In this cohort of HIV-1 infected Kenyan children, we observed a virologic failure rate of 34% during a median of 49 months on first-line NNRTI-based cART. This is comparable to the viral failure rates seen in other pediatric cohorts in similar settings.7,8,20
The median time to virologic failure on first-line treatment in our study was 9 months, and it is notable that 82% of those that experienced virologic failure did so during the first 2 years on cART. Thus, the rate of failure was low in children who maintained viral suppression at 2 years. The major strength of our study is the long follow-up, which demonstrates that durable virologic response is an achievable goal in at least two-thirds of HIV-1 infected children treated with first-line cART in similar settings. Nevertheless, the proportion of children who experienced early virologic failure is cause for concern and indicates the need to further optimize adherence, especially in the initial months of treatment.
In this cohort, younger children had a higher likelihood of virologic failure even after controlling for baseline viral load, similar to previous findings.21
This could result from sub-therapeutic drug levels in younger children due to lower adherence or differences in pharmacokinetics. We did not monitor drug levels and therefore cannot confirm the possibility of sub-therapeutic treatment. However, younger children are fully dependent on a caregiver for drug administration and only 36% of our cohort reported disclosure to other family members, implying that the pool of potential caregivers able to administer medication in the event of the primary caregiver’s absence was limited. Although we inquired about missed doses and spitting out of medications, this information was based on self-report and may not be accurate. A study in the same facility found that self-report overestimates true adherence when compared to pharmacy records.22
In addition, pharmacokinetic data for most antiretroviral drugs is poorly defined for young children and under-dosing may occur. These findings suggest that younger children should be prioritized for virologic testing in settings where access to viral monitoring is available on a limited basis.
Aside from age, no other baseline characteristics were associated with viral failure, however our sample size was limited to 100 children total, and only 34 experienced viral failure, thus power was limited. In contrast to findings from a study in Uganda, we did not find that low baseline CD4 or type of NRTI backbone predicted virologic failure.8
In the Ugandan cohort (n=222), with shorter follow-up (12 months), lower baseline CD4 counts, male sex, and use of D4T-based treatment was associated with virologic failure. The smaller size of our study and homogeneity of baseline CD4 may explain the lack of detecting a similar association.
Two-thirds of the children with viral failure had resistance detectable at the point of failure, the majority of whom had 2 or more clinically relevant mutations resulting in multi-class resistance. At the point of virologic failure, the two most common mutations found were M184V, which confers high-level resistance to lamivudine, and K103N, which confers resistance to all first-generation NNRTIs. This is similar to findings from studies in Uganda, Central America, and Cote d’Ivoire and is in part due to the low genetic barrier to resistance for lamivudine and NNRTIs.23–25
The virus that bears the M184V mutation has been found to be relatively unfit, incapable of rapid replication, and has increased susceptibility to zidovudine, which may explain why a number of children in our cohort who remained on ZDV in the presence of virologic failure were clinically stable.26
In fact, in children in our cohort on ZDV-based cART, viral load was significantly lower at rebound compared to baseline in children with detectable M184V compared to children whose mutations did not include M184V (data not shown, P=0.01). The WHO guidelines were recently revised to retain lamivudine in second-line pediatric regimens due to the high prevalence and poor replicative capacity of M184V, and our findings confirm the relevance of these guidelines for Kenya.
TAMs and K65R were found at viral failure in 4 and 1 child, respectively, which was less frequent than the prevalence of NNRTI-associated mutations in our cohort, but higher than the prevalence observed in a large cohort in South Africa.27
These mutations limit the choice of second-line regimens and therefore present a challenge to children failing thymidine-based first-line.27
The Kenyan national guidelines were revised to give preference to abacavir over zidovudine in first-line ART to lower the potential for development of TAMs.16
One-third of children in our cohort who experienced virologic failure had no detectable resistance at the initial point of viral failure. The most plausible explanation for lack of viral suppression in these children is poor adherence. In the absence of resistance, it is possible for children to achieve virologic suppression if adherence is improved. Previous studies provide evidence that targeted counseling can lead to viral suppression, averting the need for second-line regimens.28,29
Therefore, as virologic testing becomes increasingly available in these settings, optimizing adherence should be the first approach to addressing viral failure when resistance testing is not available.
In our study, 22 children who did not meet the clinical criteria to switch to second-line cART had evidence of viral failure upon retrospective testing. However, only 12 (55%) of these children had evidence of antiretroviral resistance at viral failure. Thus, for 10 (45%) children viral suppression could possibly have been achieved with better adherence. These findings underscore the importance of resistance assays, which when available, add critical information to viral load assays to guide treatment.
Switch to 2nd
-line treatment was based on clinical or immunologic failure, which lagged viral failure by an average of 12 months. This extended period on first-line treatment in the presence of unrecognized viral failure resulted in the accumulation of additional resistance mutations in 18 of 23 children, and multi-class resistance often developed in children who had only single-class or no resistance at the onset of viral failure. There is evidence from other studies that this lag in switching to second-line treatment is associated with increased mortality rates, particularly when the first-line is NNRTI-based.30
A recent study found viral loads of >5000 copies/ml was associated with a nearly doubled risk of developing a WHO stage 3–4 event, independent of CD4 count, hemoglobin level and body mass index,31
. Thus, our study suggests that increased access to virologic testing may be useful for early detection of treatment failure and could improve treatment outcomes.
The number of children in our cohort who were switched to PI-ART was relatively small (n=14). However, a long follow-up (median 28 months after switch) showed that persistent virologic failure on second-line was rare. This was true although 10 of the 14 children that switched to PI-ART had detectable resistance to both NRTIs and NNRTIs prior to the switch, suggesting that PI-monotherapy may be effective in some children as shown in recent studies.32,33
Despite the lag following virologic failure on first-line, most sustained viral suppression on PI-therapy well beyond 2 years, and the emergence of detectable protease resistance was rare. This is reassuring in settings where third-line regimens, including second-generation boosted PIs or integrase inhibitors, are not feasible due to high cost.
Limitations of this study include the fact that the cohort was established primarily for research, which may somewhat limit generalizability. Resistance was assessed by population-based sequencing, which only detects resistant virus that comprises >20% of the viral population, and therefore it is possible that we missed resistance mutations present at lower frequencies in these children. In addition, the cohort was established in the pre-PEPFAR period, when access to ART was critically limited and therefore may represent very sick children and self-selected survivors. Baseline CD4% at cART initiation in this treatment program has progressively risen from 5%, when this cohort was established, to about 13% currently. Finally, this cohort did not have children with perinatal antiretroviral exposure and hence the findings may be less relevant to children with prior PMTCT exposure. Strengths of the study include the long follow-up with serially detailed viral and resistance data.
In summary, approximately a third of long-term cART-treated children experienced virologic failure during ~4 year follow-up, the majority of whom had antiretroviral drug resistance. Viral load assays may decrease the lag to treatment switch, and thus lessen the accumulation of additional mutations. However, without resistance assays it is not possible to distinguish failure due to non-adherence from viral rebound due to resistance. Children had excellent suppression on second-line therapy despite the lag in detection of viral failure.