Introduction

By 2013, women accounted for almost half of all 24.6 million people who reported past-month illicit drug use in the United States [1]. Of the HIV diagnoses attributed to injection drug use, more than 40% were among women [2]. Women who use drugs (WWUD) are at elevated risk for HIV acquisition [2] for various reasons, such as, receptive syringe sharing and drug-influenced disinhibited sexual risk behaviors, as well as by the high prevalence of transactional sex and syndemic risk factors observed in this population (violence, poverty, disempowerment and lack of care access) [3,4,5] One study conducted in 20 United States (U.S.) cities indicated that two-fifths of 2305 women who inject drugs reported selling sex for money, goods or drugs at least once in the past year, with the majority of sex exchange encounters involving condomless sex [6]. An immediate and effective tool is urgently needed to supplement behavioral prevention approaches (e.g., condom use, syringe exchange) to prevent HIV acquisition and transmission among WWUD [7,8,9,10,11].

Pre-exposure prophylaxis (PrEP) via once daily oral administration of Truvada (emtricitabine/tenofovir disoproxil fumarate), a U.S. Food and Drug Administration (FDA)-approved HIV anti-retroviral medication, has been shown to be highly efficacious in reducing HIV incidence among high-risk populations [12,13,14,15]. However, the benefits of PrEP remain largely unrealized in real-world settings due to various individual and contextual barriers across the PrEP care continuum (i.e., PrEP awareness, willingness to use PrEP, linkage to PrEP care, PrEP uptake, retention, and adherence) [16, 17]. As the entry stage to the PrEP care continuum, PrEP awareness/willingness to use is essential; no further actions can be taken to achieve sustained PrEP adherence and risk reduction benefits unless this stage has been achieved. Recent studies have revealed a moderate level of PrEP awareness or willingness to use among several key populations, including men who have sex with men (MSM) and sero-discordant couples [18,19,20,21,22].

Furthermore, Choopanya et al. [12] conducted a clinical trial to assess the efficacy of PrEP among Thai men and women who inject drugs (PWID), and found significantly lower HIV incidence among those taking daily Truvada (0.35 vs. 0.68 per 100 person-year), with showing more effective among women as they had lower incidence compared to men in the treatment group (0.20 vs. 0.39 per 100 person-year). Despite the demonstrated efficacy of PrEP among PWID and the elevated risk of HIV acquisition among WWUD, few studies have assessed the implementation of PrEP in this group or explored the challenges/opportunities across the PrEP care continuum to inform future prevention intervention efforts among WWUD [23]. Although no studies have revealed the proportion of WWUD who are currently using PrEP, limited PrEP research involving WWUD in the U.S has revealed that participants have been highly satisfied with PrEP and considered PrEP as an essential tool for HIV prevention [24]. PrEP recipients have reported no adverse impact on quality of life as a result of taking PrEP [25], and other participants would accept PrEP with caution [26]. In spite of these limited observations, characterization of the key stages of the PrEP care continuum among WWUD remains poorly understood in the U.S.

Following the preferred reporting items for systematic reviews and meta-analyses (PRISMA) guideline (http://www.prisma-statement.org/), we gathered evidence from published studies and meta-analytically described key stages along the PrEP continuum among WWUD in the U.S., including all available studies with rigorous study designs. This systematic review and meta-analysis is meant to provide empirical guidance for future targeted PrEP programs and interventions among this particularly high-risk group.

Methods

Eligibility Criteria

Inclusion Criteria

Published articles were included if they: (a) sampled women or reported data separately for women and men who reported either injection or non-injection illicit drug use (e.g., marijuana, cocaine, heroin, methamphetamine, hallucinogens); (b) used quantitative inferential study designs (e.g., randomized control trials, cross-sectional/cohort studies); (c) reported quantitative measures (e.g., proportion) on any stage of the PrEP care continuum (i.e., PrEP awareness, acceptance/willingness to use PrEP, PrEP uptake/use) or provided sufficient information to calculate pooled estimates; (d) were peer-reviewed and published in English between January 1, 2012 until July 5, 2018; and (e) can be searched from indexed databases or published sources. Our search start date parameter was selected to coincide with U.S. FDA approval of PrEP for HIV prevention among individuals aged 18 or older [27].

Exclusion Criteria

The exclusion criteria included: (a) descriptive studies (e.g., case reports) or qualitative studies without quantitative measurement; (b) studies that did not report results for a segregated subgroup of WWUD (e.g., a mixed sampling of people who use drugs and people who do not use drugs); (c) reviews/meta-analysis; and (d) theoretical/modeling studies without original data.

Information Sources, Search and Study Selection

Following the PRISMA guidelines, we conducted a comprehensive literature search of the following databases: PubMed/MEDLINE, Web of Science, PsycINFO, EMBASE, and GOOGLE SCHOLAR. Our Boolean search queries were: (“HIV” OR “human immunodeficiency virus” OR “AIDS” OR “acquired immunodeficiency syndrome”) AND (“PrEP” OR “pre-exposure prophylaxis”) AND (“woman (women)” OR “female(s)” OR “girl(s)”) AND (“drug use” OR “drug-using” OR “drug abuse” OR “drug dependence” OR “substance abuse” OR “substance dependence”). We also searched through conference proceedings, as well as references from articles/reviews that met our inclusion criteria. Two reviewers (CZ and YL) independently reviewed articles identified in the initial search. The interrater reliability was > 90%, and disagreement between reviewers was resolved by discussion.

Data Collection and Quality of Evidence Assessment

Two reviewers independently extracted data from eligible articles using a standardized form to record the following information: (a) location and year of the study conducted, (b) demographic characteristics of participants (e.g., age, race/ethnicity), (c) sample size, (d) study design, (e) recruitment strategy, (f) key measurements (e.g., PrEP awareness and willingness to use), and (g) key findings and notes or other informative features. For studies with duplicate publications [28, 29], we reported the study only once in the analyses, with the most complete data included. Three studies reported PrEP awareness [30,31,32], and six studies reported willingness to use PrEP [28,29,30,31,32,33] specifically among WWUD. In addition, we employed the GRADE rating scheme to evaluate the quality of evidence from each individual study using recommended criteria (i.e. risk of bias, precision, consistency, directness) [34].

Statistical Analysis

Measures

Prevalence of PrEP awareness (i.e. proportion of those who had ever heard of PrEP among all interviewed WWUD) and willingness to use PrEP (i.e. proportion of those who self-reported willingness to take PrEP among all interviewed WWUD) were the key measures in the current analysis.

Assessment of Heterogeneity and Publication Bias

We used the I2-statistic and corresponding 95% confidence intervals (CI) to depict heterogeneity. The I2-statistic describes the percentage of the variability in effect estimates that is due to heterogeneity rather than sampling error, with higher percentages indicating higher heterogeneity [35, 36]. Heterogeneity statistic (i.e., tau-squared) and its corresponding p value were also reported for each analysis. Publication bias was assessed by funnel plots and Egger’s test [37]. Funnel plots were employed to visually assess the asymmetry, while Egger’s test was used to test the asymmetry statistically.

Data Synthesis, Subgroup and Sensitivity Analyses

We employed the DerSimonian-Laird random-effects model to weight and pool the individual estimates [38], as all included studies were conducted among populations across heterogeneous settings [35]. Unlike the fixed effects model that assumed all studies shared identical true effect sizes, the random effects model was designed to capture variances of estimates across studies [35, 39]. Forest plots were used to illustrate the aggregate findings. We also performed subgroup analyses to examine the pooled estimates if the pooled estimates differed by sample (e.g., women who use drugs vs. women who inject drugs only, men and women [mixed] who use drugs vs. men and women (mixed) who inject only) because drug use behavior may correlate with HIV risk and PrEP utilization. Sensitivity analyses were employed to examine the stability of the pooled estimates by evaluating whether the overall pooled estimates were sensitive to exclusion of any individual studies (e.g., study with highest or lowest weight, studies with lowest rating, and with smallest or largest sample size). We performed all statistical and meta-analysis using STATA 15 (College Station, TX). The STATA Metaprop command was used to pool the data as it was particularly designed for binomial data, using the binomial distribution to model within-study variances [40]. Based upon this strategy, the Method of Moments is used to estimate the mean, but Maximum Likelihood is used to estimate heterogeneity. [38, 40] The Metaprop command computes 95% CI using the score statistics (for smaller sample size) that allows incorporation of the Freeman-Tukey double arcsine transformation [41] of proportion to generate admissible pooled estimates within the range of [0, 1], which has been validated by multiple studies [40, 42, 43].

Narrative Synthesis

In order to identify factors that may affect PrEP use among WWUD, we employed a narrative synthesis approach to identify predisposing (e.g., patient-level factors) and enabling factors (e.g., available resources that may facilitate PrEP use) as well as needs components (e.g., health status barriers, belief or perceptions) based upon the Behavioral Model of Healthcare Utilization for Vulnerable Populations [23, 44].

Results

Study Selection, Characteristics, and Risk of Biases Across Studies

Our initial search yielded 203 citations. After title/abstract screening, 109 references were full-text reviewed, with seven studies meeting all inclusion criteria (Fig. S1). Among these publications, findings from one abstract [28] overlapped with another article [29], and two articles reported findings from the same study [33, 45]. Another study reported separate information for women who inject drugs in New York City and Long Island, and we treated these two data points separately [32]. A total of eight articles and one abstract, reporting findings from seven individual studies, were retained after rigorous selection.

Details about each included study are presented in Table 1. We pooled the numbers of WWUD from all selected studies. Our pooled sample size is 755, including 370 Black (49%), 126 Hispanic (16.7%), and 259 (34.3%) White women. All studies used cross-sectional designs; three were conducted in New York [31, 32, 46], and one each in Maryland [30], Washington DC [28, 29], Connecticut [33, 45], and Massachusetts [47]. Three studies (reported by three articles and one abstract) were secondary data analyses of the National HIV Behavioral Surveillance (NHBS) data [28, 29, 31, 32]. Four studies used convenience sampling to recruit participants from health services settings, including syringe exchange and sexual health services [30], opioid detoxification programs [47], or methadone maintenance programs [33, 45]. Among the included studies, six reported prevalence of PrEP awareness [28, 30,31,32,33, 45, 46] with three reporting women-specific data; [30,31,32] four reporting prevalence of PrEP use willingness and all four reporting women-specific information [30, 33, 47]. Most studies used self-reported HIV status [28, 29, 31,32,33, 45,46,47], and some studies did not report ethnicity/race specific data in relation to PrEP use [28, 29, 31,32,33, 45,46,47].

Table 1 Summary of included studies for drug-using women (n = 7 studies)
Full size table

PrEP was usually referred to as “a pill to prevent HIV” in these included studies. In most studies, PrEP awareness was assessed by a single question, such as “HIV Pre-exposure prophylaxis, or PrEP, is a way for people who do not have HIV to prevent HIV infection by taking a pill every day?”; or “Have you heard of HIV PrEP before today?”; or “Before today, have you ever heard of people who do not have HIV taking antiretroviral medicines, to keep from getting HIV?” [28,29,30,31,32,33, 45,46,47]. For willingness to use PrEP, all studies assessed participants’ willingness by asking general questions, such as “If a daily HIV pill to prevent you from getting HIV was available in DC for free or was covered by your health insurance, how likely would you be to take it?”, or “How interested would you be in taking a pill every day to prevent HIV infection?” [28, 29]. One study assessed participants’ willingness to take PrEP using specific risk reduction information (i.e., “Would you be willing to take a once a day pill every day to lower your risk 90% (or 40%) of becoming HIV positive?”) [47]. In addition to these two key measures (awareness and willingness to use), several studies also reported participants’ perception of potential risk compensation (PrEP users may conduct more risky behaviors as they feel protected by taking PrEP). Questions assessing risk compensation included “I will no longer need to sterilize or use clean needles/use condoms or practice safe sex if I am taking pills to prevent HIV infection” or “Would you take PrEP if you still had to use condoms to be fully protected from HIV?” [28,29,30, 33, 45, 47].

Synthesis of Results

In studies that involved WWUD samples, the pooled prevalence of PrEP awareness was 20.6% (95% CI 8.7%, 32.4%). However, once these participants were informed of PrEP, willingness of PrEP use was 60.2% (95% CI 55.2%, 64.1%) among injectors and non-injectors and 57.3% (95% CI 48.8%, 65.9%) among injectors only, respectively. Very few women reported PrEP uptake (i.e., ever taking PrEP) across included studies [31, 33, 45]. For example, in the study by Walters et al. (2017), < 1% of injectors (1 of 118) reported ever using PrEP [31]. In another study involving mixed men and women who use drugs, only 1.8% ever used PrEP [33, 45]. No data for PrEP care retention and adherence were reported in any included studies (Figs. S2, S3). In addition to measurements of the PrEP cascade, four studies reported respondents’ perception of potential risk compensation under the hypothetical scenario of initiating PrEP (Table 2) [28,29,30, 33, 45, 47]. The pooled proportion of perceived risk compensation among all men and women (mixed) who use drugs was 26.1% (95% CI 6.9%, 45.2%), and among men and women (mixed) who inject drugs was only 18.4% (95% CI 9.9%, 26.8%). No gender-specific data regarding risk compensation perceptions were available (Table 2; Fig. 1, Fig. S6).

Table 2 Summary of meta-analyses among drug-using people
Full size table
Fig. 1
figure 1

PrEP awareness and willingness among drug-using people by different types of drug usea,b. apooled prevalence of PrEP willingness and PrEP awareness among different types of drug users (e.g., all injected drug users [IDU], all drug user including men and women, female IDU only, all drug using women) and its corresponding 95% confidence interval; bx-axis is the pooled prevalence (ranged from 0 to 0.70)

Full size image

Risk of Biases Within and Across Studies

The quality of evidence for most outcomes was scored as low or very low, primarily due to the nature of observational study design, limited sample size, and generalizability (Table 1). Publication bias was assessed by funnel plot, where the standard error of the effect size was plotted against the effect size. Examining the funnel plot revealed that publication bias was present, as the graph showed slight asymmetry within the funnel (Figs. S4, S5, S7). We were unable to conduct further tests on the funnel plot asymmetry, as there were fewer than ten studies included and the power was too low to distinguish chance from statistically significant asymmetry [37, 48]. We further examined the heterogeneity using I2-statistic (ranged from 50.8 to 91.7%) across different subgroup analyses; this indicated moderate to high heterogeneity of the studies. p-values of heterogeneity Chi squares also indicated significant heterogeneity across included studies (Table 2).

Additional Analyses

Sensitivity analyses were conducted by removing studies with highest and lowest weight, studies with largest and smallest sample size, and studies rated as “very low” by GRADE scoring scheme, respectively. Comparing outcomes from sensitivity analyses with the original outcome, no difference was found from the original sensitivity analyses (not shown).

From the narrative analysis, predisposing factors of PrEP willingness among people who use drugs (PWUD) included experience of intimate partner violence, younger age and multi-partnership. Enabling factors included receiving information from their doctors, and needs components were composed of increased perceived and behavioral HIV risk, depression, and neurological impairments. For PrEP awareness among PWUD, route of drug administration and sexual orientation were considered as predisposing factors, while discussing PrEP at a syringe exchange program was considered as an enabling component (Table 3).

Table 3 Summary of factors associated with PrEP awareness and willingness among people who use drugs
Full size table

Discussion

Our meta-analytic review included 755 WWUD from seven studies (604 from four women-only samples). The low awareness of PrEP among WWUD (21%) raises concern for this high-risk group, especially when considering this group may have frequent contact with harm reduction and health professionals via various encounters (e.g., settings providing needle exchanges and sexual health services [30], opioid detoxification programs [47], and methadone maintenance programs). However, once WWUD were informed of PrEP, their willingness to take PrEP increased considerably (60% for WWUD, and 57% for women who inject drugs). The discrepancy between lower awareness and higher willingness represents a missed opportunity for initiating PrEP among WWUD who are willing to take PrEP if they are well-informed. After including male drug users in the analyses, both PrEP awareness (20.6% vs. 17.2%) and willingness to use (60.2% vs. 52.1%) decreased slightly. Our findings suggest that WWUD’s PrEP awareness were higher compared to male counterparts who use drugs, but lower compared to MSM (PrEP awareness ranged from 27 to 46% among MSM) [49,50,51]. On the other hand, WWUD’s willingness was similar or higher than their male counterparts, but PrEP uptake among WWUD much lower than MSM [49,50,51]. This higher level of awareness and uptake among MSM may stem from various PrEP initiatives primarily focusing on this population; such initiatives have been lagging for individuals who use drugs. In settings where PWUD usually seek help (e.g., drug treatment services), it is highly plausible that health professionals could engage their patients/clients about PrEP use through counselling and treatment referrals. For instance, in New York State the Department of Health has successfully initiated a PrEP service in established syringe exchange programs and sexual health clinics to assist high-risk individuals access PrEP care [52].

In addition to examining PrEP awareness and willingness to use PrEP, potential risk compensation was examined among men and women (mixed) who use drugs. Although < 20% of PWUD indicated they would “no longer need condoms during sexual episodes or sterilized needles during injections after PrEP use”, these estimates were based on complete relinquishment of other forms of protection (e.g., elimination of condoms/clean needles after taking PrEP). If participants’ perception of risk compensation was evaluated conservatively (e.g., reduction in the use of condom/clean needles), a majority of participants expressed potential tendency for risk compensation while taking PrEP [33, 45]. Though PrEP clinical trials have yet to observe any risk compensation behaviors [14, 53], increased prevalence of risky sexual behaviors (e.g., condomless sex, increased sexual partners) and sexually transmitted infections (STI) have been reported among PrEP users in several observational studies [54, 55]. However, none of the included studies contained data depicting gender-specific perceptions of potential risk compensation in these included studies. Future studies should closely monitor actual risk compensation behaviors among PWUD and are taking PrEP, and future PrEP implementation interventions targeting individuals who use drugs should include education on both the benefits and limitations of PrEP to avoid potential risk compensation behaviors.

Our meta-analysis has several strengths, including (a) it is the first study to synthesize PrEP awareness and willingness to use among WWUD as well as among all PWUD in the U.S. by subgroup analyses; (b) we employed the STATA Metaprop command with random effect models to accurately accommodate the within-study variances of binominal data; and (c) the PRISMA guideline was strictly followed throughout all procedures.

Several limitations of our review and findings should be noted, most of which derive from the limited number, geographic scope, and specificity of existing studies on PrEP among WWUD. These limitations constrained the representativeness and generalizability of our findings. First, of the seven included studies, all were conducted in urban areas in the Northeast U.S., no studies have been conducted among WWUD in the Deep South or rural areas, which have among the highest rates of HIV incidence and prevalence across the U.S. [56,57,58]. In addition, half of the included studies failed to provide gender-specific data [28, 29, 33, 45,46,47]. Even for studies with gender-specific data, WWUD were only a small proportion of all recruited participants [33, 45]. Thus, regional and gender under-representativeness may not reflect the full scope of sexual risk and HIV transmission dynamics among WWUD in the U.S.

Second, there is a lack of data regarding PrEP uptake, retention, and adherence among this at-risk group as a whole. Among all included studies, few participants reported PrEP use [31, 33, 45]. It is estimated that only 5.3% to 15.2% of high-risk individuals achieve sustained PrEP use, due to individual, social, and structural barriers interrupting the continuum of PrEP care [16, 59]. The scarcity of data regarding the PrEP care continuum among WWUD requires future studies to address this gap.

Third, several included studies failed to report demographic-specific data regarding PrEP use among WWUD. Only one study examined racial differences in HIV risk factors, including PrEP awareness and willingness to use among individuals with opioid use disorders [46]. Lack of sufficient demographic-specific data regarding PrEP use may hinder the deployment of interventions tailored to WWUD with various cultural characteristics.

Fourth, half of the included studies employed secondary data analyses [28, 29, 31, 32], and those studies may not have been originally designed to assess PrEP awareness and willingness. In addition, the rest of the included studies employed convenience sampling strategies to recruit participants from settings providing health services (syringe exchanges and sexual health services [30], opioid detoxification programs [47], and methadone maintenance programs), which may have resulted in under-sampling of hidden and hard-to-reach subpopulations. Omission of these key subgroups may leave important gaps in our understanding regarding HIV epidemics and PrEP implementation in this population [60].

Lastly, the accuracy and reliability of the synthesized effect sizes of PrEP awareness and willingness to use may be affected by the significant heterogeneity, publication and self-reported biases, as well as inconsistent measurements and misclassification of key indicators across studies. For instance, some studies categorized “very likely and somewhat likely” as “willingness to use PrEP” [30, 47], whereas another study categorized “somewhat likely” as “no willingness to use PrEP” [28, 29]. Furthermore, some studies referred to PrEP simply as “a pill” or “anti-HIV medicines” [29, 30, 47, 61], whereas other studies specified PrEP as “Truvada” or “an antiretroviral medication” [31, 32, 46]. Furthermore, all measurements on “PrEP awareness” and “willingness to use” were based upon one single item, and were inconsistent across different studies: some studies asked the likelihood of taking PrEP in the future [33, 45], while another evaluated willingness based upon different hypothetical scenarios [47]. Inconsistent descriptions of PrEP, variations of the cut-off points, and the measurement’s lack of comprehensiveness and consistency may lead to biased synthesized estimates.

Our systematic review revealed several critical deficiencies with regard to current evidence on the PrEP care continuum among U.S. WWUD, including the limited number, scope and specificity of extant studies, as well as limitations in study design. Despite these shortcomings, our meta-analysis provided empirical estimates for the initial two stages along the PrEP care continuum: “PrEP awareness” and “willingness to take PrEP.” Since using PrEP can prevent women from sexual and blood-born transmission of HIV, WWUD, in particular, would benefit from PrEP. Each phase of the PrEP care continuum cascade represents a potential barrier and critical intervention point to achieve sustained PrEP protection. In order to address limitations identified in the current review, we call for comprehensive assessment tools and standardized measures to evaluate each step of the PrEP care continuum among WWUD in future studies.