The HIV, HCV, and HBV prevalence detected among Kabul IDUs from 2007 to 2009 was not substantially different from prevalence estimates obtained in 2006 . Though factors believed to predispose to a rapid epidemic, such as poverty, unemployment, and insecurity have been consistently present, the explosive increase in HIV prevalence noted between annual cross-sectional studies in Pakistan and other settings has not yet occurred [12–14].
The association between HIV and HCV and lifetime needle or syringe sharing is consistent with 2006 findings and potentially indicates a sustained level of risk behavior among the population. However, reported recent sharing of either needles/syringes or injecting works were not associated with either infection. The reported levels of lifetime and recent sharing among Kabul IDUs in 2006 were much higher, as 50.4% reported ever sharing needles or syringes and 31% reported sharing injecting equipment in the six months prior to enrollment [7, 15]. We hypothesize that harm reduction service expansion and inclusion of field-based NSP services has both decreased sharing and increased HIV transmission knowledge. However, the large decrease in reported lifetime sharing may indicate also that sharing is becoming a stigmatized behavior and is subsequently under-reported. The national drug use survey from 2009 reported lifetime needle sharing of 87% among IDUs; this survey included rural areas and noted a concentration of injecting in the southern provinces . Few harm reduction programs operate in these provinces, where the population is predominantly rural, limiting the ability to disseminate information on the dangers of needle sharing. Since needle and syringe sharing remained strongly associated with prevalent HIV and HCV, there remains room for improvement in terms of harm reduction expansion in Kabul.
Initiating injecting outside Afghanistan was also associated with both HIV and HCV infection, likely reflecting increased likelihood of transmission in neighboring countries with higher HCV and HIV prevalence (e.g. Iran, Pakistan) [13, 16]. The 2005 Kabul study did not assess where drug use and injecting were initiated, though injecting was assumed to be a behavior acquired outside Afghanistan . As Afghans living as refugees may not have the same access to harm reduction/NSP services in countries of refuge, sharing needles and syringes may have been more likely, as observed in Iran .
HCV was associated with history of injection site abscess, potentially reflecting rushed injecting which may be a proxy measure for sharing of equipment [18, 19]. In formative work, IDUs stated that relative speed of administration and ability to conceal use, particularly from police, were possible reasons for initiating injecting . This same rationale may exist for rushed injecting. Having had an abscess at the injection site within the last year was associated with HCV antibody among IDUs in the United Kingdom in a cross-sectional study,  but a similar association was not observed in a Canadian study .
HBV infection was not associated with sharing needles/syringes or injecting in prison, as in 2006, but was negatively associated with current NSP use. As HIV and HCV were not associated with NSP use, it seems less likely that this may be a proxy for fewer risky sharing behaviors. However, the recent expansion of NSPs in Kabul and the greater virulence and ease of transmission associated with HBV may provide an earlier sign of positive behavior change associated with NSP use; forthcoming incidence data will be assessed for impact of NSP use on HBV and other infection incidence over time.
There was higher reported lifetime use of harm reduction programs, specifically NSP services, than measured in the 2006 study. As data presented here and in 2006 are cross-sectional and utilized convenience sampling, some of this difference may be attributed to undersampling NSP users in 2006 or oversampling this group during cohort recruitment. However, an increased number of harm reduction programs and introduction of primary exchange services in the field, now the preferred method for NSP delivery, have debuted and may contribute to this change .
Current NSP users represent IDUs who, in many respects, are at greater risk for blood-borne infection due to initiating drug use with injecting, recent sharing of injecting equipment, and more frequent injecting daily. That IDUs using NSP services are more likely to have risky injecting practices has been noted in other settings, such as the United Kingdom and Canada [23–25]. The data indicate that these services are reaching their target clientele; however, service delivery in this setting is a work in progress and efficacy remains an open question. NSP users were substantially less likely to perceive great need for treatment, potentially indicating NSP use is an entrée to addiction treatment. This relationship has been established in other settings and several harm reduction programs in Kabul have incorporated abstinence-based treatment programs [26, 27]. As opioid substitution treatment (OST) is scaled up, harm reduction programs will need to ensure that linkages remain in place to treatment [26–28]. Further, programs should incorporate a low threshold model consistent with harm reduction rather than an abstinence based approach which has unrealistic expectations regarding the recovery process. The negative association between NSP use and having lived outside the country in the last five years may represent difficulties experienced by recently-repatriated IDUs in accessing services of which they may not be aware.
This study has several limitations. Participants were enrolled over a lengthy period by convenience sampling and may not be representative of IDUs in Kabul. Further, no data were recorded on those ineligible or declining entry, potentially leading to under-representation of hidden or isolated groups. We elected not to use respondent-driven sampling or other chain-referral methods in this setting for two reasons. First, based on qualitative work preceding cohort enrollment in Kabul, drug users perceive that injection drug use results in better services and greater opportunities for compensation . Memorably, we were informed by one participant that press photographers would pay IDUs to be photographed or videotaped while injecting. By choosing and compensating seeds and recruits, we were concerned that drug users might inadvertently initiate injecting due to compensation, which may weigh on perceived ability to meet basic needs in this impoverished setting . Similar concerns have been voiced by researchers in Lebanon . We did not monetarily compensate our participants and also ensured that medical care, including VCT services and naloxone, were available to non-participants upon request. Next, in insecure environments where frequent migration outside and within the city occurs, networks are transient and assignment of seeds may not guarantee recruit derivation from the same network. This network fragility may then prompt "seeds" to recruit IDUs unknown to them, compromising the validity of RDS and increasing risk for coercive practices between seed and recruiter.
Socially desirable response may have occurred, particularly with sensitive behaviors; we attempted to reduce this through staff familiar to the participant population and providing choice as to location of the study interview. Though audio computer-assisted survey interviewing has been noted to improve self-reporting among IDUs, this technology was not financially possible or feasible in this environment where there was often no power and the team was largely field-based . HIV-related analyses were underpowered due to low prevalence, potentially masking some associations. There were no female IDUs enrolled, precluding characterization. Female IDUs exist in Kabul but access to this hidden group has been elusive as the few women using harm reduction services are not injectors and require home visits due to cultural proscriptions.