Blinded sample size re-estimation in 2 x 2 crossover trial

Blinded sample size re-estimation is often considered to have realistic estimate of the required sample size, during the mid-course of a trial, without compromising the trial's integrity, to save a trial from being under or over-powered. We consider blinded sample size re-estimation, in the context of $ 2 \times 2 $ 2x2 crossover trial, following permuted block randomization. The underlying likelihood function suffers from identifiability problems. Blinded re-estimation procedures are prescribed under this setting and it is shown that the associated risk of unblinding may not always be negligible.