Treatment interruption for virological failure or as sparing regimen in children with chronic HIV-1 infection

Objective: To assess both benefits and risks related to treatment interruption (TI) in children with chronic HIV-1 materno-fetal infection.

Design: A multicentre, retrospective analysis in five university hospital pediatric departments in France.

Methods: Clinical events, plasma HIV-1 RNA, CD4 cell counts, CD4 percentages (CD4%) and genotypes were recorded in 24 patients before and during TI. Patients were classified as sparing regimen or virological failure groups according to the main reason for treatment interruption. Clinical events, immuno-virological evolution and genotype reversions were monitored.

Results: After a median of 40 weeks of TI, none of the patients presented with an AIDS-defining event. For the whole cohort, median viral load variation from baseline, measured during TI was +1.26 log₁₀ copies/ml (range, -0.22, +4.3 log₁₀) with large inter-individual variability, median absolute CD4 cell loss was 32.5% (range, -82, +17%). These variations were not different in the two patient groups. The mean number of mutations conferring resistance to nucleoside reverse transcriptase inhibitors, non-nucleoside reverse transcriptase inhibitors and protease inhibitors at baseline and last evaluation did not differ significantly. Few mutation reversions to wild type were noted in our cohort.

Conclusions: Treatment interruption in children with chronic HIV-1 infection is associated with higher viral load increases than observed in adult patients. The CD4 cell loss is comparable. Although no clinical AIDS-defining event was noted close monitoring is required when TI is proposed to HIV-infected children. Very few reversion mutations were observed during treatment interruption.

Introduction

In industrialized countries, increasing use of highly active antiretroviral therapy (HAART) is associated with a reduced frequency of mortality, opportunistic infections and other HIV-associated complications in HIV-1-infected adults and children [¹]. For younger children, drugs are now increasingly available as syrup or powder formulations. New drugs and therapeutic classes are currently under investigation and pharmacokinetics and pharmacodynamics are quickly available in the pediatric population [^2-5]. As in adult patients, most naive infected children treated with protease inhibitor (PI)-containing regimens achieved optimal viral suppression and restoration of immune function a few months after treatment initiation [^5-8].

In addition to common toxic effects, serious metabolic complications have been reported in HIV-1-infected adults. Recently the DAD group published a 26% increase risk of myocardial infarction per year of exposure to combination antiretroviral therapy (ART) in a prospective multinational cohort study [⁹]. Hypercholesterolemia, hypertriglyceridemia and insulin resistance are believed to be associated with the use of protease inhibitors and severe lactic acidosis with the use of nucleoside analogue reverse transcriptase inhibitor regimens. Such complications are now currently described with increasing frequency in HIV-1-infected children [^10-15]. In children, as experience has accumulated, changes in body fat distribution and osteopenia have become apparent. Moreover, some authors suggest that children may be more vulnerable than adults because of the potential effect of treatment on growth and the children's likely greater cumulative exposure to drugs [^10,13].

Because of the difficulty of full long-term adherence and the increasing frequency of treatment-related complications more and more infected children present with virological failure [^6,8]. For some of them, therapeutic options become difficult because of the accumulation of mutations in the reverse transcriptase and/or protease genes conferring resistance to multiple drugs or therapeutic classes. In heavily pre-treated adult patients, structured treatment interruption (STI) has been investigated [^16-20]. The background for this strategy is to re-establish a predominant population of drug-susceptible virus [^21,22]. On the other hand STI has been proposed in chronic HIV-1-infected patients with controlled and stable plasma HIV-1 polymerase chain reaction (PCR)-RNA [^23-25]. In this context, the objective of sparing regimens is to study whether it can prevent or reduce HAART-associated toxicity and improve quality of life without jeopardizing the efficacy of future treatment.

Treatment interruption had not yet been fully investigated in children with HIV-1 infection. An open randomized controlled phase II study is planned in the Pediatric European Network for Treatment of AIDS (TICCH Trial - PENTA II). In order to investigate the current practice of pediatricians experienced in management of HIV infection in pediatric patients, we conducted a retrospective analysis in five university hospitals in France. Physicians were asked to report their own experience. Data on treatment interruption were collected and analyzed retrospectively.

Methods

Study design

We conducted a retrospective multicentre study to analyze treatment interruption (TI) in children with chronic HIV-1 infection (study design - Fig. 1). The main aim of this study was to evaluate retrospectively the following parameters: clinical events, immuno-virological variations and genotype reversion during TI. Responsible pediatricians with experience in management of HIV infection in pediatric patients from five university hospitals in France (Marseille, Montpellier, Nantes, Nice and Toulouse) were asked about their practice in TI. Query forms were sent in May 2003. Epidemiological, clinical, biological and therapeutic data were collected. The physicians were asked if TI was justified by virological failure (VF) or sparing regimen (SR). VF was defined as uncontrolled plasma HIV-1-RNA with levels above 5000 copies/ml during the past 6 months and at baseline, most of the time associated with decreased CD4 cell counts. SR was defined as total antiretroviral treatment wash-out in patients with stable plasma HIV-1-RNA (less than 0.5 log₁₀ variation) and CD4 cell count in the past 6 months. Clinicians were asked for the baseline and highest plasma HIV-1 PCR-RNA values, baseline and lowest values for CD4 cell count and percentage measured in each patient during the TI. Baseline values were set as close as possible before TI.

Fig. 1

At the time of treatment interruption all the children were receiving ART with at least two different drugs. For each patient, clinicians were asked about previous exposure to nucleoside reverse transcriptase inhibitor (NRTI) drugs (azidothymidine, lamivudine, didanosine, dideoxycytosine, stavudine, abacavir and tenofovir), non-nucleoside reverse transcriptase inhibitors (NNRTI) (nevirapine and efavirenz) and/or protease inhibitors (PI) (indinavir, ritonavir, nelfinavir, saquinavir, amprenavir and lopinavir).

On 31 December 2003 a second query form was sent asking for a clinical and biological update (last evaluation). Children were classified according to ongoing treatment interruption or resumed ART. For patients who had resumed ART, details on therapy and biological data (including genotype) as mentioned earlier where obtained just before and 4 weeks after resumed ART.

Genotype analysis

Genotypes were analyzed at baseline and before resuming ART. For patients with ongoing TI, a last evaluation was requested in December 2003. The number of mutations conferring resistance to NRTIs (M41L, K65R, D67N, T69D/N/S, K70R, V75M/S/A/T, L74V, Q151M, M184V/I, L210W, T215Y/F/A/C/D/EG/H/I/L/N/S/V and K219Q/E), NNRTIs (L100I, K101E, K103H/N/S/T, Y181C/I, Y188C/L and G190A/C/E/Q/S/T/V) and PIs (L10F/I/V, K20I/M/R, L24I, D30N, M46I/L, I54L/V, L63P, A71I/L/V/T, G73A/C/S/T, V77I, V82A/F/T/S, I84V, N88S/D and L90M) were recorded. For thymidine-associated mutations (TAMs; M41L, D67N, K70R, L210W, T215Y/F, K219Q/E) data were pooled. Genotypes with three or more TAMs were noted. Genotypes were compared for each patient at baseline and at the last evaluation. The number of reversions from mutated to wild-type genotype for each codon was evaluated. According to the genotype profile and the French algorithm of resistance analysis [²⁶], a percentage of susceptibility to the drug combination was calculated for each patient.

CD4 T-lymphocyte count and plasma HIV-1-RNA assay

In the five university hospitals, peripheral blood CD4 T-lymphocyte assays were conducted by flow cytometry using commercial monoclonal antibody. Plasma HIV-1 RNA values were determined using the Cobas Amplicor HIV Monitor assay (Roche Diagnostics, Bassel, Switzerland) according to the manufacturer's instructions.

Statistical analysis

Patients were grouped according to the main reason for TI as VF or SR as described. Qualitative demographic, clinical, genotypic and therapeutic factors were compared between these two groups using Fisher's exact test or the χ² test. Median values were compared using the Mann-Whitney U-test. The non-parametric Wilcoxon rank sum test for paired values was used to determine the statistical significance of difference in viral load, absolute CD4 cell count and percentage rate between different periods or groups of patients. Significance was established at P < 0.05.

Results

Patient characteristics

Twenty-four patients were enrolled in the study. Ten treatment interruptions were justified by VF and 14 for SR. Baseline characteristics for all the patients and sub-groups are given in Table 1. Most patients were symptom-free [N or A in the Centers for Disease Control (CDC) classification]. The median age at the time of TI, duration of TI and baseline median CD4 cell count did not differ between the two groups. Median baseline plasma PCR-RNA was higher in the VF group (P = 0.0045). There was a trend for a lower median percentage of CD4 cells in that group of patients (P = 0.06). At the time of TI, all the patients had been previously exposed to NRTI, 46% to NNRTI and 54% to PI. Neither the median number of drug exposures, nor the median number of class exposures was different between the two groups. At the time of TI, 29.4% of the drugs prescribed were considered to be effective according to the last genotype. The difference between the two groups was not significant (SR, 40.9% versus VF, 16.7%).

Table 1

Clinical events

Minor clinical events were observed in three patients. Two patients in the VF group suffered from upper respiratory infection (tracheitis, bronchitis). One patient suffered from influenza in the SR group. The clinical course was unremarkable. Two patients from the SR group experienced transitory failure to thrive, enteritis, anemia and hepatic cytolysis. The latter symptom disappeared with resumed treatment. None of the patient presented with an AIDS-defining event.

Immunological and virological consequences of treatment interruption

Median highest plasma HIV-1 RNA-PCR values, median lowest CD4 cell count and percentage values; median viral load and absolute CD4 variations from baseline are presented in Table 2. Median measured viral load variation was +1.26 log₁₀ copies/ml for the whole group, with large inter-individual variability (range, -0.2 to +4 log₁₀ copies/ml). For most of the time the highest viral load values were observed during the first month of TI. The higher median HIV-1 PCR-RNA value measured during TI was observed in the VF group. However, the viral load variation observed was not different between the two groups. Globally nearly one-third of CD4 cells were lost during TI. With a median of 40 weeks of TI for the whole group, median CD4 cell decrease was -11.25 per week. The nadir of CD4% was significantly lower in the VF group in comparison with the SR group.

Table 2

Genotype variation between baseline and last evaluation

Genotypes could be evaluated for 17 patients and compared at baseline and at the last evaluation. The mean baseline number of mutations conferring resistance to NRTI, NNRTI and PI were 3.26 (range, 0-7), 0.35 (range, 0-2) and 3.48 (range, 0-8), respectively, for the whole group. The difference was significant between VF and SR groups for number of NRTI mutations (mean, 4.4 versus 2.4, respectively; P = 0.01) and for number of PI mutations (mean, 4.7 versus 2.54, respectively; P = 0.05) at baseline. At the last evaluation the mean numbers of mutations were 2.9 (range, 0-6), 0.53 (range, 0-2), 3.35 (range, 1-8) for NRTI, NNRTI and PI, respectively. The differences were not significant for the latter between the two sub-groups. Figure 2 shows the number of mutations (box-and-whisker plot with median, 25th, 75th interquartile values and extremes) at baseline and last evaluation for the three classes. The variation between baseline and last evaluation was not significant for any class. Amino acid substitutions in the reverse transcriptase and protease genes were noted at baseline and last evaluation for the 17 patients. For most codons, mutation frequencies and number of reversions from mutated to wild-type amino acid were evaluated (Table 3). T215Y/F A/C/D/EG/H/I/L/N/S/V and M41L were the most frequently observed mutations for the reverse transcriptase gene (85.7 and 66.7% of the observed mutations, respectively). Few reversions to wild type were seen. For the thymidine-associated mutations (TAMs) only two reversions to wild type were seen at the M41, D67 and K70 codons and one reversion at each of the L210, T215 and K219 codons. None of the children had complete reversion of their TAM. One patient had a multi-drug resistance-associated mutation (Q151M). This mutation was not observed in the genotype evaluated just before resuming ART. For the NNRTI-associated mutations, nearly 20% of genotypes presented with the K103 mutation at baseline. No reversion was noted. In the protease gene the frequency of mutations ranged from 9.5% (L24) to 66.7% (L63) at baseline. Few reversions to wild type were observed. Only one out of 10 mutations at the L10 codon, three out of the nine at the M46 codon and two out of seven at the K20 codon reversed to wild type. None of the 14 mutations noted at the polymorphic L63 codon reverted.

Fig. 2

Table 3

Patients who resumed ART

Clinical and biological data at baseline and 4 weeks after resumed ART were available for seven patients who stopped treatment because of virological failure. The main reason for resumed ART was a drop in CD4% below 15% (six of seven) or rapid decline of CD4 cell count (one of seven). For those patients, the median baseline characteristics were not different than for the whole group with median age at TI of 9.8 years and median duration of TI of 25.7 weeks. The median baseline plasma HIV-1 PCR-RNA was 4.58 log₁₀ copies/ml and the median CD4 cell count was 601 × 10⁶ cells/l (median CD4%, 25; range, 9-36%). At the end of TI all patients received a triple combination (two NRTI + NNRTI, four patients; two NRTI + PI, one patient; NRTI + NNRTI + PI, one patient; NRTI + two PI, one patient) with a mean number of 2.3 new drugs (range, 1-3). According to the genotyping carried out just before resumed ART, a median of 74% (range, 17-100%) of the new drugs prescribed were considered as genotypically effective. The median viral load variation measured between baseline and 4 weeks after resumed ART was -1.01 log₁₀ copies/ml (range, -2.9 to -0.29). One out of seven patients achieved undetectable viral load (< 40 copies/ml) at 4 weeks after resuming ART. Median CD4 cell count variation between baseline values and those measured 4 weeks after resuming ART was -12.5% (range, -84 to +28%) (P = NS), and between nadir absolute CD4 cells × 10⁶ cells/l and 4 weeks after resuming ART it was +72% (P = 0.03).

Discussion

The case of the 'Berlin' patient, published in 1998, who was able to control his viral replication for 176 days after cycling on and off therapy focused attention on STI [¹⁹]. This patient was treated very early after the acute phase and treatment was interrupted without viral load rebound. It was suggested that the patient's HIV-1-specific cytotoxic T-lymphocyte response was responsible for the long-term virus suppression after treatment discontinuation as previously described in the simian immunodeficiency virus model [²⁷]. Following this report, treatment interruption in adult HIV-1-infected patients has been postulated to offer some benefits in three distinct clinical settings: acute treated primary infection, chronic stage of the disease with controlled viremia and salvage therapy [^17,19,20]. Numerous trials analyzing the clinical and biological consequences of STI in these different settings have been published in adult HIV-1-infected patients [^22,23,28-35]. To date, no data have been published concerning children with HIV-1 infection.

In our retrospective study no children presented with acute retroviral syndrome (ARS) or an AIDS-defining clinical event. The clinical risk of TI in children is hypothetic and can only be partially deduced from experience in adults. Actually, very few instances of retroviral syndrome have been observed [²⁹]. A case of aseptic meningitis and ARS was reported by Worthington in a recent issue of the journal [³⁶]. Some AIDS-defining events have been reported during STI mainly in advanced adult patients. Three out of 16 patients had such events in the Deeks series (Pneumonocystis carinii pneumonia, neuropathy and thrombocytopenia) [³³]. In larger cohort studies, data are more confused. In the CPCRA 064 randomized trial comparing immediate versus deferred change in HAART regimen after a 4-month treatment interruption in heavily pre-treated patients, STI was associated with greater progression of the disease [²⁹]. Conversely, in the GigHAART study, there was no excess of clinical events during treatment interruption [³⁵]. The difference may result from a shorter period of STI proposed in the latter clinical trial.

STI in patients with uncontrolled viral load resulted in a switch among predominant viral quasi-species from multi-drug resistant to wild-type in 30-60% of patients, theoretically improving viral susceptibility to subsequent salvage therapy [^21,22]. Several factors have been associated with a shift to archival wild type. STI duration did not differ among patients in the shift versus no shift groups. Only the baseline CD4 cell count at STI and previous duration of exposure to ART were associated with a shift to wild type in the cohort published by Miller et al. [³⁴]. Clinical stage was the only factor associated with genotype reversion in the study published by Izopet et al. [²¹]. The small number of AIDS patients (CDC stage C) in our cohort did not allow analysis of clinical stage influence.

In general this shift in resistance occurs abruptly and typically for all the drugs at the same time in 2-15 weeks [^21-23]. Meanwhile, reversion to wild type requires a drug-free period with both a CD4 cell decrease and a HIV RNA level increase, and a potential AIDS-defining event. It induces the predominance of a wild-type virus with greater replicative capacity, which is responsible for larger changes in CD4 cell count. The fall in CD4 cell count is thus more rapid than the average decrease in CD4 cell count observed in untreated patients. This is especially worrying because some patients already have low CD4 cell counts at the time of STI.

In our cohort, the decrease in CD4 cells during TI (-11.25 cells per week) is not different than those observed in adult studies. In the latter, the slope of CD4 cell loss varies according to the circumstances of treatment interruption (post-acute phase, sparing regimen or before salvage therapy), baseline CD4 cell count values, viral load increase and replicative capacity, level of drug resistance and the patient's specific cytotoxic T-lymphocyte response. A median decrease of 128 × 10⁶ cells/l was observed over the 12 weeks of treatment discontinuation reported by Deeks et al. [³³], and of 89 × 10⁶ cells/l in 48 patients in the Frankfurt cohort [³⁴], who had interrupted treatment for more than 2 months.

As viral rebound is not immediate, some researchers have postulated that short STI alternating with treatment periods might maintain undetectable viremia and prevent a drop in CD4 cell count. Unfortunately, trials have indicated that STI initiated in this setting, although relatively safe, failed to significantly reduce the viral load set point linked to the equilibrium between cellular immune-boosted response and viral replication [^25,37]. In most cases, although STI induced quantitative boosting of the patient's HIV-specific cellular immunity, no substantial change in viral load replication was observed [³⁷]. In the STACCATO study, 19 of the 36 evaluable patients had HIV-RNA concentrations > 500 copies/ml at the end of the week off therapy [²⁵]. In the Swiss-Spanish Intermittent Therapy Trial (SSITT) ART was stopped for 2 weeks and resumed thereafter for 8 weeks (four cycles) in 14 patients with plasma viral load < 50 copies/ml and CD4 cell count > 300 × 10⁶ cells/l at enrollment. Twelve out of 14 patients had detectable HIV-RNA at day 14 of the STI. Moreover, a continuous increase of viremia above the cycle baseline values was observed [³²]. Thus, it appeared that STI may not initiate immune control of viral replication during long-term interruption in the majority of chronically infected patients. Unfortunately, the circumstances of our study did not allow a thorough analysis of the control of viral replication by the patients' own immune systems but we must emphasize that 11 out of the 14 children in the SR group were still on ongoing TI at a median of 48 weeks after interruption. Thus, it seems that for some of them TI may offer the opportunity to safely reduce drug exposure and the associated toxicities. Moreover, in children and adolescents it also improves quality of life by providing an appreciable break from therapy.

In adult HIV-1-infected patients, median viral load increase varied from 0.4 to 0.84 log₁₀ copies/ml [^24,28,32,37] during STI with large inter-individual variability. Interruption of antiretroviral treatment in our HIV-1 chronically infected children resulted in a median of 1.26 log₁₀ copies/ml viral load increase (ranging from -0.2 to 4), 33% CD4 cell loss but no significant reversion mutation as observed in adult patients. The reasons for this discrepancy remain unclear. The virus populations from all the patients with detectable resistance reported by Deeks et al. became susceptible to PI within the 16 weeks following STI [³³]. Sixty-four percent of patients in the group assigned to STI showed partial or complete shift to wild-type genotype at month 4 in the CPCRA 064 study [²⁹]. A reversion was observed in 16 out of 31 patients in the deferred arm in the GigHAART study [³⁵]. Sixty-one percent of patients (23 of 38) displayed a shift from drug-resistant to predominantly wild-type genotype in the series of Izopet et al. [²¹]. The high degree of CD4 cell loss similar to adult data, in the absence of genotype reversion observed in our series is somewhat surprising. Indeed, in adult patients CD4 cell depletion observed during STI is mainly due to the increase in viral replicative capacities in the context of shift to wild-type reversion. [^17,18,33].

In spite of the low frequency of genotype reversion to wild type, ART was resumed in seven patients of the VF group. The gain in CD4 cell count measured 4 weeks after resuming ART was significant. Clinical stage (CDC stage A), greater number of new drugs, previous exposure to few drugs, shift to wild-type phenotype or genotype and adequate drug concentrations have been associated with a positive response to re-initiation of treatment in adults [^28,35,37-39]. The median of 1 log₁₀ copies/ml viral load decrease measured 4 weeks after resuming ART in the sub-group of seven patients is logical considering that more than 70% of the new resumed drugs were genotypically suitable. These data are in agreement with those published in adult cohorts. Sixty-three percent of patients in the Deeks cohort achieved durable virological response to salvage therapy [³³]. In the GigHAART randomized study, 62% of the deferred-arm patients achieved a 1 log₁₀ copies/ml reduction in plasma HIV level compared with 26% in the immediate treatment group [³⁵]. However, incomplete viral load suppression is of concern because low levels of persisting resistant viruses may rapidly be re-selected upon renewed exposure to treatment.

Data suggest that despite the presence of reduced drug susceptibility and partial viral load control, PI-containing HAART regimens can provide immunologic and virologic benefit in some patients [⁶]. On the other hand, recent data from PACTG 152 suggest that the presence of mutations associated with resistance after nucleoside ART is significantly associated with increased risk of clinical disease progression [⁴⁰]. TI in HIV-1-infected children is therefore disputable. Our retrospective study shows that surrogate markers during TI in HIV-1-infected children do not strictly behave as in adult cohorts. Prospective studies are mandatory to fully analyze clinical, metabolic, virological and immunological consequences of TI in children. Thus, in the event that a clinician decides to proceed with TI in HIV-1-infected children, we strongly recommend frequent monitoring, optimal opportunistic infection prophylaxis, close follow-up and inclusion in clinical trials.

Acknowledgements

We are indebted to Dr Jacqueline Cottalorda and Dr Jacques Durant for very helpful discussions and to Dr Frederic Berthier for his help in statistical analysis.