Safety margins and variability in the composition of glucose and maize-based oral rehydration therapy (ORT) prepared by mothers in a rural district of Western Kenya, are reported here. In a 2-year longitudinal field study, packets containing glucose ORS and a home solution made with maize and table salts, were provided to the mothers of 6180 children in Kakamega District in two separate sub-locations. Experienced and trained field workers supported by community registered nurses provided training for the preparation and use of ORT during weekly visits to every household. On four occasions separated by 4-6 months, a 5 per cent random sampling was done of the home-prepared solutions actually used for the treatment of children with diarrhoea. The water used for preparing ORT was also sampled. Samples of 174 glucose-based ORT, 148 maize-salt ORT, and 201 samples of water were analysed. Only 2 per cent of the maize-based ORT were above 120 meq/1 sodium (i.e., the safe range of sodium concentration) compared to 17.8 per cent for glucose solutions (p < 0.001). Home water samples contained substantial amounts of salt, which could unpredictably affect the final composition of the ORT solutions. We conclude that maize-salt ORT had a better margin of safety than glucose-based ORS.