The provision of non-needle/syringe drug injecting paraphernalia in the primary prevention of HCV among IDU: a systematic review

Background Sharing drug injecting paraphernalia other than needles and syringes (N/S) has been implicated in the transmission of Hepatitis C virus (HCV) among injecting drug users (IDU). We aimed to determine whether the provision of sterile non-N/S injecting paraphernalia reduces injecting risk behaviours or HCV transmission among IDU. Methods A systematic search of seven databases and the grey literature for articles published January 1989-February 2010 was undertaken. Thirteen studies (twelve observational and one non-randomized uncontrolled pilot intervention) were identified and appraised for study design and quality by two investigators. Results No studies examined the association between the provision of non-N/S injecting paraphernalia and incident HCV infection. One cross-sectional study found that individuals who frequently, compared to those who infrequently, used sterile cookers and water, were less likely to report prevalent HCV infection. Another found no association between the uptake of sterile non-N/S injecting paraphernalia and self-reported sharing of this paraphernalia. The remaining observational studies used attendance at needle and syringe exchange programmes (NSP) or safer injection facilities (SIF) that provided non-N/S injecting paraphernalia as a proxy measure. Eight studies presented adjusted odds ratios, ranging from 0.3 to 0.9, suggesting a reduced likelihood of self-reported sharing of non-N/S injecting paraphernalia associated with use of NSP or SIF. There was substantial uncertainty associated with these estimates however. Three unadjusted studies reported a reduction in the prevalence of sharing of non-N/S injecting paraphernalia over time among NSP users. Only one study reported an adjusted temporal trend in the prevalence of sharing non-N/S injecting paraphernalia, finding higher rates among non-NSP users than NSP users at each time point, and a greater reduction in sharing among non-NSP than NSP users over time. Study limitations included the use of convenience samples, self-reported exposure and outcome measures, flawed classification of the exposed and unexposed groups, and inadequate adjustment for potential confounding variables. Conclusions The evidence to demonstrate that the provision of sterile non-N/S injecting paraphernalia reduces HCV transmission or modifies injecting risk behaviours is currently limited by an insufficient volume and quality of studies. Further research is required to inform practice and policy in this area.


Background
The sharing of drug injecting paraphernalia, equipment used in the preparation and administration of drugs for injection, is common and has been implicated in the transmission of Hepatitis C (HCV) [1][2][3]. There is compelling evidence that HCV transmission occurs through the sharing of contaminated needles and syringes (N/S). This is biologically plausible; in the act of preparing and administering drugs for injection, N/S will generally come into direct contact with blood. The prevalence of HCV is high among those who report injecting drug use [4] and the risk of HCV transmission has been independently documented in studies of needle-stick injury in health care workers [5,6].
Evidence relating to the risk of HCV transmission associated with sharing non-N/S injecting paraphernalia is less well established however. Ethnographic studies have highlighted numerous opportunities for crosscontamination to occur when individuals share drug preparation equipment other than N/S [7][8][9][10]. Laboratory studies have isolated HCV RNA from injecting equipment including spoons used as drug cookers (also called stericups), filters (also called cottons or sterifilts) and water samples [11]. The self-reported sharing of such items of paraphernalia is high, irrespective of the setting or population studied. Between 65% to 84%, 50% to 77%, and 15% to 83% of intravenous drug users (IDU) report sharing drug cookers, filters and water, respectively [12].
To date, few longitudinal observational studies have detected an association between HCV incidence and the sharing of non-N/S injecting paraphernalia. A recent systematic review concluded that evidence relating the sharing of such drug injecting paraphernalia to HCV transmission was 'not overwhelming' and highlighted several methodological limitations of the studies that contribute to this evidence base [13]. A high baseline prevalence of HCV among IDU and poor study retention has limited the statistical power that many studies have had to detect an association between incident HCV infection and isolated, often highly correlated, injecting practices.
In the United Kingdom (UK) primary and secondary health care is provided, free at point of access, by the National Health Service (NHS). Local authorities are responsible for meeting nationally agreed targets on health and social care but retain considerable autonomy over the allocation of resources. Following legislative change in 2003, N/S exchange programmes (NSP) in the UK are now able to provide clients with sterile items of drug injecting paraphernalia in addition to N/S. There is however wide local and regional variation in the provision of such paraphernalia by NSP [14,15]. The effectiveness, and indeed cost-effectiveness, of providing of sterile injecting paraphernalia other than needles and syringes in the primary prevention of HCV among IDU has yet to be established [16]. To inform future public health policy and practice in the UK and elsewhere, we conducted a systematic review to determine whether the distribution of sterile drug injecting paraphernalia other than needles and syringes reduces the risk of HCV transmission or modifies injecting risk behaviours.

Methods
We searched MEDLINE, MEDLINE In-Process & Other Non-Indexed Citations, Cochrane Central Register of Controlled Trials, Cochrane Database of Systematic Reviews, Database of Abstracts of Reviews of Effects, EMBASE and PsycINFO for articles published between January 1989 and February 2010 using the search strategy outlined in Appendix 1. The grey literature was searched using the terms 'Hepatitis C' or 'HCV' and 'paraphernalia'. Reference lists of selected articles were reviewed and citation checks carried out to identify further potentially relevant studies. Two investigators (MG, NP) reviewed the abstracts of potentially eligible articles to determine relevance; where relevant full texts were retrieved and examined.

Inclusion and exclusion criteria
Injecting paraphernalia was defined as equipment used in the preparation or administration of drugs for injection. For the purposes of this review we limited this definition to drug cookers, filters and water, as these are the items of injecting paraphernalia most likely to be contaminated with blood in the course of preparing or administering drugs [7][8][9][10]. The exposure of interest was the distribution of drug cookers, filters and/or water; however, the self-reported uptake of these items or the use of an NSP or safer injection facility (SIF) that provided these items was accepted as a proxy measure of exposure. Studies that did not provide one or more of these items of paraphernalia or that did not explicitly state which items of paraphernalia were provided, were excluded. The outcomes of interest were (i) incident HCV infection (ii) prevalent HCV infection and (iii) injecting risk behaviours, namely the self-reported sharing of drug cookers, filters and/or water. The study population of interested was current IDU. Descriptive studies, qualitative studies, review articles, editorials and opinion pieces were excluded. Only primary research studies were considered. Where two or more articles presented data from the same cohort during overlapping time frames, the publication with the largest sample size was selected for inclusion.

Study Appraisal
We used the STROBE [17] criteria as a guide to consider key aspects of study design: sampling methods, definition of study population, inclusion and exclusion criteria, assessment of exposure, assessment of outcome (s), completeness of follow-up, treatment of missing data and treatment of possible confounding variables. Each item was assessed in terms of how well the area was addressed and reported by investigators in relation to the aims of this review. Study quality was assessed independently by two reviewers (MG, NP) with disagreement resolved by a third reviewer (SH).

Cohort Studies
Five cohort studies examining over 5,000 IDU between 1994 and 2005 were included in this review [19][20][21][22][23] ( Table 2). The largest study followed 2,814 IDU, recruited in Seattle, USA from 1994 through 1997, over a one year period [19]. The authors found no significant difference in the sharing of 'cookers or cottons' in the month prior to follow-up interview between those who reported 'ever' compared to 'never' using an NSP at study enrolment, with an adjusted odds ratio (AOR) of 0.79 (95% CI 0.58 -1.08). Huo et al documented a high prevalence of sharing 'other injecting paraphernalia' in the 30 days prior to interview in both NSP users (60%) and nonusers (70%) at baseline in their cohort of 901 IDU recruited in Chicago, USA between 1997 and 2000 [20]. Across the 3 year study the self-reported prevalence of sharing 'other injecting paraphernalia' fell to almost identical levels in both NSP users and nonusers (approximately 22%). Following adjustment, NSP users were less likely, across all study visits, to report sharing 'other injecting paraphernalia' in the preceding month compared to non-users (AOR 0.7, (0.52 -0.95)). Sears and colleagues compared the self-reported sharing of 'cottons, cookers or water' in the 30 days prior to interview at baseline, 6 months and 1 year follow up in NSP users (n = 132) and non-users (n = 97) recruited in San Francisco, USA in 1993 [21]. Due to differential attrition rates according to NSP status, bivariate analyses were limited to cases that completed all three interviews (n = 101, 44% of the study population). NSP users were "significantly" more likely than non-users to report sharing a 'cotton, cooker or water' at baseline (61% NSP users vs. 40% non-users, p value not reported). A decline in the prevalence of sharing a 'cotton, cooker or water' at follow up was reported in both NSP users (to approximately 55% at 6 months and 45% at 1 year) and nonusers (to approximately 40% at 6 months and 30% at 1 year); p-values were not provided. In multivariate analyses there was an interaction between NSP status by study visit for the sharing of a 'cotton, cooker or water' was reported. AOR stratified by NSP status were not presented, however the authors concluded "nonexchangers showed a larger decrease than exchangers in this behaviour over time." More contemporaneous data (2004 -2005) from Vancouver, Canada explored changes in injection practices associated with use of a medically supervised safer injection facility (SIF) [22]. Consistent SIF users (n = 433) -individuals that reported using the SIF for greater than 25% of all injecting episodes -were more likely than inconsistent users (n = 327) to report a change to using clean water for injecting since they started using the SIF (AOR 2.99 (2.13 -4.18)). Lastly, Vlahov and colleagues carried out a before -after study following a cohort of 422 IDU    from enrolment in a newly established NSP in Baltimore, USA over a six month period [23]. The authors observed a significant reduction in the proportion of participants that reported sharing cotton filters in the two weeks prior to interview, from 46% at baseline to 31% at six month follow up visit (P < 0.001). This trend was also observed for cooker sharing, for which the corresponding values were 59% and 39%, respectively (P < 0.001). Adjusted analyses were not presented.

Serial Cross-sectional Studies
Bluthenthal et al conducted serial cross-sectional studies in seven semi-annual waves over a three year period in two neighbourhoods in California, USA in which an illegal NSP distributed injecting kits containing alcohol wipes and cotton filters [24]. Between the first and last waves there was a statistically significant reduction in the reported prevalence of sharing 'other injection supplies' (cookers, filters or rinse water) in the preceding  six months (from an estimated 62% to 40% respectively; P < 0.001) and a concurrent increase in the proportion of those who reported receiving 'other injection supplies' from an HIV prevention provider (from an estimated 18% to 24%, P < 0.004) as well as an increase in reported NSP use (from 5% to 36% respectively; P < 0.001). However, the study found no statistically significant difference in the odds of sharing 'other injection supplies' in the preceding 30 days among those who reported receiving sterile injecting paraphernalia in the preceding 30 days compared to those who did not (AOR 1.11, (0.89-1.39)) ( Table 2). Nor was there a statistically significant difference in the sharing of 'other injection supplies' in the preceding 30 days between those who reported NSP use and those who did not (AOR 0.85 (0.68-1.07)).

Cross-sectional Studies
The only study identified in this review that correlated the use of sterile non-N/S injecting paraphernalia to (self-reported) HCV infection examined a convenience sample of predominantly male, cocaine injecting IDU (n Percentage sharing cooker during last day of injection 16% vs. 6% b a cohort studies did not report risk ratios: see Results section for details, b not significant, c P < 0.05, d P < 0.001, e odds ratios presented here use the group with lower frequency of NSP use as the referent group (as compared to those presented in the text), f intervention site were recruited from an area in which a secondary exchange operated, comparison site were recruited from another location near which two other NSP operated, g result not presented, h practices absent at baseline therefore p values for test against null hypothesis that proportions at follow up < 15% = 275) recruited from NSPs in Montreal, Canada [25]. Almost one quarter of the study population reported using sterile cookers (23%) and filters (23%) and three quarters (75%) using sterile water for half or more of all injecting episodes in the preceding month. Over a third (37%) reported sharing any non-N/S drug preparation equipment (cookers, filters or water) in the 6 months prior to interview. IDU who reported frequent ('half the time or more' vs. 'less than half the time') use of sterile cookers, filters and water in the month prior to interview were less likely to report sharing these items of paraphernalia in the preceding 6 months (AORs 0.42 (0.22 -0.80)), 0.52 (0.28 -0.97)) and 0.51 (0.29 -0.90)) respectively). The overall prevalence of self-reported HCV infection was 66%. Individuals reporting frequent, compared to infrequent, use of sterile cookers (AOR 3.92 (1.58 -9.70)) and water (AOR 2.93 (1.12 -7.68)), but not filters (AOR not provided), in the preceding month were significantly more likely to be self-reported HCV-negative. Of the remaining cross-sectional studies, three [26][27][28] undertook multivariate analyses, although only one appropriately adjusted for potential confounding variables and presented the results [26]. Longshore et al examined injecting risk behaviour according to frequency of NSP attendance (once per month or less, two to four times per month and more than four times per month) [26]. Infrequent NSP users (once per month or less) were significantly more likely to report sharing a cooker in the preceding six months than frequent NSP attendees (more than four times per month) (AOR 2.55 (1.05 -6.17)). However there was no significant difference between infrequent and frequent NSP attendees in self-reported sharing cotton filters in the preceding six months (AOR 1.57 (0.66 -3.75)), despite the NSP providing this item of paraphernalia. Kipke et al compared injecting risk behaviours in the preceding six months in a convenience sample of young IDU described as NSP users and nonusers, although these groups were poorly defined [27]. Following adjustment for age, gender and ethnicity, NSP users were significantly less likely to report 'sharing cotton, cookers or water' in the preceding six months than nonusers (AOR 0.53 (0.28 -0.99)). Sears et al examined injecting risk behaviours in young homeless IDU recruited from two areas in San Francisco, USA, an intervention site at which a peer-led secondary exchange programme was in operation (n = 67) and a comparison site near which two sanctioned primary NSPs and one clandestine NSP were in operation (n = 55) [28]. Participants recruited from the intervention site were less likely to report using someone else's cotton in the 30 days prior to interview than participants recruited from the comparison site (27% vs. 49%, p = 0.003). This difference did not, however, persist following adjustment for potential confounding variables (AOR not reported).
Two further cross-sectional studies were reviewed [29,30]. In their multisite study, Heimer et al reported a statistically significant difference between the sharing of water for drug preparation (3% vs. 9% respectively; p < 0.001) and syringe rinsing (3% vs. 9% respectively; p < 0.005) in the 30 days prior to interview between NSP users and nonusers, but no difference in the sharing of cookers (9% vs. 12% respectively) or filters (6% vs. 8% respectively) between groups (p values not provided) [29]. Adjusted analyses according to NSP use were not presented and it was noted that NSP had recently aggressively begun distributing sterile water; the other items of paraphernalia provided by NSP were not described in detail. Finally, Guydish et al described the frequency of NSP use and the mean percentage of syringes obtained from NSP in relation to self-reported sharing of rinse water (a proxy measure for the sharing of cookers, cottons or rinse water) in the preceding 30 days [30]. The authors reported no significant difference in the mean number of NSP visits in the preceding 30 days among participants reporting sharing rinse water compared to those who did not (4.36 vs. 3.75 visits to the NSP in the preceding 30 days respectively). However, those who shared rinse water reported obtaining significantly less syringes from an NSP (86%) compared to non-sharers (89%) (P < 0.05).

Intervention study
Colon et al evaluated a non-randomized, uncontrolled pilot study of a 16 week intervention that aimed to reduce contamination of injecting paraphernalia through the promotion of novel drug preparation practices and devices (a water bottle with dropper for preparing drug solution and rinsing syringes, a N/S pre-fitted with a sterile filter and a hand sanitizer for cleaning hands and injection sites) [31]. Two weeks following the end of the intervention, 56% of study participants reported using water from the water bottle in preparing drug solution, 53% using water from the bottle to rinse a syringe, 66% using the hand sanitizer and 34% using the syringe with pre-fitted filter for at least 75% of all injecting episodes in the day prior to interview and 'most of the time' or 'always' in the 7 days prior to interview; levels indicating potential for self-sustaining change in behaviours. There was no statistically significant reduction in the reported sharing of cookers in the week prior to interview between baseline (16%) and follow up visits (6%) (p = 0.453). However a 66% reduction in the likelihood of detecting red blood cells on items of paraphernalia (water containers and cookers) collected from shooting galleries during the intervention, compared to prior to the intervention, was reported.

Summary of key findings
This review aimed to explore the evidence base around the provision of non-N/S sterile injecting paraphernalia as an intervention in the primary prevention of HCV among IDU. We did not identify any studies that examined the relationship between the supply of injecting paraphernalia other than N/S and biological measures of HCV infection. One study related selfreported HCV infection to the use of sterile cookers, filters and water, finding that those who frequently used sterile cookers and water but not filters were more likely to self-reported negative HCV status than those who infrequently used them [25]. Inferences about temporality cannot be drawn from this study due to its cross-sectional design. The remaining observational studies reported behavioural outcomes. Eight studies presented adjusted odds ratios for the association between exposure to an NSP or SIF and sharing injecting paraphernalia other than N/S [19][20][21][22]24,[26][27][28]. Effect size estimates were suggestive of a reduction in the odds of sharing injecting paraphernalia other than N/S associated with exposure to NSP or SIF, but confidence intervals were wide and often included unity. One study found no significant association between the receipt of non-N/S sterile injecting paraphernalia and the sharing of these items [24]. Studies that examined unadjusted temporal trends in the prevalence of sharing non-N/S injecting paraphernalia [20,21,23,24] reported significant reductions over time, usually coinciding with an increase in NSP use. However, the only study to report an adjusted temporal trend found that prevalence rates of sharing injecting paraphernalia other N/S were lower at each time point in non-NSP users compared to NSP users [21]. The authors report a greater decline over time in non-NSP users compared to NSP users although data to support this statement are not provided in the manuscript.

Limitations of the current literature
The only non-observational study included in this review was a non-randomized, uncontrolled pilot intervention study with few participants and no long-term follow up [31]. Whilst providing encouraging data to support larger scale intervention trials, these results should be interpreted with caution. Of the observational studies, seven employed cross-sectional designs [24][25][26][27][28][29][30], which are limited with regards to the causal inferences that can be drawn. It is possible that IDU who engaged in high risk behaviours were also less likely to use NSP or SIF. In studies examining temporal trends in the prevalence of sharing non-N/S injecting paraphernalia [20,21,23,24], changes in observed injecting practices may be attributable to contemporaneous interventions, such as the delivery of harm reduction advice through NSP, SIF or other service providers.
A constraint exists between the aim of this review and the primary aim of the studies we identified. Examining the effect of providing sterile non-N/S paraphernalia to IDU was often a secondary rather than primary aim. This is reflected in the design, analyses and reporting of these studies. Few studies actually measured the provision or uptake of paraphernalia and, with two notable exceptions [24,25], exposure to NSP was invariably adopted as a proxy measure for uptake of the injecting paraphernalia offered by NSP. Three studies measured the frequency [22,26] or probability [19] of NSP (or SIF) use, but several studies used a more crude binary measure of NSP "users and non-users". The distinction between NSP users and non-users was often unclear and in some cases, both individuals designated as NSP users and non-users had access to the sterile paraphernalia under study [20,24,27,28]. For example, in the study by Huo et al, clean filters and water were distributed to all participants [20], whilst in the study by Sears and colleagues participants were classified according to the site of recruitment rather than reported NSP use. In the latter study, use of "any NSP" was almost ubiquitous among participants recruited from the intervention site and as high as 80% of the participants recruited from the comparison site [28]. Misclassification of exposure status may have attenuated any observed association between NSP use and injecting risk behaviours.
As has been shown in the studies included in this review, the sharing of individual items of injecting paraphernalia is often highly correlated. Half of all studies selected the sharing of a single or multiple items of injecting paraphernalia to represent all indirect sharing in their analyses [19][20][21]24,27,30]. Huo et al for example investigated the association between NSP use and the sharing of 'other injecting paraphernalia', denoting sharing of either a cooker, filter or rinse water [20]. Thus, it is not possible, in this review, to isolate the impact of paraphernalia provision on the sharing of any specific item of paraphernalia.
All studies used interviewer administered questionnaires or structured interviews to ascertain injecting risk behaviour and may therefore be subject to recall and social desirability bias [19][20][21][22][23][24][25][26][27][28][29][30]. It is reported that IDU reliably describe injecting risk behaviours [32] however ethnographic studies suggest that IDU may not be aware of episodes of indirect sharing [7]. Differential reporting of sharing practices could result in residual confounding which would affect any reported association.
Only two studies recruited a random sample of IDU [19,22]; the remaining studies adopted a targeted [24,29,31], systematic sampling strategy [21,23,30] or recruited convenience samples [20,[25][26][27] and often offered financial incentive to participants [19,[21][22][23][25][26][27][28]30]. All the studies identified were conducted in North America and several examined specific groups of IDU (for example homeless youth [28] or predominantly male, cocaine injectors [25]), in particular settings or contexts (for example SIF [22] or secondary syringe exchange programmes [28]). IDU are not a homogeneous group. Differences in the patterns of sharing of drug injecting equipment, the perceived risks associated with this and use of NSP have been reported according to key sociodemographic such as gender [33]. In addition important differences in structural and organisation components of NSP, the provision of and access to health care and the wider socio-political climate that may limit the generalisability of these findings to other populations.
A number of studies failed to adjust for potential confounders [23,29,30], or adjusted for only socio-demographic characteristics in statistical analyses [27]. It is likely therefore that any effect from these studies will be subject to residual confounding. Effective sample sizes of the included studies ranged from 32 to 1582. The larger studies [19,20,24] were able to estimate effect sizes with more precision, but given that no sample size calculations were presented, it is uncertain whether studies were adequately powered to detect an effect. Additionally, given that injecting behaviours were highly correlated, it is possible that even the largest studies would have had insufficient power to detect an association between an isolated injecting risk behaviour and NSP attendance. Indeed the largest of the cohort studies, Hagan et al experienced substantial losses to follow-up and analysed data on only 56% of all participants enrolled at baseline which will have further reduced ability to detect an association [19].

The wider context
There is limited evidence that HCV transmission occurs through the indirect sharing of injecting paraphernalia [13]. Nevertheless the theoretical risks of HCV transmission through this route have been recognised for over a decade [11] and high rates of sharing other injecting paraphernalia have consistently been reported. Indeed the sharing of cookers, filters and water is often reported to be much more common than the sharing of N/S. Many of the challenges of undertaking research in this area have been highlighted in the methodological limitations of the studies included in this review. Harm reduction interventions such as NSP are complex and as such demonstrating the effectiveness of each active component of the intervention is challenging. In addition, access to sterile injecting paraphernalia other than N/S at NSP or other settings does not necessarily translate into uptake or use of paraphernalia by IDU. Qualitative studies from the UK and elsewhere suggest that many IDU do not fully appreciate the risks associated with sharing non-N/S injecting paraphernalia [33]. Situational (convenience, ease of access) and social factors (knowledge of injecting partners, pooling of resources) [25,33], and levels of satisfaction with, and perceived ease of use of [25], sterile injecting equipment are also important in determining use. A further challenge lies in demonstrating the impact of an intervention on HCV transmission: the prevalence of HCV in IDU populations is generally high, therefore identifying a large cohort of HCV naïve IDU to prospectively study, and ensuring low rates of attrition, can be difficult. However new developments in the serosurveillance of IDU [34,35] offer the potential to examine the impact of interventions on recently acquired HCV infection using a cross-sectional design [34,35].

Limitations of this review
Given the heterogeneity in study designs, settings, populations and outcomes, we have not been able to present an overall measure of effect in this review. We conducted a thorough search of the literature, including the grey literature, to identify all potentially relevant material. However a number of studies were excluded because a detailed description of the injecting supplies provided by NSP was not available; improved reporting in primary studies may have yielded more evidence. This review examined biological (prevalent and incident HCV infection) outcomes, although this did not yield any studies, and behavioural (injecting risk behaviours) outcomes. It is important to acknowledge that the use of sterile non-N/S injecting paraphernalia may have additional benefits that have not been captured in outcomes considered in this review; for example, the potential to reduce bacterial infections or to attract and engage IDU who may not otherwise be in contact with health services [18]. Policy decisions on the distribution of non-N/S paraphernalia should not therefore be made on the basis of evidence relation to HCV transmission or injecting risk behaviour in isolation.

Conclusions
Current evidence suggests that attendance at NSP providing sterile non-N/S injecting paraphernalia may be associated with reduced sharing of non-N/S injecting paraphernalia. However, the evidence is limited by the number and quality of the studies. Therefore robust conclusions cannot be drawn in relation to the impact of providing non-N/S paraphernalia on injecting risk behaviours. No studies were found to have examined the impact of providing sterile non-N/S injecting paraphernalia on the transmission of HCV. We have identified a critical gap in the literature. Future studies would benefit from longitudinal designs, accurate measurements of the exposure variable as opposed to the use of proxy variables, adjustment for known and potential confounders and should be adequately powered to detect changes in injecting risk behaviour or incident HCV infection between exposed and unexposed groups. These data are critical to inform future public health policy and guide decision making by service providers and users.