OBJECTIVES: We examine long-term retention of adults, adolescents and children on antiretroviral therapy under different HIV treatment guidelines in Malawi. DESIGN: Prospective cohort study. SETTING AND PARTICIPANTS: Adults and children starting ART between 2005 and 2015 in 21 health facilities in southern Malawi. METHODS: We used survival analysis to assess retention at clinic level, Cox regression to examine risk factors for loss to follow up, and competing risk analysis to assess long-term outcomes of people on antiretroviral therapy (ART). RESULTS: We included 132,274 individuals in our analysis, totalling 270,256 person years of follow up (PYFU; median per patient 1.3, interquartile range (IQR) 0.26-3.1), 62% were female and the median age was 32 years. Retention on ART was lower in the first year on ART compared to subsequent years for all guideline periods and age groups. Infants (0-3 years), adolescents and young adults (15-24 years) were at highest risk of LTFU. Comparing the different calendar periods of ART initiation we found that retention improved initially, but remained stable thereafter. CONCLUSION: Even though the number of patients and the burden on health care system increased substantially during the study period of rapid ART expansion, retention on ART improved in the early years of ART provision, but gains in retention were not maintained over 5 years on ART. Reducing high attrition in the first year of ART should remain a priority for ART programs, and so should addressing poor retention among adolescents, young adults and men.