LEVEL-B AND LEVEL-C IN-VIVO IN-VITRO CORRELATIONS - STATISTICAL CONSIDERATIONS

An approach to in vivo/in vitro correlation is proposed based on the s uggestion that formulations are acceptable, or equivalent to a referen ce product, if the confidence interval for the in vivo variable lies w ithin a therapeutically acceptable range. An acceptable in vivo/in vit ro correlation is defined as a correlation for a range of formulations for which the confidence interval of the predicted in vivo variable l ies within the therapeutically acceptable range. Causes of failure of correlations are discussed, with emphasis on the need to understand th e mechanism of release from the formulation and the need to control cr itical variables in experimental studies. (C) 1997 Elsevier Science B. V.