Biological networks exhibit intriguing topological properties such as small-worldness. In this paper, we investigate whether the topology of a metabolic network is related to its robustness. We do so by perturbing a metabolic system in silico, one reaction at a time and studying the correlations between growth, as predicted by flux balance analysis, and a number of topological metrics, as computed from three network representations of the metabolic system. We find that a small number of metrics correlate with growth and that only one of the network representations stands out in terms of correlated metrics. The most correlated metrics point to the importance of hub nodes in this network: so-called "currency metabolites". Since they are responsible for interconnecting distant functional modules in the network, they are important points in the networks for predicting if reaction removal affects growth. Source code and data are available upon request.