Background:  We examined plasma viral load dynamics to predict virological failure and to define viral evolution in heavily pretreated patients receiving genotypic resistance testing (GRT) -guided HAART.

Methods:  Prospective study of 30 heavily pretreated (3.9 previous failing regimens, mean) and of 8 naive patients receiving GRT-guided HAART. Study visits were scheduled at week 0, 1, 2, 3, 4, 8, 12, 16, 20, 24, 32, 40, and 48 for plasma viral load, CD4 count, GRT (if plasma viral load >50 copies/mL), and self-reported adherence. Patients were considered responders (reaching and maintaining plasma viral load <50) or non-responders (never reaching plasma viral load <50 or 2 plasma viral loads >50 after suppression). A mathematical model was used to fit the experimental data. Non-linear square fitting algorithm was used to estimate the clearance of free virions (c), activated infected CD4 cells (δ), and long-lived infected cells (m).

Results:  Of 8 naive, 8 (100%) and of 30 pre-treated patients, 22 (73%) were responders. Neither first nor second phase dynamics differed between pre-treated and naive responders. Of 30 pre-treated patients, 8 (27%) were non-responders. First phase plasma viral load dynamics, and week 0 to 2 plasma viral load reductions (–1.8 vs 1.6 log copies/mL, respectively) did not differ between non-responders and responders. By contrast, the second phase dynamics differed significantly between non-responders and responders, allowing a clear separation between the δ and the m calculated fits. Non-responders had higher plasma viral loads at weeks 4, 8, 12, 16, 20, and 24 (p <0.001) and unfavorable weeks 2 to 4 and 2 to 8, plasma viral load changes (p <.001). Positive plasma viral load changes between weeks 2 to 4 or weeks 2 to 8 identified non-responders with a 87.5% sensitivity, 100% specificity, 100% positive predictive value, 94.7% negative predictive value, and 96.0% accuracy (p <0.0001). A progressively evolving GRT pattern was seen in all non-responders. The mean number of emergent resistance-associated mutations increased, at an apparently constant rate, from 0.7 (range 0 to 2) at week 4 to 3.0 (range 2 to 6) at week 24. The estimated time to the emergence of a single new mutation was 83 (range 23 to 140) days.

Conclusions:  In heavily pretreated patients failing to respond to GRT-guided HAART, while first phase dynamics did not differ from responders, virological rebound was associated with failure to fit with the second phase dynamics, and with a rapid emergence of resistance-associated mutations. We propose that plasma viral load should be measured 2, 4, and 8 weeks after changing therapy for virological failure to achieve an early recognition of virological rebound and to prevent mutation accumulation.