[Latin square vs twin crossover design for mouse blood glucose assay to estimate insulin potency].

According to parallel line analysis, a Latin square design was used for estimating insulin potency in mouse blood glucose assay. Four to six groups of 4 x 4 Latin squares were used to estimate 80%, 100% and 120% standard preparations and the recovery rates were 95-106%. Meanwhile, in comparison with twin crossover design, 13 batches of variant preparations including some beyond expiry dates were checked by Latin square design. The results showed that the potencies were about the same in two designs. The average fiducial limit rates were still less than 25% but 40% or so of animal numbers were saved. Therefore, Latin square design in a precise and accurate assay for the estimation of insulin.