A systematic review of BCG vaccination policies among high-risk groups in low TB-burden countries: implications for vaccination strategy in Canadian indigenous communities

A literature search of the Medline and Embase databases was conducted in January 2018, using the detailed search strategy shown in Additional file 1. The search was filtered to return only studies published after 1988, and written in English or French. Studies pertaining to BCG vaccine efficacy, TB incidence under specific vaccination policies, or BCG-associated adverse events, as well as general BCG vaccination policy guidelines in low-burden countries were included. Detailed inclusion and exclusion criteria regarding publication time frame, language, study population, study design, and reported outcomes are outlined in Table 1. The methodology of this review is presented according to the PRISMA guidelines for the reporting of systematic reviews and meta-analyses [24].

(The review protocol was not registered on PROSPERO).

The initial search returned 1109 abstracts after duplicate removal, and three additional studies were included via manual searches. The 1112 records were then screened and evaluated against the above inclusion criteria. Following 834 exclusions, full texts for the remaining 278 records were retrieved and read. Following full text review of 278 articles, 229 records were excluded (For reasons such as: not taking place in a low-burden country (n = 36), not providing information relevant to BCG vaccination policy (n = 46) (e.g. assessing healthcare provider’s knowledge of policy), or being editorials, commentaries, letters or conference abstracts (n = 52)), leaving 49 studies remaining in the final review. Of these, 32 were primary (original research) studies, 12 were policy reports, and 5 were reviews. Figure 1 shows the study screening process and provides detailed exclusion reasons. Study de-duplication, screening, and inclusion was tracked using the Covidence systematic review software (Veritas Health Innovation, Melbourne, Australia), and relevant data regarding vaccine efficacy, adverse events, future policy recommendations and implications of withdrawal were extracted in Microsoft Excel (version 14.5.5).

As studies reported varying TB incidence data and BCG vaccination policies depending on publication date, for the policy overview, national TB incidence per 100,000 population per year was calculated for each country included in the review, based on the 2016 incidence reported in the World Health Organization Tuberculosis country profile database [23], and current BCG vaccination policies were extracted from the recently updated BCG World Atlas [4]. On the other hand, information regarding vaccination policy and associated outcomes regarding TB incidence and adverse events over particular time periods was extracted from each specific included study. Based on these results, recommendations are made regarding BCG vaccination policy in the Canadian Indigenous context, specifically in Canadian Indigenous communities.

The methodological quality of the 32 primary research studies included was assessed at the study level using risk of bias assessment tools appropriate to the corresponding study design. RCTs (n = 4) were assessed using the National Heart, Lung and Blood Institute’s (NHLBI) Quality Assessment Tool for Controlled Intervention Studies [25], case-control or case-cohort studies (n = 3), were evaluated using the NHLBI’s Quality Assessment Tool for Case-control Studies [25], quasi-experimental studies (n = 5) were evaluated via the Joanna Brigg’s Institute (JBI) Critical Appraisal Checklist for Quasi-Experimental Studies [26], and cross-sectional or observational cohort studies (n = 15) were assessed using the NHLBI’s Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies [25]. Studies are classified as having low, unclear or high risk of bias for each of the quality criteria assessed and the results presented graphically. The case report was descriptively evaluated using the JBI Critical Appraisal Checklist for Case Reports [26], and modelling studies (n = 4) were assessed based on the ISPOR Principles of Good Practice for Decision Analytic Modelling in Health-Care Evaluation [27], scored depending on adherence to these principles.

Of the 49 included studies, 32 were primary (original research) studies, 12 were policy reports, and 5 were reviews. The primary studies had a variety of study designs, including RCTs (n = 4), case-control or case-cohort studies (n = 3), quasi-experimental studies (n = 5), cross-sectional studies (n = 8), retrospective or prospective cohort or observational studies (n = 7), modelling (simulation) studies (n = 4), and one case report.

Among the 12 policy reports and 32 primary studies, in terms of countries of conduct (all of which must have had a TB incidence rate of < 100 cases per million population as a criterion for inclusion), most took place in European countries (France n = 6, UK n = 6, Denmark n = 4, Sweden n = 3, Netherlands n = 2, Finland n = 2, Norway n = 2, Czech Republic n = 2, and Ireland n = 1). Further studies were conducted in Canada (n = 3), the USA (n = 3), Australia (n = 3), Saudi Arabia (n = 1), the United Arab Emirates (n = 1), and Egypt (n = 1). The remaining 4 studies spanned multiple countries (with one including the USA, Canada and Greenland¹), whilst others provided information on BCG vaccination across low-incidence countries in general. The 5 reviews focused either on the UK (n = 1), France (n = 1), or general low-incidence settings (n = 3).

Of the 32 primary studies, 16 reported on vaccine efficacy or TB incidence under different BCG vaccination policies, 15 reported on adverse events associated with BCG use, or provided other recommendations regarding vaccination policy (14 studies and one case report), and 1 study reported both [10].

A summary of policy recommendations in included study countries, by national TB incidence, is provided in Table 2. As the included studies report different policies and incidence rates depending on publication date, current policies in this summary table are extracted from the BCG World Atlas [4], which was last updated 2017, unless otherwise indicated. National incidence rates per 100, 000 population for each country were calculated based on 2016 WHO incidence estimates, whereas incidence rates in specific targeted risk groups (where available) as well as policy recommendations based on study observations are taken from individual studies included in the current review.

As shown in Table 2, incidence rates in the included study countries ranged from 0.7/100,000 population/year (in the UAE) to 9.8/100,000 population/year (in the UK) (as per the review’s exclusion criteria, only countries with incidences < 10/100,000 (i.e. < 100/1,000,000) are included). Although almost all of the 15 countries had at some point implemented a form of mass or routine vaccination previously, mass vaccination continues in only 4 of these countries today (Egypt, Saudi Arabia, the UAE, and, in some regions, Ireland). All 11 remaining countries have since moved to targeted vaccination of selected risk groups, instead of universal vaccination. It is interesting to note that whilst three of the four countries continuing mass vaccination have relatively high incidence rates in comparison to the other low-burden countries (Saudi Arabia: 9.3/100,000 Egypt: 8.6/100,000 Ireland: 6.8/100,000 population/year), the fourth, the UAE, continues mass vaccination despite having the lowest TB incidence of all 15 countries (0.7/100,000 population/year).

In countries currently implementing targeted vaccination, risk groups identified for targeting most often included children with parents from TB endemic countries, or children with a family history of TB or contact with a TB case, who had a non-reactive TST. Common reasons cited for the shift from a universal to a targeted vaccination program were the low risk of infection among the general population and thus the high number of vaccinations needed to prevent one case in the context of low-incidence settings (for example, an estimated 21,699–25,125 vaccinations needed among the Norwegian adolescent population to prevent one case) [11, 34], and the concentration of the majority of cases among a specific risk group [36]. In countries where mass vaccination was deemed to still be justified, reasons for continuing mass vaccination included the low incidence of serious adverse events associated with the vaccine [28], and findings regarding a significant correlation of BCG vaccine coverage with reduced TB incidence and TB-associated mortality rates (in Egypt) [45].

Notably, among studies in countries that have shifted from universal to targeted vaccination, all found this change in policy appropriate and did not recommend re-introduction of universal vaccination, however, possible dangers of withdrawal of mass vaccination were also highlighted. These included not only a potential rise in TB cases, (as occurred, for example, in the Czech Republic, where a rise in TB incidence was observed in the region in which mass vaccination was discontinued in 1986 [9]), but also the concern of incomplete coverage of risk groups after discontinuation of universal vaccination [35]. It was therefore generally agreed that for successful withdrawal of universal vaccination, effective identification of high-risk children and strict adherence to guidelines regarding their vaccination is needed [14, 35], and a strong TB control program (including sufficient screening and diagnostic strategies) must be in place prior to withdrawal [19, 32].

Study findings regarding vaccine efficacy and the effect of changes in vaccination policies on TB incidence are shown in Table 3 (grouped by country). Although two-arm studies comparing vaccinated to non-vaccinated groups (or areas with higher vs. lower vaccine coverage) generally report a higher incidence of TB in non-vaccinated (or low-coverage) groups compared to vaccinated (or high-coverage) groups [8, 30, 42, 55, 56]. Despite this, reported vaccine efficacy however varied widely across studies, ranging from 49% (95%CI: 14–62%) in a study of Asian children living in the UK [53] to as high as 87.5% (95%CI: 30–98%) in a study of the general French population [50]. Findings that supported targeted vaccination among selected risk groups rather than universal vaccination included a high number of vaccinations needed to prevent one TB case in low-burden areas [34], along with a potentially high number of adverse events per prevented case [52], and the persisting low risk of infection in low-incidence settings making mass vaccination unnecessary [10, 11, 13]. It was also emphasized that a further benefit of the discontinuation of mass vaccination in low-incidence settings was the restoration of the diagnostic utility of the TST for the identification of LTBI [10]; an important consideration given that detection and prophylaxis of LTBI continues to be a component of TB control in the Canadian Indigenous context [57], and because interferon-γ-release assays are not widely available [58].

Primary studies reporting adverse events are shown in Table 4. Of these, one reported BCG-associated osteomyelitis [10], three reported BCGitis [10, 44, 63] (including one case study in a French 4-month old [63], not shown in table), and 3 reported lymphadenitis [9, 37, 41]. Although the occurrence of serious adverse events as a result of vaccination was generally rare across studies, it was found that the use of unregistered vaccine strains was associated with a higher incidence of adverse events in comparison to registered vaccines, suggesting that particularly during vaccine shortages, when the implementation of unregistered vaccines is increased, improved surveillance and management of possible adverse events may be needed [37, 38].

Regarding non-specific effects of the BCG vaccine, studies investigated its association with the development of inflammatory bowel disease (IBD) [40], child psychomotor development [39], and the incidence of childhood non-TB infections [59], allergic disease (such as asthma and eczema) [61], and atypical mycobacterial disease [62]. No significant association was found between BCG use and child psychomotor development [39] or incidence of IBD [40], non-TB infections [59], or asthma [61], whilst there was a lower incidence of atypical mycobacterial disease [62] and lower use of medication for eczema [61] in groups receiving BCG compared to those who did not.

A summary of recommendations from the 12 included policy reports is provided in Table 5. In general, mass vaccination was not recommended in low-incidence countries, where instead, TB control strategy should focus on the identification and rapid treatment of active cases as well as the control of LTBI [29]. In addition, it was emphasized in multiple reports that readiness for withdrawal necessitates a strong TB surveillance system to allow timely assessment of policy consequences [6, 19, 32], and comprehensive coverage of the selected high-risk groups should be ensured [65]. Re-vaccination was not recommended in any report, due to the lack of evidence for its efficacy [6, 32, 36, 65].

Among the 5 included reviews, the 3 that made recommendations for general low-incidence settings all concluded that a universal vaccination policy is of limited value in low-incidence countries, with targeted vaccination among high-risk groups being recommended instead [5, 67, 68]. The 2 remaining reviews focused specifically on the UK and France, and were both published prior to the discontinuation of mass vaccination in both countries, with the French review arguing that mass vaccination at the time (2003) continued to be justified based on the number of incident TB meningitis cases, although a precise cut-off for the number of cases at which vaccination would no longer be considered beneficial was not provided [69], whilst the British study stated that (as of 1988), routine vaccination of 10–14 year olds within British school programs continued to be justified, but that this policy is subject to revision in future years [70]. As previously mentioned, mass vaccination policies in both countries have since been revised, in favour of the adoption of targeted vaccination [12, 13].

The risk of bias in included RCTs (n = 4) was assessed using the National Heart, Lung and Blood Institute’s (NHLBI) Quality Assessment Tool for Controlled Intervention Studies [25]. The included RCTs were generally found to have a low risk of bias, as the risk of cross contamination (i.e. vaccination occurring in the non-vaccinated group) was low, sample sizes were justified, and group allocation was appropriately randomized in most studies. Potential sources of bias in the included RCTs however include the fact that, due to the nature of the intervention (i.e. administering the vaccination) study staff could not be blinded to group assignment, and group allocation could not be concealed from participants. Risk of bias assessment results for the included RCTs is shown in Fig. 2.

Observational and cross-sectional studies (n = 15) (assessed via the NHLBI’s Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies [25]) were also generally of high methodological quality, meeting most of the quality assessment criteria. However, lack of blinding of outcome assessors (or lack of reporting thereof) was a potential source of bias in most studies, and there was, in most studies, inadequate or unclear reporting of adjustment for confounders, as shown in Fig. 3 below. The included case-control and case-cohort studies (n = 3) (evaluated using the NHLBI’s Quality Assessment Tool for Case-control Studies [25]) also insufficiently reported blinding of outcome assessors to the status of participants (as cases or controls), although other sources of bias were appropriately addressed in all studies, such as providing clear case definitions and differentiating them clearly from controls. The risk of bias assessment for case-control studies is shown in Fig. 4.

In the case of quasi-experimental studies (n = 5), which were evaluated using the JBI Critical Appraisal Checklist for Quasi-Experimental Studies [26], considerable sources of bias were the fact that most studies did not provide sufficient information regarding participants’ baseline characteristics and whether there were any significant differences in these characteristics in the exposed vs. unexposed groups, as well as insufficient or no comparison between the characteristics of participants lost to follow-up compared to those who completed the study. Figure 5 displays the risk of bias assessment for the included quasi-experimental studies, and individual study scores for each quality assessment criterion in Figs. 2, 3, 4, 5 are provided in Additional file 2.

The case report [63] was evaluated using the JBI Critical Appraisal Checklist for Case Reports [26], and was found to have a low risk of bias, as it clearly described the patient’s demographic characteristics and clinical history, as well as the patient’s current condition, the administered diagnostic tests and the treatment outcome.

Lastly, the 4 included modelling studies were assessed using the ISPOR Principles of Good Practice for Decision Analytic Modelling in Health-Care Evaluation [27]. All modelling studies used a model structure that was consistent with the nature of the disease, and in the case of Markov models (n = 2), health states relevant to the disease were included and the probabilities of transition between these states took into account patient history and treatment. However, most did not clearly state that their model inputs were derived from a comprehensive review of the literature, and only 2 conducted sensitivity analyses [30, 51]. In addition, none provided information on whether or where the source code for the model is accessible for peer-review.

Given the variation in its reported efficacy and the high number of vaccinations needed to prevent one TB case in a low-incidence setting, there was general consensus among the reviewed studies that the replacement of universal screening with targeted screening of high-risk groups is justified in low-burden countries [5, 10, 11, 67, 68]. Such targeted screening programs however require establishing precise guidelines or incidence cut-offs on which to base the categorization of high-risk groups. In this regard, it is relevant to note that a recent study among Canadian, Alaskan and Greenland Indigenous populations found that population-based vaccination of infants, along with screening and treatment for LTBI, was significantly associated with a decrease in TB incidence [55], which suggests that population-level vaccination in high-risk communities as a whole, rather than vaccination based on individual risk factors, may be beneficial in the Canadian Indigenous community context. In terms of the strength of this evidence however, any causal inferences should be made with caution in the case of population-based studies such as this, given the uncertainty in individual exposure. The benefit of population-level vaccination however is supported by findings of a simulation study of optimal BCG vaccination strategies among children in low-incidence countries, which concluded that community-wide vaccination is beneficial in scenarios of communities with a prevalence of approximately 30 smear-positive TB cases / 100,000 [52]. The strength of this evidence is judged to be considerable, given that this modelling study adhered to most of the quality assessment criteria in the aforementioned ISPOR guidelines for decision-analytic models (see Table 6). Therefore, as the Canadian Indigenous population has a similar incidence of TB (23.5/100,000) [18], community-level vaccination may be beneficial in this setting.

Given the findings of these two studies, it may therefore be of interest to consider community-level incidence as an indicator of the potential utility of targeted BCG vaccination in any particular Indigenous community. With this in mind, according to the Canadian Immunization Guidelines (based on those by the International Union Against TB and Lung Disease [71]), although routine vaccination is not recommended in any Canadian population, risk groups in which vaccination can be considered include infants in First Nations or Inuit communities with an average annual incidence of smear-positive pulmonary TB > 15 per 100,000 population over the past 3 years, or infants in communities with an annual risk of TB infection > 0.1% [70]. .Unfortunately however, as noted by a previous policy report on BCG vaccination in Canadian Indigenous communities, due to the small population sizes of First Nations communities, incidence rates easily fluctuate above or below specific incidence cut-offs provided in vaccination guidelines, and therefore may make these of limited use in determining appropriate vaccination policy [33]. In addition, it is relevant to note that re-introduction of BCG vaccination in a given community will limit the value of the TST as a useful screening tool for LTBI [10], which is especially relevant in the Canadian indigenous setting, where screening for LTBI remains a component of the ongoing TB control strategy and access to interferon-γ-release assays is limited [58].

Additional considerations include that if routine BCG vaccination is ultimately re-introduced to Canadian Indigenous communities, based on the findings of studies in the current review, it is recommended that the age of vaccination be increased from at birth to 6 months of age, given that higher protective efficacy has been shown with vaccination at 6 months compared to before 6 months of age (63% vs. 42% efficacy, respectively), (in a study with moderate strength of evidence due to a small sample size, but otherwise having low risk of bias) [48] and that vaccination at 6 months rather than at birth allows sufficient time for the identification of any potential underlying immune-deficiencies that could result in disseminated BCGitis following vaccination (in a study with a large sample size (n = 139,000) and generally low risk of bias (see Fig. 3) [44]. This decision would however need to be considered in light of the logistical challenges of conducting TST testing prior to offering BCG at 6 months, given the need for follow-up 48 to 72 h after TST administration.

The heterogeneity of the included studies, in terms of study design, study populations, and outcome measures, represents a limitation of this review, given that it precluded the conduct of a meta-analysis. In addition, although included studies generally had a low risk of bias, due to the overt nature of the intervention (receiving the BCG vaccine), many primary studies could not take measures to blind outcome assessors to participants’ intervention assignment (vaccinated or non-vaccinated). A further limitation of this review and the resulting policy recommendations includes that although the findings of the included studies are potentially relevant to the Canadian Indigenous community context in so far as they focus on high-risk populations in low-incidence settings, Canadian Indigenous communities represent a unique setting with additional challenges including their remoteness and vast, sparsely-populated areas. These unique characteristics therefore may make some of the findings regarding optimal vaccination policies in other high-risk groups in low-incidence countries less applicable to the Canadian Indigenous context (for example the aforementioned limited applicability of incidence-based vaccination policies in the context of small populations with easily fluctuating incidence rates).

Where BCG vaccination is implemented, delivery strategies and potential barriers to achieving adequate coverage should be considered, which is particularly relevant in Canadian Indigenous communities and other hard-to-reach settings. Furthermore, given that vaccination at birth has been shown not to allow sufficient time to identify possible immune-deficiencies in infants, consequently putting them at risk for disseminated BCGitis, increasing the age at which BCG is administered from at birth to at 6 months of age should be considered (although keeping in mind the logistical challenge of then needing to administer a TST prior to vaccination at 6 months) [19, 44]. In addition, in the context of the varied history of BCG vaccination in Canada and other low-incidence countries, and potentially ongoing differences in implementation across communities, education of healthcare providers regarding the interpretation of positive TSTs is needed in order to ensure continued adequate identification and subsequent treatment of LTBI.

The withdrawal of universal BCG vaccination, even when replaced with a targeted vaccination policy, may have adverse implications, including a potential increase in TB incidence. This is particularly true if the high-risk groups recommended for vaccination under a targeted program are not effectively identified and subsequently vaccinated. This implication is particularly alarming considering that 3 studies reported subsequent incomplete vaccine coverage of high-risk groups following withdrawal of mass vaccination [14, 35, 60], (Table 4) highlighting potential negative implications of discontinuation of universal programs, and the need to improve strategies to identify and reach the selected at-risk groups when shifting to a targeted vaccination policy. Therefore, in cases where an existing vaccination program is discontinued, it is imperative that an effective TB surveillance system is in place and well established before the withdrawal of the program, and that case-finding, screening, and diagnostic efforts are strengthened in order to ensure continued TB control [19, 20]. This is particularly relevant to highlight in reference to the Canadian Indigenous setting, where significant barriers persist regarding surveillance, screening, and diagnostic efforts in remote and under-served communities.