Share this article on:

Social Network–Based Recruitment Successfully Reveals HIV-1 Transmission Networks Among High-Risk Individuals in El Salvador

Dennis, Ann M. MD*; Murillo, Wendy PhD; de Maria Hernandez, Flor MD, MPH; Guardado, Maria Elena MD, MPH§; Nieto, Ana Isabel MD; Lorenzana de Rivera, Ivette MS; Eron, Joseph J. MD*; Paz-Bailey, Gabriela MD, MSc, PhD‡,§

JAIDS Journal of Acquired Immune Deficiency Syndromes: May 1st, 2013 - Volume 63 - Issue 1 - p 135–141
doi: 10.1097/QAI.0b013e318288b246
Epidemiology and Prevention

Objective: HIV in Central America is concentrated among certain groups such as men who have sex with men (MSM) and female sex workers (FSWs). We compared social recruitment chains and HIV transmission clusters from 699 MSM and 787 FSWs to better understand factors contributing to ongoing HIV transmission in El Salvador.

Methods: Phylogenies were reconstructed using pol sequences from 119 HIV-positive individuals recruited by respondent-driven sampling (RDS) and compared with RDS chains in 3 cities in El Salvador. Transmission clusters with a mean pairwise genetic distance ≤0.015 and Bayesian posterior probabilities =1 were identified. Factors associated with cluster membership were evaluated among MSM.

Results: Sequences from 34 (43%) MSM and 4 (10%) FSW grouped in 14 transmission clusters. Clusters were defined by risk group (12 MSM clusters) and geographic residence (only 1 spanned separate cities). In 4 MSM clusters (all n = 2), individuals were also members of the same RDS chain, but only 2 had members directly linked through recruitment. All large clusters (n ≥ 3) spanned >1 RDS chain. Among MSM, factors independently associated with cluster membership included recent infection by BED assay (P = 0.02), sex with stable male partners (P = 0.02), and sex with ≥3 male partners in the past year (P = 0.04).

Conclusions: We found few HIV transmissions corresponding directly with the social recruitment. However, we identified clustering in nearly one-half of MSM suggesting that RDS recruitment was indirectly but successfully uncovering transmission networks, particularly among recent infections. Interrogating RDS chains with phylogenetic analyses may help refine methods for identifying transmission clusters.

*Division of Infectious Diseases, University of North Carolina, Chapel Hill, NC;

Department of Microbiology, National Autonomous University of Honduras, Tegucigalpa, Honduras;

Center for Health Studies - HIV Surveillance, Del Valle University of Guatemala, Guatemala City, Guatemala;

§Tephinet, Inc, Atlanta, GA; and

National HIV Program, Ministry of Health, San Salvador, El Salvador.

Correspondence to: Ann M. Dennis, Division of Infectious Diseases, 130 Mason Farm Road, Suite 2115, CB# 7030, Chapel Hill, NC 27599-7030 (e-mail:

Supported by the National Center for Research Resources and the National Center for Advancing Translational Sciences, National Institutes of Health, through Grant KL2TR000084, Sida/SAREC on behalf of a bilateral collaboration with the National Autonomous University of Honduras, the Network for Research and Training in Tropical Diseases in Central America (NeTropica) under the project No 06-R-2010, the Training Programs in Epidemiology and Public Health Interventions Network (TEPHINET) through a cooperative agreement (#6D43GH000014-05) from the Centers for Disease Control and Prevention (CDC) and the United States Agency for International Development (USAID).

Parts of this work were presented at the 19th Conference on Retroviruses and Opportunistic Infections (CROI), March 2012, Seattle, WA (Abstract #F-148).

The authors have no conflicts of interest to disclose.

Received September 05, 2012

Accepted January 18, 2013

Back to Top | Article Outline


Despite considerable improvements in access to testing and treatment for HIV/AIDS throughout Central America over the past decade, targeted prevention efforts among high-transmission settings remain at low-levels.1 Prevention in these settings, particularly among men who have sex with men (MSM) and female sex workers (FSWs) is hampered by the continued stigma and discrimination surrounding HIV among the general population. In El Salvador, although the HIV prevalence is low overall (<1% in the general population), a significant concentrated epidemic was documented in 2002 among MSM (15.5%) and FSWs (3.2%).2 Repeat surveillance in 2008 revealed little change in HIV prevalence among these high-risk groups.3

Recently, more innovative recruitment methods have been employed in El Salvador to reach high-risk groups. Respondent-driven sampling (RDS), a chain referral method often used to recruit hard-to-reach or “hidden” populations,4 has been successfully implemented for HIV biological and behavioral surveillance in international settings.5 In 2008, RDS was implemented in El Salvador to estimate the size of the MSM and FSW population,6 assess HIV and sexually transmitted infection (STI) prevalence rates, and to gain information on high-risk behaviors in these groups.3 Although many studies have utilized RDS for HIV surveillance, few have integrated molecular data from HIV-positive participants to better understand relationships between social recruitment chains and HIV transmission networks.7,8 Typically, in RDS, recruitment is performed from social rather than from sexual networks. However, the RDS social recruitment chain takes advantage of potential networks where HIV and other STIs are transmitted. HIV-1 sequence data can be exploited to reconstruct transmission networks based on sequence similarity and inference of common ancestors using phylogenetic methods.9

A better understanding of the relationship between social recruitment and groups of individuals who share related HIV strains (transmission clusters) may assist in the design of intervention programs including the assessment or refinement of recruitment methods. The objective of this study was to reconstruct transmission networks among HIV-positive MSM and FSWs in El Salvador who were recruited by RDS and compare the relationship between these networks to their RDS recruitment chains. Clinical and demographic data were evaluated to assess factors associated with membership in HIV transmission clusters among MSM.

Back to Top | Article Outline


Study Design and Population

We conducted a cross-sectional analysis of HIV-positive individuals who were recruited through RDS in El Salvador during 2008 and had a specimen available for genotypic sequencing. From March to September 2008, MSM and FSWs residing in 3 cities in El Salvador were recruited during an integrated biological and behavioral survey implemented to assess HIV/STI prevalence rates and to collect behavioral data. Recruitment and data collection methods for this survey, the Encuesta Centroamericano de Vigilancia de VIH y Comportamiento en Poblaciones Vulnerables, have been previously described in detail.3 MSM were men who had anal sex with another man in the past 12 months; they were recruited in San Salvador and San Miguel. FSWs were women who reported exchanging sex for money in the past 30 days; they were recruited in San Salvador and Sonsonate. Individuals in both groups were 18 years or older, signed informed consent, and also lived and worked in the city in which they were recruited.

To initiate RDS recruitment of peers, various “seeds” diversified with respect to age, self-identification of sexual orientation, and income levels were located through local organizations working with the populations. Eleven seeds were used in San Salvador and 5 in San Miguel for MSM. Ten seeds were selected in San Salvador and 5 in Sonsonate for FSW. The participants were asked to recruit 3 other individuals from their social network or peers, and each new recruit was asked to continue this process, or “waves” of recruitment. Recruits who presented a valid coupon to a study site were screened for eligibility and provided information about the study. They were then asked to provide written informed consent. They responded to questions in a computer-based and in face-to-face interview and were asked to submit blood, urine, and anal or vaginal swab samples. The participants were not directly asked about sexual contact with the individuals they recruited. All the participants received promotional materials, condoms, and a choice of other incentives (T-shirt, towel, make-up kit, lubricants, and additional condoms) for each recruited peer they referred (up to 3). Recruitment was tracked using RDS coupon numbers; names were not collected.

HIV testing was conducted with 2 rapid HIV tests (Determine HIV-1/2 and OraQuick Rapid HIV-1 Antibody Test); positive results were confirmed with an HIV enzyme-linked immunosorbent assay and/or Western blot. All reactive sera were analyzed with the BED test HIV-1 EIA capture assay (BED-CEIA; Calypte Biomedical Corp, Lake Oswego, OR) to estimate the probable duration of infection.10,11 Recent infections were defined as a confirmatory OD-n of <0.8 on the BED-CEIA, indicative of a mean seroconversion period of 162 days (95% confidence limit, 146–179 days).12 Patients were also screened for STIs using polymerase chain reaction for Trichomonas vaginalis, Mycoplasma genitalium, Neisseria gonorrhoeae, and Chlamydia trachomatis. Active syphilis was defined as a reactive rapid plasma reagin >1:8 and positive Treponema pallidum hemagglutination assay.

Back to Top | Article Outline

HIV-1 Sequences and Phylogenetic Inference

Population-based sequencing of the full protease and partial reverse transcriptase gene was performed by previously described in-house methods.13,14 Sequences were aligned using Clustal X and edited manually with removal of gapped positions.15 Subtypes and recombination were assessed using Subtype Classification Using Evolutionary Algorithms (SCUEAL).16 Subtyping was confirmed with phylogenetic analysis using Mega v4 and the neighbor-joining method with general time-reversible–Γ estimated distances and 1000 bootstrap replicates.17 The sequences most closely related to each one of our sequences were downloaded from the Los Alamos National Laboratory HIV database ( using the Basic Local Alignment Search Tool (BLAST). At the time of our search in 2011, no sequences from El Salvador were deposited in the Los Alamos National Laboratory. The BLAST search generated 30 unique sequences, and these were incorporated in the phylogenetic analyses. Additionally, 4 more HIV-1 B-subtype references were included, accession numbers: IT_AF251949, FR_L31963, FR_D86069, and B.FR.83.HX.

Potential transmission clusters involving highly related sequences were first identified as clades with ≥2 sequences on the neighbor-joining tree with bootstrap support ≥90% and mean intracluster pairwise genetic distance ≤0.015 nucleotide substitutions per site. Neighbor-joining trees were reconstructed in MEGA 4.0 with the Tamura–Nei model.17 Clusters were then confirmed using Bayesian methods in MrBayes18 using the general time reversible–Γ model with a proportion of invariant sites. The model was run to 40 × 106 generations, sampling every 1000th generation, and with a 10% burn-in. Convergence of the estimates was evaluated with generation vs log probability plots in Tracer v.1.519 using an Effective Sample Size >200. A maximum clade credibility tree was generated using TreeAnnotator in BEAST.20 Clusters with posterior probabilities =1 were considered robust. Analyses were also repeated after removing sites associated with major drug-resistance mutations to avoid theoretic convergent evolution, which showed similar topology (data not shown). Mutations associated with transmitted drug resistance were identified using the 2009 standardized surveillance list from the World Health Organization.21

Back to Top | Article Outline

Social Network and Statistical Analyses

Social network data were visualized using NetDraw (Analytic Technologies, Lexington, KY) to visualize recruitment waves among participants originating from each seed. The outcome of interest was membership in phylogenetic clusters. Descriptive statistics were used to assess the relationship between the RDS recruitment chains and membership in phylogenetic clusters. Bivariate associations between characteristics of interest including demographics, prevalence of STIs and selected behaviors among MSM in clusters compared with those not in clusters were obtained using χ2 for categorical variables and Kruskal–Wallis for continuous variables. Independent predictors of membership in MSM clusters were assessed using log-linear binomial regression. Prevalence ratios (PRs) were calculated instead of odds ratios to avoid overestimation of the relative risk because our outcome was common (>10%).22 Data were analyzed with Stata version 11.0 (StataCorp, College Station, TX). All analyses were done unweighted without adjusting for RDS recruitment, and therefore findings are not generalizable beyond the sample.

Back to Top | Article Outline


Characteristics Study Population

In total, 824 MSM and 848 FSWs were recruited during the RDS in 2008. Of those participants consenting to HIV testing, 95/699 (13.6%) of MSM and 46/787 (5.8%) of FSWs tested HIV-1 positive. Among these 141 HIV-positive participants, 119 (84%) had specimens available for genotypic sequencing. The participants who did not have sequences were similar to those with sequences by risk group (66% vs 67% MSM) and residence (83% vs 81% from San Salvador). Individuals without sequences seemed to be randomly distributed in the recruitment chains; approximately one-half of the chains included at least 1 individual without a sequence.

Of the 119 participants with sequences, 90 (67%) were MSM and 39 (33%) were FSWs. All sequences were of subtype B and have been deposited in GenBank (Accession # KC412724-KC412842). The majority of both risk groups were recruited in San Salvador; however, the groups differed by several characteristics (Table 1). MSM were more likely to be younger {median 24 years [interquartile ratio (IQR) 21–29] vs 32 [IQR 26–38] for FSW} and have recent infection (24% vs 0%) than the FSW. Rates of STIs at the time of recruitment were higher among FSW (59% vs 23% MSM), likely due to high rates of M. genitalium and Trichomonas. However, only MSM had active syphilis (11%). Only 2 participants reported the use of antiretroviral therapy (ART). The prevalence of transmitted drug resistance was similar (9% MSM vs 10% FSW) with K103N being the most frequent mutation (6 of the 11 sequences with drug resistance).



Back to Top | Article Outline

Identification and Characteristics of Phylogenetic Transmission Clusters

We identified 14 transmission clusters involving 38 (32%) sequences on the neighbor-joining tree with >90% bootstrap and ≤1.5% mean intracluster genetic distance. All clusters remained robust with high support of the nodes in the Bayesian analysis (Fig. 1). The mean cluster size was 3 members (range 2–6) with 72% in dyads. Median intracluster genetic distance was low at 0.005 substitutions per site (IQR 0.003–0.012), indicating that sequences were highly related. Clusters were defined by risk group and recruitment city (Table 2). Not surprisingly, MSM were more likely to be in clusters (43% vs 10% FSW; P < 0.001). However, 4 sequences from FSW formed 2 separate cluster dyads and 1 dyad spanned 2 cities. Among MSM, similar rates of cluster membership were seen between men recruited in San Salvador (44%) and San Miguel (35%) though no clusters spanned both cities. Although no sequences from both MSM and FSWs clustered together at the 1.5% genetic distance cutoff, more distantly related clusters are noted where both risk groups share a robust common ancestor (Fig. 1).





Back to Top | Article Outline

Comparison to Social Recruitment Chains

In the RDS recruitment chains, 12/16 (75%) MSM and 13/15 (87%) of the FSW seeds recruited HIV-positive individuals. Of these, all 12 MSM and 3 of the FSW recruitment chains included at least 1 participant who was also a member of a transmission cluster. Recruitment in an earlier wave of the chain may indicate more network ties compared with recruitment in later waves, which may be expected for members of transmission clusters. However, the median number of waves until recruitment into the sample was similar for HIV-positive individuals who were members of the transmission clusters compared with nonmembers [5 waves (IQR 3–9) vs 6 (IQR 4–9); P = 0.73]. Among the 14 transmission clusters (labeled A–N), half included at least 2 members recruited from the same social chain, and 4 MSM dyads were composed of members of the same chain (Table 2). However, the 2 members were directly linked through the RDS recruitment in only 2 of these dyads. The most successful RDS chain at identifying clusters had 65% (11 of 17) HIV-positive MSM who were members of clusters; however, these spanned 6 discrete clusters (Fig. 2). Overall, participants with recent infection tended to be in clusters that included members of their same RDS chain compared to those with chronic infection (67% vs 35%, respectively; P = 0.07). Finally, to assess whether RDS tended to enrich the network sample for MSM clusters, we evaluated the association between clustering and the participants reported personal network size (number of MSM the participant estimates to know). Men who were cluster members were significantly more likely to report larger networks than nonmembers [median 13 persons (IQR 7–20) vs 5 (IQR 3–10); P = 0.006].



Back to Top | Article Outline

Characteristics Associated With Cluster Membership Among MSM

Among MSM, 43% were members of phylogenetic clusters. On bivariable analysis, these patients were significantly more likely to be younger, have recent infection, and report any sex (protected or unprotected) with a stable male partner over the past year (Table 3). In the multivariable regression analysis (adjusted for variables P < 0.1 in the bivariable analysis), recent infection {PR 1.59 [95% confidence interval (CI) 1.07–2.34], P = 0.02]}, sex with a stable male partner [PR 1.69 (95% CI 1.08–2.54), P = 0.02], and sex with ≥3 male partners in the past year [PR 1.70 (95% CI 1.02–2.80; P = 0.04)] were independently associated with cluster membership. No associations were found with illicit drug use and, of note, injection drug use was uncommon with only 1 MSM reporting injection of cocaine or heroin in the past year.



Back to Top | Article Outline


Our study demonstrates that RDS not only identified many HIV-positive individuals but by using phylogenetic analysis also it provides an important insight into local HIV transmission in El Salvador that would not have been apparent from participant history alone. HIV sequences from one-third of the study population grouped into 14 highly related transmission clusters using stringent criteria including short genetic distance and Bayesian posterior probabilities. For MSM, 44% were found in clusters, indicating that RDS recruitment was successfully identifying transmission networks. Cluster members were more likely to be recently infected and report both stable partners and a higher number of partners than men who were not in clusters. Altogether, this combination of social and molecular data could help target prevention interventions for these high-risk groups and assist in the refinement of social recruitment methods to better enable detection of transmission networks.

To our knowledge, compared with previous studies, we have analyzed phylogenetically the largest number of HIV sequences with RDS chains to date. We found very few individuals who were members of the same HIV transmission cluster and who were directly linked in the social recruitment chains. These results are similar to the few other studies that compared RDS chains to HIV7,8 and/or hepatitis C viral phylogenies.8,23 Among 18 HIV-positive MSM recruited in Croatia, no direct linkage with RDS recruitment and genetic relatedness was found.7 Similarly, among injection drug users recruited in Canada, very few of the 18 HIV-positive individuals with clustered sequences were directly linked in the RDS chains.8 The discordance between direct social recruitment and genetic relatedness studies that use HIV-1 phylogenetics to interrogate RDS chains is likely due to imperfect alignment of social and sexual/transmission networks. Social networks are dynamic—the RDS represents a snapshot of the current social network, whereas phylogenetic clusters represent historical transmission networks. RDS studies also limit the number of recruits for each participant, and usually recruits are not restricted to sexual partners, but to any social contact such as in our study. These limitations may result in a partial description of the social or sexual network. Unlike contact networks, the reported social or sexual connections between all the participants in the RDS recruitment are not collected; another limitation of our study which prevents us from evaluating the overlap between sexual and social networks. Thus, many participants in our study who were not directly connected in the RDS chain may have indeed been direct sexual contacts, as is indicated by the phylogenetic analysis. Limiting RDS recruitment to sexual partners rather than to social contacts may enhance the ability to find recruitment chains more closely aligned to phylogenetic clusters.

Nonetheless, the high degree of phylogenetic clustering seen among MSM in our study supports the view that RDS may be a choice method for penetrating the networks HIV-positive MSM. This is suggested by the finding that larger personal networks were associated with increased cluster membership among MSM. However, due to the lack of a comparison group in our study, it is unclear whether RDS would be superior to other recruitment methods at finding phylogenetic clusters. In our study, this method seems particularly beneficial in the case of identifying incident HIV infections as cluster members were more likely to be recently infected compared to nonmembers. The phylogenetic reconstruction of HIV sequences offers a unique view on the movement of the virus within the population, which provides an estimate of the transmission network structure. The fact that several highly related clusters were found shows that the population sampled is sexually networked and that ongoing transmission is occurring locally. However, there are several limitations to be noted when interpreting phylogenetic trees. The direction of transmission cannot be determined in these types of analyses and there may also be unrecognized third parties involved in the transmission chain. Transmission cluster members may go unrecognized through incomplete sampling because of unavailability of HIV sequences or if individuals are undiagnosed, refuse HIV testing, or are not linked to care. For example, the FSWs in our study were very unlikely to have sequences that clustered. This is expected because partners of FSWs were not included in the study, so half of each FSW transmission event is missing from the phylogenetic analyses. However, identifying clusters is possible where FSWs share similar transmission sources. We describe 2 such cases in our study, of which one cluster spanned cities of Sonsonate and San Salvador.

The HIV and STI prevalence among high-risk groups such as MSM in El Salvador, as in other Central American countries, remains unacceptably high. Nearly 14% of MSM recruited in the RDS were HIV-infected and many were likely recently infected. Most of these recent infections were among young men, a group of high concern for an expanding epidemic. Similar HIV prevalence rates have been reported among MSM in the surrounding countries of Guatemala (12.1%) and Honduras (12.4%).2 The high level of stigma surrounding both homosexuality and HIV in Latin America24,25 presents a significant barrier to reaching this group for prevention. Fortunately, ART coverage for eligible patients is high in El Salvador with 96% reported coverage in 2008.26 The prevalence of transmitted drug resistance in our sample is strikingly similar to developed countries,27,28 reflecting this high ART coverage. Altogether, this information further emphasizes the need to diagnose and link patients to clinical care (especially young men) to take advantage of the morbidity and mortality benefits of ART,29,30 decrease onward HIV transmission,31 and limit onward spread of transmitted drug resistance.

Better methods are needed to effectively control HIV transmission at the population level. For these methods to be successful, at-risk persons must to be identified either before or early in the course of infection to take advantage of interventions known to be successful at the individual level such as early ART, circumcision, and condom use. Social recruitment methods offer ways at finding those persons who may not only be at highest risk of infection, but may also be the hardest to reach for prevention campaigns. Integrating molecular tools, such as the phylogenetic reconstruction of HIV sequences, with social recruitment data can be a complimentary method to detect the transmission networks driving local epidemics. This integrated approach has the potential to be a powerful tool in assessing the success of different types of recruitment methods, including time location or venue-based recruitment, to detect HIV transmission networks.

Back to Top | Article Outline


The authors thank study participants, the Ministry of Health of El Salvador, and local organizations who assisted with the implementation of the Encuesta Centroamericano de Vigilancia de VIH y Comportamiento en Poblaciones Vulnerables (ECVC).

Back to Top | Article Outline


1. Chan M, Sidibe M, Lake A. Global HIV/AIDS response: epidemic update and health sector progress towards universal access—progress report 2011. Available at: Accessed July 7, 2012.
2. Soto RJ, Ghee AE, Nunez CA, et al.. Sentinel surveillance of sexually transmitted infections/HIV and risk behaviors in vulnerable populations in 5 Central American countries. J Acquir Immune Defic Syndr. 2007;46:101–111.
3. Guardado ME, Criswell J, Monterroso E, et al.. Encuesta Centroamericana de Vigilancia de Comportamiento Sexual y Prevalencia de VIH/ITS en Poblaciones Vulnerables, Hombres que Tienen Sexo con Hombres, Trabajadoras Sexuales y Personas con VIH, ECVC EL Salvador. San Salvador, El Salvador: Minisiterio de Salud de El Salvador, Universidad del Valle de Guatemala, USAID, CDC, PASMO; 2010.
4. Magnani R, Sabin K, Saidel T, et al.. Review of sampling hard-to-reach and hidden populations for HIV surveillance. AIDS. 2005;19(suppl 2):S67–S72.
5. Malekinejad M, Johnston LG, Kendall C, et al.. Using respondent-driven sampling methodology for HIV biological and behavioral surveillance in international settings: a systematic review. AIDS Behav. 2008;12(suppl):S105–S130.
6. Paz-Bailey G, Jacobson JO, Guardado ME, et al.. How many men who have sex with men and female sex workers live in El Salvador? Using respondent-driven sampling and capture–recapture to estimate population sizes. Sex Transm Infect. 2011;87:279–282.
7. Lepej SZ, Vrakela IB, Poljak M, et al.. Phylogenetic analysis of HIV sequences obtained in a respondent-driven sampling study of men who have sex with men. AIDS Res Hum Retroviruses. 2009;25:1335–1338.
8. Pilon R, Leonard L, Kim J, et al.. Transmission patterns of HIV and hepatitis C virus among networks of people who inject drugs. PLoS One. 2011;6:e22245.
9. Hue S, Clewley JP, Cane PA, et al.. HIV-1 pol gene variation is sufficient for reconstruction of transmissions in the era of antiretroviral therapy. AIDS. 2004;18:719–728.
10. Parekh BS, Kennedy MS, Dobbs T, et al.. Quantitative detection of increasing HIV type 1 antibodies after seroconversion: a simple assay for detecting recent HIV infection and estimating incidence. AIDS Res Hum Retroviruses. 2002;18:295–307.
11. Dobbs T, Kennedy S, Pau CP, et al.. Performance characteristics of the immunoglobulin G-capture BED–enzyme immunoassay, an assay to detect recent human immunodeficiency virus type 1 seroconversion. J Clin Microbiol. 2004;42:2623–2628.
12. Parekh BS, Hanson DL, Hargrove J, et al.. Determination of mean recency period for estimation of HIV type 1 Incidence with the BED-capture EIA in persons infected with diverse subtypes. AIDS Res Hum Retroviruses. 2011;27:265–273.
13. Lindstrom A, Albert J. A simple and sensitive 'in-house' method for determining genotypic drug resistance in HIV-1. J Virol Methods. 2003;107:45–51.
14. Murillo W, de Rivera IL, Parham L, et al.. Prevalence of drug resistance and importance of viral load measurements in Honduran HIV-infected patients failing antiretroviral treatment. HIV Med. 2010;11:95–103.
15. Larkin MA, Blackshields G, Brown NP, et al.. Clustal W and Clustal X version 2.0. Bioinformatics. 2007;23:2947–2948.
16. Pond SLK, Posada D, Stawiski E, et al.. An evolutionary model-based algorithm for accurate phylogenetic breakpoint mapping and subtype prediction in HIV-1. PLoS Comput Biol. 2009;5:e1000581.
17. Tamura K, Dudley J, Nei M, et al.. MEGA4: Molecular Evolutionary Genetics Analysis (MEGA) software version 4.0. Mol Biol Evol. 2007;24:1596–1599.
18. Huelsenbeck JP, Ronquist F. MRBAYES: Bayesian inference of phylogenetic trees. Bioinformatics. 2001;17:754–755.
19. Tracer v1.4 [computer program]. Available at: Accessed April 14, 2011.
20. Drummond AJ, Rambaut A. BEAST: Bayesian evolutionary analysis by sampling trees. BMC Evol Biol. 2007;7:214.
21. Bennett D, Camacho R, Otelea D, et al.. Drug resistance mutations for surveillance of transmitted HIV-1 drug-resistance: 2009 update. PLoS One. 2009;4:e4724.
22. McNutt L-A, Wu C, Xue X, et al.. Estimating the relative risk in cohort studies and clinical trials of common outcomes. Am J Epidemiol. 2003;157:940–943.
23. Paintsil E, Verevochkin SV, Dukhovlinova E, et al.. Hepatitis C virus infection among drug injectors in St Petersburg, Russia: social and molecular epidemiology of an endemic infection. Addiction. 2009;104:1881–1890.
24. Caceres CF. HIV among gay and other men who have sex with men in Latin America and the Caribbean: a hidden epidemic? AIDS. 2002;16(suppl 3):S23–S33.
25. Jacobson JO, Creswell J, Elena Guardado M, et al.. Coverage of HIV prevention components among people with long-standing diagnosed HIV infection in El Salvador. Sex Transm Dis. 2012;39:694–700.
26. Ministerio de Salud Publica y Asistencia Social de El Salvador. Resumen de Casos de VIH-Sida: Notificacion en El Salvador. San Salvador, El Salvador: Ministerio de Salud; 2009:1984–2008.
27. Vercauteren J, Wensing AMJ, van de Vijver DAMC, et al.. Transmission of drug-resistant HIV-1 is stabilizing in Europe. J Infect Dis. 2009;200:1503–1508.
28. Poon AFY, Aldous JL, Mathews WC, et al.. Transmitted drug resistance in the CFAR network of integrated clinical systems cohort: prevalence and effects on pre-therapy CD4 and viral load. PLoS One. 2011;6:e21189.
29. Egger M, May M, Chene G, et al.. Prognosis of HIV-1-infected patients starting highly active antiretroviral therapy: a collaborative analysis of prospective studies. Lancet. 2002;360:119.
30. Hogg RS, Yip B, Chan KJ, et al.. Rates of disease progression by baseline CD4 cell count and viral load after initiating triple-drug therapy. JAMA. 2001;286:2568–2577.
31. Cohen MS, Chen YQ, McCauley M, et al.. Prevention of HIV-1 infection with early antiretroviral therapy. N Engl J Med. 2011;365:493–505.

HIV-1; molecular epidemiology; phylogenetic; respondent-driven sampling; transmission networks; El Salvador

© 2013 Lippincott Williams & Wilkins, Inc.