Targeting cancer via ROS-based mechanism has been proposed as a radical therapeutic approach. Cancer cells exhibit higher endogenous oxidative stress than normal cells and pharmacological ROS insults via either enhancing ROS production or inhibiting ROS-scavenging activity can selectively kill cancer cells. In this study, we randomly chose 4 cancer cell lines and primary colon or rectal cancer cells from 4 patients to test the hypothesis and obtained following paradoxical results: while piperlongumin (PL) and β-phenylethyl isothiocyanate (PEITC), 2 well-defined ROS-based anticancer agents, induced an increase of cellular ROS and killed effectively the tested cells, lactic acidosis (LA), a common tumor environmental factor that plays multifaceted roles in promoting cancer progression, induced a much higher ROS level in the tested cancer cells than PL and PEITC, but spared them; L-buthionine sulfoximine (L-BSO, 20 μM) depleted cellular GSH more effectively and increased higher ROS level than PL or PEITC but permitted progressive growth of the tested cancer cells. No evident dose-response relationship between cellular ROS level and cytotoxicity was observed. If ROS is the effecter, it should obey the fundamental therapeutic principle - the dose-response relationship. This is a major concern.