BACKGROUND AND OBJECTIVE: Although not designed for research purposes, medical charts can be a unique source for obtaining information on long-term adverse drug reactions. This study aimed to assess the availability of key information on paper-based patient medical records needed to detect long-term adverse reactions to antiretroviral therapy (ART). METHODS: This is an ongoing historical cohort study carried out in three public HIV/AIDS referral centers in Belo Horizonte, Brazil. Medical charts of treatment-naïve HIV-infected adult patients initiating ART between 2001 and 2005 were reviewed for a follow-up period of up to 5 years after the first ART prescription. Descriptive analysis was performed by estimating the absolute and relative frequencies of selected variables. The Naranjo algorithm was employed to assess the availability of data on long-term adverse outcomes in medical charts. RESULTS: A total of 233 medical charts were eligible for study and 26.1% contained at least one long-term adverse reaction, corresponding to 45 cases of dyslipidemia (19.3%), 16 (6.9%) of lipodystrophy and 5 of type 2 diabetes mellitus (2.1%). Temporal relationship and ART switch could be better documented from medical charts. Information on reasons for ART switching and alternative causes for adverse reactions was very lacking. CONCLUSIONS: Specific tools should be developed and included in medical routines to improve adverse reaction reporting by physicians and other health professionals. This could be implemented simultaneously with the transition from paper to electronic medical charts in Brazil, facilitating the identification of long-term adverse reactions to antiretroviral drugs in epidemiological studies and in clinical practice.