BACKGROUND: We examined clinical outcomes, patient characteristics and trends over time of non-medically supervised treatment interruptions (TIs) from a free-of-charge antiretroviral therapy (ART) programme in British Columbia (BC), Canada. METHODS: Data from ART-naïve individuals > or =18 years old who initiated triple combination highly active antiretroviral therapy (HAART) between January 2000 and June 2006 were analysed. Participants having > or =3 month gap in HAART coverage were defined as having a TI. Cox proportional hazards modelling was used to examine factors associated with TIs and to examine factors associated with resumption of treatment. RESULTS: A total of 1707 participants were study eligible and 643 (37.7%) experienced TIs. TIs within 1 year of ART initiation decreased from 29% of individuals in 2000 to 19% in 2006 (P<0.001). TIs were independently associated with a history of injection drug use (IDU) (P=0.02), higher baseline CD4 cell counts (P<0.001), hepatitis C co-infection (P<0.001) and the use of nelfinavir (NFV) (P=0.04) or zidovudine (ZDV)/lamivudine (3TC) (P=0.009) in the primary HAART regimen. Male gender (P<0.001), older age (P<0.001), AIDS at baseline (P=0.008) and having a physician who had prescribed HAART to fewer patients (P=0.03) were protective against TIs. Four hundred and eighty-eight (71.9%) participants eventually restarted ART with male patients and those who developed an AIDS-defining illness prior to their TI more likely to restart therapy. Higher CD4 cell counts at the time of TI and unknown hepatitis C status were associated with a reduced likelihood of restarting ART. CONCLUSION: Treatment interruptions were associated with younger, less ill, female and IDU participants. Most participants with interruptions eventually restarted therapy. Interruptions occurred less frequently in recent years.