The effectiveness and cost-effectiveness of screening for active tuberculosis among migrants in the EU / EEA : a systematic review

Christina Greenaway1,2, Manish Pareek3, Claire-Nour Abou Chakra4, Moneeza Walji2, Iuliia Makarenko2, Balqis Alabdulkarim2, Catherine Hogan1,2, Ted McConnell2, Brittany Scarfo2, Robin Christensen5, Anh Tran6, Nick Rowbotham6, Teymur Noori7, Marieke J van der Werf7, Kevin Pottie8, Alberto Matteelli9, Dominik Zenner10,11, Rachael L Morton6 1. Division of Infectious Diseases, Jewish General Hospital, McGill University, Montreal, Canada 2. Centre for Clinical Epidemiology of the Lady Davis Institute for Medical Research, Jewish General Hospital, Montreal, Canada 3. Department of Infection, Immunity and Inflammation, University of Leicester, Leicester, United Kingdom 4. Department of Microbiology and Infectious Diseases, Université de Sherbrooke, Québec, Canada 5. Musculoskeletal Statistics Unit, The Parker Institute, Bispebjerg and Frederiksberg Hospital, Copenhagen, Denmark 6. National Health and Medical Research Council (NHMRC) Clinical Trials Centre, University of Sydney, Sydney, Australia 7. European Centre for Disease Prevention and Control, Stockholm, Sweden 8. C.T. Lamont Primary Health Care Research Centre, Bruyère Research Institute, Ottawa, Ontario, Canada 9. Clinic of Infectious and Tropical Diseases, University of Brescia and Brescia Spedali Civili General Hospital, World Health Organization Collaborating Centre for TB/HIV and TB Elimination, Brescia, Italy 10. Respiratory Diseases Department, Centre for Infectious Disease Surveillance and Control (CIDSC), Public Health England, London, United Kingdom 11. Department of Infection and Population Health, University College London, London, United Kingdom


Introduction
Tuberculosis (TB) is a public health priority in the European Union (EU) and European Economic Area (EEA), and countries have committed themselves to the World Health Organization (WHO) End TB Strategy with an ambitious goal to end TB [1][2][3][4].The foreign-born population make up an increasing and considerable number and proportion of all TB cases in countries with low TB incidence (< 10 Methods cases/100,000 population) and challenge TB elimination efforts in the EU/EEA [3,5].More than one quarter of reported TB cases in 2015 in the EU/EEA occurred in the foreign-born population [5].This proportion has been increasing steadily; in 2007, 13.6% of TB cases occurred in migrant populations whereas in 2013, they accounted for 21.8% [6].In many low TB incidence countries in the EU/EEA, more than half of all TB cases occur among foreign-born individuals [5].Between 2007 and 2012, the EU/EEA received on average 1.5 million migrants from outside of the EU/EEA, and larger numbers in 2015 and 2016 [7,8].As a result, the foreignborn population now makes up 11.4% of the population in the EU/EEA and exceeds 15% in many low TB incidence countries [7,8].A considerable proportion of these migrants were born in countries with a high TB burden [9,10].
Given the disproportionate TB case noti-㰊cations in migrant populations and the faster decline of TB rates in host populations, enhanced TB control strategies among migrants will be necessary to achieve TB elimination in the EU/EEA (de-㰊ned as achieving a rate of less than one case of TB per 1,000,000 population) [1][2][3][4]11,12].Countries have generally focused on two targeted control strategies among migrants: (i) identi-㰊cation of active TB with chest radiography (CXR) before or soon after arrival in the host country to detect prevalent TB cases and limit onward transmission and (ii) more recently, identifying and treating latent TB in migrants from high TB burden countries to prevent TB reactivation [13].Many EU/EEA countries with low TB incidence screen migrants for active TB on or soon after arrival.The migrant groups targeted for screening and the location of screening are different for each country because screening guidelines for active TB in migrants are lacking at the EU/EEA level [13][14][15].We conducted a systematic review on the effectiveness and a second systematic review on the costeffectiveness of screening for active TB among migrants in the EU/EEA region with the aim of informing migrant screening guidelines.

Overall approach and key questions
This review supports a project of the European Centre for Disease Prevention and Control (ECDC) to develop guidance on screening for six infectious diseases (chronic hepatitis C, hepatitis B, HIV, TB (active and latent) and intestinal parasites) in newly arrived migrants to the EU/EEA.The project followed the new Grading of Recommendations Assessment, Development and Evaluation (GRADE)-ADOLOPMENT approach to conduct systematic reviews on screening migrant populations for these six infectious diseases [16].The review protocol and the methods of GRADE-ADOLOPMENT guideline development have been published [16,17].All reviews followed a Cochrane methodological approach and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) methods for reporting systematic reviews [18].For each review, we developed two research questions (using a population, intervention, comparison and outcome (PICO) framework), an analytic framework to illustrate the screening evidence pathway, and identi-㰊ed and prioritised clinically important outcomes, following the evidence-based review methods described by the United States (US) Preventative CXR: chest radiography; NNS: number needed to screen; TB: tuberculosis.
Task Force [19,20].We sought to answer two research questions: (i) what is the effectiveness of screening migrants arriving and living in the EU/EEA for active TB and (ii) what is the resource use, cost and cost-effectiveness of screening migrants for active TB?We developed an analytic framework that identi-㰊ed the evidence chain to address the effectiveness and cost-effectiveness of active TB screening among migrants (Figure 1) [17].We developed the following key questions along this evidence chain: (i) what is the yield of active TB screening with CXR in migrants, (ii) what are the test performance characteristics of CXR to detect active TB, (iii) how effective is active TB therapy and what are the associated harms, (iv) what is the uptake of active TB screening by migrants, and (v) how cost-effective is screening for active TB in migrants [17]?
Figure 1 Analytic framework of the evidence chain for active tuberculosis screening in migrants Search strategy and selection criteria Following the GRADE-ADOLOPMENT process, we identi-㰊ed an evidence review that assessed the effectiveness of latent TB infection (LTBI) screening among migrants, published in 2011 by the Canadian Collaboration on Immigrant and Refugee Health (CCIRH), and used this as a starting point for our literature search (anchoring review) [16,21].The CCIRH review included systematic reviews on the effectiveness of LTBI screening in migrants up to 2008 but did not review costeffectiveness.We therefore conducted two separate searches to address our research questions.The -㰊rst search updated the CCIRH evidence review and identi-㰊ed systematic reviews and guidelines on the effectiveness and costeffectiveness of TB screening programmes in migrant populations from 2005 to 2016.The second search identi-㰊ed individual studies on the resource use, costs and cost-effectiveness of TB screening programmes for migrants over a longer period, 2000 to 2016, given these topics were not covered in the CCIRH evidence review.For the -㰊rst search, MEDLINE via Ovid, Embase, the Cumulative Index to Nursing and Allied Health Literature (CINAHL), Epistemonikos and Cochrane

Study selection and quality assessment
We identi-㰊ed and included systematic reviews and evidence-based guidelines that directly addressed each key question along the active TB screening evidence chain and prioritised documents focusing on newly arrived (< 5 years in the host country) migrants.Migrant populations included were non-forced economic migrants, and refugees, asylum seekers and illegal migrants who may have been forced to ‾䐴ee con‾䐴ict, natural disaster, or economic peril [17].We only included studies published in full and in English or French.If more than one version of a systematic review was identi-㰊ed, the most recent was considered.Studies were excluded if they were not relevant to the key questions, if they were not a systematic review or guideline, if the study methodology was unclear, and if they focussed only on non-generalisable subgroups (such as healthcare workers or HIV-positive people) or addressed only latent TB screening.Two authors screened the titles and abstracts, assessed selected full-text articles for eligibility and extracted data from included articles.Disagreements were resolved by consensus or by a third author.The methodological quality of systematic reviews was assessed using the AMSTAR tool (A Measurement Tool to Assess Systematic Reviews) and the quality of individual studies was assessed with the Newcastle-Ottawa scale [22,23].The GRADE criteria were applied to assess the quality and certainty of the evidence for the individual studies included in the systematic reviews [24].

Data extraction and synthesis
The following information was extracted from each study: study design, objectives, analyses, quality assessment of the individual studies included in the systematic review, population examined, number of included studies, total number of participants included, intervention, outcome and results.We created GRADE evidence pro-㰊les and summary of -㰊ndings tables for each outcome where

Results
appropriate.Numbers needed to screen (NNS) were estimated by calculating 1/mean prevalence of active TB found through CXR screening strati-㰊ed by TB incidence in the country of origin as reported in the study by Aldridge et al. [25].
For each of the cost-effectiveness studies, we extracted the following data: economic methods used (e.g.micro-costing study, within-trial cost-utility analysis, Markov model), description of the case base population, the intervention and comparator, the absolute size and relative difference in resource use and costeffectiveness (e.g.incremental net bene-㰊t (INB) or incremental cost-effectiveness ratio (ICER)) [26].The certainty of economic evidence in each study was assessed using the relevant items from the 1997 Drummond checklist [27].All currencies were converted to 2015 Euros using the Cochrane web-based currency conversion tool: https://eppi.ioe.ac.uk/costconversion/default.aspx.
In the -㰊rst search, we retrieved 3,375 studies through database searching and 22 additional studies identi-㰊ed through other sources on the effectiveness of TB screening in migrant populations (Figure 2).After removal of duplicates, 2,884 studies were screened by title and abstract.A total of 127 studies underwent full text assessment.We did not identify any single study on the effectiveness of active TB screening in migrants.We therefore included seven studies that addressed the active TB screening evidence chain: the yield of detecting active TB among migrants in CXR screening programmes (n = 3) [25,28,29], the performance characteristics of CXR to detect active TB (n = 2) [30,31], the effectiveness of TB therapy in those born in the EU/EEA and the foreign-born population (n = 1) [6], and the uptake of active TB screening by migrants (n = 1) [32].In the second search, 2,856 articles were retrieved through database searching and an additional 13 articles identi-㰊ed through other resources (Figure 3).After removal of duplicates, 2,740 studies were screened by title and abstract.A total of 37 studies underwent full text assessment and three individual studies were included for analysis [33][34][35].

Yield of chest radiography to detect active tuberculosis
Three systematic reviews assessed the yield of detecting active TB among migrant populations in CXR screening programmes performed before and after arrival in the EU/EEA and low TB incidence countries outside the EU/EEA [25,28,29].The yield of active TB was heterogeneous across studies, varied by migrant type and the setting in which the screening was done and was consistently higher with higher TB incidence in the country of origin (Table 1).No overall estimates were presented but the yield increased steadily with the TB incidence in migrant source countries.The yield was 19.6 per 100,000 in migrants originating from countries with a TB incidence lower than 50 per 100,000 and 336 per 100,000 in migrants originating from countries with a TB incidence greater than 350 per 100,000 [25].The quality of the data in studies included in these three systematic reviews was very low to low (GRADE).

Accuracy of chest radiography to detect active tuberculosis
We identi-㰊ed two systematic reviews that assessed the performance of CXR to detect active TB [30,31].Van't Hoog et al. showed that CXR (presence of any abnormality) was highly sensitive (98%) and moderately speci-㰊c (75%) to detect active TB [30].Screening for active TB with symptoms alone had lower sensitivity (78%) and speci-㰊city (68%) [30].Pinto et al. also found that CXR to detect active TB was highly sensitive 95% (range: 81-100%) but less speci-㰊c 42% (range: 22-72%) [31].Focussing on the presence of upper lobe in-㰊ltrates and cavities increased the predictive value for diagnosing active TB.The certainty of the evidence of these two studies was judged to be very low (Table 1).

Numbers needed to screen
Using inputs of the yield of CXR reported by Aldridge in the pre-arrival programmes we estimated the NNS to detect one case of active TB in migrants strati-㰊ed by TB incidence in source countries (Table 2) [25].We found that the NNS decreased dramatically with increasing TB incidence in source countries and ranged from 5,076 in countries with a TB incidence between 50 and 149 per 100,000 to 298 in countries with a TB incidence greater than 350 per 100,000.The yield of active TB detection in pre-arrival CXR screening programmes for migrants by TB incidence in country of origin from Aldrige et al. [25].NNS = 1/mean prevalence of active TB found through CXR screening strati-㰊ed by TB incidence in the country of origin.

Effectiveness of active tuberculosis treatment
In an ECDC report on TB surveillance from 2007 to 2013, TB treatment outcomes were similar or better in those born outside the EU/EEA than in those born in the EU/EEA [6].Treatment success was as high in the foreign-born (for all regions of origin) compared with those born in the EU/EEA (77.4% vs 74.6%); however, their failure rates (0.2% vs 2.4%) and default rates (5.4% to 6.6%) were lower.This European surveillance data was judged to be high-quality evidence (Table 1).

Acceptability of screening
Mitchell et al. conducted a review to determine the acceptability of targeted TB screening and active case -㰊nding among vulnerable and at-risk groups and found that TB screening was well accepted by the majority of risk groups, including migrants (85%; range: 55-96%).Lower acceptability was found among persons living with HIV/AIDS and individuals in refugee camps and internally displaced persons [32].Overall, the study found that simple TB screening (at point of care) was more acceptable than referral requiring multiple visits.The evidence in this study was judged to have considerable bias (Table 1).

Cost-effectiveness of active tuberculosis screening programmes
There was very little information on the cost-effectiveness of active TB screening in migrant populations as only three studies were identi-㰊ed.These studies demonstrated that the most cost-effective CXR screening strategies were among high-prevalence groups, close contacts of those with known TB, and migrants at entry if they originated from intermediate (60/100,000) and high (> 120/100,000) TB incidence countries [33][34][35] (Table 3).The Drummond Criteria [27]: (i) Was a well-de-㰊ned question posed in answerable form?(ii) Was a comprehensive description of the competing alternatives given (i.e. can you tell who did what to whom, where and how often)?(iii) Was the effectiveness of the programme or services established?(iv) Were all the important and relevant costs and consequences for each alternative identi-㰊ed?(v) Were costs and consequences measured accurately in appropriate physical units (e.g.hours of nursing time, number of physician visits, lost working days, gained life years)?(vi) Were the cost and consequences valued credibly?(vii) Were costs and consequences adjusted for differential timing?(viii) Was an incremental analysis of costs and consequences of alternatives performed?(ix) Was allowance made for uncertainty in the estimates of costs and consequences?(x) Did the presentation and discussion of study results include all issues of concern to users?All currencies were converted to 2015 Euros using the Cochrane web-based currency conversion tool: https://eppi.ioe.ac.uk/costconversion/default.aspx.Resource use was expressed in cost per person and classi-㰊ed as low (savings or ≤ USD 1,000/person (EUR 808)), moderate (USD 1,000-100,000/person (EUR 808-80,845)) or high (USD ≥ 100,000/person (EUR > 80,845)).
Two studies demonstrated that CXR screening of migrants was cost-effective compared with no screening: Oxlade et al. determined that the ICER of CXR relative to no screening was CAD 30,000 (Canadian dollars in 2004; EUR 23,690) per case averted in migrants from intermediate TB incidence source countries, and less than CAD 1,000 (EUR 789) per case averted in the high-incidence group [35].Similarly, CXR compared with no screening in immigrants with a risk of reactivation of more than 5% was cost-effective.Dasgupta et al. reported that close-contact screening resulted in net savings of CAD 815 (EUR 758) for each active case detected and treated and of CAD 2,186 (EUR 2,033) for each future active case prevented, compared with passive case detection [34].The certainty of the evidence in these studies ranged from low to moderate (Table 1).
There were no single studies that directly addressed the overall effectiveness of active TB screening programmes on the health outcomes of migrant populations.We therefore evaluated the screening chain of evidence.The yield of detecting active TB through CXR screening of migrants was heterogeneous across studies and varied by migrant type and the setting in which the screening was done, but consistently increased with higher TB incidence in the country of origin [25,28,29].The NNS to detect one case of active TB decreased and cost-effectiveness increased with increasing TB incidence in source countries [25,34,35].CXR is a highly sensitive and moderately speci-㰊c screening tool to detect active TB [30,31].CXR screening is highly acceptable to most foreign-born populations [32].
The yield of CXR to detect active TB varied widely among migrant sub-groups in the three systematic reviews (120 to 2,340/100,000) however the overall yield (350 cases/100,000) in the post-arrival setting was consistent between studies [28,29].There was also consistency in the increase in yield with increasing TB incidence in source countries in both pre-and post-arrival setting [25,28,29].The majority of studies in the post-arrival setting were carried out in various EU/EEA countries whereas pre-arrival screening was done in migrants arriving in the United Kingdom.The wide range in yield of post-arrival screening programmes re‾䐴ects the heterogeneity of the programmes and the composition of migrants screened.Post-arrival programmes differed widely between countries with respect to timing of screening (port of arrival, in reception areas, in the community or ad hoc), countries of origin of migrants received, the type of migrants targeted (all migrants, asylum seekers only or undocumented migrants), and the threshold of TB incidence in the countries of origin at which screening was performed.Although 31 EU/EEA countries have an active TB screening programme for migrants, the absolute and attributable impact on active TB rates in those countries is unknown [37,38].Extrapolating from the impact of the wellestablished pre-migration TB programme in the US, there may be bene-㰊t of active TB screening in migrants on TB control in the host country.An evaluation of this programme demonstrated that detecting prevalent active TB before arrival in the US reduced TB noti-㰊cation rates among migrants in the -㰊rst years after arrival [39].
Higher NNS and lower cost-effectiveness with higher TB incidence in countries of origin suggests that active TB screening programmes will be most eㄊ橔cient when targeting migrant populations from high TB incidence countries.This is consistent with WHO recommendations to focus active screening on the highest risk groups [40].The heterogeneity of the estimates from these studies, however, limits the ability to provide more precise guidance on which type of migrants to target, the best timing to screen or the optimal threshold of TB incidence in countries of origin.Although screening migrants from the highest TB incidence countries is most eㄊ橔cient, the impact on TB incidence in the host country might be limited since many cases occur in migrants from countries with lower TB incidence and in migrants who entered the country many years before TB diagnosis [41,42].
Although the CXR is a good screening test for active TB and is highly sensitive (78%), con-㰊rmatory sputum culture for TB is essential to improve speci-㰊city and is the gold standard for diagnosing active TB [30,31,43].Screening for symptoms of active TB may be a reasonable -㰊rst screening tool in certain situations such as in an emergency setting with no on-site CXR facilities.These situations include the reception centres in Italy and Greece and/or when the receipt of a large number of migrants overwhelm health systems (as occurred in Europe in 2015) [8].Those with symptoms would need referral for CXR.The choice of the screening algorithm will need to be determined by the availability, feasibility and cost of the tests.
Active TB case -㰊nding in at-risk populations is an important TB control strategy as it allows for early detection and treatment, reduces individual morbidity and prevents TB spread to others.Active screening programmes are, however, limited by the fact that the yield is low (0.31-1.21%) and that they do not capture or prevent the majority of incident TB cases occurring in the EU/EEA that are primarily due to reactivation of latent TB or new acquisition during travel [13].Furthermore, the epidemiology of TB in the EU/EEA is heterogeneous.While migrants make up the majority of TB cases in low TB incidence EU/EEA countries, they make up a minority of cases in member states with higher TB incidence (Supplement 2).Screening for active TB in migrants will therefore need to be tailored to the local TB epidemiology in host countries, and the healthcare capacity in each setting [2,3].Finally, many migrant sub-groups are vulnerable and face barriers in accessing heath care and treatment in the EU/EEA [44].Addressing

Conclusions
Acknowledgements barriers in accessing care and treatment for all migrants, including the right to healthcare access for all and programmes tailored to address unique needs, will be essential to ensuring the most effective active TB screening and treatment programmes.

Study limitations
Our study was limited by the fact that we did not retrieve any studies that directly estimated the effectiveness of active TB screening and by the very limited data on the cost-effectiveness of active TB screening.The search was limited by the fact that it was conducted up until May 2016 and that we only included studies published in English or French.A recent narrative review of the effectiveness and cost-effectiveness, however, reports similar literature and -㰊ndings as our study [45].Our -㰊ndings are further limited by the quality of the original studies that were included in the systematic reviews.Study quality was low or very low, as almost all included studies were observational studies.

Evidence gaps and future directions
Robust studies on the yield of active TB screening among migrants by age group, migration type, timing of screening, threshold of TB incidence in source countries and the associated cost-effectiveness will be required to design the most effective active TB screening programmes.Additional studies are needed that determine the absolute and attributable impact of active TB programmes on TB control in low-incidence countries in the EU/EEA and the optimal threshold of incidence in source countries at which to screen.Finally, evidence on the comparative effectiveness and cost-effectiveness of different TB control strategies (active vs latent TB screening) for migrants will be required to prioritise TB control efforts for this population.
Active TB screening programmes that target migrants from high TB incidence countries will provide the highest yield and will be the most cost-effective.The heterogeneity of the estimates from the studies identi-㰊ed and the small number of studies addressing both the effectiveness and cost-effective of active TB screening in migrants limits the ability to provide precise guidance on which type of migrants to target, the best timing to screen or the optimal threshold of TB incidence in countries of origin.This highlights the need for further data to inform active TB screening programmes for migrants in the EU/EEA.
Numbers needed to screen to detect one case of active tuberculosis a b a b

Funding:
This work is supported by the European Centre for Disease Prevention and Control (ECDC); FWC No ECDC/2015/016; Speci-㰊c Contract No 1 ECD.5748.Dr Manish Pareek is supported by the National Institute for Health Research (NIHR Post-Doctoral Fellowship, Dr Manish Pareek, PDF-2015-08-102).The views expressed in this publication are those of the author(s) and not necessarily those The search terms and strategy in Ovid MEDLINE are included in Supplement 1.We also searched grey literature websites for published guidelines and reports from the US Centres for Disease Control and Prevention (CDC), ECDC, WHO and the International Union Against Tuberculosis and Lung Disease (IUATLD).We did not apply language restrictions to the search.Additional guidelines and studies were identi-㰊ed by our co-authors and through searching bibliographies of included studies.In the second search, using the search terms 'tuberculosis', 'screening', 'costs' and 'cost-effectiveness', we searched MEDLINE, Embase, the National Health Service Economic Evaluation Database (NHS EED), the Database of Abstracts of Reviews of Effects (DARE), the Tufts Medical Center Cost-Effectiveness Analysis Registry and Google Scholar for entries between 1 January 2000 and 31 May 2016.