Twenty-nine studies involving a total of 2,215 patients were included.
Lactose-containing versus non-lactose-containing diets.
Thirteen studies (873 patients) compared treatment failures. The RR of treatment failure for children on lactose-containing diets was 2.1 (95% confidence interval, CI: 1.6, 2.7), compared to children who did not receive lactose (p<0.0001). The significant heterogeneity between studies was explained by analysing studies on the basis of initial severity of dehydration (moderate or severe and mild), year of the study (reflecting the management approach), or definition of treatment failure. Children with mild or no dehydration, and those who are managed according to appropriate treatment protocols, have similar treatment failure whether they receive lactose or not.
Four studies (387 patients) compared stool frequency and 4 studies (209 patients) compared stool amount. There was significant heterogeneity between studies, and any effects identified were not clinically significant.
Nine studies (826 patients) compared duration of diarrhoea. There was significant heterogeneity between studies, even when studies allowing the addition of solid foods were excluded.
The data on weight gain were insufficient to be reliably assessed.
Undiluted lactose-containing milk versus the same milk at reduced concentration or introduced later during therapy.
Fourteen studies (934 patients) compared treatment failures. The RR for undiluted milk was 1.3 (95% CI: 0.9, 1.8) and non significant; the results of the heterogeneity test were also non significant (p=0.45). Analysing by initial severity of dehydration, the RR was significantly higher for the more severe cases (2.0, 95% CI: 1.2, 3.3), but non significant for milder cases.
Six studies (480 patients) compared stool frequency and 3 studies (272 patients) compared stool amount. The analyses found a slight increase in stool output with undiluted diets, but the differences were of minor clinical importance.
Ten studies (876 patients) compared duration of diarrhoea. There were no significant differences by dietary group. Removal of those studies permitting consumption of other foods did not alter this conclusion.
Seven studies (509 patients) compared change in body weight. There was a significant advantage of undiluted milk, with an average effects size of 0.25 standard deviations (p=0.002). The effect was homogeneous across studies (p=0.16).