Although it is widely recommended that serum levels of aminoglycoside antibiotics be monitored by assay, the justification for this approach has not been clearly presented. A number of studies indicate that serum levels of these agents cannot be predicted reliably on the basis of simple dosage formulae; the major confounding factors being abnormalities of renal function and of extracellular fluid volume in addition to less well defined variables such as fever and anaemia. The influence of haemodialysis and concomitant administration of carbenicillin further complicate dosage estimations in patients with renal insufficiency.On the basis of currently available data, it is reasonable to suggest an optimum range of 5 to 8&mgr;g/ml for peak serum levels of gentamicin. There are no reliable studies from which to derive a comparable value for trough (pre-dose) concentrations. The relative importance of peak and trough values for nephrotoxicity and ototoxicity is an unresolved subject of controversy. However, it seems possible that neither of these individual values, but rather the ‘area under the time-concentration curve’ is the major risk factor for toxicity.In view of the unpredictability of serum levels, especially in seriously ill patients in a fluctuating physiological state, periodical serum gentamicin assays should be performed. The main objective of these assays is to ensure that the peak serum levels attained are adequate, but not unnecessarily high.