Retention along the elimination of Mother to Child Transmission (eMTCT) cascade in Uganda remains poor as only 62.7%-69.5% are followed up to 18months. The objective of this study was to determine the rates of retention of mother-baby pairs at two levels of the eMTCT cascade (12 and 18 months) and associated factors. This was a longitudinal analysis of 368 mother-baby pairs who were enrolled into the eMTCT program in Kaberamaido district from January 2013 to December 2018. Data was extracted from early infant diagnosis (EID) and mothers' ART registers, entered into Microsoft Excel and then exported to Stata statistical software package version 14.0 for management and analysis. Descriptive statistics such as mean and frequencies were computed at univariate level. At the bivariate level, Cox proportional hazard regression was performed to assess the level of association between the primary outcome and each independent variable, while Cox proportional hazard regression model was built at multivariate level to determine the factors independently associated with retention of mother-baby pairs in the eMTCT program. Of the 368 mothers enrolled into the study, their average age was 29.7years (SD = 6.6). Nearly two-thirds of the mothers were married/cohabiting, (n = 232, 63.0%). The 368 mother baby pairs were observed for a total time of 6340 person months, with majority, 349 (94.8%, 95%CI = 92.0-96.7) still active in eMTCT care, while 19(5.2%, 95%CI = 3.3-8.0) were lost to follow up at 12months. At 18 months, 323 (87.8%, 95%CI = 84.0-90.8) were active in eMTCT program while 45(12.2%, 95 CI = 9.2-16.0) were lost to follow up. At bivariate level, marital status, health facility level of enrolment, mothers' ART treatment supporter, and mothers' ART enrolment time were significantly associated with survival/lost to follow up (LTFU) of mother-baby pairs along the eMTCT cascade. At multivariable level, the mothers' time of ART initiation was significantly associated with survival/lost to follow up (LTFU) of mother-baby pairs at along the eMTCT cascade, with mothers-baby pairs who were initiated during the antenatal/post-natal periods having higher hazards of LTFU compared to those who initiated ART before Antenatal period (before pregnancy), aHR = 4.37(95%CI, 1.62-11.76, P = 0.003). Mother-baby pairs who were enrolled into the eMTCT program after the implementation of HIV test and treat policy (year 2017 and 2018) had higher hazards of LTFU as compared to those enrolled before the implementation of test and treat policy in Uganda (year 2013-2016), aHR = 2.22(95% CI, 1.15-4.30, P = 0.017). All the other factors had no significant association with lost to follow up and cascade completion at multivariate level. There was high level of retention of mother-baby pairs in the eMTCT program in Kaberamaido at 12 months, but it was suboptimal at 18months. ART initiation during the antenatal and/or post-natal period was significantly associated with suboptimal retention of mother-baby pairs along the eMTCT cascade.