The hypothesis that reactive oxygen species (ROS) can be not just associated with but causally implicated in disease was first made in 1956, but so far, the oxidative stress theory of disease has not led to major therapeutic breakthrough, and the use of antioxidant is now confined to the field of complementary medicine. This chapter reviews the lack of high-level clinical evidence for the effectiveness of antioxidants in preventing disease and the epistemological problems of the oxidative stress theory of disease. We conclude on possible ways forward to test this hypothesis with approaches that take into account personalized medicine. The previous oxidative stress model has helped neither to diagnose nor to treat possibly ROS-related or ROS-dependent diseases. The redox balance concept that low ROS levels are beneficial or tolerable and high levels are disease triggers and best reduced is apparently wrong. Physiological ROS signalling may become dysfunctional or a disease trigger by at least five mechanisms: a physiological source may appear at an unphysiological site, a physiological source may be underactivated (less common) or overactivated (more common), a new source may appear, a physiological source may be overactivated or underactivated, and a toxifying enzyme may convert an ROS signal molecule into a more reactive molecule. The latter three mechanisms may reach a physiological or nonphysiological target. All of these dysregulations may be the direct and essential cause of a disease (rarely the case) or just a secondary epiphenomenon, which will disappear once the non-ROS-related cause of the disease is cured (much more common). Importantly, these mechanisms are the same for almost every signalling system. Causal target validation (sources, toxifiers and targets) is essential in order to identify effective drugs and therapies for ROSopathies.