Diarrhoeal disease is a leading cause of childhood morbidity and mortality worldwide. A large part of this disease burden is attributed to enteric viruses. This was a two-year study of viral gastroenteritis among children (outpatients and inpatients) receiving care at the Queen Elizabeth Central Hospital in Blantyre, Malawi. Enzyme immunoassays were used to examine faecal samples obtained from children with acute, dehydrating gastroenteritis for the presence of rotavirus, as well as two other recognised viral causes of paediatric gastroenteritis, astrovirus and enteric adenovirus (serotypes 40 and 41). Maternal infection with the human immunodeficiency virus (HIV) is common in Malawi, and maternal-infant transmission of HIV has resulted in the birth of large numbers of HIV-infected infants. A group of children with rotavirus gastroenteritis, with and without HIV infection, was selected for short-term follow-up (maximum 1 month) after discharge from hospital. The effect of HIV infection on the duration of rotavirus shedding in stool, and serum immuneresponse to rotavirus, could therefore be determined. Furthermore, the effect of rotavirus on host HIV disease (measured by CD4 T lymphocyte counts and plasma HIV load) was explored. A second major objective of this work was to examine the extent of rotavirus strain diversity in Blantyre. Rotaviruses were therefore examined by polyacrylamide gel electrophoresis (PAGE) for rotavirus electropherotype, and by hemi-nested reverse transcription-polymerase chain reaction (RT-PCR) for rotavirus VP7 genotype (G type) and VP4 genotype (P type). Between July 1997 and June 1999, a total of 1186 children were enrolled into the study, comprising 786 inpatients and 400 outpatients. Rotavirus was detected equally among inpatients (35.8%) and outpatients (35.0%), but was identified less frequently among HIV-infected children (98/336, 29.2%) than among HIV-uninfected children (323/850, 38.0%) [p = 0.005]. Significant differences in rotavirus severity were not observed between hospitalised children with and without HIV infection. Astrovirus and adenovirus 40/41 were each detected in less than 5% of children, and their distribution did not differ by HIV status. Among children with rotavirus gastroenteritis selected for follow-up, HIV-infected children were significantly more likely to die than HIV-uninfected children, and death was related to low CD4 count at presentation. Of 74 children with rotavirus infection who completed follow-up, HIV-infected children shed rotavirus longer than HIV-uninfected children, but this was not associated with diarrhoea. Three-quarters of children exhibited a four-fold rise of IgG or IgA to rotavirus, which did not vary by HIV status. The duration of shedding, and seroresponse to rotavirus, were unrelated to CD4 count and HIV load. In addition, neither CD4 count nor HIV load were significantly affected by rotavirus infection. A total of 414 rotaviruses were characterised by electropherotype, G type and P type. The following strain types were identified: P[8], G1 (n= 111, 26.8%), P[6], G8 (n = 110, 26.6%), P[8], G3 (n = 93, 22.5%), P[4], G8 (n = 31, 7.5%), P[8], G4 (n = 21, 5.1%), P[6], G3 (n = 12, 2.9%), P[6], G1 (n=7, 1.7%), P[6], G9 (n = 3, 0.7%), P[6], G4 (n = 3, 0.7%), P[4], G3= 1, 0.2%), mixed (n = 15, 3.6%) and non-typeable (n = 7, 1.7%). The majority of serotype G8 strains, and all serotype G9 strains, had short electropherotype profiles. All remaining typeable strains had long electropherotypes. Surprisingly, G8 was the most frequently identified VP7 type (n =144, comprising 34.8% of strains) and was associated with VP4 types P[6] and P[4], thus representing novel combinations of the two, surface neutralisation proteins VP7 and VP4. Two short electropherotype, serotype G8 strains were therefore characterised further. The VP7 gene of a representative P[6], G8 strain (MW23) and of a representative P[4], G8 strain (MW333) exhibited greatest homology to human and bovine serotype G8 rotaviruses. The VP4 gene of MW23 was closely related to genotype P[6] rotaviruses identified recently in the United States, and the MW333 VP4 gene was most closely related to published genotype P[4] strains from the Philippines and Australia. The gene encoding NSP4 of both strains was classified in NSP4 genetic group I. Antigenic analysis by subgrouping and virus neutralisation confirmed complete antigenic characterisation of MW23 as subgroup I, P2A[6], G8 and MW333 as subgroup I, P1B[4], G8. Both prototype strains may have been derived by genetic reassortment between rotaviruses of human and bovine origin. Considerable diversity was also identified within other G and P types. Because of multiple mutations in the typing primer binding regions of the VP7 gene of type G1 strains, and the VP4 gene of type P[8] strains, alternative primers were designed to type many divergent strains in this collection. In particular, extensive diversity was recognised among the VP8* gene fragment of P[8] strains, which clustered phylogenetically into three distinct lineages. While children infected with P[6],G8 rotaviruses were younger than children excreting other rotavirus strains, other features (HIV status, diarrhoea severity, duration of virus shedding, and seroresponse), did not differ significantly by strain type. However, these data are of great epidemiological importance since current rotavirus vaccines are not designed to protect against disease caused by serotype G8 strains. Furthermore, the ability of HIV-infected children to clear rotavirus from the stool and mount an immune response to the infection, regardless of immune status and strain type, should encourage the evaluation of rotavirus vaccines in this vulnerable group of children. This would represent a significant step forward in attempts to reduce the dreadful burden of diarrhoeal disease that afflicts children in Malawi and other developing countries.