Gentamicin nephrotoxicity: application of multivariate analysis

A retrospective study was conducted to identify patient risk factors associated with gentamicin nephrotoxicity using multivariate analysis. Patients were evaluated for nephrotoxicity using an algorithm that considers serum creatinine, previous general experience with the drug, alternative etiologic candidates, timing of events, drug concentrations or evidence of overdosage, dechallenge, and rechallenge. There were 220 patients who met the criteria for inclusion in the study (103 men and 117 women). Variables found not to be significant discriminators between toxic and nontoxic patients included height, weight, sex, dose, estimated peak gentamicin concentration, age, and initial renal function. Variables that were found to enter into the discriminant analysis included presence of complicating factors with nephrotoxic potential, number of concurrent potentially nephrotoxic drugs, intensity index or estimated trough gentamicin concentration, and the total dose or treatment duration. Two models were developed; the probability of correctly classifying a patient as toxic was 63.5% and 65.4%, and as nontoxic was 72.1% and 68.8% for the two models, respectively. Although the discriminant models developed were statistically significant, the discrimination between toxic and nontoxic patients was not satisfactory to allow development of a generalized predictive model. Multivariate techniques were found to be essential in allowing for the differentiation of variables causing gentamicin nephrotoxicity.