Within the context of the SWEN trial to prevent breast-milk HIV transmission, we found that prevalence of NVP-R was higher for Indian infants who became subtype C HIV-infected in the first six weeks of life despite receiving daily NVP than for those who had received SD-NVP, a relationship that held even when low-frequency variants were considered. Furthermore, we found a substantial proportion of infants infected through breast-milk transmission during the first year of life had NVP-resistant HIV when low frequency variants were considered. The types of NVP-resistance mutations were similar between the two intervention groups but differed by timing of infection, where the Y181C variant was predominant in infections occurring in the first six weeks of life compared to the Y188C/H variant associated with late breast-milk infections. The pattern of persistence of NVP-R was similar at both population and low frequency levels at a median of one year of age, although only a small number of infants were studied.
The high prevalence of NVP-R in this study is not surprising given the long half-life of the drug 
and the rapid selection of NVP-R mutations observed in 45–87% of infants after SD-NVP 
. In our study, a significantly higher prevalence of NVP-R mutations was found by standard population sequencing among infants receiving SWEN (92%) and diagnosed with HIV infection by six weeks of age than those receiving only SD-NVP (38%). The prevalence of NVP-R remained higher with the SWEN regimen with low frequency clonal analysis, but because of small sample sizes we were not powered to detect a statistical difference. Others have reported higher prevalence of NVP-R mutations among Ugandan infants receiving extended-dose NVP while enrolled in the SWEN trial and diagnosed with subtype A or D infection in the first six weeks of life than for infants receiving SD-NVP 
. In our analysis, daily dosing of NVP did not alter the dominant NVP-resistant mutation selected in infants infected in the first six of weeks of life. The Y181C variant remained dominant as observed in studies of NVP-R following single-dose exposure in infants 
and in earlier studies of daily NVP monotherapy for treatment of infected individuals 
Among infants born to SD-NVP exposed women and who escape in-utero
or peripartum/early-breastfeeding, the potential for acquiring NVP-resistant HIV through breastfeeding continues, as NVP-R mutations are present in breast-milk in up to 65% of women exposed to SD-NVP 
. In our study, NVP-R did not differ by receipt of SD-NVP or SWEN among infants infected through late-breastfeeding. In both groups, 15% of infants had NVP-R detectable at population levels, with 31% of SWEN and 40% of SD-NVP exposed infants acquiring NVP-R when low frequency variants were considered. The predominant transmitted variants, here, were Y188C/H and K103N, reflective of the most common mutations found in women following SD-NVP exposure 
. Maternal HAART exposure also did not play a role in the resistance outcomes we observed in these infants, as genotyping was done prior to initiation of maternal HAART and therefore, not due to exposure to suboptimal NVP concentrations in breast-milk. Furthermore, in the few samples available for analysis from women who transmitted HIV through breast-milk, the mutations present in plasma by standard population sequencing or clonal analysis often differed from those detected in their respective infants, highlighting the potential compartmentalization of HIV variants within the breast and the complexity of breast-milk transmission 
While it is well established that the use of single antiretroviral agents in HIV-infected individuals virtually guarantees selection of drug resistant virus, the efficacy of administering HAART to HIV-infected women for prevention of breast-milk transmission is currently under study 
. However, preliminary data from one study suggest that even with the administration of maternal HAART, the risk of transmission of drug-resistant HIV through breastfeeding remains. In the Kisumu Breastfeeding Study (KiBS) in Kenya where NVP or protease-inhibitor based HAART was administered to women from 34 weeks gestation through 6 months postpartum, 67% of 24 infants had either lamivudine and/or NVP-resistant HIV 
. Since continued breastfeeding improves child survival, even for HIV-infected infants 
, use of antiretroviral drugs in the mother and or infant to prevent breast-milk transmission is currently the only available option for preventing HIV infection to the infant in resource-constrained settings, but, as seen here, this practice carries an inherent risk of drug resistance.
NVP-R as a consequence of prophylaxis to prevent MTCT is highly relevant because treatment in resource-constrained settings relies on NVP as a key component of HAART. NVP-R variants arising after SD-NVP are eventually replaced in blood by wild-type virus 
, and so NVP-based HAART remains the first-line treatment regimen. In the context of extended NVP prophylaxis, a higher proportion of SWEN-exposed Ugandan infants had detectable NVP-R at population levels at six months of age than did SD-NVP exposed infants (100% of 7 vs. 17% of 6) 
. However, in our study, NVP-R variants decayed in 70% of infants exposed to SWEN or SD-NVP by a median of 12 months of age, but remained at low frequencies in 86% of infants. The clinical implications for infants of this NVP-R and its persistence at low frequencies, however, are unclear. One small study showed a higher risk of virologic failure for SD-NVP-exposed than for unexposed infants on NVP-based HAART initiated at a median of 8 months 
. Time to treatment initiation following prophylaxis in infants may also be critical, as SD-NVP exposed women who initiated NVP-based HAART less than six months after delivery had a higher risk of virologic failure than those who started treatment after six months 
. Recent evidence showed that early therapy in HIV-infected infants can improve survival 
, which has led the WHO to recommend early HAART for all HIV-infected infants less than 12 months of age 
. Therefore, if the findings on persistence of NVP-R from the Indian and Ugandan SWEN trials are confirmed in a large number of subtype A, C, or D HIV-infected infants, then this will influence the use of NVP as part of first line therapy, if extended NVP prophylaxis is implemented.
Our study has some limitations. First, unlike other studies of SD-NVP prophylaxis, we were unable to test for NVP-R at a fixed time point, although the median age at genotyping for infants infected in utero or through peripartum/early-breastfeeding was similar to other published studies. Assessing NVP-R at low frequencies using clonal analysis is limited by the number of clones required to detect NVP-R variants and may underestimate resistance at frequencies <5%. However, cloning did allow comprehensive assessment at all eight known sites of NVP-R in HIV reverse transcriptase. In addition, due to the limited number of maternal samples available for genotyping at six months and lack of breast-milk samples, drug resistance mutations present in plasma and breast-milk could not be determined or correlated with HIV transmission.
In summary, our results suggest the extended-dose NVP regimen used in the SWEN trial to prevent breast-milk transmission of HIV does carry a higher risk of infection with drug resistant virus than does SD-NVP in subtype C-infected infants, and that the risk of infection with NVP-R variants at low frequencies continues through the first year of life with breastfeeding. Given the high mortality associated with pediatric HIV infection in the first year of life, the complexities of early infant diagnosis, and limited resources for initiating HAART in infants in settings where breastfeeding itself is critical for survival, the primary goal should be to prevent peripartum and breast-milk transmission with infant antiretroviral prophylaxis. As other prevention options become available, this increased risk could be revisited.