Understanding the prevalence and correlates of treatment interruptions (TIs) in resource-limited settings is important for improving adherence. HIV-infected adults on highly active antiretroviral therapy (HAART) in Bangalore, India, were enrolled into a prospective cohort study assessing HAART adherence. Participants underwent a structured interview assessing adherence, including occurrence of TI > 48 hours since HAART initiation, length of TI, and self-reported reasons for TI. Serum HIV viral load (VL) and CD4 was measured at 6-month intervals. Baseline data are presented in this article. For the 552 participants mean age was 37.8, 32% were female, 70% were married, 45% earned < $2/day. Eighty-four percent were on nevirapine-based antiretroviral therapy; median duration on HAART was 18 months (range: 1-175) and median CD4 count was 318 cells/µl (IQR: 195-460) at time of study enrollment. Twenty percent (n=110) reported at least one TI; of these, 33% (n=36) reported more than one TI. Median length of most recent TI was 10 days (range: 2-1095). TI was associated with a higher probability of having VL > 400 copies/ml (43% versus 12%; p<0.001). After controlling for time on HAART, TI was more likely among those who were unmarried (OR: 1.9; CI: 1.2-3.1), those treated in a private clinic setting (OR: 2.7; CI: 1.6-4.6 compared with public, and OR: 4.1; CI: 1.9-9.0 compared with public-private setting), and those on efavirenz-based therapy (OR: 2.0; CI: 1.1-3.6). The most common self-reported reason for TI was "side effects" (n=28; 25%), followed by cost of therapy (n=24; 22%). We discuss implications for both individual and structural level interventions to reduce TIs.