Screening for tuberculosis infection and effectiveness of preventive treatment among people with HIV in low-incidence settings

Objective: To determine the yield of screening for latent tuberculosis infection (LTBI) among people with HIV (PWH) in low tuberculosis (TB) incidence countries (<10 TB cases per 100 000 persons). Design: A systematic review and meta-analysis were performed to assess prevalence and predictive factors of LTBI, rate of TB progression, effect of TB preventive treatment (TPT), and numbers needed to screen (NNS). Methods: PubMed and Cochrane Library were searched for studies reporting primary data, excluding studies on active or paediatric TB. We extracted LTBI cases, odds ratios, and TB incidences; pooled estimates using a random-effects model; and used the Newcastle–Ottawa scale for bias. Results: In 51 studies with 65 930 PWH, 12% [95% confidence interval (CI) 10–14] had a positive LTBI test, which was strongly associated with origin from a TB-endemic country [odds ratio (OR) 4.7] and exposure to TB (OR 2.9). Without TPT (10 629 PWH), TB incidence was 28/1000 person-years (PY; 95% CI 12–45) for LTBI-test positive versus 4/1000 PY (95% CI 0–7) for LTBI-test-negative individuals. Among 625 PWH (1644 PY) receiving TPT, 15 developed TB (6/1000 PY). An estimated 20 LTBI-positive individuals would need TPT to prevent one case of TB, and numbers NNS to detect LTBI or prevent active TB varied according to a-priori risk of LTBI. Conclusion: The relatively high prevalence of LTBI among PWH and the strong correlation with origin from a TB-endemic country support risk-stratified LTBI screening strategies for PWH in low-incidence countries and treating those who test positive.


Introduction
Globally, tuberculosis (TB) is a leading cause of death in people with HIV (PWH).TB preventive treatment (TPT) effectively prevents progression to TB disease among PWH [1], especially among those who screen positive for latent tuberculosis infection (LTBI) [2].Therefore, the WHO recommends TPT for all PWH who screen positive, after exclusion of TB disease [3].In low TBincidence countries, testing for LTBI using tuberculin skin test (TST) and/or interferon gamma release assay (IGRA) is usually advised, often selectively to PWH deemed at higher risk such as immigrants [4].However, a survey among low TB burden countries reported that only 75% had a national policy on LTBI, and that 66% provided LTBI testing and treatment for PWH [5].The potential benefit of LTBI screening and TPT for PWH in low TBincidence settings remains contentious [6], and adherence to these guidelines is low in many low-incidence countries such as the Netherlands [7], Belgium [8], the UK [9] and New Zealand [10].One reason for this is that clinicians feel that PWH are at low risk for TB once effective antiretroviral therapy (ART) is started [7,11,12].With the introduction of early and effective ART, the risk of TB has probably gone down over the years [13,14].Still, also in TB low-endemic countries, many PWH present late with low CD4 þ counts, and individual studies suggest that the risk of TB, even with early ART and proper immune reconstitution, remains higher for PWH than for people without HIV [15].As such, the yield and outcome of LTBI screening and TPT for PWH in these settings is uncertain and it remains unclear what the benefit of screening and treatment of LTBI in HIV-infected patients in these settings is [6].To address this knowledge gap, we conducted a systematic review and meta-analysis aiming to establish the LTBI prevalence and associated risk factors among PWH in low TB-incidence settings, and assess risk of progression to TB, and effectiveness of TPT.

Methods
Search strategy and selection criteria Three separate research questions were addressed.Firstly, we wanted to establish the LTBI prevalence and associated risk factors among PWH in low TB-incidence settings (category: general, prevalence, predictive factors).Secondly, we wanted to assess the risk of progression to TB in LTBI in PWH in low TB-incidence settings (category: disease progression).Thirdly, we wanted to determine the effectiveness of TPT in LTBI in PWH on the progression to TB (category: prophylactic therapy).This third research question corresponds to the following PICO format.P: PWH with LTBI, I: TPT, C: no TPT, O: progression to TB.
We searched PubMed and Cochrane Library databases for relevant peer-reviewed studies (search performed 6 October 2020, rerun 9 December 2021).No filters were applied, and the full search strategies for the separate questions -indicated by corresponding category -are available in the supplementary material (Supplementary Table 1, http://links.lww.com/QAD/D2).References of relevant articles and guidelines were also checked.Studies were included if: full text was available in English or Spanish; the article reported primary data; the study was conducted in a low TB-incidence setting (<10 TB cases per 100 000 persons) [1]; if TST and/or IGRA data were reported; and if at least four PWH had been included in follow-up.Studies about active or paediatric TB were excluded.Search criteria and methods used are described in the research protocol (Supplementary Table 2, http:// links.lww.com/QAD/D2).

Data abstraction and quality assessment
Data were abstracted by one reviewer and included author, year, study design, setting, sample size, LTBI prevalence, method used to diagnose LTBI, demographic characteristics of the study population and HIV-related factors such as blood CD4 þ cell count and ART status, person years (PY) of follow-up, and diagnosis of TB during follow-up.Quality assessment of data extraction was performed by two other authors.PRISMA-guidelines were followed for analysis and presentation of results.Quality assessment was performed using a tailored Newcastle-Ottawa scale to classify studies as low quality (<4 stars), moderate quality (4-5 stars), or high quality (>5 stars) (Supplementary Table 3, http://links.lww.com/QAD/D2)Data analysis and statistics STATA version 16 was used for data processing, analyses, and producing figures.Heterogeneity was assessed for all meta-analyses using I 2 statistic [16].As there was substantial heterogeneity, all meta-analyses were carried out using a random-effects model.A funnel plot was used to assess small study effects.To assess LTBI prevalence, proportions of positive test results and total participants and their 95% confidence interval (CI) were pooled and presented in a forest plot with sub-analyses based on year of publication and originating continent.For each individual potential predictive factor, odds ratios and their 95% CI were calculated by test result or abstracted from the articles whenever available.Log-transformed odds ratios were pooled independently for every factor and back transformed.P less than 0.05 was used as a cutoff for significance.
Incidence of active TB during follow-up was assessed according to TST/IGRA test result and TPT.Incidence was expressed as the number of cases divided by personyears of follow-up whenever available, otherwise mean or median time of follow-up was used.To estimate progression rates, incidence rates of TB disease for each group were calculated per study.Subsequently, the incidence rate differences and ratios between these groups were calculated, which were then pooled with their 95% CI.Due to zero event studies, a continuity correction was applied.Studies with no events in either group were excluded from meta-analysis on incidence rate difference and ratio.
We calculated the number of individuals with a positive LTBI test that would need TPT to prevent one case of TB [number needed to treat (NNT)] as the inverse of the absolute risk reduction (ARR) assessed by our metaanalysis on the effectiveness of treatment [17].Using pooled prevalence data, we also calculated the number needed to screen (NNS) to detect one case of LTBI stratified for predictive factors identified as described earlier.Finally, by dividing the NNT by the pooled prevalence, we estimated the NNS and NNTof PWH to prevent one case of TB, assuming those not screened would remain unaware of their LTBI status and all those with a positive test result would receive TPT.

Search results
After removal of duplicates, the search on the prevalence and predictive factors for LTBI among PWH in low TB-incidence settings yielded 1030 articles that were screened by title and abstract; 72 publications remained for fulltext assessment and 51 were included in the analysis (Fig. 1).To assess TB disease progression, 497 articles were identified and screened by title and abstract, 35 were assessed by full-text and a total of seven were included in analysis.To examine the effectiveness of TPT, 928 articles were screened by title and abstract, 30 were eligible for full-text assessment and five were included in the analysis.

Prevalence of latent tuberculosis infection and predictive factors for a positive test result
To assess LTBI prevalence, we included 39 cohort studies [10, and 12 cross-sectional studies [56][57][58][59][60][61][62][63][64][65][66][67], with a total of 65 930 PWH, as summarized in Table 1.Studies were published after 1991; they were conducted in Canada, the United States, Australia, New Zealand, and several Western and Southern European countries.These studies were aiming to either compare diagnostic test performance [43,48,63], provide data on TB and LTBI during routine medical check-ups [22,24,39,68], estimate the prevalence of LTBI [44], or evaluate LTBI screening and treatment [41].Most studies excluded TB disease, either by clinical assessment [28], sputum smear and  The decision on inclusion of low-incidence countries (<10 TB cases per 100 000 population) was based on the year of publication of the study.In some countries, for example, the United Kingdom, the TB incidence was greater than 10/100 000 at the time of study but had progressed to the low-incidence state at the time of publication.We kept these studies to allow sufficient number of publications for analysis.
The median age of patients ranged from 36 [58,68] to 46 [10,63] years.Most studies had an overrepresentation of PWH originating from TB-endemic countries, varying from 22% [53] to 72% [43].Also, PWH with low CD4 þ cell counts were more likely to be screened for LTBI [31,68].LTBI screening was performed less often among those on successful ART [68].Median baseline CD4 þ cell counts, reported in 16 of 22 studies, ranged from 272 [23] to 648 [56] at enrolment; six studies reported proportions of those with less than 200 CD4 þ cells, ranging from 25% [21] to 48% [45].Median plasma HIV-RNA (reported in 14 studies) ranged from less than 50 copies/ml [29,53] to 4.9 log 10 copies/ml [31].Use of ART, reported in 14 studies, ranged from 0% [22] to 86% [44], but only 10 studies reported the proportion of patients with virological suppression, ranging from 0% [48] to 72% [44].Eight out of 17 studies with data on date of HIV diagnosis had only included newly diagnosed patients; in the other studies, time between HIV diagnosis and LTBI screening ranged from less than 3 years [45] to a median of 8.6 years [42].Further systematic analysis of LTBI screening in relation to time of HIV diagnosis, CD4 þ cell count, ART use, and plasma HIV-RNA was not possible as individual data were not reported.
Two studies found a significantly higher risk of TB disease for PWH with a low CD4 þ cell count [39,69].One study reported a relatively low median CD4 þ cell count (208 cells/ml) at the time of TB diagnosis [68].The proportion of patients on ART at enrolment, reported in four studies, varied, and ranged from 40.8% [68] to 100% [69].Five studies found that disease progression was significantly more common in PWH not on ART and/or with detectable plasma HIV-RNA [18,29,48,68,70].However, specific analysis of TB risk according to CD4 þ cell count or use of ART could not be done in the absence of individual data.

Effectiveness of tuberculosis preventive treatment
Our literature search identified four cohort studies [23,39,43,68] and one randomized controlled trial [71], which reported on the effectiveness of TPT in PWH with a positive LTBI test (Supplementary Table 5, http://links.lww.com/QAD/D2).In these studies, published between 2007 and 2014, a total of 625 test-positive PWH (range: 67-411 for individual studies) who received TPT were followed for 1020 person-years (range: 78-648 person-years), while 469 test-positive individuals (range: 15-246) who did not receive TPT were followed for 1423 person-years (range: 50-1005 person-years).One study lacked a control group [72] and three studies reported no cases in the treatment group [23,43,68] of which one [47] had no TB cases in either arm during 128 person-years of follow-up, so a continuity correction was applied for these groups.The study without TB cases in either group [43] and the one without a control group [72] were excluded from metaanalysis on incidence rate difference and ratio.
Incidence rates between individual studies ranged from 0 to 17.5 cases per 1000 person-years for those who received TPT, and from 0 to 183.5 for those who did not.The pooled incidence rate derived from random-effects meta-analysis in the TPT group was 6 cases per 1000 person-years (95% CI 5-7 Subtotal (I^2 = 96.13%,p = 0.00) Sester

Discussion
Our systematic review and meta-analysis revealed that one in eight PWH living in low TB-incidence countries who were screened for LTBI tested positive.LTBI prevalence was strongly associated with origin from a TBendemic country, sub-Saharan African ethnicity, and close TB contact.Second, our study showed that the risk for TB progression was seven times higher among PWH with a positive LTBI test, while risk of TB appeared to be lower among individuals with higher CD4 þ cell counts and those using ART.Third, TPT effectively prevented progression to TB in PWH with LTBI.These findings suggest that a risk-stratified approach of LTBI screening among PWH in low TB-incidence countries, with TPT for all those testing positive, may be appropriate in these settings.
LTBI prevalence varied between studies, most likely because of differences in epidemiological background of PWH included, and differences in LTBI screening strategies and tests.It is well known that TST and IGRA lack concordance, and both poorly reflect the risk of progression to TB [73].Furthermore, test sensitivity is reduced in advanced HIV disease and low CD4 þ cell counts, with more frequent indeterminate IGRA tests reported [74].Positive LTBI tests were more common among PWH from TB-endemic countries or sub-Saharan Africa, as noted previously [75], though few studies stratified results among foreign-born according to TB burden in the country of origin [18].
We found that the risk of TB in PWH was seven times higher among individuals with a positive LTBI test than in those with a negative test.As the overall incidence appeared approximately 100-fold higher than the TB incidence in low-incidence countries (<10 cases per 100 000 persons), screening appears favourable in this population.The TB risk for PWH on ART may be lower nowadays as current international guidelines recommend PWH to start ART shortly after diagnosis of HIV, irrespective of their CD4 þ cell count [15].In an analysis that also included data from medium TB-incidence countries (defined as <100/100 000 person-years) the incidence appeared to have decreased more recently [76].
Although early and effective ART preserves and restores anti-TB-specific immune responses in PWH, the risk of progression from LTBI to TB remains higher than in people living without HIV [48].
We found TPT to be very effective in PWH with LTBI as it led to an approximately 90% reduction in TB incidence.Numbers of PWH NNS to detect one case of LTBI and numbers NNT to prevent one case of TB varied between studies and the NNS and NNT derived from our metaanalyses should be interpreted with caution because of the pooling of data [77].As our analysis was based on a metaanalysis of only three studies regarding TPT, all reporting very few TB cases and relatively short follow-up, confidence intervals in our analysis were large, which demonstrates another limitation.Also, for our estimations, we assumed that all test-positive individuals would receive TPT, although a recent cascade-of-care analysis reported that among LTBI test-positive PWH in lowendemic settings, a pooled 86.3% initiated TPT [12].Lastly, adherence and TPT completion was not evaluated in all studies, which could also result in an underestimation of the effectiveness of TPT.
Our study was limited by the heterogeneity between studies, inclusion of small studies, selective inclusion of patients, and a lack of individual patient data on ART, CD4 þ cell count, and other determinants in included studies.Included studies suffered from selection bias, with an overrepresentation of individuals at increased risk for LTBI (because of migrant status or social determinants, and TST more often performed in these groups) and likely overestimation of LTBI prevalence among PWH in low TB-incidence settings [10,18,23,68].Misclassification because of self-report may have led to inappropriate inclusion of individuals previously treated for TB or LTBI.Higher anergy rates in older studies may reflect suboptimal uptake of ART or poor control of HIV, which would underestimate the LTBI prevalence.The meta-analysis was influenced by sparse data bias because of low TB incidence in these settings, which makes it harder to assess risk factors of TB progression.Moreover, the time between HIV diagnosis and assessment of LTBI, and the time to TB diagnosis was not always available, which also complicates interpretation.Due to applying a continuity correction for studies that had zero events in one group in combination with low incidence rates in other groups, we might have overestimated the rate of disease progression and underestimated the effectiveness of preventive treatment.
Our results show that TB remains more common among PWH than the general population in low TB-incidence settings and that TPT can effectively prevent TB.However, it should be noted that this result may be overestimated because of the inclusion of those more at risk for TB and by inclusion of countries that used to have a higher TB incidence during the study period compared with now.Our findings suggest that targeted LTBI screening among PWH in low TB-incidence countries could be more efficient and cost-effective than current strategies, as suggested by a recent interventional study [78].Certain groups among PWH could be prioritized for screening, such as those from TB-endemic countries or with a history of exposure.This should be accompanied by interventions to increase uptake of TPT for people with positive LTBI test results, especially among those deemed at the highest risk for developing TB [79].
Current guidelines vary between countries, and adherence to these guidelines is often poor [4,[7][8][9][10][11].Some low TB-incidence countries have adopted the WHO recommendation to screen all PWH for LTBI, whereas others use more selective screening approaches [6,80,81].However, nowadays people with newly diagnosed HIV usually immediately start potent ART, resulting in faster restoration of CD4 þ counts, leading to lower TB risk (irrespective of LTBI status) but with still significantly increased TB incidences [15,82].Future studies should establish the effectiveness and cost utility of targeted LTBI screening among PWH in low TB-incidence countries where there is widespread ART use; and promote strategies to improve uptake and completion of TPT among those with positive LTBI tests.

Table 1 .
Individual studies included.
a Range or IQR not reported.bDone according to the STARD-checklist, with the following classification: reporting less than 10 items as low quality, 10-16 items as moderate quality and more than 16 as high quality.NA, not applicable.

Table 2 .
factors associated with a positive latent tuberculosis infection test result.American in US studies and participants with origin in sub-Saharan Africa in European studies.
a Africanb Derived from random-effects meta-analysis.c TB contact or previous TB exposure.

Table 3 .
Tuberculosis incidence according to baseline TST or IGRA test result.
CI, confidence interval; IGRA, interferon gamma release assay; PWH, people with HIV; PY, person-years; TST, tuberculin skin test.aDerived from random-effects meta-analysis.bIncidencerate ratio is not expressed per 1000 PY, but as a ratio.