In this large ART roll-out program in South Africa we identified patients with confirmed virologic failure on second-line ART. Despite the absence of major PI resistance at first-line ART failure, most individuals failed second-line ART quickly (e.g. within a mean of 10 months) and two-thirds failed with wild-type virus. Furthermore, while patients remained on second-line ART with continued virologic failure, drug resistance did not develop over the follow-up period.
Given the known potency of second-line ART regimens and the low frequency of drug resistant virus found, it appears that medication non-adherence is an important cause of second-line ART failure. While access to pharmacy refill, pill count, and patient-level pharmacokinetic data were not available, there are plausible patient-level, regimen-specific, and structural explanations for adherence problems on second-line ART. Drug toxicities related to lopinavir and didanosine and the buffered formulation of didanosine (the enteric-coated formulation was unavailable at the study site) likely hindered adherence 
. Social and structural obstacles to adherence can include inaccessible clinic location or lack of access to transportation, work/child-care responsibilities, and decreased health care provider to patient ratio as a consequence of the rapid growth in ART roll-out programs 
. ART shortages, though not problematic at this site, can also be a challenge.
Most patients had substitutions in PR at sites that are known to be polymorphic in HIV-1 subtype C. These minor resistance mutations were also present at first-line ART failure in samples from the 10 patients who had samples available for sequencing from that time point. This observation confirms a prior genotype study from this clinic of HIV-infected individuals who were ART-naïve and who had failed first-line where the same polymorphisms in PR were present 
. It is also consistent with other studies that performed genotypic analysis of individuals infected with HIV-1 subtype C who were ART-naïve 
. The impact of these polymorphisms has been debated, without clear clinical evidence that they affect drug susceptibility 
Despite the absence of NNRTI exposure on second-line ART, virus from almost 30% of patients had at least one major NNRTI resistance mutation and virus from 15% had a mutation at the K103N codon, which suggests ongoing resistance to nevirapine and efavirenz. The persistence of NNRTI resistance is consistent with genotypic analysis of two other public sector South African cohorts of patients that failed PI-based ART 
. The low frequency of etravirine-associated mutations suggests excellent susceptibility to this next-generation NNRTI if it were to become available.
There are several therapeutic implications for the absence of NRTI resistance at second-line failure in this study. While other studies have suggested a higher rate of emergence of K65R at the time of stavudine-based ART failure in HIV-1 subtype C infection, we found that K65R did not appear after exposure to first-line stavudine or second-line didanosine; this may hold promise for the increasing use of tenofovir in resource-limited settings with predominance of nonsubtype-B HIV clades 
. M184V mutation and TAMs became undetectable supporting the hypothesis that lack of drug exposure due to ART non-adherence was the most likely cause of ART failure. In addition, consideration of archived resistance mutations is important in selection of second-line ART 
Females represented 85% of the patients whose viruses were sequenced, which is slightly higher than the distribution of females commencing second-line ART (75%)
. While the demographic and clinical characteristics of the cohort are largely consistent with other large ART programs, the small sample size may limit generalizability 
. Limited data were available on ART adherence, pharmacokinetic data, and contextual co-variates that may have an important impact on ART use and effectiveness. For example, those who receive treatment for other conditions, such as tuberculosis, may experience decreased levels of lopinavir/ritonavir if they are concomitantly taking rifampicin 
. Because data on prior ART exposure, such as for prevention of mother-to-child transmission were not available, we limited the analysis to individuals who were documented as ART naïve at initiation of first-line ART in the clinical record. This study did not test for minority drug-resistance variants that are not detected by conventional genotypic testing but contribute to ART failure 
The low frequency of drug resistance to boosted-lopinavir at the time of virologic failure in this large ART-roll out program in Cape Town confirms other studies in Johannesburg and Soweto, South Africa, as well as in North America and Europe 
. In vivo
resistance data suggest that the accumulation of accessory mutations in PR occurs rapidly only after major protease resistance mutations are established 
. The absence of new PR mutations over a short interval of virologic failure on PI-based second-line ART is consistent with this observation.
In those who experience virologic failure on second-line ART in South Africa, ART failure occurs quickly and most often with wild-type virus. However, rapid development of resistance does not occur. Interventions that successfully improve adherence could allow patients to continue to benefit from second-line ART therapy even after initial failure.