Human immunodeficiency virus infection and related risk behavior in people who inject drugs in Nepal

This observational study retrospectively analyzed IBBS surveys of PWID in Nepal, conducted from 2002 to 2015. IBBS surveys were cross-sectional in design and were conducted in Kathmandu Valley, Pokhara Valley, eastern Terai (3 districts), and western-to-far-western Terai (7 districts). PWID were defined as ‘males current injectors aged 16 years and above who have been injecting illicit drugs for at least three months prior to the date of the survey’. However, no data from PWID younger than 18 years old was included in the present study. IBBS surveys employed respondent-driven sampling (RDS) to recruit PWID in the Kathmandu valley and Pokhara valley. However, RDS was used in eastern Terai (3 districts) in 2002 and 2005 and in west to far western Terai (7 districts) in 2003. Two stage cluster sampling was used as the alternative to RDS in eastern and western-to-far-western Terai from 2007 onwards, because of the lack of strong network among the PWID in these regions. IBBS surveys were conducted in compliance with both ethical and human rights standards. Ethical approval for the surveys was obtained from Nepal Ethical Review Board, of Nepal Health Research Council, which approved the current study (registration number 03/2016; Ref No. 1232). In the previous IBBS surveys, documented and witnessed informed verbal consent (to preserve anonymity) was obtained from PWID before their interview and the collection of blood samples for the IBBS. Study centers with laboratories and clinics were established at easily accessible locations in all the study districts. Individual interviews, clinical examinations, and blood collection were conducted in separate rooms in each of the study centers.

IBBS datasets for year and region were available as computer files comprising background characteristics, knowledge on HIV and AIDS, drug injecting practices, sexual behavior and access to HIV services. The independent variables selected were background characteristics (age, year, region, education, and marital status), drug injecting practices (age of first drug injection, duration of drug use and injecting drugs, used needle/syringe previously used by someone else in past week, used syringe/needle left in a public place in past week, and shared needle/syringe with someone after using in past week) and sexual behaviors (age at first intercourse, number of sexual partners, number of FSWs, consistent condom use with regular partner, FSW and nonregular female sex partner). The sex partners of the PWID were categorized as regular female sex partners, FSWs, and nonregular female sex partners. A ‘regular female sex partner’ is defined as spouse or any sexual partner living together with PWID. FSWs were defined as those who sell sex in exchange for cash, kind, or drugs. ‘Non-regular female sex partners’ were defined as those with whom the PWID was not married or living together. These may include girlfriends or other female friends with whom they have sexual relationship.

Statistical analyses were performed using the R program. Bivariate analyses were performed to estimate the association of demographic and behavioral variables with HIV prevalence using χ² tests. Logistic regression [12,13] analyses were performed to determine variables associated with HIV infected proportion defined by combinations of the determinants, using the additive model:

In this model, P is the expected probability of HIV infection, α is the intercept, x_i through x_k are determinants variables, β_i are regression coefficients. This model also provides confidence intervals for HIV infection for levels of each risk factor adjusted for other risk factors using sum contrast methods [14,15,16]. The confidence intervals based on sum contrasts has the advantage that they provide a simple criterion for classifying levels of the factor into 3 groups according to whether each corresponding confidence interval exceeds, crosses, or is below the overall percent. The confidence intervals compared percent of the specified cause group in each category with the overall percent. To compare models for prevalence we used the area under the receiver operating characteristic (ROC) curve as measure of goodness-of-fit and a constructed mosaic plot. ROC curve sensitivity against the false positive rate to show how well the model predicts a binary outcome. For a range of decision choices, it plots sensitivity (probability of finding an outcome when it is there) against the false positive error rate (probability of finding an outcome when it is not there). It also provides a mosaic plot for comparing logistic regression models similar to an r² decomposition plot.

Data from a total of 7,073 PWID from 2002 to 2015 were included in the analysis, of whom 17.8% (1,257) tested HIV positive. Table 1 examines the association between background characteristics and HIV infection. Year, region, age, education, and marital status were significantly associated with HIV (P <0.001). HIV infection has significantly decreased from 68% in 2002 to 6.2% in 2015. By region, Kathmandu Valley had highest prevalence (30.5%) followed by Eastern Region (18%). HIV infection was found to be associated with older age; the prevalence being 32% in those aged ≥30 years. HIV infection also varies by education; PWID who had no education (28.5%) and primary education (23.8%) had higher prevalence than secondary and above (15%). HIV infection among the married was 24%, which exceeds that of the unmarried (13.3%).

Table 1

Associationbetween characteristics and HIV infection among people who inject (illicit) drugs

Table 2 shows the prevalence of risky behaviors in this population and the results of bivariate analyses between HIV infection and drug injecting behaviors. Risky behaviors are highly prevalent in this population. Age of first drug initiation, duration of drug use. duration of injecting drugs, previously used needle/syringe in past week, syringe/needle left in a public place in past week, and sharing a needle/syringe with someone after using it in past week, were significantly associated with HIV infection (P <0.001). HIV infection was higher among the PWID who had their first drug initiation when ≥30 years, followed by those aged 20–29 years. HIV infection significantly increased with duration of drug use and injecting practices. PWID whose duration of drug use exceeded 10 years had prevalence of 32%, and those who had injected drugs for more than 10 years had a prevalence of 35%. HIV infection prevalence was higher among the PWID who used unsafe sharing practices in past week, for example, PWID who used needle/syringe previously used by someone else was 31%, used syringe/needle left in a public place had 39%, and those who shared a needle/syringe with someone after using it had 30%.

Table 2

Association between drug injecting behaviors and HIV prevalence among people who inject (illicit) drugs

Age of first sexual intercourse, number of sexual partners, number of FSWs, consistent condom use with regular partners, FSWs and nonregular female sex partners were significantly associated with HIV infection, as shown in Table 3 (P <0.001). HIV was higher among PWID who had first sexual intercourse at 20-29 years (25%) followed by those below 20 years (17%). Similarly, HIV infection was higher who have no sexual intercourse in past year (25%) and no sex with FSW in past year (20%). HIV infection was higher among PWID who had consistent condom used with regular female sex partners (23%) and had no sex with FSW (20%) and had no sex with nonregular female sex partners (22%).

Table 3

Association between sexual behaviors and HIV prevalence among people who inject (illicit) drugs

In bivariate analysis, all determinants were found to be significantly associated with prevalence of HIV infection. Therefore, these determinants were included for multivariate logistic regression analysis. In the multivariate logistic regression, year, region, age, education, duration of drug use was significantly associated with HIV infection. However, the study found an interaction between year and region and age and education. Therefore, year and region were combined. There were 22 levels in the year-region group factor depending on year and region. Age and education were also combined. The number of levels in the age group-education factor depends on the age and education of HIV infection. For HIV, we chose 7 levels of age group-education factor with 3 education level and 3 age groups (16–20 years, 21–30 years, and >30 years). However, PWID who had no education were merged in one group because of the small sample size.

In the logistic regression, year-region factor, age group-education factor, and duration of injecting drugs were significantly associated with HIV infection. The graph below shows results from fitting logistic model for HIV infection prevalence, with year-region factor, age group-education factor, and duration of injecting drugs as determinants. The model also highlights a substantial decrease on HIV infection prevalence from 2002 to 2015. Kathmandu Region had highest prevalence compared to other regions of Nepal. HIV infection prevalence increases with age. The prevalence was highest for those aged ≥30 years. Education was significantly associated with HIV infection as the highest prevalence was observed among the illiterate in all age groups. Moreover, the prevalence significantly increased with duration of injecting drugs. Higher prevalence was found among PWID who had a duration of injecting drugs >10 years.

Receiver operating characteristic (ROC) curve

Figure 2 shows a ROC curve for final model fitted to HIV infection prevalence. ROC curves provide a mosaic plot for comparing logistic regression models. The area under the curve (AUC) as 0.70, indicating model performance is fit. ROC curves show that model containing year-region factor, age-education factor and duration of injecting drug fits the prevalence data extremely well.

Figure 1

Figure 2

IBBS surveys were analyzed to confirm that PWID in Nepal are indeed at elevated risk of HIV infection. The study found year-region factor, age group-education factor, and duration of injecting drugs were associated with HIV prevalence in a multivariate logistic regression. This study showed notable success in the prevention of HIV infection among PWID. HIV infection prevalence among PWID has significantly decreased from 2002 to 2015. Factors such as access to HTV infection intervention programs, needle sharing programs, safe drug injecting practices, safe sexual practices with partners, and increase knowledge of HIV had contributed to decrease in HIV infection prevalence in recent years. Data from the IBBS surveys and Nepal Demographic Health Survey (NDHS) surveys delineate major improvements in several factors that may have attributed to this decline in HIV infection among these groups [4,5,6,7,8, 12]. Moreover, the prevalence remains similar in the latest three IBBS surveys of 2011,2012, and 2015. Although, HIV infection prevalence in Nepal remains stagnant in recent years, PWID still remain the most vulnerable subpopulation compared to other KAPs with regard to HIV infections [4,5,6,7,8].

There was pronounced spatial variation of HIV infection in four zones of Nepal. High HIV prevalence was found in Kathmandu and the Eastern region compared with Western Region and Pokhara. The highest concentration of drug user population is confined in the Kathmandu Valley and in the locations along the Eastern Highway. According to mapping and size estimation among drugs users, around 4,341 to 4,758 drug users were living in Kathmandu [17,18]. In Kathmandu, PWID has been widespread in the shooting galleries that are typical common places for PWID to congregate, and these are found in clandestine locations that often provide opportunities for buying, renting, or borrowing needles/syringes and other items for injection. Higher prevalence in the Eastern Region may also be a consequence of cross-border issues with India. The open and porous border between these two neighbors has provided an ideal passage for smugglers for decades, and districts mostly in the Terai region of the country turned into a major transit points for peddling drugs.

HIV infection prevalence was positively associated with increasing age and low education status. A previous study has shown that not only are young PWID at increased risk of HIV, but also that those who use drugs in shooting galleries, or places with other and older PWID, are more likely to start injecting drugs early, and are at increased risk of HIV, partly because of the high prevalence among the older subgroup [19]. HIV infection was also associated with low education level. A study in Nepal among female drug users also shows similar results, i.e., low education is strongly associated with HIV infection prevalence [9]. Low or limited education relates to poor awareness about HIV or acquired immunodeficiency syndrome (AIDS), and contributes to their vulnerability. In Nepal, PWID often came from rural areas and have limited education, making them extremely vulnerable to drug involvement and unsafe sex. Their powerlessness and poor understanding of their own risk probably also make them more likely to engage in high-risk drug use and unsafe sexual practices [4,8].

Consistent with previous studies, injection behavior showed a strong association with HIV infections in this population [9,10,11]. HIV infection prevalence was higher among PWID who had a long duration of injecting drugs (>10 years). This association could be causal because the sharing of injection needles over long duration was associated strongly with HIV infection rates, and in a dose-dependent manner. Sharing needles or syringes and other equipment for injection is the most frequent drug-related risk behavior, which puts the PWID at risk of HIV transmission. In Nepal, the sharing practice is deeply rooted within the PWID social and cultural context. In addition, other frequently cited reasons for sharing include limited resources; lack of clean needles/syringes, and fears of arrest by law personnel, all of which reinforce reusing and sharing equipment [5,8].

PWID provide an effective epidemiological bridge for a wider epidemic, through unsafe sexual practices with their regular partners (wives) and with nonregular partners. Unprotected sex behavior also increased HIV risk among the PWID. The risk was significantly increased by sex with commercial or casual sex partners. In the present study, we found the number of sexual partners and the inconsistent use of condom with a regular partner were significantly associated with HIV infection in bivariate analysis. However, in multivariate analysis, no statistical association was found. HIV infection was higher among PWID who do not have sex or had one sexual partner in past year. This may be the result of PWID were aware of their status and limited their sexual intercourse with partners. Moreover, PWID perceive that the risk of HIV transmission is less likely through sexual contact than through sharing injection items, and were therefore less likely to adopt safe sex than safe injection practices. Similarly, HIV infection was higher among those consistently using a condom with their regular female sexual partners. The reason may be the PWID were using condoms to either prevent HIV transmission to their clients, or as self-protection from infection with HIV, or sexually transmitted diseases. In addition, previous studies among drug users have found safer sex practices strongly associated with HIV infection. A study in the United States found that self-reporting as HIV infected was the strongest factor associated with consistent condom use in the past 6 months [20]. Another study in Puerto Rico found that HIV-positive drug users were nearly five times more likely to use condoms during vaginal sex [21].

Consistent with earlier findings [22,23,24], the HIV risk behaviors include sharing injecting equipment and unprotected sex, and the risk of contamination from sharing practices and nonsanitized usage of injecting equipment is present in considerable proportions, albeit in varying degrees. Programs to target HIV infection prevention and treatment should be urgently developed and implemented for this population. Provision of clean needles and syringes will be useful for the prevention of HIV transmission through networks of PWID.

In conclusion, the decrease in HIV infection prevalence among the PWID gives the impression that HIV infection prevention interventions have been successful in Nepal. Data from the upcoming IBBS survey rounds are expected to provide more specific insights into the level of impact by interventions in the study sites. This study also guides policymakers on designing HIV and sexually transmitted infection intervention programs based on the risk factors.

The study has a few limitations. IBBS surveys are cross-sectional by their design and cannot provide evidence of causal relationships between the determinants and HIV infections. Moreover, this study covered only HIV infection and risk-related behaviors, while other issues related to HIV intervention programs and drug policies have not been discussed. Despite such limitations, the positive association between background characteristics and drug injecting practices that increase the risk of HIV infection have important implications.