Review of the evidence base for the ‘F’ and ‘E’ components of the SAFE strategy for trachoma control

Authors


correspondence Paul Emerson, Medical Research Council Laboratories Fajara, PO Box 273, Banjul, The Gambia. E-mail: pemerson@mrc.gm

Abstract

Summary Community control of trachoma as a blinding disease is based on the SAFE strategy of Surgery, Antibiotic therapy, Facial cleanliness and Environmental improvement. Surgery and antibiotic therapy currently dominate most programmes. Blindness from trachoma results from frequent infections repeated over many years, so ultimate success requires the reduction of transmission. This is only likely to be sustainable through the F and E components of SAFE. Environmental improvement with access to water, enhanced hygiene and better sanitation reduces trachoma transmission and the blinding sequelae eventually disappear. Transmission routes and factors that cause this are not known and consequently no single specific tool for F and E is in place. Evidence from intervention studies shows that the promotion of face-washing gave modest gains for intense effort and a pilot study showed that trachoma transmission was reduced in the absence of eye-seeking flies. Other studies have shown that latrines and improved access to water are associated with a lower prevalence of active trachoma. There is likely to be a long-term beneficial effect of a combination of improved water supplies, provision of latrines, facial hygiene promotion through established infrastructure and control of eye-seeking flies. Each of these interventions offers additional public health and other benefits in its own right. Further research on the routes of transmission, the role of hygiene and means of sustainable fly control should be a priority.

Introduction

Trachoma is a contagious eye disease caused by ocular infections with the bacterium Chlamydia trachomatis. It is the commonest cause of infectious blindness worldwide with hundreds of millions at risk and an estimated 5.8 million blind ( Thylefors et al. 1995 ). Blindness from trachoma is linked to poverty and is two to three times more likely to occur in women than men ( Tabbara & Ross-Degnan 1986; West et al. 1991 ). Trachoma afflicts the most deprived people in the world, people marginalized and without political voice ( Mecaskey 1998). It causes disability, dependency and poverty and is a barrier to development.

The challenge posed by the control of this disease has resulted in the formation of the WHO alliance for the Global Elimination of Trachoma as a blinding disease by the year 2020 (GET 2020). The alliance aims to control blindness from trachoma through the SAFE strategy, that is the provision of Surgery, Antibiotic therapy, Facial cleanliness and Environmental improvement. Corrective lid surgery is effective for delaying the onset of blindness in people with trichiasis ( Bog et al. 1993 ) and is a priority where there is a backlog of people at immediate risk. Regrettably surgical compliance is usually poor: only 18% of 205 women offered free treatment and transport had actually undergone an operation after two years in one study from Tanzania ( West et al. 1994 ). Treating trichiasis by mechanically sticking the lashes to the outside of the lid has been proposed as an alternative to surgery ( Graz et al. 1999 ), but the technique is yet to be adopted by the GET 2020 alliance. Rational use of antibiotics relies either on frequent mass treatment or targeted treatment of ‘at risk’ groups such as all children under 10 in hyperendemic areas or, because ocular chlamydial infection is transmitted within the family, cases and their families. Treatment is with either a single oral dose of azithromycin or topical application of tetracycline for six weeks. Treatment is effective on an individual basis ( Bailey et al. 1993 ) and mass treatment can reduce infection and transmission in the short term, but has not given lasting control ( West et al. 1993 ; Schachter et al. 1999 ). The frequency of mass azithromycin administration required to eliminate active trachoma from an area has been modelled by Lietman et al. (1999) , suggesting that in hyperendemic areas (where > 50% of children are active cases), biannual treatment of all people would be required for the elimination of active cases. In areas where trachoma is moderately prevalent (< 35% active cases in children), annual treatment with 100% coverage would be sufficient.

Blindness from trachoma is a result of frequent infections repeated over many years and the ultimate success of GET 2020 relies on blocking transmission and reducing the community prevalence of infection. It is logistically impractical to achieve this by surgery and antibiotic therapy alone and control programmes need to be enhanced by strategies that reduce transmission – this is the F and E part of SAFE. Overall development of an area causes the prevalence of trachoma to decline and the blinding sequelae to eventually disappear ( Nichols et al. 1967 ; Bobb & Nichols 1969; Dolin et al. 1997 ) and community-based programmes can reduce the prevalence of active trachoma ( Sutter & Ballard 1978, 1983), but we do not know the specific mechanisms by which this occurs.

This paper is a response to a call from affected countries at the third meeting of the WHO Alliance for the Global Elimination of Trachoma for a review of the evidence base linking facial cleanliness and environmental improvement with trachoma control, two aspects of the SAFE strategy that need to be incorporated into national trachoma control programmes. We review the potential public health contribution of available strategies. Papers were identified by referring to the earlier works of Prost and Négrel (1989) and Marx (1989), by a multilingual MEDLINE and BIDS search of publications from 1975 to 1999 and by following up citations noted in these studies.

Trachoma as a disease

For simplicity of diagnosis and consistency of reporting between areas, clinical trachoma and its complications have been divided into grades ( Figure 1) ( Thylefors et al. 1987 ). Throughout this review the term ‘active trachoma’ refers to grades TF and TI and is used as an indicator of current trachoma prevalence. ‘All trachoma’ also includes TS, TT and CO, which demonstrate past infection and are a measure of the historical impact of trachoma. The WHO grading system describes the condition of the upper conjunctiva at one point in time and the grades should only be considered sequential in that TS, TT and CO occur only after a history of the active grades. Each active infection is self-limiting and does not necessarily lead to an inexorable decline to blindness. A history of past infection manifests itself as scars on the upper conjunctiva which start to become prevalent among older children in hyperendemic areas. If the scarring is severe it can cause the lid margin to become distorted, turning the lashes inwards so they rest against the globe. In addition to the intolerable discomfort this must cause, the constant abrasion and secondary bacterial infections can lead to opacity of the cornea, loss of vision and eventually to blindness. The prevalence of scarring and its blinding sequelae increase with age and are most commonly seen in older adults ( Muñoz & West 1997).

Figure 1.

WHO simplified system for grading trachoma. The upper eyelid is everted and the tarsal conjunctiva visually examined with a × 2.5 binocular loupe and adequate light.

Many active infections are asymptomatic and appear to cause little or no discomfort ( Mabey et al. 1991 ). Each individual active infection probably presents insignificant long-term risk, but the contribution is not known. Community prevalence rates of active disease of up to 20% can be associated with levels of trichiasis below 1% ( Luna et al. 1992 ; Sukwa et al. 1992 ), but because of demographic changes the population in many affected areas is ageing, implying that more people will survive to old age when blindness from trachoma develops ( Schachter & Dawson 1990). To diminish the risk of blindness from trachoma, it may not be necessary either to find and treat all cases, or to eradicate the infection from the population. Transmission intensity need only be reduced such that the community prevalence falls below an as yet undefined threshold for the development of blinding complications.

Geography vs. poverty

The current geographical distribution of trachoma is largely, but not entirely, associated with hot and arid parts of the world such as the Sahel, the interior of Australia and the highlands of Ethiopia and India. This has led to a general perception that trachoma is linked with dry environments and scarcity of water. Trachoma has not always followed this distribution: in the past it was a major public health concern in more temperate climates. The disease was so widespread in London at the start of the 19th century that Moorfields Eye Hospital was founded in 1805 largely to deal with it ( Jones 1980). In 1931 MacCallan included Latvia, Poland, Estonia, Finland and Czechoslovakia in a list of countries where trachoma was considered ‘very common’, and special trachoma clinics continued to be held until 1949 in Finland ( Prost & Négrel 1989).

Within a geographical area the prevalence of trachoma is not necessarily greater where there is poor access to fresh water. Mann (1967) reported the prevalence of signs of trachoma as 8.6% in people from the Marshall Bennett islands, who had ‘no access to drinking water’– and instead were ‘relying on the water from green coconuts’, and 45.9% in the wet and humid capital of Papua New Guinea, Port Moresby. Studies from India ( Taylor et al. 1958 ), Brazil ( Luna et al. 1992 ), Australia ( Royal Australian College of Ophthalmologists 1980; Tedesco 1980), Japan ( Marshall 1968), Malawi ( Tielsch et al. 1988 ) and Saudi Arabia ( Barenfanger 1975) have all shown a higher prevalence of trachoma in households with lower income than their neighbours and where the head of the household is poorly educated (reviewed in Marx 1989). Trachoma is a problem among the disadvantaged and dispossessed, whether they be nomads eking out an existence in the Australian desert, subsistence farmers on the edge of the Sahara or impoverished workers in preindustrial cities. Hot and arid may simply be a proxy for poverty in the modern world.

The transmission of trachoma

Reservoir

Cases of active trachoma that shed the elementary body stage of C. trachomatis are believed to be the main source of infection. The prevalence of active trachoma is greatest among pre school-age children and then decreases with age ( Kupka et al. 1968 ; Dawson et al. 1976 ; Taylor et al. 1985 ; Tielsch et al. 1988 ; West et al. 1991 ). Infections in children persist longer than in adults ( Bailey et al. 1999 ) suggesting that infected children form a reservoir of infection. C. trachomatis can also be found in the genital tract and in the naso-pharynx. Cervical infections, caused by C. trachomatis serotypes D–K, can be transmitted vertically during birth and this was previously believed to be an important route of transmission ( Jones 1975). It is now known that these genital serotypes are only rarely transmitted between eyes ( Brunham et al. 1990 ). Naso-pharyngeal C. trachomatis may be important in trachoma transmission but could be a result of ocular infection and probably does not constitute a separate reservoir. West et al. (1993) found that children with active trachoma were also likely to have nasal specimens positive for C. trachomatis. However, the rate of reinfection after topical eye treatment was similar in children with a positive and those with a negative nasal swab at baseline, when faster reinfection in those with positive swabs may have been expected if there were a naso-pharyngeal reservoir. Systemic antibiotic treatment would remove a nonocular reservoir, so we might expect that it would offer longer protection from reinfection than topical treatment. Bailey et al. (1993) showed that the rate of reinfection in children treated with a systemic antibiotic was the same as in those treated topically. No animal reservoir has been found, and extra-ocular sites of infection do not appear important in trachoma epidemiology. The eyes of infected people, particularly children, should be considered as the main reservoir of infection.

Clustering of cases

The prevalence of signs and symptoms of trachoma is not uniform within a geographical area and can vary greatly between villages. Within a village trachoma is strongly clustered by household ( Katz et al. 1988 ; Tielsch et al. 1988 ; Mabey et al. 1992 ), suggesting that the type of prolonged close contact found within families is required for high levels of transmission. Strong evidence implying that intimate contact is important has come from multivariate studies which identify sharing a sleeping room with an active case as a major risk factor for trachoma ( Bailey et al. 1989 ; Courtright et al. 1991 ; West et al. 1991 ). Crowded living conditions increased the risk of trachoma in some studies ( Assad et al. 1969 ; Barenfanger 1975), but not others ( Bailey et al. 1989 ). It is unclear whether this is the result of crowding per se, or because crowded families are likely to be larger with more pre school-age children and less well-off than their neighbours.

Transmission routes

Knowledge of the mechanisms of transmission has not come a long way since the seminal work of MacCallan (1931) who proposed four routes over 50 years ago. Possible routes of transmission include:

  • Direct spread during play or when sharing a bed;

  • Conveyance on fingers;

  • Indirect spread on fomites (shared handkerchiefs, towels, etc.);

  • Eye-seeking flies;

  • Coughing/sneezing.

The relative importance of these routes may differ over time and according to place, and it is unlikely that any single route is responsible for all transmission in a particular area. Establishing the relative importance of these routes of transmission is difficult, even with multivariate analyses of risk factors, which control for the effects of age and clustering, or through intervention trials that seek to examine the effect on transmission of blocking a suspected route. In the absence of a single route of infection, the findings of such studies may only be of local relevance and it is doubtful that a single globally applicable ‘magic bullet’ to stop transmission can be found. Nevertheless, attempts to elucidate transmission may be important for the formulation of locally appropriate ‘F’ and ‘E’ strategies.

Trachoma and water

Although trachoma is currently associated with arid parts of the world where access to water is limited and the standard of personal hygiene low as a consequence, the published evidence linking trachoma with water is not conclusive. The mechanisms by which improved access to water is supposed to influence trachoma transmission are seldom explained and are not entirely clear. Increased frequency of laundry – specifically bed sheets and handkerchiefs – may reduce transmission on fomites ( Taylor et al. 1958 ). If transmission on fingers is important, the use of water for personal hygiene purposes may reduce the frequency of infection ( Wilson et al. 1991 ). Face-washing may decrease the accessibility of infectious discharge to flies ( Taylor et al. 1989 ) and make infected eyes less attractive to eye-seeking flies, since the presence of ocular or nasal discharge increases the number of flies attracted to a face ( Emerson et al. 2000 ). Whilst stating that face-washing ‘obviously has no effect on the course of disease’, Muñoz and West (1997) conjecture that it ‘may reduce the likelihood of autoreinfection’, presumably by flushing the elementary body form of the pathogen from the eye when rinsing.

The distance to a water source has been used as a predictor of disease; some studies have shown that an increased distance to water increases the risk of trachoma ( Mathur & Sharma 1970; Tielsch et al. 1988 ; West et al. 1989 ), whereas others found no effect of distance ( Kupka et al. 1968 ; West et al. 1991 ). One study from Ethiopia reports that people living more than 15 min walk from water had less active trachoma than those living less than 15 min away ( Zerihun 1997). These apparently contradictory results may be explained by the observation that per capita water consumption is remarkably constant between households when the round trip to a water source is 30 min or less ( Figure 2). The ‘water use plateau’ ( Cairncross & Feachem 1993) has been documented in East, West and Southern Africa, Asia and Central America and suggests that water consumption will differ from the cultural norm only where the supply is in-house or further than 15 minutes' walk away.

Figure 2.

Domestic water consumption and time required for water collection.

The quantity of water brought into a household may be of more importance than the distance it has come. In Morocco the quantity of water brought into a house was found to be independent of the distance, and children from households with a greater quantity of water had less active trachoma ( Kupka et al. 1968 ). Conversely two other studies controlling for distance have found that the total quantity of water used has no effect on the prevalence of active trachoma ( West et al. 1989 ; Bailey et al. 1991 ). One of these studies ( Bailey et al. 1991 ) observed how the water was used in the household and found that trachoma-free households used more water for washing children than households with trachoma cases. The evidence suggests that the availability of water may well have an impact on the epidemiology of trachoma, but it is difficult to disentangle the effect of water availability from other indicators of socio-economic development and the standard of living of study participants. In areas with poor water availability, the provision of water not only provides the means for better hygiene and cleanliness, it also releases the time needed for these activities to take place – the two greatest inhibitory factors to facial cleanliness identified in a survey conducted in Tanzania ( McCauley et al. 1990 ).

Literature on hygiene and environmental risk factors for trachoma – problems of methodology

Data from the literature ( Table 1) come from two types of study: observational and intervention. Most information comes from observational studies that compare the trachoma situation in different locations in an attempt to identify risk factors. There are only two well-conducted intervention studies ( West et al. 1995 ; Emerson et al. 1999 ) that have sought to investigate the effect of a specific change on trachoma transmission whilst keeping potentially confounding variables the same. In common with the literature on the impact of water supply on diarrhoeal diseases ( Blum & Feachem 1983) the trachoma literature is beset with a number of problems of methodology that make interpretation difficult.

Table 1.  Literature providing evidence for the F and E components of the SAFE strategy for trachoma control Thumbnail image of

1. Lack of adequate controls: an environmental change or intervention can have a positive, neutral or deleterious effect on the prevalence of trachoma in a community. In order for meaningful conclusions to be made it is vital that adequate controls are in place. Problems of control arise in several forms: the comparability of groups not established at baseline ( Portney & Hoshiwara 1970; Cairncross & Cliff 1987); obvious differences between groups making them incomparable ( De Sole 1987); surveys conducted in different seasons ( Marshall 1968; Royal Australian College of Ophthalmologists 1980), or even years apart ( Kupka et al. 1968 ).

2. Trachoma clusters by household which results in a lack of independence between cases that should be taken into account during analyses at the individual level ( Taylor et al. 1958 ; Marshall 1968; Assad et al. 1969 ; Royal Australian College of Ophthalmologists 1980; Tedesco 1980; Zerihun 1997).

3. Confounding variables such as sex and age not taken into account. Active trachoma is more prevalent in children and blinding trachoma is more prevalent in women. Comparisons of active trachoma prevalence should be standardized by age and blinding trachoma by age and sex ( Kupka et al. 1968 ; Marshall 1968; Mathur & Sharma 1970; Royal Australian College of Ophthalmologists 1980; Taylor et al. 1985 ).

4. Repeatability and validity of survey methods: It is reasonable to assume that if a single observer screened all subjects, systematic errors would occur evenly between groups. When two or more observers screen groups, the interobserver variation should be investigated to determine the reliability of the results ( Marshall 1968; Assad et al. 1969 ; Mathur & Sharma 1970). Observations of categoric details such as the presence/ absence of a latrine and distance to water should also be validated ( Majcuk 1966). Measuring behaviour-related variables such as face-washing, the frequency of fly eye contact or latrine use present difficulties of validation and need to be carefully defined to ensure repeatability.

5. Behaviour reporting: relying on questionnaires to measure behaviour is problematic. There may be a tendency to over-report what is considered positive behaviour ( Traore et al. 1994 ) and to exaggerate for effect if the respondent feels that he may benefit from his suffering. Hence frequency of washing may be over-reported and long distances to water sources exaggerated ( Tielsch et al. 1988 ; Taylor et al. 1989 ; West et al. 1989 ).

Facial cleanliness

A questionnaire-based study conducted in two rural villages in Mexico ( Taylor et al. 1985 ) found that children whose faces were reportedly washed more than seven times per week had a significantly lower risk of contracting active trachoma than those for whom face-washing was reported less often (relative risk = 3.1). This finding suggested that the promotion of face-washing would be a suitable tool for intervention deliverable through primary health care and education systems. However, the two villages in the study were significantly different at baseline and the relative risk was calculated from pooled data. Pooling data from different populations can lead to spurious results, an effect known as Simpson's Paradox ( Rothman 1986): in this instance the data should have been stratified and not pooled. Mabey et al. (1992) reanalysed after stratifying for village and showed a more modest relative risk of 1.85 (95% CI 1.15–2.8).

In a large (= 3832) multivariate study of Tanzanian preschool children ( Taylor et al. 1989 ), active trachoma was associated with unclean face (defined ‘by general facial appearances, especially the presence or absence of dirt, dust, or crusting on the cheeks and forehead’)(OR 1.30; 1.11–1.54). In the same multivariate model two other hygiene-associated factors also appeared protective of active trachoma: handkerchief use (OR 0.67) and towel use (OR 0.76). The same sample were additionally questioned about the frequency of face-washing and no effect was found – doubts were later raised about the validity of self-reporting of what was considered a desirable practice ( Muñoz & West 1997). 1085 children from the same population went on to have laboratory tests for trachoma ( West et al. 1991 ); 589 were clinically positive (i.e. were grade TF, TI or both) and 354 had a positive laboratory test for Chlamydia. Regression analyses revealed a positive association between clinical signs and unclean face for all children (= 1085) which did not reach statistical significance (OR 1.13 (0.83, 1.54)). The same pattern persisted for those either clinically or laboratory positive: clinical trachoma (= 589) OR for unclean face 1.21 (0.77, 1.90), positive laboratory test (= 354) OR for unclean face 1.01 (0.72, 1.23). Studies from Brazil ( Luna et al. 1992 ) and Malawi ( Tielsch et al. 1988 ) failed to show an association between frequency of face-washing and prevalence of active trachoma, but again these were based on self-reporting and may have been biased.

To test if face-washing was a suitable intervention for reducing the prevalence of active trachoma among children, an intensive participatory strategy to change hygiene behaviour was undertaken in central Tanzania ( McCauley et al. 1990 ; Lynch et al. 1994 ). This succeeded in increasing the prevalence of clean faces sevenfold (from 4% to 27%) over 12 months. Facial cleanliness was assessed on two days for each survey; clean faces were defined as having only one or no ‘sleep’ in the eye and no nasal discharge or flies on the two days. This definition correlated with objective evidence of face-washing based on the use of invisible fluorescent cream applied to the subject's forehead ( Lynch et al. 1994 ).

Based on this result an intervention trial on the effect of face-washing on active trachoma in children was conducted in three pairs of Tanzanian villages ( West et al. 1995 ). One year after baseline children in the intervention villages were more likely to have a sustained clean face than those in the control villages (defined as the presence of a clean face on at least two of three post intervention surveys); OR 1.61 (0.94, 2.74). The 95% confidence interval for the difference includes one indicating that the difference was not statistically significant. There was no difference in the prevalence of all active trachoma cases (TF and TI) between intervention and control villages, but face-washing was associated with a lower prevalence of severe trachoma (TI) OR 0.62 (0.40, 0.97). When all participants from intervention and control villages were pooled, children who had a sustained clean face were less likely to have active trachoma than those who ever had a dirty face OR 0.58 (0.47, 0.72). This is the only published hygiene-based intervention trial against trachoma and it shows that even under intensive conditions, and after painstaking preparation, sustainable changes in hygiene-based behaviour were difficult to achieve. For this reason, face-washing may be better incorporated into the general promotion of hygiene rather than advocated as a control measure in its own right. The trial indicated, after controlling for confounders, that children who had a sustained clean face were less at risk of trachoma than those who had dirty faces. The factors that predict a sustained clean face need to be elucidated.

Environmental improvement – flies, cattle and latrines

The presence of flies has been associated with trachoma for at least 400 years. Duke-Elder (1965) cites the 1598 memoirs of the Bohemian Baron Harant of Poljitz and the papers of Lucien Howe (1888) as the first anecdotal and scientific evidence (respectively) of the involvement of flies in trachoma transmission. High fly densities have been associated with outbreaks of trachoma in Morocco ( Reinhards et al. 1968 ) and Egypt ( Maxwell Lyons & Abdine 1952; Hafez & Attia 1958) but they have been absent from other areas ( Taylor et al. 1985 ). When combined with antibiotic therapy, fly control reduced reinfection in Morocco, but was of little effect on its own ( Reinhards et al. 1968 ). Jones (1975) has shown that flies were capable of transferring fluorescein between children's eyes in Iran and later asserted that they were ‘most exquisite passive vectors of ocular discharge from one person to another’ ( Jones 1980). C. trachomatis has been cultured from flies after they were fed on heavily infected laboratory cultures ( Forsey & Darougar 1981) and identified by PCR on 2/395 Musca sorbens caught from the eyes of swab-positive children ( Emerson et al. 2000 ). The PCR technique used was a modification of that developed for use with ocular swabs ( Bailey et al. 1994 ) and this result is interpreted as evidence that it is possible for wild-caught flies to be carrying the bacteria and not an indication of vectorial capacity.

The presence of flies around the house was associated with a greater risk of trachoma in Tanzania ( Taylor et al. 1989 ; West et al. 1991 ), but the sugar-board method used to estimate the size of the fly populations ( Taylor 1988) was later shown to be unreliable ( Brechner et al. 1992 ). Working in Ethiopia and Somalia, De Sole (1987) has suggested that there is an association between cattle herding and trachoma on the basis that cattle encourage flies and flies transmit trachoma. There were obvious differences of ethnicity, behaviour and geographical location in the study groups rendering them incomparable, and the findings should be interpreted with care. Cattle herding as an occupation vs. physical presence of cattle in a village, appeared to be a risk factor for active trachoma in the Tanzanian study ( Taylor et al. 1989 ; West et al. 1991 ). This could be a marker for other socio-economic factors since children that herd cattle probably live a more traditional way of life than those who are involved in other activities such as schooling and may represent the poorer segments of society or marginalized ethnic groups. There are many species of fly in the domestic environment, but only the Bazaar fly, Musca sorbens, has been proposed as a vector of trachoma ( Hafez & Attia 1958; Saccà 1964; Jones1980; Emerson 2000). M. sorbens is a species complex, and although no true distribution maps exist, it has been recorded in Australia and all the African and Middle Eastern countries affected by trachoma ( Paterson & Norris 1970; Pont 1991). It is strongly attracted to pus, mucus, open sores and eye secretions on which it feeds aggressively. Catches of flies from children's eyes ( Hafez & Attia 1958; Emerson et al. 2000 ) have shown that a disproportionate number of females are caught. Emerson hypothesized that pregravid or part-gravid females that require protein-rich food for egg development are attracted to the discharge from infected eyes. This subset of flies may move appetitively between eyes, which would greatly enhance transmission potential.

A pilot intervention study was conducted in The Gambia to assess the contribution of eye-seeking flies to the transmission of trachoma. The study was conducted for three months in two pairs of villages; one in the wet season (= 566), the other in the dry season (= 358). Baseline characteristics of the village pairs were similar. One village from each pair was randomly assigned the intervention and flies were controlled with insecticide. Fly populations were monitored with multiple attractant traps and by catches of eye-seeking flies directly from children under five in the dry season. Spraying reduced the population of M. sorbens by 75% and reduced child eye-fly contact by 96%. Where flies were controlled, the number of new prevalent cases of trachoma at three months was 75% lower than in the comparison group (OR 0.25; 0.09, 0.64) ( Emerson et al. 1999 ). Fly control in this intervention was based on the use of insecticide, which was expensive and labour intensive, hence unsuitable as a long-term community-based control measure.

The larval medium for M. sorbens is faeces, and it shows a marked preference for human faeces over any other type ( Hafez & Attia 1958). It only uses human excreta available in the environment: larval stages have not been found in latrines and adults have not been caught emerging from them ( Curtis & Hawkins 1982). Since latrines effectively remove the larval habitat from the environment, they may be a suitable method to control M. sorbens and hence reduce trachoma transmission. Pit latrines have been shown to be protective against trachoma, at household level, in risk factor analyses in Egypt ( Courtright et al. 1991 ), Malawi ( Tielsch et al. 1988 ), Australia ( Tedesco 1980) and Ethiopia ( Zerihun 1997), but not in Tanzania ( Taylor et al. 1989 ). If the observation is explained in terms of diminished larval habitat for M. sorbens, it would imply that the flies do not move between households, which does not seem reasonable. Community-wide protection would seem more likely in the presence of latrines as the local population of M. sorbens would decline. The studies in which latrines were protective observed existing latrines built to fulfil a perceived need by the user. Providing latrines to communities where they do not exist without generating the perceived need may not have the desired effect of reducing faecal contamination of the environment.

The role of F and E in the SAFE strategy

A review of the evidence presented in the literature shows that there is currently no single specific tool for trachoma control that can be recommended for inclusion in the F and E part of SAFE. There is good evidence that children with sustained clean faces are less likely to be at risk of trachoma but the results of an intervention based on the promotion of face-washing show that on its own this had limited success ( West et al. 1995 ). Other studies have shown that the presence of an adequate water supply is associated with a reduced community prevalence of active trachoma, but although necessary, this is only a part of what is required and not the whole answer ( Prost & Négrel 1989). Latrines appear to be associated with a lower prevalence of active trachoma and the evidence that flies are vectors of trachoma has recently been strengthened by a pilot intervention study.

The fact remains that for GET to be a success there has to be an impact on transmission to reduce the community prevalence of active disease. This calls for a community approach to both the perception of the disease and its management. Trachoma is a community disease clustered by individual families, not a series of isolated cases ( Bailey et al. 1989 ). The need for provision of surgery and antibiotics is not in doubt, but on their own they are unlikely to be successful. When used rationally, antibiotics have a positive effect, but it is unrealistic to think that this can be maintained ( Mabey et al. 1991 ). The majority of cases of active trachoma are asymptomatic or mild, and affected people rarely seek treatment. This means that antibiotics need to be delivered to cases identified by health personnel, which has huge implications for health budgets, or via mass treatment programmes. When used as the only method to combat trachoma, antibiotics have not achieved lasting control ( Reinhards et al. 1968 ; Dawson et al. 1976 ; Bailey et al. 1993 ), and the projected required frequencies of mass administration of azithromycin ( Lietman et al. 1999 ) are daunting when applied to trachoma-endemic regions.

Success of GET will require a combined approach that couples surgery and the use of antibiotics with hygiene promotion and environmental improvement. The long-term solution will not be arrived at quickly and it will not be cheap. The need for trachoma control has to be incorporated into government policy in the affected areas and should impact on the ministries responsible for water and education in addition to health. The evidence suggests that it is likely that the provision of latrines and adequate water to the affected areas would be beneficial. Trachoma should be included with other diseases such as malaria and taught about in schools. Simple messages about the importance of good hygiene, the role of flies and how to identify trachoma could be incorporated in the science, home economics and civics curricula of both primary and secondary schools.

Simple, cost-effective and sustainable measures to control trachoma are urgently needed to support the GET programme. Research into the modes of trachoma transmission and meaningful risk factor analyses may suggest further areas for investigation. The reasons why latrines and clean faces are protective need to be clarified and the recent evidence that Musca sorbens is a likely vector needs to be verified and sustainable methods to control it must be developed.

Acknowledgements

This review was funded by the Edna McConnell Clark Foundation.

Ancillary