HIV Incidence among People Who Inject Drugs, San Francisco, 2022

Abstract

Background:

Overall progress in getting to zero new HIV infections in San Francisco has been made since 2012. However, among people who inject drugs (PWID) there has been no clear downward trend in new HIV infections in recent years. Direct measures of the rate of HIV acquisition and characterization of factors associated with higher rates among PWID are needed to help achieve HIV elimination.

Methods:

This study is a secondary analysis of the National HIV Behavioral Surveillance survey of PWID in San Francisco in 2022. HIV incidence is estimated using age of first injection drug use and date of first HIV-positive test or interview date as the exposure period. Factors associated with HIV seroconversion were identified by Cox regression.

Results:

Of 518 PWID, 12 newly tested HIV positive in the survey and 38 reported a prior positive test. HIV incidence was calculated as 0.46 per 100 person-years (95% CI 0.35–0.61). HIV incidence rates were significantly higher among PWID who were men who have sex with men (adjusted hazard ratio [aHR] 15.98, 95% Cl 7.77–32.87, p<0.001), transgender (aHR 8.64, 95% Cl 2.76–27.08, p<0.001), and Hispanic (aHR 2.85, 95% Cl 1.23–6.63, p<0.001).

Conclusions:

We found a moderate HIV incidence of ~5 per 1,000 person-years among PWID in San Francisco, with significantly higher rates among sexual, gender, and ethnic minority groups of PWID. Further progress in getting to zero new HIV infections will require more vigorous scale-up of effective prevention interventions, such as PrEP, specifically reaching vulnerable groups of PWID.

INTRODUCTION

From 2012 to 2022, San Francisco witnessed a substantial decline in reported new HIV diagnoses (from 478 to 157), signaling progress toward the goal of “Getting to Zero” new infections.¹ Unfortunately, the last three years of data suggest progress has stalled in reducing HIV incidence among people who inject drugs (PWID), with 27 cases reported in 2020, 40 in 2021, and 30 in 2022.¹ The number of new HIV diagnoses, however, may not reflect underlying incidence if there are delays in diagnosis and reporting of persons acquiring HIV. Direct measures of the rate of HIV seroconversion (i.e., incidence) are therefore needed to track true progress in eliminating transmission and identify groups at highest risk for acquiring infection.

Unfortunately, HIV incidence estimates are difficult to obtain, and all available methods have potential biases.² Prospective cohort studies, considered the gold standard for measuring incidence, are time-consuming, expensive, and subject to limitations including selection and participation biases, loss of follow-up, and a prevention effect on HIV incidence due to long-term study participation. Participation in longitudinal studies and loss to follow-up may be particularly challenging for PWID who experience multiple structural barriers, including homelessness, incarceration, food insecurity, and competing health care needs as well as high mortality.^1,2 Serological assays that identify persons recently infected in cross-sectional studies^2,3 require very large sample sizes to provide precise estimates, a feat that is difficult for marginalized populations such as PWID. There is therefore need for a simple, readily available method to estimate HIV incidence from cross-sectional surveys of hard-to-reach populations, particularly PWID. Our objective for the current study is to provide an updated estimate of HIV incidence among PWID in San Francisco using available, community-based behavioral survey data.

METHODS

Our estimation of HIV incidence among PWID was a secondary a secondary analysis of the National HIV Behavioral Surveillance (NHBS) survey conducted in San Francisco in 2022. The NHBS is a biobehavioral surveillance system operating since 2003 in approximately 20 US cities with high incidence of AIDS.⁴ The full protocol is available online.⁴ In brief, PWID are recruited through respondent-driven sampling, which entails the recruitment of initial eligible “seed” PWID who are then instructed to recruit other eligible PWID to the study. Long chains of recruitment grow through the social networks of PWID eventually branching out into different groups independently of the starting seeds including those who may not access health and prevention services or are unstably housed. For example, 67% of PWID participating in NHBS had been homeless in the last year.⁵ An anonymous, interviewer-administered questionnaire collects information on demographic characteristics, risk and preventive behaviors, HIV testing, and drug use history. After the interview, participants undergo rapid HIV testing regardless of whether they self-report as previously testing positive. Eligibility criteria are presenting a recruitment coupon from another participant, living in San Francisco or San Mateo counties, age ≥18 years, and having injected drugs without a prescription in the past 12 months. Participants were given a $75 cash incentive for completing the survey and $10 for each eligible peer they referred to the study, for up to five referrals.

Osmond and colleagues developed an approach to estimate HIV incidence rates in cross-sectional surveys calculated by dividing the number of HIV-positive cases detected by the person-years (PY) at risk determined by the date of onset of risk behavior, such as initiating injection drug use.³ The approach has been used to estimate HIV incidence among men who have sex with men in San Francisco and elsewhere in the US and Puerto Rico,^6,7 Amsterdam,⁸ Bangkok,⁹ the Republic of Georgia,¹⁰ Vietnam,¹¹ and six cities in China.¹² We applied the method to estimate HIV incidence in the cross-sectional data collected for the 2022 round of NHBS among PWID in San Francisco. PY at risk was estimated from participants’ self-reported age they first started injecting drugs to the date of the NHBS interview for HIV-negative participants. PY at risk for HIV-positive participants was estimated from the age of first injection to the midpoint (i.e., to account for interval-censored data) of one of four possible dates: (a) the date of last negative HIV test and first positive test if the participant reported the dates of both tests, (b) the date of last negative HIV test and interview date if the participant reported the date of the last negative HIV test only, (c) the date of first HIV positive test if the participant reported the date of first positive HIV test only, or (d) the interview date if participant did not report testing or recall the date of HIV testing. For participants who tested positive within one year of their age at first injection, time at risk was considered as one year. To identify groups of PWID with higher HIV incidence, we examined seroconversion rates by demographic characteristics, sexual orientation, behaviors, and drug use histories using the log-rank test and Cox regression analysis (Stata version 17, Stata Corp LLC, College Station, TX). Proportional hazards (PH) assumptions were assessed using Schoenfeld residual statistics. P-values <0.05 were considered statistically significant.

RESULTS

Overall, 527 PWID participated in the survey. Of these, 9 (2.1%) were excluded because they reported that their first HIV-positive test was before they started injecting drugs. Of the 518 included in the current analysis (Table 1), 68.0% identified as male, 29.9% as female, and 2.1% as transgender. Of note, the transgender category did not record sex assigned at birth, although anecdotally most, if not all, were trans women. The sample was diverse with respect to race/ethnicity, with a plurality (47.5%) identifying as non-Hispanic White, followed by Black (31.3%), and Hispanic (12.3%). By behavior and reported sexual orientation, 4.9% were men who had sex with men (MSM).

Table 1.

HIV incidence rates by background characteristics among people who inject drugs in the National HIV Behavioral Surveillance (NHBS), San Francisco, 2022

Characteristic	Total n (%)	HIV incidence		Incidence per 100 person-years (95% CI)	P-value
		HIV-positive n (%)	Person-years of exposure
Total	518 (100)	51 (100)	11051	0.46 (0.35–0.61)	--
Age group (years)					0.597
18–29	25 (4.8)	1 (2.0)	190	0.52 (0.07–3.73)
30–39	108 (20.9)	12 (23.5)	1306	0.91 (0.52–1.61)
40–49	143 (27.6)	11 (21.6)	2365	0.46 (0.25–0.84)
50+	242 (46.7)	27 (53.0)	7189	0.37 (0.25–0.54)
Age at first drug injection (years)					0.98
≤18	157 (30.3)	18 (35.3)	4786	0.38 (0.24–0.59)
19–30	236 (45.6)	21 (47.1)	4952	0.48 (0.32–0.72)
30+	125 (24.1)	9 (17.7)	1313	0.68 (0.35–1.3)
Year of first drug injection					0.365
<1990	139 (26.8)	18(35.3)	5220	0.34 (0.21–0.55)
1990–99	109 (21.0)	9 (17.7)	2851	0.32 (0.16–0.60)
2000–09	126 (24.3)	14 (27.5)	2002	0.70 (0.41–1.18)
2010–14	67 (12.9)	8 (15.7)	639	1.25 (0.63–2.50)
2015–21	77 (14.9)	2 (3.9)	337	0.59 (0.15–2.37)
Gender					<0.007
Male	352 (68.0)	36 (70.6)	7581	0.47 (0.34–0.66)
Female	155 (29.9)	11 (21.6)	3277	0.33 (0.19–0.60)
Transgender1	11 (2.1)	4 (7.8)	193	2.07 (0.78–5.51)
Education completed 2					0.273
Below secondary school	86 (16.7)	13 (25.5)	2099	0.62 (0.36–1.06)
Secondary	210 (40.7)	15 (29.4)	4727	0.31 (0.19–0.52)
Some college	187 (36.2)	19 (37.3)	3582	0.53 (0.33–0.83)
Bachelor’s degree	33 (6.4)	4 (7.8)	502	0.66 (0.24–1.76)
Race/ethnicity 2					<0.001
Non-Hispanic Black	161 (31.3)	11 (21.5)	4004	0.27 (0.15–0.49)
Non-Hispanic White	244 (47.5)	23 (45.1)	5064	0.45 (0.30–0.68)
Hispanic	63 (12.3)	14 (27.5)	968	1.44 (0.85–2.43)
Non-Hispanic others	46 (9.0)	3 (5.9)	934	0.32 (0.10–0.99)
Sexual orientation/behavior 2					<0.001
Heterosexual men and women	377 (73.8)	27(53.0)	8106	0.33 (0.23–0.48)
Men who have sex with men (MSM)	25 (4.9)	16 (31.3)	270	5.92 (3.62–9.66)
Bisexual, other	109 (21.3)	8 (15.7)	2516	0.31 (0.15–0.63)

¹The category did not record sex assigned at birth.

²Categories do not add to total due to missing data.

A total of 51 (9.8%) were HIV positive in the survey. Of these, 38 (74.5%) reported a previous positive test; 12 (25.5%) had been unaware of their HIV-positive status prior to testing in the survey. Calculating the exposure period as described above produced 11,051 PY of exposure for an HIV incidence of 0.46 per 100 PY (95% Cl 0.35–0.61) experienced by this 2022-recruited sample of PWID. Highest incidence was observed among PWID who were MSM at 5.92 per 100 PY (95% Cl 3.62–9.66, p<0.001 compared to heterosexual and bisexual identified PWID), followed by transgender identity at 2.07 per 100 PY (95% Cl 0.78–5.51, p=0.007 compared to participants who identified as cis- male or female), and PWID of Hispanic origin at 1.44 per 100 PY (95% Cl 0.85–2.43, p<0.001 compared to other ethnic groups). Multivariate Cox regression analysis confirmed that higher HIV incidence rates were independently associated with PWID who were MSM (adjusted hazard ratio [aHR] 15.98, 95% Cl 7.77–32.87, p<0.001), transgender (aHR 8.64, 95% Cl 2.76–27.08, p<0.001), and Hispanic (aHR 2.85, 95% Cl 1.23–6.63, p<0.001).

DISCUSSION

Our analysis produced a moderate HIV incidence rate of approximately 5 per 1,000 person-years among PWID in San Francisco using cross-sectional data from the NHBS in 2022.⁴ Similar to previous studies in San Francisco,^13.14 we found significantly higher HIV incidence among PWID who were MSM, highlighting the especially high risk of HIV acquisition in this population with potential intersecting risks with high prevalence populations through sexual and needle-sharing networks. At least one longitudinal study also found HIV incidence higher among Hispanic PWID compared to other racial/ethnic groups.¹⁴ New to our study, we detected that transgender persons who inject drugs experienced significantly higher incidence of HIV compared to PWID of other gender identities.

Although methods to estimate incidence differ, our data fit into a picture of the course of the HIV epidemic among PWID in San Francisco when aligned with previous studies. HIV incidence estimated from applying a recency assay to serological specimens collected from PWID from 1987 to 1998 suggested a significant decline in HIV incidence from 2.7% per year in 1987/88 to 1% per year and stable from 1989 through 1998.¹³ The investigators attributed the decline in HIV incidence during this phase of the epidemic as due to the early adoption of harm reduction programs in San Francisco, including community outreach education, syringe exchange programs, and treatment on demand. A longitudinal study conducted among young PWID in San Francisco showed a continuing incidence rate near 1 per 100 PY from 2000 to 2014, fluctuating between 0.93 and 1.06.¹⁴ Our data infer that PWID living in San Francisco in 2022 experienced an average annual HIV incidence roughly half that observed in this study. Although our cross-sectional survey data cannot definitively conclude why there has been a drop in HIV incidence in the current decade compared to the previous, there are several possible reasons. First, higher levels of antiretroviral treatment and viral suppression were achieved among PWID by 2022 compared to the previous decade, potentially resulting in more effective suppression of HIV transmission at the population level.¹⁵ For example, viral suppression had scaled-up to 58% of PWID living with HIV by 2014,¹⁶ while reaching a plateau of 82% to 84% from 2017 to 2022.¹ Second, it is possible that preventive behaviors associated with injection practices have been lower since 2014. For example, in prior rounds of NHBS, never sharing injection equipment increased from 74% of PWID in 2005 to 85% in 2012 and sustained at 86% through 2018.^17,18 Another hypothesis is that the uptake of fentanyl use could decrease injection frequency with subsequent decrease in HIV transmission risk. However, this causal mechanism has not been shown and the relationship between fentanyl use and opioid injection appears complex. At least one study has found that increased reporting of fentanyl use was associated with higher frequency of opioid injection.¹⁹

Further gains in reducing HIV incidence among PWID may be challenging. Surveillance data find little change in the number of new HIV diagnoses reported among PWID in the last three years and little change in viral suppression among those on treatment.¹ Increase use of PrEP may hold the greatest potential for reducing HIV incidence among PWID, particularly given the nearly negligible use of PrEP by PWID so far.²⁰ Data indicate considerable “upstream” barriers, such as limited discussion of PrEP between healthcare providers and PWID patients and lack of awareness of PrEP’s ability to prevent transmission through injection practices.²⁰ There are likely multiple other structural barriers to PrEP use by PWID, including housing instability, incarceration, and mobility.²¹ Studies are needed on the most effective ways to increase uptake and persistence on PrEP by PWID, including the potential impact of long-acting PrEP.

We recognize limitations to our study. First, data on the exposure period and past HIV test results were obtained by self-report, which is subject to measurement error, particularly given the long periods of time entailed for many. Second, as noted above, the long exposure periods and intervals between HIV tests make the timing of seroconversion imprecise, limiting interpretation of the current HIV incidence and trends over time. Third, the number of seroconversions for many categories of predictor variables (e.g., transgender identified PWID) lead to imprecise incidence estimates and limit the number of variables we were able to consider. Despite the small numbers for some categories, we highlight the efficiency of estimating HIV incidence from cross-sectional surveys. A longitudinal study of PWID to prospectively identify a comparable number of HIV seroconversions may be difficult to recruit or prohibitively long to follow.

In conclusion, we were able to produce a credible estimate of HIV incidence among PWID that is consistent with the long-term downward trend in new cases, past studies of PWID, and in identifying groups of PWID at higher risk for HIV acquisition. Unfortunately, we also demonstrate that HIV transmission continues at a moderate rate and may be hitting a nadir if short-term trends in new cases prevail. Further gains will especially need to deliver effective HIV prevention to PWID in sexual, gender, and ethnic minority populations where transmission is occurring fastest if we are to achieve zero HIV infections by 2030.