Various hormonal parameters and the best logistic regression model to predict disease probability were evaluated in women with polycystic ovary syndrome (PCOS). Concentrations of LH, FSH, LH/FSH ratio, testosterone, free testosterone, SHBG and insulin in serum were recorded in 32 women with PCOS and in 25 controls. A model including LH/FSH ratio, insulin and testosterone measurements yielded the best goodness of fit for classification of women with and without PCOS in the logistic regression analysis. Only LH/FSH ratio and insulin were retained as significant variables. The diagnostic characteristics of LH/FSH ratio and insulin for PCOS when compared by receiver-operator characteristic analysis were found to be equally effective. By combining these two variables a higher area under curve was obtained. LH/FSH ratio, insulin or the combination of these two can predict the disease probability in women with PCOS.