The bit error rate (BER) performance of interleave division multiple access (IDMA) based systems can be predicted by a semi-analytical method referred to as signal-to-noise ratio (SNR) evolution. SNR evolution tracks the average symbol SNR at each iteration and provides a faster solution than brute-force simulation. In this paper a revised SNR updating formula is proposed for orthogonal frequency division multiplexing interleave division multiple access (OFDM-IDMA) systems in Rayleigh fading channels. By alternating the order of expectations and division of random variables when updating the average symbol SNR, a more accurate approximation of the expected SNR is obtained compared with the existing formula. Hence improved BER prediction performance can be achieved, which is verified by simulations.