Although immunocompromised persons may be at increased risk for gastrointestinal illnesses, no trials investigating drinking water treatment and gastrointestinal illness in such patients have been published. Earlier results from San Francisco suggested an association (OR 6.76) between tap water and cryptosporidiosis among HIV + persons. The authors conducted a randomized, triple-blinded intervention trial of home water treatment in San Francisco, California, from April 2000 to May 2001. Fifty HIV-positive patients were randomized to externally identical active (N = 24) or sham (N = 26) treatment devices. The active device contained a filter and UV light; the sham provided no treatment. Forty-five (90%) of the participants completed the study and were successfully blinded. Illness was measured using 'highly credible gastrointestinal illness' (HCGI), a previously published measure. There were 31 episodes of HCGI during 1,797 person-days in the sham group and 16 episodes during 1,478 person-days in the active group. The adjusted relative risk was 3.34 (95% CI: 0.99-11.21) times greater in those with the sham device. The magnitude of the point estimate of the risk, its consistency with recently published observational data, and its relevance for drinking water choices by immunocompromised individuals support the need for larger trials.