Predicting fat androstenone concentration using a rapid and cheap method applied to live pigs is needed for efficient genetic selection against boar taint. Piétrain x (Large White x Landrace) boars were slaughtered either at 119 ± 4 kg live weight and 168 ± 1 d of age (S1, n = 48) or at 116 ± 4 kg and 174 ± 1 d of age (S2, n = 42). Blood and saliva samples were collected on D 0 (S1: 29 d, S2: 35 d before slaughter) and D 27 (S1: 2 d, S2: 8 d before slaughter) to measure plasma concentrations of estradiol and testosterone and salivary concentration of estrone by EIA. A backfat sample was collected on live pigs by biopsy on D 27 and on carcasses at slaughter to measure concentration of androstenone by HPLC. Testes and Cowper glands were weighed at slaughter after tissue trimming. Data were analyzed using the R MASS package. When necessary, variables were log or square root transformed for normalization. Predictive models of fat androstenone at slaughter were established with the modlin function. Plasma estradiol and fat androstenone on D 27, taken separately, were good predictors of fat androstenone at slaughter (mean R2 between measured and predicted values was 0.48 and 0.43, respectively for each predictor) in S1 pigs. The quality of the prediction decreased between S1 and S2 pigs despite inclusion of salivary estrone in the equation with estradiol. To conclude, plasma estradiol can be considered as a good predictor of fat androstenone at slaughter if the delay between blood collection and slaughter is short.