Individual and clinical variables associated with the risk of Buruli ulcer acquisition: A systematic review and meta-analysis

Background Buruli ulcer (BU) is a necrotizing skin disease, caused by Mycobacterium ulcerans, with poorly understood acquisition risk factors. This review aims at evaluating the importance of individual–sex, age, family ties with history of BU, gene variants–and clinical–Bacillus Calmette-Guérin (BCG) immunization, Human Immunodeficiency Virus (HIV) infection–variables in this process. Methods A systematic review was performed considering the following databases: ClinicalTrials.gov, Cochrane Controlled Register of Trials (CENTRAL), Current Contents Connect, Embase, MEDLINE, SciELO, Scopus and Web of Science. Eligible studies were critically appraised with The Joanna Briggs Institute checklists and heterogeneity was assessed with Cochran Q-test and I2 statistic. Published demographic data was descriptively analysed and clinical data pooled within random-effects modelling for meta-analysis. Results A total of 29 studies were included in the systematic review. Two randomized controlled trials (RCTs) and 21 case-control studies were selected for meta-analysis. Studies show that BU mainly affects age extremes, more preponderately males among children. Data pooled from RCTs do not reveal BCG to be protective against BU (odds ratio (OR) = 0.63; 95% CI = 0.38–1.05; I2 = 56%), a finding case-control studies appear to corroborate. HIV infection (OR = 6.80; 95% CI = 2.33–19.85; I2 = 0%) and SLC11A1 rs17235409 A allele (OR = 1.86; 95% CI = 1.25–2.77; I2 = 0%) are associated with increased prevalence of the disease. No definite conclusions can be drawn regarding the influence of previous family history of BU. Discussion While available evidence warrants further robustness, these results have direct implications on current interventions and future research programs, and foster the development of more cost-effective preventive and screening measures. Registration The study was registered at PROSPERO with number CRD42019123611.


Introduction
Buruli ulcer (BU) is a neglected tropical chronic skin disease caused by Mycobacterium ulcerans. The archetypal BU case begins with a painless small nodule or ulcer, less commonly a plaque or edema [1]. Over a period of weeks to months, these lesions may progress to extensive ulcerative lesions, which in some circumstances can afflict the underlying bone [1]. Due to its multiple presentations and the prospect of confounding superinfections, WHO recommends, without compromising the beginning of antibiotherapy, that the diagnosis of BU should include at least one, ideally two, of the following laboratorial methods: a) direct smear examination of acid-fast bacilli (AFB) with Ziehl-Neelsen from a swab or a biopsy; b) histopathology; c) culture on Löwenstein-Jensen medium at 32ºC; and d) Polymerase Chain Reaction (PCR) targeting the IS2404 insertion sequence [2]. The latter is currently considered the gold standard test, mostly due to its higher sensitivity and specificity (>90%), despite not being available in all laboratories [2].
BU has been reported in 34 countries worldwide [3,4]. As of 2018, there were two major foci of the disease, one mainly involving sub-Saharan African countries Ghana, Nigeria, Ivory Coast and Benin, and the other located in Australia, accounting for a total of 1892 cases, more than half of all recorded cases that year [3]. However, although there is a degree of consensus regarding the role of some environmental and behavioral factors associated to the disease, the literature has been less clear regarding the individual determinants of BU [5][6][7]. In fact, the only systematic reviews that have covered the intrinsic and clinical variables are either almost a decade-year old or largely inconclusive, prompting for an update on the state of the art [7,8].
Very little is known on the clinical features that predispose to BU. Sociodemographic variables, such as sex or age, have repeatedly failed to be associated to changes in the incidence of the disease, although some studies reported a higher number of cases amongst children and older adults, as well as a varying predominance of sex across the difference age groups [7,[9][10][11]. Interestingly, while the reasons for such differences are still elusive, some studies have implicated BU as having a genetic component associated to its development. Whilst one group found that previous family history of BU significantly increased the odds for new cases in the family, other authors reported associations of alleles in SLC11A1, IFNG, amongst others, with increased chances of contracting the disease [12][13][14][15]. However, not only are many of these findings still pending validation, but others authors have also reached opposite conclusions, thus remaining the question of what role does genetics play in BU [13,[16][17][18].
Likewise, the impact of immunological determinants, such as Bacillus Calmette-Guérin (BCG) vaccination or Human Immunodeficiency Virus (HIV) co-infection, on the development of BU is also a matter of uncertainty [19]. A recent meta-analysis concluded that although the original randomized controlled trials (RCTs) testing BCG vaccination showed a protective role for the vaccine, later case-control studies did not confirm such effect [8]. HIV co-infection, on the other hand, appears to promote the clinical manifestation of BU, but the low prevalence (<2-3%) of HIV-infected patients in many of the BU-endemic countries undermines the statistical power of the analyses and the reliability of the results [16,20,21].
Hence, considering the ambiguity governing the importance of the above-mentioned individual factors in the acquisition of BU, a systematic review of the available body of literature was herein undertaken. Eight different medical and scientific databases were screened for experimental and observational studies worldwide, from which data was extracted for metaanalysis whenever applicable. Critical analysis of such information is expected to yield valuable information on the understanding of the risk factors for BU development, with potential implications for future studies and intervention measures.

Methods
This study complies with The Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement (PRISMA; S1 Checklist) and its protocol is published and accessible at PROSPERO with the number CRD42019123611.

Eligibility criteria
For the present systematic review and meta-analysis, data derived from human RCTs, casecontrol and cohort studies were considered, without any temporal or language constraints. Studies were selected if they met all of the following criteria: (1) patients with BU from any part of the world, excluding studies restricted to specific age groups or presentations of the disease to avoid selection biases, as well as RCTs with recurrent cases of BU to avoid immunological memory confounding; (2) analysis of variables age, sex, BCG vaccination, HIV status, family history of BU and gene alleles, with exclusion of methodological studies and studies mainly focusing on outcomes of the disease process, subjective perceptions or behaviors of individuals and environmental risk factors; (3) primary outcome being the acquisition of BU.
(1900 -present) was last performed on February 1, 2019, following the structure presented in the supporting information (S1 Text).

Study selection
Eligibility assessment was performed independently by JF and NS according to the abovementioned criteria, with disagreement between the two authors being reviewed by PT. Titles and abstracts regarded as inconclusive were considered for full text inspection. Studies were included only when all the eligibility criteria were met.

Data collection process and summary measures
Extraction was performed by JF, NS and AGF and information independently cross-checked among these authors, namely: sample size, frequency and incidence metrics on individual characteristics of participants-age, age group, sex and genetic alleles-, as well as on the number of individuals vaccinated with BCG, infected by HIV or with previous family history of BU. Whenever the frequency of any group could not be calculated from the available data, corresponding authors were asked, via e-mail contact, to provide the necessary information.

Statistics
Meta-analyses were conducted using Review Manager 5.3 and MetaGenyo [22,23]. Heterogeneity was tested with the Cochran Q-test and I 2 statistic, calculated as I 2 = [(Q-degree of freedom)/Q] × 100, where Q is the Cochran's statistic. I 2 values of 30, 50 and 75 represent low, medium and high heterogeneity, respectively [24]. In the presence of high heterogeneity, subgroup analysis with laboratory-BU confirmed cases and/or low risk bias studies was attempted. Random-effects model (the Restricted Maximum-Likelihood method) was used to calculate the summary of pooled frequency estimates [25].

Risk of bias across studies
Studies were independently classified by JF and NS as to the risk of biases, using the adequate critical assessment tools developed by the Joanna Briggs Institute [26]. Whenever disagreements between authors were found, a third call to PT was solicited. In addition, meta-analyses comprising at least 10 studies were checked for the presence of potential publication bias through the visual inspection of funnel plot asymmetry.

Study selection
The study selection process is illustrated in Fig 1. The initial search yielded a total of 3575 records, of which 2238 were initially removed using the Zotero duplicate detection tool [27], followed by a manual independent screening by JF and NS. During the title and abstract screening, 1240 references were excluded mainly due to: a) being a non-original study or having a non-eligible design; b) not addressing Mycobacterium ulcerans infection; c) focusing on other outcomes than the acquisition of BU; d) restricting to specific age groups or forms of the disease. Ninety-seven references were selected for full-text review, at which stage 68 studies were additionally excluded mainly due to not having a compatible study design or providing relevant information for the goal of this review. Contacting of authors was attempted for the retrieval of seven studies, but at the moment of manuscript submission these were still not available.

Study characteristics and demographic characterization
Demographic characterization of the 29 eligible studies is presented in Table 1. These account for a total of 9265 BU cases, with a generally well-balanced male:female ratio (average of studies: 0.99). Following this line, five of the included sex-unmatched case-control studies [11,16,[28][29][30] (S1 Table), showed comparable sex frequencies between groups (OR = 0.90;       Table). Remarkably, most cohort studies equally converged on a higher incidence of the disease in this age group, particularly among males, with two additionally reporting another peak of incidence in the elderly [10,31] (Table 1).

Family history of BU
One cohort study found higher rates of BU in family members of BU patients (5.69/1000 person-years) when comparing the estimated rates for the general population of the region (0.85-4.04 cases/year/1000 population; [41] (Table 1). Among the five case-control studies registering this variable, only one adjusted the analyses for non-consanguineous bonds [15], while two included household relationships [33,40] and the other two did not provide any description on the variable [16,18] (S3 Table). Pooling of available data from studies comprising probable (OR = 1.14; 95% CI = 0.

BCG vaccination
Only two RCTs attempted to scrutinize the role of BCG in protecting against BU [43,45], both concluding on a short-term positive effect for the vaccine (S5 Table). However, studies were moderately heterogeneous, demonstrating a negligible effect for the vaccine when analyzed under a random-effects model (OR = 0.63; 95% CI = 0.38-1.05; I 2 = 56%; Fig 3).

Risk of bias
The summary of the risk of bias among studies can be found in S7-S9 Tables. Unclear effects of selection and detection biases should be considered when analyzing all studies, mainly due to the lack of incorporation of the social history and BCG vaccination and HIV infection statuses of BU patients, or the use of the unreliable BCG scar as a source of vaccination status check [47]. Considering a mean incubation period of M. ulcerans of 4.8 months, it is also impossible to warrant that the studies reviewed only included individuals free-of-outcome ad initium [48]. This timeframe was also considered in the evaluation of exposure periods, which was here established as minimum for the correct sorting of controls. Furthermore, literature supports that BU clinical diagnosis can be correctly achieved in over 90% of cases, although this value is dependent on the experience of the physician documenting the event [49]. This reinforces the assertions made by WHO and other authors to confirm BU cases by any of the

Discussion
Demographic variables of BU patients have been reviewed by many authors, with no strict consensus having been achieved as to their role as a risk modifier for disease acquisition [7,50]. Results here obtained point to a bipartisan role of age and sex in the susceptibility to BU, as the disease is more frequently observed among children, particularly males, and in the elderly. Interestingly, these patterns can also be observed in other studies not fulfilling the inclusion criteria for this systematic review (S10 Table). In younger generations, it has been previously argued that these differences could be explained based on increased exposure to M. ulcerans, mainly due to the more erratic behavior that makes children more prone to skin lesions, as well as their likelihood to stay near aquatic environments [50]. However, more recent data encompassing worse outcomes of BU among the elderly casted a stronger suspicion on age as a modulator of intrinsic processes, likely immunological, a phenomenon that could be transversal across generations [41,51]. As such, while age and sex have the ability to influence the disease process, it is still uncertain at this point why.
Suspicion on a role for inheritance in susceptibility to BU can be traced to a case series from 1990, fostered by the question: why, in a community of individuals with similar environmental exposure, do some develop the disease while others do not [15,52]? Unfortunately, not all studies have used the same criteria to assess this variable, significantly increasing the chance of incorporating the confounding effects of environmental exposure (S6 Table). In this regard, when considering exclusively non-consanguine relationships, Sopoh et al. found the odds of acquiring BU to increase the susceptibility to the disease (adjusted OR = 5.5; 95% CI = 3.0-10.0), which altogether with previous evidence suggests BU resistance/susceptibility can be inherited [15,17,41].
Genetic association studies performed so far also point to a role for genetics in BU. More importantly, not only does the study of SNPs help define further risk factors for BU development, but it also reinforces findings from cellular and molecular research, as seen, for instance, with interferon-gamma [13,53]. From the SNPs that were addressed in more than one study, data revealed that the minor allele from SLC11A1 rs17235409 was associated with decreased prevalence of the disease. Besides its role as a divalent cation transport, SLC11A1 (solute carrier family 11, member 1) possesses several other immunomodulatory properties, having also been significantly associated with other mycobacterioses in another meta-analysis [54,55]. Likewise, parkin, encoded by PARK2, can mediate an antimycobacterial response through protein ubiquitination, a process critical for autophagic targeting of intracellular pathogens [56]. Despite the absence of effect estimated for PARK2 rs1040079 in BU acquisition, Capela et al. found that the allelic variant in PARK2 rs1333955 conferred susceptibility to the disease, when modelling for dominant model of genetic inheritance [12]. As of now, these SNPs could be considered for population screening and stratification. However, in the future, larger scale genomic studies may point to new variants that open avenues for novel therapeutic targets.
The WHO recently advocated BCG vaccination in BU endemic countries, based on the results of a systematic review and meta-analysis from Zimmermann et al. demonstrating a protective effect for BCG, at least when considering the RCTs included in the study [8,19]. By analyzing data of RCTs and case-control studies within a random-effects modelling, the present study suggests that BCG does not have a meaningful impact in protecting against BU. In fact, and similar to what was already observed in tuberculosis, the protection conferred by the vaccine appears to wane over time, being in BU apparently restricted to the first 6 months postvaccination [43,57]. Moreover, from the 4-year follow-up made by Smith et al. it is possible to estimate that it is necessary to vaccinate at least 125 individuals in order to prevent one case of BU within this timeframe [43,45]. Still, whether stratification of these results by BCG strains could have exposed different results remains an interrogation, as this is a factor known to influence immune cellular response [58]. Thus, based on current evidence, the usefulness of BCG in BU appears to be essentially restricted to protecting against osteomyelitis forms of the disease [59,60].
Contrary to tuberculosis, in which HIV coinfection is well-known to foster active infection, evidence for such role in BU has been scarcer [7,61]. The studies gathered in this systematic review allow to draw a similar conclusion regarding BU development, further supported by estimates of a generally higher HIV prevalence among BU patients [21,62]. Indeed, not only does HIV favor the acquisition of BU, but Christinet et al. also described increased lesion size and severity, together with more prolonged wound healing periods in BU HIV + patients with low CD4 T cell counts [62]. Thus, immune compromise by HIV infection hampers control of M. ulcerans, findings that not only have consequences at the populational level but may prove helpful to advance the understanding of the disease mechanisms.
This study represents the first meta-analysis on clinical risk factors for BU beyond BCG vaccination. Its major strengths rely on the broad search criteria employed, inclusion and ponderation of results from a large number of variables from both RCTs and observational studies, as well as the consideration of the BU-confirmation status of cases in the analyses. In contrast, it is limited by the level of evidence provided by the studies, the actual number of studies that addressed some of the risk factors here reviewed, as well by a lack of uniformity among definitions and study designs seen across the literature. Thus, some recommendations for more robust epidemiological studies are here proposed: consideration of the confirmation-status of BU patients throughout analyses; reporting and/or more thorough evaluation of the exposure variables here addressed; and the use of consistent cut-off values, namely for age groups, to allow data comparison.
Overall, this work contributes with the notion that there is a bimodal peak of incidence in BU, namely among children and older adults, although the underlying causes are yet to be unraveled. Furthermore, evidence suggests BCG vaccination is not suitable to prevent BU in the mid-long term, prompting even further for the investigation of better vaccination alternatives. Additionally, this review found a higher prevalence of the disease in HIV-infected individuals and also carriers of genetic polymorphisms in SLC11A1, both of which can be easily detected by low-cost methods and thus be used for BU risk stratification. Altogether, these findings contribute to the understanding of BU and should be regarded when advocating for new global health policies against this neglected tropical disease.