Background Several million children are killed each year by diarrhoeal diseases ; preventive strategies appropriate for developing countries are vital.

Despite strong circumstantial evidence that flies are vectors of diarrhoeal diseases, no convincing studies of the impact of fly control on diarrhoea incidence in developing countries have been reported.

We undertook a randomised study of the effect of insecticide spraying on diarrhoea incidence.

Methods Six study villages were randomly assigned to two groups.

Flies were controlled through insecticide application in group A in 1995 and in group B in 1996.

In 1997 the effectiveness of baited fly traps was tested in group A villages.

Diarrhoea episodes were monitored in children under 5 years through mothers'reports during weekly visits by a health visitor.

Fly density was monitored by use of sticky fly-papers hung in sentinel compounds.

Findings During the fly seasons (March-June) of both 1995 and 1996, insecticide application practically eliminated the fly population in the treated villages.

The incidence of diarrhoea was lower in the sprayed villages than in the unsprayed villages in both 1995 (mean episodes per child-year 6.3 vs 7.1) and 1996 (4.4 vs 6.5) ; the reduction in incidence was 23% (95% Cl 11-33, p=0.007).

At times other than the fly season there was no evidence of a difference in diarrhoea morbidity between sprayed and unsprayed villages. (...)