The effectiveness of postcoital contraception can only be estimated. The most commonly used method of calculation compares the expected pregnancy rate in the exposed population to the resultant pregnancies after treatment. Estimation of the fertile period and the day of ovulation are critical to calculate the expected pregnancies. The aim of this study was to improve the accuracy of calculations by evaluating the hormonal status on the day of contraceptive treatment. A total of 483 consecutive women requesting postcoital contraception was included in a prospective observational trial. A blood sample was obtained at the moment of consultation to measure serum luteinizing hormone, estradiol, and progesterone concentrations. An ethinylestradiol-levonorgestrel combination (100 micrograms/500 mg for two doses, 12 h apart) was then prescribed. The fertile period was estimated according to previous hormonal studies in the normal cycling population. Of 483 women, 64 (13.25%) women were excluded because they presented irregular menstrual cycles and 37 (7.6%) women were lost to follow-up. Two pregnancies occurred in the remaining 382 women. Following Wilcox's and Trussell's methods, 21.1 and 17.75 pregnancies should be expected, yielding an overall treatment effectiveness of 90.52% (95% confidence interval [CI] 62.58%-97.6%) and 88.73% (95% CI 55.93%-97.12%), respectively. Hormonal data were available in 356 women; 303 of whom presented with regular cycles. Hormonal information in this group restricted the number of exposed cases to 88 women. Of the women included in Trussell's method of analysis, only 51 (51.5%; p < 0.05) were at risk using hormonal data. Fifty-six percent (95% CI 34.9%-75.6%) of women with luteinizing hormone levels > 20 IU/L were not between days-1 and +1 of the cycle. Hormonal studies suggest that methods based on pregnancy risk calculated by cycle day do not faithfully reflect the real exposure.