Mississippi Mud No More: Cost-Effectiveness of Pharmacokinetic Dosage Adjustment of Vancomycin to Prevent Nephrotoxicity

William Darko, Pharm.D., Joseph J. Medicis, Pharm.D., Adrienne Smith, Pharm.D., Roy Guharoy, Pharm.D., David F. Lehmann, M.D., Pharm.D.


Pharmacotherapy. 2003;23(5) 

In This Article


A decision analysis framework was used to assess the cost-effectiveness of pharmacokinetic monitoring and dosage adjustment of vancomycin compared with no monitoring. Decision analysis is a quantitative economic method of estimating clinical and financial outcomes of alternative clinical management strategies. By calculating the value of each possible outcome and weighing the probability of each outcome, decision analysis evaluates each intermediate point of the model and identifies the sequence that will maximize benefits, minimize costs, or offer the cost/outcome produced as well as the probability of each outcome.

To evaluate the cost-effectiveness of pharmaco-kinetic dosage adjustments of vancomycin to prevent nephrotoxicity, we developed a decision analysis model using DATA 3.5 for Healthcare decision software (TreeAge Software, Inc., Williamstown, MA). The reference case was determined in part by a retrospective review of 200 patients randomly selected from a 3-year period of pharmacokinetic consultation service. Patients selected were aged 18 years or older and had received intravenous vancomycin for at least 48 hours, with at least two -- one peak and one trough -- vancomycin serum concentrations obtained during therapy. The decision analysis did not consider patients who were younger than 18 years, admitted while receiving intravenous vancomycin, receiving intravenous vancomycin for prophylaxis, or administered only oral vancomycin. Patient demographics are presented in Table 1 .

The model was based on the health-system perspective. For each patient, the frequency of vancomycin serum trough concentrations less than or equal to 10 mg/L and greater than 10 mg/L was determined. All initial vancomycin regimens were prescribed without pharmacokinetic intervention. The study group consisted of patients whose pharmacokinetic monitoring and dosage adjustment were performed and implemented after the first set of peak and trough levels was obtained. The frequency of trough concentrations less than or equal to 10 mg/L and greater than 10 mg/L was then calculated. For the control group, patients served as their own controls through application of pharmacokinetic calculations to the initial regimen prescribed with the assumption that no dosage changes would be made without monitoring.

Through application of patient-specific pharmacokinetic parameters to the initial regimen, expected peak and trough serum concentrations were calculated based on the initial vancomycin regimen prescribed for each patient and assumed to be carried throughout their treatment course. We believe this assumption is valid because dosage would not be adjusted in the absence of serum drug concentration monitoring. In both groups, pharmacokinetic calculations allowed for determination of frequency of serum trough concentrations less than or equal to 10 mg/L and greater than 10 mg/L ( Table 2 ). Probabilities of nephrotoxicity for patients with such serum levels were derived from results of published clinical trials ( Table 3 ).[15,16,17,18,19,20,21,22,23]

The probabilities of nephrotoxicity data were derived solely from more recent clinical trials in which vancomycin was purified. A MEDLINE search from January 1985-December 2001 was performed to identify all controlled studies of nephrotoxicity caused by vancomycin alone or in combination with other nephrotoxic agents ( Table 3 ). Analysis was performed by subtracting the difference in the rollback for probabilities of nephrotoxicity from the control and study group. These data on nephrotoxicity allowed for a marginal analysis to measure the efficacy of achieved trough serum concentration by the denominator of nephrotoxic episode prevented. A marginal analysis calculates the difference between the final probability for the control and study groups with which a nephrotoxic episode would occur. A sensitivity analysis was then performed using the extremes (worst and best case) of probabilities of nephrotoxicity and in costs to calculate the range of cost-effectiveness.

Calculation of the cost of nephrotoxicity was performed based on charges derived from the hospital database for drug-induced renal failure. All diagnostic-related groups associated with drug-induced renal failure were included to calculate the mean and range of possible costs. Cost accrued in monitoring vancomycin levels was calculated from the following expenses: pharmacist time spent monitoring vancomycin pharmacotherapy, calculated from the annual salary of a clinical pharmacist at our institution; laboratory costs for analyzing peak and trough vancomycin serum concentrations; nursing time spent drawing blood and administering vancomycin; treatment of acute renal failure at our institution.

The cost of monitoring serum concentrations was derived from the cost of the serum vancomycin assay (drug assay kit, calibration, quality control, technician time, and other medical supplies) and the costs based on the amount of time spent by the nurses and clinical pharmacists performing these monitoring activities. Nursing time spent on each task (blood extraction and administration of drug) was assumed to be 5 minutes. Time spent by the clinical pharmacists was 40 minutes, which involved chart review, interpretation of results, pharmacokinetic calculations, and follow-up evaluation. Wages were derived from the institutional human resources database. All costs, calculated in U.S. dollars, were determined using the current charges at our institution. The standard 50% of charges was used as a proxy for costs. Personnel costs were calculated by multiplying the average wage/minute by the total number of minutes spent. Table 4 shows the details of the microcosting method.

Subgroup analysis identical to the analysis described above was performed for patients in the intensive care unit, those on the oncology service, and those receiving concomitant nephrotoxins (aminoglycosides, amphotericin, and acyclovir).