Dosing of vancomycin in patients with continuous-flow left ventricular assist devices: A clinical pharmacokinetic analysis

BACKGROUND

Continuous-flow left-ventricular assist devices (CF-LVADs) have emerged as a life-saving therapeutic option in end-stage heart failure patients. Outcomes with these devices are continuing to improve and overall 24-month survival from recent registries has been above 70% (1).

Infections are a leading complication in these patients, particularly driveline infections which occur at a rate of up to 20% per patient year of support (2). Vancomycin is often required to treat Gram-positive pathogens (3). Device implantation, by improving organ perfusion and reducing congestion, may alter the pharmacokinetics of drugs. To date, there is no published literature describing drug disposition of any medication in patients after CF-LVAD implantation. The purpose of this study is to describe the pharmacokinetics of vancomycin in patients with chronic CF-LVAD support.

METHODS

Patients ≥18 years implanted with a Heart Mate II device (Thoratec, Pleasanton, CA, USA) who received vancomycin between January 2008 and April 2012 and who had at least two evaluable vancomycin serum concentrations were eligible for inclusion in this retrospective analysis. Exclusion criteria included unstable renal function prior to or during vancomycin treatment defined as ≥ Stage 1 acute kidney injury by AKIN criteria (4); hemodynamic instability during vancomycin treatment defined as vasopressor or acute inotrope requirements, cardiac surgery within five days of vancomycin treatment, and acute decompensation of heart failure (including right-ventricular failure). Waiver of informed consent was granted by the institutional review board.

The study objective was to compare the calculated first-order elimination rate constant (Ke) and volume of distribution (Vd) of vancomycin in compensated CF-LVAD patients to the population-derived estimates which are often used to determine initial vancomycin dosages. Calculated pharmacokinetic parameters were derived from peak and trough serum vancomycin concentrations drawn under steady state conditions using one-compartment model equations and body habitus definitions used in routine clinical practice (5-6-7).

Various methods for estimating Ke and Vd were tested in order to potentially identify the most accurate prediction for true values in CF-LVAD patients. Method 1 (EstKe1) used actual body weight (ABW) for underweight patients and ideal body weight (IBW) for all other patients for the Cockcroft-Gault (CG) equation (7). Method 2 (EstKe2) is a simplified variation on method 1, whereby a weight of 72 kg was used for all patients with the CG equation (8). For obese patients, Method 3a (EstKe3a) used the Salazar-Corcoran equation 9 and Method 3b (EstKe3b) utilized adjusted body weight (AdjBW) for the CG equation (7, 8). For estimating Vd, various methods were used according to the patient’s weight: Method 1 (EstVd _ABW), Method 2 (EstVd_AdjBW), and Method 3 (EstVd_IBW).

The primary analysis included pairwise comparisons between measured and estimated values for Ke and Vd. Planned subgroup analysis classified patients according to body habitus. Paired t-test or Wilcoxon signed-rank test was used for related continuous data, as appropriate. All eligible patients were included as a convenience sample. P values <0.05 were considered statistically significant.

RESULTS

Reasons for exclusion are shown in Figure 1. Baseline demographic and clinical characteristics for all patients and according to body habitus are reported in Table I. Estimated and measured pharmacokinetic parameters with mean paired differences for the various prediction methods are shown in Table II. All methods for predicting Ke provided significant overestimates. On average, Method 1 (EstKe1) and Method 2 (EstKe2) were more than 51% higher than the measured Ke (p = 0.003 for both comparisons). These findings remained consistent when stratified according to body habitus.

Methods for predicting the measured Vd tended to provide overestimates, although not with the same consistency as the EstKe methods (Tab. II). EstVd_IBW was the most accurate method regardless of body habitus, with only 10% error on average.

DISCUSSION

Our study suggests that population-based equations routinely used in clinical practice for estimating initial doses of vancomycin may result in excessive vancomycin doses for patients with CF-LVADs. Calculation of creatinine clearance according to the CG equation consistently overestimated elimination rate for all study patients regardless of body habitus. The Salazar-Corcoran equation for obese patients also provided significant overestimates of elimination rate. All population-based estimates of Vd resulted in overestimates, particularly those using actual body weight.

Several prior studies have described altered pharmacokinetics of various drugs as a result of the decreased organ perfusion and congestion associated with heart failure (11-12-13-14-15-16), but to the authors’ knowledge, our study is the first to examine medication disposition in patients with CF-LVADs. Recent literature confirms that vancomycin clearance decreases linearly with ejection fraction in patients with heart failure (17). We hypothesized that the improvement in hemodynamics and volume status from CF-LVAD implantation would impact drug disposition (18). However, similar to previous literature in non-LVAD patients (17), the CF-LVAD patients in our cohort demonstrated a reduced vancomycin clearance. In addition, the reduced Vd seen with other drugs in non-LVAD heart failure patients (11-12) from other studies was seen in our cohort of patients those supported by CF-LVADs. The reasons for this abnormality are not completely clear, but based upon these data it may be prudent to confirm these findings with a formal pharmacokinetic analysis.

Limitations of our study should be mentioned. Pharmacokinetic parameters were calculated using a one-compartment model from serum vancomycin samples collected during routine clinical monitoring. Although this method is less reliable than a two-compartment model using multiple serial serum concentrations (19), our results may be more reflective of the “real world” clinical setting where dosing is routinely based on one-compartment model, population-based equations. Additionally, this retrospective pilot study was neither designed nor able to evaluate novel dosing strategies or clinical outcomes. Also, the lack of a non-LVAD control group of heart failure patients prevents direct comparison with this population and our CF-LVAD cohort. Finally, as we took care in only evaluating compensated patients, these results should not be applied to decompensated heart failure patients with CF-LVAD support.

CONCLUSIONS

Our findings suggest that despite improvement in organ perfusion and relief of heart failure symptoms that result from CF-LVAD support, vancomycin pharmacokinetics are altered in these patients. Clinicians caring for CF-LVAD patients should anticipate reduced clearance and Vd when determining initial vancomycin dosing, and monitor serum drug levels very closely in these patients. Additional study using multiple serum vancomycin concentrations with two-compartment modeling and including non-LVAD heart failure patients is warranted in order to confirm these findings.