .

[Annika M.-D.]

Thanks Tjalling!
I think you made some good points. I would be interested if you can dig something up. Or maybe someone else reading this remembers a study example.

Another angle to this question is, whether it makes sense to use "acute" data for validation. I guess this should not be such a big problem, once the parameters are properly estimated. Any thoughts on that?

[Tjalling]

For validation, you can be a bit more flexible since you would not need to fit the model to the data set. However, for a convincing proof that the calibrated model can predict something, you would still need to be able to see effects. If the test is too short to really spot effects on growth and reproduction, it would also not help in validation.

I have given this whole issue a bit more thought, as I think it is an interesting one. I think the problem with effects on growth and reproduction is that we cannot measure the rate of the process. We can only measure the integrated result over time (body size, number of produced eggs). So a change in a process rate always needs a bit of time to show up as a measurable effect in the integrated, observable, endpoint. Now, if we could measure process rates, the situation would be different. For example, we can measure respiration rate of an animal. This is a bit difficult to interpret since multiple processes contribute to respiration. However, it should be easier to work with short experiments to get meaningful results.

Maybe this is also a place where 'biomarkers' may play a role. If such a marker can be linked to a mass flow in a DEB model ...