*1 started stavudine and substituted zidovudine at 12 weeks, 2 started abacavir and did not change (both prescribing errors) **2 started zidovudine and did not change (1 prescribing error and 1 child changed regimen to match twin sibling) aincludes 1 not seen after randomisation Not randomised(n=154) Protocol exclusion criteria(n=151) WHO criteria for ART not met(n=59) on ART, VL>50 copies/ml(n=28) laboratory abnormality (n=26) unwilling to attend (n=20) acute TB treatment(n=5) HIV negative (n=4) <1 year, peri-natal ART(n=4) acute opportunistic infection (n=3) contra-indicated medication(n=2) Other reason(n=3) did not return(n=3) Screened (n=634) Randomised (n=480) Figure 1. Trial profile Stavudine (n=156) No exclusions (n=156) Allocation Analysis Follow-up Abacavir (n=165) 1 exclusion (n=164) subsequently found to be HIV uninfected Zidovudine (n=159) 1 exclusion(n=158) <3 years on induction TB treatment Given trial drugs(n=156) 1 withdrawn and not seen after enrolment Not known to have died, last seen before study end (n=11) 8 lost to follow-up a 3 withdrew consent Died before trial end (n=7) Given trial drugs(n=164) 2 not as randomised ** 1 lost to follow-up and not seen after enrolment Not known to have died, last seen before study end (n=6) 4 lost to follow-up a 2 withdrew consent Died before trial end (n=9) Given trial drugs(n=158) 3 not as randomised * Not known to have died, last seen before study end (n=8) 5 lost to follow-up 3 withdrew consent Died before trial end (n=3)

2a. 2b. Figure 2, Adverse Events 2a, primary endpoint (clinical grade 2+, lab grade 3 confirmed, lab grade 4 all) 2b, first grade 3/4 event Months from randomisation Proportion without endpoint StavudineZidovudineAbacavir ART experienced ART naïve Number at risk (events) 13060(66)44(16)38(6)30(3)10(0)2(1)abacavir 11250(60)37(11)32(5)22(3)10(3)2(1)zidovudine 12347(75)34(11)30(4)23(3)12(1)1(0)stavudine 3427(7)23(4)22(1)20(1)12(0)3 abacavir 4637(9)30(7)28(2)27(1)13(1)5(0)zidovudine 3329(4)25(3)23(2)21(1)9(0)2 stavudine ART naive ART experienced p=0.63 heterogeneity p=0.41 p=0.48 heterogeneity p=0.59

3a. 3b. Figure 3, Viral Suppression 3a, <400 copies/ml 3b, <100 copies/ml abacavir zidovudine stavudine 3432abacavir 45 zidovudine 32 stavudine ART naive ART experienced N ART naive ART experienced ART naive P=0.51P=1.00 P=0.58 P=0.46 P=0.82P=0.33 P=0.61P=0.79 S: 67% Z vs S: +6.3% (-5.9,18.5) A vs S: +4.0% (-7.9,15.9) A vs Z: -2.3% (-14.3,9.6) S: 73% Z vs S: +1.9% (-10.1,13.8) A vs S: +3.9% (-7.6,15.5) A vs Z: +2.1% (-9.5,13.6) S: 85% Z vs S: -5.0% (-15.1,5.1) A vs S: -4.0% (-13.6,5.6) A vs Z: +1.0% (-9.5,11.5) S: 75% Z vs S: +1.0% (-10.6,12.6) A vs S: +6.5% (-4.4,17.4) A vs Z: +5.5% (-5.4,16.4) S: 97% Z vs S: -8.0% (-19.0,3.0) A vs S: +0.2% (-8.1,8.5) A vs Z: -+8.2% (-2.6,19.0) S: 94% Z vs S: +4.0% (-5.4,13.5) A vs S: +3.1% (-7.2,13.5) A vs Z: -0.9% (-8.3,6.5) S: 97% Z vs S: -1.3% (-9.8,7.2) A vs S: -+0.2% (-8.1,8.5) A vs Z: +1.5% (-6.8,9.8) S: 97% Z vs S: +3.1% (-2.9,9.2) A vs S: +0% (-8.5,8.5) A vs Z: -3.1% (-9.2,2.9)

4b. Figure 4, Resistance 4a, Major NRTI mutations in those >500 copies/ml at week 96 4b, Predicted drug susceptibility in those >500 copies/ml at week 96 4a Percentage with mutation (95% CI) 41L65R67N70E/r74V115F151M184V/I210W215F/Y219E/Q StavudineZidovudineAbacavir Percentage with int/high resistance according to Stanford (95% CI) 3TCFTCABCDDIZDVD4TTDFEFVNVPETRRPV StavudineZidovudineAbacavir