Management and control of tuberculosis control in socially complex groups: a research programme including three RCTs

Article history

The research reported in this issue of the journal was funded by PGfAR as project number RP-PG-0407-10340. The contractual start date was in September 2008. The final report began editorial review in September 2018 and was accepted for publication in January 2020. As the funder, the PGfAR programme agreed the research questions and study designs in advance with the investigators. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The PGfAR editors and production house have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the final report document. However, they do not accept liability for damages or losses arising from material published in this report.

Copyright statement

© Queen’s Printer and Controller of HMSO 2020. This work was produced by Story et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

2020 Queen’s Printer and Controller of HMSO

Evolution of the programme components

In view of these challenges, we designed a programme of research that was submitted to NIHR for consideration of funding. This included the following studies:

• study 1 – cross-sectional survey on the prevalence of latent tuberculosis infection, human immunodeficiency virus (HIV), hepatitis B and hepatitis C in prisoners, drug users and people experiencing homelessness to inform a mathematical model (study 7) of the impact and cost-effectiveness of screening socially complex groups

• study 2 – observational study of the effectiveness of a new prison radiographic screening service

• study 3 – cluster randomised controlled trial (CRCT) evaluating interventions to increase the uptake of mobile radiographic screening in homeless populations

• study 4 – establishment and evaluation of a rapid diagnostic pathway to reduce loss to follow-up prior to diagnosis

• study 5 – individually randomised controlled trial of clinic-based DOT versus community-based DOT to inform optimal treatment delivery for socially complex groups

• study 6 – evaluation of a specialist clinical service providing in-reach high-quality tuberculosis care in prison

• study 7 – qualitative research to ensure that proposed interventions are acceptable and accessible to the target population and to inform understanding of barriers to services

• study 8 – development of a dynamic transmission model to predict public health impact of the interventions and inform economic analysis comparing costs with the savings made through averting future cases.

Following peer review, the funders requested the following changes to this programme of work:

• Removal of studies 6 and 7 in response to the programme being perceived as overambitious and the team having insufficient qualitative research expertise.

• Simplification of study 3, which was originally conceived as a three-arm CRCT comparing usual practice with financial incentives and with peer educators as interventions to improve mobile radiographic screening uptake. We were requested to remove one study arm and, therefore, simplified this to a proposed trial of incentives versus usual practice.

• Addressing concerns about the safety of conducting smear microscopy on the mobile radiographic unit. To address this issue, we proposed evaluating how the pan-London service would use a single NHS trust for rapid sputum smear microscopy.

• Cost-effectiveness analyses – it was requested that the cost-effectiveness of the adherence interventions also included a societal perspective.

Subsequent to funding, a number of developments led to further modifications of the study proposal, which were agreed with funders at regular reviews. These are summarised below.

Cost-effectiveness analyses

During the course of the programme grant, a cost-effectiveness analysis of screening for latent tuberculosis infection and active tuberculosis in prisoners was conducted by the National Institute for Health and Care Excellence (NICE). In view of this new analysis, this aspect of the economic evaluation was removed. As the peer intervention to improve screening uptake showed no evidence of effectiveness, cost-effectiveness analyses were considered unnecessary. Similarly, as we were unable to demonstrate that the use of a rapid diagnostic approach improved outcomes, a cost-effectiveness analysis was deemed unnecessary. Finally, as the VOT intervention proved substantially more effective and cheaper than DOT, no cost-effectiveness analysis was required. This work package, therefore, focused on the cost-effectiveness of screening for latent tuberculosis infection alongside mobile radiographic screening in the homeless, and cost comparison of DOT and VOT.

Following these changes, we divided the work into six work packages as outlined in Figure 1.

FIGURE 1.

Overview of work packages. a, Testing of prisoners for blood-borne viruses and latent tuberculosis infection was originally included in WP1 but for clarity and improved contextualisation of the results we have included the description of this activity in WP2. WP, work package.

Study data handling

Study nurses with relevant honorary NHS contracts had access to a look-up table between patient identifiers and study numbers to enable them to make referrals and follow up patients who had consented to this. This look-up table was held on University College London’s data safe haven. Anonymised data sets were used by other staff for analyses and were kept on secure servers with access to named individuals only.

Introduction

Individuals experiencing homelessness have high rates of tuberculosis infection and often present late to health-care services. 3 Latent tuberculosis infection is common in homeless populations in low-burden countries,4,5 although limited data are available in the UK.

Homelessness and tuberculosis in homeless populations are both increasingly important problems in London. In 2014, it was estimated that 3.6% (89/2498) of tuberculosis cases with social risk factor information available had a history of homelessness. 6 The point prevalence of active tuberculosis in people experiencing homelessness in London has been estimated at 788 per 100,000, compared with 27 per 100,000 in the overall London population. 3

Developments in testing for latent tuberculosis infection and treatment of blood-borne viruses provide new opportunities to improve the health of people experiencing homelessness. 7,8 Poor adherence and potentially severe hepatotoxicity exacerbated by high rates of alcohol- or viral-related liver disease remain concerns for latent tuberculosis infection treatment in people experiencing homelessness. 9 Health services may not be well tailored to deliver care for identified infections in people experiencing homelessness, potentially limiting the value of screening.

We undertook a cross-sectional survey to estimate latent tuberculosis infection and blood-borne virus prevalence among individuals in homeless hostels in London and examined outcomes of referral to health-care services after 12 months.

Methods

Results

Study population

We invited 804 individuals to participate following Find&Treat mobile radiographic screening. A total of 542 out of 804 (67.4%) individuals consented to take part (Figure 2). We subsequently excluded 51 (9.4%) individuals, mainly because of a lack of venous access for blood sampling (n = 31); therefore, we included 491 individuals in the analysis.

FIGURE 2.

Recruitment flow chart for work package 1. a, It was operationally extremely intensive to collect data on the number of individuals who were eligible, approached and refused screening, therefore these data were collected only at the start of the study for the first 474 eligible patients. The numbers presented are estimates based on this initial sample.

Most participants were men (437/491, 89.0%) aged 30–49 years (257/491, 52.3%) who were born in the UK (305/491, 62.1%), were current tobacco smokers (394/491, 80.2%) and reported being homeless for ≥ 1 year (443/491, 90.2%). Of those not born in the UK, 45 were from other Western European countries, 18 were from Central Europe, Eastern Europe or Central Asia, 23 were from the Eastern Mediterranean, Middle East, North, East or West Africa, 41 were from Central or sub-Saharan Africa, 25 were from South or South East Asia or the Indian subcontinent, five were from East Asia, 18 were from Latin America or the Caribbean and 11 were from other regions. Just over half (263/481, 54.7%) had spent time in prison. Drug use was common, with 107 out of 491 (21.8%) having ever smoked heroin or crack cocaine and 86 out of 491 (17.5%) having ever injected either crack cocaine or heroin. A high proportion (202/477, 42.3%) had ever been concerned about their drinking, or had had a health worker express concern about this. Latent tuberculosis infection and blood-borne virus results are shown in Tables 1 and 2.

TABLE 2 - Baseline demographic and clinical characteristics for participants stratified by test results for latent tuberculosis infection, hepatitis B and hepatitis C

Sum of current and past hepatitis B.

Sum of current and past hepatitis C.

HIV data not included to reduce the risk of deductive disclosure.

Reproduced from Aldridge et al. 2 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/. The table includes minor additions and formatting changes to the original text.

Characteristic	All	QuantiFERON positive		Hepatitis B positivea		Hepatitis C positiveb
	n	n	%	n	%	n	%
All	491	81	16.5	58	11.9	64	13.0
Age (years)
18–29	69	8	11.6	6	8.7	3	4.3
30–49	257	39	15.2	28	10.9	43	16.7
≥ 50	165	34	20.6	24	14.5	18	10.9
Sex
Female	54	4	7.4	5	9.3	3	5.6
Male	437	77	17.6	53	12.1	61	14.0
Born in the UK
Yes	305	29	9.5	29	9.5	50	16.4
No	186	52	28.0	29	15.6	14	7.5
Total time spent homeless
< 1 year	48	8	16.7	6	12.5	4	8.3
1 year	135	18	13.3	16	11.9	13	9.6
2–3 years	141	28	19.9	19	13.5	11	7.8
> 3 years	167	27	16.2	17	10.2	36	21.6
Has ever spent time in prison
No	218	35	16.1	27	12.4	12	5.5
Yes	263	45	17.1	30	11.4	50	19.0
Missing	10	1		1		2
Illicit drug use
Neither smoked nor injected	298	44	14.8	27	9.1	13	4.4
Has ever smoked heroin/crack cocaine	107	20	18.7	14	13.1	5	4.7
Has ever injected drugs	86	17	19.8	17	19.8	46	53.5
Currently smokes cigarettes
No	97	18	18.6	10	10.3	2	2.1
Yes	394	63	16.0	48	12.2	62	15.7
Participant or health worker ever been concerned about drinking
No	275	51	18.5	30	10.9	24	8.7
Yes	202	28	13.9	25	12.4	36	17.8
Missing	14	2		3		4

Blood-borne viruses

Seven out of 489 patients (1.4%, 95% CI 0.4% to 2.5%) had current hepatitis B infection as confirmed by HBsAg neutralisation. In addition, 10.4% (51/489) (95% CI 7.7% to 13.1%) of patients had evidence of past hepatitis B infection. A total of 59.5% (291/489) (95% CI 55.1% to 63.9%) of patients were non-immune to hepatitis B; this was lower for those who had ever injected drugs (27.1%, 23/85) (95% CI 17.4% to 36.7%; Figure 3). Most non-immune individuals (77.7%, 226/291) did not recall whether or not they had been previously vaccinated for hepatitis B, and 29 out of 291 (10.0%) patients reported having never received vaccination. A total of 41.2% of these non-immune individuals (120/291) had spent time in a UK prison. Only four non-immune individuals reported being vaccinated against hepatitis B more than once.

FIGURE 3.

Work package 2 flow diagram. BBV, blood-borne virus; CXR, chest X-ray; LTBI, latent tuberculosis infection.

Among a total of 64 out of 491 (13.0%, 95% CI 10.0% to 16.0%) participants with anti-HCV seropositivity, 51 (10.4%, 95% CI 7.8% to 13.1%) tested positive for hepatitis C RNA, indicating current infection. The remaining 13 participants (2.7%, 95% CI 1.2% to 4.1%) showed confirmed anti-HCV reactivity in the absence of hepatitis C RNA, indicating a resolved infection. Individuals who had ever injected drugs had a much higher prevalence of hepatitis C (46/86, 53.4%, 95% CI 42.4% to 64.3%) than those who reported no injecting history (12/405, 3.0%, 95% CI 1.3% to 4.6%). In those diagnosed with latent tuberculosis infection, the frequency of co-infection with either hepatitis B or hepatitis C (past or current) was 37.0% (95% CI 26.3% to 47.8%), and co-infection with both hepatitis B and hepatitis C (past or current) was 16.2% (95% CI 9.7% to 24.7%).

The prevalence of HIV seropositivity was 1.02% (95% CI 0.1% to 1.9%); all cases were due to HIV-1 and all participants were previously aware of their diagnosis.

Discussion

This study demonstrates a high burden of latent tuberculosis infection and blood-borne virus infections in a London homeless population at levels that are substantially higher than in the general population. For example, the prevalence of latent tuberculosis infection in inflammatory bowel disease patients screened for latent tuberculosis infection before the initiation of anti-tumour necrosis factor alpha (anti-TNFα) therapy in the UK is 1.6% (95% CI 0.2% to 5.7%)16 and the prevalence of hepatitis C infection in the general population is 0.4%. 17

Although the highest latent tuberculosis infection prevalence was in foreign-born participants, around 10% of UK-born patients experiencing homelessness were infected. Prison history increased the risk threefold in UK-born participants. During the study, referral rates for treatment for latent tuberculosis infection were low because of the criteria in operation at the time. Under the most recent (2016) NICE guidelines,11 all those with a positive test aged up to 65 years would be eligible for referral for treatment; therefore, instead of three people (4%; 3/81) being referred, 76 (93.8%; 76/81) would now be eligible.

Significantly higher levels of current and past hepatitis B infection were seen in this study than in the general population (1.4% and 10.4%, respectively). Hepatitis B vaccination was higher in those reporting injecting drug use, possibly as a result of targeted vaccination in this population, but there remained a substantial proportion of this homeless population who were non-immune and who would benefit from vaccination. No patients were initiated on treatment; however, this is not necessarily unexpected given the prolonged clinical assessment (typically two or three appointments spaced out by a few months) that was required before treatment initiation for hepatitis B.

Hepatitis C prevalence was high (13%). This was mainly among those who reported injecting drug use, but even those without such a history had higher levels than the general population. Engagement with health services was poor in those diagnosed with current hepatitis C infection, with just over half of those referred either not attending appointments or being lost to follow-up. In only a minority of those referred was antiviral therapy initiated within 12 months. Until recently, hepatitis C care in general has been characterised by a small number of treatment initiations relative to the number of people needing and accessing care. 17 With the introduction of interferon-free regimens of short duration (typically 12 weeks), there is a new emphasis on increasing treatment coverage, but the impact on this vulnerable population has not yet been formally investigated. In individuals diagnosed with latent tuberculosis infection, co-infection with either hepatitis B or hepatitis C (past or current) was high, at 37.0%, as was co-infection with both hepatitis B and hepatitis C, at 16.2%. The implications of this for latent tuberculosis infection treatment and risk of hepatotoxicity need to be carefully considered.

We used convenience sampling as it was not possible to use a formal sampling framework in a study conducted alongside a very-high-throughput NHS clinical service dealing with complex groups. The requirement for individuals to be able to consent meant that our results do not include individuals who were intoxicated (by drugs or alcohol) and, therefore, are likely to under-represent those at highest risk of blood-borne virus infection. Despite this, the homeless population in this study included a high proportion of previous rough sleepers, people with either current or previous high-risk drug use, and people with harmful and hazardous alcohol use. The population sampled is broadly demographically comparable to homeless populations nationally according to the Find&Treat data collected from extensive screening outside London, and Homeless Link’s (London, UK) health needs audit. 18

To our knowledge, this study provides the first measure of latent tuberculosis infection and blood-borne virus prevalence in people experiencing homelessness in the UK. Previous studies in other high-income countries (including Italy, Japan, South Korea and the USA) have reported latent tuberculosis infection prevalence in homeless populations and found rates varying from 16% to 75.9%. 5,19–22

A meta-analysis of active tuberculosis and blood-borne virus prevalence in homeless populations internationally found that hepatitis C virus (HCV) infection ranged from 3.9% to 36.2%, and HIV infection ranged from 0.3% to 21.1%. 23 None of the studies testing for HIV was conducted in the UK, but one hepatitis C study, which recruited individuals experiencing homelessness from shelters, special projects and medical centres in Oxford, found 26.5% of individuals to be HCV infection positive using oral fluid testing. 24

Low levels of hepatitis B immunity indicate inadequate access to vaccination in this high-risk group. This finding is consistent with our previous work demonstrating the inverse care law with respect to influenza vaccination. 25 Our data demonstrated that the eligibility of people experiencing homelessness for influenza vaccination as a result of clinical risk factors was 38.9%, compared with 13.0% in the general population, but only 23.7% of those eligible were vaccinated compared with national levels of 53.2%. Given the unmet need for hepatitis B vaccination, there is a strong rationale for offering universal provision of hepatitis B vaccination to people experiencing homelessness through existing services engaged with this group. 26 The work also indicates the need to strengthen the prison hepatitis B vaccination programme and vaccination alongside drug treatment services. Individuals who tested HBsAg positive generally maintained links with services after referral, whereas those diagnosed with hepatitis C infection showed suboptimal retention by care services. Further studies are required to determine whether or not expanded availability of interferon-free regimens of short duration will increase engagement in this population.

Since 2010, the number of people seen rough sleeping has doubled nationally. 27,28 People experiencing homelessness represent the extreme end of health inequalities in high-income countries and they experience a high burden of preventable morbidity and mortality from infectious and non-infectious disease. 29,30 We demonstrated the high prevalence of undiagnosed latent tuberculosis infection, hepatitis B infection and hepatitis C infection in homeless populations in the UK, and a large unmet need for hepatitis B vaccination. We also demonstrated the need for intensive case management and ongoing support to ensure that testing can translate into treatment opportunities. High rates of co-infection highlight the importance of service integration through combined testing and treatment pathways. 31 NICE now recommends that persons accessing targeted mobile radiology should be offered tests for blood-borne viruses11 and our data provide the basis to estimate the cost-effectiveness of this approach. The recent national collaborative tuberculosis strategy32 commits to new investment in a national outreach service in line with the proven Find&Treat outreach model. 10

Deviations from the protocol

The protocol for work package 1 and other studies as submitted to NIHR are provided in Report Supplementary Material 1. The key deviation from the protocol was in the numbers of patients screened for latent tuberculosis infection and blood-borne viruses. The original target had been 3000 (1500 in prison and 1500 in homeless hostels). This target number had been based on power calculations that would enable the progression rates from latent to active tuberculosis to be measured with accuracy through data linkage to national surveillance data. We were able to recruit only one-third of this target (491 individuals in homeless hostels and 515 individuals in prison). In the homeless populations, we conducted the screening alongside the mobile radiographic unit, which has a high turnover. It was not possible to approach, consent and take blood samples from a high proportion of eligible participants under this arrangement. As completion of radiographic screening (to help rule out active disease in those with evidence of latent tuberculosis infection) was an eligibility criterion, this limited the recruitment. In addition, it often proved difficult to identify private spaces in which to take blood. Similarly, in the prison we were restricted to screening those who had undergone radiographic screening. This greatly limited numbers as the prison radiographic screening that was due to be delivered by the NHS had major implementation problems, as described in work package 2. As the number of recruits was lower than anticipated, the recruitment number was renegotiated with NIHR as it was agreed that the lower number would still enable accurate measurement of latent tuberculosis infection and blood-borne viruses; however, the second objective of measuring rates of progression through data linkage was abandoned because of insufficient numbers.

Successes/added value

Findings from this study contributed to the NICE public health programme guidelines Identifying and Managing Tuberculosis Among Hard-to-Reach Groups. 33 Citing high rates of infection has, in part, influenced the recommendation to screen people experiencing homelessness for tuberculosis and to refer latent tuberculosis infection cases for further clinical investigation if the individual is aged < 35 years. Our research showing high levels of latent tuberculosis infection in those screened on the mobile radiographic ray unit has led to latent tuberculosis infection screening being commissioned on the mobile radiographic unit to migrants experiencing homelessness across London who are from high-incidence countries and to the commissioning of hepatitis B vaccination on the mobile radiographic unit.

Further work, alongside this study, identified a high unmet need for influenza vaccination among people experiencing homelessness. 25 The mobile radiographic unit has now been commissioned to provide influenza vaccination. Work carried out alongside this study also showed the high prevalence of chronic disease including cardiovascular disease in people experiencing homelessness screened by the mobile radiographic unit. 29,34 This has recently provided support for a successful application for a NIHR programme development grant on Cardiovascular Disease in People Experiencing Homelessness (Chief Investigator – Ami Banerjee). The work also led to a Department of Health and Social Care policy research programme, the HALT Hepatitis Study, a randomised control trial of the use of peers to support full diagnosis and treatment completion for hepatitis B and hepatitis C among socially complex groups (URL: www.isrctn.com/ISRCTN24707359; accessed 25 September 2020). The charity Groundswell (London, UK) is now commissioned in parts of London to support patients experiencing homelessness to attend HCV appointments.

The study also contributed to the successful procurement of a European Union-funded grant on hepatitis C screening in Europe, which is funding hepatitis C screening and portable liver FibroScan^® (Echosens, Paris, France) screening alongside the mobile radiographic unit to assess disease severity. 35

Although the objective of data linkage was dropped from this study, the work was used to fund the development of probabilistic data linkage methodologies at PHE. The algorithms developed have been validated by members of our team. 36 They have also been used to link data from the mobile radiographic unit to national tuberculosis surveillance in order to confirm which patients who were screened went on to develop active tuberculosis. This has enabled members of our group to measure the sensitivity and specificity of mobile radiographic screening for tuberculosis26 and to evaluate the accuracy of computer-aided detection of tuberculosis from radiographic images. 37 The Find&Treat team is currently purchasing a licence to allow the use of computer-aided diagnostics technology on its radiographic database. The linkage has also been used to link data from the pre-entry migrant screening programme to national tuberculosis surveillance data, enabling the measurement of the incidence of tuberculosis post migration. This was published in The Lancet38 and the lead author (RW Aldridge) based a successful Wellcome Trust (London, UK) postdoctoral fellowship application on this approach.

Background/rationale

Prison populations are at an increased risk of infectious diseases including tuberculosis and blood-borne viruses. 40 In London, a large continued outbreak of drug-resistant tuberculosis has had many cases that are linked to imprisonment and substance misuse,41 and research using a mobile radiographic unit for intermittent screening in prisons found a rate of tuberculosis of > 208.4 per 100,000 prisoners. 3 Prisons have disproportionally large numbers of people who are infected with HIV,42 involved in substance misuse,42–44 homeless,45 and with known clinical and social risk factors for infection and progression to active disease. Prisoners with tuberculosis are also more likely to have infectious pulmonary disease, drug resistance and poor treatment outcomes. 46 The prevalence of HIV and hepatitis C among prisoners is also higher than in the general population, and prisoners are among the most vulnerable population to contract hepatitis B virus (HBV) in the UK. 47 In the UK, unlinked anonymous blood-borne virus monitoring showed that 10.1% of those in specialist drug services and 9.4% of those in prisons tested positive for hepatitis C, compared with 2.2% overall in primary care in 2013. 48 Of all hepatitis C diagnoses, 90% are associated with injecting drug use, and 40% of those who are injecting drugs are estimated to be living with hepatitis C infection. 49 The prevalence of HIV in those injecting drugs remains low, at 1.1%, and HBV prevalence has been continuously decreasing (now 17%). This has been attributed to the HBV vaccination programme in specialist drug services and in prisons, where 75% of individuals report having received HBV vaccination.

The National Institute for Health and Care Excellence recommends that prisons and immigration removal centres should have a tuberculosis liaison lead and a tuberculosis policy that is developed in conjunction with tuberculosis services and Public Health England. In prisons with Department of Health and Social Care-funded static digital radiographic facilities for tuberculosis screening, all new prisoners, detainees and transfers should have a chest radiograph within 48 hours of arrival unless they have had a chest radiograph in the last 6 months. 33

In addition, NICE advises that prisons in high-incidence areas, or which receive prisoners from high-incidence areas, offer latent tuberculosis screening [interferon gamma release assay (using IGRA)] to prisoners who are aged < 35 years and in regular contact with support services, and where continued support is provided on release, including directly observed preventative treatment (DOPT). 33

This part of the guidance has not been widely implemented. Little is known about the prevalence of latent tuberculosis infection in prisons. No such studies have been conducted in the UK, although a Cochrane Review found that the average incidence rate of latent tuberculosis infection in prisons in Brazil and the USA was 26.4 times that of the general population. 40 The co-infection rate with latent tuberculosis infection and blood-borne viruses also needs to be established because of the potentially increased risk of hepatotoxicity in the treatment of latent tuberculosis infection with pre-existing liver disease. Furthermore, DOT is currently provided to, on average, only 60% of eligible patients (in London) with active tuberculosis,50 and the impact on health-care services of adding DOPT for prisoners who are released on preventative treatment needs to be assessed.

This study aimed to evaluate the effectiveness of a new prison radiographic screening programme using a teleradiology network of static digital radiographic units to reduce the risk from tuberculosis in prisons, and to determine the prevalence of latent tuberculosis infection and co-infection with blood-borne viruses as well as outcomes of referrals for onward care.

Methods

The primary objectives were to evaluate the effectiveness of a new prison radiographic screening programme using a teleradiology network of static digital radiographic units to reduce the risk from tuberculosis to those in prisons, and to determine the prevalence of latent tuberculosis, HIV, hepatitis B and hepatitis C infection among prisoners in London, with clinical outcomes recorded for participants referred to health-care services. Secondary objectives were to determine interobserver reliability for the interpretation of chest radiographs, and the prevalence of co-infection with latent tuberculosis, HIV, and hepatitis B and hepatitis C in prisoners in London and how this varied according to demographics and lifestyle risk factors such as alcohol and drug use.

Results

A total of 1491 chest radiographs were taken for 1484 individuals. There were 3032 new arrivals during the study, of whom 618 (20%) arrived in the substance misuse wing. The overall screening coverage of new prisoners was 43% (1302/3032). The vast majority of radiographs that were taken were from new arrivals (87%), 1242 (84%) were from prisoners on the admission wing, 158 (11%) were from the substance misuse wing and 91 (6%) were from other wings. The total number of active screening days was 112, with an average of 13 radiographs per day.

From the initial reporting by the NHS trust radiology department, 141 out of 1491 (9%) radiographs showed some kind of abnormality (see Figure 3). Ninety-seven (69%) of these were acted on by the prison health-care team, including 29 prisoners (29/141, 20%) who had further investigations for tuberculosis. Of the 29 suspected tuberculosis cases, 15 (52%) had one or more sputum sent off for acid-fast bacillus smear and culture and sensitivity. Of these, two had a positive smear result for acid-fast bacillus; none was culture positive for Mycobacterium tuberculosis but one prisoner started tuberculosis treatment and completed a full course of antibiotics under DOT in the prison (see Figure 3). This translates into a tuberculosis rate of 67 per 100,000 prisoners. Of the other 14 cases, one was referred to the tuberculosis clinic for prophylactic treatment, three were reviewed and discharged by the prison GP, three had a computerised tomography scan (from which two cases did not suggest tuberculosis and were discharged and one case was referred to the chronic obstructive pulmonary disease nurse), three had a history of tuberculosis or atypical tuberculosis and required no action, two were lost to follow-up and two had no interventions initiated.

From the secondary reporting, 56 chest radiographs (49 individuals) were reported as suggestive of active tuberculosis. When compared with the initial reporting, 14 of the 56 radiographs (or 49 individuals) had been identified as suggestive of tuberculosis in both reports and 42 radiographs (or 35 individuals) had not been identified as suggestive of tuberculosis in the initial reporting (Table 5). The NHS trust radiology reporters classified 1.9% of chest radiographs as potentially indicating active tuberculosis, whereas the Find&Treat reporter classified 3.8% as active tuberculosis. The inter-rater reliability for suspicion of active tuberculosis was fair (κ = 0.312, 95% CI 0.181 to 0.442).

A total of 15 of the 35 individuals who had not initially been identified as having concerning radiographs had also consented to and participated in the latent tuberculosis infection screening and were referred to local chest clinics. Of these individuals, six were discharged following a review or repeat chest radiography showing no changes, seven had an unknown outcome and two were lost to follow-up. For the 15 participants who were to be contacted by the prison health-care team, no outcome data were available.

Of the 1484 prisoners screened with a chest radiograph, 595 prisoners were approached by the research team and 88% of them consented to participate in the study. The final analysis included 513 individuals (see Figure 3 for full details).

All participants were male, and the majority were born in the UK (336/513, 66%) and aged 16–39 years (404/513, 79%) (Table 6). Just over half (274/506, 54%) of participants had a history of homelessness, slept rough, sofa surfed, squatted or stayed in a homeless hostel, and 16% (81/513) had slept at a place where people used illicit drugs. The vast majority (408/513, 80%) were current smokers and around one-third (170/513, 31%) had smoked crack cocaine or heroin [with crack cocaine more commonly used (160/513, 31%) than heroin (99/513, 19%)]. Injecting drug use was less frequent (40/513, 8%), and heroin use (39/513, 8%) was higher than crack cocaine use (26/513, 5%). Four per cent of participants (21/513) reported having shared needles.

The estimated overall prevalence of latent tuberculosis infection was 13% (65/513); among this group 5% (3/65) of participants were co-infected with HCV and 6% (4/65) were co-infected with HBV. Univariable analysis demonstrated higher rates of latent tuberculosis infection in non-UK-born participants than in UK-born participants (21.5% vs. 6.3%), and in those aged 30–39 years than in those aged 16–29 years (13.5% vs. 8%).

Thirty-seven participants with a positive IGRA result were eligible for preventative treatment and referred to the local hospital chest clinic. A total of 40.5% (15/37) of participants started preventative treatment, of whom 60% (9/15) completed their treatment course and 40% (6/15) had incomplete treatment. The remaining 22/37 (59.5%) were lost to follow-up or discharged as they did not attend appointments. Prevalence of current hepatitis C was 4% (22 individuals), which in 45% of cases (10/22) was a new infection. None of these individuals was referred to a hepatitis clinic, half (5/10) were seen by a prison or community GP and the other half were lost to follow-up. Of the 12 participants with known hepatitis C infection, the majority (58%) were lost to follow-up or discharged because of non-attendance, and 5 out of 12 (42%) were under medical review by their GP or the hepatitis clinic. None of the 22 participants started hepatitis C treatment.

Prevalence of current hepatitis B was low (2%), and 70% (7/10) of the individuals with current hepatitis B had new or previously unknown infections, of whom 43% were seen by prison GPs or community GPs and 57% were lost to follow-up or discharged because of non-attendance. Of the three individuals with known HBV infection, one was referred to a hepatitis clinic, one was deported before referral and one had no interventions. Sixty-five per cent of all participants had insufficient or no immunity to HBV, and none had HIV infection.

Discussion

Our study of the tuberculosis radiographic screening programme found a tuberculosis rate of 67 per 100,000 prisoners but, as this is based on a single case, CIs are wide. In combination with the high prevalence of latent tuberculosis, and compared with the London rate of 35.5 per 100,000 Londoners,57 this indicates that prisoners remain at high risk of tuberculosis in the UK.

The lack of classification system in the initial reporting meant that radiographs were reported in a variety of ways by the different radiologists from the NHS trust. As a consequence of this, we changed a secondary objective from looking at the outcome of radiographs reported as suggestive of tuberculosis to looking at the outcome of radiographs where investigations for tuberculosis were instigated by the prison.

Prisoners were out of their cells for 3.5 hours per day and screening competed with paid prison work, attending courses that could lead to transfer to a lower-category prison or earlier release on probation, exercise and social visits. Access outside these times was very limited. During lockdown (security alert or staff meetings), no movement among prisoners was allowed, which had a further impact on the screening uptake; this, and restricted operational capacity within the research team, meant that we were unable to approach all eligible participants in the studies. Owing to the study setting, a sampling framework was not possible for recruiting patients, and we used convenience sampling.

We consider that the sample size obtained in both studies was a strength in our research. The prison population is often considered challenging to engage with, and our research nurses had extensive experience of working with prisoners as well as substance misusers. We found that both chest radiography and latent tuberculosis infection/blood-borne virus screening were highly acceptable among prisoners. We were also able to collect individual biological data, including information on self-reported clinical and social risk factors for tuberculosis, on all participants in the latent tuberculosis infection screening. These questionnaires were developed with prisoners, piloted and altered following user feedback at the start of the study. A limitation of the research was that the questionnaires were not validated for risk factors for prisoners. We used robust data collection methods, and obtained a high level of clinical outcome data.

The levels of hepatitis C and hepatitis B were lower than expected. This is probably explained by the small number of substance misusers (those with the highest risk of tuberculosis) who participated in the study as a result of the remote location of the chest radiography equipment, which was a prerequisite for the latent tuberculosis infection/blood-borne virus screening.

Many other studies have reported rates of tuberculosis in prisons. In London, sporadic radiographic screening in prisons found a rate of 208 per 100,000 prisoners. A survey from the WHO European region58 found a median notification rate of 232 (range 0–17,808) per 100,000 prisoners. Although the type and timing of screening varied, most countries performed radiographic screening at prison entry. 58 The cost-effectiveness of chest radiographic screening for this high-risk population has also been demonstrated. 59–61

We are not aware of any published data on tuberculosis and blood-borne virus co-infection among UK prisoners. A study of tuberculosis and blood-borne viruses in a male prison in Pakistan demonstrated higher prevalence of HIV, HBV and hepatitis C (2%, 5.9% and 15.2%, respectively), but also a higher prevalence of illicit drug use and slightly more injecting drug use. 62 Other prevalence studies on sexually transmitted infections and blood-borne viruses found rates of hepatitis C between 4.9% and 29.7%, HIV of 0% and 6.6% and HBV of 2.4% and 25.2%. 63–65

Other studies on latent tuberculosis infection prevalence in prisons have been carried out in various countries, although, to our knowledge, not in a UK setting. Latent tuberculosis infection rates vary considerably: 17% in the USA,15 40.3–53.3% in Spain,66 48% in Pakistan,67 49% in Brazil and 87.6–88.8% in Malaysia. 68,69 Comparisons with our study results are challenging because of the heterogeneity between the studies including the different tests used for latent tuberculosis infection, varying background rates of tuberculosis, prevalence of HIV infection, eligibility criteria and levels of uptake. Chan et al. 70 used QuantiFERON IGRA to assess the lifelong effect of bacillus Calmette–Guérin (BCG) vaccination and found a latent tuberculosis infection prevalence of 25% in Taiwan.

We found a continued high rate of active tuberculosis (albeit with wide CIs). Although lower than previous London estimates, this still supports the use of systematic screening for active tuberculosis among prisoners, although how this should be arranged needs further development to maximise uptake including in the most vulnerable groups, such as substance misusers. Funding for training and allocation of prison staff to carry out regular radiographic screening need to be provided, and the timing of radiography considered. Screening prisoners immediately on arrival may maximise uptake before prisoners are allocated to different wings. Another option would be to offer chest radiographic screening as an opt-out part of the ‘next-day’ health screening offered to all new prisoners. For future large-scale tuberculosis chest radiographic screenings in a prison setting, we would recommend using an established classification for the reporting of chest radiographs, such as the American Thoracic Society or the Public Health England pre-departure tuberculosis screening, to enable follow-up, analysis and comparison with other studies.

Our study also demonstrated previously unknown high levels of latent tuberculosis infection and blood-borne virus, as well as low levels of HBV vaccination despite a target vaccination programme. This indicates a high level of unmet need and highlights the inequalities in health seen within this population. The prison setting with DOT facilities provides an excellent opportunity to treat individuals with latent tuberculosis who have social risk factors, such as homelessness and substance misuse, that frequently prevent them from being offered latent tuberculosis infection treatment in the community. However, our study demonstrates extremely poor outcomes across all three infections for those referred to NHS health services following a positive result. This may partly be explained by the nature of prison services: a high and fast turnover of prisoners with limited notice, complicated follow-up and onward referral to local NHS services, and incomplete contact details in health-care records, which meant that some prisoners did not receive their results and missed out on onward care upon release. New and better performance indicators are required to improve patient pathways through the prison; however, improved co-ordination of services and follow-up support in the community is also essential to ensure continued care and treatment completion upon release. Alternative models using multidisciplinary teams and appointment sites acceptable to patients have been trialled for hepatitis C71–73 and have improved outcomes for tuberculosis in London, which has been recognised by NICE. 10,26 Further research and health and social care service development are required to improve health outcomes and address the wider determinants of health in this population.

Deviations from the protocol

The planned work had been based on the assumption that the prison would start a routine radiographic screening service; as this did not happen, we provided radiographic resources to the prison to undertake screening. This reduced the numbers going through screening, preventing full evaluation. Owing to the small numbers, the planned linkage to surveillance data was abandoned as only one active tuberculosis case was identified. Participants were recruited into the study of latent tuberculosis infection and blood-borne viruses only if they had undergone radiographic screening, which also limited recruitment.

Introduction

Tuberculosis rates in London are among the highest in Europe, and account for nearly 40% of UK cases. 75,76 Rates are highest in so-called ‘hard-to-reach’ groups (people experiencing homelessness, prisoners and drug users),3 many of whom have infectious tuberculosis, delays to diagnosis, poor adherence to treatment and high levels of loss to follow-up before treatment completion. 3 Congregate living and reluctance to engage with services complicate attempts to identify active and latent tuberculosis in these populations. 77

The pan-London Find&Treat tuberculosis service includes a mobile digital radiographic unit and has been shown to be cost-effective among hard-to-reach groups (people experiencing homelessness, substance misusers and prisoners). 10 Mobile radiographic unit screening can detect tuberculosis early, often before people become infectious. 26,78 Historical data (Dr Alistair Story, Find&Treat, 2019, personal communication) show low uptake rates in eligible individuals resident in homeless hostels (around 50%).

Increased screening uptake has the potential to increase cost-effectiveness, reduce the transmission of tuberculosis and improve the health of this vulnerable population. Peer educators (with lived experience of homelessness, problem drug or alcohol use and tuberculosis) may be able to engage more effectively with those experiencing these issues than health service staff. Peer educators have been shown to improve knowledge about health conditions and increase the use of health services in several areas including HIV, smoking and condom use. 79,80 Previous qualitative research has found that volunteer peers working alongside tuberculosis services derive value from their involvement with the service and the opportunities this provides for meaningful, structured activity. 81

Using a CRCT, we aimed to compare current practice in central London (where Find&Treat and hostel staff encourage people experiencing homelessness to be screened for tuberculosis) with the use of peer educators who have direct experience of tuberculosis and homelessness.

Methods

Results

There were 59 hostels considered; 12 were excluded because of uptake rates of > 80% and one was excluded because it would not allow peers into the venue. Forty-six hostels with a total of 2342 residents were randomised, 1192 in control hostels and 1150 in intervention hostels (Figure 4). Randomisation appeared well balanced (chi-squared p-value > 0.47 for all baseline characteristics; Table 7).

FIGURE 4.

Work package 3 flow diagram. ITT, intention to treat. Reproduced from Aldridge et al. 74 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/. The figure includes minor additions and formatting changes to the original text.

Reproduced from Aldridge et al. 74 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/

Median screening uptake was 44% [interquartile range (IQR) 26–59%)]: 45% (IQR 33–55%) in control hostels and 40% (IQR 25–61%) in intervention hostels. Using Poisson regression to account for the clustered study design, size of hostel and previous screening uptake, there was no evidence of peer educators affecting the uptake of screening, with an adjusted risk ratio of 0.98 (95% CI 0.80 to 1.20) (Table 8). These results did not differ significantly in secondary analyses examining the effect by hostel size, historical screening uptake and intervention fidelity (Table 9).

The study team noted no adverse events.

Discussion

This CRCT showed no evidence that peer educators increased the uptake of screening.

The study included most homeless hostels being screened by Find&Treat, a project with broad geographical coverage across London. Cluster randomisation was used as individual randomisation would not have been possible. The study was not powered to detect a difference in tuberculosis cases identified by the two arms as this would require a considerably larger sample. We did not collect individual data as part of the study as this would have required individual consent and would have interrupted the flow of screening and caused unacceptable delays for service users.

Evidence of the effectiveness of peer education interventions is mixed. One review found evidence for peer education increasing physical activity, decreasing smoking and increasing condom use, but no evidence for breastfeeding, medication adherence, women’s health and participation in general activities. 79 A further review of the use of peers educators in HIV also found mixed evidence of effects. 80 This same review attempted to examine what implementation factors, including peer educator recruitment, supervision and training, improved effectiveness, but found a lack of evidence to support any of these approaches, partly as a result of low sample size.

Several factors may have influenced results. Most sites were not naive to the peer intervention as a result of the increased use of peers alongside Find&Treat in the time between the study design and its conduct; thus, at several sites we, in effect, withdrew the ‘intervention’ from the control arm. The prior use of peers may have led to an increased awareness among hostel staff and residents, reducing any effect during the trial. The primary outcome of screening uptake does not take account of peers’ previously reported ability81 to engage the more difficult-to-reach and vulnerable cases, which could be assessed only by the collection of individual-level data. Such an effect (if it exists) could ultimately lead to an increase in the rate of detection of active tuberculosis. Having peers available at the time of screening may also help with the engagement of those who are screened and require further health-care management. Peers are then able to accompany people to follow-up appointments, based on a relationship that began on the mobile radiographic unit during the peer education work. Finally, the presence of peers may have reduced the intensiveness with which Find&Treat staff promoted uptake but may have freed them to focus on other activities. Peers in this study were associated with minimal costs as they were volunteers and may benefit from involvement in structured, meaningful activity as suggested by previous qualitative studies. 81

We found no evidence for an increased uptake of screening in this study; however, peer educators may have contributed to other, unmeasured factors in the screening process. Additionally, involvement of peers in the screening is likely to have directly benefited the peers as a result of the training and skills that were learnt during the research process, factors that were not measured during this study.

Acknowledgements

We would like to acknowledge the invaluable help and support of Groundswell and all the peers who volunteered for this trial.

Deviations from the protocol

With the exception of the revised power calculations that were required because of the change in average hostel size and the exclusion of one hostel as a result of the refusal of the hostel to allow peer workers on site, there were no recorded protocol violations.

Added value

Find&Treat and Groundswell continue to highly value the role of peers; however, partly as a result of this study, Find&Treat has evolved the peer model, increasingly moving volunteer peers into paid roles providing support across the diagnostic and treatment pathway. There is an emphasis on matching peers to clients’ backgrounds to support their care. There is also an emphasis on training peers to undertake clinical activities, such as obtaining oral swabs, conducting near-patient tests for blood-borne viruses and undertaking liver fibroscanning. Partly as a result of our involvement in this study, we are now collaborating on a NIHR Research for Patient Benefit grant assessing the effectiveness of the Groundswell peer advocacy model (Chief Investigator Lucy Platt). A Wellcome Trust Clinical Research Fellow (Neha Pathak) is also using Groundswell peer educators in a survey of cervical cancer screening uptake and sexually transmitted infection prevalence among women experiencing homelessness.

Introduction

Experience of tuberculosis screening among socially complex groups and other tuberculosis patients by the pan-London Find&Treat tuberculosis outreach service has shown that current service provision in the UK rarely achieves same-day diagnosis in smear-positive pulmonary cases. In the UK, patients often have to wait for a follow-up appointment to be given the results of smear microscopy. Patients who are smear negative and clinically well often have to wait several weeks for the results of the culture prior to commencing treatment and there is further delay in the availability of drug sensitivity tests to inform clinical management. This carries an important risk of loss to follow-up care, especially among socially complex groups, and inappropriate arrangements for isolation and infection control in hospital and prison settings. The use of PCR-based molecular technologies allows a high proportion of smear-negative cases (later confirmed by culture) to be diagnosed within 48 hours. 84 In addition, these technologies enable the identification of mutations specific to rifampicin drug resistance (a key marker of multidrug resistance). The reagents used in the test are effective in sterilising samples, making them suitable for testing outside category 3 laboratories and suggesting a potential role as a point-of-care test. 85 Although these tests are available on request in several London laboratories, there is a paucity of evidence to support their use as a near-patient test in field conditions.

Methods

Results

During the study period, 95 patients met the inclusion criteria for the study, 37 of whom were recruited (Figure 5). In the 18 patients in whom Cepheid was used, 14 had negative results, two were positive for Mycobacterium tuberculosis and two had indeterminate results. Six of these patients were ultimately diagnosed with active tuberculosis, 10 were classified as not having tuberculosis and two were lost to follow-up. In the control arm, of the 19 patients identified, five were ultimately diagnosed with active tuberculosis, two were lost to follow-up and 12 were classified as not having active tuberculosis. It did not prove possible to obtain more detailed primary and secondary outcome measures from the NHS services. Owing to low recruitment and difficulties collecting primary and secondary outcome data (see Challenges and deviations from the protocol) and following discussion with the funders, the trial was abandoned. The intention was to continue the evaluation as an observational study of the technology on the mobile radiographic unit, but the mobile radiographic unit stopped using the technology soon after trial abandonment.

FIGURE 5.

Work package 4: Cepheid recruitment update/outcomes – from August 2013 to July 2015. CXR, chest X-ray; LFU, lost to follow-up; MXU, mobile X-ray unit; PHE, Public Health England; TB, tuberculosis.

Challenges and deviations from the protocol

This study, which sought to evaluate the use of molecular diagnostics on a mobile radiographic unit screening highly complex patients, was subject to many challenges. Optimising the Cepheid technology to function in transit was a major challenge. After initial installation, it became apparent that brief interruptions of power supply disrupted its use, requiring the installation of a custom-made back-up battery. We overcame practical issues in implementation, which included the development of sputum pots with one-way valves to allow the safe handling of sputum specimens outside a category 3 laboratory. We also trained non-specialist staff in the use of the equipment and developed procedures and permissions for the local microbiology service to accredit the mobile diagnostic service. Recruitment commenced but became very slow as the mobile radiographic unit was a highly pressured area for the consenting and randomisation of participants alongside routine screening plus other screening tests being offered in the unit that had an important impact on the recruitment rate. Staff reported that it was extremely challenging to explain the randomised nature of the study to patients in a highly pressurised clinical environment. Conducting the test interrupted their other clinical activities. They also reported that the test took too long to produce results (up to 1 hour), which made it hard for patients to wait for results. In addition, in many cases the mobile unit’s schedule meant that it needed to move to another screening location before test results were completed. In light of these challenges, it was decided to terminate the study. Although it was planned to establish this as a service evaluation study rather than a RCT, we found that, following cessation of the trial, the Cepheid system was used only twice (both results were negative). We conclude that routine use of this diagnostic technology has limited utility in the current context. The Find&Treat team now uses the technology in selected cases and during outbreak investigations. More technologically simple and rapid confirmatory tests are required.

Introduction

Directly observed treatment has been the standard of care for tuberculosis since the early 1990s. 87,88 DOT was developed following observations that irregular treatment could threaten clinical and public health outcomes through the generation of drug resistance, relapse and transmission of infection. 89 DOT is recommended by WHO1 and the American Thoracic Society. 90 In England, DOT is recommended for patients at high risk of poor adherence. 11 This includes those with clinically complex disease, MDRTB, mental health problems, previous tuberculosis treatment or previous poor adherence. 11 It also includes socially complex groups (persons with a history of homelessness, imprisonment, drug use or alcohol problems).

Directly observed treatment can be administered in clinic, community or home settings but is a substantial inconvenience to patients and service providers. Regimens of three times per week have also been approved with DOT and are used in England,11 although they are not currently recommended by WHO owing to increased risk of treatment failure. 1 Seven-day regimens are, therefore, generally administered through DOT 5 days per week and through self-administered treatment at the weekend.

Advances in technology have raised the possibility of remote VOT. 91 Initially, this required a live video call (synchronous VOT). 92–97 More recently, apps have been developed that enable video clips to be recorded and forwarded for later viewing (asynchronous VOT). 98 Asynchronous VOT is currently used by some clinics in the USA and has high reported levels of patient acceptability, decreased costs and programmatic evidence of effectiveness. 98–102 The WHO, therefore, recently conditionally recommended VOT as an alternative to DOT, but the evidence was graded as weak because of a lack of RCTs. 1 In addition, VOT has yet to be assessed in patients from socially complex groups.

Methods

Results

Trial population

Recruitment began on 1 September 2014 and continued until 1 October 2016, when the study’s independent Trial Steering Committee advised that recruitment should be stopped based on interim analysis results. Follow-up continued until 31 December 2016. Flow through the study is summarised in Figure 6.

FIGURE 6.

Enrolment and randomisation. a, The most common ‘other reason’ for failing to enrol patients (32 out of 45) was that the clinic staff considered that the patient needed intensive face-to-face support owing to emotional, medical or physical reasons or because of imminent risk of loss to follow-up. Reproduced from Story et al. 86 © 2019 Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license. See: http://creativecommons.org/licenses/by/4.0/. The figure includes minor additions and formatting changes to the original text.

Reproduced from Story et al. 86 © 2019 Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license. See: http://creativecommons.org/licenses/by/4.0/. The figure includes minor additions and formatting changes to the original text.

The ITT analyses included 114 patients randomised to DOT and 112 patients randomised to VOT. Baseline characteristics of patients are shown in Table 10. Patients were mainly young adults born outside the UK. A high proportion (131 out of 226 patients; 58%) had a history of homelessness, imprisonment, drug use, alcohol problems or mental health problems. The baseline characteristics were similar in the two arms.

TABLE 10 - Characteristics of the patients at baseline, by allocated intervention

History of homelessness, imprisonment, drug use or alcohol problems, and mental health problems.

There were no significant differences in the baseline characteristics between allocated groups.

Reproduced from Story et al. 86 © 2019 Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license. See: http://creativecommons.org/licenses/by/4.0/. The table includes minor additions and formatting changes to the original text.

Characteristic	DOT		VOT
	n	%	n	%
Total	114		112
Age group (years)
16–34	61	53.5	64	57.1
35–54	45	39.5	35	31.3
≥ 55	8	7.0	13	11.6
Sex
Male	83	72.8	82	73.2
Female	31	27.2	30	26.8
Born in UK
No	83	72.8	93	83.0
Yes	31	27.2	19	17.0
Previous tuberculosis
No	82	71.9	85	75.9
Yes	30	26.3	27	24.1
Pulmonary tuberculosis
Yes	73	64.0	69	61.6
No	41	36.0	43	38.4
Social risk factora
Never	48	42.1	47	42.0
> 5 years ago	19	16.7	19	17.0
Within 5 years	47	41.2	46	41.1
Homeless
Never	77	67.5	70	62.5
> 5 years ago	14	12.3	16	14.3
Within 5 years	23	20.2	24	21.4
Prison
Never	93	81.6	97	86.6
> 5 years ago	9	7.9	8	7.1
Within 5 years	11	9.6	7	6.3
Drug use
Never	96	84.2	89	79.5
> 5 years ago	2	1.8	4	3.6
Within 5 years	15	13.2	18	16.1
Alcohol problems
No	91	79.8	92	82.1
Yes	21	18.4	17	15.2
Mental health problems
No	94	82.5	94	83.9
Yes	18	15.8	14	12.5

In the ITT analysis, 78 out of 122 (70%) VOT patients successfully achieved the primary outcome (≥ 80% scheduled observations successfully completed during the first 2 months), compared with 35 out of 114 (31%) DOT patients [adjusted odds ratio (aOR) 5.48, 95% CI 3.10 to 9.68; p < 0.0001]. In the restricted analysis, the proportions with the primary outcome were 78 out of 101 patients (77%) for VOT and 35 out of 56 patients (63%) for DOT (aOR 2.52, 95% CI 1.17 to 5.47; p = 0.019).

High observation rates were maintained in the VOT arm, but they rapidly declined in the DOT arm. Over the full follow-up period (up to 6 months), 12,422 out of 16,230 (77%) scheduled observations were completed in the VOT arm, compared with 3884 out of 9882 (39%) scheduled observations in the DOT arm (p < 0.0001). In the restricted analysis over the full follow-up period, 12,422 out of 14,907 (83%) scheduled observations were completed in the VOT arm, compared with 3884 out of 6351 (61%) scheduled observations in the DOT arm (p < 0.0001). Observation completion rates were higher for VOT than DOT in all subgroups.

There were no significant differences in positive sputum cultures at 2 months following treatment onset, treatment completion and loss to follow-up or numbers of hospital admissions between trial arms.

Average staff time per dose observed was 56 (SD 54) minutes for community-based DOT (including travel time), 15 (SD 12) minutes for clinic-based DOT and 3.2 (SD 0.5) minutes for VOT. Patients on DOT spent a mean of 29 (SD 48) minutes per week on treatment observation (including travelling to/from clinics, waiting for appointments and appointment time). Those on VOT spent a mean of 1.8 (SD 2.2) minutes setting up and recording each video.

Discussion

Video-observed treatment was substantially more effective than DOT at ensuring that scheduled doses were observed. A large part of this effect was due to patients failing to engage in DOT over the first week of treatment. After restricting analyses to those patients who engaged with observation in the first week, VOT remained significantly superior at ensuring that scheduled observations took place. This effect was maintained throughout treatment. VOT observations involved substantially less health-care worker time and patient time than DOT.

Systematic reviews and meta-analyses draw differing conclusions about the effectiveness of DOT;105,106 however, a recent review showed that DOT increased treatment success, adherence and 2-month sputum conversion, and decreased loss to follow-up and acquired drug resistance compared with self-administered treatment. 1 Community DOT was more effective than hospital-based or clinic-based DOT, demonstrating the importance of making DOT convenient for patients. The review also showed that DOT was more effective when delivered by health staff or lay workers than when delivered by family members. 1 VOT provides the technology to support professional treatment observation that is potentially more convenient and cheaper than in-person DOT.

Video-observed treatment, as used in this study, has a wide range of components beyond convenience of observation. The intervention included personal support: patients met the VOT observer for training and received regular personalised messages as reminders, confirmation of receipt of video clips or follow-up when clips were not received. The observers also supported onward referral to deal with reported adverse events. Patients were provided with a smartphone with a data plan and free domestic calls and text messages. This acted both as an enabler (as it facilitated easy communication and improved access to care providers) and as a material incentive that was valued by patients. Phones were reused throughout the trial.

The trial had a number of limitations. It was not possible to blind patients or treatment observers to the intervention, although the investigators and analysts were blinded. We could not distinguish between doses that were taken but were not observed and doses that were not taken.

The primary outcome (observing 80% of scheduled doses) could be considered to be biased towards DOT; it required substantially more VOT doses (scheduled 7 days per week) to be observed than DOT doses (scheduled three or five times per week). The restricted analysis further favoured DOT as it included only the subset of patients allocated to DOT who were willing to be observed.

The study was not powered to detect differences in culture conversion rate, treatment completion, loss to follow-up, relapse or development of drug resistance; nevertheless, it is reasonable to assume that improved adherence might improve all of these outcomes.

Another limitation was the exclusion of MDRTB patients from the randomised trial. They were excluded because they had a range of treatment regimes, some of which included multiple scheduled doses to be observed per day. It would be impractical for all of these observations to be made through face-to-face DOT and we therefore did not consider there to be equipoise between VOT and DOT for MDRTB patients; however, 26 MDRTB patients were offered VOT in a non-randomised arm of the trial. In this arm, it was harder to be certain about which doses were scheduled for VOT because of the higher frequency of dosing, the fact that some doses (particularly injectable doses) continued to be administered at regular clinic visits and the higher rates of hospitalisation. Nevertheless, all but one of the 26 MDRTB patients accepted VOT and 16 MDRTB patients had > 80% doses observed in the first 2 months. There was also evidence that the intervention could be sustained for the long-term regimens required for treatment completion in MDRTB; VOT was maintained for a maximum of 796 days.

There is an urgent global need for more effective and cheaper alternatives to DOT to enable effective ambulatory care of both drug-sensitive tuberculosis and MDRTB. In particular, it is critical that new opportunities for shorter regimens for MDRTB are not lost as a result of insufficient attention to adherence. 107

The WHO now recommends that VOT can be a ‘suitable alternative’ to DOT and has published guidance on its implementation. 1 The results of this trial may allow WHO to make firmer recommendations about the use of DOT and VOT.

Deviations from the protocol

The full trial protocol and analysis plan are presented in Report Supplementary Material 3 and Report Supplementary Material 4. Some eligible patients (45/548 patients assessed for eligibility) were not recruited for a range of reasons that were not within the original exclusion criteria (see Figure 8). With the exception of the trial being terminated following an interim analysis requested by the funder, there were no clear protocol violations. No adverse effects were reported.

Challenges

In our original submission to NIHR, we proposed a trial comparing community DOT with clinic DOT; however, in the time between study conception and delivery, the use of community DOT in routine practice had increased considerably. We also found it very challenging to recruit community DOT providers for this work; therefore, with the permission of NIHR, we redesigned the trial to assess VOT. This followed a small pilot by Find&Treat that had found the approach to be acceptable to patients. Considerable delays were introduced by the multicentre aspect of the study, which required research and development approvals across 22 sites in four cities. Local principal investigators were required in each site and, frequently, these principal investigators (primarily specialist tuberculosis nurses) had not previously been involved in research and needed to undertake training in Good Clinical Practice before the trial could start. When sites were up and running, we found recruitment to be slower than anticipated. This was, in part, due to the decreasing numbers of tuberculosis cases in London and, in part, due to a reluctance by staff in clinics to randomise people who, despite being eligible for DOT, they thought would not accept DOT or for whom they thought that it would be difficult to arrange. In view of these delays, NIHR requested an interim analysis and, following this, based on overwhelming evidence of the superiority of one intervention arm, the external steering group recommended cessation of the trial.

Added value

At the time of writing, the VOT trial has already had considerable impact. The work has led to a collaboration with the Behavioural Insights Team to undertake a trial of video-observed therapy for tuberculosis in Moldova, which is now complete and being analysed. Discussion of the findings at meetings of the WHO Digital Health for the End Tuberculosis Strategy (for which the co-chairperson and co-applicant is A Story) has contributed to WHO recommending VOT as a suitable alternative to DOT in countries with suitable technological infrastructure. The London Clinical Commissioning Groups have commissioned Find&Treat to provide a VOT service for London, and other parts of the country also commission VOT from Find&Treat on an ad hoc basis.

Latent tuberculosis infection screening in the homeless: transmission dynamic and health economic analysis

Economic analysis of video-observed treatment versus directly observed treatment

Parts of this section have been reproduced or adapted from Story et al. 86 © 2019 Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license. See: http://creativecommons.org/licenses/by/4.0/. The text includes minor additions and formatting changes to the original text.

Recommendations for future research

• How can the uptake of treatment of latent tuberculosis infection be maximised in homeless populations?

• How can radiographic and latent tuberculosis infection screening programmes in prisons best be operationalised?

• How can identification and treatment of homeless populations, prisoners and drug users for hepatitis C best be maximised?

• How can prison and community tuberculosis care best be integrated to reduce loss to follow-up and treatment interruption?

• What is the effectiveness of VOT in high-incidence countries?

• Can VOT be used to support adherence in other conditions in socially complex groups (e.g. hepatitis C and mental health management)?

Contributions of authors

Alistair Story (https://orcid.org/0000-0003-3853-1715) (Honorary Senior Lecturer and Find&Treat Clinical Lead) contributed to the co-design of the programme grant, was co-lead for study implementation, and contributed to analysis and interpretation.

Elizabeth Garber (https://orcid.org/0000-0002-5962-3232) (Project Manager) contributed to the analysis, interpretation and drafting of reports.

Robert W Aldridge (https://orcid.org/0000-0003-0542-0816) (Research Fellow) contributed to the study implementation, was the lead for analysis, and contributed to analysis and interpretation.

Catherine M Smith (https://orcid.org/0000-0003-3959-8479) (Research Associate) contributed to the analysis of trial-based data and the compilation of the final report.

Joe Hall (https://orcid.org/0000-0002-7681-6124) (Research Associate) contributed to data collection and study implementation, and to the analysis and interpretation.

Gloria Ferenando (https://orcid.org/0000-0001-8580-3037) (Research Nurse) contributed to data collection and the VOT trial implementation, and to the analysis and interpretation.

Lucia Possas (https://orcid.org/0000-0001-9758-3449) (Study Data Manager) contributed to the study implementation, analysis and interpretation.

Sara Hemming (https://orcid.org/0000-0003-3487-0517) (Research Nurse) contributed to the study implementation, analysis and interpretation.

Fatima Wurie (https://orcid.org/0000-0003-4802-2308) (Research Assistant) contributed to the study implementation, analysis and interpretation.

Serena Luchenski (https://orcid.org/0000-0002-6260-4606) (Research Fellow) contributed to the study analysis and interpretation.

Ibrahim Abubakar (https://orcid.org/0000-0002-0370-1430) (Professor of Infectious Disease Epidemiology) contributed to the study design, implementation, analysis and interpretation.

Timothy D McHugh (https://orcid.org/0000-0003-4658-8594) (Professor of Microbiology) contributed to the study design, implementation, analysis and interpretation.

Peter J White (https://orcid.org/0000-0002-6644-3512) (Reader) contributed to the transmission modelling and economics.

John M Watson (https://orcid.org/0000-0002-2595-405X) (Honorary Professor of Epidemiology) contributed to the study design, implementation, analysis and interpretation.

Marc Lipman (https://orcid.org/0000-0001-7501-4448) (Consultant in Respiratory Medicine) contributed to the study design, implementation, analysis and interpretation, and was the clinical lead.

Richard Garfein (https://orcid.org/0000-0003-3663-7153) (Professor of Global Health) contributed to the VOT study design, implementation, analysis and interpretation.

Andrew C Hayward (https://orcid.org/0000-0002-3549-6232) (Professor of Infectious Disease Epidemiology and Inclusion Health) was the chief investigator, contributed to the co-design of the programme grant and was co-lead for study implementation, contributed to analysis and interpretation, and collation of the final report.

Publications

The following publications have arisen directly or indirectly (e.g. as a result of methodological developments and collaborations that would not have otherwise happened) from this programme grant.

Data-sharing statement

All data requests should be submitted to the corresponding author for consideration. Access to anonymised data may be granted following review.

Patient data

This work uses data provided by patients and collected by the NHS as part of their care and support. Using patient data is vital to improve health and care for everyone. There is huge potential to make better use of information from people’s patient records, to understand more about disease, develop new treatments, monitor safety, and plan NHS services. Patient data should be kept safe and secure, to protect everyone’s privacy, and it’s important that there are safeguards to make sure that it is stored and used responsibly. Everyone should be able to find out about how patient data are used. #datasaveslives You can find out more about the background to this citation here: https://understandingpatientdata.org.uk/data-citation.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, CCF, NETSCC, PGfAR or the Department of Health and Social Care. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the PGfAR programme or the Department of Health and Social Care.

Introduction

We analysed the cost-effectiveness of adding screening and treatment for latent tuberculosis infection to screening and treatment for active tuberculosis in people experiencing homelessness in London, using an integrated transmission dynamic and health economic model. 108 Dynamic transmission models are preferred over health-state transition models as they can take account of onward transmission chains more effectively. The model used is an individual-based model in which individuals are ‘followed’ while they are homeless and after re-entering the general population, to capture the long-term health consequences of tuberculosis infection, for case finding and for treatment while homeless; infection acquired while homeless can result in active tuberculosis disease after re-entering the general population (which causes QALY loss and incurs treatment cost), and the treatment of latent tuberculosis infection while homeless may avert this.

Analysis

In the model, each individual is in one of a number of health states: uninfected (naive), uninfected (recovered), latently infected (fast progressing or slow progressing), with active tuberculosis disease (sputum smear negative or positive), on treatment for latent tuberculosis infection or on treatment for active tuberculosis. Figure 7 shows transitions between states that can occur in the model. Uninfected individuals who become infected move into the latent infection state. Individuals with tuberculosis infection history (recovered) are assumed to have partial immunity and, therefore, have a reduced probability of (re-)acquiring tuberculosis infection compared with naive individuals. Individuals in the latent state can progress to active tuberculosis disease, which is infectious and can be diagnosed through active or passive case finding, resulting in individuals being placed on tuberculosis treatment. Active case finding for active tuberculosis occurs through chest radiography on the mobile radiographic unit, with cases of abnormal radiology being referred to hospital for tuberculosis diagnosis. Passive case finding for active tuberculosis occurs when individuals present for care. In scenarios in which latent tuberculosis infection screening occurs, a proportion of diagnosed individuals are treated. Patients who are successfully treated for latent tuberculosis infection or for active tuberculosis enter the recovered state. Patients who are not successfully treated return to their former state of latent tuberculosis infection or active tuberculosis disease. Individuals whose IGRA test was false positive remain in the naive or recovered state regardless of whether they are treated or not, but costs are incurred for those who are treated. Active tuberculosis disease causes QALY loss through morbidity (which is reduced by treatment) and mortality.

FIGURE 7.

Flow diagram showing health states of and treatment outcomes for patients in the model of tuberculosis. ‘Untreated smear positive’ and ‘Untreated smear negative’ compartments denote infectious health states. Different lines represent different transitions: infection (dark-blue lines), disease progression (mid-blue lines), tuberculosis treatment initiation (orange lines), tuberculosis treatment failure (dashed grey lines) and recovery from tuberculosis disease (light-blue lines). LTBI, latent tuberculosis infection; TB, tuberculosis.

Economic analysis

The economic analysis is from an NHS perspective. In the baseline (current practice) scenario, a mobile radiographic unit provides people experiencing homelessness with screening for active tuberculosis; individuals with abnormal radiographs are referred to hospital for diagnosis and those with tuberculosis are treated. Individuals with active tuberculosis can also be found passively, when they present for care.

We compared the baseline scenario with intervention scenarios in which people experiencing homelessness are offered testing for latent tuberculosis infection by IGRA, and offered treatment if they test positive. As treatment was not offered in the empirical study, there is uncertainty regarding the proportion of IGRA-positive patients who would accept and complete latent tuberculosis infection treatment, so we compared scenarios in which this proportion varied (i.e. 25%, 50% and 75%). The intervention is applied for 10 years, and the model considers the lifetime of the population cohort. Analysis was from the perspective of the NHS. The total net costs incurred (including staff time, drugs and diagnostic tests) and the QALYs accrued by the patient cohort are calculated. The cost-effectiveness of different options is compared using incremental ICERs relative to current practice. Each scenario (baseline and interventions) was run 1000 times. Costs and health utilities are discounted at 3.5% per annum. Parameter values were obtained primarily from sources used to inform NICE guidance11,33 or this study, as appropriate (Tables 13 and 14).

Results

In our analysis, screening for and treating latent tuberculosis infection had a net cost, as the intervention cost was greater than the averted costs of active tuberculosis disease; however, the intervention was beneficial to health, with a QALY gain.

Whether or not the intervention would be considered cost-effective depends on the value of 1 QALY and the proportion of IGRA-positive individuals who are treated for latent tuberculosis infection (Figures 8 and 9). The higher the value of 1 QALY, and the greater the proportion of IGRA-positive patients who are treated for latent tuberculosis infection, the more likely it is that the intervention would be considered cost-effective.

FIGURE 8.

Cost-effectiveness plane for latent tuberculosis infection screening and treatment compared with current practice over a 10-year intervention, with discounting at 3.5% per annum. LTBI, latent tuberculosis infection.

FIGURE 9.

Cost-effectiveness acceptability curves for latent tuberculosis infection screening and treatment compared with current practice for a 10-year intervention, with discounting at 3.5% per annum. Three different proportions (25%, 50% and 75%) of IGRA-positive individuals being treated for latent tuberculosis infection are considered. LTBI, latent tuberculosis infection.

In the three scenarios that we considered regarding the proportion of IGRA-positive patients who are treated for latent tuberculosis infection, we found that if this proportion was 25% then introducing latent tuberculosis infection screening and treatment would be cost-effective when 1 QALY was valued at £30,000 (see Figure 9). If the proportion was 50% or 75%, then introducing latent tuberculosis infection screening and treatment would be cost-effective if 1 QALY was valued at £20,000; however, there is substantial uncertainty about the cost-effectiveness (Table 15).

The points in Figure 8 show incremental costs and incremental QALYs from individual model realisations comparing the effect of latent tuberculosis infection screening and treatment with current practice for different proportions (25%, 50% and 75%) of IGRA-positive individuals being treated for latent tuberculosis infection. Square markers show the mean for each scenario. The diagonal lines in Figure 8 show the cost-effectiveness thresholds if 1 QALY is valued at £30,000 (thin line) or £20,000 (bold line).

Limitations

There are limited sensitivity analyses conducted and estimates of uncertainty are not presented. We do not report cost per case averted. We do not consider the adverse effects of latent tuberculosis infection and tuberculosis treatment.

Discussion

Successful treatment of latent tuberculosis infection prevents progression to infectious active tuberculosis disease and thereby averts transmission, which benefits population health and averts future costs of treatment in others, as well as benefiting the individual patient. However, not all patients with latent tuberculosis infection would progress to active tuberculosis disease without treatment and, therefore, several patients with latent tuberculosis infection have to be treated to avert a case of active tuberculosis disease. We find that the cost-effectiveness of adding screening and treatment for latent tuberculosis infection to screening and treatment for active tuberculosis in people experiencing homelessness depends on the proportion of IGRA-positive patients who accept and complete latent tuberculosis infection treatment, and that the greater this proportion is, the more likely that the intervention would be considered cost-effective. As treatment for latent tuberculosis infection was not recommended at the time of the study, it was not offered to patients and, therefore, we do not know what proportion of patients would accept and complete treatment. As this is a key determinant of cost-effectiveness, further empirical study is required to determine if this intervention should be recommended.

The probability that latent tuberculosis infection screening is cost-effective would be increased if the cost of latent tuberculosis infection testing were lower than in our analysis. With latent tuberculosis infection screening and treatment for immigrants from high-burden countries now recommended, bulk procurement of test kits may enable cost savings that would make latent tuberculosis infection screening and treatment in the homeless more likely to be cost-effective.