Background Plasma viral load (pVL) is a key indicator of therapeutic response in HIV-infected patients receiving combination antiretroviral therapy (cART), but is often unavailable in routine clinical care in resource-limited settings. Previous model-based simulation studies have suggested that the benefits of routine pVL monitoring among patients on first-line regimens in resource-limited settings are modest, but this needs corroboration in well-defined study populations. Methods We investigated virological suppression levels and identified predictors of detectable viraemia among 870 randomly selected patients who started cART between May 2009 and April 2012 in 10 health-care facilities in Addis Ababa, Ethiopia. A total of 656 (75.4%) patients, who were alive, were retained in HIV care and receiving cART for at least 6 months provided a blood sample for pVL measurement. Predictors of detectable viraemia were identified in a multivariate logistic regression model. Results In on-treatment analysis, 94.5% (95% CI 92.5, 96.1) of the patients achieved virological suppression below 400 copies/ml after a median (IQR) of 26 (17–35) months on cART. When patients who were lost to follow-up, dead or stopped were assumed to have had detectable viraemia, the proportion of patients with virological suppression ≤400 copies/ml decreased to 74.6% (95% CI 71.5%, 77.4%). Younger age, lower educational status, ≤95% medication adherence, lower CD4⁺ T-cell count at cART initiation and/ or the diagnosis of immunological failure thereafter significantly predicted detectable viraemia. Conclusions Virological suppression levels can be high in an established ART programme in a resource-limited setting, even without the availability of routine pVL monitoring. Efforts to improve treatment outcomes should focus on younger and illiterate patients, earlier detection of HIV-positive status and cART initiation before patients are severely immunocompromised, and improving retention in care.