Background
Trachoma remains the most common infectious cause of blindness worldwide. It is estimated that 84 million people, mostly young children, have active disease, and 1.2 million people are blind from trachoma [
1]. Blindness occurs as a result of repeated infection of the ocular surface with the bacterium
Chlamydia trachomatis. The infection results in chronic, recurrent conjunctival inflammation (active trachoma), which then progresses to scarring [
2]. This scarring leads to in-turning of the eyelashes, known as trachomatous trichiasis (TT), which traumatises the eye surface leading to corneal opacification and blindness [
2].
To prevent sight loss from trachoma, effective long-term control of
C. trachomatis infection and transmission is required [
2]. The World Health Organization (WHO) aims to eliminate trachoma as a public health problem by 2020. To help achieve this, the Alliance for the Global Elimination of Trachoma by the year 2020 (GET 2020) was formed. The Alliance has two major elimination targets, the first of which is to reduce the prevalence of trachomatous inflammation, follicular (TF) in children aged 1–9 years old to < 5 % in all endemic regions by 2020. To help reach this goal, the WHO recommends the use of the SAFE strategy: surgery for trichiasis, antibiotics to treat infection, facial cleanliness and environmental improvements to suppress transmission [
2]. Programmatic success is evaluated through follow-up surveys assessing prevalence of TF in 1–9-year-olds, following several annual antibiotic treatment rounds.
The current antibiotic of choice is oral azithromycin, which is distributed to entire endemic districts meeting the intervention criteria. The WHO recommends that treatment is repeated annually for an initial 3–5-year period, depending on the baseline disease prevalence [
3]. Unfortunately, particularly in areas with a high baseline prevalence, the response to treatment at the community level has been relatively disappointing. In some areas which have been studied in detail, such as Kongwa, Tanzania [
4], and Gurage, Ethiopia [
5], more than seven rounds of treatment have been deployed; however, prevalence of active disease has not dropped below the target level of < 5 % in children aged 1–9 years to eliminate trachoma as a public health problem. This is supported by findings from a recent study, in which the authors suggested that countries should prepare for extended antibiotic timelines [
6]. As the target date to eliminate trachoma as a public health problem by 2020 fast approaches, there is a pressing need to identify alternative, more effective antibiotic distribution strategies for highly endemic regions.
Individuals with the highest ocular chlamydial loads are most commonly young children, under 10 years of age [
7]. Available data suggest that individuals with higher bacterial infection loads are less likely to resolve infection upon treatment than children with a lower starting load. Therefore, children with the highest bacterial loads pre-treatment can remain a source for re-emergent infection in the community [
8]. Additionally, re-emergence of disease clusters have been reported around households of children identified as having high bacterial loads [
9,
10]. Furthermore, simply increasing the dose of azithromycin for those with severe active trachoma (as a surrogate marker for higher infection loads) was not found to be more effective than a standard dose in resolving infection [
9]. Similarly, treatment of individuals with severe active trachoma with two standard antibiotic doses given on consecutive days was also not found to be more effective than the standard single dose [
11]. However, it is possible that the second dose applied in this trial on the second day did not have additional benefit, as it was administered when an individual probably still had a high bacterial load [
11].
Therefore, in this study we suggest the implementation of an alternative double-dose antibiotic distribution strategy, where two doses of antibiotic are administered 2 weeks apart from one another. The rationale for this approach is that the probability of resolving infection following treatment is related to the load. The available data suggest that while some high-load infections do fully resolve following a single dose of treatment, a proportion of these high-load infections do not; however, they do transition into a lower-load category [
8]. Here, the first dose of antibiotic treatment acts to reduce an individual’s bacterial load. Individuals are then re-treated 2 weeks after the first dose. Those that are still infected after the first treatment are likely to be in a lower-load state, and as such, they have a higher probability of fully resolving their infection following the second antibiotic dose.
Treatment of infection with antibiotics represents only one arm of the SAFE strategy. Indeed, a recent analysis suggested that particularly for highly endemic regions, repeated annual single-dose mass drug administration (MDA) alone is insufficient to eliminate infection, without lasting environmental improvements [
6]. The quantitative impact of non-pharmaceutical interventions (NPIs) on trachoma transmission remains unclear. One recent study attempted to quantify the impact of facial cleanliness (F) and environmental improvements (E) in helping to reduce transmission and the odds of acquiring infection [
12]. The authors suggested that the impact of improved sanitation and hygiene in helping to reduce trachoma transmission is currently underestimated and that it can help to reduce the odds of acquiring infection [
12]. However, the mechanisms of transmission remain poorly defined, and there are few randomised control trials in this area to support or quantify the role of F&E on trachoma.
Here, we propose a development of the existing F&E interventions, which we refer to as ‘enhanced F&E’. In this trial scenario, enhanced F&E would be developed through intensive observational studies, which document human behaviour, its determinants and the hygiene practices adopted by individuals. This focused ethnographic approach will thus document in detail key behaviours and possible transmission routes relevant to trachoma transmission. Through this, targeted F&E interventions will be developed in partnership with the communities to generate promotional activities to prevent these pathways occurring. It is this specific targeting of interventions and promotional activity informed by ethnographic research that defines enhanced F&E.
Previous mathematical modelling studies have suggested that in areas of high transmission intensity at least two rounds of single-dose antibiotic treatment at 6-monthly intervals may be needed to help control infection [
5,
13,
14]. However, one clinical trial in high prevalence communities reported that despite biannual treatment for 3 years infection re-emerged [
5]. The logistics and cost of distributing treatment at 6-monthly intervals requires considerable resources. If the second dose is given in quick succession there are likely to be some savings in cost relative to giving MDA at 6-monthly intervals; for example, less overall preparation and management costs, it would only be necessary to sensitise and mobilise the community once a year, and there may also be some transport cost savings if the distribution team is still in the same area.
Here, we explore the potential community-level impact of the double-dose antibiotic distribution strategy administered on an annual basis to help reduce transmission of the pathogen and burden of disease within the community. We examine the projected impact of reductions in transmission that may occur as a result of the introduction of enhanced F&E interventions, as defined previously. We assess our findings across a range of key parameters, including: frequency of treatment, coverage of treatment, differential treatment efficacy and a range of reductions in the transmission rate.
Discussion
This study explored the projected effectiveness of different antibiotic distribution strategies and levels of transmission reduction on prevalence of infection and disease in three different transmission settings, with the ultimate aim of achieving the GET 2020 goal of reducing prevalence of TF in 1–9-year olds to < 5 %. It has previously been demonstrated that the single-dose efficacy of azithromycin treatment is not 100 %, particularly for those with high bacterial loads [
8,
10]. In the proposed double-dose strategy, in cases where the infection does not resolve, the first dose of antibiotic treatment reduces an individual’s bacterial load, thus when they are re-treated 2 weeks after the first dose, they have a higher probability of resolving their infection following the second dose. The implementation of this new strategy may also help to increase the overall level of coverage achieved across the two rounds, due to increased awareness, along with catching individuals who may have missed the first round of treatment. Although the number of antibiotic doses required initially would be higher than when single-round annual MDA is implemented, it is likely that the total number of doses required for long-term control would be fewer under the double-dose strategy as the treatment programme would need to run for a shorter time period.
To date, very few clinical studies have explicitly monitored the impact of water, sanitation and hygiene (WASH) or F&E on human behaviours; nor have they attempted to quantify the possible impact of these interventions on trachoma transmission within a clinical trails setting. Further to this, these interventions have, to our knowledge, not been previously assessed in the trachoma mathematical modelling literature. A key feature of the proposed clinical trial is the enhanced F&E component. This component seeks to draw ethnographic information on human behaviours and practices to identify key routes of trachoma transmission within the community, facilitating the formulation and development of highly specific and targeted F&E interventions, which aim to directly reduce trachoma transmission within the community. This will for the first time allow estimates of highly targeted transmission reduction measures for trachoma to be made.
We acknowledge that particularly for single-dose annual MDA, the rates of infection rebound in the model following treatment are generally higher than those observed in the field [
5,
18,
19]. However, high variance in the number of cases reported following several rounds of MDA is cited [
18]. Furthermore, many antibiotic treatment studies may have an F&E component, which is not always explicitly monitored or quantified. Therefore, we may expect an underlying degree of transmission reduction in these trials as well. As such, we may expect findings from trials to look like the simulations which model MDA with a small amount of transmission reduction. Indeed, we see that the output from these simulations show a much slower rebound effect.
We found that the GET 2020 goal of reducing prevalence of TF in children aged 1–9 years old to < 5 % may be achievable with the interventions explored for a number of transmission settings; however, monitoring will be needed (and is recommended) to check that the prevalence remains below the elimination threshold, as our results suggest that in the absence of transmission reduction interventions, infection may re-emerge. This observation is supported by MDA clinical trial data [
5,
18,
19]. One recent theoretical modelling study suggested that if a community experienced 10 annual rounds of MDA and rapid transmission reduction, shifts in population immunity may be observed resulting in short-term increases in R0 [
20]. However, the modelling results presented here suggest that with a smaller number of annual rounds of MDA and transmission reduction through F&E it is possible to eliminate infection in all transmission settings.
For hyperendemic communities, the new double-dose treatment regime may have a large impact on reducing prevalence of infection, in comparison to single-dose annual MDA. With a reduction of greater than 10 % in the transmission parameter β through enhanced F&E, it may be possible to control infection in hyperendemic communities. The annual double-dose strategy was projected to reduce prevalence of infection and disease more quickly than biannual treatment at 6-monthly intervals. However, as coverage and treatment efficacy reduced, the projected effect of impact was also reduced. For example, in the hyperendemic community, if coverage was only 60 %, neither strategy in the absence of enhanced F&E was sufficient to lower infection. This reinforces the importance of attaining high levels of treatment coverage to successfully control infection.
For hypo- and mesoendemic settings, the double-dose distribution strategy was projected to provide an advantage in helping to achieve the GET 2020 goal of reducing the prevalence of active disease in 1–9-year olds. The double-dose strategy reduced prevalence of infection more quickly in comparison to annual MDA, suggesting that it may help to maintain a lower prevalence of infection and active disease for longer periods. However, when efficacy of treatment and coverage level reduced, infection may be more likely to remerge.
Our study has a number of limitations. Firstly, we have assessed the impact of interventions using a deterministic mathematical model, which at high levels of transmission intensity is reasonable. However, as transmission reduces our model does not allow for the stochastic fade-out of infection that may occur at low levels of prevalence and transmission [
21‐
23]. As such, we may have overestimated the effort required to control infection at lower levels of transmission intensity. Equally, re-emergence of infection at very low transmission intensity may only be an artefact of the deterministic model structure used here. However, it is also important to consider that stochastic re-emergence of infection within communities may occur. Secondly, the quantitative impact of enhanced F&E remains poorly understood; therefore, we have modelled this effect through direct reductions in the transmission rate parameter
β. However, these interventions may act through indirect means not explicitly accounted for in the model. We do not know what these levels of transmission reduction would correspond to in terms of F&E control measures, nor do we know whether F&E would be equally accessed by all members of the population, though this seems a reasonable assumption for a model. Additionally, the reduction in transmission may not be instantaneous, but more gradual as hygiene and sanitation conditions improve over time. However, sensitivity analysis to this assumption suggested that it had relatively little impact in the meso- and hypoendemic settings (Additional file
1: Figure S35–S36). For hyperendemic communities, year-on-year reductions in infection prevalence were slightly lower when
β declined more gradually over the 5-year period. However, as the total level of transmission reduction in the community achieved was the same in both situations, the results at the end of the intervention period were the same. Here, 10–20 % reduction in transmission was sufficient to eliminate infection with the double-dose treatment strategy (Fig.
2 and Additional file
1: Figure S34).
Thirdly, while we have assessed the impact of our different intervention scenarios across a range of coverage levels, it is possible that coverage may vary between years of intervention implementation—this may impact long-term trends in prevalence. Fourthly, as with many modelling studies we have made a number of simplifying assumptions, including assuming treatment efficacy was the same for all individuals who were low loader and those who were high. In reality, it is likely that efficacy between all individuals will vary to some extent. However, capturing the heterogeneity in response to treatment through high and low bacterial loads is one further step in the development of biologically plausible models. Further to this, we have assumed the primary mechanisms underlying the re-emergence of infection within the community are due to the failure of one antibiotic dose to fully clear infection in those who have high bacterial loads; however, the picture is likely to be more complex. We have assumed that individuals are missed from treatment at random, though there may be small groups of people or households who are systematically not treated, resulting in continuous transmission within these groups. However, ongoing transmission within the household or the community is to some extent accounted for in the re-emergence of infection within the community under certain scenarios.
In addition, for a given level of treatment efficacy we have assumed that for low loaders infection is fully cleared. However, the microbial cure rate with azithromycin for urogenital chlamydia is estimated to be 97–98 % [
24]. This ‘latent chlamydia’ is believed to not be susceptible to treatment, and is reported to be the reason behind the imperfect cure rate for urogenital chlamydia. This observation may also extend to ocular chlamydia, but to our knowledge, data to support this in the ocular context remains limited. However, if there is a background level of latent chlamydia in the community that is untreatable, the rate of infection re-bound within MDA-treated communities may be higher than currently modelled here. The ability of individuals with latent chlamydial infection to act as a reservoir source of trachoma infection within the community remains unknown and warrants further investigation.
We have assumed that individuals with active disease are immune to re-infection, although a small amount of clinical data has suggested that this might not be true. However, very limited quantifiable information is available to parameterise this. Nevertheless, sensitivity analysis to this assumption suggested that an increase in the susceptibility to re-infection in the AD state could lead to a larger impact of the modelled interventions for both the single- and double-dose treatment regimens across all transmission settings (Additional file
1: Figure S37–S39). Lastly, we have assumed that immunity to infection is acquired exponentially, and high bacterial loads fall after the first ten infections. However, immunity to infection may take a different functional form, and it may require more or less infections to qualify as a high or low bacterial loader. A recent study in Kongwa, Tanzania, suggested that bacterial load prior to MDA was not predictive of continued infection in individuals following MDA [
25]. However, the authors acknowledge that this is not consistent with their previous findings, and that the calculation of infection load between the two studies limits their analysis [
25]. We feel that two published studies, which present conflicting results, do not provide sufficient evidence to refute the hypothesis evaluated in this article. Moreover, there is also support in the urogenital literature to suggest that organism load may be associated with treatment failure [
26‐
28], highlighting that the role of bacterial load and its association with treatment failure is, at this point, not clear-cut.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
AP, MJB and MG conceived and designed the study. AP prepared the first draft and performed the simulation analysis. All authors provided interpretation on the modelling output. All authors contributed to the writing of the manuscript and approved the final version.