Bioequivalence and Bioavailability Forum

Hi Hötzi,

❝ At the 2^nd GBHI conference (Rockville 2016) body weight as a covariate was a hot topic. Body weight changes quite often in studies of anti-cancer drugs. The FDA was happy with that. No way for the EMA – as usual.

So odd.

"The statistical analysis should take into account sources of variation that can be reasonably assumed to have an effect on the response variable."

A larger person has larger plasma volume and will, all other factors equal, be expected to have lower AUC and Cmax. Thus, body weight is definitely a source of variation that can be assumed to have some effect. And this is how it is in practice; those metrics correlate significantly with body weight in many trials albeit not always in a p<0.0000001 fashion when doing e.g. Spearman's rank or Pearson (but hey, period for example is usually not significant but has to be included in the model).

The fact that one can do potency correction in the right circumstances is in fact an introduction of a kind of ANCOVA principle. So why they decide to shut the door on body weight is not clear to me, is it clear to anyone? Does this really have to do with concern for sources of variation, or are we in an altogether different territory here?

Dear All,
I have one query, In PK bio-similarity study in healthy volunteer if the recommended dose is as per bodyweight (say 5mg/kg) then in this case subject with higher bodyweight will receive more total dose as compared to the subject with lower bodyweight. Will this going to impact the serum drug level? If yes then what corrections we need to apply?

Thanks in advance

Hello balakotu,

❝ When i keep totaldose is a factor along with normal fixed effects (seq sub(seq) per trt) the model is not execute, Please let us know what are the fixed effects is required in ANOVA model.

❝

❝ SAS code:

❝ model Cmax AUCt AUCi= seq sub(seq) per trt totaldose / ss3;


❝ test h= seq e=sub(seq)/htype=3 etype=3;

I am not a SAS user, so I googled this one. It is important to tell the program which ones are fixed factors and which ones are covariates.
Am absolutely not sure, but I believe you might need to use class for seq, sub(seq), per and trt to indicate that those are the fixed effects. For trt and seq this may not be an issue if the stats program does not coerce them to integers or other other numerics. I have no idea if SAS does this in the absence of a class statement.

Otherwise it could have no way to know if you did not mean per and sub(seq) as covariates. Bear in mind the levels of those are typically coded with integers like 1, 2, 3...