Predictive values (positive and negative) of screening tests for six illicit drugs of concern in the workplace were estimated based on published information on test sensitivity and specificity and survey data on prevalence. Predictive value (positive) was the percentage of screenees with positive tests who really had the condition being screened (true positives). Predictive value (negative) was the percentage of screenees who tested negative who did not have the drug being screened (true negatives). Predictive value (positive) was strongly influenced by the specificity of a test and the prevalence of users (decreasing sharply at prevalences below 10 percent), and less strongly by sensitivity. The midpoints of the ranges of estimated sensitivity for amphetamines, barbiturates, cocaine, hallucinogens, marijuana, and opiates were 50, 53, 50, 98, 69, and 50 percent, respectively. The midpoints of the ranges of estimated specificity for the drugs were 50, 95, 97, 100, 86, and 95 percent, respectively. However, the actual ranges of both sensitivity and specificity were wide in most cases, except for hallucinogens. Prevalence of use varied from 0.5 percent of the population for opiates and hallucinogens to 11 percent for marijuana. Estimated most probable predictive values (positive) were 1, 8, 35, 100, 38, and 5 percent for the six drugs, respectively. However, the ranges of the estimated values were wide for all drugs except hallucinogens: 0 to 100 percent for amphetamines, cocaine, and opiates, 14 to 100 percent for marijuana, and 100 percent for hallucinogens. Predictive values (negative) were 96 to 100 percent, with narrow ranges except for the amphetamines. The authors conclude that predictive value will be low in unselected populations, given the current variability of laboratories, and that improvements in predictive value can come from more accurate confirmatory tests, ensuring laboratory proficiency, and restricting mass screening to populations of high prevalence.