You are on page 1of 3


Effect of fly control on trachoma and diar rhoea

Paul M Emerson, Steve W Lindsay, Gijs E L Walraven, Hannah Faal, Claus Bøgh, Kebba Lowe, Robin L Bailey

Summary infections over many years. In less-developed countries,

Background Domestic flies are accepted vectors of diarrhoea, young children are the reservoir of infection, and
but their role in trachoma transmission has never been transmission is clustered within villages,10–12 but the
mechanisms by which infection is spread are poorly
quantified and no study has shown that fly control decreases
understood. An intervention that lowers the frequency of
the prevalence of trachoma. We assessed the effect of fly
transmission is likely to lower the prevalence of trachoma-
control on public health in a pilot study in Gambian villages.
related blindness in the future.
Methods We studied two pairs of villages—one pair in the The WHO initiative for the global elimination of
1997 wet season, and one pair in the 1998 dry season. For trachoma by the year 2020 (GET 2020) aims to control
each pair, deltamethrin was sprayed for 3 months to control trachoma by eyelid surgery, antibiotic treatment, facial
flies in one village whilst the other was used as a control. Fly cleanliness, and environmental improvement (the SAFE
populations were monitored with traps. We surveyed trachoma strategy). However, the implementation and sustainability
at baseline and at 3 months, and collected daily data on of antibiotic therapy is beset by poor compliance with
diarrhoea in children aged between 3 months and 5 years. topical application of ophthalmic tetracycline eye
Findings Fly control decreased numbers of muscid flies by ointment13 and by the prohibitive cost of oral azithromycin.
around 75% in the intervention villages compared with Trachoma is associated with poverty, and largely
controls. Trachoma prevalence was similar at baseline (wet disappears under improved environmental conditions and
season, prevalence in intervention village 8·8% vs control access to better sanitation.14 Nonetheless, the prospect of
12·2%; dry season, 18·0% vs 16·0%), but after 3 months of rapid global improvement remains bleak. The SAFE
fly control there were 75% fewer new cases of trachoma in the strategy would be strengthened by the inclusion of realistic
and sustainable methods to reduce trachoma transmission.
intervention villages (wet season 3·7% vs 13·7%; dry season
We undertook a pilot study to investigate the role of
10·0% vs 18·9%; rate ratio and relative risk of pooled data
domestic flies in the transmission of trachoma and
0·25 [adjusted 95% CI 0·09–0·64], p=0·003). There was 22%
less childhood diarrhoea in the wet season (14% vs 19%,
period prevalence ratio 0·78 [0·64–0·95], p=0·01), and 26% Methods
less diarrhoea in the dry season (6% vs 8%; 0·74 Study population
[0·34–1·59], p=0·60) compared with controls. Our study took place in four villages over 3 months in the Sanjal
Interpretation Muscid flies are important vectors of trachoma area of The Gambia.15 One pair of villages was studied from
and childhood diarrhoea in The Gambia. Deltamethrin spray is September, 1997, to December, 1997 (the wet season), the other
pair was studied from January, 1998, to April, 1998 (the dry
effective for fly control and may be useful for reducing
season). The two pairs of villages were selected from 23 available
trachoma and diarrhoea in some situations, but further small villages with populations of less than 400 in the Sanjal
research on sustainable fly-control methods is needed. district on the basis of trachoma prevalence, lack of immediate
Lancet 1999; 353: 1401–03 access to primary health-care, homogeneous ethnic composition
See Commentary page 1376 (Wolof), and willingness to participate. The village pairs were
3 km and 12 km apart, respectively. One village from each pair
Introduction was arbitrarily assigned the study intervention (fly control), and
Flies act as mechanical vectors of many pathogens, but the other acted as a control. The study was approved by the joint
they are commonly overlooked in public health ethics committee of The Gambia government and the Medical
interventions. Childhood diarrhoea causes 3·3 million Research Council, and we obtained informed consent from the
villagers and guardians of children.
deaths worldwide per year, with an estimated 1 billion
cases of diarrhoea annually in children under 5 years.1 Flies Fieldwork
are known vectors of diarrhoea-causing pathogens, Flies were controlled by ultra-low-volume application (Hudson
particularly those requiring a minute inoculum such as Portapac, Chicago, USA) of 0·175% volume to volume
shigella.2–5 Chlamydia trachomatis causes trachoma, and has deltamethrin (Aqua K-Othrine, Agrevo, UK) within and up to
been found on flies fed on heavily infected laboratory 20 m outside each village. Control began with an “attack” phase
culture media.6 There is an association between flies and of spraying every 2 days for 2 weeks, and continued with a
trachoma,7–9 but as yet there are no field-based or “maintenance” phase of spraying twice weekly in the wet season
when the fly population was largest, or once weekly in the dry
epidemiological studies that show that flies are direct
season when the fly population was smaller. No adverse effects of
vectors of trachoma. spraying were recorded.
Trachoma is the main cause of preventable blindness Fly populations were monitored by four fish-baited traps placed
worldwide.10 Such blindness is more common in women in each village at an animal-tethering area, in a latrine, at the
than men11 and results from repeated ocular C trachomatis centre of a domestic compound, and at the main meeting point for
24 h every 2 weeks. To measure fly-eye contact in the dry season,
Medical Research Council, The Gambia (P M Emerson MSc, hand-net collections of eye-seeking flies were made fortnightly
G E L Walraven MD ); University of Durham, UK (S W Lindsay PhD); from ten seated children for 15 min. Flies that touched the
National Eye Care Programme, The Gambia (H Faal FRCOphth, children’s eyes were collected and taken to the laboratory for
K Lowe); Danish Bilharziasis Laboratory, Denmark (C Bøgh MSc, identification.
S W Lindsay); and London School of Hygiene and Tropical Medicine, The whole of each village community was screened for
UK (R L Bailey MRCP) trachoma at baseline and at 3 months by the same community
Correspondence to: Paul M Emerson, Medical Research Council ophthalmic nurse, who was unaware of the treatment status of
Laboratories, PO Box 273 Banjul, The Gambia each village. Screening was done by everting the upper eyelid and

THE LANCET • Vol 353 • April 24, 1999 1401


visually examining the tarsal plate with a torch and a 32·5 Wet season Dry season
binocular loupe. Eyes were graded according to the WHO Control Intervention Control Intervention
simplified scale,16 which classifies active trachoma as the presence village village village village
of follicular trachoma (five or more follicles >0·5 mm visible) or
Census population in 1993 285 334 185 217
intense trachoma (50% of tarsal plate obscured by inflammation). Number aged <10 years* 112 136 79 87
People with symptoms of trachoma were offered tetracycline eye Number aged 3–60 months 82 77 51 53
ointment, and those with trichiasis were referred to the district eye Number aged 3–60 months per family
clinic at which surgery was available. Mean 3·9 3·1 3·6 3·1
Median (range) 4 (1–11) 3 (1–7) 3 (1–7) 3 (1–9)
We collected data on diarrhoea for all children in the villages
Ocular examination
aged between 3 months and 5 years. Diaries were completed by At baseline 319 381 227 207
the mother or guardian of each child, who was asked to record At follow-up (%) 271 (85) 295 (77) 189 (91) 169 (74)
daily whether the child had had diarrhoea. We analysed the period % days diarrhoea data 90·9 91·6 93·7 90·3
prevalence (proportion of days on which diarrhoea reported) as a *In both ocular examinations.
measure of morbidity. Oral rehydration salts were given to Characteristics of study villages
children with diarrhoea, and any child with diarrhoea on
5 consecutive days was referred to the local health centre. 1993 census was 1020. 1124 people of all ages were
screened for trachoma at baseline, of whom 924 (82%)
Statistical analysis
were also screened at 3 months. Loss to follow-up, mainly
Our statistical analyses used EpiInfo (version 6) and SPSS
(version 6.1). Fly-count data, which were not normally owing to inclusion of temporary migrants in the baseline
distributed, were described by the median or adjusted geometric data, was similar for intervention and control groups (rate
mean. Counts from the four traps set on the same day were ratio for intervention vs control 1·13 [0·83–1·54]).
obtained for each village, and we calculated the adjusted geometric Data on trachoma prevalence (figure) shows that there
mean for the catch from these traps by adding 1 to each count, was no difference in the community prevalence of active
calculating the geometric mean of the resulting four counts and trachoma at baseline in either village pair (wet season
then subtracting 1, to avoid the difficulty of a count score of zero. intervention 26/295 [8·8%] vs control 33/271 [12·2%]; dry
Variables that were not normally distributed were compared by the season 34/189 [18·0] vs 27/169 [16·0]). The prevalence of
non-parametric Kruskal Wallis H test and the Wilcoxon matched- trachoma after 3 months was lower in the intervention
pairs signed-ranks test. Categorical outcomes, such as prevalence
village than in the control village in both seasons (wet
rates, were initially compared by x2 with Yates’ correction. We
used Mantel-Haenzel estimates of relative risk and rate ratios for season intervention 11/295 [3·7%] vs control 37/271
stratified data. Trachoma cases and episodes of diarrhoea cannot [13·7%]; prevalence ratio 0·27 [adjusted 95% CI
be assumed to occur independently within households, so we 0·08–0·93], p=0·04; dry season 19/189 [10·0%] vs 32/169
adjusted confidence intervals and p values for these outcomes to [18·9%]; prevalence ratio 0·53 [95% CI 0·25–1·12],
allow for correlation within households by Miettinen’s test-based p=0·09). Overall, there was a 61% lower community
approach.17 The Kruskal Wallis H statistic derived from ranking of prevalence of active trachoma in the intervention villages
household or family disease rates was used in the adjustment. associated with fly control (relative risk=0·39 [95% CI
0·20–0·77], p=0·007).
Results The relative risk of becoming a new active trachoma case
Entomology at 3 months from baseline was lower in the intervention
Muscid flies (Musca sorbens—bazaar fly: M domestica— villages than the control villages (wet season relative
housefly) were more abundant in the wet-season control risk=0·20 [95% CI 0·05–0·82], p=0·026; dry season 0·33
village than in the dry-season control village: the median [0·1–1·06], p=0·064). Overall the risk was 75% lower in
(range) of the adjusted geometric mean numbers of flies per villages where fly control was practised than in control
trap per day was 9·80 (2·29–37·3) versus 5·98 (2·41–38·3), villages (relative risk=0·25 [0·09–0·64], p<0·003).
respectively, for M sorbens; and 10·2 (2·5–102·9) versus
4·24 (1·14–12·05) respectively, for M domestica. Spraying Diarrhoea
resulted in 76% fewer M sorbens (2·15 [0·19–5·62], p=0·02) All resident children aged between 3 months and 60
and 57% fewer M domestica (4·44 [1·21–36·25], p=0·04) in
the wet-season intervention village than in the control
village. In the dry-season intervention village, 75% fewer
M sorbens (1·51 [0–5·17], p=0·002) and 71% fewer
M domestica (1·21 [0–5·31], p=0·006) were caught during
the spraying period than in the control village.
Handnet collections of eye-seeking flies were virtually
100% efficient when there were few flies, but up to 5% of
flies escaped when there were more flies in total. In the dry
season during the spraying period, 96% fewer flies were
collected from the eyes of children in the intervention
village than from those in the control village (adjusted
geometric mean of number of flies per child in 15 min 0·06
vs 1·54, 4/54 fly positive catches vs 38/49, x2=49·8,
p<0·001). 92% of eye-seeking flies were M sorbens and 8%
were M domestica. Other fly species, predominantly
Chrysomyia albiceps, were caught in the traps, but were
never recorded on eyes or faces.
Village communities were of similar size, age composition Prevalence of active trachoma before and after fly control in
(table), and ethnicity (Wolof). The total population in the study villages in wet and dry seasons

1402 THE LANCET • Vol 353 • April 24, 1999


months were recruited for diarrhoea surveillance, 263 for a whole year. The development and assessment of other
children in total. Data were obtained for over 90% of sustainable and cost-effective methods for the control of
possible child-days of observation, with no significant muscid flies, such as the provision of latrines, identification
differences between villages or age groups. The distribution and clearance of breeding sites, and assessment of the
of numbers of children under 60 months per family was feasibility of locally made traps, should be a research
not significantly different in the four villages (table). priority.
Children in the dry season had 58% less diarrhoea than Contributors
those studied in the wet season (period prevalence All investigators contributed to design and implementation of the study.
ratio=0·42 [0·23–0·76], p=0·003). The proportion of days Paul Emerson and Kebba Lowe did the fieldwork. Analyses were done by
on which diarrhoea was reported was significantly lower Robin Bailey, Steve Lindsay, and Paul Emerson. Paul Emerson and
Steve Lindsay wrote the paper, which all investigators edited.
with increased age (p for trend <0·001) in all the villages,
but there was no association with family size. Period Acknowledgments
We thank Keith McAdam for his support and for use of MRC facilities in
prevalence of diarrhoea was lower in the intervention
The Gambia; the people of Kumbija, Sinchu Palen, Pasi Mut, and
village than the control village in both seasons (wet season Samba Sotor for their willingness to participate; Dawda Joof and the North
14% vs 19%, relative risk=0·78 [95% CI 0·64–0·95], Bank Division District Health Team for their support; and Shabbar Jaffar
p=0·01; dry season 6% vs 8%, 0·74 [0·34–1·59], p=0·60). and Linda Williams at the London School of Hygiene and Tropical
Medicine for statistical advice. Deltamethrin was donated by Agrevo
Overall, in the villages with fly control there was 22% less Environmental Ltd. The study was funded by the UK Department for
childhood diarrhoea in the wet season and 26% less in the International Development (DFID).
dry season than in the control villages.
Discussion 1 Benn C, Martines J, de Zoysa I, Glass RI. The magnitude of the global
In our pilot study the control of muscid flies lowered the health problem of diarrhoeal disease: a ten year update. Bull World
transmission of trachoma and the period prevalence of Health Organ 1992; 70: 705–14.
diarrhoea in children. However, our study was done in 2 Cohen D, Green M, Block C, et al. Reduction of transmission of
shigellosis by control of houseflies (Musca domestica). Lancet 1991; 337:
only four small settlements and may have been affected by 993–97.
bias—for example, realisation that there were fewer flies at 3 Fotedar R, Banerjee U, Singh S, Shriniwa S, Verma AK. The housefly
certain times might have made the mothers report less (Musca domestica) as a carrier of pathogenic micro organisms in a
diarrhoea and led the nurse to make differential hospital environment. J Hosp Infect 1992; 20: 209–15.
observations of trachoma. We cannot exclude this 4 Monzon RB, Sanchez AR, Tadiaman BM, et al. A comparison of the
role of Musca domestica (Linnaeus) and Chrysomya megacephala
possibility, but it does not seem likely to have affected our (Fabricius) as mechanical vectors of helminthic parasites in a typical
findings, since the diarrhoea data show similar age trends slum area of metropolitan Manila. S E Asian J Trop Public Health 1991;
in all villages and the nurse worked consistently at 22: 222–28.
validation sessions. 5 Wright EP. The isolation of Campylobacter jejuni from flies. J Hyg 1983;
91: 223–26.
Collections of flies directly from children’s eyes 6 Forsey T, Darougar S. Transmission of chlamydiae by the housefly.
confirmed that despite comprising less than 10% of the Br J Ophthalmol 1981; 65: 147–50.
total fly population, M sorbens made over 90% of fly-eye 7 MacCallan AF. The epidemiology of trachoma in Egypt. Br J
contacts and is therefore the most likely insect vector of Ophthalmol 1931; 15: 369–411.
trachoma. M domestica made the remainder of the eye 8 Reinhards J, Weber A, Nizetic B, Kupka K, Maxwell-Lyons F. Studies
in the epidemiology and control of seasonal conjunctivitis and trachoma
contracts and must be considered a potential vector. We in Southern Morocco. Bull World Health Organ 1968; 39: 497–545.
tested 395 flies caught from the eyes of swab-positive 9 Taylor HR. A simple method for assessment of association between
children by PCR for the C trachomatis common cryptic synanthropic flies and trachoma. Am J Trop Med Hyg 1988; 38: 623–27.
plasmid,18 and we identified successfully two M sorbens 10 Thylefors B, Negrel AD, Pararajasegaram R, Dadzie KY. Global data
carrying the bacteria—supporting evidence of their role as on blindness. Bull World Health Organ 1995; 73: 115–21.
11 West SK, Muñoz B, Turner VM, Mmbaga BBO, Taylor HR. The
vectors. Our field records of fly behaviour show that M epidemiology of trachoma in central Tanzania. Int J Epidemiol 1991; 29:
domestica and M sorbens are strongly attracted to human 1088–92.
faeces and prepared food, and are probably mechanical 12 Bailey RL, Osmond C, Mabey DCW, Whittle HC, Ward ME. Analysis
vectors of diarrhoea in The Gambia. of the household distribution of trachoma in a Gambian village using a
Ultra-low-volume spraying with deltamethrin decreased Monte Carlo simulation procedure. Int J Epidemiol 1989; 18: 944–51.
13 Mabey DCW, Downes RM, Downes B, Bailey RL, Dunn DT. The
fly-eye contacts by 96%, although if fly-catching was less impact of medical services of trachoma in a Gambian village: antibiotics
efficient in control villages with more flies, this proportion alone are not the answer. Ann Trop Paediatr 1991; 11: 295–300.
may be an underestimate. Fly control resulted in 75% 14 Dolin PJ, Faal H, Johnson GJ, et al. Reduction in trachoma in a sub-
fewer new cases of trachoma in intervention villages than in Saharan village in absence of a disease control programme. Lancet 1997;
349: 151–12.
control villages, and a 61% (95% CI 23–80) decrease in
15 Greenwood BM, Pickering HA. A review of the epidemiology and
point prevalence of trachoma in intervention villages, control of malaria in The Gambia, West Africa. Trans R Soc Trop Med
which compares favourably with supervised face-washing.19 Hyg 1993; 87 (suppl 2): S2–11.
However, fly control with insecticides is unlikely to be a 16 Thylefors B, Dawson CR, Jones BR, West SK, Taylor HR. A simple
sustainable routine public-health measure in countries in system for the assessment of trachoma and its complications. Bull World
Health Organ 1987; 65: 477–83.
which trachoma prevalence is highest. Spraying would be a
17 Rothman KJ. Modern Epidemiology. Boston: Little Brown, 1986:
useful adjunct to trachoma-elimination strategies or during 145–47.
trachoma outbreaks, and our study suggests that spraying 18 Bailey RL, Hayes LJ, Hampton TJ, Ward ME, Mabey DCW.
would also decrease morbidity from diarrhoea. A definitive Polymerase chain reaction for the detection of ocular chlamydial
study on the role of flies in the transmission of trachoma infection in trachoma-endemic communities. J Infect Dis 1994; 170:
would need to include external assessment of trachoma,
19 West S, Muñoz B, Lynch M, et al. Impact of face-washing on trachoma
perhaps with photographs, a larger study population, and a in Kongwa, Tanzania. Lancet 1995; 345:
larger number of villages. Such a study would need to run 155–58.

THE LANCET • Vol 353 • April 24, 1999 1403