Demographic, risk behaviour and personal network variables associated with prevalent hepatitis C, hepatitis B, and HIV infection in injection drug users in Winnipeg, Canada

The likelihood of exposure to bloodborne pathogens is a multifactoral process primarily dependent on the risk behaviours an individual practices and the likelihood that a susceptible individual will come into contact with an infected individual (thus increasing or decreasing the risk actually associated with a risk behaviour). The interactions that bring susceptible and infected individuals into contact with each other occur within the context of social networks, the overall structure of which can also affect the rate of pathogen spread [1, 2]. Chance events also play a role as a social network may contain a group of individuals whose behaviours favour transmission, but until such time as a bloodborne pathogen enters that network, they face no risk of acquiring that pathogen.

Individual risk behaviours for injection drug users (IDU) include those directly associated with transmission, such as the use of syringes previously used by another IDU or those behaviours which can act as markers of the above types of behaviours. Examples of risk markers that are positively associated with disease prevalence include drug scene roles, such as dealing drugs or injecting others as a service (street or hit doctors) [3‐6]. Other behaviours, such as obtaining clean needles from questionable sources such as drug dealers, shooting gallery owners, or on the street, can also act as a marker of an increased probability of using a contaminated needle, as doctoring of used needles to make them appear new has been reported [7].

Social network analysis is a technique which measures and analyzes the interactions that occur between people and its application can enhance an otherwise overly simplistic interpretation of individual risk behaviours. In Brooklyn, injecting with a used syringe was associated with HIV only if syringes were being shared with someone who was 10 or more years older than the interviewed subject, or was a daily injector [8]. By incorporating network variables indicative of a higher probability of coming into contact with an infected individual, a better indication is obtained of who is at risk of infection and what actually constitutes a high-risk behaviour.

Subsequent research on IDU and their network characteristics have identified other network variables associated with transmission risk. High-risk injection practices have been linked to network characteristics such as the number of network members [9, 10]; the presence of family members or spouses within the network [4, 11]; higher network density [10]; the setting where injection takes place [10]; turnover of network members [12]; and the pooling of financial resources within networks for the purpose of obtaining drugs [9]. Racial/ethnic differences in HIV prevalence have also been at least partially explained by taking into account the differing network characteristics of different ethnic groups [13].

Some individual behaviours or characteristics may also be proxy markers of network behaviours. The type of drug an IDU chooses to inject can be measured as a characteristic of the individual injector and itself can influence risk, as some drugs, like cocaine, are prepared at room temperature and hence are more conducive to pathogen survival [14]. In addition to these more risky drug-specific practices, IDU may form networks based on drug type, which mark the broader social network within which an individual is a member [15, 16]. Therefore, network members are more likely to come into contact with whichever pathogen(s) happen to be circulating within that network. Similarly, moving to a new city within the past year can also be an indicator of higher risk as individuals create a social bond through the sharing of drug equipment to try and establish themselves in new networks [12, 17].

We have recently described networks centered on downtown hotels in the core area of Winnipeg (manuscript in press), which demonstrate that the setting where an individual injects drugs can also act as a proxy marker of a social network in which some of the members may not know each other, but are nonetheless linked by a common venue. Within these venues accidental or intentional use of contaminated syringes is more likely. Injection at a shooting gallery is another example of a geographic place-based network that has also frequently been linked to increased risk of infection by a bloodborne pathogen [6, 18, 19].

The majority of the network research described above has focused on HIV and relatively little on hepatitis B and C (HBV and HCV, respectively). Comparative studies of network variables associated with the transmission of HIV, HCV and HBV within the same setting have also not been conducted. Our goal was to determine whether respondent and network variables (including some proxy markers of networks as described above) are associated with the prevalence of HCV, HBV, and HIV in our setting and to what extent similar or dissimilar patterns emerge for the different pathogens.

A total of 435 IDU were enrolled in the study. Some IDU were unable or unwilling to provide a blood specimen. Additionally, five individuals who self-identified as transgender and 10 individuals who self-identified as belonging to ethnic groups other than Caucasian or Aboriginal, were excluded from the analysis. For the remaining 369 respondents, summary statistics for the egocentric network are as follows: mean number of members, 8.8; median, 8; range, 1–20, lower quartile, 5; upper quartile, 11; SD, 4.6. Given that the IDU risk network was truncated at 5 IDU, we also report these summary statistics for the total number of people within the egocentric network identified as IDU: mean number of IDU in the egocentric network, 3.5; median, 3; range, 0–20; lower quartile, 1; upper quartile, 5. In total, 17% of respondents (66/369) reported more than 5 IDU in their egocentric network. The lack of data on these individuals is discussed further in the limitations section of this paper. Of the 369 respondents considered for analysis, several indeterminate results were identified for each pathogen therefore, the final sample sizes for HCV, HBV, and HIV were 365, 364, and 360, respectively.

Our two primary goals in this investigation were to determine which respondent and/or network variables were associated with HCV, HBV, and HIV in our setting and to compare and contrast these variables for the different pathogens. Only two respondent variables were positively associated with all three pathogens: injection with syringes previously used by another IDU and injection at a shooting gallery. The former represents a direct route for parenteral transmission of blood borne pathogens and is clearly a common factor for transmission of all three of the pathogens examined in our study area. The latter, given the high-risk activities characteristic of this type of venue [6, 18, 19], would be associated with an increased likelihood of exposure through purposeful or accidental use of contaminated equipment. Although this variable is a marker for high-risk behaviour on the part of individuals, it is also a proxy marker for membership in high risk, largely anonymous networks of IDU. Networks form around specific places where geographic place is the common connecting tie between individuals [22‐24]. Including place in network analysis adds explanatory power for understanding disease transmission regardless of the infection. Here shooting galleries are not simply high risk environments for individual users, but, by acting as repositories for contaminated equipment, they create bridging opportunities for pathogens between individuals who may not know each other and who may never otherwise meet.

Like shooting galleries, certain hotels in Winnipeg are also important venues for injection and favor formation of place-based networks of IDU. The likelihood of HBV infection increased proportional to the number of IDU the respondent knew who had injected at a hotel. Association with HBV, but not with HCV or HIV, could result from HBV entering the hotel network(s) earlier than HCV and HIV, resulting in more extensive spread. Alternatively, there may be specific behaviours associated with this group of IDU that favour specific transmission of this pathogen (e.g. hotel rooms, as opposed to shooting galleries, may offer more privacy and a greater likelihood of engaging in sexual activities, thus favouring HBV transmission over HCV). Regardless of the causal factors involved, like others [8, 25], we found that incorporating interactions with risk network members provided a clearer understanding of what actually constitutes high-risk behaviour.

Respondents' injection of talwin/ritalin and their use by IDU network members was a marker for HCV and HBV. The type of drug used can loosely demarcate distinct networks of IDU [15, 16]. Within these networks specific behaviours may favour transmission (e.g. the room temperature production of cocaine vs. the need to heat some other types of drugs [14]). To our knowledge, extensive use of talwin/ritalin is relatively rare outside of the Canadian provinces of Manitoba and Saskatchewan. It is notable that in our analysis, injection with used syringes interacted with talwin/ritalin such that the prevalence of HCV was high amongst talwin/ritalin users regardless of whether or not they reported injection with used syringes. Like cocaine, talwin/ritalin is typically prepared at room temperature and frequently involves communal use of the prepared drug and filters, which may account for the high prevalence of HCV regardless of syringe sharing practices (typically pills are crushed in room temperature water; IDU then use a common filter to draw drug into their own respective syringes). It is also notable, that HIV does not show a positive association with talwin/ritalin in contrast to HCV and HBV. As above, this may reflect a different temporal framework for pathogen entry to the network of talwin/ritalin users. If this is the case, the same behaviours which facilitated spread of HBV and HCV may ultimately lead to high HIV prevalence in this group.

Two IDU risk network variables were negatively associated with either HCV or HBV. For HBV, there was a negative association with the number of IDU in the respondent's risk network who had initiated them to injection drug use, while for HCV there was a negative association if the respondent reported that there were IDU in their risk network who had shown them how to inject drugs. For HCV, there was also a weak interaction between the number of IDU members who had shown the respondent how to shoot drugs and syringe sharing, such that syringe sharing was no longer positively associated with prevalence. In general, these network variables may mark small, relatively closed networks of IDU who may share equipment and syringes, but rarely admit new members. Low network turnover has been associated with lower risk [12] as opportunities for entry of pathogens to these networks (via infected individuals) would be correspondingly reduced. An area worthy of research would be whether the lower prevalence of infection is a chance byproduct of the social interactions within these networks or whether the individuals within closed networks are actively and knowingly attempting to isolate themselves from infection risk.

The greater likelihood of HIV and HBV being transmitted through sexual contact was consistent with sex partner-related variables being retained in the final models for these two pathogens. However, while a positive correlation for HIV occurred with same-sex client partners, the directionality for HBV and opposite-sex casual partners was opposite. Although the direction of this relationship appears counter-intuitive, it may reflect a greater likelihood to engage in safer sex harm reduction activities with sex partners of this type. This question could be addressed in more detailed research on sexual contact behaviours in this population.

Aboriginal ethnicity and a recent move to Winnipeg were associated with HIV and HBV, respectively. HIV was the only pathogen where ethnicity remained as an independent predictor of risk in the final model. Research in other cities have identified specific social network interactions that can account for ethnic differences in disease prevalence [13]. Further research would be necessary within our study population to identify what risk behaviour(s) may be associated with HIV infection in the aboriginal population in Manitoba that may not have been captured by our survey. The relatively small number of HIV positive individuals in our dataset also may have limited the analytic potential for this pathogen in regression analysis. A recent move to Winnipeg was associated with a lesser likelihood of HBV positivity. This pattern may reflect a lower prevalence of HBV and/or earlier or more extensive deployment of vaccination programs in the provinces from which these IDU originated (e.g. the province of British Columbia began vaccinating approximately 6 years prior to Manitoba).

A number of limitations for the study should be noted. First, the study was cross-sectional and hence does not provide any data on trends or whether currently observed behaviours were a result of, rather than the cause of, an infection by one of the pathogens studied. Second, the data were egocentric and hence no independent validation of the behaviours of IDU in the risk network was possible. Third, network members were elicited based on contact in the previous 30 days, but questions regarding individual behaviours and interactions with network members were typically collected for the previous 6 months (or longer periods for some variables such as "ever-use of someone else's syringe"). Time frames are relevant, given the noted lack of correlation between some known risk behaviours and serostatus in our analysis (see paragraph below). Any comparisons of our results with those from other areas would need to take note of the time frames used. Fourth, the egocentric and IDU risk networks were truncated at 20 and 5 people, respectively. The implications of this data truncation are greatest for the IDU risk network variables as they formed a large part of the analysis. Overall, 17% of respondents reported more than 5 IDU in their egocentric network. We felt that full data collection on all IDU in the egocentric network was impractical in terms of time, complexity of data collection and accurate recall by respondents, however, it should be noted that this additional data could have revealed additional patterns not apparent in our results.

Finally, some limitations regarding specific variables should be noted. First, data on sharing of drug-preparation equipment and transfer of prepared drugs between syringes was available for only the six months prior to interview. In contrast, sharing of used syringes was based both on six-month and "ever" data. Only the latter was a predictor of risk, therefore, if more extensive data had been collected for equipment sharing and drug transfers, they may also have been significant predictors of risk of infection for some, or all, of the pathogens studied. Second, only a limited number of sexual behaviour questions were included in the study and no specific questions on type of sex were asked. Collection of data on anal sex, in particular, could have revealed other correlations not evident with the dataset available. Third, only "injecting someone else as a service" was analyzed; "injecting someone else as a favor" or "receiving injections from others" can be considered distinct traits and could also be related to the common use of syringes and pathogen transmission.

Our findings reinforce the clear link between respondent behaviours (e.g. injection with someone else's used syringe) and risk of infection. However, they also highlight the importance of defining the types of personal networks that form within an area and identifying the different patterns of transmission of pathogens within those networks. In our setting, pathogen transmission is associated with loosely defined, sometimes anonymous, high-risk networks formed around specific venues (shooting galleries, hotels) or between users who are linked by their drug use preferences. Smaller, somewhat isolated pockets of IDUs appear to exist within the larger population where behavioural patterns pose a lesser risk, unless, or until, a given pathogen enters those networks. Temporal differences with respect to pathogen introduction to the different networks likely influences which specific behaviours appear as protective or as high-risk, highlighting the importance of verifying the extent to which risk behaviours or markers identified as important in one setting or to one pathogen apply to other settings or pathogens. The lack of any apparent correlation between the prevalent infections for the three pathogens, noted in the results section, reinforces the need to identify the different social and temporal patterns associated with the various pathogens that may be circulating within a population. Both individual- and network-level concepts assist in characterizing and comparing the transmission of HCV and HBV, as well as, HIV.

A potential hierarchy of risk emerges, defined by space and time. Individuals may have some independence in their decisions to use drugs or not, or share syringes or not. However, the broader personal network in which they find themselves, whether defined by living near a hotel where drug use is common or by their ethnic or age group, is no less important in contributing to their potential for initiation of specific behaviours or exposure to infectious agents. The smaller, more specific social interactions with their IDU risk network forms an additional step in a progression which may determine an individual's choice of drug; their norms, rituals and/or practices of injection; and their likelihood of exposure to infected individuals. Both individual- and network-level concepts assist in characterizing and comparing the transmission of HCV, HBV, and HIV and determining the underlying patterns that drive the social connections between individuals that may favour or hinder transmission of specific pathogens.