In this study, we determined the frequency and patterns of drug resistance mutations in children failing first-line ART in a rural primary health care ART programme where care is delivered largely by nurses and counselors. For older children on NNRTI-based regimens, drug resistance mutations were detected in the majority. Whilst in most cases the mutations would be unlikely to significantly compromise a second-line regimen based on a ritonavir-boosted PI, five patients had complex mutation patterns (three or more TAMs or Q151M complex) that might substantially limit the future activity of the NRTI class of drugs. In contrast, the younger children on PI-based regimens more often had no drug resistance mutations (six of 16 cases) and all but one had an absence of major protease mutations. This suggests a potential need for drug resistance genotyping, particularly in this group on PI-based regimens, to determine the appropriateness of regimen switch and to preserve first-line regimens where possible.
These data represent one of the largest drug resistance studies of paediatric patients failing ART undertaken thus far in South Africa [12–18]. In addition, this study was unique in being from a rural decentralized primary health care programme where the delivery of ART to adults and children has scaled up rapidly. The characteristics of the children included in the study highlight some of the challenges of ART delivery in this setting, with very long time spent on failing regimens, a finding also reported amongst our adults on ART . This suggests not only challenges to long-term adherence but also deficiencies in following the protocol for virological monitoring and switch to second-line ART guidelines. Problems with delayed switching have been well documented in South Africa [1, 9, 17] and these may be particularly problematic with children as many nurses and counselors are not confident managing paediatric ART and receive insufficient support and training in these issues. Some of the barriers to adherence in this setting are complex and difficult to address with the resources available within the health system. Furthermore, in rural areas with largely paper-based systems, results may simply be lost or misfiled and therefore overlooked.
First-line ART in South Africa using NNRTI-based regimens is challenging given the low genetic barrier of currently available drug options such as efavirenz and nevirapine and the fact that only one or two key DRMs are required to confer high-level resistance or cross-resistance to a drug class . High-level resistance mutations and cross-resistance severely compromises future ART options, a dire consequence for paediatric patients who require lifelong ART. We detected high levels of drug resistant mutations in the group failing NNRTI-based regimens. The majority (80%) had both NRTI and NNRTI resistance mutations. The proportion that had complex NRTI resistance patterns, such as three or more TAMs or Q151M complex, was lower (5.5%) than another recent study of 38 children on ART from the same province where 39% had three or more TAMs  despite a shorter duration (median = 2.8 yrs; IQR = 1.9- 2.3 yrs) on ART compared with our cohort (median = 3.3 yrs; IQR = 2.5-4.4 yrs). This suggests that the majority of children failing first-line NNRTI-based regimens in our cohort should retain susceptibility to a second-line regimen consisting of two alternative NRTIs and a ritonavir-boosted PI.
It was noteworthy that only one of 17 children on a PI-based regimens had a major PI resistance mutation. The low prevalence of PI mutations has been previously described in a number of manuscripts [17, 21, 22] This does raise the possibility of differential adherence to different components of the ART regimen (lopinavir/ritonavir syrup can be poorly tolerated), to problems with dosing of lopinavir/ritonavir syrup or possibly to drug-drug interactions particularly for those co-infected with TB, all issues we were unable to explore in detail for this study but which are subject to on-going research.
The finding that around nine in ten children with virological failure had at least one drug resistance mutation is consistent with other studies from South Africa [22–24] and a systematic review of studies from low- and middle-income countries, which reported a pooled proportion of 90% of children on ART with any DRM  The proportion with TAMs (23%) was similar to that seen in a similar paediatric programme in the Western Cape (19%) , although lower than the 56% reported in a systematic review of first-line failure of paediatric patients . The proportion with TAMs was surprisingly low given the long duration of virological failure, and was also lower than the 40% reported from adults in our programme with a similar duration of ART and similar time on a failing regimen in an adult cohort from the same region . This might suggest that adherence levels were either too low or too variable for the accumulation of TAMs over time. Alternatively, there was differential adherence to components of the regimen, with avoidance or suboptimal dosing of stavudine (d4T). The lack of major PI mutations in the young children on LPV/r-based regimens is consistent with other studies from the region which have shown PI mutations to be much more commonly associated with full dose ritonavir-based regimens [16, 21–23].
Currently our genotyping costs are approximately 50 US$ at reagents cost and less than 100US$ when staff and transport costs are added on. The normal cost of genotyping is 250–300 US$ in the public sector. In order to facilitate large-scale genotyping and in the interest of reducing costs, we did not perform a pre-genotype confirmatory viral load, yet we successfully genotyped 88% (89/101) of our cohort. Our genotyping system and reagents are likely to be affordable to upper middle-income countries like South Africa and Botswana but further cost reductions would be required to make drug resistance testing affordable in lower middle-income and low income countries within Africa. An additional feature of our study was that genotypes directed subsequent clinical care where a doctor, social worker and other clinic staff managed patients from enrolment to implementing an intervention post-genotyping as was carried out in an adult cohort from the same area. The results from our study addressed three of the ten goals of the Department of Health 2010 ART Guidelines in that we present a means of achieving the best outcomes for HIV-infected patients receiving ART in a cost-effective manner; we employ existing infrastructure, that of a decentralised rural public health clinic facility for patient management; and by identifying DRMs early we ensure patient retention on lifelong ART by instituting early interventions to halt ART failure and prevent morbidity and mortality.
Interpretation of these data should be subject to some limitations of the study. This was a cross-sectional study and whilst we identified as many children with first-line ART failure as possible, we were unable to accurately estimate what proportion of all children on ART had virological failure and we cannot be certain that we included all children meeting the eligibility criteria. Further, we did not have accurate information on prior exposure to pMTCT regimens either in the mother or infant therefore have no means of assessing its impact on the spectrum of DRMs we observed in this cohort. We can speculate that the patterns of NNRTI mutations we observed are suggestive of and may arise from pMTCT exposure, however we cannot determine with certainty from our data whether DRMs in these children were acquired or transmitted via MTCT. We note this as a limitation of the present study. Although these data represent the largest group of genotypes for children failing first-line ART in South Africa, the numbers remain small. This highlights the need for collaborative studies and surveillance from multiple representative sites in order to inform national policies. The data regarding protease inhibitor resistance mutations could be limited by exploring only the protease gene and it is possible that we missed mutations at other sites, e.g. Gag cleavage sites .