Objective To investigate the optimal timing for surgery in early-onset exotropia （XT） and the long-term postoperative effect. Methods One hundred eleven cases diagnosed with XT before 3 years old who were followed up for at least 2 years after surgery were retrospectively analyzed. Logistic regression was used to study the possible factors of the postoperative orthotropia rate and postoperative synoptophore/Titmus stereopsis， such as the age of onset， age at surgery， the type of strabismus， preoperative synoptophore/Titmus stereopsis and postoperative eye position. Results Based on a chi-square test and Spearman correlation analysis， classification of intermittent and constant XT had a statistically significant correlation with the rate of long-term orthotropia （χ²=4.125， P<0.05）. Classification of intermittent and constant XT （χ²=3.951， P<0.05） and the postoperative orthotropia （χ²=4.269， P<0.05） had a statistically significant correlation to postoperative synoptophore stereopsis. Classification of intermittent and constant XT had a statistically significant correlation to postoperative Titmus stereopsis （χ²=6.988， P<0.01）. Logistic regression analysis showed that the classification of intermittent and constant XT was the main factor affecting the long-term orthotropia rate （OR=0.445， P<0.05）. Preoperative synoptophore stereopsis was the main factor affecting postoperative synoptophore stereopsis （OR=10.500， P<0.01）. Preoperative Titmus stereopsis and the classification of intermittent and constant XT were the factors affecting postoperative Titmus stereopsis （OR=11.480， 0.175， P<0.05）. Conclusion It is suggested that patients with early-onset XT undergo surgery before the loss of binocular vision function. The loss of stereopsis function and the loss of the ability to control deviation were indications for exotropia surgery.