We question the validity of the subgroup analysis in the recent report by Ioannidis and colleagues [1] and disagree with the authors' assertion that there is a trend toward delayed progression of human immunodeficiency virus (HIV) disease in long-term trials of early compared with delayed zidovudine treatment. The inclusion of the Azidothymidine Collaborative Working Group (AZTCG) follow-up study [2] in the subgroup analysis contradicts the eligibility criteria initially established by the authors. A valid comparison of the long-term effectiveness of zidovudine from this trial is impossible because both the former placebo group and the treatment group received zidovudine.

We have done a subgroup analysis according to CD4 cell count at study entry in nine randomized trials of early compared with delayed zidovudine therapy. We excluded the AZTCG follow-up study and integrated additional information as collected from the authors of the original trials.

The risk ratio for progression to the acquired immunodeficiency syndrome or death in long-term trials (> 21 months' duration), regardless of the patients' initial CD4 status, was 0.96 (95% CI, 0.81 to 1.13). For short-term trials (< 14 months' duration), risk for progression was 0.43 (CI, 0.32 to 0.59) (Table 1). The risk ratio for progression to AIDS or death in short-term trials for patients with CD4 counts of 200 to 499 cells/mm³ was 0.42 (CI, 0.29 to 0.60). For patients with CD4 counts less than 200 cells/mm³, the risk ratio was 0.57 (CI, 0.34 to 0.95). In long-term trials, however, the corresponding risk ratio was 0.91 (CI, 0.61 to 1.36) for patients with more than 500 CD4 cells/mm³, 0.94 (CI, 0.78 to 1.13) for patients with 200 to 499 CD4 cells/mm³, and 1.27 (CI, 0.96 to 1.66) for patients with fewer than 200 CD4 cells/mm³.

Ioannidis and colleagues' meta-analysis more adequately shows the transient effectiveness of zidovudine in various stages of HIV infection but shows no beneficial effect in long-term trials. In addition, we are concerned that long-term treatment with zidovudine in patients with advanced HIV infection (< 200 CD4 cells/mm sup 3) may even be harmful.