OBJECTIVE To determine prevalence and predictors of antiretroviral drug resistance in newly diagnosed individuals with HIV-1 infection, using a systematic approach to avoid selection bias. METHODS Plasma samples from all persons diagnosed HIV-1 seropositive at a large London centre between April 2004 and February 2006 underwent sequencing of HIV-1 reverse transcriptase (RT) and protease genes. Subtype was assigned by phylogenetic analysis. Resistance was scored according to the IAS-USA list (2005) modified to include T215revertants and exclude isolated E44D or V118I and minor protease mutations. Recent seroconversion was identified by HIV antibody avidity testing. RESULTS The cohort of 239 included 169 (70.7%) males, 126 (52.7%) homosexuals, 118 (49.5%) persons of white ethnicity and 144 (60.0%) persons born outside the UK. Subtypes included B 134 (56.1%), C 46 (19.2%), A 17 (7.1%), other non-B 42 (17.6%). The prevalence of resistance mutations was 17/239 (7.1%; 95% confidence interval 4.5-11.1%), comprising 10/239 (4.2%) nucleoside/nucleotide RT inhibitor (NRTI); 4/239 (1.7%) non-nucleoside RT inhibitor (NNRTI) and 4/239 (1.7%) protease inhibitor (PI) associated mutations. Dual-class (NRTI + PI) resistance mutations were detected in 1/239 (0.4%) person. The prevalence of resistance mutations was 7/85 (8.2%) and 10/154 (6.5%) in persons with recent and established infection, respectively. In multivariate analysis, having been born in the UK and high CD4 count, but not gender, age, risk group, ethnicity or subtype, were independent predictors of resistance. CONCLUSIONS In an unselected UK cohort, subtypes other than B accounted for 43.9% of new HIV-1 diagnoses. The prevalence of resistance mutations was 7.1% and highest in those born in the UK.