There is interest in the part played by free radicals and glutathione in HIV-1 infection. Because glucose-6phosphate dehydrogenase (G6PD) is the main enzyme responsible for maintaining glutathione in its reduced form (GSH), we compared the survivals of HIV-1-infected people carrying the normal G6PD allele (Gd) or the Mediterranean variant (Gd). People with Gd have almost undetectable erythrocyte G6PD activity, and mononuclear cells from peripheral blood express 20–25% of the normal amount of enzyme. People with the defect are unable to maintain adequate concentrations of GSH and should therefore be more prone to oxidation of GSH after an induced oxidative stress such as that reported to occur during HIV-1infection. From 1983 to 1996, 323 consecutive HIV-1-positive men from Northern Sardinia, an area where about 8% of men have the Gd mutation, were screened for G6PD deficiency. 27 were Gd. No significant difference was found between those with and without the mutation in age at diagnosis, risk factors for HIV-1-infection, CD4 T cells count, or incidence and type of opportunistic infections. No haemolytic crises were observed in the Gd group during the follow-up period. Disease progression was estimated from the first HIV-1-positive test until December, 1996, or death. The median survival of the patients was 93 months for both groups (range 6–166 for the Gd and 8–151 for the Gd, respectively); -test for trend: p=0·18 (figure). Overall, 55 people with Gd (18·6%) and eight with Gd (29·6%) were dead by the end of the follow-up period. 98 people in the Gd group (33·1%), and ten in the Gd (37·0%) developed AIDS. Those with Gd did not progress to AIDS more rapidly than those with Gd (64·1 months for the Gd and 61·2 for the Gd, p=0·83). Likewise, G6PD deficiency did not affect survival in those who developed AIDS: they had a median survival of 82 months with Gd (range 6–159) and 71 months with Gd (range 8–124); test for trend: p=0·27 (figure). At least under the conditions of our study, oxidative stress is unlikely to play a large part in progression of HIV-1infection. The low concentrations of GSH reported by some census. As the ISTAT procedure could provide information on parents’ place of birth only for children living with both parents (92·6% of the number of children under 15) all the denominators were adjusted for this figure. The children were classified in three groups according to place of birth of the parents, as both born in Sardinia, only one born in Sardinia, or both born in other region of Italy. We also considered an independent source: the demographic files of Milan for the years 1989–93 including all the residents (around 11% of the region’s residents), and for each year we estimated the fraction of children under 15 with at least one Sardinian parent. The average figure was 3·6%, very close to the 3·8% regional ISTAT estimate. The figure shows the observed annual rates of IDDM incidence for children according to parents’ birth place: 26·1/100 000 (CI 7·1–66·8) for the four children with both parents born in Sardinia; 15·6/100 000 (CI 8·9–25·4) for the 16 children with only one parent born in Sardinia; and 7·8/100 000 (CI 7·1–8·5) for children with no parents born in Sardinia. Children with one parent born in Sardinia had a relative risk of developing IDDM of 2·0 (CI 1·2–3·3) compared with children with both parents born outside Sardinia. The relative risk for children born from two Sardinian parents was 3·4 (CI 1·3–9·0). The expected number of cases, assuming the same rate as the Sardinian population, is 5·3 for children of both parents born in Sardinia. On the other hand, the number of observed IDDM cases with only one Sardinian parent is about less than half the number of expected cases (35·2) by the Sardinian incidence rate. By means of a telephone interview with the cases we found only a very low exposure to the Sardinian environment (time spent in the island, supply of Sardinian food before the onset of IDDM, and diabetes family history). The same data were used to provide a Lombardy incidence rate of IDDM among children less than 15 years, estimated after validation by two other independent sources. The expected incident cases, obtained by the capture-recapture method, were 586·2, with an annual incidence rate of 9·5/100 000 (CI 8·8–10·3/100 000). Our results confirm a high IDDM incidence in Sardinianheritage children who live in a low IDDM incidence region.