Estimating calcium (Ca) and phosphorus (P) requirements of gilts using the factorial method includes body nutrient accretion for the respective net growth requirement and maintenance requirements. As body composition data of gilts is often lacking, their net growth requirement has been based on extrapolated data from growing-finishing pigs. To improve prediction of P and Ca requirements of gilts, the body composition of 24 gilts was assessed by dual-X-ray absorptiometry (DXA) at mating, 40 and 80 days post-mating and 2 days after farrowing. Gilts were fed restrictively a gestation diet (per kg: 12.1 MJ digestible energy; 2.6 g digestible P; 7.8 g Ca). Ten body composition models based on live weight, empty body weight, back fat thickness, and body protein were used to evaluate their ability to predict P and Ca body composition generated in gilts. The accuracy of prediction was assessed by mean squared prediction error (MSPE) between observed values and those predicted by the models. Most models had low MSPE (6-16%) and underestimated P and Ca content. The prediction equations for P and Ca based on empty body weight and back fat thickness showed the best prediction accuracy, with 6% and 9% error respectively, mainly (>70%) due to the error of disturbances. These results show the potential to improve predictions of Ca and P requirements in gilts. However, it is important to note that our gilts were leaner than those use to fit the models and some of the equations included back fat thickness. Thus, the conclusions of this study must be adjusted according to the body composition of the sows.