During the year 1974, urinary beta 2-microglobulin (beta 2mu) was measured at monthly intervals using the first-morning urine sample of randomly selected individuals from the BEN affected village of Petka (416 persons) and from the nearby situated control village of Stubica (216 persons). Initial compliance was complete; over 90% of villagers had at least 10 tests performed. beta 2mu, as assessed by radial immunodiffusion (RID), was repeatedly (at least twice) positive in 12% and 1.4% of the populations of the endemic and control villages, respectively. Over the 15 years of follow-up (1974 to 1988), none from the control village developed BEN, while many medical records of the cohort exposed to BEN contained data suggestive of BEN. Death from/with BEN was used as a measure of outcome. Incidence density of 12 was 3.3 per 1000 person/years of observation (19/5723). A single positive beta 2mu test was a sensitive predictor of BEN death (sensitivity = 89.5%). Selecting two or more positive tests as the cut-off point, the specificity and positive predictive value were considerably increased. Using the sulfosalicylic acid test for detection of significant proteinuria, a similar level of validity indices was reached only by four testings.