Rapid Decline in HIV Incidence Among Persons Who Inject Drugs During a Fast-Track Combination Prevention Program After an HIV Outbreak in Athens

Abstract

Background.

A “seek-test-treat” intervention (ARISTOTLE) was implemented in response to an outbreak of human immunodeficiency virus (HIV) infection among persons who inject drugs (PWID) in Athens. We assess trends in HIV incidence, prevalence, risk behaviors and access to prevention/treatment.

Methods.

Methods included behavioral data collection, provision of injection equipment, HIV testing, linkage to opioid substitution treatment (OST) programs and HIV care during 5 rounds of respondent-driven sampling (2012–2013). HIV incidence was estimated from observed seroconversions.

Results.

Estimated coverage of the target population was 88% (71%–100%; 7113 questionnaires/blood samples from 3320 PWID). The prevalence of HIV infection was 16.5%. The incidence per 100 person-years decreased from 7.8 (95% confidence interval, 4.6–13.1) (2012) to 1.7 (0.55–5.31) (2013; P for trend = .001). Risk factors for seroconversion were frequency of injection, homelessness, and history of imprisonment. Injection at least once daily declined from 45.2% to 18.8% (P < .001) and from 36.8% to 26.0% (P = .007) for sharing syringes, and the proportion of undiagnosed HIV infection declined from 84.3% to 15.0% (P < .001). Current OST increased from 12.2% to 27.7% (P < .001), and 48.4% of unlinked seropositive participants were linked to HIV care through 2013. Repeat participants reported higher rates of adequate syringe coverage, linkage to HIV care and OST.

Conclusions.

Multiple evidence-based interventions delivered through rapid recruitment in a large proportion of the population of PWID are likely to have helped mitigate this HIV outbreak.

(See the editorial commentary by Chan and Flanigan on pages 1491–2.)

In 2011, an outbreak of human immunodeficiency virus (HIV) infection occurred among persons who inject drugs (PWID) in the Athens metropolitan area [1, 2]. Until then, Greece had had a low-level HIV epidemic, mainly among men who have sex with men, with only a few new diagnoses reported annually among PWID [1, 3]. The 2011 outbreak occurred in the context of low levels of harm-reduction service provision [4] and a severe economic recession, which have been proposed as macro-level causes [5, 6]. The Greek Organisation Against Drugs substantially increased the number of centers providing opioid substitution treatment (OST) (from 25 to 58 OST units between August 2011 and December 2012) and, in cooperation with nongovernmental organizations and the Hellenic Centre for Disease Control and Prevention, scaled up needle and syringe programs (NSPs) soon after the outbreak was recognized (from 7 to 45 syringes per estimated PWID per year between 2010 and 2012) [7]. Despite these improvements, during the outbreak there was a waiting time of 3.6 years for opioid dependents to enter OST in Athens [7] and syringe programs were underperforming, ranging below the guideline of 200 syringes per PWID per year [8].

In addition to these efforts, the ARISTOTLE combination intervention was implemented according to the “seek-test-treat” paradigm, with the aim to rapidly identify as many PWID as possible, test them for HIV infection, distribute prevention materials (syringes, condoms), provide counseling, and link those who were HIV seropositive to antiretroviral therapy (ART) and OST programs, to decrease HIV-1 transmission [9, 10]. Respondent-driven-sampling (RDS) was used to recruit PWID during 2012–2013. This article describes trends in (1) HIV-1 incidence and factors associated with incident HIV infection, (2) HIV-1 prevalence, (3) risk behaviors, (4) access to OST and NSPs, and (5) linkage to HIV care and ART, during an evolving outbreak.

METHODS

Design of the Program

The ARISTOTLE intervention program is described in detail elsewhere [10]. In short, respondents were eligible to participate if they presented a valid RDS coupon, had injected drugs without a prescription in the past 12 months, were ≥18 years of age, and resided in the Athens metropolitan area. Participation included (1) computer-assisted personal interviewing (Supplementary Materials); (2) blood sample collection; (3) receipt of 3 coupons to recruit other PWID, a pack of 20 syringes with injection paraphernalia, condoms, and leaflets with prevention information, and monetary incentives for participating; and, a few days later, (4) collection of HIV test result, counseling, and linkage to the Greek Organisation Against Drugs and/or to Infectious Diseases Units for OST, HIV care, and ART. ARISTOTLE was implemented in 5 consecutive RDS rounds with a short break in-between rounds. PWID could participate in multiple rounds but only once in each round. Dedicated staff supported linkage to HIV care, ART, and OST programs.

Ethical Issues

The survey protocol and informed consent form were approved by the Institutional Review Board of Athens University Medical School. Eligible persons were asked to provide written informed consent. The informed consent form included information about the program and explained that confidentiality would be protected and that participants were free to withdraw their consent at any point in the process.

Laboratory Testing

HIV tests were performed with a microparticle enzyme immunoassay anti–HIV-1/2 (AxSYM HIV-1/2 gO; Abbott) and HIV-1/HIV-2 confirmation by Western Blot (MP Diagnostics). In the case of indeterminate results, HIV status was determined by measurement of HIV RNA levels.

Linkage to HIV Care During the Program

In our analysis, a seropositive participant was considered to be linked to HIV care if he or she had ≥1 visit to an infectious diseases unit. Information on the date of first visit to the infectious diseases unit and of initiation of ART was collected from the treating physicians and the Hellenic Centre for Disease Control and Prevention. After completion of the program, additional data collection was performed by a research nurse who visited the infectious diseases units in Athens.

Statistical Analysis

We used the Respondent-Driven Sampling Analysis Tool to produce adjusted estimates of population proportions of characteristics of interest (version 7.1.38, Cornell University, Ithaca, NY, USA) [11]. HIV incidence was estimated using data from PWID with ≥2 samples who tested negative for anti-HIV-1 at first participation. PWID with a negative HIV test result followed by a positive result were defined as HIV-1 seroconverters. The seroconversion time for these subjects was estimated by the midpoint of the seroconversion interval, that is, the interval between the last negative and the first positive sample. Cox proportional hazards regression models were used to identify predictors of HIV seroconversion. Variables included in the multivariable model were all those with P < .25 in the univariable analyses as well as age and sex. Except for baseline age, sex, education, and country of origin, all variables were modeled as time-dependent covariates.

To assess trends in risk behaviors as well as access to prevention and care, we evaluated these proportions overall, as well as separately for first-time and repeat participants, across the 5 rounds of the program. Access to NSPs, OST, and HIV testing, either through ARISTOTLE or through other community-based activities, was based on self-reported data. Linkage to HIV care and ART was evaluated from data collected from infectious diseases units.

To assess trends in access to free syringes, taking also into account potential changes in the frequency of injections, we have estimated individual syringe coverage percentage as the ratio of the number of new syringes participants reported having received from prevention activities in the past month to the total number of injections in the same period (for participants who had injected drugs within the past month). To allow comparisons with European national-level syringe coverage data that refer to the yearly number of syringes per PWID [12], we multiplied the number of free syringes received in the past month by 12 (separately for 2012 and 2013).

RESULTS

Participants, Questionnaires, and Blood Samples

The 5 rounds of the program took place from August 2012 through December 2013 (a total of 13 months, excluding breaks) (Table 1). During the program, 7113 questionnaires and blood samples were collected from 3320 unique participants; 1533 PWID participated only once, and the rest (n = 1787; 53.8%) participated in multiple rounds, 681 (20.5%) in 2 rounds, 469 (14.1%) in 3 rounds, 374 (11.3%) in 4 and 263 (7.9%) in all 5 rounds (details on repeat participants provided in Supplementary Materials).

Table 1.

Five Rounds of RDS in ARISTOTLE: Dates, Number of Participants, and Sociodemographic Characteristics (Unweighted Estimates)

Characteristics of Rounds	Round A	Round B	Round C	Round D	Round E
Dates of recruitment	20 August to 29 October 2012	4 December 2012 to 1 March 2013	19 March to 7 June 2013	17 June to 11 September 2013	19 September to 16 December 2013
Duration of recruitment period, wk	10.0	12.4	11.4	12.3	12.6
All participants, No.	1415	1444	1434	1413	1407
First-time participants, No. (%)	1415 (100.0)	766 (53.1)	458 (31.9)	370 (26.2)	311 (22.1)
New injectors, No. (%)a	150 (10.7)	135 (9.4)	100 (7.0)	114 (8.1)	110 (7.8)
Injectors of drugs within the last 1 mo, No. (%)	1216 (86.3)	1132 (78.7)	1108 (77.7)	1108 (78.8)	1120 (79.7)
Age, mean (SD), y	35.5 (7.8)	35.9 (8.2)	36.2 (8.1)	36.3 (8.0)	36.3 (7.9)
Male sex, %	85.2	83.0	82.6	81.6	82.0
Greece as country of origin, %	85.0	85.7	88.8	88.9	87.6
Currently homeless, %	24.2	21.2	21.7	23.9	22.0
History of imprisonment, %	52.4	48.0	46.4	47.9	48.0

Abbreviations: RDS, respondent-driven sampling; SD, standard deviation.

^aDuration of injecting ≤2 years.

ARISTOTLE Coverage

The capture-recapture estimate for 2012 of the population size of problem drug users in Athens metropolitan area who had injected in the last month is 3069 (95% confidence interval [CI], 2520–3797) [13]. At first participation, 2689 PWID had injected within the last month; thus, the estimated overall coverage of the program was 88% (71%–100%).

HIV Prevalence Per Round

Of 3320 participants, 547 (16.5%) were found to be HIV positive at any participation during the program (95% CI, 15.2%–17.8%). HIV prevalence per round ranged between 12.0% and 16.2% (test for trend P = .27) (Figure 1 and Supplementary Table 1). HIV prevalence was also assessed among “new” injectors (defined as those who had been injecting for ≤2 years) and longer-term injectors. Among new injectors, HIV prevalence declined from 26.1% in round A to 5.8% in round E (Supplementary Materials; Supplementary Figures S1 and S2).

Figure 1.

Human immunodeficiency virus (HIV) prevalence per round among persons who inject drugs participating in ARISTOTLE (respondent-driven sampling–weighted estimates with 95% confidence intervals).

HIV Incidence

In the longitudinal analysis, 1521 PWID with ≥2 samples and who tested negative for HIV at their first visit were followed up for a total of 1004 person-years. During follow-up, 45 seroconversions were recorded, yielding an overall incidence rate per 100 person-years of 4.5 new infections (95% CI, 3.3–6.0). During the program, incidence per 100 person-years decreased from 7.8 (95% CI, 4.6–13) in August–December 2012 to 1.7 (.55–5.3) in August–December 2013 (78% decrease; test for trend P = .001) (Table 2 and Figure 2).

Table 2.

HIV Incidence Rate During ARISTOTLE

Dates	Seroconversions, No.	Time at Risk, y	Incidence/100 PY (95% CI)a
August–December 2012	14	180	7.76 (4.60–13.11)
December 2012 to April 2013	18	306	5.88 (3.70–9.33)
April–August 2013	10	344	2.91 (1.57–5.41)
August–December 2013	3	175	1.71 (.55–5.31)

Abbreviations: CI, confidence interval; HIV, human immunodeficiency virus; PY, person-years.

^aTest for trend P = .001.

Figure 2.

Number of reported human immunodeficiency virus (HIV) type 1 cases per month in Greece (vertical bars), incidence of HIV-1 infection (solid black line), and timing of ARISTOTLE and other interventions (January [Jan] 2010 April 2013). Boosts in needle and syringe programs (NSPs) and opioid substitution treatment (OST) included increased numbers of centers providing OST and initiatives of the Greek Organisation Against Drugs in cooperation with the Hellenic Centre for Disease Control and Prevention and nongovernmental organizations to scale up NSPs. Jul, July; Mar, March; Nov, November; Sep, September.

Risk Factors for HIV Seroconversion

The only variables that remained statistically significant in the multivariable analysis were frequency of injecting, homelessness, and a history of imprisonment (Table 3). The adjusted hazard ratio (HR) among PWID for injection at least once per day compared with less than once per week was 10.0 (95% CI, 2.96–33.9; P < .001). Marginally significant adjusted associations were also identified with history of imprisonment (HR for history vs no history of imprisonment, 1.99; .98–4.04; P = .056) and homelessness (HR for currently homeless versus not homeless, 1.96; .98–3.85; P = .051).

Table 3.

Predictors of HIV Seroconversion in PWID Participating in ARISTOTLE, 2012–2013^a

Predictor	Crude HR (95%CI)	P Value	Adjusted HR (95% CI)	P Value
Sociodemographic characteristics
Age, y
≤35	1		1
>35	1.15 (.64–2.07)	.63	0.97 (.51–1.83)	.92
Sex
Male	1		1
Female	0.69 (.29–1.62)	.39	0.68 (.26–1.76)	.42
Country of origin
Greece	1		1
Middle Eastb	3.64 (1.30–10.21)	.01	1.70 (.56,5.14)	.35
Other	1.26 (.45–3.54)	.66	0.64 (.19–2.15)	.48
Highest completed level of education
Up to secondary school	1
High school or higher	1.13 (.63–2.04)	.68	…	…
Currently homelessc
No	1		1
Yes	1.75 (1.30–2.36)	<.001	1.96 (.98–3.85)	.051
History of any imprisonmentc
No	1		1
Yes	2.40 (1.27–4.53)	.007	1.99 (.98–4.04)	.056
Size of participant’s network, PWIDc
1–10	1		1
11–30	1.12 (.44–2.91)	.81	0.74 (.28–1.96)	.54
>30	2.64 (1.14–6.11)	.02	1.38 (.58–3.28)	.46
Currently in an OST programc
Yes	1		1
No	2.29 (1.13–4.66)	.02	1.35 (.64–2.86)	.43
Injecting drug use behavior
Main substance of usec
Heroin/Thai	1		1
Cocaine/speedball	0.52 (.22–1.22)	.13	1.10 (.45–2.70)	.83
Injecting drug use in past 1 moc
No	1		1
Yes	3.96 (1.22–12.8)	.02	0.81 (.21–3.05)	.75
New injector (≤2 y injecting drug use)c
No	1
Yes	1.66 (.65–4.21)	.29	…	…
Frequency of injecting drug usec,d
Less than once weekly	1		1
At least once weekly	5.55 (1.98–15.6)	.001	5.94 (1.69–20.9)	.005
At least once daily	11.8 (4.5–30.6)	<.001	10.0 (2.96–33.9)	<.001
Sharing syringesc,d
Never or rarely	1		1
About half the time or more	1.72 (1.03–2.88)	.04	0.98 (.22–4.30)	.98
Use of drugs divided with a syringe that someone else had already used for injectionc,d
Never or rarely	1		1
About half the time or more	1.57 (.87–2.82)	.13	1.53 (.45–5.17)	.50
Syringe coverage in past 1 moc
Adequatee	1
Inadequate	1.04 (.52–2.08)	.91
Sexual behavior			…
Same-sex contacts (men)c,d
No	1
Yes	1.85 (.25–13.5)	.54	…	…
Received money/drugs in exchange for sex (women, any partner)c,d
No	1
Yes	1.02 (.77–1.43)	.90	…	…

Abbreviations: CI, confidence interval; HIV, human immunodeficiency virus; HR, hazard ratio; OST, opioid substitution treatment; PWID, persons who inject drugs.

^aThis analysis included 1521 participants who were HIV negative at first participation and provided ≥2 samples; there were 45 seroconversions.

^bAfghanistan, Iran, Iraq, or Pakistan.

^cCovariates varying over time.

^dIn the past 12 months.

^eReceiving as many or more syringes in a month as they needed for their frequency of injection.

Trends in Risk Behaviors During the Program

RDS-weighted proportions of PWID reporting risky injection and sexual behaviors per round are reported in Table 4. There were remarkable changes over time, such as the decline in the proportion of PWID reporting injecting drug use at least once per day in the past 12 months (from 45.2% in round A to 18.8% in round E; P < .001), sharing syringes in the past 12 months (from 36.8% to 26.0%; P = .007), and receipt of used syringes in the last month from ≥2 different persons (from 4.5% to 1.5%; P = .002). The prevalence of high-risk behaviors (frequency of injection, sharing syringes) was similar among first-time and repeat participants (Supplementary Figure S3). There were no changes over time concerning sexual behavior (Table 4).

Table 4.

Homelessness, Imprisonment, Injecting Drug Use, Sexual Behavior, and Access to Prevention and Treatment in the 5 RDS Rounds of ARISTOTLE

Risk Behaviors and Access to Prevention and Treatment	RDS-Weighted Estimate of Proportion (95% CI), %
	Round A	Round B	Round C	Round D	Round E
Homelessness and imprisonment
Currently homeless	15.3 (11.8–18.8)	14.3 (11.1–17.6)	16.0 (12.4–19.7)	15.4 (12.2–18.6)	17.7 (13.8–21.7)
Ever been imprisoned	44.3 (38.5–50.1)	40.7 (35.8–45.6)	39.5 (34.3–44.8)e	41.3 (35.9–46.7)	43.6 (38.6–48.7)
Injecting drug use behavior
Main substance of usea
Heroin/Thaia	85.4 (81.6–89.2)	78.0 (73.7–82.4)	79.7 (75.6–83.8)	70.8 (65.5–76.0)	75.2 (71.0–79.4)
Cocaineb	11.9 (8.6–15.3)	17.9 (13.7–22.0)	17.3 (13.4–21.2)	22.3 (17.7–27.0)	20.0 (16.1–23.8)
New injector (≤2 y injecting drug use)	11.9 (8.0–15.8)	12.8 (9.0–16.5)	8.1 (5.1–11.0)	7.7 (5.2–10.2)	8.1 (5.4–10.8)
Injecting drug use in past 1 mo	77.8 (72.3–83.4)	71.7 (67.0–76.5)	73.8 (69.1–78.5)	72.9 (67.7–78.1)	67.8 (62.4–73.1)
Injecting drug use at least once dailya	45.2 (39.4–51.0)	20.3 (16.2–24.3)	21.3 (16.7–25.9)	18.9 (14.7–23.0)	18.8 (15.2–22.4)
Sharing syringes
In past 12 mo: about half the time or more	8.4 (5.9–11.0)	3.7 (2.4–5.1)	4.9 (2.1–7.7)	6.5 (3.1–9.9)	6.2 (3.3–9.0)
In past 12 moc	36.8 (31.2–42.4)	29.5 (25.2–33.9)	26.0 (21.3–30.7)	29.5 (24.3–34.7)	26.0 (21.9–30.2)
At last injection	19.4 (14.9–24.1)	19.6 (15.4–23.7)	17.8 (13.5–22.1)	24.4 (19.1–29.6)	18.6 (14.9–22.4)
Took used syringes in past 1 mo from ≥2 different personsd	4.5 (3.1–6.0)	2.1 (1.3–2.9)	2.2 (1.0–3.3)	2.6 (1.4–3.8)	1.5 (.9–2.1)
Use of drugs divided with a syringe that someone else had already used for injection: about half the time or more	6.3 (4.5–8.1)	4.4 (2.4–6.4)	2.7 (.8–4.7)	4.3 (1.4–7.2)	3.8 (1.7–5.8)
Sexual behavior
Male-to-male sex	1.2 (.1- 2.3)	2.0 (.9–3.2)	0.8 (.1–1.4)	1.8 (.4–3.1)	1.9 (.7–3.2)
Use of condoms in past 12 mo: usually or alwayse
Male with female partners	58.8 (51.7- 65.8)	59.3 (53.2- 65.4)	54.3 (47.3–61.2)	56.1 (49.3–62.9)	56.5 (50.4–62.6)
Female with male partners	27.0 (16.0–38.0)	46.0 (31.6–60.5)	36.7 (26.8–46.6)	38.0 (24.3–51.7)	26.8 (16.0–37.7)
Women: sex in exchange for drugs/money in past 12 moe	13.8 (6.3–21.3)	25.5 (13.1–37.8)	24.6 (15.9–33.4)	12.0 (6.3–17.7)	11.1 (6.3–15.9)
Access to testing, treatment, and prevention
Participated in drug treatment program (past 12 mo)	45.6 (38.8–53.1)	52.8 (46.5–59.2)	50.5 (44.0–57.0)	44.5 (38.2–50.9)	41.7 (35.9–47.5)
Currently receiving OSTa	12.2 (7.9–16.4)	18.2 (13.8–22.7)	21.4 (16.9–25.9)	24.9 (20.0–29.9)	27.7 (22.8–32.6)
Received syringes through prevention activities (past 12 mo)	49.7 (43.8–55.5)	42.8 (37.9–47.8)	45.9 (40.6–51.3)	46.5 (41.0–51.9)	46.9 (41.8–51.9)
Received syringes through prevention activities (in past 1 mo) (among those who had injected in past 1 mo)f	42.7 (36.7–48.8)	32.7 (27.5–37.8)	43.6 (37.3–49.8)	31.7 (26.3–37.0)	32.3 (27.7–36.8)
Τested for HIV within past 1 ya	50.0 (44.1–55.9)	66.7 (62.0–71.4)	78.7 (74.1–83.3)	81.5 (77.0–86.1)	87.1 (83.5–90.7)
Aware of being HIV infecteda	2.2 (1.1–3.4)	8.2 (5.3–11.2)	11.6 (7.8–15.4)	11.3 (7.6–15.0)	10.3 (6.9–13.7)
Among ΗΙV-positive participants: fraction with undiagnosed HIVa	84.3 (76.2–92.5)	48.8 (34.8–62.8)	28.4 (16.0–40.9)	16.1 (4.3–28.0)	15.0 (8.1–21.8)

Abbreviations: CI, confidence interval; HIV, human immunodeficiency virus; OST, opioid substitution treatment; RDS, respondent-driven sampling.

^a P < .001 (P for trend).

^b P = .001 (P for trend).

^c P = .007 (P for trend).

^d P = .002 (P for trend).

^eAmong those reporting having sex in the past 12 months.

^f P = .01 (P for trend).

Trends in Access to Treatment and Prevention

There was an increase in the proportion of PWID reporting being currently receiving OST: from 12.2% among round A participants to 27.7% in round E (P < .001) (Table 4 and Figure 3A). The proportion of PWID reporting having received free syringes through prevention activities in the past month (either from ARISTOTLE or from elsewhere) decreased from 42.7% in round A to 32.3% in round E (P = .01) (Table 4). To account for the decline in injection frequency during the program, we also estimated the proportion of participants with adequate syringe coverage in the past month (ie, syringe coverage ≥100%). There was an increase from 15.0% in round A to 28.8% in round C, but in the last 2 rounds coverage rebounded to its initial levels (Figure 3A). The extrapolated average self-reported yearly number of free syringes received per PWID in Athens was 243 in 2012 and 161 in 2013.

Figure 3.

Access to prevention and linkage to human immunodeficiency virus (HIV) care during ARISTOTLE. A, Proportion of persons who inject drugs (PWID) reporting currently receiving opioid substitution treatment (OST), proportion reporting adequate syringe coverage in the past month (ie, receiving as many or more syringes in a month as they needed for their frequency of injection), and proportion of HIV-1–positive PWID reporting being undiagnosed per round (respondent-driven sampling–weighted estimates). B, Cumulative proportion of previously unlinked HIV-positive PWID who were linked to HIV care and started highly active antiretroviral therapy (HAART) by the end of each round (unweighted estimates from data collected directly through infectious diseases units).

A considerable increase was observed in the proportion of participants reporting having been tested for HIV within the past year in previous rounds of ARISTOTLE or elsewhere (from 50.0% in round A to 87.1% in round E; P < .001). The fraction of HIV-positive participants who reported having undiagnosed HIV infection at the time of blood sampling declined from 84.3% in round A to 15.0% in round E (P < .001) (Table 4 and Figure 3A).

The proportion of PWID who returned and obtained their HIV test result ranged between 81% and 92% per round. Of 547 HIV-positive PWID identified in total, 435 (79.5%) obtained their HIV test result. Of the remaining 112 PWID who did not return, 28 had reported being aware of their HIV positivity.

Concerning linkage to HIV care, 179 (32.7%) seropositive PWID were already linked before their participation to ARISTOTLE. Of the remaining 368 previously unlinked PWID, by the end of round E, 178 ( 48.4%) visited an infectious diseases unit for the first time after their participation, and 90 (24.5%) initiated ART (Figure 3B).

Trends concerning access to prevention and treatment were more favorable among repeat participants than among first-time participants (Supplementary Figure S4). In the final round, repeat ARISTOTLE participants reported higher rates of adequate syringe coverage in the past month (30.7% vs 21.8%; P = .02) and linkage to ART (among HIV-positive participants, 30.4% vs 6.1%; P = .004) and OST (22.4% vs 15.2%; P = .006) than first-time participants.

DISCUSSION

During an evolving outbreak of HIV among PWID, the ARISTOTLE program was able to reach a very high proportion (88%; 71%–100% based on the uncertainty of the PWID population size estimate in Athens) of the estimated total number of PWID in the last month. The program recorded a strong decline in HIV incidence (from 7.8 to 1.7 per 100 person-years), high HIV prevalence (16.5%), and risk factors for seroconversion. Over the course of the program, injection frequency, sharing used syringes, and undiagnosed HIV infection declined in the population, and OST increased. Repeat participants reported higher rates of adequate syringe coverage, linkage to HIV care, and OST, despite coverage levels remaining suboptimal and overall syringe coverage even declining over time. Our results suggest that multiple evidence-based interventions delivered through rapid RDS recruitment in a large proportion of the target population are likely to have contributed to mitigating this HIV outbreak.

The ARISTOTLE program recorded a decline of 78% in HIV incidence between 2012 and 2013. Despite this large decline, endemic transmission was still occurring at the end of 2013 compared with preoutbreak years, in agreement with molecular analysis conducted on HIV-1 sequences from PWID [14]. Based on data collected from PWID accessing OST programs and low-threshold centers in Athens, HIV prevalence increased from 0.8% in 2010 to 18.3% in 2013 [13]. The latter estimate synthesizes the relatively low HIV prevalence among PWID attending low-threshold services and the high prevalence among those receiving OST (HIV-positive PWID are prioritized to enter OST). During ARISTOTLE, HIV prevalence remained relatively stable (12.0%–16.2%), with an overall HIV prevalence of 16.5%.

Daily injecting drug use and, to a lesser extent, homelessness and a history of imprisonment were independent risk factors for HIV seroconversion in multivariable analysis. In 2015, an increase in the number of HIV cases reported in Dublin occurred among homeless drug users [15]. In our study, migrants from the Middle East (Afghanistan, Iran, Iraq, and Pakistan) were at increased risk of acquiring HIV compared with PWID of Greek origin. This risk was mediated by homelessness and frequency of injection. The vulnerability of these migrants was recognized soon after the outbreak [2], and ARISTOTLE was designed to attract this particularly hard-to-reach subpopulation of PWID, employing cultural mediators from Iran and Afghanistan. These findings underline the need for adequate and timely access to prevention and treatment services for this population, with a specific focus on preventing homelessness, injection risks, and imprisonment in the wider population of PWID.

After adjustment for the frequency of injection (and, to a lesser extent, for injection in the past month), the protective effect of OST on HIV risk was diluted. PWID receiving OST may have had a reduced risk of HIV infection because they reduced their frequency of injection; thus, when frequency was adjusted for, OST was no longer significant. In our analysis, adequate syringe coverage in the past month was not associated with reduced risk of HIV infection, which may be attributed to the inadequate or irregular delivery of NSPs over time.

To explain the decline in HIV incidence and the potential impact of ARISTOTLE, we assessed longitudinal trends in risky behaviors and in OST, NSPs, and ART uptake per round. We recorded a decrease in risky injection behavior, such as daily drug injecting (from 45.2% to 18.8%) and sharing syringes (from 36.8% to 26.0%). Participation in OST, as reported by ARISTOTLE participants, doubled over the 5 rounds (from 12.2% to 27.7%). There was an increase in HIV test uptake (from 50% to 87.1%), and the fraction of undiagnosed HIV-positive participants declined from 84.3% in round A to 15.0% in round E. Given ARISTOTLE’s high coverage of the estimated target population, it can be inferred that the fraction of undiagnosed HIV-positive PWID is likely to have decreased significantly among the overall population of seropositive PWID in Athens. Concerning linkage to HIV care, almost half (48.4%) of previously unlinked seropositve PWID visited an infectious diseases unit, and one-fourth (24.5%) initiated ART after their participation and by the end of the last round. When trends in injection behaviors and access to prevention and care were evaluated separately for first-time and repeat participants, we found a general shift to less risky behaviors among PWID in Athens in both groups. However, repeat ARISTOTLE participants reported higher rates of adequate syringe coverage, linkage to HIV care, and OST, as well as awareness of their serostatus compared with first-time participants.

Despite these favorable changes, coverage of prevention, harm reduction, and treatment during the outbreak in some cases failed to reach acceptable levels. The proportion of PWID who had obtained free syringes from prevention activities in the month before their participation, remained low (31.7%–43.6% per round), with variations in observed coverage from round to round coinciding with periods when there was a general shortage of syringes. In ARISTOTLE, participants obtained syringes during their participation, and as a result the program could not contribute to a large scale-up of NSP coverage. Similarly, levels of OST coverage remained suboptimal (27.7% of PWID reported being receiving OST in the last round). Finally, despite the intensity of this program, linkage to HIV care and to ART remained suboptimal (48.4% and 24.5%, respectively) and far from 90:90:90 targets [16]. Although the success of the program in linking seropositive PWID to HIV care is likely to have been underestimated, because those identified in the last months of the program might have been linked to care after the program was stopped, future interventions should put more effort into linkage to care and ART.

It is difficult to assess to what extent ARISTOTLE may have played a role in the large decline in incidence observed in Athens. Figure 2 shows the number of reported HIV cases among PWID per month during 2010–2014 in Greece [17] and the calendar time periods of ARISTOTLE and of changes in other interventions. A potential explanation could be that the observed decline was part of the natural course of the epidemic, owing to saturation with HIV infection of the core group of highest-risk individuals [18]. In other HIV epidemics among PWID, the reported annual decline in HIV incidence was lower, ranging between 3.6% and 48% [19–34], suggesting that the strong decline in Athens may perhaps not fully be explained by saturation effects. Moreover, apart from the strong decline in incidence, the outbreak in Athens was contained at a relatively moderate HIV prevalence of about 15%, lower than the prevalence recorded in any of a number of countries with similar outbreaks, where peak HIV prevalence ranged between 26% and 66% [24, 35–42].

Although we cannot prove in this analysis whether the stronger decline and lower final HIV prevalence in Athens are due to highly specific local differences in population network characteristics or other unknown factors, we may to some extent be able to control for these by comparing HIV and hepatitis C virus (HCV) prevalence, where HCV prevalence is likely to be determined by similar factors as HIV (assuming that sexual transmission is not large in the population of PWID). It has been found that the levels of HIV prevalence at endemic equilibrium correlate with the levels of HCV prevalence in a given PWID population in which HIV is introduced [43, 44]. In Athens, HCV prevalence among PWID is very high (79.1% in 2012) [13]. In a concurrent HIV outbreak in Romania, a country with equally high HCV prevalence (82.9% in 2009) [45], HIV prevalence reached approximately 50% [37]. This suggests that ARISTOTLE might have had an impact in reducing HIV incidence and keeping prevalence at relatively low levels, despite conditions and behaviors in this population that would have favored higher HIV transmission and prevalence.

The limitations of our study should be considered when interpreting the reported findings. A limitation is that risk behaviors and some key data (eg, OST coverage but not ART linkage) are based on self-reports. For example, the findings that sharing syringes or other paraphernalia were not found to be associated with HIV seroconversion may be attributed to potential underreporting of this risk behavior. Furthermore, the denominator that was used to estimate the coverage of the population of PWID injecting during the last month is the official estimate of the population size in Athens obtained through capture-recapture and is subject to the usual limitations of this method.

ARISTOTLE was designed as a community-based peer-driven outreach program. Participants used the study site not only for HIV testing but also as a drop-in center and a place to receive counseling and social and psychological support. These characteristics may have contributed to the rapid recruitment of PWID and the high coverage of the target population and may have facilitated their engagement with the program, demonstrated by the multiple participations of more than half of recruits. ARISTOTLE was designed as an intensive intervention to reduce transmission during an ongoing outbreak; a staff of 7 full-time and 5 part-time persons (social workers, psychologist, physician, and cultural mediators) were dedicated to interviewing, blood sample collection, counseling, and linkage to care of approximately 28 PWID per day [10], resulting in 7000 visits during a period of 13 months.

In conclusion our findings suggest that prevention of HIV transmission among PWID in an HIV outbreak may be relatively successful, potentially with a high impact on HIV incidence, when using high-coverage, fast-track combined approaches [46, 47]. Future studies and similar intervention programs should pay attention to attaining full coverage of all interventions provided, as well as providing specific and culturally appropriate services to high-risk or hard-to-reach groups, such as migrant and homeless PWID, as well as those with a prison history. The experience of the ARISTOTLE program may be useful for countries that need to provide rapid and massive interventions, particularly in light of ongoing recent HIV outbreaks among PWID in several countries [12, 24].

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Acknowledgments. We acknowledge the contribution of the following persons: from the ARISTOTLE staff, C. Bagos, M. Esmaili, M. Hasan, E. Karamanou, F. Leobilla, C. Mourtezou, E. Sidrou, M. Zigouritsas, M. Dimitropoulou, N. Kaguelari, M. Michail, S. Papadopoulos, and A. Vlahos; from Athens University Medical School, V. Benetou, A. Gkegka, E. Hatzitheodorou, P. Iliopoulos, M. Kantzanou, M. Katsimicha, A. Katsoulidou, G. Kokolesi, V. Milona, H. Papachristou, C. Rocca, M. Souvatzi, S. Tripou, and A. Vassilakis; from the Greek Organisation Against Drugs, A. Panopoulos and K. Gazgalidis; and the volunteers of the nongovernmental organizations Positive Voice and Praksis.

Financial support. This work was supported by National Strategic Reference Framework 2007–2013 (MIS 365008) and cofunded by the European Social Fund and national resources. Additional financial support was provided by the Hellenic Scientific Society for the study of AIDS and STDs and the project “Preventing HIV Transmission by Recently-Infected Drug Users” (NIH National Institute of Drug Abuse, National Institutes of Health; grant DP1 DA034989).

Potential conflicts of interest. All authors: No reported conflicts. All author has submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.