A Systematic Review of Alternative Nicotine Delivery System (ANDS)-Based Interventions on Smoking-Related Health Outcomes Among Marginalised Population Groups.

Background: Smoking and smoking-related disease is endemic amongst many marginalised populations such as people experiencing homelessness, people who use drugs, people living with HIV, sex workers and members of the LGBTQ+, Indigenous, Traveller and migrant communities. Alternative Nicotine Delivery Systems (ANDS)-based interventions including those using Nicotine Replacement Therapy (NRT) and Electronic Nicotine Delivery Systems (ENDS) show promise in supporting people to reduce their smoking. However, little is known about ANDS-based smoking reduction interventions with marginalised populations. This systematic review provides a best evidence synthesis of ANDS-based smoking reduction interventions to address this gap. Methods: A systematic review protocol was registered on PROSPERO (CRD42020158832) and literature review conducted through MEDLINE/PubMed, CINAHL, EMBASE, OVID SP, ScienceDirect and Google Scholar. The primary outcomes of cigarette smoking, and biochemical validation of abstinence were reported. Secondary outcomes reported included physical health, mental health and other outcomes. use of ENDS with marginalised populations. Future studies should aim to address these decits.


Introduction
Although tobacco smoking has been decreasing in many higher income countries, it remains endemic amongst many marginalised populations in both lower and high income countries. 1,2 Many of these groups are disproportionately impacted by smoking-related diseases such as Chronic Obstructive Pulmonary Disease (COPD) and lung cancer and smoking can dramatically reduce their quality of life and life expectancy. 2 However, there are limited reviews of evidence-based smoking reduction interventions outside speci c marginalised groups such as people experiencing homelessness 3,4 A wide range of marginalised populations have been associated with high smoking rates including people who use drugs 5 , people living with mental illness 5 and people involved with the criminal justice system 7 . Such minority populations are often excluded from smoking reduction interventions despite their high levels of smoking and smoking-related disease. 8 The lack of engagement with these groups may be related to a need to identify effective smoking reduction interventions with these groups. 8 Alternative nicotine delivery systems (ANDS) including nicotine replacement therapy (for example, gums, lozenges and inhalers) and the provision of electronic nicotine delivery systems (ENDS) show some promise in helping a range of marginalised populations to quit or reduce smoking. 2,3 The systematic review reported in this paper aimed to provide a comprehensive summary of intervention research that has examined the use of ANDS-based interventions in reducing tobacco smoking and other key outcomes in marginalised populations. Within this review marginalised populations de ned were people with experience of homelessness; criminal justice involvement; drug use; severe mental illness; sex work; forced displaced or migrant people; people living with HIV and members of the Indigenous, LGBTQ+ and traveller communities.
The key outcomes of interest in this inquiry were: (I) Self-reported cigarette consumption; (II) Biochemically validated cessation (II) Measures of nicotine dependence (II) Mental outcomes (IV) Physical outcomes (VI) Other outcomes Methodology Our review protocol was registered with PROSPERO (CRD42020158832) and followed the PRISMA guidelines. 9 We conducted the systematic review to summarise evidence for the use of ANDS interventions among various marginalised population groups which are linked to high smoking rates. We de ned ANDS as including all nicotine-based interventions including various forms of nicotine replacement therapy (for example, nicotine gum, patches and inhaler), ENDS devices and snus.
Studies which were included were those published in peer-review academic journals in English. This study focused broadly on a range of marginalised populations, which were de ned by expert consensus. These included people with lived experience of homelessness; criminal justice involvement; drug use; severe mental illness; sex work; forced displacement or migration; people living with HIV and members of the Indigenous, LGBTQ+ and traveller communities.

Inclusion criteria
To be included in the review, studies were required to have a (i) primary ANDS intervention either controlled or uncontrolled with pre and post measures, (ii) be subject to peer review publication in an academic publication and (iii) include primarily marginalized population group (s) in the intervention -as per the search strategy described below.

Search methods and strategy
A systematic search of the literature was conducted to identify all intervention studies where ANDS-based interventions had been implemented with 10 separate marginalised population groups as shown below. Table 1 Search Terms 1. homeless OR houseless OR "precariously housed" OR unhoused; 2. prisoner OR imprisoned OR prison; 3. "person who uses drugs" OR PWUD OR "person who injects drugs" OR PWID OR "drug user" OR "substance use disorder" OR "drug dependence"; 4. Indigenous OR "Native American" OR Maori OR "First Nations"; 5. Roma OR gypsy OR Traveller; . "serious mental illness" OR "severe mental illness" OR "severe and persistent mental illness"" OR "common mental disorder" OR "common mental illness" OR schizophrenia OR psychosis OR psychotic OR "major depression" OR depression OR dysthymia OR bipolar OR anxiety OR "posttraumatic stress disorder"; 7. "sex worker" OR prostitute OR "rent boy"; . "Asylum seeker" OR migrant OR refugee; 9.
LGBT OR lesbian OR gay OR bisexual OR transgender OR intersex OR queer 10. "People living with HIV" OR PWLHIV OR "HIV patients".
The primary outcomes of cigarette smoking, and biochemical validation of abstinence were reported. Secondary outcomes included physical, mental and other outcomes. Articles from inception to 2020 were included. Articles were reviewed from MEDLINE/PubMed, CINAHL, EMBASE, OVID SP, ScienceDirect and Google Scholar.

Study selection process
Three reviewers worked independently to screen articles for inclusion eligibility using Covidence. Where disagreement occurred, a discussion was held and decisions were put to a third independent author (MG, JSW) for a nal decision. The data were extracted using the predetermined outcomes. These outcomes were classi ed under domain areas to allow for summary data extraction and analysis. The generated domain areas are shown in Table 1 below. Controlled Trials was adopted for use in this process. 10 In analysing the results, a best evidence synthesis was conducted, whereby studies with a control group and studies without control measures were considered separately and data, as per Table 1, were extracted from each study. A meta-analysis of data was precluded given the diverse heterogeneous nature of the studies that were included in the nal review.

Study selection
Following the removal of duplicates, a total of 2463 articles were considered at the title and abstract level. One-hundred-and-seven full texts were reviewed, and 77 were excluded with reasons (see Fig. 1 and Appendix 1), with 29 unique studies meeting the eligibility criteria. Full details of the search results are summarized in Fig. 1. PRISMA Flowchart.

Details of included studies and participant's characteristics
Thirteen of the included studies were randomised controlled trials (RCTs) (n= 2828) and 16 studies were quasi controlled studies (n=1172). The sample size across the studies ranged from 9 36 to 775 participants. 21 The mean age of participants ranged from 32.4 years 18 to 56.9 38 with an overall mean age of 39.55 years (three studies did not report mean age impact on the 7-day point prevalence quit rate (both p = .03). 12 However, this effect was no longer statistically signi cant at 38 or 52 weeks (ibid.). A study using an ENDS device 18 found that the intervention was effective in reducing self-reported cigarette consumption with a biochemically veri ed (<8ppm CO) end of treatment smoking abstinence rate of 8.6% (intent-to-treat) and 10.7% for the per protocol group (n=28). Twenty-eight percent of those who completed the program reported a 50% reduction in cigarette consumption, however, this did not seem to have been corroborated by biochemical veri cation (ibid.). A NRT based intervention, which included a computerised feedback component, did not increase motivation to quit or increase long term abstinence. 20 Two RCTs reported mental health outcomes of interventions with people living with mental illness. One study 11 investigated the effect of NRT in reducing agitation in people living with schizophrenia as measured by the Mean Agitated Scale score over a 24-hour period. The mean Agitated Behaviour Scale score for the group receiving NRT was 33% lower at 4 hours and 23% lower at 24 hours than the placebo group (ibid.). Another study, 13 also in an inpatient unit, compared the effect of patches, gum and inhalers, and found that the groups using inhalers or gum had signi cantly reduced psychiatric symptom severity symptoms at weeks 4 and 12 with the inhaler group showing the greatest reductions. Only the inhaler group was found to have a signi cant reduction in CO at weeks 4 and 12 (P < 0.001 and P = 0.032, respectively) but both the nicotine inhaler and the patch groups showed signi cant decrease in nicotine withdrawal symptoms from baseline at both weeks 4 and 8 (ibid.).
Two studies considered NRT pre-loading (pre supplementation with nicotine) to increase treatment e cacy for people living with mental illness. 14, 17 One study 14 which investigated the effect of NRT pre-loading with a 21mg nicotine patch found that those receiving the intervention were more likely to be abstinent (3 months 8.5% vs. 1.0%, respectively, p = .01) and were more likely to accept four or more cessation sessions (50.8% vs 29.2%, respectively, p < .001). However, a study 17 exploring NRT pre-loading with another population of people with serious mental illness found no statistically signi cant reduction in smoking or craving during the preloading period and no differences in terms of 6 week or 6-month abstinence when compared to the control group.
A 10 week group intervention with NRT with women in prison found that point prevalence quit rates for the intervention group, as compared to a waitlist control, were 18% at end of treatment, 17% at 3-month follow-up, 14% at 6-month follow-up, and 12% at 12-month follow-up. 15 An intervention with men in prison found that Motivational Interviewing (MI) + NRT was more effective than MI only in reducing smoking as biochemically validated with CO (p = 0.02). 22 An intervention with pregnant Indigenous women with online sessions, NRT and CO monitoring, evaluated through a step cluster design, led to biochemically validated (<6ppm) 12 week 7-day point-prevalence abstinence in 13.% (n=3) of participants. 19 A study exploring the effect of NRT contacts with people living with HIV found that the number of contacts was signi cantly associated with 6 month abstinence. 23 Another study with people living with HIV found that a 8 week NRT + counselling intervention found that the treatment group smoked less cigarettes per day(η2 = .07; p = .13); had lower physical nicotine dependence(η2 = .18, p = .01) and lower smoking urges (η2 = .16, p = .01). 16 A novel study 21 compared two very low nicotine cigarettes with one higher nicotine cigarettes in an intervention with individuals with affective disorders, individuals with opioid use disorder and women smokers with socioeconomic disadvantage. The very low nicotine cigarettes led to reductions in cigarettes per day but there were no signi cant differences between the two very low nicotine cigarette types (ibid.).

Quality Appraisal of Controlled Trials
The overall quality of the controlled trials, as evaluated using the JBI Appraisal Tool, was found to be low with an average of 5.92/13 with a range from 3-9/13. Almost fty percent of studies did not use or did not clearly describe true randomisation procedures (6/13 studies). All studies did not clearly describe concealing allocation to the treatment group (13/13 papers). Many of the studies (9/13 studies) reported that the treatment groups were similar at baseline. Only two of the 13 studies, clearly outlined that their participants were blind to treatment assignment whilst only one study clearly outlined that those delivering treatment were blind to treatment allocation; four studies stated assessors were blind to treatment. Most studies (11/13 studies) clearly outlined that treatment groups were treated identically other than the intervention of interest. Six of the studies completed follow up or adequately explained loss to follow up. Seven of the studies analysed participants in the groups to which they were randomised. Most studies (12/13) clearly described that outcomes were measured in the same way between treatment groups. Four of the studies delineated how their measurements were performed reliably. Five of the studies accurately described how they had used the correct statistical techniques. Eight of the studies utilised an appropriate trial design and satisfactorily reported any deviations from the trial design.

Summary of uncontrolled trials
The results from the included uncontrolled trials are summarized in

Summary of Uncontrolled Trial Results
An intervention with people living with HIV consisting of 5 sessions of smoking cessation counselling with NRT led to 28% biochemically veri ed abstinence from smoking at 6 months. 24 An intervention with men who have sex with men consisting of 7 weekly sessions facilitated by peers with NRT was found to be signi cantly associated with point prevalence abstinence (p=0.00) with 38.6% reported smoking abstinence over the previous 7 days at 6 months. 29 A study which evaluated a decreasing regiment of NRT patches with people in prison was found to lead to biochemically validated point prevalence abstinence and continuous abstinence rates of 26% and 22% respectively. 34 A partially student run intervention which utilised NRT with people experiencing homelessness led to 13.7% of participants contacting the smoking cessation line. 25 A small scale study, using NRT patches, with people accessing mental health services found that the intervention led to a reduction in the number of daily cigarettes number of cigarettes, dependency levels and average weekly expenditure on tobacco at 3 months. 31 Several studies evaluated interventions related to ENDS devices. An intervention with people experiencing homelessness using information advice, group sessions, NRT and providing a choice of ENDS device found an 18% increase in odds of reporting 7-day, biochemically veri ed, point-prevalence abstinence. 27 All study participants reporting abstinence used an ENDS device and those using an ENDS device were found to have a 44% reduction in smoking intensity and a 1.2-point reduction in dependence (ibid.) Another study, with people experiencing homelessness and using an ENDS device, found biochemically validated sustained abstinence 3/48 (6.25%) at 24 weeks in the treatment group compared to 0/32 (0%) for the usual care group. 28 A small ENDS intervention with people experiencing homelessness found a statistically signi cant decrease in reported cigarettes smoked and decrease in Fagerstrom scores (p < 0.001). 36 However, decreases in biochemically validated abstinence were not statistically signi cant (ibid.). A six week ENDS intervention with people with serious mental illness was found to lead to a (≥50%) reduction in cigarettes consumed per day between baseline and week 6 (p<0.001). 30 Another ENDS based intervention with veterans with mental health issues was found to lead to signi cant reductions in breath CO ( p < 0.02) and self-reported cigarettes smoked ( p < 0.001). 38 A study evaluating a 21mg NRT patch intervention with people using drugs stated that seven subjects (14.3%) self-reported tobacco abstinence at 21 days, and 5 (10.2%) self-reported abstinence as outpatients at 6 weeks. 35 An ENDS-based intervention, with a population recruited from a methadone clinic, was found to lead to a reduction in the number of cigarettes smoked at 3 weeks (P < 0.001), 5 weeks P < .001) and 9 weeks (P <0 .01) 37 .Similarly , a NRT intervention with people in a residential treatment centre reported extremely high in-treatment (85%) and posttreatment (99%) abstinence rates. 39 Mental health and other bene ts are reported by several studies. A study with people experiencing homelessness, using an ENDS device, found that GAD-7, PHQ-9 and AUDIT scores declined from baseline. 28 Another study using an ENDS device with people experiencing homelessness found a decrease in "poor concentration" over the treatment period. 36 An intervention using NRT and weekly peer led sessions with men who have sex with men was found to increase mental health between baseline and follow up (p= 0.00). 29 However, there was no difference between depression and anxiety scores between those who had relapsed and those who did not (p > 0.05). A community-based study with people who use drugs found that an intervention including a NRT component led to a considerable reduction in self-reported illicit substance use (18.8%), including a reduction in the opioids heroin (6.3%), fentanyl (2.6%) and Oxycontin (3.8%). 33 Other reported bene ts included improved health, return to work and greater community engagement (ibid.)

Quality Appraisal of Uncontrolled studies
The overall quality of the uncontrolled trials, using the JBI Appraisal Tool, was found to be low with an average of 2.5/9 with a range from 0-6/9. The cause and effect of the study was clearly articulated in most studies (11/16 studies). Only one study included a control group and there were differences between the treatment and control group. Four studies had completed follow up and/or used intention to treat analysis. Four studies included comparisons but only one clearly outlined that outcomes were measured in the same manner. No study clearly outlined that measurements were performed in a reliable way. Eight of the studies clearly outlined that they had used the appropriate statistical test.

Controlled studies
Study Primary or secondary aim To investigate the e cacy of an intensive tobacco cessation intervention for alcohol-dependent smokers in early recovery.
At 12 and 26 weeks the 7 day point prevalence quit rates were signi cantly higher in the intensive care group (both p = .03).
However, the quit rates were not statistically signi cant at 38 or 52 weeks Veri ed 30-day alcohol abstinence were not statistically signi cant at any of the follow up points.

3/13
Chan et al., 13 To compare effectiveness and acceptability of 3 different forms of NRT Nicotine inhaler group showed statistically signi cant reduction in CO level from baseline at both weeks 4 and 8 (P < 0.001 and P = 0.032, respectively).
The nicotine inhaler and the patch group showed signi cant decrease in nicotine withdrawal symptoms from baseline at both weeks 4 and 8.
The nicotine inhaler and the gum groups showed signi cant decrease in psychiatric symptom severity from baseline at both weeks 4 and 8.

8/13
Christiansen et al. To examine the feasibility, acceptability, and effect of e-CIG on smoking behavior among smokers with opiate use disorder. S1: End of treatment smoking abstinence was 8.6% (intent-to-treat; n=35) and 10.7% for per protocol group (n=28). 28% of those who completed the program were able to achieve a 50% reduction to their baseline cigarette consumption.

Discussion
There is limited evidence that many NRTs (for example, gum, patches and inhalers) and ENDS based interventions have a positive effect on self-reported smoking with the marginalised groups focused upon in this review. Evidence that does exist is often not corroborated with biochemical validation. Clear standards for CO cut off points for biochemical validation seem to be lacking with studies using cut-offs as low as 4ppm CO and as high as 10ppm CO and various different devices. Additionally, the evidence for these interventions is negatively impacted by the poor design of many of the reviewed trials.
Overall, the methodological quality of the studies was very poor with an average score of 5.92/13 in the RCT group and an average score of 2.5/9 in the quasi-experimental group. Few studies were identi ed where outcomes were measured in a reliable manner.
It was di cult to ensure that statistical analysis was appropriate using the JBI checklist, due to a dearth of information given in the majority of studies. None of the controlled trials and only one uncontrolled trial 37 identi ed that they carried out statistical power analysis; therefore the assumption that other studies were underpowered could not be dismissed. The studies did not use power analysis to determine the effect size of different forms of NRT or ENDS or complementary psychological interventions as part of the treatment arms. Studying the reliability of ANDs intervention statistics was therefore a complex task and using more robust methods such as intention to treat in larger and statistically powered trials would improve future studies and thereby enable comparison of outcomes. The importance of the studies included in determining the feasibility and effects of ANDs in minority groups cannot be dismissed.
Preliminary evidence suggests that engagement with smoking related interventions may have a positive effect on mental health 11,13,24,[28][29]36 though this is not reported in all trials 17 . This may not be related to meeting smoking abstinence outcomes 29 although there is consistent evidence that smoking cessation is associated with mental health bene ts. 40 There may also be bene ts in relation to substance use, general health and employment. 33 Although there were several ENDS-based interventions found in the quasi-experimental studies 26-28, 30, 36-37 only one included RCT 18 evaluated this type of intervention. In the context of potential bene ts for smoking reduction 41 and the ongoing debates 42-43 around the potential bene ts and harms of these interventions, further study in properly powered and randomised clinical trials is warranted. Such research should evaluate the bene ts and risks of such interventions in comparison to available smoking reduction options for speci c groups.
This review is signi cantly limited due to heterogeneity regarding study type, intervention type and populations. This prevents the possibility of generalising any results. Although several quasiexperimental studies investigated interventions with people experiencing homelessness, no randomised controlled trials were identi ed.
Despite the fact that smoking remains extremely high in many marginalised groups, 1,2 there is a relative absence of properly controlled trials with these groups. In this context it is important to note that this review included no studies of members of the Traveller, sex worker or migrant communities. This highlights a signi cant gap in the research literature that should be addressed in the future.

Conclusion
Future studies should aim to adopt best practice standards in relation to CO measurement as the large variance in cut off points reduces the possibility of comparing study outcomes. Several studies suggest potential mental health bene ts arising from participation in smoking reduction-related interventions; such bene ts should be explored in properly controlled trials. There is a current lack of RCTs related to smoking related interventions with people experiencing homelessness. Similarly, there is a lack of properly controlled RCTs exploring the use of ENDS with marginalised populations. Future studies should aim to address these de cits.

Declarations
Ethics approval and consent to participate N/A Consent for publication N/A Availability of data and materials Quality Appraisal and Statistical appraisal les are available.

Competing interests
One of the reviewed studies was co-authored by FS, JW, KM and AM. No con ict of interest from AJ Funding Figure 1 PRISMA Flow Chart