Stochastic processes are very important in statistics and probability, where finding upper and lower bounds of mean-square stochastic integral has led to a basic problem. In this paper we show that for mean-square differentiable stochastic process, the convexity condition in previous well-known results can be replaced by weaker conditions.