Study participants

Potential participants were recruited at various sites around New York City and enrolled at the Lower East Side Harm Reduction Center (LESHRC). The LESHRC is a community based, non-profit organization, also acting as a syringe service program (SSP), focused on harm reduction and reducing transmission of HIV/AIDS and HCV amongst PWID in the Lower East Side. Individuals unaware of their HCV status were offered free testing for HCV antibody and polymerase chain reaction (PCR) testing. Eligible participants were above the age of 18 years, spoke English or Spanish, had injected illicit drugs for a minimum of 1 year, had injected at least 1 illicit drug within the last 90 days, and were HCV RNA positive within the last 90 days. Those who were pregnant, diagnosed with advanced liver disease (defined as decompensated cirrhosis and/or hepatocellular carcinoma), and those already receiving HCV care (defined as having attended 2 or more visits with a HCV clinician in the last 6 months) were excluded from the study.

Study design

Eligible patients with positive HCV RNA tests were randomized 1:1 to the AC or UC arms. All participants provided written informed consent, and the study was conducted in accordance with Good Clinical Practice and the ethical principles that originated in the Declaration of Helsinki. The study was approved by the Weill Cornell Medicine and NYU School of Medicine institutional review boards.

Statistical analysis

The baseline characteristics of participants were reported as frequencies with percentages and mean values with standard deviations. Comparisons between the study arms were tested with of chi-square, Fisher exact, t-test, or Wilcoxon ranked sum tests when appropriate.

Injection frequencies and frequency of equipment sharing were categorized into ordinal variables, rather than continuous variables, to mitigate the impact of extreme outliers, address the highly skewed variable distribution, and to distinguish between key outcomes such as no injection days and daily injections (30 injections days).

At baseline (Table 2) and at each study timepoint (Table 3), participants were grouped into ordinal categories of a range of number of injection events/days in the last 30 days based on their interview answers. The ranges for the ordinal categories were chosen to proportionately capture the distribution of answers from all participants while highlighting no injection days and 30 injection days. The categories for injection days in the last 30 days are as follows: 10 days (no injection), 1–8 days (2 or less times a week), 9–29 days, and 30 days (daily injection). The categories for injection events in the last 30 days were the following: 0 times (no injection), 1–29 times (on average, less than once a day), 30–89 times, and 90+ times (on average, more than 3 times daily). The same ordinal categories were used for equipment sharing events in the last 30 days.

To assess the intervention (AC versus UC) effects observed during the study period (baseline, 3-, 6-, 9-, 12-months), we modeled the data using generalized linear mixed model (GLMM) to evaluate the main effect of intervention, time, and the interaction effect of intervention and time. Additionally, initiation to treatment and its interaction with intervention were included as covariates, accounting for potential confounding effect among those who initiated treatment vs those who did not. Given our focus on testing for trends, time was treated as a numeric variable. For the treatment of opioid use disorder, since the variable is dichotomous, a binary distribution with logit link function was used. For injection frequency and equipment sharing frequency variables, a multinomial distribution with cumulative logit link function was used. All models included a random intercept.

The missingness of the data was also assessed via logistic regression for the five outcome variables at 12 months, employing the outcomes at the previous four data points as predictors [22]. This analysis aimed to examine if the patterns of missingness could be explained by the observed values. In addition, regardless of whether the missingness depended on the observed values, we conducted a sensitivity analysis using multiple imputation.

Statistically significant results were evaluated at p<0.05. Analyses were conducted using SPSS 25 (IBM Corp., 2017) and GLMM was performed using PROC GLIMMIX and multiple imputation using PROC MI/MIANALYZE in SAS software (Release 9.4 SAS Institute Inc., Cary, NC USA).

Frequency of injection drugs

In the current study, patients in the AC arm reported a significantly greater decrease in the total number of injection events and the number of days using IV substances within the last 30 days throughout the 12 months of enrolment. While previous studies have looked at the independent impacts of HCV treatment and patient education on high-risk injection behaviors, our study is the first to analyze the collective impact of accessible HCV treatment, psychosocial support, and patient education for harm reduction delivered in a multi-disciplinary setting [16, 23, 24].

Our parent study showed that the higher rate of SVR12 achieved in the AC arm was not due to differences in treatment response but from higher rates of advancement along the care cascade prior to treatment initiation [9]. This is likely attributable to decreased psychosocial and logistical barriers to care [9]. It unfortunately remains commonplace for healthcare providers to attribute negative qualities such as “weak mindedness” to PWID [11, 20]. Experiences of such external stigma often translate into internal stigma and individuals begin to apply these negative qualities to themselves [11, 20]. In a 2020 survey, 78.1% of PWID reported stigmatizing experiences with healthcare providers (HCPs), highlighting the magnitude of the issue of demoralization that PWID experience secondary to stigmatization [25]. The emphasis that the AC arm placed on interactions devoid of stigmatization and paternalism likely increased the self-efficacy of participants, encouraging them to seek out medical care and make health-conscious decisions which resulted in the increased advancement through the care cascade and the decrease in injection frequency. Qualitative studies of HCV-treated PWID suggest that receiving treatment on its own increases one’s sense of responsibility over their wellbeing [20, 26]. The knowledge of having made a health-conscious decision paired with the measurable outcome of a cleared viral load can act as a motivating point of “transformation” for further behavioral changes including injection frequency and method of injection [20, 26]. Thus, the higher percentage of patients initiating treatment and achieving SVR12 in the AC arm could be a contributing factor to the greater decrease in injection frequency seen in patients in the AC arm.

High-risk injection behaviors

There was no statistical difference in the rate of receptive sharing of syringes over the 12 months between those receiving AC and UC care. There was a decrease in the rate of cooker sharing in the AC group compared to the UC group. The statistical significance of this depended on whether missing data was considered or not. One possibility is that objects used as cookers are more readily available than syringes and so there was less of a barrier for participants to apply safe injection practices they had been taught against sharing cookers in the AC arm. Another possibility is that there was no difference between both syringe or cooker sharing rates between the AC and UC groups as implied when missing data is factored in. The decision to inject may be the rate-limiting step in the process of making a health-conscious decision, and that once the decision to inject has been made, the threshold for overlooking the health-risks has already been crossed.

A study led by Garfein et al. in 2002 compared the impact of peer education intervention (PEI) with video discussion intervention (VDI) on high-risk injection behaviors [23]. PEI comprised of 6 two-hour group sessions where PWID received education on harm reduction, community-based resources, and peer-education from an in-person facilitator [23]. Those in the VDI group were asked to watch videos addressing psychosocial issues, with less emphasis on harm reduction, with the same number of cumulative hours of education as the PEI group [23]. This trial looked at rates of receptive sharing of used syringes, use of a new syringe to divide drugs, sharing of cookers/cotton/rinse water, and the proportion of partners that participants shared paraphernalia with [23]. Interestingly, there was only a significant change in decrease in the composite index of all six outcomes in the PEI group compared to the VDI group. There was no significant difference in the rates of the six outcomes when individually compared between the two groups [23]. Contrastingly, Bertrand et al. conducted a study comparing individuals receiving motivational interviewing (MI) versus educational intervention (EI) [24]. At the 6 month mark, individuals in the MI group were 50% less likely to share paraphernalia excluding syringes (OR = 0.50; CI = 0.09–0.90, p = 0.017), and 53% less likely to share containers (OR = 0.47; CI = 0.11–0.84, p = 0.011) [24].

Data regarding change in high-risk injection behaviors is sparse. However, the data available, including our study and two studies mentioned above, is supportive of a trend towards decreased rates of high-risk injection behaviors when PWID receive interactive and personalized education on harm reduction and safe injection practices [28].

Enrolment in MOUD programs

The rate of enrolment in MOUD programs fell in both arms throughout the 12 months with no statistical difference between the AC and UC arms. The magnitude of the decreases in both arms is consistent with the MOUD retention rates seen in prior studies [17, 27, 28]. Given our low-threshold care model did not require (or even endorse) abstinence or engagement in MOUD as a prerequisite to HCV treatment, it was not surprising to see no statistical difference between AC and UC in MOUD retention.

As our study took place amidst the pandemic, the impact of social distancing, given pre-COVID policies requiring in-person appointments for MOUD administration, was considered as a possible confounder. A sub-study of the parent study looked at the impact of the COVID-19 pandemic on drug service utilization by HCV-positive PWID in New York City [29]. Subjects were categorized into pre-COVID versus post-COVID groups as well as AC versus UC groups. There was no significant decrease in MOUD enrolment post-COVID except for a significant decrease in buprenorphine-based program enrolment post-pandemic in the UC group [29]. However, the majority of the participants receiving MOUD in the current study were in methadone-based programs.

Limitations

One of the limitations of our study is the non-attendance rates at the different follow-up time points during which interviews were held. This could be partially attributed to the fact that the study took place during the COVID-19 pandemic. The greatest rate of non-attendance occurred at three months with thirty-five and twenty-four participants from the UC and AC group respectively not attending the follow-up interviews. However, many participants who did not attend a follow-up visit attended subsequent visits and participated in questionnaires then. At the 12-month mark, there was no significant difference in the number of those lost to follow-up in the AC and UC cohorts.

Another limitation that we faced were the outliers when it came to injection frequency. To circumvent this issue, we presented and analyzed the data as ordinal categories that represented the distribution of frequencies rather than with mean values. This unfortunately could not be done due to the dichotomous distribution of answers we got regarding injection behaviors. Thus, the mean values presented for frequency of syringe and cooker sharing should be interpreted with caution as they account for outliers on both ends of the spectrum.

Given that the frequency of injection and high-risk behavior frequencies were self-reported during interviews, the possibility of recall bias and social desirability bias is present. In this study, questionnaires were administered by interviewers. Assuming that individuals in the AC arm had stronger rapport with the care team, it is important to consider whether this led to underreporting due to concerns of disappointing the interviewer or if it led to more transparency.