INTRODUCTION
Hepatitis C virus (HCV) is a contagious blood-borne disease that is considered a global public health threat [Reference Bruggmann1]. The World Health Organization (WHO) estimates that about 130–150 million people are infected with chronic hepatitis C globally and that about 35–50 million die from HCV-related diseases [Reference Walsh2, Reference Gower3]. There has been an increasing trend in the prevalence of HCV infection in East-Central Asia, with East Asia reported to have an anti-HCV antibody prevalence of 1.2% and a viraemic HCV prevalence of 0.7% [Reference Walsh2, Reference Gower3]. Studies examining the genetic diversity of HCV have indicated that six genotypes and more than 80 sub-genotypes exist in the HCV-infected population [Reference Dong4].
People who inject drugs (PWID) are a high-risk population for HCV infection [Reference de Vos, Prins and Kretzschmar5, Reference Hagan, Pouget and Des Jarlais6]. The estimated prevalence of HCV in PWID in China is 61.4–70.0%, with the heavily endemic areas being the Hubei, Yunnan, Guangxi, Hunan and Xinxiang provinces. A previous study reported that China has the largest population of PWID worldwide, with about 1.33 million PWID [Reference Zhuang7]. Considering the large number of PWID in China, the challenge of controlling HCV transmission in PWID is great.
Epidemiological studies have recently been increasing their focus on the social networks of PWID [Reference Puerta8]. Social network analysis (SNA) allows for the identification and targeting of groups of people with high-risk behaviours and the investigation of the mode of disease transmission in a scientific manner, based on traditional contact tracing using structured interviews and questionnaires [Reference Gardy9]. The target groups can include highly active participants, high-risk behaviour groups and individuals serving as a ‘bridge’ connecting people. A number of previous studies have used the SNA approach to explore HIV transmission in different countries including China [Reference Gyarmathy10–Reference Rice12]. A study derived from a cohort of PWID confirmed that HCV phylogenetic clustering is associated with the social injecting network, highlighting the importance of the injecting network in HCV transmission [Reference Sacks-Davis13]. Rolls et al. [Reference Rolls14] created an empirical network of PWID and developed a detailed model of HCV transmission in this network, providing valuable insights for controlling high-risk groups. Additionally, researchers from Australia determined that HCV transmission clusters correlated with reported injecting relationships [Reference Hellard15].
To our knowledge, limited information exists regarding SNA of PWID infected with HCV in China. No studies have used SNA and gene sequencing together to explore the transmission routes of HCV in Chinese patients under methadone maintenance therapy (MMT) who are infected with HCV. It was hypothesized that social network linkages would exist in PWID attending different MMT clinics in a single city (Wuhan, China), as a previous study had identified social network linkages between PWID attending different street drug markets in Melbourne, Australia (a city of slightly larger area than Wuhan) [Reference Sacks-Davis13]. Using a cohort study performed in China in MMT users infected with HCV [Reference Zhou16], this study aimed to: (1) explore the social networks in PWID in central China; (2) reveal the HCV genotypes of the target groups; and (3) assess how certain network structural characteristics such as node degree, network density and betweenness centrality are related to HCV infection in PWID. This study will provide useful insights into HCV transmission in PWID, which potentially could be used by public health services to help control HCV epidemics.
METHODS
Study design and participants
This study comprised of three parts: a baseline survey, HCV viral sequencing of a case group (HCV seroconversion), and SNA.
The clinical records of all patients visiting the 20 MMT clinics in Wuhan (China) were reviewed. The distances of the other MMT clinics from the most central clinic ranged from 1.1 km to 23.6 km (mean 10.3 ± 7.5 km). In 2013, Wuhan city had a population of ~10 220 000, and 0.27% of the population (~27 950) were PWID (according to the Public Security Bureau of Wuhan City). Around 14 000 PWID attended MMT clinics in 2013, and the number of people taking drugs each day was estimated to be 4000.
Of the 16 085 patients visiting between May 2006 and June 2011, 12 755 (79.3%) received a HCV test at MMT entry. Of the 3558 MMT patients who tested HCV-negative at MMT entry, 47.8% (1702/3558) agreed to participate in the prospective follow-up study. To ensure that the investigated HCV antibody seroconversion was the result of transmission during the MMT programme, only participants who had two HCV-negative blood specimens (at MMT entry and at least 3 months later) and who completed at least one follow-up interview were included in the final study sample; therefore, 502 (29.5%) of 1702 eligible patients were excluded. Consequently, 1200 participants were included in the prospective follow-up sample. Their characteristics have been published recently [Reference Zhou16]. Of the 1200 participants, seroconversion was observed in 555 between May 2006 and June 2011. By the beginning of the present study (April 2014), 32.07% (178/555) of patients who showed seroconversion were still attending their MMT clinic and 31 participants agreed to supply a blood sample. Consequently, the case group included these 31 participants. Of the 645 non-seroconversion participants, only 128 continued to attend their MMT clinic, of which 49 agreed to participate in the study. Ultimately, the case group included 31 participants and the control group included 49 participants (see Supplementary Fig. S1).
Epidemiological investigation
The socio-demographic data (including gender, age, education, marital status, employment status, living status) and self-reported behavioural data (such as illicit drug use history, needle/equipment sharing history, drug rehabilitation, treatment history) were assessed using a questionnaire. This survey was conducted during 1–15 April 2014, with the cooperation of MMT doctors (face-to-face method). All 80 patients completed the questionnaire, as required for inclusion in the final analysis.
Specimen collection and HCV genotyping
Fasting whole blood specimens (10 ml) were collected in K3 EDTA tubes, immediately stored at 4 °C, and transferred to the laboratory of Wuhan CDC for processing on the same day. Plasma samples were stored in aliquots at −20 °C until analysed.
All specimens were tested for anti-HCV antibodies using the third-generation enzyme-linked immunoassay (EIA-3) system (Kehua Biotechnology Inc., China), according to the manufacturer's instructions. Specimens found to be initially positive by EIA-3 were retested in duplicate. Repeat reactive specimens were assumed to be seropositive. Specimens shown to be reactive after repeat testing were used to conduct further HCV RNA analyses.
HCV RNA was extracted using the care HCV RT–PCR assay v. 2 kit (Qiagen, Germany), according to the manufacturer's instructions. The amount of HCV RNA was measured by UV spectrophotometry. RNA (1 µg) was used for real-time polymerase chain reaction (PCR) analysis using the 7300 real-time PCR system. The isolated RNA was used to synthesize cDNA using random primers. RNA (1 µg), the random primers and ddH2O were mixed in a reaction volume of 12 µl and incubated at 65 °C for 5 min in the PCR instrument (Eppendorf, Germany). The mixture was immediately placed on ice and dNTPs and reverse transcriptase (RT) were added. After incubation at 30 °C for 10 min and 42 °C for 20 min, cDNA was obtained. The cDNA was then amplified in a reaction volume of 20 µl using RT–PCR and the SYBR Green reaction system (Applied Biosystems, USA) using two sets of primers (Ac2 and Sc2; S7 and A7) targeting the 5’-UTR and core (335 bp) regions [Reference Ohno17]. The sense sequencing primer was designed at nucleotide positions −12 to 8 of the 5’-UTR and core regions. The antisense primer was located at nucleotide positions 319–343 of the core region. The reaction conditions were pre-denaturation at 95 °C for 1 min, denaturation at 95 °C for 15 s, renaturation at 56 °C for 20 s, and extension at 72 °C for 45 s, for 40 cycles. The fluorescence signal was collected during the extension phase. The RT–PCR amplified DNA products were analysed by agarose gel electrophoresis (2%) containing 0.5 µg/ml ethidium bromide. The DNA PCR products were purified from the agarose gel and sequenced by a sequencing company (Sangon Biotech Co. Ltd, China).
A phylogenetic tree was constructed using the 5’-UTR/core region sequences of 31 HCV strains aligned to randomly selected published HCV sequences that were retrieved from the GenBank database. The statistical reliability of this tree was determined using a bootstrap analysis that provides a percentage relatedness score for each cluster predicted by the tree. A cluster with a score of >95% was considered as significantly related.
The resulting sequences were aligned using CLUSTAL X 2.1 (University College Dublin, Ireland). Phylogenetic trees were constructed with MEGA 5.05 software (USA) using the maximum-likelihood method, and the reliability of the clustering was assessed using the bootstrap test (1000 replicates). Genetic distances were calculated using the Poisson model with MEGA 5.05. Reference sequences were used to construct the phylogenetic tree on the basis of 15 HCV isolates selected from GenBank (http://www.ncbi.nlm.nih.gov/genbank/). The 15 HCV isolates were D11443 (Japan), D49374 (Japan), GU814263 (Italy), NC009824 (Japan), NC009827 (Japan), EU246930 (Vietnam), FJ410172 (USA), NC004102, D10750, FJ390398, NC009826, NC009823 (USA), D10922, GU814265 (Egypt) and NC009825 (Middle East).
The obtained sequences were submitted to GenBank and given the following accession numbers: JX944384–JX944467 and KC348398–KC348438.
SNA investigation
Social network data were obtained using a social network questionnaire (SNQ) following the SNA design principle, which stipulates that the questionnaire must include a roster (i.e. a list of the names, including given names and commonly used nicknames, of all 80 participants), anchored choice (e.g. ‘face-to-face contact’ and ‘have never met’) and rating (a scale between the two anchored choices to indicate the frequency of contact) [Reference Scott18]. Participants were asked to examine the roster and indicate which participants they had contacted frequently in the past 3 months, which participants they had a close relationship with, and which participants they had never known (excluding themselves). The contact dimension was divided into two levels: regular contact (defined as ‘face-to-face contact >6 times in the past 3 months’ or ‘face-to-face contact <6 times in the past 3 months but have close relationship with the other participant’) and minimal contact (‘face-to-face contact <6 times in the past 3 months and do not have close relationship with the other participant’ or ‘have never met’).
To ensure the quality of the research, each participant was shown only the names/aliases of the other participants in the roster, in order to avoid subjective selections unrelated to the study objective. Other information (such as gender, MMT clinics attended, home address and contact information) was available only to the researchers. The collection of SNA data was conducted during 15–30 April 2014. SNA data were collected for all 80 patients.
Statistical analysis
The epidemiological data were entered into EpiData software (www.EpiData.dk). The statistical analyses were performed using SPSS v. 17.0 (SPSS Inc., USA). Continuous data are expressed as the mean ± standard deviation (s.d.) and were analysed using Student's t test. P < 0.05 was considered statistically significant.
All network analyses were conducted using UCINET software [Reference Borgatti, Everett and Freeman19]. The networks were constructed with each study participant represented by a node (or vertex), and with frequent contact or familiarity between two participants represented by an edge between two nodes. The networks were described on the basis of the following properties: network density, node degree, number of cliques, network centralization index, number of components, average distance and betweenness centrality. The network density (range 0–1) describes the general level of cohesion in the network, and was calculated as the number of edges in the network (i.e. direct links between nodes representing frequent contact or familiarity) divided by the maximum number of edges possible [which is given by n(n – 1)/2, where n is the total number of nodes]. The node degree for each node was defined as the number of other nodes that connected to that node (i.e. the number of incident edges). A clique in a social network describes a subset of participants who are more closely tied to one another than to other members of the network. A clique was defined as a subset of participants such that every two vertices were adjacent (i.e. all nodes within the clique were directly connected by edges to all other nodes within the clique); the minimum size of a clique was set to three nodes. Network centralization index (range 0–100%) measures the degree of dispersion of all the node degree centrality scores in a network from the maximum degree centrality score obtained in the network, where the degree centrality score of a node is an indicator of how connected that node is to all other nodes. A high network centralization index indicates that a network contains a small number of central nodes that are highly connected to most other nodes, and a larger number of less central nodes that tend to connect only to the small number of central nodes. The average distance was calculated as the average length of all the shortest paths (geodesics) between all pairs of vertices. The components of the network were defined as subgraphs (each of which contains a group of connected nodes) that were disconnected from each other [Reference Freeman, Borgatti and White20]. The betweenness centrality measures the extent to which a node lies on paths between other nodes, and thus measures the ability of a participant (represented by that node) to potentially influence the interaction between other participants in the network. The betweenness centrality of a particular node is calculated as the number of shortest paths from all nodes to all others that pass through that particular node. In the context of this study, a node with high betweenness centrality would be considered to have a large influence on HCV transmission through the network, under the assumption that HCV transmission follows the shortest paths.
Ethical approval
Ethical approval was given by Institutional Review Boards (IRB) of the Wuhan Centers for Disease Control and Prevention (CDC) (reference no. WHCDCIRB-K-2014001). All participants signed an informed consent form that included a detailed description of the potential benefits of the study. The patients were aware and understood that for SNA their identity would be provided to the other participants.
RESULTS
Characteristics of the participants
The demographic and socioeconomic characteristics of the participants are summarized in Table 1. Of the 31 case participants, 71% were male and 91% were in the 30–49 years age group; 67% of the controls were male and 73% of the controls were in the 30–49 years age group. All participants reported low levels of education and did not have steady or continuous employment. Within the case group, 48% were married or had a regular partner, and most (90%) lived with their family. Over half (59%) the control group was married or had a regular partner; 78% lived with their family. A comparison between the case and control groups regarding drug injection within the past 30 days revealed significant differences (71% vs. 47%, respectively, P < 0.05).
Table 1. Characteristics of the participants
* P < 0.05.
HCV genotypic distribution
The genotypes of 31 HCV sequences were determined by phylogenetic analysis against a panel of HCV reference sequences. In total, three HCV genotypes covering five different subtypes were identified, as shown in Figure 1. The largest group, comprising 13 sequences on the 5’-UTR/core tree, clustered with the subtype 3b reference sequence D11443. The second largest group, comprising 10 sequences on the 5’-UTR/core tree, clustered with the subtype 6a reference sequence. The other groups clustered with subtypes 3a, 1b and 1a. The proportions for each genotype were: 51.6% (16/31) for genotype 3 (three with subtype 3a and 13 with subtype 3b), 32.3% (10/31) for genotype 6 (all with subtype 6a), and 16.1% (5/31) for genotype 1 (two with subtype 1a and three with subtype 1b).
Fig. 1. The genotypes of 31 HCV sequences were determined by phylogenetic analysis against a panel of HCV reference sequences. In total, three HCV genotypes covering five different subtypes were identified. Thirteen sequences (QK4, QK20, WC2, WC10, WC11, WC12, WC14, WC27, WC32, WC38, WC127, WC129, WC135) clustered with the subtype 3b reference sequence, ten sequences (QK3, QK19, WC3, WC5, WC21, WC24, WC34, WC35, WC36, WC37) clustered with the subtype 6a reference sequence, three sequences (WC23, WC31, WC40) clustered with the subtype 3a reference sequence, three sequences (WC8, WC9, WC29) clustered with the subtype 1b reference sequence, and two sequences (WC15, WC126) clustered with the subtype 1a reference sequence.
Characteristics of the social networks
The overall descriptive statistics for the two groups are presented in Table 2. The densities of the social networks for the whole sample (N = 80), case group (n = 31) and control group (n = 49) were 0.038, 0.054 and 0.008, respectively. The mean node degree was 3.025 ± 2.525 (range 0–17) for the whole sample, 3.000 ± 2.680 (range 0–14) for the case group and 0.776 ± 0.789 (range 0–3) for the control group (P < 0.05, Kruskal–Wallis test, Table 2). The most notable result, as shown in Table 2, was that 49 cliques between the 49 ego networks in the control group were identified. This number is 1.69 times greater than that in the case group, which suggests a structural difference between the two groups. Similarly, the network centralization index of the overall population (28.05%) was higher than that of the control group (0.04%). The network for all participants contained eight components, the largest of which contained 72 nodes.
Table 2. Social network analysis indicators in the two groups
Visualizing social networks is another way of determining the structural differences between them (Fig. 2). It was evident that the larger the node, the higher the value of betweenness for the participant. The network presented with five central points that were active in the network: participant 1 (936.26), participant 2 (411.26), participant 13 (312.97), participant 66 (456.74) and participant 70 (491.45). Additionally, clustering was prevalent for the case group, and it was particularly noticeable that PWID infected with HCV mainly maintained contacts with participants within the same group. This is clear in Figure 2, from which it is evident that participants in the case group had frequent contact with other members of their group, whereas contact was much less frequent between members of the control group.
Fig. 2. Socio-meric graph of the study population. Large nodes depict higher betweenness centrality.
The distribution of the number of participants from each MMT clinic is shown in Supplementary Table S1. Figure 3 presents an alternative visualization to emphasize the within-clinic structure of the social network, with lines between points representing familiarity/frequent contact between participants in the same MMT clinic. The results depicted in Figure 3 suggest that participants within the same MMT clinic have more familiarity or frequent contact, although it was not possible to determine whether this contact between participants within a clinic was initiated before or after registration at the clinic.
Fig. 3. Visualization of the within-clinic structure of the social network. Red symbols identify individuals who were HCV-positive. Lines between participants represent familiarity/frequent contact between participants in the same MMT clinic.
Construction of the susceptible population and participants with a high-probability of transmission based on gene sequencing and SNA
Participants 1–31 belonged to the case group, while participants 32–80 belonged to the control group. Based on the interesting findings regarding the characteristics of the social networks described above, we attempted to determine the potentially susceptible population and the participants with a high probability of transmission. First, four pairs of nodes with a genotypic distance of 0.000 were identified, these clustered in subtypes 6a and 1b. The four pairs of nodes were participants 24 (6a) and 5 (6a), participants 7 (1b) and 6 (1b), participants 26 (6a) and 2 (6a), and participants 25 (6a) and 3 (6a). Further analysis of the social network revealed that each pair of subjects was linked and existed in one clique. Second, three of the five most active nodes were infected with HCV; they were determined to be participants 1 (6a), 2 (6a) and 13 (3b). The above three nodes served as a bridge, contributing to the connection of the other nodes that were identified as high-probability transmission participants with the highest betweenness centrality. Correspondingly, the negative individuals who were in close connection with the above three participants and existed in the same clique were regarded as the principal susceptible population.
DISCUSSION
Social networks facilitate the transmission of HCV in PWID, but no studies have used SNA and gene sequencing to explore the transmission of HCV in MMT patients infected with HCV in China. Therefore, this study aimed to assess how certain network structural characteristics are related to HCV infection in PWID and to determine which individual PWID are susceptible to HCV transmission. Three HCV genotypes were identified, covering five subtypes. The densities of the social networks for the whole sample, case group and control group were 0.038, 0.054 and 0.008. PWID infected with HCV were in frequent contact with others within their group. There were four pairs of nodes with genotypic distances of 0.000 that were identified and clustered in subtypes 6a and 1b; each pair of subjects was linked and found in one clique. Three of the five most active nodes were infected with HCV. These three nodes served as a bridge, contributing to the connection of the other nodes.
This study showed that there are differences in injection risk behaviours between HCV-infected PWID and those not infected. Gene sequencing further showed the genetic types and distribution in HCV-positive participants. Using SNA we identified highly central individuals. The combination of SNA and gene sequencing was used in order to circumvent the recognized limitations of traditional epidemiological investigations [Reference El-Sayed21]. SNA can improve case-finding in vulnerable populations using specific questionnaires that identify the activity density, activity diameter, cliques and high-risk behaviours [Reference Rolls14, Reference Scott18].
In this exploratory study, we first determined the epidemiological characteristics of PWID in central China. Most of the participants were middle-aged men aged 30–49 years. The study population was similar to that of a previous investigation [Reference Hahn22]. In addition, the case group had a low level of education and was mostly unemployed. In terms of injection risk behaviours, HCV-infected PWID generally reported a high rate of drug injection in the past 30 days (71%), similar to the findings of another report (68.8%) [Reference Clatts23]. Despite the substantial decline in the incidence of acute HCV infection observed over a 25-year period (1982-2006) [Reference Walsh2], being a PWID remains a relevant risk factor for acquiring a new HCV infection. Nonetheless, in the context of the high seroconversion rate in PWID, the injecting risk behaviours were extensive in this population [Reference Tsui24]. Thus, it is possible that the control participants in this study who did not seroconvert were interacting with fewer PWID since they were injecting less frequently.
In this study, the HCV genes found in PWID were sequenced and genotyped. The results showed that 3b (41.94%, 13/31) and 6a (32.26%, 10/31) were the predominant subtypes, which is similar to the genotypic distribution in the PWID population in southern Chinese provinces, where the main HCV subtypes included 6a (38%) and 3b (37%) [Reference Garten25]. However, the genotypic distribution of HCV in patients receiving MMT in this study was not entirely consistent with previous results in PWID not receiving MMT, which showed the predominant HCV subtypes to be 6a (50%) and 3b (32.2%) [Reference Peng26]. This may be due, in part, to the small sample size in this study. Sequence analysis provides a powerful tool that may be used to investigate the spread of HCV within a community. Interestingly, four pairs of PWID presented genetic distances of 0; their HCV genes showed a 99% similarity. We found that the proportion of patients in transmission clusters with a genotypic distance of 0.000 accounted for up to 25.8% (8/31) of the participants.
The limitations of traditional epidemiological research preclude the exploration of whether PWID with genetically similar HCV genes knew each other and were in frequent contact, and of how this seroconverted population actually transmits HCV. However, SNA affords the ability to determine answers to the above questions [Reference Hellard15]. The use of simulated contact networks generated from an empirically grounded network model to obtain data regarding HCV transmission and treatment represents a recently developed statistical approach [Reference Rolls27]. In this study, SNA of HCV infections suggests that HCV may be more easily spread in small groups of PWID, where one individual most likely acts as the main source of the spread of HCV. The case group network was much closer, numerically denser and more centralized than the control group network. PWID in the case group were more likely to resemble a small world, showing clustering within a few cliques. Highly central individuals have been targeted with prevention campaigns to become peer leaders in preventative interventions [Reference Latkin28]. The present study also highlights the potentially dual importance of people with high betweenness centrality. First, SNA can identify the HCV-infected PWID that have the highest probability of transmitting HCV to other participants, and can recognize the population at highest risk of transmission. In a sense, the central points can be referred to as potential ‘super-spreaders’. Second, people with high betweenness centrality could be used as effective peer educators in preventative interventions in order to reach various high-risk populations. Therefore, the use of questionnaires to identify contacts and the application of SNA to these data could potentially detect HCV-infected PWID with high betweenness centrality as well as identifying their network of contacts (i.e. other PWID with whom they interact). This would facilitate the targeting of interventions to high-risk populations in order to reduce HCV transmission.
The present study has a slightly smaller sample size than that typically used in the field of epidemiology. This was due to the recruitment of a susceptible group and represents one of the limitations of the study. Some of the participants from the previous follow-up study [Reference Zhou16] appear to have been lost to follow-up for a variety of reasons. Since drug trading is illegal in China, some PWID were in prison or in compulsory detoxification. Many PWID refused to participate because of the non-confidential nature of the SNA. Because of these limitations, the present study is merely exploratory. Another limitation is that SNA does not provide direct information about who injected with whom. Indeed, because drug injection is illegal, the response rate for this specific type of enquiry may be very low. However, since each injecting drug use territory was relatively small, the participants were likely to share needles only with people they knew. Furthermore, although the results suggest that participants from the same MMT clinic had closer contact, it was not possible to determine whether the familiarity started before or after registration at the clinic, and hence what role the clinic played in the establishment of the contact. Another limitation is that the socio-demographic, self-reported behavioural and SNA data were collected in April 2014, whereas blood tests for seroconversion were only performed between May 2006 and June 2011; thus, it is possible that some of the participants classified as not showing seroconversion (by June 2011) may in fact have shown seroconversion by the time the questionnaires were administered (April 2014). In addition, since the sample may have had different characteristics from the larger patient population from which it was taken, further work is needed to establish the generalizability of these findings. Additional, larger-scale studies are necessary to address these issues.
CONCLUSION
This is the first study of HCV-infected PWID in China using SNA and gene sequencing. It provides reliable and scientific information that may be useful for controlling HCV in PWID through social-network visualization and gene distance. It also motivates new research that will be able to use this innovative method of combining spatial contact with gene linkage. Due to the high prevalence of HCV in PWID compared to HIV, this study clearly describes the susceptible population and participants with high probability of transmission. These results might be of great importance to public health agencies. The findings also highlight the significance of future research on the effectiveness of interventions based on social networks in PWID, and on the importance of targeting gatekeepers and ‘bridges’.
SUPPLEMENTARY MATERIAL
For supplementary material accompanying this paper visit http://dx.doi.org/10.1017/S0950268816001333.
ACKNOWLEDGEMENTS
We thank Professor R. S. Schottenfeld and Professor M. C. Chawarski from Yale University School of Medicine for providing valuable suggestions for this programme.
DECLARATION OF INTEREST
None.
