Chapter  156:  Antiretroviral (ARV) Drug Resistance in the Developing World

Review of Published Overviews on Incidence and Prevalence

Several overviews on the incidence and prevalence of ARV resistance were identified; however, most provided data only on the developed world. The WATCH study found that the rate of resistance (to any drug) was 5.5 percent in Africa, 7.4 percent in East Asia, 5.7 percent in Southeast Asia, and 6.4 percent in Latin America, lower than in North America (11.4 percent) and Europe (10.6 percent). The patterns of resistance world-wide appear to reflect trends in use of ARV medications: In developed areas, where medications have been more widely available, rates of resistance to NNRTI range from three to four percent, whereas, in developing regions with limited resources, the rates of resistance are much lower.

Resistance Among Treatment-naïve Persons in Developing Countries

We analyzed resistance prevalence separately for the following areas: India, other developing nations in Asia, Latin America, and the continent of Africa.

Studies of ARV resistance among treatment-naïve patients in India were rare. Published data were found from only three studies;¹⁹–²¹ these studies varied in their results. A study of 60 persons in Northern India found that at least 80 percent of patients had resistance to ZDV.¹⁹ A study of 50 persons in South India showed that fourteen percent were resistant to NNRTIs, 6 percent were resistant to NRTIs, and 20 percent were PI resistant although all had mutations that conferred partial resistance.²⁰ A study of 128 persons in Mumbai found that only two were resistant to NRTIs.²¹

Only three studies were found that reported on rates of resistance in other developing regions of Asia.²²–²⁴ None reported NNRTI resistance. NRTI resistance ranged from four to seven percent; primary resistance to PIs ranged from two to three percent.

In Latin America, eight studies identified rates of ARV resistance. Resistance to NRTIs ranged from 2 to 14 percent among treatment-naïve groups.²⁵–³² Reported rates of resistance to NNRTIs were low, ranging from zero to two percent. In a handful of studies that included Brazilian and Argentine populations, rates of secondary PI mutations were high, presumably among persons with HIV clade B. No studies assessed HIV resistance in Central America.

In sub-Saharan Africa, 14 studies identified rates of ARV resistance among the treatment naive. NNRTI resistance rates ranged from 0 percent to 7.7 percent, and were associated with infection with most HIV clades.³³–⁴⁶ NRTI resistance rates ranged from zero percent to eight percent for all clades. Primary PI mutations were rare (less than three percent) among Africans; however, there were high levels of secondary PI mutations in most studies.

Resistance among Pregnant Women and Children in Developing Countries

Thirty-one studies were identified that addressed the emergence of ARV resistance in women in developing countries subsequent to the administration of therapy to prevent pMTCT of HIV-1 infection. These studies were highly heterogeneous with respect to study design and outcomes.

Studies of resistance following SD-NVP. Seven studies assessed the effects of SD-NVP on postpartum NNRTI resistance; resistance rates ranged from 9 to 69 percent.⁴⁷–⁵³ Seven studies that assessed the effect of SD-NVP among women infected with clade C on resistance reported rates of prevalence that varied from over 70 percent at 2 weeks postpartum to less than nine percent at six weeks postpartum; these findings suggest a decline in resistance over time.^{49, 52, 54}–⁵⁸ In studies of women infected with other clades,^{48, 49, 59}–⁶¹ resistance ranged from 16 percent⁶⁰ to 36 percent.⁴⁹

Complex patterns of NNRTI resistance were observed following the receipt of SD-NVP. The most common mutations that resulted in amino acid changes were K103N and Y181C. Y181C was the most commonly detected resistance mutation in plasma collected 2 weeks postpartum from women infected by HIV-1 clade C.^{54, 55} In contrast, the K103N resistance mutation was most common at later time points in persons infected by clade C and at all times in persons infected by other HIV-1 clades. Among clade-C women, from one study in South Africa, V106/A/M was the most common amino acid change.⁵⁴ One report found that the rates of detection of K103N, Y181C, and Y188C were significantly higher in women infected by HIV-1 clade C than by HIV-1 clades A or D.⁴⁹

Several longitudinal analyses support the observation of declines in NNRTI resistance with time and suggest that these declines may differ by clade.^{47, 48, 52, 55, 62}

Studies of resistance following other pMTCT regimens. Ten studies from the developing world were identified that studied the emergence of ARV resistance following other pMTCT regimens, including adding ARV therapy before or after SD-NVP. Two studies were done in Thailand,^{63, 64} one was conducted in Argentina,⁶⁵ and the other eight studies were done in Sub-Saharan Africa.^{50, 60, 66}–⁷¹ A randomized comparison of outcomes in women receiving pMTCT with SD-NVP alone with those of women who received SD-NVP followed by a “tail” of 3 or 7 days of ZDV and 3TC found that the prevalence of NNRTI resistance in these three groups was 57 percent, 13 percent, and 9 percent, respectively.⁴⁶

In one study, women who received either intrapartum SD-NVP or placebo following daily ZDV during their third trimester were put on three-drug ARV postpartum. Intrapartum SD- NVP was statistically associated with NRTI and NNRTI mutations at both 3 and 6 months. In another study, in which women were put on three-drug ARV after intrapartum SD-NVP or placebo, resistance was strongly associated with intrapartum SD-NVP among women who started ARV within 6 months postpartum, but not among women who started ARV 6 months or more postpartum,⁷² suggesting development of resistance may be associated with shorter interval between receipt of SD-NVP and initiation of full ARV.

Several trials suggest that when infants receive SD-NVP at birth in the background of ZDV, maternal SD-NVP does not provide any additional protection from HIV transmission.^{71, 73}

Effect of test sensitivity. The rates of NNRTI resistance are consistently greater when more-sensitive assays are employed.^{51, 58, 65, 74} For example, the prevalence detected with RT-PCR (real time polymerase chain reaction) of plasma RNA can be up to double that detected with standard testing at six or seven weeks post-partum.

Resistance in HIV-positive Infants Born To Mothers Receiving pMTCT

Children are often born HIV infected despite pMTCT. Resistant viruses can be transmitted or can emerge independently in either mother or infant. Eight studies^{50 54 55 57 60 62 69 75} evaluated the emergence of ARV resistance in children born to mothers in sub-Saharan Africa who received pMTCT. Most studies examined the impact of SD-NVP. In most studies, infected infants were tested for resistance at 6 or 7 weeks of age. Rates of NNRTI resistance among untreated HIV-positive infants of mothers receiving only SD-NVP ranged from 36 percent to 50 percent. One study found that detectable resistance fades over time in infants exposed to SD-NVP.⁶²

Several studies compared the effects of multiple drugs on infant drug resistance. One study compared resistance rates among mother-infant pairs infected with various HIV clades who were treated with ZDV (N = 48) or SD-NVP (N = 61).⁶⁰ The HIV transmission rates were similar in the two groups: 16.4 percent and 16.7 percent respectively. Using the oligonucleotide DNA ligation assay for resistance at 6 weeks postpartum, the researchers found NNRTI mutations in half the infected pairs who received SD-NVP. No ZDV mutations were present in pairs treated with that drug. Resistance data for mothers and infants were not presented separately. In a second study, researchers gave mothers ZDV (300mg) twice daily starting at 36 weeks gestation in addition to oral SD-NVP at onset of labor.⁶⁹ Only 23 percent of their HIV-positive children were resistant to NNRTIs at 4 weeks postpartum. In a third study, researchers compared resistance among infants who received SD-NVP with those who received SD-NVP plus either 4 or 7 days of ZDV + 3TC after birth.⁵⁰ Infected infants who received the additional 7-day ZDV + 3TC showed no resistance at 6 weeks, compared to 78 percent of those who received SD-NVP only. Finally, a fourth study recently compared resistance rates in infants who received various combinations of intrapartum SD-NVP, SD-NVP (2 mg/kg) immediately postpartum, and ZDV (4 mg/kg) twice daily for one week.⁷⁵ At 6 to 8 weeks, the rate of resistance was much lower (27 percent) in the infants who received the postpartum SD-NVP and ZDV without the intrapartum SD-NVP. The study also showed that when infants received ZDV, the transmission rate did not differ between mothers who received SD-NVP and those who did not. When infants received SD-NVP plus one week of ZDV and the mother received SD-NVP, transmission was 17 percent and resistance was observed in 74 percent; when infants received SD-NVP plus 1 week of ZDV and mothers did not receive SD-NVP, transmission was 17 percent but resistance was only 27 percent. Data from the MASHI trial also suggest that when the infant receives SD-NVP at birth in the background of ZDV, maternal SD NVP does not provide any additional protection.^{51 65 72 74}

Predictors of Resistance. Several studies evaluated predictors of ARV resistance among persons receiving SD-NVP. A multivariate analysis demonstrated an increased risk of resistance detectable by standard genotyping assays in persons infected with clade C (compared with clade A or clade D) or with increased viral load at delivery, but not with increased age or number of births.⁴⁹ One additional study supports the predictive value of increased viral load.⁵⁴ Plasma HIV RNA level⁶⁴ has also been shown to be predictive of resistance, and increased duration of prepartum use of ZDV+3TC for pMTCT was associated with an increased maternal prevalence of the M184V resistance mutation.⁷⁶

Adherence and Resistance

Adherence to HIV treatment has long been associated with treatment outcomes. Most studies in the developed world have found evidence that at very low levels of adherence, there is insufficient selective pressure for mutations to evolve. At high levels of adherence, viral replication, and thus, viral evolution, should cease.

Few studies could be identified that were designed to assess the effect of health services delivery factors or medication adherence on the development of resistance in patients in developing countries. Three studies assessed the impact of treatment interruptions on ARV resistance in the developing world: two were conducted in Kampala, Uganda.^{77 78} In the first study, 137 patients received a three-drug regimen; most included NVP (77 percent) or efavirenz (14 percent). At a median of 38 weeks, 91 of the 137 patients had undetectable viral loads. Of the 137 patients, 32 experienced an unplanned treatment interruption of more than 4 days. In multivariate analysis, this treatment interruption was statistically associated with resistance, as was being treatment-naïve at study entry. Of 36 patients for whom genetic data were available, 26 were resistant to NNRTIs; the most common mutation was K103N. NRTI resistance was measured in 23 patients, attributable to the M184V/I mutation. A second study found that among 95 participants, the 33 without treatment interruption had no drug resistance, compared to eight of the 62 participants who did interrupt treatment. A third African study that assessed the relationship between adherence and the presence of resistance in an Ivory Coast AIDS clinic found no association between missed treatments and resistance.⁷⁹ Finally, one additional study from the UK studied black African children with non-clade B virus.⁸⁰ All children whose adherence was classified as good or intermediate had resistance, compared with five of eight children whose adherence was poor. Of critical importance, unlike many studies conducted in the developed world, the latter studies were not initially designed or powered to assess the relationship between adherence and resistance.

Stanford University HIV Drug Resistance Database

The HIV RT and Protease Sequence Database is an on-line relational database that catalogs sequence variation in HIV reverse transcriptase (RT) and protease enzymes, the molecular targets of anti-HIV therapy. The database was developed in order to help physicians and researchers determine clinical cross-resistance among current and experimental anti-HIV drugs and to identify any drug regimens that retain their effectiveness against drug-resistant HIV-1 isolates. The database links HIV-1 RT and protease sequence data, drug treatment histories, drug susceptibility, and clinical parameters.

One part of the database provides copies of or links to published papers, meeting abstracts, and GenBank entries that have not yet been published as a paper or abstract. We searched the publications list in January, 2006, and downloaded relevant studies. We did not conduct our own analysis of the patient-level data stored in the RT and Protease Sequence database.

WATCH: Worldwide Analysis of Resistance Transmission over Time of Chronically and Acute Infected HIV-1 infected persons

Located in Europe, the WATCH project aims to collect HIV-1 RT and protease sequences from treatment-naïve participants all over the world and analyse them together in a standardised manner, in order to be able to make a good comparison of resistance figures. We found two WATCH conference abstracts and contacted the authors about additional research in progress.

Presentation - Preventing Mother-to-Child HIV Transmission and ARV Drug Resistance

Dr. Lynne Mofenson of the National Institute of Child Health and Human Development provided us with a copy of an overview she presented on March 16, 2006; this excellent overview contained nearly 100 slides, from which we obtained all relevant studies cited.

Article - The Global Status of Resistance to ARV Drugs

In 2005, Vella and Palmisano published an overview in the journal Clinical Infectious Diseases.¹⁸ They searched for published studies on the incidence and prevalence of drug resistance in persons recently infected with HIV in all regions of the world. We retrieved the one study from a developing country (Ivory Coast).

We also searched the proceedings of the following conferences for the past 2 years. (We chose a 2 year cut-off because we expected that in general, abstracts and presentations would be published in journals within 2 years of the conference.)

Conference on Retroviruses and Opportunistic Infections (CROI)

The most recent CROI conference was held in Denver, Colorado on February 5–8, 2006. The CROI is a scientifically based meeting of leading HIV/AIDS researchers from around the world, and the 2006 conference was sponsored by the Foundation for Retrovirology and Human Health, the National Institute of Allergy and Infectious Diseases, the Centers for Disease Control and Prevention, and the University of California, San Diego School of Medicine. Participants included scientists actively engaged in basic science or clinical studies examining retroviral diseases and their complications or full-time academic clinician-teachers responsible for HIV/AIDS training and research programs. The aim of the conference was to enable researchers to present, discuss, and critique their investigations with the ultimate goal of translating their research into progress in the fight against HIV/AIDS.

International AIDS Conference

The most recent International AIDS Conference was held in Toronto, Canada on August 13–18, 2006. The conference is the largest HIV/AIDS meeting in the world; participants include scientists; health care providers; political, community and business leaders; journalists; government, non-governmental and intergovernmental representatives; and people living with HIV/AIDS. The goal of the conference is to allow for the exchange of ideas, research, and knowledge to help strengthen HIV/AIDS programs worldwide, with an emphasis on evidence, outcomes, and best practices. The theme of the 2006 conference was A Time to Deliver, underscoring the still urgent need to bring effective prevention and treatment programs to communities worldwide.

IAS Conference on HIV Pathogenesis and Treatment

The International AIDS Society (IAS) is a professional society with over 6,000 members, including scientists, health care workers, and others involved in the prevention and treatment of HIV/AIDS. Every 2 years, IAS sponsors a conference to disseminate knowledge that can accelerate the response to HIV/AIDS worldwide. We examined the proceedings from the July 24–27, 2005 conference, which took place in Rio de Janeiro, Brazil and was co-sponsored with the Universidad de Federal do Rio de Janeiro and the Brazilian Society of Infectious Diseases. The specific aim of the conference was to focus on new insights into HIV disease development, prevention, and care and their practical applications in developing countries.

Meetings of the Infectious Diseases Society of America (IDSA)

The meetings of the Infectious Diseases Society of America (IDSA) are geared towards physicians, scientists, and other health care professionals involved in research, patient care, public health, disease prevention, and education in the field of infectious diseases (ID). The meetings are designed to disseminate current knowledge and advancements in the field, bridge the gaps among various medical disciplines, and promote multidisciplinary dialogues and communications that will facilitate the prevention and treatment of ID. The 2006 meeting of the IDSA took place on October 12–15 in Toronto, Canada, and included four program tracks: investigative infectious diseases (ID), adult ID, HIV, and pediatric ID. The IDSA website also provides a meetings archive that enabled us to examine the proceedings.

Study Inclusion

Our initial search was unrestricted by study design. The studies included in the review are of one of the following types of designs.

Review articles identified by the search were classified as either systematic (including meta-analyses) or nonsystematic. Systematic reviews were identified by reading the methods section of the article to determine whether an acceptable method was employed to identify evidence (such as a description of the name of the computerized database searched and the full set of search terms used, as well as details about the method for accepting and rejecting identified articles).

Randomized controlled trials (RCTs) are studies where the participants are definitely assigned prospectively to one of two (or more) alternative forms of intervention, using a process of random allocation (e.g., random number generation, coin flips).

Controlled clinical trials (CCT) s are studies where participants (or other units) are either

• (a)

  definitely assigned prospectively to one of two (or more) alternative forms of health care using a quasi-random allocation method (e.g., alternation, date of birth, patient identifier)

  OR

• (b)

  possibly assigned prospectively to one of two (or more) alternative forms of health care using a process of random or quasi-random allocation.

Observational studies (such as cohort, case-control, and cases series) are those where the investigators do not control who gets the interventions. Much of the data included in this report comes from observational studies. To avoid reviewing potentially numerous case reports, we initially set a threshold of 50 or more participants per series for inclusion in our review. After discussion with our TEP, we reduced this number to 20 for studies of populations in developing countries, so as not to exclude potentially important data.

Screening

Using a single-page “screening form” (included in Appendix C ^*), we reviewed the studies retrieved from the various sources against our exclusion criteria. Two reviewers, each trained in the critical analysis of scientific literature, independently reviewed each study and resolved disagreements by consensus. The lead investigators resolved any disagreements that remained unresolved after discussions between the reviewers.

To be included in our report, a study had to focus on human resistance to ARV. Studies of cell lines or animals were excluded. There were no language limitations; we included many non-English studies Studies of obsolete therapies were also excluded at this time, if it was very unlikely that such therapies would never be used again, even in resource-limited regions.

After the TEP further narrowed the scope of the project, we created a second one-page screening form. Using this form, we tracked articles on treatment-naïve patients, pregnant women and children. We also used this form to identify studies of entire geographic regions, large hospitals, and specific medications.

During this phase of screening, we were asked to exclude studies not set in the developing world. “Developing” generally refers to economic growth and social development in areas such as health care, literacy, life expectancy, and fertility. Some less-developed regions could actually be considered non-developing, as a number have experienced years of decline rather than development. However, the United Nations (UN) allows each country to decide for itself whether it will be designated as “undeveloped” or “developing.” For the purposes of our report, we chose to include the following “undeveloped” and “developing” regions: Latin America, sub-Saharan Africa, India, and all other areas of Asia except Hong Kong, Macau, Singapore, and Japan.

India

In recent years, ARV treatment in India has expanded. However, studies on resistance are quite rare. We found only three studies conducted in India that reported resistance in treatment-naïve HIV positive persons, all published in peer-reviewed journals (Table 1). All studies conducted genotypic testing; none conducted phenotyping. None of the studies stratified data by mode of transmission, and none estimated time from seroconversion to resistance testing. Each study tested consecutive patients entering treatment.

One study¹⁹ from North India assessed 60 men, women, and children from 2000-2002 using a sensitive nested amplification refractory mutation (ARMS) PCR test as well as a sequence-based mechanism. Of the NRTI mutations, about 32 percent occurred at codon 184, 80 percent at codon 70, and fewer than 2 percent at codon 215. NNRTI and PI mutations were not reported. Most participants showed a mixture of wild-type and mutant virus. CD4 count and gender were not statistically associated with mutations. Frequency of mutations at codon 70 was higher in adults than children.

A second study²⁰, collected data on 50 male and female adults in South India in 2002 and 2003. All were infected with HIV clade C. Fourteen percent were resistant to NNRTIs and 6 percent were resistant to NRTIs. The most common PI mutations were naturally occurring polymorphisms, including M36, L63, and I93. Twenty percent had PI mutations at major positions; the most common was V82.

The final study collected data in 2003 from 128 participants in Mumbai (mean age 30.4 years).²¹ Over 96 percent were infected with HIV clade C. Two participants were resistant to NRTIs, as the M184V mutation was present. Polymorphisms such as M36 and L63 were common, but no major PI mutations were reported.

In sum, the populations assessed were very small compared to the number of persons living with HIV in India. We are not sure how well they represent the regional populations. Results varied, and data were not usually stratified by variables of interest. More research on ARV resistance in India is recommended.

Other Developing Nations in Asia

We found only three studies that reported resistance patterns in treatment-naïve persons in other developing regions of Asia (we excluded Japan, Singapore, and Hong Kong) (Table 2). The first two studies did not stratify results by gender; none of the studies stratified by mode of transmission. All studies conducted genotypic testing; no phenotyping was conducted. All were published in peer-reviewed journals.

The first study²² reported data for 50 adults, including MSMs and IDUs, in South Korea. Data were collected between 1998 and 2002. It is unclear how participants were selected for testing. Ninety-four percent of the sample was infected with HIV clade B. Participants were tested for resistance an average of 2 months after testing positive for HIV. None were resistant to NNRTIs, 6.5 percent were resistant to NRTIs, and 2.6 percent were resistant to PIs. The most common NRTI mutations were D67N, K70R, V118I, T215F, and T219Q.

Another study²³ reported data on 200 adults and adolescents in Vietnam; 42.5 percent were IDUs. Data were collected in 2001 and 2002. HIV clade was primarily CRF01_AE. None of the participants were resistant to NNRTIs; 4.5 percent were resistant to NRTIs, and 2.0 percent were resistant to PIs. The most common NRTI mutations were M41L, K219Q, and M184I. PI mutations found were D30N and L90M.

The last study²⁴ collected data on all pregnant women (N = 124) and other adults (N=22) who consecutively entered a treatment program in Cambodia in 2003 and 2004. They were tested for resistance within 1 year of a known date of seroconversion. HIV clade was primarily CRF01_AE. NRTI mutations were found in 3.4 percent; the most common mutations were K70R, V75M, and K101E. Major PI mutations were found in two participants (1.4 percent); one had M46I, the other N88D.

The samples in these studies were very small compared to the numbers of persons living with HIV in the region. Resistance data were not stratified by HIV risk factors, although that information was collected in at least two of the studies. Rates of resistance were fairly low among both clade B and recombinant clade CRF01_AE; rates were similar to those reported in Southeast Asia by the WATCH study.{Bowles, Wensing, et al. August 11–12 #2102}⁹⁸ The specific mutations in the clade B South Korean participants were quite different from those found in the Southeast Asian CRF01_AE participants.

Latin America

Nine studies reported data on resistance to ARV medications among treatment-naïve populations in Latin America. We found one study from Argentina, three from Brazil, one from Peru, one from Venezuela, and two from Mexico (one study⁹⁹ combined data from several countries-Brazil, Portugal, Ivory Coast, Nigeria, Guinea Bissau, Lebanon, and the U.S. -but did not stratify data by region or HIV clade, so it was excluded). Data were collected between 1997 and 2004. None of the studies reported time from seroconversion to assessment. The Peru and Argentina studies did not report HIV clade, nor did one of the studies from Mexico. None of the studies stratified resistance data by gender or mode of transmission, although one study included only MSM. All studies conducted genotypic assessment; two conducted additional phenotyping. Many studies did not report how participants were selected for resistance testing; statements were often vague as to whether all, or only a sample of patients entering treatment, were tested.

The study from Argentina²⁵ assessed resistance among adults, some of whom were IDUs. HIV clade was not reported. The authors compared persons with primary infection (N=13) to those with established infection (N=86). Of those with primary infection, one person was resistant to NRTIs and one was resistant to PIs. Of those with established infection, 1.2 percent had mutations to NRTIs and 1.2 percent had primary PI mutations (83.7 percent had secondary PI mutations). Line probe assays (LiPA), which identify specific mutations by a reverse hybridization strategy, were conducted on 52 of the established infection samples;¹⁰⁰ 11.5 percent were resistant to NRTIs.

In Brazil, a study of 409 treatment-naïve women (mostly infected with HIV clade C), MSMs, IDUs, and others showed very low resistance levels.²⁶ Of the group, 2.06 percent were resistant to NNRTIs, 2.36 percent were resistant to NRTIs, and 2.24 percent were resistant to PIs. Another study of 49 adults, mostly infected with clade B HIV, reported that none were resistant to NNRTIs, 2.0 percent were resistant to NRTIs, and 85 percent had secondary PI mutations.²⁷ Another study of 50 adults (mostly infected with clade B), including MSMs, reported that 14 percent were resistant to NRTIs and 85.7 percent had secondary PI mutations.²⁸ The secondary mutations in these studies were most likely naturally occurring polymorphisms with no known relationship to resistance.

A study of MSM in Peru²⁹ compared resistance in 359 treatment-naïve and 16 treatment-experienced patients. Resistance to drugs was low among the treatment naive - 3.3 percent overall, compared to 31.3 percent overall among the treatment experienced. HIV clade was not reported.

A United Nations (UN) AIDS clinic in Venezuela³⁰ reported data on adults including IDUs, MSM, and women. All but two persons had HIV clade B. The authors compared resistance rates of PI-naïve (N=56) and PI-experienced (N=44) patients. Some of the PI-naïve patients had experience with mono- and dual therapy. The rate of resistance was low among the PI naive - only one person was resistant to any drug. Among the PI-experienced patients, 40.9 percent were resistant to NRTIs and 9.1 percent were resistant to PIs.

Finally, studies in Mexico reported that 5.0 percent of 20 adults infected with clade B who enrolled in treatment at a Jalisco clinic were resistant to PIs,³¹ while of 45 patients who enrolled for treatment in Central Mexico, 11.0 percent were resistant to NRTIs and 2.0 percent were resistant to PIs.³² Again, stratified data were not reported.

In sum, rates of resistance to NRTIs ranged from 2.0 percent to 14 percent among treatment-naïve groups in Latin America. Rates of resistance to NNRTIs were low, ranging from zero to two percent. Rates of PI polymorphisms were high in Brazil and Argentina, presumably among persons infected with HIV clade B. Resistance data were rarely stratified by mode of transmission or gender, although those data were available in many cases.

The number of persons tested for resistance in these studies is small relative to the number of people taking HIV ARVs in Latin America. Some studies tested all consecutive patients entering clinic^{26, 29}, while others did not explicitly state how participants were selected. Therefore, it is difficult to judge how representative the data are. We found no published reports of resistance rates in Central American countries, despite the high HIV prevalence rates in some of those nations.

Africa

We found 14 studies that reported resistance data on treatment-naïve populations in sub-Saharan Africa. They are displayed in Table 4. Data were collected between 1995 and 2002. Studies were conducted in Cameroon, Senegal, Republic of Congo, Ivory Coast, Zambia, Gabon, Ghana, Malawi, South Africa, Zimbabwe, and Botswana; all reported HIV clade. The most common clades were CRF02 AG, which predominates in west Africa, and C, which is predominant in southern Africa. All studies conducted genotypic testing; three conducted additional phenotypic assessments³³–³⁵. Two studies included only pregnant women^{36, 37}; another included only men³⁵. The other studies did not stratify results by gender. None of the studies stratified results by mode of transmission; the majority of studies described their participants as heterosexual adults.

Of the studies that described participant selection, two studies^{38, 39} tested random samples of patients, and one study tested all pregnant women entering a clinical trial.³⁷ The other studies did not clarify how they selected participants for resistance testing. Many implied that all incoming HIV positive persons were tested; however, none explicitly stated this.

Regarding treatment-naïve pregnant women, one study³⁶ performed genotypic testing on 28 women who entered treatment in Zambia in 2000; 92.8 percent of the women were infected with HIV clade C. None had any primary mutations; 89.3 percent had secondary NRTI mutations, and all had secondary PI mutations. Most secondary mutations were polymorphisms. Testing for RT mutations was also performed on samples from 37 pregnant women who entered a clinical trial in South Africa the same year.³⁷ No women had resistance to NRTIs or NNRTIs; all were infected with clade C.

The study that included only males³⁵ found that although none of the 37 Ghanian men (mostly infected with mostly clade CRF02 AG) had primary PI mutations, secondary mutations were “commonly observed.” Exact figures were not reported. Testing for RT mutations was not performed.

Two studies compared resistance among small samples of people infected with rarer HIV clades. Vergne⁴⁰ reported data collected from 1995 to 1999 in Cameroon, Senegal, the Congo, and France. Regarding NNRTIs, one of five samples from participants infected with type J showed resistance, as did all four type O samples, compared to none of the type A, B, C, D, F, G, K, or recombinant types. Other than one participant infected with clade B, none of the others were resistant to NRTIs. No participants had primary PI mutations; all samples except those from participants infected with clade B and D had secondary PI mutations. Another study³⁹ reported on resistance among participants infected with clades A, C, D, F2, G, H, and J in addition to the dominant CRF02-AG clade. Data were collected from adults and adolescents in Cameroon from 2000 to 2002. All samples had secondary PI mutations. Tests for RT mutations were not reported.

Many of the studies conducted among individuals with highly prevalent HIV clades showed a high rate of secondary PI mutations in CFR02-AG. The same was true for HIV clade C. Rates of resistance to NNRTIs ranged from 0 percent to 7.7 percent in clade C-infected individuals and 0 percent to 2 percent in those infected with CRF02-AG. Resistance to NRTIs also ranged from 0 percent to 7.7 percent in clade C and from 0 percent to 5.1 percent in CRF02-AG.

In summary, rates of NNRTI resistance ranged from 0 percent to 7.7 percent in most HIV clades. Prevalence was higher in clades B and O in one study, but samples were too small to allow generalization to the population at large. NRTI resistance rates ranged from 0 percent to 8 percent for all clades. Primary PI mutations were rare (<3 percent) among Africans. There were high levels of secondary PI mutations reported in most studies; there is no evidence that these affect resistance in the absence of primary mutations. It was unclear how some studies selected persons for genetic testing and most studies had small samples. Therefore, the generalizability of the data is limited.

Overview

We identified 31 studies in developing countries that addressed the emergence of ARV resistance in women subsequent to the administration of ARV therapy to prevent mother to child transmission (pMTCT) of HIV-1 infection. These studies were heterogeneous in terms of the duration of ARV treatment, the selection of the ARV regimen, the timing and frequency with which ARV resistance tests were done, the sensitivity of the assay for ARV resistance, and whether clade-specific data were reported. Many studies were reported only as conference abstracts; these studies tend to include less information than articles in peer-reviewed journals.

In this section, we review the data on the prevalence of resistance in this group according to the prophylactic regimen, the infecting viral clade, the timing of the resistance assay (i.e., how long after the cessation of ARV therapy), and the assay that was used to detect ARV resistance.

Resistance and pMTCT regimen

Single dose Nevirapine (SD-NVP) alone. The greatest number of evaluable reports studied the impact of SD-NVP alone on the emergence of resistance (N= 14). Several of these reports provided data derived from the HIV-1NET 012 trial, which left three evaluable reports that provided non-overlapping data from the study,⁴⁷–⁴⁹ two reports that provided non-overlapping data from one South African study, ^{50 51} and two reports that provided non-overlapping data from another South African study.^{52 53} All studies were done in sub-Saharan Africa.

The earliest time that resistance was assessed was 7 days postpartum.⁴⁷ Most studies assessed resistance 6 – 8 weeks postpartum (N = 10). Two studies evaluated the prevalence of resistance approximately 6 months postpartum. Eleven of the fourteen reports provided information by viral clade and three did not. The rate of NNRTI resistance 6 to 8 weeks post SD-NVP ranged from 8.5 percent to 69.2 percent.

Table 5 shows the results of standard genotypic resistance assays in the main evaluable studies. Studies that address specific questions in subpopulations of the main study (e.g., data comparing the results of highly sensitive assays for the detection of ARV resistance with standard genotypic resistance assays or studies of resistance over time¹³) are not included in this table.

Analysis of SD-NVP resistance data by clade. Thirteen studies of SD-NVP contrasted the emergence of ARV resistance in mothers infected by differing clades of HIV-1. Aside from clade C, for which 8 reports provided information regarding the emergence of resistance in seven clinical trials (data from one South African study of persons infected by clade C were provided in two reports^{52 53}), no other clade was analyzed in more than two clinical trials.

As shown in Table 6, the reported prevalence of NNRTI-resistance as assessed by standard genotypic resistance tests, is highly variable, ranging from over 70 percent at 2 weeks postpartum^{54 103} to 8.5 percent at 6 weeks postpartum in one study performed in South African women.⁵⁶ Many studies showed a decline in resistance over time; this phenomenon will be discussed in more detail in a separate section.

Following SD-NVP, the prevalence of ARV resistance in women infected by HIV clades other than clade C ranged from 16 percent (6 weeks postpartum, clade A⁶⁰) to 36 percent (6 – 8 weeks postpartum, clade D⁴⁷). A summary of results from studies of ARV resistance following administration of SD-NVP for pMTCT to women infected by other clades of HIV-1 is provided in Table 7.

One formal cross-clade analysis of the prevalence of ARV resistance in female recipients of SD-NVP was identified. Using the results of the HIV-1NET 012 and NVAZ trials, Eshleman demonstrated that the proportion of women with clade C with at least one NVP resistance mutation (69 percent) 6 – 8 weeks postpartum was significantly higher than found in women infected by clade A (19 percent, p < 0.0001) or clade D (36 percent, p < 0.001).⁴⁹ In the HIV-1NET 012 trial, nevirapine resistance was greater in clade A than clade D.

Specific NNRTI amino acid changes associated with resistance in persons receiving SD-NVP. Complex patterns of NNRTI resistance were observed following the receipt of SD-NVP. The most common mutations that resulted in amino acid changes were K103N (lysine to asparagine at position 103) and Y181C (tyrosine to cysteine). Y181C was the most commonly detected resistance mutation in plasma collected 2 weeks postpartum from women infected by HIV-1 clade C.^{54, 55} In contrast, the K103N resistance mutation was most common at later time points in persons infected by clade C and at all times in persons infected by other HIV-1 clades. Among clade-C women from South Africa, V106/A/M (valine to alanine or methionine) was the most common amino acid change.⁵⁴ One report found that the rates of detection of K103N, Y181C, and Y188C were significantly higher in women infected by HIV-1 clade C than by HIV-1 clades A or D.⁴⁹ Still, there has not been enough research to assess whether there are clade-specific differences in the prevalence of resistance.

The findings from studies that provided data on the prevalence of various NNRTI resistance mutations are displayed in Table 8.

Effect of single-dose nevirapine on resistance over time. A number of considerations suggest that the prevalence of readily detectable ARV resistance, particularly resistance that results from the use of NNRTIs, may vary after the discontinuation of ARV administration. First, significant serum concentrations of NVP may persist for prolonged periods of time after a single dose, leading to delayed (i.e., beyond 7 days) emergence of resistance. The prolonged presence of subtherapeutic serum concentrations may lead to ongoing selection of mutants for 2 to 3 weeks. Conversely, if resistant virus is less fit than wild type virus, the prevalence of readily detectable resistant virus may gradually decline. Therefore, resistance testing done immediately after the administration of single dose NVP may underestimate the subsequent emergence of NNRTI-resistance due to this treatment strategy, whereas delayed ARV resistance testing may underestimate the prevalence of previously detectable resistance.

Several studies, using a variety of assays including very sensitive ones, assessed the presence of NNRTI resistance over time among women who took SD-NVP to prevent mother-to-child transmission of HIV. Study data are presented in Table 9.

Eshleman⁴⁸ conducted genotypic testing on participants in randomized trial HIV-1 NET 012 at 7 days postpartum. She found that 24 percent of women infected with HIV clade A had NNRTI resistance, compared to 19 percent of women infected with clade D. The same women were tested at 6 to 8 weeks postpartum. At that point, 28 percent of clade A women had evidence of resistance, compared to 42 percent of clade D women. The authors reported rapid fading of variants with Y181C in clade A but not in clade D. In the developed world, studies in individuals infected with clade B have shown that Y181C confers high-level resistance to NVP and impairs viral fitness. These results suggest that Y181C may have less effect on viral fitness of clade D than on fitness of clade A. Eshleman⁴⁷ also did follow up on 11 women who had resistance at the 6- to 8-week time point. When genotyping was conducted between one and two years postpartum; none of the women had evidence of NNRTI resistance.

Flys¹⁰⁴ also reported recently on women from HIV-1 NET 012. (It is unclear how her sample overlaps with those in the previous publications.) She reported that 34.1 percent of 88 clade A participants had the K103N mutation at 6 to 8 weeks, compared to 53.6 percent of 56 clade D participants. Using follow-up tests and survival analysis, she estimated that the prevalence of the mutation declined to 3.6 percent in clade A and 29.8 percent in clade D at 2 years. By 4 years, she estimated that no clade A participants and only 2.4 percent of clade D participants had the mutation.

The prevalence and decline of the most common NVP mutant, K103N, were also examined using allele-specific PCR in 65 women infected with HIV clade C in South Africa⁵⁸ after SD-NVP. The rate of resistance, assessed through a highly sensitive assay, declined steadily over 1 year, from 87 percent at 6 weeks post-partum to only 11 percent at 1 year, according to RT-PCR of plasma RNA. PCR was used to detect cellular proviral DNA; these results were used to generate estimated rates of resistance of 52 percent at 6 weeks and only 4.2 percent at 1 year. Of note, the number of women available for testing at each follow-up ranged from 26 to 53 of the original 65. Samples were available at all time points for only 16 women.

We also found two conference abstracts that assessed the association of resistance with time following SD-NVP in a pMTCT cohort. Martinson⁵² followed up on 456 women; all but four were infected with HIV clade C. NNRTI resistance was defined as any of the following mutations: K103N, V106 A/M, Y181C, Y188C, and G190A. It appears that each woman was tested once for resistance within 36 weeks of birth. Compared to women tested within 10 weeks postpartum, far fewer women who were tested between 10 and 36 weeks showed evidence of resistance (44 percent versus 24 percent, respectively). Baseline CD4 cell count, viral load, time from birth to resistance testing, and number of times a mother took NVP during pregnancy were significantly associated with development of resistance. Martinson also measured resistance in infants;⁶² the results of this assessment are discussed below in the section on infants.

Kantor⁵⁵ assessed resistance over time in 34 women in Zimbabwe who received SD-NVP. All were infected with HIV clade C. NNRTI resistance rates dropped from 75 percent at two weeks postpartum to 34 percent at 8 weeks and 13 percent at 24 weeks. (The number of women available for testing dropped from 32 at eight weeks to only eight at 24 weeks, limiting the validity of the findings.) Prevalence of the Y181C mutation dropped from 57 percent at two weeks to six percent at week eight, while the K103N mutation was present in 25 percent of women at 2 weeks and 28 percent at 8 weeks. Similarly, Loubser⁵⁴ studied a group of women in south Africa who had received SD-NVP; all but two were infected by HIV clade C. These investigators found that the prevalence of Y181C decreased from 49 percent at week 2 (N = 35) to 23 percent at week 6 postpartum (N = 164). The prevalence of the K103N mutation was 29 percent at 2 weeks and 31 percent at 6 weeks.

In sum, evidence provided by longitudinal analyses suggests that the prevalence of detectable NNRTI resistance may vary with time since treatment. Fading of variants with Y181C were reported in clade A but not in clade D. Prior studies show that Y181C confers high-level resistance to NVP and impairs viral fitness in persons infected by HIV clade B. Thus it is possible that Y181C may have less effect on viral fitness of clade D than of A. This observation warrants further investigation in future studies.

Other pMTCT regimens, including therapy pre and post SD-NVP. Ten studies from the developing world were identified that studied the emergence of ARV resistance following other pMTCT regimens, including adding ARV therapy before or after SD-NVP. Two studies were done in Thailand,^{63, 64} one was conducted in Argentina,⁶⁵ and the other eight studies were done in Sub-Saharan Africa. The findings are presented in Table 9.

In women using NRTIs alone,^{60, 63, 66, 67} the rates of NRTI resistance ranged from zero to four percent. Rates of HIV transmission to infants varied by regimen. In a study of pregnant women taking ZDV and DDI from 36 weeks gestation, 5.4% of the infants were HIV positive. In two studies of ZDV alone, transmission rates were 16.5% and 16.7%. (In the former study, women were breastfeeding.) Another study of ZDV alone did not report transmission rates.

An RCT conducted in the U.S. and France (not displayed) assessed whether NVP resistance would occur after intrapartum NVP (or placebo) in women receiving standard ARV treatment during pregnancy.¹⁰⁵ The study included mostly (62 percent) non clade B women, who were of African origin. 15 percent of the women in the NVP group developed a new NVP mutation at 6 weeks postpartum. The development of a new NVP mutation did not correlate with CD4 cell count or viral load at delivery, or with the type of ARV after delivery. Similar trials were conducted in the developing world; rates of 32 percent and 33 percent for NNRTI resistance were observed in two trials using ZDV monotherapy in the pre-partum period followed by SD-NVP during labor.⁶⁹ In another study⁷⁰ 6.3 percent of women who took ARV followed by SD-NVP showed NVP resistance at a mean of 48 days. A very recent study by Shapiro⁷¹ randomized pregnant women on ZDV-based ARV to either SD-NVP or placebo at birth. 45 percent of the SD-NVP group had NVP resistance at one month post-partum. Importantly, the difference in HIV transmission from mother to child was insignificant between the SD-NVP and placebo groups, suggesting that SD-NVP may be unnecessary in the context of ARV treatment of pregnant women

The World Health Organization (WHO)'s revised guidelines recommend a combination of ZDV and lamivudine (3TC) at the “tail” end (post-partum) be used to reduce the risk of resistance from NVP.¹⁰⁶ We found two trials that support these new guidelines. One prospective randomized trial compared the prevalence of NVP resistance in persons given SD-NVP followed by 0, 4 or 7 days of ZDV plus 3TC.⁴⁶ The rates of NNRTI resistance in these three groups, as detected by standard genotypic resistance assays, were 57 percent, 13 percent, and 9 percent respectively. Similarly, in a non-controlled trial, Chaix et al. found that the rates of NNRTI resistance were less when SD-NVP was followed by the administration of ZDV plus 3TC for 3 days post-partum than had been observed in previous trials of SD-NVP alone.⁶⁸

Recent reports from sub-Saharan Africa indicate that risk of virologic failure (the failure to achieve measurable decreases in serum viral load after treatment) in women initiating ARV after SD-NVP is associated with the interval between receipt of SD-NVP and initiation of full ARV. Lockman⁷² showed that viral response was similar between women who had SD-NVP and those who received no SD-NVP (placebo) if they started treatment more than 6 months after SD-NVP, while viral failure was more common with SD-NVP if ARV was started less than 6 months after exposure. The CD4 response was the same with or without SD-NVP, regardless of time of initiation. Data from Coovadia,¹⁰⁷ where women started ARV a median of 2 years after SD-NVP exposure, found no difference in viral response between women with SD-NVP exposure and those without. Chi⁷³ also found no difference between SD-NVP and no SD-NVP in viral response or CD4 count response. In this study, median time from SD-NVP to ARV initiation was 503 days; 81% started ARV at least six months post-partum. Additionally, several abstracts have found that SD-NVP is as effective in second pregnancy as in first pregnancy. ^{108 109} None of these abstracts reported resistance rates. We strongly suggest that future studies of this nature include resistance testing.

There are currently two RCTs funded by the NIH that will assess response to NNRTI- and PI-based therapy in persons with and without prior SD-NVP exposure. These trials, one in women and one in infants, will add significantly to our body of knowledge on this issue.

Measuring the presence of resistance using standard vs. more-sensitive techniques. HIV resistance is usually assessed by performing population-based genetic sequencing of HIV RNA collected from virus found in plasma and determining the presence of specific mutations. These assays are insensitive to minor viral sub-populations that account for less than 20 – 30 percent of circulating virus. In contrast, by the use of specific genetic probes, RT-PCR, allele-specific PCR, or multiple single genomes, it is possible to detect viral subpopulations that constitute as little as 0.1 – 0.2 percent of circulating virus in plasma. Other, more sensitive, assays include tests for the presence of HIV proviral DNA in mononuclear cells.

As shown in Table 11, the rates of NNRTI resistance are consistently greater when more-sensitive assays are employed.^{51, 58, 65, 74} For example, Eshleman first reported rates of NNRTI resistance of 69.2 percent, 36.1 percent, and 19.4 percent, respectively for women infected with clade C, D, and A, 6 to 8 weeks after SD-NVP.⁴⁹ According to Flys,¹¹⁰ Lig Amp assays on the same participants showed that in reality, 69.8 percent, 54.3 percent, and 41.7 percent of women with clade C, D, and A, respectively, had the K103N resistance mutation.

Some of the studies also show the quantitative frequency of viruses with resistance as part of the quasispecies (closely related but non-identical mutant and recombinant viral genomes subjected to continuous genetic variation, competition, and selection; thought to be the mechanism by which RNA viruses persist).¹¹¹ Loubser, 2006⁵⁸ showed that the actual frequency of resistant viruses is low and decreases further over time. Twenty seven of 31 women had the K103 N mutation at 6 weeks postpartum, but the median proportion of the virus that contained the K103N variant was 11 percent. At 3 months, in women who still had detectable resistance, the frequency was 6 percent; at 7 months, frequency was 1.2 percent; and at 12 months, the frequency was 0.7 percent. Thus, it is likely that time following SD-NVP exposure is very important in determining response to additional ARV therapy.

Resistance in HIV-positive infants born to mothers receiving pMTCT. Children are often born HIV positive despite pMTCT. Resistant viruses can be transmitted or can emerge independently in either mother or infant. Eight studies^{50, 54, 55, 57, 60, 62, 69, 75} evaluated the emergence of ARV resistance in children born to mothers in sub-Saharan Africa who received pMTCT. Most studies examined the impact of SD-NVP. Three studies reported on participants infected by HIV clade C; one other combined data from those with clades A, D, C, and recombinant types. The other studies did not report clade.

We also found a study conducted in Thailand that reported 12-month postpartum resistance rates in children who, themselves, received ART for unspecified lengths of time.⁶³ As the resistance could not be attributed to pMTCT alone, and specific pMTCT regimens were not mentioned, this study will not be discussed.

In most studies, HIV-infected infants were tested for resistance at 6 or 7 weeks of age. Rates of NNRTI resistance among untreated infants of mothers receiving only SD-NVP ranged from 36 percent to 50 percent. The most common mutations were Y181C and K103N.

Martinson⁶² showed that detectable resistance fades over time in infants exposed to SD-NVP. Of the 53 HIV-positive infants studied, 45.3 percent had NNRTI resistance at their first visit (range 4 to 12 weeks). Those with resistance were tested again every 6 months. At 18 months, only one still had the Y181C mutation.

Troyer⁶⁰ compared resistance rates among mother-infant pairs infected with various HIV clades, who were treated with ZDV (N = 48) or SD-NVP (N = 61). The HIV transmission rates were similar in the groups: 16.4 percent and 16.7 percent respectively. Using a sensitive assay for ARV resistance (oligonucleotide DNA ligation assay) at 6 weeks, NNRTI mutations were found in half of the infected pairs who received SD-NVP. ZDV mutations were not present in pairs treated with that drug. Resistance data for mothers and infants were not presented separately.

Chaix⁶⁹ gave mothers ZDV (300mg) twice daily starting at 36 weeks gestation in addition to oral SD-NVP at onset of labor. Only 23 percent of their HIV-positive children were resistant to NRTIs at 4 weeks postpartum. McIntyre⁵⁰ compared resistance among infants who received SD-NVP with those who received SD-NVP plus either 4 or 7 days of ZDV + 3TC after birth. Infected infants who received the additional 7-day ZDV + 3TC showed no resistance at 6 weeks, compared to 78 percent of those who received SD-NVP only.

Finally, Eshleman⁷⁵ recently compared resistance rates in infants who received various combinations of intrapartum SD-NVP, SD-NVP (2 mg/kg) immediately postpartum, and ZDV (4 mg/kg) twice daily for one week. At 6 to 8 weeks, the rate of resistance was much lower (27 percent) in the infants who received the postpartum SD-NVP and ZDVwithout the intrapartum SD-NVP. The Eshleman study also showed that when the infant received ZDV, the transmission rate did not differ between mothers who received SD-NVP and those who did not. When the baby received SD-NVP plus one week of ZDV and the mother got SD-NVP, transmission was 17 percent and resistance was seen in 74 percent; when the baby got SD-NVP plus 1 week of ZDV and the mother did not get SD-NVP, transmission was 17 percent but resistance was only 27 percent. Data from the MASHI trial also suggest that when the infant receives SD-NVP at birth in the background of ZDV, maternal SD NVP does not provide any additional protection.⁷² As mentioned earlier in this report, Shapiro⁷¹ randomized pregnant women on ZDV-based treatment to either SD-NVP or placebo at birth; the difference in HIV MTCT between the SD-NVP and placebo groups was insignificant at one month.

These findings imply that research should be conducted into the need for maternal SD-NVP when an infant gets NVP at birth, or when the mother receives ZDV regularly before birth.

Predictors of resistance. Several studies evaluated the predictors of ARV resistance among persons receiving SD-NVP. In the most comprehensive analysis, a multivariate analysis performed by Eshleman and colleagues demonstrated an increased risk of resistance detectable by standard genotyping assays in persons infected with clade C vs. clade A or clade D, or increased viral load at delivery, but not with increased age or number of births.⁴⁹ Similarly, Loubser found that an increased viral load was associated with an increased rate of resistance.⁵⁴ Jourdain also showed that plasma HIV RNA level was predictive of likelihood of ARV failure and NPV resistance.⁶⁴

Chaix and colleagues found that increased duration of prepartum use of ZDV+3TC for pMTCT was associated with an increased maternal prevalence of the M184V resistance mutation.⁷⁶ Another study⁵⁰ found a decrease in the rate of NNRTI resistance with longer duration of ZDV+3TC post-partum. In both studies, SD-NVP was given during birth.

The role of adherence in resistance is discussed in the next section.