OBJECTIVE To identify risk factors for dental fluorosis that cannot be explained by drinking water fluoride concentration alone. METHODS Two hundred eighty-four Tanzanian children ages 9 to 19 (mean 14.0+/-SD 1.69), who were lifetime residents at differing altitudes (Chanika, 100 m; Rundugai, 840 m; and Kibosho, 1,463 m; Sites 1, 2, and 3 respectively) were examined for dental fluorosis and caries. They were interviewed about their food habits, environmental characteristics and use of a fluoride-containing food tenderizer known locally as magadi. Meal, urine, water and magadi samples supplied by the participants were analyzed for fluoride content. Urine samples were also analyzed for creatinine concentration. Four magadi samples from Sites 1 and 3 were analyzed for complete element composition. RESULTS Of the 13 water samples from Site 2, 10 contained > or =4 mg/L F, ranging from 1.26 to 12.36 mg/L with a mean+/-SD of 5.72+/-4.71 mg/L. Sites 1 and 3 had negligible water fluoride of 0.05+/-0.05 and 0.18+/-0.32 mg/L respectively. Mean TFI fluorosis scores (range 0-9) for Site 2 were high: 4.44+/-1.68. In Sites 1 and 3, which both had negligible water fluoride, fluorosis scores varied dramatically: Site 1 mean maximum TFI was 0.01+/-0.07 and Site 3 TFI was 4.39+/-1.52. Mean DMFS was 1.39+/-2.45, 0.15+/-0.73 and 0.19+/-0.61 at Sites 1, 2, and 3, respectively. There were no restorations present. Urinary fluoride values were 0.52+/-0.70, 4.34+/-7.62, and 1.43+/-1.80 mg/L F at Sites 1, 2, and 3, respectively. Mean urinary fluoride values at Site 3 were within the normal urinary fluoride reference value range in spite of pervasive severe pitting fluorosis. Meal and magadi analyses revealed widely varied fluoride concentrations. Concentrations ranged from 0.01 to 22.04 mg/L F for meals and from 189 to 83211 mg/L F for magadi. Complete element analysis revealed the presence of aluminum, iron, magnesium, manganese, strontium and titanium in four magadi samples. There were much higher concentrations of these elements in samples from Site 3, which was at the highest altitude and had severe enamel disturbances in spite of negligible water fluoride concentration. An analysis of covariance model supported the research hypothesis that the three communities differed significantly in mean fluorosis scores (P<0.0001). Controlling for urinary fluoride concentration and urinary fluoride:urinary creatinine ratio, location appeared to significantly affect fluorosis severity. Urinary fluoride:urinary creatinine ratio had a stronger correlation than urinary fluoride concentration with mean TFI fluorosis scores (r=0.43 vs r= 0.25). CONCLUSIONS The severity of enamel disturbances at Site 3 (1463 m) was not consistent with the low fluoride concentration in drinking water, and was more severe than would be expected from the subjects' normal urinary fluoride values. Location, fluoride in magadi, other elements found in magadi, and malnutrition are variables which may be contributing to the severity of dental enamel disturbances occurring in Site 3. Altitude was a variable which differentiated the locations.