The Swedish Road Administration (SRA) performs cost-benefit studies for road infrastructure projects utilizing a specialized computer program known by its Swedish acronym EVA. This report describes three possible approaches to accomplish sensitivity studies of the SRA's costbenefit studies. A simple example explains how these approaches are applied in the context of the SRA's cost-benefit studies. First, the popular one-factor-at-a-time method is described. The major drawback of this approach is that interaction of factors cannot be addressed. The method of Monte Carlo simulation allows for this possibility. This method uses brute force computational power to calculate a large number of outcomes based on randomly altered values for the included factors. However, Monte Carlo simulation necessitates the assessment of the multivariate stochastic distribution of considered factors. It is generally very difficult to assess multivariate distributions and it is consequently difficult to utilize the full potential of the Monte Carlo technique. We suggest therefore that the SRA's sensitivity studies are based on few simulation runs where the factors are altered simultaneously in each run. The number of runs is kept low by appropriate selection of factor combinations in each run, utilizing statistical design theory for experiments. The so obtained "observations" are used to estimate a response surface, which is essentially a simplified regression relationship between the factors that matter in the context and the resulting ratio between a project's net present value and its cost. This method handles possible interaction between factors, requires from the investigator merely the determination of reasonable upper and lower bounds for included factors, and does not require a large number of simulation runs.