Search for a command to run...
Low-level viral rebound (LLVR) was identified in 704 of 2055 patients on effective antiretroviral treatment. It was followed by a value of 50 copies/ml, a ‘blip', in 490 patients, whereas two consecutive values between 51 and 500 copies/ml, a ‘bump', were observed in 155. Hazard ratios of viral failure were 2.01 for blips and 5.80 for bumps. LLVR is frequent and generally resolves spontaneously, but is associated with an increased risk of viral failure. The occurrence of viral rebound to a moderate level (51–500 copies/ml) raises concern regarding the durability of response to antiretroviral therapy (ART) as it has been correlated with sequence evolution [1–3], with failure to reduce the number of latently infected cells [4], and with drug resistance mutations in the latent reservoir [5]. However, low-level viral rebound (LLVR) could be caused by test variability [6], polymerase chain reaction contamination, non-compliance, intercurrent infection, or vaccination [7]. When evaluating the nature of ‘blip', clinicians should be aware of the variability of the ultrasensitive amplification methods. Inter-assay and intra-assay coefficient variability of up to 35.4 and 40%, respectively, has been reported for this assay [6]. In most instances, the next viral load is again below the level of detection without a treatment change. Episodes of intermittent low-level viral load, so called ‘blips', have been reported in up to 40% of patients receiving the combination of zidovudine, lamivudine, and indinavir [8]. It is of interest to assess the significance, prevalence, and predictive value of blips. To explore these issues, we analysed data on LLVR among 2055 individuals achieving viral suppression to below 50 RNA copies/ml and followed in two large observational cohort studies, the Swiss HIV Cohort Study and the Frankfurt HIV Clinic Cohort. We included for analysis all participants who had, while being treated with potent ART, at least two consecutive values below 50 copies/ml within a 24 week interval on samples taken between 1 January 1998 and 31 December 1999, and a minimum of two additional viral determinations taken within 24 weeks after the previous viral load. The plasma HIV viral load was determined using Roche Amplicor (Roche, Basel, Switzerland), with a level of detection of 50 copies/ml or lower. Viral failure was defined by one HIV viral load value above 500 copies/ml. LLVR was defined as a viral rebound of 51–500 copies/ml. The next viral load taken within 24 weeks after a LLVR was used to classify the viral event into two categories: ‘blip', the next viral load below or equal to 50 copies/ml; and ‘bump', the next viral load between 51 and 500 copies/ml. Logistic regression was used for the analysis of predictors of the next viral load level for patients after their first LLVR. The variables included in the model were age, sex, intravenous drug use, pretreatment (use of mono- or bitherapy before potent ART), CD4 cell count at the time of viral rebound, viral load at the initiation of potent ART, a non-nucleoside reverse transcriptase or protease inhibitor-containing regimen, treatment change, and the interruption of treatment between viral rebound and next viral load. Cox's regression for time-to-event analyses were used to assess the prognostic value of any episode of blip or bump. The model used the two most recent viral load values. The time to viral failure was measured since the fourth viral load. Observations were censored at the time of the last available viral load. The variables included in Cox's models were age, sex, intravenous drug use, AIDS at baseline, CD4 cell count, non-nucleoside reverse transcriptase or protease inhibitor-containing regimen, and previous treatment. The CD4 cell count was included as a time-dependent variable. SAS (version 6.12) and STATA (version 7.0) software were used for analysis. A total of 2055 patients achieved HIV viral suppression. The median follow-up after the first viral load below 50 copies/ml was 17.7 months. HIV viral load values were available on average 11.6 times per subject during the study period. A total of 1175 subjects (57.2%) maintained viral suppression during the whole study period, 704 patients (34.3%) presented a LLVR (51–500 copies/ml) and 176 patients (8.6%) had an initial viral rebound greater than 500 copies/ml. A blip was observed in 490 out of 704 subjects (69.6%) with a first LLVR, and a bump was observed in 155 of them (22.0%). A return to viral load below or equal to 50 copies/ml was also observed in 71 of 176 patients (40.3%) with a first rebound to more than 500 copies/ml (Fig. 1a).Fig. 1.: (a) Next viral load (in copies/ml) for 704 patients presenting a low level viral rebound (51–100 copies/ml and 101–500 copies/ml), and 176 patients presenting a viral failure (500 copies/ml). ░ 0–50; ▒ 51–100; ▒ 101–500; ▓ > 500 copies/ml. (b) Hazard ratio of having a viral failure (one viral load greater than 500 copies/ml) adjusted for all listed variables. aCompared with intravenous drug user; bcompared with protease inhibitor-sparing regimen; ctime-dependent; dcompared with optimally controlled subjects.The variables associated with an episode of viral rebound evolving into a blip were the level of initial viral rebound [odds ratios (OR) 0.85, 0.75–0.98, per additional 0.5 log copies/ml], and pretreatment with mono- or bitherapy before potent ART (OR 0.60, 0.38–0.94). Treatment change and the type of treatment were not associated with the probability of blip. After a median follow up of 17.7 months, 157 out of 2055 patients (7.6%) experienced viral failure. In Cox's regression models, the hazard ratio (HR) of viral failure was 2.01 [95% confidence interval (CI) 1.51–2.91; P < 0.0001] for blip, and 5.80 (95% CI 4.26–7.90; P < 0.0001) for bump compared with viral suppression (Fig. 1b). This cohort analysis thus identified a high incidence of LLVR. For most people, LLVR will resolve without treatment change. Nevertheless, it increases the risk of subsequent viral failure by approximately two-fold compared with patients with sustained viral suppression. The magnitude and persistence of the viral rebound predicted the viral outcome. In particular, the occurrence of bumps carried a worse prognosis than blips for the durability of viral suppression. The observed increased risk of viral failure associated with an episode of blip is in contradiction with the results of a recent analysis by Havlir et al. [8], in which the identification of a blip did not translate into different rates of virological failure (HR 1.28, 95% CI 0.59–2.79). This discrepancy could be explained by the smaller number of subjects included in the study by Havlir et al. [8] (n = 343), and the limited time of follow-up to detect failures in this group of patients with a very low risk of failure [9], or by differences in the study design and patient populations studied: a randomized controlled trial versus a prospective observational cohort study. The main limit of the present study is the short follow-up of 17.7 months (as a result of the recent introduction of the ultrasensitive plasma HIV-RNA assay). The choice of a limit of detection of less than 50 viral RNA copies/ml employed in this study can also be challenged. This limit was chosen to allow the harmonization of results between cohorts and among various laboratories. The use of a lower limit of detection, i.e. less than five or less than 20 viral RNA copies/ml could have provided more insight into the various issues discussed above [10]. Future studies should aim at identifying the determinants for the diverse evolution after a viral rebound. This includes a detailed analysis of the immune determinants of the control of residual viral replication, an assessment of the role of resistance mutations by the development of more sensitive and rapid methods to assess drug resistance in samples with a low level of viraemia [5], and the determination of drug penetration in various compartments. Such studies may provide insight into the observed differences in the durability of optimal suppression in patients receiving potent ART. Acknowledgements Sponsorship: This study was financed in the framework of the Swiss HIV Cohort Study, supported by the Swiss National Science Foundation. (grant no. 3345-062041).