In this South African setting, HAART containing efavirenz, lamivudine, and zidovudine was well tolerated with a minimal increase in clinical and laboratory events from the pre-HAART level. Furthermore, most events occurred within the first 12 weeks of therapy, and were effectively managed by either symptomatic care or, in a few cases, change to another regimen.
Perhaps most importantly, despite a concern for potentially worsening anemia, we found the opposite to occur in our cohort. On-HAART, grade 3 and 4 anemia occurred with an incidence similar to pre-HAART. In addition, we observed a sustained overall rise in hemoglobin after week 6. Our finding of increasing hemoglobin contrasts with studies conducted in high-income countries. The Women's Interagency HIV Study demonstrated an association between zidovudine and persistent anemia; subjects who received a regimen not containing zidovudine had resolution of anemia, whereas those who received zidovudine did not [13
]. It is unclear whether sex played a role as theircohort was entirely female. Another study from a male cohort also reported a decline in hemoglobin among those subjects treated with zidovudine, however, but not those receiving stavudine [14
]. A meta-analysis of six clinical trials conducted in high-income settings also identified a decline in hemoglobin, 0.4 g/dl at 24 weeks and 0.2 g/dl at 48 weeks after the initiation of a regimen containing zidovudine but not stavudine [15
]. The difference between our observations and reports from high-income country cohorts may reflect more advanced HIV disease and higher levels of concomitant disease in our African population [16
On the other hand, zidovudine probably contributed to neutropenia. Despite the high rates of grade 3 or 4 neutropenia, however, we did not detect evidence of severe infection, consistent with a previous study [18
]. The one neutropenia-related death occurred in a subject with disseminated tuberculosis and advanced HIV disease. In that subject, the neutropenia may have been a result of those conditions rather than HAART.
The large proportion of subjects with grade 1 and 2 neutropenia suggests that the current ACTG grading scale is not optimal for a southern African population. This finding is consistent with several studies that have reported lower means and ranges for neutrophil counts among healthy individuals of African compared with European origin. For example, the neutrophil reference ranges based on studies of individuals of African descent in the United Kingdom (900–4200 cells/μl) and in Uganda (840–3370 cells/μl) overlap ACTG ranges for grade 1 and 2 neutropenia [19
]. In contrast, the reference range derived from a population of European descent is 1500–7800 cells/μl, well above grade 1 neutropenia. Therefore for a European population, the ACTG grading for neutropenia with grade 1 neutropenia starting at a neutrophil count of 1000 cells/μl may be suitable, whereas a scale with different ranges for grade 1 and 2 neutropenia may be more appropriate for other populations.
Possible peripheral neuropathy occurred at a similar rate before and after HAART initiation and within previously reported ranges [21
]. We were unable to determine whether any cases of peripheral neuropathy occurred as a result of antiretroviral toxic neuropathy. Most subjects with peripheral neuropathy during HAART, however, also had peripheral neuropathy identified before HAART initiation, making such an association unlikely.
Efavirenz was probably responsible for the large proportion of subjects with neurocerebellar symptoms. Neurocerebellar toxicity affected 22% of subjects assessed at the week 2 clinical evaluation. The high proportion of subjects with early neurocerebellar toxicity and rapid decline in the number of subjects with this symptom is consistent with previous reports from high-income countries of efavirenz-containing HAART regimens [22
We noted more neutropenia, rashes, nausea or vomiting, and neurocerebellar symptoms among subjects receiving cotrimoxazole prophylaxis while on-HAART. Cotrimoxazole is known to cause neutropenia, can cause a variety of hypersensitivity rashes, and is associated with nausea providing a potential explanation for the increase in these events among individuals receiving cotrimoxazole. It is unclear how cotrimoxazole may have been associated with neurocerebellar effects and this could have been a chance finding. Given the reported survival benefits of cotrimoxazole, we do not suggest any reduction in its use [24
We believe the value of this study is enhanced by the comprehensiveness of our laboratory follow-up and the structured history and physical examination forms. To apply our findings to other settings, however, limitations must be recognized. Notably, our population was predominantly male, and malaria, a known risk factor for anemia, was uncommon. Some subjects with low baseline hemoglobin were not started on zidovudine, and therefore a small number of subjects who may have had a higher risk of severe anemia were not included. An additional limitation is the lack of a comparator group on a different HAART regimen. We believe comparing pre-HAART to on-HAART data allows some estimation of adverse events associated with the regimen. We may, however, have underestimated pre-HAART adverse events as ascertainment was limited by a shorter assessment time, assessments were not as frequently scheduled, laboratory testing was mostly obtained on the basis of clinical indication rather than routine monitoring, and subjects may have been sicker around the time of starting HAART than during the pre-HAART period. Equally intense pre-HAART and on-HAART monitoring may have found a higher rate of clinical events pre-HAART. This would lead to the finding of an even safer on-HAART profile. We also raise the issue that some subjects with clinical events were switched to other regimens or may have left care as a result of side effects, thus the decline in period prevalence of a given adverse event may partly reflect the loss of the most susceptible subjects. We expect the impact of any such bias to be small because of the small number of subjects who switched regimen for adverse events.
Our findings regarding clinical events occurring during the first 12 months of therapy with a regimen consisting of zidovudine, lamivudine, and efavirenz may be of use both for selecting first-line HAART regimens as well as scheduling clinical and laboratory investigations after starting HAART. Further evaluations of the tolerability of zidovudine in other African settings are important, however, especially among women and in regions with a high prevalence of endemic malaria. In addition, direct comparisons conducted in African settings between regimens containing zidovudine, stavudine, tenofovir, or abacavir would add valuable information for designing future HAART regimens.