We were surprised to see Mocroft and colleagues question our findings of improved survival among treatment initiators with CD4⁺ cell counts above 0.200x10⁹ cells/L rather than our suggestion of possible treatment benefit above a CD4⁺ cell count of 0.350x10⁹ cells/L. The former is a well-accepted minimum treatment threshold, and the latter is a finding of interest and potential import.

The calendar time for our observations (1994–2002) included periods before wide availability of HAART. However, numbers did not permit analyses stratifying by calendar time. We presented data on HAART recipients (by definition, those treated after 1995) in our Tables 1 and 2; for both antiretroviral therapy and HAART, groups starting therapy at higher CD4⁺ cell counts had fewer deaths.

Mocroft and colleagues note that treatment delayers in the group with CD4⁺ cell counts of 0.201 to 0.350x10⁹ cells/L ultimately initiated antiretroviral therapy at a median CD4⁺ cell count of approximately 0.130x10⁹ cells/L, "considerably lower than has ever been recommended," and ask whether our results would have been different had we censored those who did not initiate antiretroviral therapy once their CD4⁺ cell counts dipped below 0.200x10⁹ cells/L. Since, by definition, those who "delayed" therapy in any stratum did start therapy while in a lower CD4⁺ cell stratum, and there was only one lower stratum (<0.200x10⁹ cells/L), censoring when CD4⁺ cell counts dropped below 0.200x10⁹ cells/L would have eliminated all delayers and makes no sense analytically. Those who never received therapy are not included in our Table 2 analyses but, contrary to what Mocroft and colleagues state, were analyzed separately. In addition, to quote from our paper, "despite shorter follow-up, we observed higher mortality rates for all CD4⁺ subgroups when comparing these patients to those who either initiated or delayed ART."

Mocroft and colleagues also question the relevance of our comparative data on rates of achievement of undetectable viral loads because "fewer than 40% of patients had a viral load below the level of detection," lower than one would expect in cohorts of HAART recipients. This may represent a misreading. In Table 1, among those with CD4⁺ cell counts of 0.201 to 0.350x10⁹ cells/L, 63.8% of those initiating antiretroviral therapy versus 45.8% of treatment delayers achieved undetectable viral loads. Second, these are data from all recipients of antiretroviral therapy, not just HAART recipients. Third, and most important, earlier initiation of antiretroviral therapy was associated with greater likelihood of an undetectable HIV viral load, both in patients with CD4⁺ cell counts of 0.201 to 0.350x10⁹ cells/L and in those with CD4⁺ cell counts of 0.351 to 0.500x10⁹ cells/L. This finding strongly corroborated the mortality benefit associated with earlier initiation of antiretroviral therapy. We determined receipt of HAART on the basis of retrospective data, so the question about how this variable was defined if it was unknown at baseline seems to reflect a misunderstanding of our study design.

Last, as covered extensively in our Discussion section, we agree that the HIV clinician must carefully weigh any clinical benefit of earlier initiation of antiretroviral therapy against the possible risks of long-term toxicity, poor adherence, and emergence of viral resistance.