Title: Relationships between body art piercing acquisition, availability of body art facilities, and
risk of hepatitis C acquisition among injection drug users
Authors : Julie Bruneau1,2, Mark Daniel1,3,4, Yan Kestens1,3, Geng Zang1
1.CRCHUM, Centre Hospitalier de l’Université de Montréal, Montréal, Québec, Canada
2 Department of Family Medicine, Université de Montréal, Montréal, Québec, Canada
3 Department of Social and Preventive Medicine, Université de Montréal, Montréal, Québec,
4 School of Health Sciences, University of South Australia, Adelaide, Australia
CRCHUM, Centre Hospitalier de l'Université de Montréal
264 René-Lévesque est, room 312
Montreal, QC CANADA H2X 1P1
TR: 514-890-8000 #35882
FAX : 514-412-7280
Manuscript : 3655 words; Abstract: 242.
Title: Relationships between body art piercing acquisition, availability of body art facilities, and
risk of hepatitis C acquisition among injection drug users
Keywords: body piercing, epidemiology, HCV, incidence, injection drug use
Background: Body art piercing (BAP) is considered to be a risk factor for hepatitis C (HCV)
infection on the basis of cross-sectional associations with HCV seropositivity among injection
drug users. The temporal basis of the relationship has not been established.
Methods: Associations between HCV seropositivity, HCV incidence, recent BAP and BAP
facility availability were evaluated among IDUs in Montreal, Canada, followed biannually
between 2004 and 2008. Statistical models included individual and neighbourhood covariates.
Logistic regression was used for analysis of HCV seropositivity. Cox proportional hazards
regression was used for analysis of HCV incidence.
Results: Of 784 IDUs, 73% were seropositive for HCV. In multivariate logistic regression, HCV
seropositivity was associated with BAP availability (OR: 1.32 95% Confidence Interval
(CI):1.1,1.6) but not recent BAP. Of 145 initially HCV-negative participants, 52 seroconverted
to HCV for an incidence of 27.7/100 person-years (95%CI: 20.9,36.0). Crude Hazard Ratios
(HR) for the association between HCV infection and BAP variables were: recent BAP, HR 0.98
(95%CI: 0.4,2.7) and BAP facilities availability, HR 1.43 (95%CI: 1.1,1.9). After accounting for
individual and neighbourhood factors, crude associations between HCV infection and recent
BAP and BAP facilities availability were: HR recent BAP, 0.96, 95%CI: 0.3,2.7; and HR BAP
facilities availability, 1.21, 95%CI:0.9,1.7.
Conclusion: BAP facility availability is a marker of neighbourhood disadvantage associated with
HCV seropositivity. Longitudinal analyses accounting for behaviour risk factors and
neighbourhood characteristics do not support a temporal association between BAP acquisition,
BAP facility availability, and HCV infection among IDUs.
Body piercing and tattoos have been used for thousands of years. Contemporary body art
involves puncturing the skin for tattooing or body piercing, collectively referred to as body art
piercing (BAP). BAP is still today most often associated with marginal groups. It is prevalent
among those reporting previous drinking and recreational drug use (Lauman & Derick, 2006),
and among prisoners with a history of injection drug use (IDU) (Hellard, Hocking, Crofts, 2004).
In recent decades however, BAP has grown in popularity among the general population in
developed countries (Armstrong, Koch, Saunders, Roberts & Owen, 2007). In the United States,
a national probability sample survey of 253 women and 247 men between 18 to 50 years of age
estimated that 24% of men and women had tattoos and 14% had body piercing (Lauman &
Since BAP involves needles and blood, it carries risks, ranging from infectious, allergic,
or granulomatous complications in connection with the tattoo pigment (Kazandjieva & Tsankov,
2007) to the acquisition of systemic blood-borne infections such as HIV and viral hepatitis
(Armstrong, 2005). Relations between BAP and hepatitis C (HCV) infection in marginal
populations have been the focus of several studies in recent years. BAP has been reported as
associated with HCV seropositivity among street youths (Roy et al, 2001), sex trade workers,
veterans (Briggs et al, 2001; Luksamijarulkul & Deangbubpha, 1997) and drug users.
(Scheinmann et al, 2007) In studies that accounted for injection drug use, BAP has been linked
to HCV prevalence among prisoners in Australia (Hellard et al, 2004), Norway (Holsen, Harthug
& Myrmel, 1993), and Germany (Backmund, Meyer, Wachtler & Eichenlaub, 2003).
IDUs are the population most at risk for HCV transmission in the developed world
(Baldo, Baldovin, Trivello & Floreani, 2008). Among IDUs, syringe sharing is the strongest
determinant of HCV seroconversion (Hagan et al, 2001). Hence for IDUs the relative
contribution of BAP to HCV acquisition may be modest. The availability of BAP facilities to
IDUs, many of whom live in high-risk inner-city areas is, however, a potentially relevant
exposure that could pattern HCV risk.
Blood-borne virus infections and high-risk injection behaviour among IDUs are not
distributed homogeneously within city boundaries (Maas et al, 2007). Individual injection drug
use patterns have been found to be associated with living in an economically disadvantaged
neighbourhood (Galea et al, 2003). Social disorder and neighbourhood crime are also associated
with spatial variation in risk for blood-borne infections (Cooper, Moore, Gruskin & Krieger,
2005; Latkin, Williams, Wang & Curry, 2005). Neighbourhood socioeconomic disadvantage
and social disorder are related (Bernard et al, 2007) and are themselves associated with other
environmental influences including quality of housing and roadways (Bernard et al, 2007),
availability of public and commercial services (Daniel, Kestens & Paquet, 2009), and,
potentially, availability of BAP facilities.
In this study, we assessed for IDUs living in an urban setting and participating in a cohort
study whether recent BAP acquisition and the availability of BAP facilities were associated with
having, and developing HCV. We proposed that BAP acquisition may be associated with HCV
prevalence and incidence. Using models that specified key individual and neighbourhood factors
we further hypothesized that HCV seropositivity would be associated with the availability of
BAP facilities, even accounting for other neighbourhood factors, but that HCV incidence would
be more strongly associated with neighbourhood disadvantage and social disorder than with
access to BAP.
The study population was drawn from the St. Luc Cohort, an open cohort of IDUs
established in Montreal in 1988, to study determinants of HIV transmission (Bruneau et al,
2001). Cohort recruitment criteria include being 18 years of age or older and having injected
drugs within the past month. In November 2004, the objectives of the cohort were expanded and
a new cohort was assembled to examine the individual-related and contextual factors associated
with HCV infection. IDUs already enrolled in the former cohort who had injected in the
previous month and who were HCV antibody negative at their first follow-up visit after
November 2004 were invited to participate, and accounted for 44% of the new cohort
participants. Other cohort participants volunteered to participate in response to street-level
recruitment or by word-of-mouth referral (30%), and through community programs (26%). All
participants provided an informed consent as approved by the institutional review board of the
Centre Hospitalier de l’Université de Montréal. Cohort visits were scheduled at 6-month
intervals. Behavioural questionnaires were administered by trained interviewers with venous
blood samples drawn at each visit for HIV and HCV antibody testing. Participants were asked to
return for their serostatus test results two weeks after their interview at which time post-test
counselling and referrals were provided. Participants were given a CAD $15.00 stipend at each
interview to compensate for their time.
Seven hundred eighty-four cohort members were included in the HCV seropositivity
analysis. One hundred forty-five participants, HCV-seronegative at enrolment, were followed up
at least once and included in the incidence analysis. All seroconverters had a documented
negative HCV test at the time of enrolment and a subsequent positive HCV test during a follow-
up visit. The first available questionnaire in the new cohort was used for analyses of
seropositivity. For analysis of HCV incidence, information from each previous visit was used
for seronegative participants whereas information from each previous visit for which HCV-
samples were negative was used for seroconverters.
The main outcome variable was HCV infection detected by the presence of HCV
antibodies. A positive HCV antibody test was determined by enzyme immunoassay assay (Abbott
Laboratories) and confirmed by RT-PCR (Roche Diagnostic Systems). Specimens with
indeterminate results were sent for confirmatory tests by dual EIA and/or RIBA (gold standard).
Recent acquisition of a BAP was assessed by the following question: “Have you had a tattoo or
body piercing in the past six months?” No information was available to assess the lifetime
acquisition of BAP. Other individual-level covariates included age, gender, recent
imprisonment, crack cocaine use and frequency of injections in the past month.
We used a geographic information system (Daniel & Kestens, 2007) to determine the
availability of BAP facilities and the nature of the neighbourhood conditions to which IDUs were
exposed. All BAP facility addresses on the Island of Montreal were located using a commercial
directory and their coordinates determined using GeoPinpoint geocoding software (DMTI Spatial
Inc, Markham, Ontario). IDU coordinates were determined as described in previous reports
(Bruneau, Daniel, Kestens, Zang & Genereux, 2008; Genereux, Bruneau & Daniel, 2010).
Briefly, IDUs were asked for the postal code corresponding to their dwelling place, defined as
the place where he/she most often slept during the past 30-days. For participants with no fixed
address, postal codes were determined for the street intersection closest to where such IDUs had
most often slept. Kernel density estimations (KDE) of BAP facilities were used to represent the
availability of BAP facilities at each dwelling location. KDE involves transforming a point
distribution into a continuous surface, thus providing density estimates at all locations of the
territory. KDE has been used in health services research and was recently proposed to help
quantify the risk environment in IDU research (Cooper, Bossak, Tempalski, DesJarlais &
Friedman, 2009). For our purposes, we used a quartic decay function and a fixed bandwidth of
1,500 metres, assuming a 20-minute maximum walking distance as relevant range for non-null
availability. This distance has previously been used as a service availability threshold among
IDUs (Bruneau et al, 2008; Hutchinson, Taylor, Goldberg & Gruer, 2000).
Two other local residential neighbourhood variables were examined: i) the percentage of
households below the low income cut-off before tax (LICO) for the census tract in which an IDU
resides, based on 2006 Canada Census data; and ii) the number of total crimes against the person
per 100 inhabitants for the police district in which an IDU resides, based on 2006 data from the
Service de police de la Communauté urbaine de Montréal. The LICO is a measure of
socioeconomic disadvantage established by Statistics Canada, representing income levels at
which a household spends 20% more than the area average of before tax income on food, shelter
and clothing (Zou, Forrester & Giulivi, 2003). The proportion of total crimes against the person
was used as a measure of social disorder (Friedman et al, 2006; Latkin et al, 2005). These
measures were expressed for the local residential neighbourhood, defined as a road network
buffer of 500-metres calculated using ArcGIS network analyst (using a trim distance of 100
meters along the road network) for each participant postal code, resulting in a buffer around each
participant’s place of residence. A weighted average of the LICO and number of crimes within
the 500-metre buffer of each participant was calculated using census tract and police district data
for areas over which the buffer overlapped, with weights corresponding to overlap area.
Logistic regression (LR) was used to estimate crude and adjusted odds ratios (OR) and
corresponding 95 percent confidence intervals (95%CI) for the relation between HCV
seropositivity and individual and neighbourhood exposures. Cox proportional hazards regression
was used to estimate crude and adjusted Hazard ratios (HR) and corresponding 95%CI to
examine the relations between recent BAP acquisition, BAP availability, and incidence of HCV
among subjects who were HCV negative at the entry in the cohort. BAP availability, crime and
LICO measures were expressed at the level of individual IDUs for their specific local
neighbourhood (defined, as noted, for a road network buffer of 500-metres around each dwelling
place). Variability in BAP exposure was accounted for in analyses by introducing a spatial term
of BAP density at each participant's dwelling place. Spatial autocorrelation measures were
computed for all models’ residuals to assess potential bias associated with spatial
autocorrelation. Spatial autocorrelation was non significant; Moran’s I ranged from 0.027-0.036
(p>0.19). As no spatial structure remained in model residuals, analyses were not biased by
spatial autocorrelation. Individual exposure measures, except gender, were modelled as time-
dependent covariates representing their most recent values. Four multivariable proportional
hazards models were estimated. All four models included the exposures of primary interest, but
different models adjusted their effects for a gradually extended set of covariates. The first model
included recent acquisition of BAP, BAP availability and other all individual-level covariates;
the second and third models included these same variables and either LICO (model 2) or crime
index (model 3). The final model included all individual and neighbourhood variables. All
analyses were conducted using SAS® v 9.1.
Of 784 cohort participants included in the HCV seropositivity analysis, 571 (72.8%) had
HCV. The majority were male (83.5%), with the mean age 38.1 years (SD=9.5). The average
duration of injection drug use was 14.2 years (SD 9.6). Analyses of potential risk factors (Table
1) revealed few statistically significant correlates of HCV seropositivity. In univariate analyses,
availability of BAP facilities was greater for HCV-seropositive IDUs compared to HCV-negative
IDUs. HCV seropositivity was not associated with either the LICO or crime index.
Figure 1 illustrates the density of BAP facilities on the island of Montreal and the
distribution of participants according to their HCV status. The highest density of BAP was in the
inner-city area of Montreal. Forty-four IDUs reported having acquired a BAP in the six months
prior to the study visit, with HCV-positive individuals having a lesser frequency of acquiring a
new BAP. No correlation was observed between the kernel density estimate of BAP availability
and recent BAP acquisition (r= -0.01; P=0.750). In the multivariable logistic model that adjusted
for neighbourhood and individual characteristics (Table 1), the association between HCV
seropositivity and recent BAP acquisition was not statistically significant.
Of 184 HCV-negative cohort members eligible for this investigation, 145 (79%)
participants were followed up at least once between November 2004 and December 2008 and
were included in incidence analyses (mean follow-up time 16 months (SD 11.4)). A total of 52
(28.2%) persons seroconverted to HCV during 187.7 person-years of observation, with the
incidence rate 27.7 per 100 person-years (95% confidence interval (CI): 20.9, 36.0). The mean
kernel density estimate of BAP availability for the dwelling place of HCV-incident IDUs was 0.9
BAP per square kilometre (inter-quartile range (IQR) = 1.8-0.1 (SD 0.9)). Positive correlations
existed between BAP availability and the LICO (r = 0.33; P < 0.0001), and crime index (r= 0.47;
P < 0.0001). As for analyses of seropositivity, BAP availability and recent BAP acquisition
were not correlated (r= -0.05; P=0.545).
Table 2 provides the results of unadjusted (crude) Cox’s regression analyses of
associations between neighbourhood and individual variables, and risk of HCV seroconversion.
Higher BAP availability was associated with increased risk of HCV seroconversion. Incident
HCV was associated with neighbourhood disadvantage (LICO measure) and neighbourhood
crime. Seventeen IDUs reported BAP acquisition at least once during the study period, and of
these, six seroconverted to HCV, with the crude hazard ratio 0.98 (95% CI: 0.4-2.7). Recent
imprisonment and a higher frequency of injection in the past month were significantly associated
with an increased risk of HCV infection.
Risk factors associated with HCV incidence in univariate analysis were examined in
multivariate models that sequentially adjusted for individual and neighbourhood variables
respectively. As shown in Table 3, the association between availability of BAP facilities and
HCV incidence hardly changed when adjusted for individual covariates, shifting from a Hazard
Ratio (CI) of 1.43 (95%CI: 1.1, 1.9) to 1.44 (95%CI: 1.1, 1.9), and remained statistically
significant. When neighbourhood variables were introduced in the model, the strength of the
association decreased substantially and the adjusted effect of BAP facilities became marginally
non-significant. On the other hand, the independent associations of HCV incidence with (i) the
frequency of injections in the last month, and (ii) recent imprisonment, remained statistically
significant even after adjustment for all individual and neighbourhood variables (model 4 in
Table 3). Finally, the LICO and crime index remained statistically significant in multivariable
models that adjusted for all other variables (models 2 and 3, respectively). However, their effects
became marginally non-significant when both variables were simultaneously adjusted for each
other (model 4), due to their correlation (r (LICO,crime index) = 0.60).
In this study, we found substantial variation in BAP availability across Montreal Island.
The inner-city of Montreal, characterized by the highest levels of socioeconomic disadvantage
and crime of all boroughs in Montreal, not unexpectedly had the highest BAP availability
(Savoie, Bédard & Collins, 2006). We observed no associations at baseline between HCV status
and neighbourhood socioeconomic disadvantage (LICO) and social disorder (crime index), or
with recent BAP acquisition or other recent risk behaviours. These results differ from those
reported for studies of younger IDUs but are comparable to those from studies of longer duration
and older IDUs (Backmund et al, 2003; Briggs et al, 2001). The lack of association between
individual behaviours, neighbourhood factors and HCV positivity is consistent with what might
reasonably be expected for a population of long-term IDUs, where HCV prevalence is close to
saturation (Falster, Kaldor & Maher, 2009).
Conversely, the distribution of HCV cases among IDUs varied with the availability of
BAP. Although we found no correlation between BAP availability and recent acquisition, the
proportion of IDUs (5.6%) engaging in BAP over a 6-month period is much higher overall,
compared to the general population. In national surveys, the lifetime prevalence of BAP was
estimated at 8% in Australia (Makkai & McAllister, 2001) and 10% in England (Bone, Ncube,
Nichols & Noah, 2008), with most BAP being acquired during early adulthood. Our data also
indicate that HCV-seropositive IDUs were less likely to have recently acquired a BAP, as
compared to seronegative individuals. This finding is not unexpected, as HCV-positive subjects
were on average 5 years older than HCV-negative subjects, and because BAP is more frequent
among younger individuals. Another explanation is the possibility that knowledge of their HCV-
positive status prevented IDUs from acquiring a new BAP to avoid transmitting HCV to others.
Studies that have addressed the potential role of HCV or HIV disclosure in changing behaviours
among IDUs have generally not, however, seen any corresponding impact on syringe sharing
practices (Brogly, Bruneau, Lamothe, Vincelette & Franco, 2002; De et al, 2009). These
findings suggest that BAP availability is a marker of neighbourhood disadvantage associated
with HCV positivity among longer-term IDUs, independent of the concentration of
socioeconomic disadvantage or social disorder, without being directly related, in this population,
to BAP acquisition itself. The location of BAP facilities may have evolved to serve a population
of marginal individuals, including older and chronic IDUs who are likely to be more frequently
Results from our incidence analysis suggest that the availability of BAP facilities is
simply one among other markers of adverse conditions that predict HCV acquisition in at-risk
IDUs. The association between BAP availability and was substantially reduced and lost its
statistical significance after adjustment for neighbourhood variables, indicating that the
availability of BAP is not an independent predictor of HCV incidence.
Studies by Rhodes have determined that social situations, place-based or geographic
influences not directly related to individual decision-making, can contribute to IDU risk (Rhodes,
Singer, Bourgois, Friedman & Strathdee, 2005). We recently demonstrated that IDUs living in
neighbourhoods with a high concentration of socioeconomic disadvantage in the inner-city of
Montreal were more likely to report high-risk injection behaviour (Genereux et al, 2010).
Results from the current study further suggest a modest effect of socioeconomic disadvantage
and social disorder in HCV transmission among IDUs.
As hypothesized, recent BAP acquisition was not associated with HCV seroconversion.
Of note, among the 17 participants who acquired a BAP within six months of their most recent
follow up, six seroconverted to HCV, and two IDUs reported having acquired their BAP while in
prison. Whether obtaining a BAP invokes an independent risk, or is rather correlated with
specific high-risk taking behaviours remains to be investigated. An investigation of this premise
would command a very large sample size as the attributable risk for injection drug use is likely to
Imprisonment in the past six months was independently associated with higher HCV
incidence. It has been demonstrated that having injected while in prison predicts HIV and HCV
infection (McGovern et al, 2006). Of 51 participants reporting recent imprisonment during the
study period, just one reported having injected drugs while in prison. While this individual did
seroconvert to HCV, this event occurred more than 18 months after his release from prison.
Besides the documented risk associated with injection drug use while in prison, heightened
vulnerability may play an important role in increasing high risk injection behaviours and HCV
acquisition after release (van Haastrecht, Bax & van den Hoek, 1998).
Consistent with previous studies, the strongest independent predictor of HCV incidence
was the frequency of injections in the past month (Tyndall et al, 2003). Importantly, the
association between the frequency of injection and HCV incidence was not affected by adjusting
in statistical models for other individual covariates, and/or for neighbourhood conditions,
indicating its robustness in predicting HCV, in contrast to the effect of the BAP availability
which became non-significant after these adjustments. Download full-text
Several limitations apply to this study. Participants were not randomly selected and thus
cannot be considered representative of IDUs in Montreal. The sample is over-represented in
terms of males and chronic cocaine-using IDUs, compared to Quebec provincial data on IDUs
(Comité permanent de lutte à la toxicomanie, 2000). Further, this study was conducted in the
specific setting of a French-speaking North-American urban city. If not having particular
equivalents elsewhere this context might still serve as a reasonable representation of availability
of BAP facilities relevant to IDUs in areas of drug use in different cities internationally. Even if
our follow-up rates were high for a drug-using population, our data could have been influenced
by losses to follow-up. No differences in age, gender, frequency of injection, recent
incarceration and recent BAP acquisition were found between participants included in analyses
and those lost to follow-up.
As demonstrated by several methodological papers, another potential type of selection
bias might have occurred in our prospective analyses of HCV incidence because most cases were
already prevalent at the time follow-up was initiated (Flanders & Klein, 2007; Hernan,
Hernandez-Diaz & Robins, 2004). This type of differential selection bias may be less likely to
apply, however, in studies such as occurs where subjects change exposure categories over time.
In the present study, a weak correlation was found between exposure levels at different times (for
example, we obtained r= 0.309 between BAP availability at the initial and the latest assessment,
corresponding to less than 10% variance in the last BAP explained by the initial BAP). It is
uncertain, therefore, as to whether those participants already seropositive at their baseline, and
thus excluded from the prospective analyses of HCV incidence, would be more likely to belong
to the ‘exposed’ or the ‘un-exposed’ during the study follow-up. Furthermore, this particular