OBJECTIVE To describe the long-term efficacy over 5 years of regimens including combinations of abacavir, lamivudine and/or zidovudine in previously untreated children in the PENTA 5 trial. METHODS PENTA 5 was a 48-week randomised controlled trial comparing three dual nucleoside reverse transcriptase inhibitor (NRTI) combinations as part of first triple antiretroviral therapy (ART). METHODS 128 ART-naïve children were randomised to zidovudine\lamivudine (n = 36), zidovudine\abacavir (45) or lamivudine\abacavir (47). Asymptomatic children (n = 55) were also randomised to nelfinavir or placebo; all other children received open-label nelfinavir. Analyses are intent-to-treat and adjusted for minor baseline imbalances and receipt of nelfinavir/placebo. RESULTS Median follow-up was 5.8 years. By 5 years, 17 (47%), 28 (64%) and 18 (39%) children had changed their randomised NRTIs in the zidovudine\lamivudine, zidovudine\abacavir and lamivudine\abacavir groups respectively, but 18%, 50% and 50% of these changes were either early single drug substitutions for toxicity or switches with viral suppression (HIV-1 RNA < 400 copies/ml; e.g. to simplify regimen delivery). At 5 years, 55%/32% zidovudine\lamivudine, 50%/25% zidovudine\abacavir and 79%/63% lamivudine\abacavir had HIV-1 RNA < 400/< 50 copies/ml respectively (p = 0.03/p = 0.003). Mean increase in height-for-age 0.42, 0.68, 1.05 (p = 0.02); weight-for-age 0.03, 0.13, 0.75 (p = 0.02). Reverse transcriptase resistance mutations emerging on therapy differed between the groups: zidovudine\lamivudine (M41L, D67N, K70R, M184V, L210W, T215Y); zidovudine\abacavir (M41L, D67N, K70R, L210W, T215F/Y, K219Q); lamivudine\abacavir (K65R, L74V, Y115F, M184V). CONCLUSIONS Five year data demonstrate that lamivudine\abacavir is more effective in terms of HIV-1 RNA suppression and growth changes, with lower rates of switching with detectable HIV-1 RNA than zidovudine\lamivudine or zidovudine\abacavir, and should be preferred as first-line NRTI backbone.