Renal function has been examined in a group of 77 subjects occupationally exposed to cadmium fume and dust, together with a referent group of 103 age- and socioeconomically matched subjects. Fourteen biochemical parameters were measured on each subject. Three different ways of combining the information from all 14 tests were used to identify those subjects with renal dysfunction. These were first to count the number of parameters in which a subject recorded an abnormal test result. Second, the z value was computed for each parameter for each person by comparison with the mean and standard deviation of a derived normal population; these z scores were then summed. Lastly a multivariate distance measure, Mahalanobis D2, was determined for each subject from the distribution of normal subjects. The three approaches showed a considerable degree of agreement in identifying subjects with renal dysfunction, but they also displayed complementary strengths and weaknesses. The consensus of the three techniques was then taken to define truly dysfunctional subjects and each of the 14 parameters, and some combinations of pairs of parameters were tested as to their sensitivity and specificity. For this group of subjects, it was not possible to improve greatly on the use of retinol binding protein on its own. Were a second parameter to be chosen, it would be desirable to choose one reflecting the glomerular filtration rate, but the absence of a suitable sensitive biological monitoring parameter precludes a firm recommendation.