A simple model for cell survival which is valid also at high dose has been developed. The model parameters can be traced back to measurable quantities from nanodosimetry. It is assumed that a cell is killed by an event which is defined by two or more double strand breaks in differently sized lethal interaction volumes (LIVs). Two different mechanisms can produce events, one-track events by one-particle track and two-track events by two. One- and two-track events are statistically independent. From the stochastic nature of cell killing which is described by the Poisson distribution, the cell survival probability was derived. The ratio of the number of one- and two-track events can be directly expressed in terms of nanodosimetry by the probability F2 that at least two ionizations are produced in a basic interaction volume (5-10 base pairs). From the model, relative biological effectiveness (RBE) can be derived which depends only on F2 and the size of the LIV. The expression for RBE fits experimental data with satisfying quality.