We present Runge-Kutta methods of high accuracy for stochastic differential equations with constant diffusion coefficients. We analyze L2 convergence of these methods and present convergence proofs. For scalar equations a second-order method is derived, and for systems a method of order one-and-one-half is derived. We further consider a variance reduction technique based on Hermite expansions for evaluating expectations of functions of sample solutions. Numerical examples in two dimensions are presented.