Résumé : Background: Although breastfeeding is common in Democratic Republic of the Congo, the proportion of women who exclusively breastfeed their babies up to 6 months remains low. This study aimed at identifying predictors of discontinuing exclusive breastfeeding before six months among mothers in Kinshasa. Methods: A prospective study was carried out from October 2012 to July 2013 among 422 mother-child pairs recruited shortly after discharge from twelve maternities in Kinshasa and followed up to six months. Interviews were conducted at each woman's house during the first week after birth, and at one, two, three, four, five and six months. Collected data included history of child's feeding and mother's socio-demographic and psychosocial characteristics. The Cox Proportional Model was used to identify predictors of discontinuing exclusive breastfeeding before six months. Results: The median duration of exclusive breastfeeding was 10.9 weeks (Inter Quartile Range 4.3 to 14.9). At six months, 2.8 % of infants were exclusively breastfed. The factors independently associated with the discontinuation of exclusive breastfeeding before six months were: not confident in the ability to breastfeed [Adjusted hazard ratio (AHR) = 3.90; 95 % CI 1.66, 9.16)], no plan on the duration of EBF (AHR = 2.86; 95 % CI 1.91, 4.28), breastfeeding problems during the first week (AHR = 1.54; 95 % CI 1.13, 2.11), low level of breastfeeding knowledge (AHR = 1.52; 95 % CI 1.08, 2.15), and experienced less than five Baby-friendly practices during the maternity stay (AHR = 1.47; 95 % CI 1.05, 2.06). Conclusions: Confidence in the ability to breastfeed and intention to exclusively breastfeed were the most important predictors of discontinuing exclusive breastfeeding before six months. To have a greater impact on the duration of exclusive breastfeeding, interventions should focus on these factors.