In oncology drug development, dose escalation to determine maximum tolerated dose is a key important part in early phase trials. Many models or processes well developed to determine escalating, staying, or deescalating a selected dose level. In actual trials, some unexpected safety signals may significantly impact the predesigned escalation rules, or terminate the investigational treatment in extreme situation. This paper, based on some actual trials, in depth discusses the unexpected safety events leading to amend the escalation decisions. The paper also discussed some inappropriate treatment administration procedures which may introduce bias to the decision rules.