These intervals of 6–10 and 10–14 months allowed for laboratory tests not being performed at regular intervals in routine practice, but would approximate to

an assessment for a discordant response as soon after 6 months as possible (a mid-point of 8 months) and again at around 12 months. Individuals with a viral load <1000 copies/mL at the time of starting HAART were excluded BAY 57-1293 research buy as they may have been misclassified as treatment-naïve. Also, to be included the viral load must have fallen to undetectable levels (<50 copies/mL on one or more occasions) at or before 6 months after starting HAART, for those assessed for CD4 response at either 6–10 or 10–14 months, or both. Rebound of viral load to above 50 copies/mL at any time prior to the point of categorizing the patient as discordant

or concordant was an exclusion criterion. As this was an observational study, only one viral load measurement was required to determine eligibility or exclusion as there would have been no clinical indication for repeat testing. This analysis focused only on individuals who were known to have achieved a satisfactory virological response to HAART and maintained this until at least 6–10 and 10–14 months, respectively. For the purposes of this analysis, HAART was defined as any STI571 concentration combination of three or more antiretroviral drugs (excluding low-dose ritonavir), including triple nucleoside combinations (or two nucleosides and the nucleotide tenofovir). The baseline CD4 cell count was taken as the count closest to the start of treatment. The CD4 cell counts taken closest to 8 and 12 months were used to define the increase in CD4 from baseline. In most

cases only a single CD4 cell count was available with which to categorize the patient. Baseline viral load was log-transformed for analysis as the distribution was heavily skewed. CD4 cell count measurements were more symmetrically distributed and were not transformed. The associations between discordancy and demographic characteristics, baseline viral load and CD4 cell count, and the type of HAART regimen were examined using the Mann–Whitney and χ2 tests, as appropriate. Multiple logistic regression Florfenicol was also used to assess associations with a discordant response. Odds ratios were calculated to investigate the effect of switching regimen on the status of a patient at 12 months, compared with their status at 8 months. Switching regimen was defined as any change in therapy except for the exchange of one NRTI for another NRTI (either nucleoside or nucleotide), which was ignored. Incidence rate ratios (IRR) were calculated separately for the effect of being a discordant responder on the time to the next AIDS event or death at any time up to the last observation recorded in the database, calculated from the time of the follow-up CD4 cell count in each of the follow-up windows. Multiple Poisson regression was also used.