Hi Ankit,
❝ Question: To address the challenges related to recruitment of cancer patients in PK BE study, we want to plan a full replicate design study. Will the EMA agency accept this approach even if we are not suspecting the drug to be highly variable?
❝ As per EMA guidance, “If an applicant suspects that a drug product can be considered as highly variable in its rate and/or extent of absorption, a replicate cross-over design study can be carried out.”
Let’s see what is stated in
4.1.1 Study design:
Standard design
If two formulations are compared, a randomised, two-period, two-sequence single dose crossover design is recommended.
Recommended does not mean that is is mandatory.
Alternative designs
Under certain circumstances, provided the study design and the statistical analyses are scientifically sound, alternative well-established designs could be considered such as […] replicate designs e.g. for substances with highly variable pharmacokinetic characteristics.
Replicate designs for HVDs are given as one of the
examples. State the recruitment issues as a justification in the protocol. Note that
any replicate study can be evaluated for conventional Average Bioequivalence (ABE) as well (see also
this article). From a purely statistical perspective, replicate designs are preferable to a simple 2×2×2 crossover and thus, there should be not regulatory concerns.
❝ As per USFDA guidance “A replicate crossover study design (either partial or fully replicate) is appropriate for drugs whether the reference product is a highly variable drug or not. A replicate design can have the advantage of using fewer subjects compared to a non-replicate design, although each subject in a replicate design study would receive more treatments”.
Correct – makes sense. Power – and thus, the sample size – depends on the total number of treatments. Therefore, a 4-period full replicate design requires about half as many subjects as a 2×2×2 crossover. An example with the
-package PowerTOST
:
library(PowerTOST)
CV <- 0.25 # assumed within-subject CV
theta0 <- 0.95 # assumed T/R-ratio
target <- 0.80 # target (desired) power
design <- c("2x2x2", "2x2x4", "2x3x3") # guess…
n.per <- as.integer(substr(design, 5, 5))
expl <- data.frame(design = design, n = NA, treatments = NA)
for (j in seq_along(design)) {
expl[j, 2] <- sampleN.TOST(CV = CV, theta0 = theta0, design = design[j],
targetpower = target, print = FALSE)[["Sample size"]]
expl[j, 3] <- sprintf("%.0f ", expl[j, 2] * n.per[j])
}
fmt <- paste0("Sample sizes and number of treatments for assumed CV = %.3g ",
"and T/R-ratio = %.3g,\npowered for at least %.0f%% ",
"and evaluation for Average Bioequivalence (ABE).\n")
cat(sprintf(fmt, CV, theta0, 100 * target)); print(expl, row.names = FALSE)
Gives:
Sample sizes and number of treatments for assumed CV = 0.25 and T/R-ratio = 0.95,
powered for at least 80% and evaluation for Average Bioequivalence (ABE).
design n treatments
2x2x2 28 56
2x2x4 14 56
2x3x3 21 63
Keep the potentially larger number of dropouts in replicate designs in mind (see
this article).
❝ We understand that widening criteria will be applied only to Cmax if we observe Swr > 0.294 or else conventional BE criteria will be applied.
Almost. For the EMA the decision is
not based on \(\small{s_\text{wR}}\) (like for the FDA) but on \(\small{CV_\text{wR}}\), where for reference-scaling \(\small{s_\text{wR}=\sqrt{\log_e(CV_\text{wR}+1)}}\).
- If > 30%:
- Reference-scaling by Average Bioequivalence with Expanding Limits (ABEL)
- Upper cap of scaling at \(\small{CV_\text{wR}=50\%}\) (maximum expansion 69.84–143.19%)
- 90% CI within expanded limits \(\small{\left\{L,U\right\}=\exp(\mp 0.76\cdot s_\text{wR})}\)
- Point estimate within 80.00–125.00%
- If ≤ 30%:
- ABE
- 90% CI within 80.00–125.00%
However, in order to apply ABEL you have to take into account whether
“a wider difference in Cmax is considered clinically irrelevant based on a sound clinical justification can be assessed with a widened acceptance range”. I have some doubts whether
that can be provided for an anticancer drug.