Peer-assisted telemedicine for hepatitis C intervention in people experiencing housing instability

Abstract

Background

The rate of hepatitis C virus (HCV) treatment initiation and cure is suboptimal among people experiencing homelessness. A recent randomized controlled trial conducted by our study team compared peer-assisted telemedicine for HCV (TeleHCV) to peer-assisted referral to local providers (enhanced usual care [EUC]) in a rural population in Oregon. We hypothesized that the TeleHCV treatment model is more effective than EUC among those with unstable housing.

Methods

In this secondary analysis of the Oregon TeleHCV randomized controlled trial in adults with HCV and injection drug use (n = 203), we explored the role of unstable housing as an effect modifier of the TeleHCV intervention using a Poisson regression analysis. Randomized arm, frequency of peer contacts, and medication lockers were exposures of interest. HCV cure, defined as undetectable HCV RNA 12 weeks after treatment initiation, is the primary outcome.

Results

Overall, attaining HCV cure was less likely for those with unstable housing (RR = 0.68, 0.49–0.96, p = 0.026). However, unstably housed participants assigned to the TeleHCV arm were more than 6 times more likely to achieve HCV cure compared to unstably housed EUC participants (RR = 6.47, 3.12–13.43, p < 0.001) with a statistically significant interaction between treatment assignment and housing status (p = 0.022). The impact of TeleHCV on HCV cure was also identified among stably housed participants, but with a smaller effect size (RR = 2.15, 1.19–3.89, p = 0.012). Among participants with unstable housing, greater peer engagement was associated with an increased likelihood of both treatment initiation (RR = 1.24, p < 0.001) and HCV cure (RR 1.26; p < 0.001).

Conclusions

The peer-assisted TeleHCV intervention was substantially more effective for achieving HCV cure among participants with unstable housing at baseline compared to those unstably housed who received EUC. The treatment model should be broadly disseminated to target populations with HCV and high rates of homelessness to achieve HCV elimination goals.

Clinical trials registration

NCT04798521.

Background

Hepatitis C virus (HCV) is one of the most common bloodborne pathogens globally, with 50 million people living with chronic HCV infection [1]. Those at highest risk of transmission include people who use drugs (PWUD), particularly those who share injection equipment. Higher risk injection drug use (IDU) practices, especially sharing injection equipment, is estimated to contribute to up to 43% of all new HCV infections [1]. Substance use disorders, IDU, and HCV transmission, are prevalent in populations with unstable housing, which includes housing challenges or a lack of fixed housing, and homelessness, defined as lacking access to minimally adequate housing [2, 3]. One systematic review found that recent homelessness and unstable housing are associated with increased HCV transmission among PWUD compared to PWUD who are stably housed [3]. Despite high HCV prevalence and the availability of safe, effective treatments, people experiencing homelessness or housing instability have lower odds of treatment initiation compared to people living with HCV who are housed [4, 5]. The risk of reinfection in those experiencing homelessness also exceed that of those stably housed, further supporting the need to target HCV treatment and prevention interventions for this population [6].

The HCV care cascade describes the steps necessary to attain HCV cure, including diagnosis, linkage to care, treatment initiation, treatment adherence and completion, and sustained virologic response 12 weeks post treatment (SVR12) [7]. Unstable housing negatively correlates with treatment initiation, adherence, treatment completion, and SVR12 attainment [8, 9]. A recent meta-analysis of the HCV care cascade found that while rates of HCV screening were high among those experiencing homelessness (78%), they were less likely to be successfully contacted by healthcare providers, linked to care, initiate treatment, complete treatment, and attain SVR12 [10]. Barriers to HCV testing, treatment, and cure disproportionally impact people with housing instability, including: lower HCV awareness, mistrust in healthcare, lack of infrastructure and transportation, and cost [11, 12]. Important facilitators of HCV treatment in people experiencing homelessness include adaptable and flexible modes of delivery, shortened wait times, utilization of peers with lived experience of substance use, and rapid, point-of-care HCV testing [11]. Although acceptable SVR rates can be achieved with missed doses, history of medication non-adherence negatively impacts HCV linkage to care, and greater total adherence days increases the likelihood of SVR [12, 13]. Additional challenges in those with housing instability include medication theft and lack of medication storage [14]. Interventions to alleviate the risk of medication loss concerns need to be evaluated, including medication storage lockers.

Few studies assess interventions to improve progress through the HCV care cascade among people experiencing homelessness. Observational studies indicate that permanent housing support, HCV treatment integrated into homeless shelters, decentralized treatment, and peer support are associated with improvements in HCV care cascade outcomes, including testing, treatment uptake, and cure [15–17]. Prior to this study, none of these interventions have been tested in a prospective randomized controlled trial with an active control. Furthermore, none were tested in a rural area. Telemedicine can expand access to HCV treatment in rural areas with provider shortages, but many rural PWUD with housing instability lack cell phone and mobile network access, elevating the need for peer assistance [18].

Our recently completed randomized controlled trial showed increased rates of treatment initiation (85% versus 12.6%; RR = 6.7, 95% CI 4.0 to 11.3, p < 0.001) and HCV cure (63% versus 15.5%; RR 4.06; 95% CI 2.52 to 6.52; p < 0.001) with peer-assisted telemedicine HCV treatment (TeleHCV) compared to peer-assisted referral to local providers among PWUD in rural Oregon communities [19, 20]. In this rural HCV population, most participants reported housing instability in the previous 6 months and co-occurring fentanyl and methamphetamine use. In this secondary analysis of the randomized controlled trial, we measure how our TeleHCV intervention affects housing instability’s impact on HCV treatment outcomes. The intervention contained components designed to improve HCV outcomes in people experiencing homelessness, including peers, local adaptability, and reduced wait times [11]. Thus, we hypothesized that the TeleHCV peer-facilitated intervention would mitigate the negative impact of unstable housing on HCV treatment outcomes. We also explored the use of medication storage lockers and high frequency of peer engagement to increase HCV treatment initiation and HCV cure.

Methods

This is a secondary analysis of the Oregon TeleHCV randomized controlled trial, which has been previously described [19, 20]. In summary, adults with HCV in seven rural Oregon counties and with recent injection drug use or non-prescription opioid use in the last 90 days were individually randomized to peer-assisted telemedicine HCV treatment (TeleHCV) or peer-assisted referral to local providers (Enhanced usual care [EUC]). Peers in the study were people with lived experience of substance use with at least two years of abstinence and certified in Oregon as peer recovery support specialists, which includes completion of a week-long state-approved training program and biannual re-training and recertification. Persons who were pregnant or breastfeeding, ineligible for health insurance, or with decompensated cirrhosis were excluded [19]. Peer support specialists provided harm reduction resources, temporary housing support, facilitated lab visits, assisted participants in navigating their medical care, and served as communication links between patients and providers for both groups. Peer support specialists were invited to participate in a monthly learning collaborative for peers statewide and participated in weekly meetings to talk through participant status, HCV specific questions, and follow up steps. The primary outcome was HCV viral clearance, defined as undetectable HCV RNA at 12 weeks post treatment initiation for those who initiated treatment, and as undetectable HCV RNA at 9 months post-randomization for those who did not initiate [19, 20]. Those randomized to the TeleHCV arm completed a peer-facilitated provider virtual visit for clinical assessment and treatment initiation. HCV providers included 2 internal medicine physicians, a nurse practitioner, and a clinical pharmacist. All participants in the TeleHCV arm were offered a medication storage locker for safe medication delivery and storage if they did not have a mailing address or safe space to store their HCV medications and utilization of lockers was tracked. Medication lockers were available at each community-based organization location serving the participants in the respective county. The lockers were programmable so participants could choose the locker combination. Peer support specialists assisted in transportation and medication procurement as needed. A study clinical pharmacist monitored and tracked medication initiation, completion of refills, adherence, and side effects [19]. At the time of this study, prior authorization was required for drug approval. Participants were offered cash incentives ranging from $15–$40 for lab draws and surveys. The Oregon Health & Science University institutional review board approved the study protocol (STUDY00020911). Detailed methods, randomization details, study assessments, and peer support activities are described in the published protocol and primary analysis publication [19, 20].

Housing instability

To test the primary study hypothesis, we considered baseline housing instability as an effect modifier of intervention-outcome associations. Housing instability was defined as participant self-report of any homelessness in the past 6 months at enrollment, including living place-to-place, “couch-surfing”, or living on the street, in a car, park, abandoned building, squatting, or shelter.

Exposures

The primary intervention of interest was randomized treatment arm, TeleHCV or EUC. Secondary exposures were frequency of peer contacts and, and among participants assigned to TeleHCV, use of a medication storage and delivery locker. Frequency of peer contacts was defined as the number of times peers provided support or assistance (e.g., goal setting, transportation, study appointment facilitation, provision of harm reduction, medication support) to participants within the first 90 days of the study. Participants in both study arms (TeleHCV and EUC) received support from peers serving their communities and frequency of peer interactions was driven by participant needs and preferences. Peer interactions were not logged for seven participants, and they were excluded from models examining frequency of peer contacts. For modeling purposes, frequency of peer interactions was treated as a continuous exposure, and comparisons were made between the 3rd and 1st quartiles of the observed data (4 or more vs. 1 or 0 peer interactions). Medication storage and delivery lockers were available as part of the telemedicine-facilitated HCV treatment model, and thus, this comparison was only made among participants randomized to the TeleHCV arm. Because peer contacts and use of medication storage lockers were not randomized, analyses of these exposures controlled for number of days of substance use in the past 30 days (range of 0–30; substances included heroin, fentanyl, prescription full-agonist opioids, buprenorphine, prescription anxiety drugs, methamphetamine, gabapentin, and kratom), treatment site, and, for the analysis of medication storage lockers, frequency of peer interactions.

Outcomes

Study outcomes included HCV cure (primary outcome) and HCV treatment initiation (secondary outcome), defined as a pharmacy-confirmed fill of first direct acting antiviral (DAA) prescription plus self-report of initiation. Among those who initiated treatment, HCV cure was defined as undetectable HCV RNA at 12 weeks after treatment. Among those who never initiated treatment, HCV RNA testing occurred at 9 months post-randomization to test for spontaneous clearance. There was no missing data for treatment initiation, but twenty-nine participants did not return for repeat HCV RNA testing [10/100 missing in TeleHCV (10%) vs. 19/103 missing in EUC (18%)]; those participants were classified as not cured. Sensitivity analysis of missing SVR data did not substantially change results. Most participants (82%) were prescribed glecaprevir/pibrentasvir for 8 weeks and the remaining received sofosbuvir/velpatasvir for 12 weeks [20].

Statistical analyses

Sample characteristics and demographics were summarized and compared by participant housing status using t and chi-square tests. The primary analysis for this study compared the effectiveness of the TeleHCV intervention on HCV cure between participants who were stably housed at baseline and those who were not. Poisson regression with robust standard error estimation was utilized with HCV cure as the outcome, treatment assignment as the exposure, and housing status the effect modifier of interest. The primary hypothesis was that there was an effect modification by housing status on HCV cure. The treatment/exposure effect was estimated within each category of housing status (unstably and stably housed). Secondary analyses were conducted in a similar manner. All analyses were conducted using R v.4.3.1. All statistical tests performed were two-sided with a significance level of 0.05.

Results

Participants (N = 203) averaged 42 years old, two-thirds were male, and 88% were White (Table 1). In the 30 days preceding baseline, 62% reported illicit opioid use, 88% methamphetamine use, and 58% both. 82% injected drugs and 15% shared syringes at least once. Over half (54%) reported at least one overdose in their lifetime, but only 11% of participants had accessed medications for opioid use disorder (MOUD) in the past 30 days. 69% (n = 141) reported housing instability in the past 6 months. Participants with unstable housing were more likely to be male and report sharing needles, but these differences were not statistically significant.

Table 1.

Sample characteristics

Characteristic	Overall N = 203	Stably Housed n = 62	Unstably Housed n = 141	p
Age [mean (sd)]	41.6 (11.2)	42.74 (11.9)	41.08 (10.9)	0.331
Sex				0.118
Female (%)	77 (37.9)	29 (46.8)	48 (34)
Male (%)	126 (62.1)	33 (53.2)	93 (66)
Race				0.642
Native American (%)	14 (6.9)	4 (6.5)	10 (7.1)
Black (%)	2 (1.0)	1 (1.6)	1 (0.7)
Mixed race (%)	5 (2.5)	3 (4.8)	2 (1.4)
Other (%)	3 (1.5)	1 (1.6)	2 (1.4)
White (%)	179 (88.1)	53 (85.5)	126 (89.4)
Health insurance (%)	199 (98)	62 (100)	137 (97.2)	0.429
Fib4 score [mean(sd)]	0.90 (0.57)	0.95 (0.53)	0.87 (0.58)	0.384
Past 30-day drug use
Any opioids (%)	126 (62.1)	42 (67.7)	84 (59.6)	0.343
Fentanyl / Heroin (%) *	117 (57.6)	39 (62.9)	78 (55.3)	0.394
Methamphetamine (%)	179 (88.2)	53 (85.5)	126 (89.4)	0.581
Polysubstance Use (opioids and methamphetamine) (%)	117 (57.6)	39 (62.9)	78 (55.3)	0.394
Injected drugs in past 30 days (%)	167 (82.3)	52 (83.9)	115 (81.6)	0.843
Shared needles (%)	30 (14.8)	7 (11.3)	23 (16.3)	0.475
Medications for opioid use disorder (%) †	23 (11.3)	7 (11.3)	16 (11.3)	1
Buprenorphine (%)	13 (6.4)	2 (3.2)	11 (7.8)	0.36
Methadone (%)	11 (5.4)	5 (8.1)	6 (4.3)	0.443
Lifetime history of overdose (%)	110 (54.2)	34 (54.8)	76 (53.9)	1

Overall, 66 of 141 (47%) participants who were unstably housed initiated HCV treatment vs. 32/62 (52%) with stable housing (RR = 0.91, 95% CI 0.67–1.22, p = 0.52). In the overall population, HCV cure was less likely among participants with unstable housing, with 48/141 (34%) cured vs. 31/62 (50%) participants with stable housing (RR = 0.68, 0.49–0.96, p = 0.026).

The effect of the TeleHCV intervention on HCV cure was more evident among participants who reported housing instability at baseline compared to those who were stably housed, p for interaction between treatment assignment and housing status = 0.022 (Table 2). Unstably housed participants who were assigned to the TeleHCV arm were six and a half times more likely to be cured of HCV compared to those in the EUC arm (61.2% cured in the TeleHCV arm vs. 9.5% EUC, RR = 6.47, 3.12–13.43, p < 0.001). Conversely, TeleHCV participants with stable housing at baseline were only about twice as likely to achieve HCV cure compared to EUC participants with stable housing (66.7% cured in the TeleHCV arm vs. 31% EUC, RR = 2.15, 1.19–3.89, p = 0.012). For treatment initiation, the interaction between housing status and randomized group was not significant (p = 0.086), although there was some evidence that the intervention resulted in a larger effect size for treatment initiation among unhoused participants (RR = 9.31; 95% CI 4.58–18.94 among unhoused participants, RR = 3.81; 95% CI 1.83–7.94 among housed, see Table 3).

Table 2.

Associations between randomized arm, peer contact frequency, and medication locker utilization and HCV cure, stratified by baseline housing instability

Randomized group	TeleHCV n = 100	EUC n = 103	RR (95% CI)	p
Interaction				0.022
Stable Housing	22/33 (66.7%)	9/29 (31%)	2.15 (1.19–3.89)	0.012
Unstable Housing	41/67 (61.2%)	7/74 (9.5%)	6.47 (3.12–13.43)	< 0.001
Peer contact frequencya	≤1 contact (1st quartile)c	≥4 contacts (3rd quartile)
Interaction				0.12
Stable Housing	0.45 (0.27–0.75)d	0.47 (0.29–0.77)	1.04 (0.83–1.30)	0.73
Unstable Housing	0.25 (0.16–0.40)	0.32 (0.21–0.49)	1.26 (1.15–1.39)	< 0.001
Medication locker utilizationb	Utilized locker n = 56	Did not utilize locker n = 29
Interaction				0.70
Stable Housing	10/12 (83.3%)	11/15 (73.3%)	0.80 (0.49–1.31)	0.38
Unstable Housing	29/44 (65.9%)	10/14 (71.4%)	0.72 (0.46–1.11)	0.14

^aModels adjusted for frequency of substance use in past 30 days and treatment site

^bModels adjusted for frequency of peer contact, frequency of substance use in past 30 days, and treatment site

^cNumber of peer contacts was included as a linear covariate in analytic models, post-estimation contrasts compare third quartile of peer contacts [4] to first quartile [1]

^dModel-predicted rates of HCV cure

Table 3.

Associations between randomized arm, peer contact frequency, and medication locker utilization and HCV treatment initiation, stratified by baseline housing instability

Randomized group	TeleHCV n = 100	EUC n = 103	RR (95% CI)	p
Interaction				0.086
Stable Housing	26/33 (78.8%)	6/29 (20.7%)	3.81 (1.83–7.94)	< 0.001
Unstable Housing	59/67 (88.1%)	7/74 (9.5%)	9.31 (4.58–18.94)	< 0.001
Peer contact frequencya	≤1 contact (1st quartile)c	≥4 contacts (3rd quartile)
Interaction				0.70
Stable Housing	0.41 (0.25–0.68)d	0.49 (0.31–0.79)	1.19 (1.01–1.41)	0.037
Unstable Housing	0.35 (0.22–0.53)	0.43 (0.28–0.65)	1.24 (1.13–1.35)	< 0.001
Medication locker utilizationb	Utilized locker n = 56	Did not utilize locker n = 29
Interaction				0.63
Stable Housing	11/12 (91.7%)	14/15 (93.3%)	0.93 (0.73–1.19)	0.58
Unstable Housing	41/44 (93.2%)	14/14 (100%)	0.88 (0.77–0.99)	0.049

^a Models adjusted for frequency of substance use in past 30 days and treatment site

^b Models adjusted for frequency of peer contact, frequency of substance use in past 30 days, and treatment site

^c Number of peer contacts was included as a linear covariate in analytic models, post-estimation contrasts compare third quartile of peer contacts [4] to first quartile [1]

^d Model-predicted rates of treatment initiation

Interactions between housing status and frequency of peer contacts did not meet the threshold for statistical significance in either model (treatment initiation p = 0.7, HCV cure p = 0.12). Increased frequency of peer contacts was associated with increased likelihood of treatment initiation among both stably (RR = 1.19; 95% CI 1.01–1.41; p = 0.037) and unstably housed (RR = 1.24; 95% CI 1.13–1.35, p < 0.001) participants. There was little evidence that additional peer interactions predicted HCV cure among stably housed participants (RR = 1.04; 95% CI 0.83–1.30, p = 0.7); however, unstably housed participants who interacted with peers more frequently were 26% more likely to achieve HCV cure (RR = 1.26; 1.15–1.39, p < 0.001) (Table 2).

There was also little evidence for a difference in the associations between locker utilization and HCV treatment initiation or cure by housing status (p values for interactions = 0.63 and 0.7, respectively). There was some evidence that locker utilization was negatively associated with HCV treatment initiation among unhoused participants (RR = 0.88; 95% CI 0.77–0.99, p = 0.049), and no significant associations otherwise (Table 2).

Discussion

The TeleHCV treatment intervention was three times more effective at increasing the rate of HCV cure among participants with unstable housing compared to stably housed participants. A similar effect was found for HCV treatment initiation but did not reach the threshold for statistical significance. These findings are particularly notable given low rates of HCV treatment and cure among unstably housed people worldwide. In a meta-analysis of 65 observational and interventional studies from North America, Europe, and Australia (n = 7,402,185), individuals experiencing homelessness had the lowest rate of HCV treatment initiation (25% of people experiencing homelessness with HCV RNA positivity) and HCV cure (17% of people experiencing homelessness with HCV RNA positivity) among six author-identified “priority populations” including PWUD, people with HIV, “baby boomers,” and “others.” [10] Similarly, a meta-analysis of four studies (n = 1,550) found lower odds of HCV treatment initiation among those with unstable housing (OR = 0.40; 95%CI = 0.26–0.92) but high heterogeneity of studies prevented meta-analysis of HCV cure [5].

Several features of the TeleHCV intervention likely contributed to its profound impact on unstably housed participants. These include unscheduled “on-demand” and same-day provider telemedicine visits, the willingness of peers to flexibly meet with participants in encampments and temporary shelters, temporary housing assistance, and the provision of telecommunications by peers [19]. Permanent supportive housing improves liver-related outcomes among people with HCV and participants anecdotally reported that temporary housing supported HCV treatment initiation, but the individual effect of these temporary housing interventions as a component of the TeleHCV model are unclear [16]. We hypothesized that two other specific components of our intervention, use of medication storage lockers and frequency of peer contacts, would improve HCV treatment and cure.

Medication lockers as part of an effective peer-facilitated telehealth program among urban people living with HIV are supported by qualitative evidence [21]. In our study, lockers had an unclear effect on HCV treatment initiation and cure, with evidence of a negative association with treatment initiation and no change to HCV cure. This negative association must be interpreted cautiously because locker use was optional and not randomized. Participants who opted into medication lockers may have had unmeasured complicating factors and follow up challenges confounding the results and locker proximity to participants may have differed between counties. While unstably housed and homeless individuals are often categorized together for research purposes, there is a wide range of social vulnerability and availability of resources within these populations, and we were not able to analytically account for these [3]. Furthermore, the analytic sample for locker use was small because lockers were only used among those randomized to TeleHCV and was further reduced by the operational definition of locker use (delivery address of the first medication fill) not being identified for all participants. Pragmatically, the specialty pharmacies that fulfilled orders of direct acting antivirals for study participants required a physical address to be listed by participants as a condition of medication delivery. Thus, lockers allowed many participants who lacked secure mailing addresses or social contacts with secure addresses to receive medications from specialty pharmacies in a system which would otherwise have excluded them.

There was evidence that increased frequency of peer contacts was related to HCV cure among participants with unstable housing. Although the interaction between housing status and peer contacts was not significant, the validity of this finding is bolstered by other studies of peers in HCV treatment settings. A randomized trial of people with HCV in the UK who were recruited from substance use and homelessness services settings (n = 101) demonstrated an increased rate of engagement with HCV treatment (OR = 2.55; 95%CI = 0.97–6.70, p = 0.06) in the group randomized to peer services versus standard care [22]. A peer-based HCV treatment program in a setting with high rates of houselessness has also demonstrated that peer-based services are effective in real-world settings [23]. A mixed methods systematic review described several mechanisms by which peers facilitate favorable HCV treatment outcomes for people experiencing homelessness, including forming trusting relationships with people being treated for HCV, peer-initiated education, and advocacy [11].

This study has several strengths. We focused on two patient-centered outcomes -- treatment initiation and HCV cure -- because these are most clinically relevant for meeting World Health Organization hepatitis C elimination goals. HCV treatment initiation is linked to HCV cure in general populations and among people experiencing homelessness [10, 13]. HCV cure is the most clinically relevant outcome given its association with decreased morbidity, cirrhosis, hepatocellular cancer, and mortality [24]. This is also the first study to demonstrate an effective intervention to mitigate the negative impact of unstable housing on the HCV care cascade in rural areas.

The study also has limitations. Primarily, analyses of effect modification often have very low power and are highly susceptible to type-II errors. We aimed to follow best practices for completing and reporting these types of analyses [25]. However, as our hypotheses were not pre-specified in the study protocol, results should be considered exploratory. We did not measure and could not analyze several features of the TeleHCV model. For example, some insurance providers in Oregon provide temporary housing throughout the HCV treatment, but this was not systematically tracked in the study. The analysis of lockers was limited in number, and some comparisons had very few participants. Lastly, there are inherent differences among the various counties included in the study, such as availability of harm reduction resources and differences in specialty pharmacy requirements, that could impact rates of treatment initiation and cure.

Conclusion

Peer-assisted TeleHCV is highly effective at improving HCV treatment initiation and HCV cure in rural participants experiencing homelessness. This finding is extremely relevant to achieving equity in national efforts to eliminate HCV. Future efforts to improve HCV outcomes among unhoused populations should preserve the key elements of the TeleHCV model, including peer facilitation and low barrier access to telemedicine.

Abbreviations

DAA

Direct acting antiviral

EUC

Enhanced usual care

HCV

Hepatitis C Virus

IDU

Injection drug use

MOUD

Medications for opioid use disorder

PWUD

People who use drugs

RNA

Ribonucleic acid

SVR12

Sustained virologic response 12 weeks post treatment

TeleHCV

Peer-assisted telemedicine for HCV

Author contributions

Conceptualization, M.H.,H.S., R.C., D.G., A.S., G.L., J.B., and P.T.K.; methodology, R.C. and D.G.; software, D.G.; formal analysis, D.G. and R.C.; investigation, M.H., H.S., G.L., A.S., J.B. and P.T.K.; resources, P.T.K.; data curation, D.G. and R.C.; writing—original draft preparation, M.H., H.S., D.G. and R.C.; writing—review and editing, all authors; visualization, D.G. and R.C.; supervision, P.T.K.; project administration, G.L.; funding acquisition, P.T.K. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by Work was supported by the National Institutes of Health, National Institute on Drug Abuse grant numbers UH3DA044831, UG1DA015815, K01 DA55130.

Data availability

De-identified data are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The Oregon Health & Science University institutional review board approved the study protocol (version 16) and provided oversight. Informed consent was obtained from all subjects in the study.

Consent for publication

Not applicable.

Competing interests

Andrew Seaman has participated in investigator-initiated research funded by Merck, Abbvie Pharmaceuticals, and the Gilead Focus Foundation not related to this manuscript. All other authors have nothing to disclose.