Identifying and responding to newly acquired HIV infections remains a critical component of HIV control in North Carolina (NC), where more than 1300 new HIV diagnoses are reported annually.1 Although only a fraction of these new diagnoses are detected during acute HIV infection (AHI) or recent HIV infection (RHI),1 incident infections represent the leading edge of the HIV epidemic. Due in part to elevated viremia during acute infection, an estimated 30% to 50% of transmissions occur in the 6 months after HIV acquisition.2,3 Thus, detection of AHI and RHI presents an important opportunity for secondary prevention via rapid initiation of antiretroviral therapy (ART) and provision of transmission prevention counseling. In pursuit of this goal, NC has actively screened routine HIV tests at public testing sites for AHI since 2002 through the NC Screening and Tracing for Active HIV-1 Transmission (NC-STAT) program.4 Individuals identified with AHI and RHI are prioritized for expedited counseling, contact tracing, and referral services. From 2003 to 2012, 3.4% of all HIV diagnoses in NC were detected as AHI by NC-STAT and 96.6% of these cases successfully reached for interview.4
Although contact tracing remains central to HIV outbreak control, detecting HIV transmission sources and networks is often limited; contact tracing is resource-intensive and subject to participant willingness and recall bias.5 In addition, networks involving anonymous or unreachable partners may go unrecognized despite contact tracing efforts. The integration of traditional contact tracing with HIV sequence analyses may offer opportunities to mitigate some of these barriers. HIV pol sequences, derived from routinely collected drug resistance genotypes, are increasingly used to reconstruct putative HIV transmission clusters through phylogenetic inference in the population of interest.6 The combination of these phylogenetic analyses with epidemiologic data provides a unique view of population-level transmission patterns7 and can be used to evaluate potential and identified HIV outbreaks.8–10
We investigated contact tracing and HIV phylogenetic networks that involved AHI cases after spatiotemporal clustering of AHI cases was detected in NC in early 2014. Between December 2013 and February 2014, more than 3 times as many AHI among men who have sex with men (MSM) were reported in 2 regions in western NC (Regions 1 and 2; Fig. 1) compared with the same time frame during the previous year. HIV transmission in NC is predominately sexual, with most new cases reported among MSM (61%), and is concentrated in urban and suburban areas.1 Metropolitan Region 2 (Charlotte) has the highest HIV cases rate, although Region 1 is among the lowest in the state.1 The apparent spatiotemporal clustering in these regions triggered an investigation by the NC Department of Health and Human Services (NC-DHHS) in February 2014 out of concern for a potentially undetected outbreak. As an exploratory component of the multidisciplinary investigation, we characterized the contact tracing networks involving index cases diagnosed during AHI and RHI in Regions 1 and 2, as well as throughout the state, to determine whether the new infections were phylogenetically linked.
MATERIALS AND METHODS
After an increase in AHI cases reported in Regions 1 and 2 in early 2014, we investigated all index cases diagnosed with AHI or RHI in NC from December 2013 to December 2014 and their reported contacts (Fig. 1). In addition, index cases reported in Regions 1 and 2 from December 2012 to November 2013 were included to assess potential genetic relatedness to these earlier cases in the putative outbreak foci. All index cases were reported to NC-DHHS through the NC-STAT program11 or by acute care settings (i.e., emergency departments, urgent or primary care clinics). Acute HIV infection is defined as (1) a positive antibody test result with seronegative documentation within the previous 30 days, or (2) reactive fourth-generation antigen/antibody combination test with a negative/indeterminate multispot assay finding and a detectable HIV RNA (since November 2013), or (3) negative/indeterminate antibody test and detectable HIV RNA (NC-STAT through RNA pooling of negative tests11 2002–2013). Cases with RHI are similarly investigated by NC-DHHS and defined as a positive antibody test result with seronegative documentation and/or symptoms compatible with AHI within 3 months of diagnosis. Medical and surveillance records are extensively reviewed by STAT program medical staff for suspected RHI cases without confirmed seronegative testing to determine if STAT investigation warranted. Because this study was initially conducted in partnership with an NC-DHHS public health outbreak investigation, no additional informed consent was needed. The institutional review board at the University of North Carolina approved the study.
Contact Tracing Networks
In NC, disease intervention specialists (DISs) perform routine contact tracing for all HIV-infected cases and their named contacts. Disease intervention specialists expedite routine follow-up for AHI and RHI cases through NC-STAT, contacting individuals within 72 hours of HIV-positive testing results. Named sexual and injection drug use (IDU) contacts within the 8 (AHI) or 12 weeks (RHI) before diagnosis are notified of HIV exposure, counseled, and tested for HIV infection within 72 hours of the case interview when possible. High-risk social contacts reported by the index case are also investigated. For cases with newly diagnosed chronic HIV infection, all named contacts reported within 12 months before diagnosis are investigated.
Relevant contact tracing, demographic, and clinical data for index and contact cases were abstracted from the NC Electronic Disease Surveillance system. We classified first-degree contacts of index cases as HIV-infected (previous or new chronic diagnosis), HIV-uninfected, or HIV-status unknown. We defined a new, chronic HIV diagnosis as a case diagnosed 30 days or less before or any time after the date of the index's HIV diagnosis. Contacts who were also diagnosed during AHI or RHI were classified as index cases. Further contact tracing data (second-degree contacts to the index) were included for first-degree contacts who had either a (1) new HIV diagnosis or (2) HIV diagnosis anytime in 2013 or 2014. Demographic data included age at the time of the investigation, race/ethnicity, sex, and county of residence. Clinical data included HIV testing and diagnosis dates, testing locations, initial HIV RNA levels for index cases, and the HIV RNA level reported to NC-DHHS for HIV-infected contacts most proximal to the time of DIS investigation.
We constructed networks using contact tracing data for index cases, their reported contacts, and applicable second-degree contacts to the index cases. Although DISs perform name-based contact tracing, all analyses were conducted on de-identified data sets. Data were symmetrized and an undirected network was constructed using UCINET 6.486 and NetDraw 2.134 (Analytic Technologies, Lexington, KY). Mean index node degrees and overall graph density were calculated for the statewide network and the subset network including contacts epidemiologically linked to index cases from Regions 1 and 2. Isolated nodes were removed before density calculations. We evaluated the demographic and clinical features of independent network components, defined as all index cases and contacts linked by any epidemiologic relationship.
HIV Sequences and Phylogenetic Inference
During the investigation, NC-DHHS requested available HIV-1 pol sequences for all index cases and HIV-infected contacts from laboratories serving clinical providers in NC. These sequences were generated by genotype assays for drug resistance including HIV GenoSure MG and GenoSure PRIme (Laboratory Corporation of America Holdings [Research Triangle Park, NC] and Monogram Biosciences [South San Francisco, CA]), or TRUGENE HIV-1 assay (Siemens Healthcare Diagnostics, Tarrytown, NY). All spanned full-length protease and partial reverse transcriptase; a minority included integrase. Additional partial pol sequences (n = 1672) from participants in the UNC Center for AIDS Research HIV Clinical Cohort were included as background references. These sequences were sampled from 1997 to 2011 and were the largest sequence data set available in NC during the investigation (GenBank JX160108-JX161480). Sequences were aligned using MUSCLE12 and edited manually, stripping gapped positions; final sequence length was 1497 nucleotides. Drug resistance mutations (DRMs) and subtype assignments were identified using The Stanford HIV Web Service (Sierra v.1.1) to query the Stanford HIVdb Program.13 Major DRMs were selected using the 2009 World Health Organization standardized surveillance list.14
We constructed maximum-likelihood phylogenies in RAxML v.7.0.415 with the general time reversible model of nucleotide substitution and assessed statistical support of clades with 1000 bootstrap replicates. We defined transmission clusters as clades with a bootstrap support of at least 98% that included at least 1 index case sequence with 0.015 or less substitutions/site pairwise genetic distance from another sequence from another index or any contact case. In addition, we identified transmission clusters that were part of larger, genetically divergent clades and those that involved index cases in clades with a reference sequence. All analyses and results were shared with NC-DHHS on a regular basis during the approximately 12-month investigation period to inform the ongoing outbreak investigation in a timely manner.
In total, 68 index cases were identified (Table 1); 65 were index cases reported statewide from December 2013 to December 2014 (55 AHI and 13 RHI) and 3 were AHI cases reported in Region 2 in from 2012 to 2013. Nearly half of the index cases (n = 32/68) were diagnosed in Regions 1 or 2. In comparison, from 2003 to 2012, an average of 23 AHI cases were diagnosed annually statewide (range, 15–32).4 Almost all index cases (n = 66/68) were reached by DIS. Index cases were predominately male (n = 58/68; 85%) and median age was 25 years (interquartile range [IQR], 22–35 years; Table 1). Most index cases were black (n = 45/68; 66%), although racial composition varied by region; 86% (6/7) of index cases in Region 1 were white, compared with only 24% (6/25) in Region 2. Among the RHI cases, most (10/13; 77%) were categorized based on documented negative HIV test results within 3 months of diagnosis.
From the 66 index cases interviewed, 210 unique contacts were reported (Table 1). Of these, 189 were first-degree contacts (yielding 197 partnerships: 3 contacts were named more than once and 4 contacts were index cases) and 21 were second-degree contacts. All partnerships were sexual (93%) or social (7%), and none were identified as IDU. Index cases reported a mean (SD) of 3.0 (2.0) first-degree contacts. Of the 189 first-degree contacts, 51 (27%) were HIV infected, 66 (35%) were HIV uninfected, and 72 (38%) had unknown HIV status. Nearly one third of the 189 first-degree contacts (33%; n = 62/189) could not be located for DIS interviews; 38 (61%) of 189 were anonymous partnerships and 24 (39%) of 189 had incomplete information or could not be found. Fifteen contacts resided outside NC.
Most HIV-infected first-degree contacts (71%; 36/51) were previously diagnosed with a median of 4.2 years (IQR, 1.4–6.8 years) elapsed between contact and case diagnoses. Two contacts were named by more than 1 index case and were assessed at the time of the first index case's diagnosis. Most previously diagnosed contacts (75%) had HIV viremia (RNA >1000 copies/mL) at their most recent care visit or had no HIV RNA reported, suggesting a lack of care engagement. The 15 first-degree contacts with new HIV diagnoses were diagnosed an average of 10 days after the corresponding index case (range, 25 days before 39 days after index case). Of the 21 second-degree contacts, 20 (95%) were HIV-infected and 1 (5%) had an unknown HIV status.
Contact Tracing Networks
In the statewide network, we identified 58 independent components (median, 4 [range, 1–28] individuals per component; Figure, Supplemental Digital Content 1, http://links.lww.com/OLQ/A199). The mean (SD) degree in the statewide network was 1.7 (1.5), and the overall network density was low at 1.2% of actual contact ties among all possible connections in the network. Within the investigation focal area (Regions 1 and 2), the overall density was slightly higher at 2.5%. Statewide, 52% components (30/58) included at least 1 individual from these regions.
The 32 index cases reported in Regions 1 and 2 were distributed in 26 components, including 2 index cases unreachable by DIS (Fig. 2). The network included 118 discrete contacts (100 first-degree and 18 second-degree) with median component size of 3 (range, 2–28). Most components (85%: 22/26) contained index cases exclusively from Region 2. However, 1 (component 6) contained index cases from both Regions 1 and 2, and 3 components (components 1, 2, and 3) contained index cases exclusively from Region 1. The mean (SD) index node degrees by contacts' HIV statuses were 0.8 (0.9) for HIV-infected contacts, 1.0 (1.2) for HIV-negative contacts, and 1.3 (1.5) for contacts with unknown HIV status.
HIV-1 Sequence Analyses and Comparison to Contact Network
HIV-1 pol sequences were available for 56% (38/68) of the index cases and 48% (34/71) HIV-infected contacts (Table 1). Sequence availability did vary by region with a higher proportion in Region 1 (71%) and Region 2 (68%). Time from diagnosis to drug resistance sequencing was shorter for index cases compared with contacts (median, 19 [IQR, 12–34] vs. 43 [IQR, 17-529] days for contacts). All sequences from index cases were sampled before ART; antiretroviral exposure data were not available for contacts. All sequences were subtype B.
We identified 13 transmission clusters involving more than 1 index or an index and any contact (Table 2; Figure, Supplemental Digital Content 2, http://links.lww.com/OLQ/A200). These clusters included 30 individuals (range, 2–4 members), involving 42% (16/38) of the index cases and 41% (14/34) of the contacts with sequences. Six clusters (46%) involved an index from Region 1 or 2; only 3 clusters spanned more than 1 region. Although most clusters contained only MSM (n = 11), 2 clusters involved women (clusters L and M). In addition, nearly one third of clusters (31%; 4/13) included a previously diagnosed contact. Many clusters were subclusters on highly supported clades (with larger genetic distance spans) with reference sequences (Figure, Supplemental Digital Content 2). In addition, 2 index cases that did not cluster with a case or contact sequence clustered closely (<1.5% pairwise genetic distance difference) with a reference sequence.
Major DRMs to non–nucleoside reverse transcriptase inhibitors were common. Six (16%) index cases had transmitted drug resistance (TDR); 5 of these were in 4 clusters where all members shared the same mutations, including all with K103N (Table 2). Among contacts, 38% (13/34) had a major DRM (nearly all to non–nucleoside reverse transcriptase inhibitors). No major DRMs in protease or integrase (among 20 integrase sequences) were identified among index or contact cases.
Although a single clustered outbreak was not identified, there are notable findings in comparing the contact and phylogenetic networks. Four clusters connected more than 1 network component revealing links not apparent from contact tracing (including MSM clusters B, F, and G in Regions 1 and 2). In addition, components 40 and 39 in Region 6, involving 2 women and 1 man, were phylogenetically linked (cluster M). The largest component (no. 8; n = 28 cases) included members identified in distinct phylogenetic clusters (clusters C and D; Fig. 2). This component included 18 HIV-infected contacts and 4 index cases from Regions 2 and 4, demonstrating geographic bridging of clusters and components.
We conducted an integrated phylogenetic and contact tracing investigation to assess for a connected HIV outbreak after the spatiotemporal clustering of AHI was reported in NC. The analyses demonstrate that the observed spatiotemporal clustering was not a product of the expansion of a single linked cluster, but rather of the concurrent expansion of several smaller clusters. These ongoing HIV transmission chains were embedded in sexual network components that appeared largely disconnected both within and between regions across the state. The rise in AHI observed may be a random data artifact or the result of improved screening or reporting of AHI. Nonetheless, our study demonstrates the feasibility and applicability of incorporating HIV sequence analyses into an AHI outbreak investigation.
This investigation represents a novel integration of routinely generated contact tracing and sequence data to assess a potential HIV outbreak. Genetic networks have been widely used retrospectively to characterize local,16 national,17 and international HIV clusters,18 and contact tracing has long served as a backbone of infection control. However, only a few studies integrated both methods during an HIV investigation8,19 or retrospectively to confirm outbreaks attributed to IDU.9 Among a cluster of HIV cases among rural MSM, phylogenetics showed HIV transmission arose from separate sources.8 In another study of MSM, sexual network components that were not connected through contact tracing were genetically linked.19 These findings are congruent with our results showing multiple transmission chains and links between network components. In addition, our analyses show significant clustering of TDR, which highlights an additional advantage of tracking sequence data. Clustering analyses may inform public health efforts in the event of a future TDR outbreak. All TDRs in our study included K103N, a mutation that is increasingly linked to ART-naïve transmission sources.20,21
Our primary objective was to test for a genetically clustered outbreak, rather than identify likely transmitters to AHI index cases. However, social and genetic networks combined can inform the use of genetic linkage to identify potential transmission partners.22 Recently proposed methods to integrate both networks are all reliant on relative completeness of both networks and minimal time delays between infection, diagnosis, contact tracing, and sequencing.23–26 Our study is limited in that not all diagnosed cases had a sequence available for analysis, thus providing a restricted view of the transmission network. Furthermore, nearly one third of contacts were not located and may be important connections in the transmission network. These analyses also cannot confirm direction of transmission, and unsampled third parties, including undiagnosed persons, may be involved in the transmission chain. Thus, observed phylogenetic linkages may be indirect in reality. Despite these limitations, more than half of the index cases in the investigation foci had sequences, and phylogenetic analyses showed multiple clusters rather than 1 cluster; adding more sequences may increase the size and number of these clusters. With availability of more complete data, future investigation into source attribution can further inform local HIV transmission dynamics and outbreak investigations. Future analyses would also benefit from the inclusion of sequencing data from neighboring states, particularly given a high number of out-of-state partners reported in contact tracing.
The network analyses revealed a low yield from HIV case finding, whether previously or newly diagnosed. Both network densities statewide and in the investigation foci were low, indicating that cases (diagnosed or undiagnosed) in the community are not being identified and that chronically infected individuals may be important contributors to ongoing transmission.27 Prior studies have shown variable success of HIV case finding from contact tracing.28,29 Contact tracing of AHI may theoretically yield higher case finding compared with chronic cases because DIS expedites the investigation and direct attention to higher-risk networks. In addition, AHI cases may have less recall bias due to the condensed infectious period for potential contacts. However, the index node degree was low at 0.8 for HIV-infected contacts and higher at 1.3 for HIV-unknown status contacts. The mean index node degree for HIV-infected contacts indicates the mean number of contacts named by index cases who were diagnosed as having HIV, excluding any potential HIV-infected contacts who could not be reached by DIS and any anonymous contacts. These findings largely stem from the substantial proportion of cases that could not be located and altogether indicate the need for improved case detection.
Our exploratory investigation revealed multiple HIV transmissions clusters, involving both new and previously diagnosed cases, which reinforces the need for a comprehensive HIV prevention strategy. Enhanced prevention measures toward transmission clusters and highly connected contact networks could be considered, including intensified, expedited partner and linkage to care services for HIV-infected cases and preexposure prophylaxis for HIV-negative cases. In addition, innovative methods are needed to better identify high-risk networks and to improve case finding. In this study, no cases reported IDU contacts, which is consistent with our earlier phylogenetic studies in NC showing predominately sexual transmission.16 However, continued surveillance is important given increased rates of opioid drug and the recent HIV outbreak associated with IDU in Indiana.9 In support of these efforts, NC-DHHS is currently establishing a formal system for routine analysis of HIV sequences, as is done through sequences reported through the National HIV Surveillance System.30 Such an expanded surveillance system will allow for more timely and complete identification of HIV transmission clusters, a limitation of this study. Combined with ongoing analysis of surveillance data, such a system could allow for improved epidemic monitoring including for TDR, outbreak detection, and DIS response.
The observed spatiotemporal clustering in this putative outbreak was a product of the concurrent expansion of several smaller clusters, rather than a single-outbreak cluster. Integration of contact tracing and phylogenetic networks may provide insights into HIV transmission dynamics and may be conducted in near real time to support outbreak investigations. Further development of methods to enhance the detection of and response to HIV transmission and contact networks is needed.
1. North Carolina HIV/STD Surveillance Unit. North Carolina HIV/STD Epidemiologic Profile 2013. Issued March 2015. Raleigh, NC: North Carolina Department of Health and Human Services; 2015. Available at: http://epi.publichealth.nc.gov/cd/stds/figures/Epi_Profile_2013.pdf
. Accessed July 12, 2016.
2. Brenner BG, Roger M, Routy JP, et al. High rates of forward transmission events after acute/early HIV-1 infection. J Infect Dis 2007; 195:951–959.
3. Lewis F, Hughes GJ, Rambaut A, et al. Episodic sexual transmission of HIV revealed by molecular phylodynamics. PLoS Med 2008; 5:e50.
4. Kuruc JD, Cope AB, Sampson LA, et al. Ten years of screening and testing for acute hiv infection in North Carolina. J Acquir Immune Defic Syndr 2016; 71:111–119.
5. Resik S, Lemey P, Ping LH, et al. Limitations to contact tracing and phylogenetic analysis in establishing HIV type 1 transmission networks in Cuba. AIDS Res Hum Retroviruses 2007; 23:347–356.
6. Hassan AS, Pybus OG, Sanders EJ, et al. Defining HIV-1 transmission clusters based on sequence data. AIDS 2017; 31:1211–1222.
7. Brenner B, Wainberg MA, Roger M. Phylogenetic inferences on HIV-1 transmission: Implications for the design of prevention and treatment interventions. AIDS 2013; 27:1045–1057.
8. Nett RJ, Bartschi JL, Ellis GM, et al. Two clusters of HIV-1 infection, rural Idaho, USA, 2008. Emerg Infect Dis 2010; 16:1807–1809.
9. Peters PJ, Pontones P, Hoover KW, et al. HIV infection linked to injection use of oxymorphone in Indiana, 2014–2015. N Engl J Med 2016; 375:229–239.
10. Poon AF, Gustafson R, Daly P, et al. Near real-time monitoring of HIV transmission hotspots from routine HIV genotyping: An implementation case study. Lancet HIV 2016; 3:e231–e238.
11. Pilcher CD, Fiscus SA, Nguyen TQ, et al. Detection of acute infections during HIV testing in North Carolina. N Engl J Med 2005; 352:1873–1883.
12. Edgar RC. MUSCLE: Multiple sequence alignment with high accuracy and high throughput. Nucleic Acids Res 2004; 32:1792–1797.
13. Liu TF, Shafer RW. Web resources for HIV type 1 genotypic-resistance test interpretation. Clin Infect Dis 2006; 42:1608–1618.
14. Bennett DE, Camacho RJ, Otelea D, et al. Drug resistance mutations for surveillance of transmitted HIV-1 drug-resistance: 2009 Update. PLoS One 2009; 4:e4724.
15. Stamatakis A. RAxML-VI-HPC: Maximum likelihood-based phylogenetic analyses with thousands of taxa and mixed models. Bioinformatics 2006; 22:2688–2690.
16. Dennis AM, Hue S, Hurt CB, et al. Phylogenetic insights into regional HIV transmission. AIDS 2012; 26:1813–1822.
17. Aldous JL, Pond SK, Poon A, et al. Characterizing HIV transmission networks across the United States. Clin Infect Dis 2012; 55:1135–1143.
18. Wertheim JO, Leigh Brown AJ, Hepler NL, et al. The global transmission network of HIV-1. J Infect Dis 2013; 209:304–313.
19. Peters PJ, Gay C, Beagle S, et al. HIV infection among partners of HIV-infected black men who have sex with men—North Carolina, 2011–2013. MMWR Morb Mortal Wkly Rep 2014; 63:90–94.
20. Mbisa JL, Fearnhill E, Dunn DT, et al. Evidence of self-sustaining drug resistant HIV-1 lineages among untreated patients in the United Kingdom. Clin Infect Dis 2015; 61:829–836.
21. Drescher SM, von Wyl V, Yang WL, et al. Treatment-naive individuals are the major source of transmitted HIV-1 drug resistance in men who have sex with men in the Swiss HIV Cohort Study. Clin Infect Dis 2014; 58:285–294.
22. Wertheim JO, Kosakovsky Pond SL, Forgione LA, et al. Social and genetic networks of HIV-1 transmission in New York City. PLoS Pathog 2017; 13:e1006000.
23. McCloskey RM, Liang RH, Poon AF. Reconstructing contact network parameters from viral phylogenies. Virus Evol 2016; 2:vew029.
24. Villandre L, Stephens DA, Labbe A, et al. Assessment of overlap of phylogenetic transmission clusters and communities in simple sexual contact networks: Applications to HIV-1. PLoS One 2016; 11:e0148459.
25. Jombart T, Cori A, Didelot X, et al. Bayesian reconstruction of disease outbreaks by combining epidemiologic and genomic data. PLoS Comput Biol 2014; 10:e1003457.
26. Zarrabi N, Prosperi MCF, Belleman RG, et al. Combining social and genetic networks to study HIV transmission in mixing risk groups. Eur Phys J Spec Top 2013; 222:1377–1387.
27. Cope AB, Powers KA, Kuruc JD, et al. Ongoing HIV transmission and the HIV care continuum in North Carolina. PLoS One 2015; 10:e0127950.
28. Golden MR, Hogben M, Potterat JJ, et al. HIV partner notification in the United States: A national survey of program coverage and outcomes. Sex Transm Dis 2004; 31:709–712.
29. Brewer DD. Case-finding effectiveness of partner notification and cluster investigation for sexually transmitted diseases/HIV. Sex Transm Dis 2005; 32:78–83.
30. Oster AM, Wertheim JO, Hernandez AL, et al. Using molecular HIV surveillance data to understand transmission between subpopulations in the United States. J Acquir Immune Defic Syndr 2015; 70:444–451.