Bioequivalence and Bioavailability Forum

Hello all,

we would like to conduct a parallel study with 80 subjects.
Gender and weight are known to affect the PK and that is why we want them to be balanced across treatment groups. Therefore, we would like to stratify randomisation by gender and weight as also suggested by ICH M13A:
2.2.3.4. Parallel design studies
The statistical analysis for randomised, parallel design studies should reflect independent samples. Demographic characteristics or other relevant covariates known to affect the PK should be balanced across groups, to the extent possible. The use of stratification in the randomisation procedure based on a limited number of known relevant factors is therefore recommended. Those factors are also recommended to be accounted for in the primary statistical analysis.

We are wondering how to exactly do stratified randomisation (with let's say blocksize of 4). If we categorize weight into groups of 5 kg (51-55 kg, 56-60 kg, 61-65 kg …) and gender of course into M and F, we theoretically get around 20 strata (realistically probably a few less because not all strata would have at least 1 subject, e.g. 53 kg male) which is most likely too many as the sample size is 80 subjects.

How do others deal in such cases?

One solution could be to decrease number of groups of factor weight so that we would e.g. have groups of 10 kg (51-60 kg, 61-70 kg ...) but then risk to not get balanced groups is a little bit higher.

In any case, is it possible to include subjects based on the screening results in order to try to get as complete strata as possible? E.g. after the screening we have just 1 male with weight in stratum 61-65 kg and therefore he would not be included in the study but other subjects would rather be included who would be in complete strata - e.g. 8 females with weight 56-60 kg would be included. On the other hand we dont know if we would still get complete strata in such case, maybe it would be a little unbalanced.
Another option would be to make "pairs of twins" (blocksize of 2) as similar as possible.

Regarding the statistical analysis, if gender and weight do affect the PK then it would probably be beneficial to include these 2 factors in ANOVA model along with factor Treatment? MSE and therefore CIs should be narrower?
Guideline on adjustment for baseline covariates in clinical trials:
The main reason to include a covariate in the analysis of a trial is evidence of strong or moderate association between the covariate and the primary outcome measure. Adjustment for such covariates generally improves the efficiency of the analysis and hence produces stronger and more precise evidence (smaller P-values and narrower confidence intervals) of an effect.
Has anyone ever included a covariate in the ANOVA model when analyzing a parallel study (or maybe even in crossover studies)? Was it beneficial (narrower CIs) or not?

I reanalyzed some older parallel study (other API) with 2 additional covariates (besides factor Tretament). The effect of these 2 covariates was statistically significant (p<0.05) and obtained CIs were much narrower. I am wondering that if these 2 covariates werent statistically significant (lets say p=0.8), then the CIs would not get narrower (instead they could even be a little wider)?

Regards
BEQool