Trials have provided conflicting estimates of the risk of gastrointestinal illness attributable to tap water. To estimate this risk in an Iowa community with a well-run water utility with microbiologically challenged source water, the authors of this 2000-2002 study randomly assigned blinded volunteers to use externally identical devices (active device: 227 households with 646 persons; sham device: 229 households with 650 persons) for 6 months (cycle A). Each group then switched to the opposite device for 6 months (cycle B). The active device contained a 1-microm absolute ceramic filter and used ultraviolet light. Episodes of "highly credible gastrointestinal illness," a published measure of diarrhea, nausea, vomiting, and abdominal cramps, were recorded. Water usage was recorded with personal diaries and an electronic totalizer. The numbers of episodes in cycle A among the active and sham device groups were 707 and 672, respectively; in cycle B, the numbers of episodes were 516 and 476, respectively. In a log-linear generalized estimating equations model using intention-to-treat analysis, the relative rate of highly credible gastrointestinal illness (sham vs. active) for the entire trial was 0.98 (95% confidence interval: 0.86, 1.10). No reduction in gastrointestinal illness was detected after in-home use of a device designed to be highly effective in removing microorganisms from water.