OBJECTIVE: To determine the use of endocervical curettage in women with low-grade squamous intraepithelial lesions (LSIL) or atypical squamous cells of unknown significance (ASCUS) on Pap smear. METHODS: A cross-sectional study evaluated women at a single center colposcopy service between January 1999 and December 2003 having LSIL or ASCUS on Pap smear and compatible lesions on colposcopic examination who underwent endocervical curettage. Associations of abnormal endocervical curettage results with lesion topography, age, and satisfactory colposcopy were evaluated. A multivariate logistic regression model was used to evaluate independent associations and estimate the probabilities of abnormal endocervical curettage for different clinical scenarios. RESULTS: Two hundred thirty-seven women were included. The estimated prevalence of abnormal endocervical curettage was 5.5% (95% CI = 2.6%-8.3%). Only 2 endocervical curettage results reported cervical intraepithelial neoplasia 2 or 3 (0.8%, 95% CI = 0%-2%). Bivariate analysis showed a significant association between abnormal endocervical curettage result and central topography of the lesion (p = .01). A multivariate logistic regression analysis confirmed that association (p = .04). The model was used to calculate probabilities of abnormal results in different scenarios. CONCLUSIONS: The overall prevalence of abnormal endocervical curettage in patients with LSIL or ASCUS on Pap smear is low. The prevalence of endocervical curettage results reporting cervical intraepithelial neoplasia 2 or 3 lesions is even lower. Therefore, the use of systematic endocervical curettage performance in this setting is questionable. Our model helps to predict the probability of abnormal results and assist in the decision of whether to perform endocervical curettage.