Neoclassical tearing modes (NTMs) will be the principal limit on performance in ITER in the standard scenario, which has beta well below the ideal kink limit. Measurements of island size from ASDEX Upgrade, DIII-D and JET in beta rampdown experiments are used to determine the marginal size for m/n = 3/2 NTM removal. This is compared with data from ASDEX Upgrade, DIII-D and JT-60U with removal of the 3/2 NTM by electron cyclotron current drive (ECCD) at near constant beta. The empirical marginal island size is consistent in both sets of removal experiments and is found to be about twice the ion banana width. A common methodology is developed for fitting the saturated m/n = 3/2 island before (or without) ECCD in all four experimental devices, ASDEX Upgrade, DIII-D, JET and JT-60U. To this is added (and model tested to experiments) the effect of un-modulated co-ECCD on the island width due to replacing the missing bootstrap current and making the tearing stability parameter Δ′ more negative. The common model is then used to evaluate the ITER ECCD system, with or without modulation, for both the m/n = 3/2 mode, which is benchmarked here, as well as the m/n = 2/1 NTM. The ITER ECCD top launch system with 20 MW of power is found to be effective in greatly reducing the size of the islands. An m/n = 2/1 mode locking model is used to show that the rotation in ITER should be sufficient for the island reduction by ECCD to avoid locking that causes loss of H-mode and disruption.