Association between Footwear Use and Neglected Tropical Diseases: A Systematic Review and Meta-Analysis

Background The control of neglected tropical diseases (NTDs) has primarily focused on preventive chemotherapy and case management. Less attention has been placed on the role of ensuring access to adequate water, sanitation, and hygiene and personal preventive measures in reducing exposure to infection. Our aim was to assess whether footwear use was associated with a lower risk of selected NTDs. Methodology We conducted a systematic review and meta-analysis to assess the association between footwear use and infection or disease for those NTDs for which the route of transmission or occurrence may be through the feet. We included Buruli ulcer, cutaneous larva migrans (CLM), leptospirosis, mycetoma, myiasis, podoconiosis, snakebite, tungiasis, and soil-transmitted helminth (STH) infections, particularly hookworm infection and strongyloidiasis. We searched Medline, Embase, Cochrane, Web of Science, CINAHL Plus, and Popline databases, contacted experts, and hand-searched reference lists for eligible studies. The search was conducted in English without language, publication status, or date restrictions up to January 2014. Studies were eligible for inclusion if they reported a measure of the association between footwear use and the risk of each NTD. Publication bias was assessed using funnel plots. Descriptive study characteristics and methodological quality of the included studies were summarized. For each study outcome, both outcome and exposure data were abstracted and crude and adjusted effect estimates presented. Individual and summary odds ratio (OR) estimates and corresponding 95% confidence intervals (CIs) were calculated as a measure of intervention effect, using random effects meta-analyses. Principal Findings Among the 427 studies screened, 53 met our inclusion criteria. Footwear use was significantly associated with a lower odds of infection of Buruli ulcer (OR = 0.15; 95% CI: 0.08–0.29), CLM (OR = 0.24; 95% CI: 0.06–0.96), tungiasis (OR = 0.42; 95% CI: 0.26–0.70), hookworm infection (OR = 0.48; 95% CI: 0.37–0.61), any STH infection (OR = 0.57; 95% CI: 0.39–0.84), strongyloidiasis (OR = 0.56; 95% CI: 0.38–0.83), and leptospirosis (OR = 0.59; 95% CI: 0.37–0.94). No significant association between footwear use and podoconiosis (OR = 0.63; 95% CI: 0.38–1.05) was found and no data were available for mycetoma, myiasis, and snakebite. The main limitations were evidence of heterogeneity and poor study quality inherent to the observational studies included. Conclusions/Significance Our results show that footwear use was associated with a lower odds of several different NTDs. Access to footwear should be prioritized alongside existing NTD interventions to ensure a lasting reduction of multiple NTDs and to accelerate their control and elimination. Protocol Registration PROSPERO International prospective register of systematic reviews CRD42012003338


Introduction
Neglected tropical diseases (NTDs) are caused by a variety of pathogens, such as parasites (e.g., ectoparasites, helminths, and protozoa), fungi, bacteria, and viruses, primarily found in the tropical and subtropical regions of the world [1]. NTDs mainly occur in rural and deprived urban areas of low-and middleincome countries, where they may exacerbate poverty by contributing to significant morbidity and mortality, impairing development, and limiting productivity [1,2]. They have multiple routes of transmission and a single intervention alone is unlikely to have major sustained impact. Population-based chemotherapy is currently the mainstay of the control of various NTDs caused by helminths (e.g., lymphatic filariasis, schistosomiasis, and soiltransmitted helminth (STH) infections) and some bacterial infections (e.g., trachoma) [3,4]. More recently, attention has been given to water, sanitation, and hygiene (WASH) as an effective and sustainable measure for NTD control [5][6][7]. WASH interventions such as face washing to prevent trachoma, or hand washing to prevent diarrheal diseases and STH infection have been well-studied [8][9][10]. However, less attention has focused on other personal preventive measures to reduce exposure to infection, such as the use of footwear. Some NTDs may be transmitted or occur through the feet, and hence, footwear could prevent this exposure. To our knowledge, there has not yet been a systematic review of the evidence to assess the role of footwear use among these NTDs [8][9][10][11][12].
There is continued debate over the role of footwear use as an additional measure of NTD control [13][14][15]. Some studies have highlighted that footwear use could reduce infection with hookworm caused by Necatoramericanus and/or Ancylostoma duodenale (which is also orally infective), but such studies are often cross-sectional and should be interpreted with caution [15][16][17]. Other experts argue that decreases in the burden of hookworm disease are based on large-scale administration of deworming drugs (a strategy termed preventive chemotherapy), socioeconomic development, and improved access to WASH rather than widespread footwear use, while newer evidence indicates that the burden from hookworm disease has not changed significantly over the past 20 years [13,18]. Furthermore, the lack of adequate change in hookworm disease burden might be due to the overwhelming focus on preventive chemotherapy over the last few decades and less emphasis on other interventions [5]. In the case of podoconiosis (non-filarial elephantiasis), footwear use is currently promoted as a prevention tool, since current evidence suggests that it is caused by barefoot exposure to red clay soil from volcanic rocks [19]. Other studies and anecdotal evidence have additionally suggested that footwear use may prevent Buruli ulcer, cutaneous larva migrans (CLM), leptospirosis, mycetoma (fungal eumycetoma and bacterial actinomycetoma), myiasis, snakebite, strongyloidiasis, and tungiasis [20,21].
Here, we first identified those NTDs for which the use of footwear might have a potential impact on the risk of infection and disease, based on an understanding of disease etiology and transmission. We next conducted a systematic review and series of meta-analyses of the association between footwear use and the risk of a range of NTDs.

Methods
NTDs were selected to be included in the study based on disease etiology and potential for infection through the feet and thus prevention using footwear (Table 1). A systematic literature review protocol strategy was developed based on the 'Preferred Reporting Items for Systematic reviews and Meta-Analyses' (PRISMA) checklist (e.g., protocol and registration, eligibility criteria, information sources, searching, study selection, data collection process, data items, risk of bias in individual studies, summary measures, synthesis of results, risk of bias across studies, and additional analyses (see: Checklist S1). This protocol is available at the National Institute for Health Research PROSPERO International prospective register of systematic reviews (identifier: CRD42012003338) (see Protocol S1).
A total of 92 known medical and colloquial disease names (see Table 1) were included in a comprehensive list of key search terms. Six terms related to footwear were also included: shoe, footwear, boot, sandal, footgear, or primary prevention. Relevant databases were searched from using these terms, including Medline (coverage from 1950), Embase (coverage from 1947), Cochrane (coverage from 2003), Web of Science (coverage from 1900), CINAHL Plus (coverage from 1937), Popline (coverage from 1970), British Library for Development Studies (coverage from 1987), ELDIS (coverage date unavailable), EPPI-Centre (coverage from 2004), WHO Library (coverage from 1948), and PAHO Library Catalogue (coverage from 1902). The search was conducted from January 1, 2013 to December 31, 2014. Experts in selected NTD areas were contacted for further citation recommendations relevant to the research question. The Brighton and Sussex Medical School (BSMS) Library was consulted for assistance with article retrieval through online databases or manual journal searching. The reference lists of all identified manuscripts were also reviewed for additional citations. Manuscripts in foreign languages (namely, French, Spanish, and Russian) were translated by investigators. No other foreign language articles were identified through this search. When potentially eligible studies did not provide sufficient data in the manuscript, authors were contacted and asked if they would be willing to provide additional data. To this end, additional data were received from authors of five studies [22][23][24][25][26].
Pre-defined eligibility criteria included: (i) all intervention and observational study designs; (ii) all study settings; (iii) all ages; (iv) all types of footwear exposures; (v) prevalence or incidence estimates of infection and/or disease outcomes; (vi) all published manuscripts and grey literature; (vii) all publication dates; and (viii) all languages. Observational studies were included because it was hypothesized that few randomized controlled trials (RCTs) had been conducted to answer the research question. Abstracts of identified studies were reviewed before appraisal of full manuscripts when possible. If a study did not explicitly investigate the association between footwear use and any of the target NTDs or

Author Summary
Consistent use of footwear may help in preventing or slowing down the progression of many neglected tropical diseases (NTDs). We conducted a systematic review and meta-analysis to assess the association between footwear use and infection or disease for those NTDs for which the route of transmission or occurrence may be through the feet. We found that footwear use reduces the risk of Buruli ulcer, tungiasis, hookworm, any STH infection, strongyloidiasis, and leptospirosis. No significant association between footwear use and podoconiosis was found and no data were available for mycetoma, myiasis or snakebite. We recommend that access to footwear should be prioritized alongside existing NTD interventions to ensure a lasting reduction of multiple NTDs and to accelerate their control and elimination.
did not meet the eligibility criteria, it was excluded. Decisions on inclusion were reached by the consensus of independent screenings conducted by two investigators (ST and KD).
A standardized Excel data extraction form was developed based on the PRISMA statement [27] and used to record the following information: study ID, author, title, journal, publication year, type of literature, research question, study design, study setting, outcome, follow-up, sample size, number of cases, descriptive case data (e.g., age, sex, and proportion wearing footwear), descriptive control data (e.g., age, sex, and proportion wearing footwear), crude and adjusted effect estimates of footwear use on disease, including 95% confidence intervals (CIs) and p-values, and study quality ratings. The methodological quality of studies was assessed by the same investigators. According to a pre-defined scale, the following question was assessed, stating ''Were the following items reported?'': (i) study population (e.g., social-ecological characteristics); (ii) selection of participants (i.e., random or convenience); (iii) sample size calculation; (iv) method of measuring footwear use and presence of NTDs; and (v) estimates adjusted for confounding. These questions were scored on a yes/no basis and proportions answering yes to each question were described to assess study quality and risk of bias in and across individual studies. None of the investigators on this review assessed the study quality of their own primary studies.
All primary data and quality ratings were extracted from identified manuscripts. STATA version 12.0 (College Station, TX, United States of America) was used to summarize the data descriptively. RevMan version 5.2 and its generic variance format was used to generate individual forest plots according to primary NTD outcomes [28].We entered odds ratio (OR) estimates of footwear use on a logarithmic scale and standard errors (calculated from 95% CIs). An adjusted OR was used if provided in the manuscript. A few studies only provided raw outcome and exposure data, so we calculated a crude OR in these cases. All calculations and data used are detailed in the footnotes of each figure. A random-effects model in RevMan was then utilized to produce individual study ORs and 95% CIs and to consider a pooled summary effect estimate (using random effects to address potential heterogeneity). Heterogeneity was assessed by the I 2 test with values greater than 50% representing moderate-to-severe heterogeneity.

Results
The electronic searches generated 427 citations and abstracts. These were screened and 374 were excluded for a range of reasons ( Figure 1). We included 53 sources: Buruli ulcer (n = 3), CLM (n = 1), leptospirosis (n = 7), podoconiosis (n = 6), any STH infections (n = 11), hookworm infection (n = 17), strongyloidiasis (n = 4), and tungiasis (n = 4). No data were found to quantify the association between footwear use and mycetoma, myiasis, and snakebite. Type of source included 50 journal manuscripts (94.3%), two unpublished pieces of work (3.8%), and one book excerpt (1.9%). Information describing the studies included are summarized in Table 2, including study design, publication year, country and outcome. We identified a total of 40 cross-sectional studies (75.4%), eight case-control studies (15.1%), three cohort studies (5.7%), and two RCTs (3.8%). The median publication year was 2003 (range: 1950-2014). Geographically, 29 studies were conducted in Africa (54.7%), 12 in Asia (22.6%), 11 in the Americas (20.8%), and one in Europe (1.9%). The median sample size was 366 individuals (range: 59-129,959). Among the 11 studies with known follow-up periods, the median follow-up time was 12 months (range: 2.5 months to 7 years). Descriptive results by individual studies including sample size, median age, and proportion of females are shown in Table 2. A summary of descriptive results by outcome, including study quality results, are provided in Table 3.

Discussion
We found that footwear use was significantly associated with a lower odds of Buruli ulcer, CLM, leptospirosis, strongyloidiasis, any STH infection, hookworm infection, and tungiasis, highlighting the important role of footwear use in the prevention of NTDs. No significant association was found between footwear use and podoconiosis. We found no data regarding the use footwear and mycetoma, myiasis, and snakebite. The results presented here have important implications for both policy and practice. Promotion of footwear use should be an important part of selected NTD control strategies.
The significant association between footwear use and the lower odds of Buruli ulcer, CLM, hookworm infection, leptospirosis, and tungiasis are consistent with the mode of transmission of these diseases [16,20,[29][30][31][32][33][34].The risk factors include presence of skin cuts or abrasions and contact with water, soil, or mud during work or recreational activities if the water is contaminated with human and animal excreta, including rodent urine [33]. Our findings are also consistent with a recent meta-analysis on WASH interventions    which included some results regarding footwear use and hookworm infection or any STH infection [7]. This review found that footwear use was significantly associated with a lower odds of hookworm infection (OR 0.29; 95% CI 0.18-0.47) and any STH infection (OR = 0.30; 95% CI 0.11-0.83) [7], as compared to the findings on hookworm infection (OR = 0.48; 95% CI: 0.37-0.61) and strongyloidiasis (OR = 0.56, 95% CI 0.38-0.83) in the current analysis.
We did not find a significant association between footwear use and the risk of podoconiosis. Two issues may explain this: first, podoconiosis is a chronic disease primarily affecting the feet, so reverse causality is likely (when an individual first notices foot or leg swelling, he or she starts wearing shoes), and second, podoconiosis requires a long period of exposure, but assessment of current use of footwear does not reflect previous exposure [35]. Studies comparing podoconiosis patients with healthy controls have found that patients tend to wear footwear more than healthy controls to protect their legs from injury or to conceal the swelling in fear of stigma and discrimination [36,37]. Other studies have suggested that age at first footwear use would be a more precise indicator of protection than current footwear use [38].
A number of our findings support integrated control strategies of NTDs. Footwear use appears to have a protective effect across multiple NTDs and thus may become an important integrated NTD control measure and should be considered by researchers, program planners, and policy makers. Footwear use interventions also have the potential to enhance sustainability of NTD control programs, similar to improved access to clean water, sanitation, and altered hygiene behavior [1,5,7,[39][40][41][42][43].Advocacy could be integrated into current efforts such as school health services, and indicators on type and frequency of footwear use could be included in NTD monitoring and evaluation. Initial investments may only be needed to create awareness and demonstrate the practical benefits of footwear use and promote it as a continued behavior. A public-private partnership model similar to that of pharmaceutical companies for population-based chemotherapy could be seen as example to leverage resources with footwear companies. However, future cost-effectiveness studies are needed to fully explore the feasibility and sustainability of these interventions.
We aimed to adhere to the PRISMA statement for the reporting of meta-analysis of observational studies. However, there were several limitations in this systematic review. First, only six out of the 56 included studies specified the type of footwear. Thus, we were unable to explore how the type of footwear may have affected the results. Type and frequency of footwear use may vary regionally due to differences in seasonality, socioeconomic conditions, occupation, and cultural practices. These differences could affect the effectiveness of footwear interventions and practical implementation of related interventions. Only one study was identified for CLM which limited our ability to conduct metaanalysis for this outcome.
Second, there was marked heterogeneity with wide CIs between some studies which may have led to imprecise summary estimates. I-squared estimates varied, including low heterogeneity (strongyloidiasis 0% and Buruli ulcer 26%) and moderate-to-high heterogeneity (tungiasis 63%, leptospirosis 69%, any STH infection 74%, hookworm infection 74%, and podoconiosis 96%). This may have been due to the different definitions (e.g., many studies used a questionnaire design without clarifying the  [44]. *Log [odds ratio] and SE calculated from raw data: Marston et al [84] and Raghunathan et al [45]. doi:10.1371/journal.pntd.0003285.g002  [46] and Sulong et al. [49]. *Inverted Log [odds ratio] and SE from effect estimate of barefoot exposure: Bovet et al. [51], Johnson et al. [46], and Leal-Castellanos et al. [47]. *Log [odds ratio] and SE calculated from raw data: Lacerda et al. [52]. doi:10.1371/journal.pntd.0003285.g003 type of footwear or consistency of use) or diagnostic methods employed to determine infection or disease across studies. However, we used a random-effects model to calculate summary measures in an attempt to address this heterogeneity. The results using a fixed-effects model did not substantially differ from the random-effects model, indicating only small study biases.
Lastly, most of the studies were observational in nature (e.g., cross-sectional surveys and case-control studies), giving rise to concerns regarding study quality. With cross-sectional surveys, we are unable to reach conclusions about the effect of shoes on the incidence of infection or disease over time and estimates may be confounded by other variables. Only a limited number of studies provided adjusted estimates, often controlling for just a few sociodemographic variables, with potential residual confounding. Case-control studies may be affected by recall bias, depending on how cases recall footwear exposure compared to those without disease. Details on the measurement of footwear use and the presence of NTDs were not always reported which also may have led to biased estimates. Prospective studies specifically designed to look at the effect of footwear use on selected NTDs are needed to answer this research question. RCTs may provide more robust evidence but can be ethically and financially challenging. A recent cluster randomized trial failed to show any association between hookworm infection and footwear use due to contamination [12]. Approaches such as a stepped wedge trial design or a robust cohort study may offer more feasible solutions.

Conclusions
NTDs have multiple routes of transmission and a single intervention alone is unlikely to completely interrupt transmission. Little attention has focused on personal preventive measures to reduce exposure to infection, such as the use of footwear. Our findings provide evidence that footwear use could help prevent a range of different NTDs, including Buruli ulcer, leptospirosis, CLM, tungiasis, any STH infection, strongyloidiasis, and hookworm infection. Although prospective data are still needed to explore the effect of footwear use on the incidence of NTDs over time, these findings support the integrated control strategies of NTDs that include footwear use. Initial investments are required to create awareness and demonstrate the practical benefits of footwear use and promote it as a continued behavior. There may also be a need to provide footwear to particular at-risk groups (e.g., school-aged children for STH infections), and a similar publicprivate partnership model to that used with pharmaceutical  [54]. * A 95% confidence interval/standard error (SE) was not available so it was not included in the forest plot: Steinmann et al. (odds ratio: 0.64) [54]. *Adjusted effect estimate: Yori et al. [53]. *Inverted Log [odds ratio] and SE from effect estimate of barefoot exposure: Yori et al. [53]. *Log [odds ratio] and SE calculated from raw data (comparing severe form of illness to chronic infection): Sanchez et al. [55]. doi:10.1371/journal.pntd.0003285.g004  [60]. *A 95% confidence interval/standard error (SE) was not available so it was not included in the forest plot: Lello et al. (odds ratio: 0.81) [60]. *Adjusted effect estimate: Phiri et al. [56]and Mihrshahi et al. [57]. *Inverted Log [odds ratio] and SE from effect estimate of barefoot exposure: Phiri et al. [56], Modjarrad et al. [25], Mihrshahi et al. [57], Kurup et al. [58], Gunawardena et al. [23], Gamboa et al. [24]. *Log [odds ratio] and SE calculated from raw data: Martinez et al. [61], Liabsuetrakul et al. [62], and Khan et al. [63]. doi:10.1371/journal.pntd.0003285.g005  [71]. *The magnitude of the odds ratio and the 95% confidence interval/standard error (SE) were not available so it was not included in the forest plot: Behnke et al. [71] (odds ratio for footwear use not significant) and Bethony et al. [14] (footwear use was not significantly associated with hookworm infection   [79]. *Stratified exposure totals were not given for the following studies: Njau et al. [80] (N = 385) and Thielecke et al. [34] (N = 147). *The magnitude of the odds ratio and the 95% confidence interval/standard error (SE) were not available so it was not included in the forest plot: Thielecke et al. [34] (marginal decrease in intensity of infection with footwear use). doi:10.1371/journal.pntd.0003285.g007  [38]. *Inverted Log [odds ratio] and standard error (SE) from effect estimate of barefoot exposure: Deribe et al. [81] and Molla et al. [38]. *Log [odds ratio] and SE calculated from raw data: Price et al. [82], Kloos et al. [35], and Yakob et al. [83]. doi:10.1371/journal.pntd.0003285.g008 companies for large-scale preventive chemotherapy might be applied to leverage resources with footwear companies. However, future cost-effectiveness studies are needed to fully explore the feasibility and sustainability of these interventions.