A total of 195 children were randomised to zidovudine (immediate) or matching placebo (deferred) in a multicentre double blind trial in vertically HIV infected children with early disease (the PENTA 1 trial). Median follow up in the blinded phase was 1.9 years. Thereafter, individual children were unblinded following the results of adult trials showing a benefit of combination antiretroviral therapy (ART) over monotherapy, but follow up continued and is reported here until December 1998 (total follow up 4.6 years). Median time to starting ART in the deferred group was 2.7 years; 19% of deferred children had not started ART by 1999. Throughout follow up, the percentage of time spent on no ART, monotherapy, dual, and triple ART was 21%, 44%, 29%, and 6% respectively for immediate and 62%, 12%, 18%, and 8% for deferred groups. During the blinded phase eight (7.8%) immediate and 12 (13.3%) deferred children developed AIDS or died (log rank p = 0.24); overall 21 immediate and 20 deferred children progressed. In an analysis including all children regardless of original allocation, the risk of progression to AIDS or death, adjusting for age and time since trial entry was significantly lower during 1997-98 (2.4 per 100 child years) than during 1992-96 (6.6 per 100 child years), most likely a result of increased use of combination ART.