Transmission Risk From Sharing Drug Injecting Paraphernalia
Transmission Risk From Sharing Drug Injecting Paraphernalia
We found that PWID who reported sharing N/S (+/− paraphernalia) and paraphernalia (only) had approximately seven and three times, respectively, the odds of recent HCV infection of those who did not share in the last 6 months. Although we could not isolate the risk associated with N/S sharing due to too few participants reporting solely this behaviour, the fact that the majority of individuals who reported sharing N/S also reported sharing paraphernalia suggests that the risk from the latter could be higher than the odds ratio of 3 that we observed.
Our effect size for sharing paraphernalia is consistent, if slightly higher, than effect sizes reported in previous studies. These studies reported RRs ranging from 1.1 to 2.7, although the confidence intervals for most of these included our estimate. In addition, previous studies found a significant association with water, which was not replicated in our analysis. It is of note that we presented ORs, rather than RRs; however, we also calculated RRs in univariable analyses, and these were comparable to the ORs.
In the sample presented in this study (2008/09 and 2010 combined), 7% reported sharing N/S in the previous month, which is relatively low in comparison with other countries with comparable injecting equipment provision (IEP) services: 17% of PWID surveyed at specialist services in England, Wales & Northern Ireland (EW&NI) in 2011 and 16% of PWID surveyed at IEP sites in Australia in 2011. By contrast, 30% of our study population reported sharing paraphernalia in the last month; the comparable figures for EW&NI and Australia are 20% and 16%, respectively. The risk associated with paraphernalia sharing determined in this analysis, combined with the high prevalence of this behaviour among Scottish PWID, means that it could potentially account for a substantial number of new HCV infections in this population. Previous studies have suggested that the proportion of infections attributable to paraphernalia sharing could be in the region of 19–51%.
During the Hepatitis C Action Plan for Scotland, there were approximately sixfold and fourfold increases in the provision of filters and containers, respectively, between 2008/09 and 2009/10 (financial years). Given these increases, our finding that the prevalence of sharing paraphernalia was significantly lower in 2010 than in 2008/09 is notable. Further NESI surveys will provide data to determine trends in this risk behaviour.
This is the first study to examine the association between paraphernalia sharing and recent HCV infection using a cross-sectional design. This study design is generally considered inferior to a longitudinal design because the former usually cannot measure incident infection, a limitation that we have overcome with our approach to detecting recent infections. Another frequently cited limitation of cross-sectional studies is that it is not known whether the exposure precedes the outcome. This may be true here, as we ascertain injecting risk behaviour (the exposure) in the 6 months prior to interview, and it is not known where in the 6-month period the behaviour occurred. However, this presents no disadvantage in comparison with longitudinal studies of risk behaviour and BBVs, for which establishing the temporal association is also a problem: these studies often follow up individuals at 6-month intervals – the infection could have been acquired at any time during the interval and the exact timing of the risk behaviour is similarly not known. Although cross-sectional studies may be subject to sampling bias, they avoid bias due to participant attrition that can arise in longitudinal cohort designs. Previous studies of paraphernalia sharing and incident HCV have lost substantial numbers of participants to follow-up, and although a description of dropouts was not always provided, in some cases, they tended to engage in riskier behaviour than those who remained in the study. The result could be an underestimate of the association between sharing paraphernalia and incident infection.
To determine the independent effects of paraphernalia sharing with regard to incident HCV, it is important to adjust for N/S sharing. Although many previous studies adjusted for N/S sharing, not all have done so. Furthermore, few studies have undertaken a stratified analysis (i.e. restricted the analysis to individuals who did not share N/S), as we did here, which can be a more reliable method of eliminating confounding due to N/S sharing than simply adjusting for it in a multivariate model.
In our study population, there were large overlaps between sharing of different paraphernalia items and thus the models examining individual items could not be adjusted for the other items. We therefore cannot state that the effect size associated with sharing containers is independent of sharing spoons/water, and vice versa. Despite this, we did not find an association between recent HCV infection and sharing water. A few epidemiological studies have detected significant associations between sharing water (for rinsing) and HCV transmission, and a recent laboratory study demonstrated that HCV can survive for up to 3 weeks in bottled drinking water and remain infective (depending on the initial dose). Water can become contaminated when it is used to flush/rinse out a syringe; it poses a transmission risk when it is subsequently reused by someone else for making a drug solution or rinsing out a needle/syringe (which is then used for injection). Twenty-six per cent of the sample reported sharing water in the last 6 months, but we do not know which of the latter two behaviours this refers to. It is plausible that the sharing of water for mixing is less risky, because a small volume of water is used, and the subsequent heating of the drug solution may deactivate some of the virus. The concentration of virus in the contaminated flush/rinse water may be correlated with the volume of residual blood left in the syringe after an injection, corresponding to the syringe's 'dead space'. The relative proportion of high/low dead space syringes used by the study population might therefore influence the transmission risk associated with water sharing. Our study did not collect information on syringe type; therefore, we are unable to examine this risk factor.
Limitations of this study include the use of self-reported measures of exposure. Evidence from studies that have compared computer-assisted self-interviewing to interviewer-administered questionnaires suggests that PWID underreport 'sensitive' behaviours, such as sharing N/S and other equipment. Although there may be a degree of underreporting, the NESI surveys have taken steps to reduce this, such as employing independent, trained interviewers, undertaking the interviews in a private area and collecting the information anonymously. Corroboration of the behavioural data collected in the NESI surveys is provided by the strength of the association between N/S and recent infection seen here, which is larger than previous studies in which inconsistencies have been observed, and by the dose–response relationship that we observed between risk behaviour and recent HCV infection, such as increasing odds of recent infection corresponding with increasing reported frequency of N/S sharing (e.g. ORs of 3.9 and 5.4 among those who shared once/twice and ≥3 times, respectively, in the last 6 months, relative to those who reported not sharing).
The preseroconversion window period for detection of recent HCV infection – with an estimated mean duration of 6–8 weeks – has implications for this study. The studies that generated these window period estimates mostly involved plasma donors as study subjects, and therefore, the validity of the estimates for PWID is less certain. Many factors can influence the duration of this acute phase of infection, including demographics and exposure frequency. However, the accuracy of the actual duration of the window period is perhaps less important in this study, where we are not concerned with estimating incidence per se. Nevertheless, the short duration of the window period means that the outcome is rare: in this study, the 35 recent infections limited the statistical power.
Misclassification of the outcome must also be considered. A small number of recent infections may have been false positives in the case of delayed seroconversion among immunocompromised persons (i.e. chronic infections that were misclassified as recent). Similarly, due to fluctuating viraemia among a small proportion of acutely infected individuals, a recent infection could have been misclassified as being RNA-negative during a low viraemic phase. Both of the latter could result in a biased estimate of the association between the exposure and outcome. While re-infections among individuals who had previously cleared the virus would ideally be included in our outcome, these would not be captured by our measure of incident infection due to the presence of antibodies from previous exposure to HCV among these individuals. This could result in a dilution of the exposure–outcome association as these individuals would likely be positive for the exposure (sharing injecting equipment) but negative for the outcome. Finally, a further source of misclassification could be the laboratory tests to detect HCV antibodies and HCV RNA on dried blood spots. Although these have high sensitivities (99% and 100%, respectively) and specificities (100% and 96%, respectively), there is nevertheless the possibility of false antibody negatives and false RNA positives.
We cannot exclude the possibility that other unmeasured risk factors (for example, tattooing) may account for some of the risk of HCV infection. There may also be a role for other items used in the injecting process – for example, swabs – in the transmission of HCV, but we did not collect information on the sharing or reuse of items other than the ones presented here. It is also notable that the ORs presented here are not related to a 'per event' risk, because we did not have information on the frequency of paraphernalia sharing; however, we attempted to address this by adjusting for variables related to frequency of injecting in the models.
In conclusion, we have demonstrated that a cross-sectional design generates results that are similar to those from longitudinal studies of the association between sharing injecting paraphernalia and incident HCV infection. We present the first European study to examine this association. Our study confirms that, as has been observed in other western countries, the prevalence of paraphernalia sharing is high and represents significant potential for HCV transmission among PWID.
Discussion
We found that PWID who reported sharing N/S (+/− paraphernalia) and paraphernalia (only) had approximately seven and three times, respectively, the odds of recent HCV infection of those who did not share in the last 6 months. Although we could not isolate the risk associated with N/S sharing due to too few participants reporting solely this behaviour, the fact that the majority of individuals who reported sharing N/S also reported sharing paraphernalia suggests that the risk from the latter could be higher than the odds ratio of 3 that we observed.
Our effect size for sharing paraphernalia is consistent, if slightly higher, than effect sizes reported in previous studies. These studies reported RRs ranging from 1.1 to 2.7, although the confidence intervals for most of these included our estimate. In addition, previous studies found a significant association with water, which was not replicated in our analysis. It is of note that we presented ORs, rather than RRs; however, we also calculated RRs in univariable analyses, and these were comparable to the ORs.
In the sample presented in this study (2008/09 and 2010 combined), 7% reported sharing N/S in the previous month, which is relatively low in comparison with other countries with comparable injecting equipment provision (IEP) services: 17% of PWID surveyed at specialist services in England, Wales & Northern Ireland (EW&NI) in 2011 and 16% of PWID surveyed at IEP sites in Australia in 2011. By contrast, 30% of our study population reported sharing paraphernalia in the last month; the comparable figures for EW&NI and Australia are 20% and 16%, respectively. The risk associated with paraphernalia sharing determined in this analysis, combined with the high prevalence of this behaviour among Scottish PWID, means that it could potentially account for a substantial number of new HCV infections in this population. Previous studies have suggested that the proportion of infections attributable to paraphernalia sharing could be in the region of 19–51%.
During the Hepatitis C Action Plan for Scotland, there were approximately sixfold and fourfold increases in the provision of filters and containers, respectively, between 2008/09 and 2009/10 (financial years). Given these increases, our finding that the prevalence of sharing paraphernalia was significantly lower in 2010 than in 2008/09 is notable. Further NESI surveys will provide data to determine trends in this risk behaviour.
This is the first study to examine the association between paraphernalia sharing and recent HCV infection using a cross-sectional design. This study design is generally considered inferior to a longitudinal design because the former usually cannot measure incident infection, a limitation that we have overcome with our approach to detecting recent infections. Another frequently cited limitation of cross-sectional studies is that it is not known whether the exposure precedes the outcome. This may be true here, as we ascertain injecting risk behaviour (the exposure) in the 6 months prior to interview, and it is not known where in the 6-month period the behaviour occurred. However, this presents no disadvantage in comparison with longitudinal studies of risk behaviour and BBVs, for which establishing the temporal association is also a problem: these studies often follow up individuals at 6-month intervals – the infection could have been acquired at any time during the interval and the exact timing of the risk behaviour is similarly not known. Although cross-sectional studies may be subject to sampling bias, they avoid bias due to participant attrition that can arise in longitudinal cohort designs. Previous studies of paraphernalia sharing and incident HCV have lost substantial numbers of participants to follow-up, and although a description of dropouts was not always provided, in some cases, they tended to engage in riskier behaviour than those who remained in the study. The result could be an underestimate of the association between sharing paraphernalia and incident infection.
To determine the independent effects of paraphernalia sharing with regard to incident HCV, it is important to adjust for N/S sharing. Although many previous studies adjusted for N/S sharing, not all have done so. Furthermore, few studies have undertaken a stratified analysis (i.e. restricted the analysis to individuals who did not share N/S), as we did here, which can be a more reliable method of eliminating confounding due to N/S sharing than simply adjusting for it in a multivariate model.
In our study population, there were large overlaps between sharing of different paraphernalia items and thus the models examining individual items could not be adjusted for the other items. We therefore cannot state that the effect size associated with sharing containers is independent of sharing spoons/water, and vice versa. Despite this, we did not find an association between recent HCV infection and sharing water. A few epidemiological studies have detected significant associations between sharing water (for rinsing) and HCV transmission, and a recent laboratory study demonstrated that HCV can survive for up to 3 weeks in bottled drinking water and remain infective (depending on the initial dose). Water can become contaminated when it is used to flush/rinse out a syringe; it poses a transmission risk when it is subsequently reused by someone else for making a drug solution or rinsing out a needle/syringe (which is then used for injection). Twenty-six per cent of the sample reported sharing water in the last 6 months, but we do not know which of the latter two behaviours this refers to. It is plausible that the sharing of water for mixing is less risky, because a small volume of water is used, and the subsequent heating of the drug solution may deactivate some of the virus. The concentration of virus in the contaminated flush/rinse water may be correlated with the volume of residual blood left in the syringe after an injection, corresponding to the syringe's 'dead space'. The relative proportion of high/low dead space syringes used by the study population might therefore influence the transmission risk associated with water sharing. Our study did not collect information on syringe type; therefore, we are unable to examine this risk factor.
Limitations of this study include the use of self-reported measures of exposure. Evidence from studies that have compared computer-assisted self-interviewing to interviewer-administered questionnaires suggests that PWID underreport 'sensitive' behaviours, such as sharing N/S and other equipment. Although there may be a degree of underreporting, the NESI surveys have taken steps to reduce this, such as employing independent, trained interviewers, undertaking the interviews in a private area and collecting the information anonymously. Corroboration of the behavioural data collected in the NESI surveys is provided by the strength of the association between N/S and recent infection seen here, which is larger than previous studies in which inconsistencies have been observed, and by the dose–response relationship that we observed between risk behaviour and recent HCV infection, such as increasing odds of recent infection corresponding with increasing reported frequency of N/S sharing (e.g. ORs of 3.9 and 5.4 among those who shared once/twice and ≥3 times, respectively, in the last 6 months, relative to those who reported not sharing).
The preseroconversion window period for detection of recent HCV infection – with an estimated mean duration of 6–8 weeks – has implications for this study. The studies that generated these window period estimates mostly involved plasma donors as study subjects, and therefore, the validity of the estimates for PWID is less certain. Many factors can influence the duration of this acute phase of infection, including demographics and exposure frequency. However, the accuracy of the actual duration of the window period is perhaps less important in this study, where we are not concerned with estimating incidence per se. Nevertheless, the short duration of the window period means that the outcome is rare: in this study, the 35 recent infections limited the statistical power.
Misclassification of the outcome must also be considered. A small number of recent infections may have been false positives in the case of delayed seroconversion among immunocompromised persons (i.e. chronic infections that were misclassified as recent). Similarly, due to fluctuating viraemia among a small proportion of acutely infected individuals, a recent infection could have been misclassified as being RNA-negative during a low viraemic phase. Both of the latter could result in a biased estimate of the association between the exposure and outcome. While re-infections among individuals who had previously cleared the virus would ideally be included in our outcome, these would not be captured by our measure of incident infection due to the presence of antibodies from previous exposure to HCV among these individuals. This could result in a dilution of the exposure–outcome association as these individuals would likely be positive for the exposure (sharing injecting equipment) but negative for the outcome. Finally, a further source of misclassification could be the laboratory tests to detect HCV antibodies and HCV RNA on dried blood spots. Although these have high sensitivities (99% and 100%, respectively) and specificities (100% and 96%, respectively), there is nevertheless the possibility of false antibody negatives and false RNA positives.
We cannot exclude the possibility that other unmeasured risk factors (for example, tattooing) may account for some of the risk of HCV infection. There may also be a role for other items used in the injecting process – for example, swabs – in the transmission of HCV, but we did not collect information on the sharing or reuse of items other than the ones presented here. It is also notable that the ORs presented here are not related to a 'per event' risk, because we did not have information on the frequency of paraphernalia sharing; however, we attempted to address this by adjusting for variables related to frequency of injecting in the models.
In conclusion, we have demonstrated that a cross-sectional design generates results that are similar to those from longitudinal studies of the association between sharing injecting paraphernalia and incident HCV infection. We present the first European study to examine this association. Our study confirms that, as has been observed in other western countries, the prevalence of paraphernalia sharing is high and represents significant potential for HCV transmission among PWID.