This work aims at proposing a higher-order moments analysis, in order to obtain an effective characterisation (in a non-Gaussian sense) of the global DS/CDMA noise in the case of few-user systems and short spreading codes. The use of the Edgeworth series expansion, whose terms are directly linked with the statistical cumulants allows to reach a good approximation of the unknown probability density function of the global DS/CDMA noise. The results concerning the DS/CDMA BER obtained by using the Edgeworth expansion are then compared with the ones provided by the Gram-Charlier series, and by other higher-order-moment-based approximations already presented in literature