Surrogate endpoint is an endpoint that is intended to relate to a clinically important outcome but does not in itself measure a clinical benefit. Surrogate endpoint was closed up in the era of accelerated approval initiated by the Food and Drug Administration in 1992. It is reasonably likely to predict the clinical benefit. Bevacizumab is an example of accelerated approval in the U.S. The E15 guideline of pharmacogennomics（PGx）in 2008 classified a surrogate endpoint into three categories：known valid, probable valid, and exploratory. Thus, the use of surrogate endpoint has been rapidly increased in many disease categories. A formal definition of valid surrogate endpoint was first shown by Ross Prentice in 1989. Two conditions were presented, namely 1）it must be correlated with the clinical endpoint；and 2）it must fully capture the net effect of the intervention on the clinical efficacy. Surrogate endpoint is usually understood to satisfy the first condition alone. The second condition cannot be verified in many situations. Furthermore, Journal of American Medical Association（JAMA）showed three principles to define a surrogate from a viewpoint of consistency between surrogate and true endpoints. In any events, surrogate endpoints are useful and efficient in new drug development. However, we must be careful about the use of surrogate endpoint since there have been many examples that led to an inconsistent finding regarding a clinical endpoint although there was a clear efficacy in a surrogate endpoint. Despite a recent attitude toward a new drug development using a surrogate endpoint, we must confirm its effectiveness on the clinically meaningful endpoint after approval.