Good question. Given the apparent issues with pre-defined point-in-time reviews, the authors suggest continuous or random-time reviews instead.