Population pharmacokinetics of valganciclovir prophylaxis in paediatric and adult solid organ transplant recipients

Abstract

Aim

Our aims were to quantify ganciclovir pharmacokinetics in paediatric and adult kidney, liver and lung transplant patients taking a range of valganciclovir doses to prevent herpes virus infections, including a 450 mg regimen, and to identify sources of pharmacokinetic variability.

Method

Plasma samples were collected at 2, 4, 8 and 12 weeks post-transplant and at 4, 6, 8 and 12 months post-transplant in subjects prescribed longer courses. Ganciclovir was measured by liquid chromatography/ultraviolet detection. Non-linear mixed effects modelling was used to analyze the concentration–time data and evaluate demographic and transplant-related covariates.

Results

A two compartment model with first order absorption best described the data. Given the range of body sizes, clearance and volume of distribution terms were scaled using standard weight-based allometric exponents. Creatinine clearance was included on apparent oral clearance. Final estimates in a standard 70 kg individual for apparent oral clearance, central volume of distribution, intercompartmental clearance and peripheral volume of distribution were 14.5 l h⁻¹, 87.5 l, 4.80 l h⁻¹ and 42.6 l, respectively. The median terminal half-life for kidney, liver and lung transplant recipients was 9.4, 9.5 and 8.2 h, respectively. Median exposure (i.e. AUC(0,∞) in subjects taking valganciclovir 900 mg or 450 mg once daily was 57.4 and 34.3 μg ml⁻¹ h, respectively.

Conclusion

Allometric scaling allowed simultaneous analysis of data from children and adults. Ganciclovir pharmacokinetics were similar among kidney, liver and lung transplant recipients. Ganciclovir exposure after valganciclovir 450 mg once daily may be suboptimal in some individuals and requires evaluation along with virologic outcomes data.

WHAT IS ALREADY KNOWN ABOUT THIS SUBJECT

• Valganciclovir is widely used to prevent herpes virus infections in adult and paediatric solid organ transplant patients.

• Ganciclovir pharmacokinetics following valganciclovir prophylaxis at standard doses have been reported; however, exposure data for novel low dose regimens are limited as are pharmacokinetic data for children and lung transplant recipients.

WHAT THIS STUDY ADDS

• Allometric scaling allowed simultaneous analysis of ganciclovir data from children and adults.

• Ganciclovir pharmacokinetics in lung transplant recipients are similar to parameters determined in other transplant types.

• Valganciclovir 450 mg once daily can lead to suboptimal ganciclovir exposure and requires evaluation along with virologic outcomes data.

Introduction

Valganciclovir is the oral prodrug of ganciclovir, a nucleoside analogue DNA polymerase inhibitor. It was developed to improve the bioavailability of ganciclovir, which is 10-fold higher compared with oral ganciclovir capsules [1]. Valganciclovir is approved for the prevention of cytomegalovirus (CMV) disease in select adult and paediatric solid organ transplant patients [1]. Valganciclovir is also often administered off label to at risk paediatric solid organ transplant patients to prevent Epstein-Barr virus (EBV)-associated tissue invasive disease. Ganciclovir has good in vitro activity against both viruses. In CMV isolates from immunocompromised patients including kidney and bone marrow transplant recipients, the ganciclovir concentrations needed to inhibit CMV replication by 50% (IC₅₀) started at 0.65 μm [2]. For EBV, the IC₅₀ was reported to be as low as 0.05 μm [3]. These in vitro concentrations are attainable at valganciclovir doses used clinically.

The recommended dose and duration of valganciclovir for CMV prevention is 900 mg once daily within 10 days of transplantation and until 100 days post-transplantation in adult heart or kidney-pancreas transplant recipients or until 200 days post-transplantation in adult kidney transplant recipients [1]. In kidney or heart transplant recipients 4 months to 16 years of age, the dose of valganciclovir is determined according to an algorithm that incorporates body surface area and creatinine clearance. Similar to adult patients, the drug is administered within 10 days of transplantation and until 100 days post-transplantation [1]. Valganciclovir is not approved for use in other organ transplant types although it is routinely used in liver and lung transplant recipients. The dose of valganciclovir for EBV prophylaxis is analogous to that given for CMV disease prevention. However, the period of administration is longer, with high risk children often receiving the drug for 1 year or more. In recent years, concerns over drug-related side effects, mainly bone marrow suppression, and late occurring CMV disease have led to studies of lower doses of valganciclovir prophylaxis (i.e. 450 mg once daily) and longer periods of administration [4–14]. The pharmacokinetics of ganciclovir following these alternative dosing strategies have not been thoroughly evaluated [15,16]. Furthermore, our knowledge of ganciclovir pharmacokinetics in lung transplant recipients is limited because few studies conducted to date have included this patient population [17,18]. The objective of the present study is to describe the pharmacokinetics of ganciclovir in a population of paediatric and adult kidney, liver and lung transplant recipients receiving valganciclovir prophylaxis and identify patient specific characteristics associated with pharmacokinetic variability. Importantly, this study characterizes ganciclovir pharmacokinetics and exposure in both children and adults, in lung transplant recipients, and in individuals receiving either standard dose or low dose valganciclovir prophylaxis regimens.

Methods

Subjects

This study was conducted at the University of Minnesota in collaboration with the Transplant Center at the University of Minnesota Medical Center, Fairview. The University of Minnesota Institutional Review Board approved the study. Written informed consent was obtained from volunteers 18 years of age or older and from parents or guardians of volunteers who were less than 18 years old. For children aged 7 to 17 years, written informed assent was also obtained. First time transplant recipients who were prescribed valganciclovir prophylaxis after transplantation were eligible for the study. The only exclusion criterion was prior receipt of a solid organ transplant. This study was conducted within a larger natural history study of viral infections and post-transplant morbidity to allow future exploration of ganciclovir exposure–response relationships.

Study design

This was a prospective population pharmacokinetic study. On four separate occasions at approximately weeks 2, 4, 8 and 12 during the standard 3 month valganciclovir prophylaxis period, a portion of blood collected at regular post-transplant clinic visits for routine monitoring of tacrolimus, ciclosporin, or sirolimus concentrations was used to measure ganciclovir plasma concentrations. Subjects taking valganciclovir prophylaxis for up to 12 months had additional ganciclovir concentrations measured at approximately months 4, 6, 8 and 12. For each blood sample, the draw time and valganciclovir dosing history, including the amount and the date and time of the last dose, were recorded using a system of mail-in postcards and telephone interviews.

Quantification of ganciclovir in human plasma

Plasma concentrations of ganciclovir were measured using a multi-drug reversed phase high performance liquid chromatography assay developed and validated in our antiviral pharmacology research laboratory [19]. Briefly, plasma (200 μl) was subjected to protein precipitation with perchloric acid. The injection volume was 25 μl of pH adjusted, protein free extract. The mobile phase consisted of 97% potassium phosphate buffer (25 mm at pH 6) and 3.0% acetonitrile (v/v). The stationary phase was a 3.0 mm × 150 mm YMC C8 reversed phase column (Waters Corporation, Milford, MA, USA). The flow rate was 0.4 ml min⁻¹ with detection at 250 nm. All analytic measurements were performed using an HP series 1100 LC system (Agilent Technologies, Palo Alto, CA, USA). The assay internal standard was penciclovir. Plasma standards ranged from 50 to 20 000 ng ml⁻¹. Accuracy and variability were determined using four quality controls (50, 100, 1000, and 10 000 ng ml⁻¹) measured in triplicate on 5 separate days. The lower limit of quantification (LLOQ) for this assay was 50 ng ml⁻¹. The limit of detection (LOD) for ganciclovir was 3 ng ml⁻¹ and was calculated using a signal : noise ratio of 3 : 1. A single factor analysis of variance was used for statistical analysis. Accuracy ranged from 89.5% to 106.4%. Within and between assay variability, expressed as CV%, ranged from 0.7% to 4.8% and 1.0% to 7.9%, respectively.

Population pharmacokinetic analysis

Ganciclovir plasma concentration–time data were analyzed using a non-linear mixed effects modelling approach to determine the pharmacokinetic parameters. nonmem® version 7.2.0 with PDx-POP® version 5.0 (Icon Development Solutions, Ellicott City, MD, USA), and Intel® Visual Fortran Compiler Professional Edition version 11.1 (Intel Corporation, Santa Clara, CA, USA) were used for data analysis. Perl-speaks-nonmem version 3.6.1 (Uppsala University, Uppsala, Sweden), R version 2.14.0 (Free Software Foundation, Inc., Boston, MA, USA) and Xpose version 4.0 (Uppsala University, Uppsala, Sweden) were used to visually explore the data, generate diagnostic plots and qualify the final model [20,21].

Data management

Ganciclovir plasma concentrations less than the LOD were excluded. However, concentrations below the assay's LLOQ but above the LOD were retained. Particular care was given to the management of outliers. As in any non-linear regression analysis, outlying concentrations can markedly influence the fit of the model and the parameter estimates. Ganciclovir plasma concentration–time data were visually inspected. A concentration was considered to be a potential outlier when the absolute conditional weighted residual (CWRES) value was above 3 standard deviations. In addition, a concentration that was highly inconsistent with the rest of that subject's data was carefully examined. All outliers and highly inconsistent concentrations were checked for data entry and documentation errors. If the data were confirmed, the model was re-run without these particular concentrations to assess their influence on the parameter estimates.

Structural model development and interindividual and residual variability

Several structural models were evaluated during the development process including one and two compartment models with and without absorption lag time. Oral doses of valganciclovir were administered into an absorption depot compartment. Absorption of valganciclovir was modelled as a first order process and conversion to ganciclovir in the central compartment was assumed to be instantaneous. A two compartment model with an absorption lag time was specified with ADVAN4 and TRANS4. The model was parameterized in terms of absorption rate constant (K_a), apparent oral clearance (CL/F), apparent central volume of distribution (V₂/F), apparent inter-compartmental clearance (Q/F), and apparent peripheral volume of distribution (V₃/F). Body weight and creatinine clearance were included as covariates in the base model in order to improve model stability. Studies in paediatric populations have shown body weight to affect both clearance and volume of distribution terms [22]. Weight was standardized to 70 kg and its effect on CL/F, V₂/F, Q/F and V₃/F was described by a fixed exponent power function using standard allometric scaling values of 0.75 and 1.0 on clearance and volume of distribution terms, respectively [22–24]. Ganciclovir is eliminated renally by glomerular filtration and active tubular secretion [1]. Creatinine clearance was calculated by the Cockcroft–Gault equation in subjects 18 years of age or older and by the Schwartz equation in subjects less than 18 years old [25–29]. After correcting for the body surface area standardization inherent when using the Schwartz equation, all creatinine clearance estimates were standardized to the approximate population median value (i.e. 60 ml min⁻¹) and adjusted by body weight. The effect of creatinine clearance on CL/F was described by a power function with the exponent value estimated. Between subject variabilities on CL/F, V₂/F, Q/F and V₃/F were modelled using an exponential error model. This imposed a log-normal distribution on the parameters with results expressed as a coefficient of variation (CV%). Residual unexplained variability was modelled by a proportional error model and expressed as a CV%. nonmem's first order conditional estimation with interaction (FOCE-I) method was implemented for all runs. Four significant digits were specified for the estimation procedure.

Covariate model development

Covariate data were obtained from the Transplant Outcomes Database and the Organ Transplant Tracking Record maintained by the University of Minnesota Transplant Information Services group. Covariate data consisted of continuous and categorical variables. The continuous variables were age, weight and creatinine clearance and were treated as piecewise constant values from the previous observation to the current data record. Weight was often obtained on the same day that blood was collected for ganciclovir determination. However, when it was not, the weights recorded immediately before and after the blood draw date were used for linear interpolation of weight on the sampling day. Serum creatinine was always measured on the same day as blood collection for ganciclovir determination; therefore, imputation of this laboratory value for creatinine clearance calculations was not needed. The categorical variables were recipient gender, transplant type (kidney, liver, lung, kidney/pancreas, pancreas or kidney/liver), donor source (living related, living unrelated or deceased), recipient race (self-identified white/Caucasian or other), induction immunosuppression (thymoglobulin, basiliximab or methylprednisolone), maintenance immunosuppression (tacrolimus, ciclosporin or sirolimus), valganciclovir formulation (tablet or liquid) and days post-transplant (≤30 days or >30 days). Transplant type was also evaluated as a dichotomous variable (kidney or non-kidney, liver or non-liver, and lung or non-lung). Age was tested as a dichotomous variable (paediatric when age was <18 years or adult when age was ≥18 years). Of note, all subjects also received mycophenolate as part of their maintenance immunosuppression regimen; therefore this drug was not included as a covariate.

Covariates were screened for their potential importance by assessing clinical relevance, evaluating graphically covariate–parameter relationships, conducting a generalized additive model (GAM) analysis, and implementing the likelihood ratio test during the process of stepwise forward inclusion and backward elimination. The likelihood ratio test, which is approximately χ² distributed, is the primary statistic used to compare two nested models, and the objective function value (OFV) is a measure of goodness of fit. During forward inclusion, each covariate is entered into the model one at a time, checked for significance and removed. The most significant covariate is kept in the model and the process is repeated with the remaining covariates. The next most significant covariate is kept in the model and forward inclusion is repeated until no significant covariates remain. A full model is constructed with all of the significant covariates from stepwise forward inclusion. During stepwise backward elimination, each covariate is eliminated from the full model one at a time, checked for significance and added back. The covariate with the smallest impact on the OFV is removed from the model first and backward elimination is repeated. The next least significant covariate is removed and the stepwise procedure is repeated until removal of any other covariate results in a statistically inferior model. The final full model consists of all significant covariates following stepwise forward inclusion and backward elimination. In the present study, covariates were added to the base model using a multiplicative power function. A statistically significant improvement during forward inclusion and backward elimination was defined as a change in the OFV of 7.88 units (P < 0.005, d.f. = 1).

Model qualification

The precision of the final parameter estimates was checked by a stratified bootstrap re-sampling technique involving 1000 runs. The original data were stratified into two groups by sampling frequency with sparse sampling defined as fewer than four samples per subject and less sparse sampling defined as four or more samples per subject. The analysis was restricted to resample within these two groups in order to produce bootstrap data sets that all contained subjects with sparse and less sparse data but with different compositions [30]. If this sparse/non-sparse stratification was not applied, the resulting bootstrap estimations often did not converge or provided parameter estimates that were extremely variable. The final model developed using the original data set was fitted to each of the bootstrap data sets to obtain median bootstrap parameter estimates and standard errors. Bootstrap runs that minimized successfully were used to compute 95% confidence intervals (CIs). The median bootstrap parameter estimates and 95% CIs were compared with the final estimates and 95% CIs from the original data set. Model performance was also assessed through standard and stratified visual predictive checks and normalized prediction distribution errors (NPDE). For the visual predictive checks, 1000 data sets were simulated using the final model parameters. No assumptions were made about the covariate distributions. The 5^th, 50^th and 95^th percentiles of the observed data along with 90% prediction intervals based on the simulated data were plotted and compared. The NPDE were plotted against the model predictions as well as against age to assess discrepancies in model fit between children and adults.

Ganciclovir exposure and half-life determinations

A range of valganciclovir doses was administered to the subjects; therefore the ganciclovir exposure metric of interest was the area under the concentration–time curve from time 0 to infinity (AUC(0,∞)). This metric was derived from the empirical Bayes estimates of an individual's CL/F parameter and the valganciclovir dose that subject received as F*DOSE/CL. In addition to AUC(0,∞), terminal half-life was calculated using the standard equation for a two compartment model.

Results

Study population

Between February 2010 and June 2011, 115 first time solid organ transplant patients (100 adults and 15 children) receiving valganciclovir prophylaxis enrolled in the study. Of these, 95 (82 adults and 13 children) provided suitable plasma samples for the population pharmacokinetic analysis. The median (range) baseline characteristics for the 95 participants are summarized in Table 1. Valganciclovir dosing among subjects was highly variable because both adults and children were included in the study and dosage recommendations for patients with impaired renal function were also followed [1]. Most subjects received the tablet formulation of valganciclovir at a dose of either 900 mg every 24 h or 450 mg every 12, 24 or 48 h. The oral solution was prescribed to eight subjects, seven of whom were children, at a dose of 350 mg, 300 mg, 270 mg, 225 mg, 150 mg or 75 mg every 24 h.

Table 1.

Baseline subject characteristics (n = 95)

	Median	Range
Children (n = 13)
Age
0 to 24 months (n = 3)	15	6–17
2 to 11 years (n = 4)	7	5–10
12 to 17 years (n = 6)	13	12–15
Weight, kg	33.0	6.9–61.1
Creatinine clearance* (ml min−1 1.73 m−2)	72.1	30.2–154
Adults (n = 83)
Age (years)	53	18–78
Weight (kg)	71.8	8.05–115
Creatinine clearance* (ml min−1)	60.7	29–108

	Number	Percent
Gender, n (%)
Male	60	63.2
Female	35	36.8
Race/ethnicity, n (%)
Caucasian/white	83	87.4
African American or Black	7	7.4
Asian	4	4.2
American Indian or Alaska Native	1	1
Organ transplant, n (%)
Kidney	54	56.8
Liver	24	25.3
Lung	11	11.6
Kidney/pancreas	4	4.2
Pancreas	1	1.1
Kidney/liver	1	1.1
Donor source, n (%)
Deceased	50	52.6
Living unrelated	20	21.1
Living related	25	26.3

^*Cockcroft–Gault method [25] for subjects 18 years of age or older: [(140 – age (years)) × weight (kg)]/[serum creatinine (mg dl⁻¹) × 72] all multiplied by 0.85 for females; Schwartz method 27–29 adjusted for body surface area for subjects less than 18 years old: [k × height (cm)]/[serum creatinine (mg dl⁻¹)] where k = 0.45 for subjects aged less than 2 years and k = 0.55 for boys aged 2 years to less than 13 years and girls aged 2 to 16 years.

Ganciclovir population pharmacokinetics

During the data screening process 64 out of 333 (19.2%) observations were removed from the dataset as described previously. Of these 64 concentrations, 21 (6.3%) were below the assay's LOD, six (1.8%) had an absolute CWRES value above 3 standard deviations and 37 (11.1%) were internally inconsistent with the rest of a subject's data collected at a similar time post-dose. Importantly, when retained in the data set these concentrations yielded parameter estimates that were completely inconsistent with values previously published in the literature. A total of 269 observations from 95 subjects comprised the final dataset and were used in the population pharmacokinetic analysis. A two compartment model with first order absorption was determined to be the most appropriate structural pharmacokinetic model for the data. The absorption lag time and absorption rate constant were fixed to values of 0.5 h and 3.0 h, respectively. Body weight and creatinine clearance adjusted for body weight were included as covariates in the base model as described previously. Inter-individual variability in CL/F was 33%. The data were not sufficient to support estimates of interindividual variability on V₂/F, Q/F and V₃/F. Residual variability was 33%. After assessing clinical relevance inspecting plots of covariate–parameter relationships and, considering Akaike values from the GAM analysis, additional covariates that appeared important to evaluate on CL/F using the likelihood ratio test were gender, age, induction immunosuppression, maintenance immunosuppression and transplant type. During stepwise forward inclusion no covariate was found to be significant, as the reduction in OFV was < 7.88 in every case. Therefore, stepwise backward elimination was not performed. Given these findings, the final pharmacokinetic model defined below included body weight-adjusted creatinine clearance on CL/F and body weight on CL/F, V₂/F, Q/F and V₃/F through standard allometric relationships.

Plots of the observations vs. the concentrations predicted under the final regression model and the observations vs. the concentration predictions based on individual parameter estimates were examined and deemed acceptable (results not shown). Eta-shrinkage for CL/F was 4.7%. Epsilon-shrinkage was 13.5%.

All parameter estimates, their relative standard errors and 95% CIs, as well as median bootstrap parameter estimates and 95% CIs are shown in Table 2. Successful estimation and covariance steps were achieved in 92% of the bootstrap runs. A visual predictive check plot for the final model is shown in Figure 1. The cluster of observations considerably outside of the upper 90% prediction interval represents one individual's data that were consistently high compared with the rest of the population. The visual predictive check plot was most useful around clusters of dense data (i.e. 0–5, 10–15 and 20–25 h post-dose) whereas bins with fewer data points (i.e. 15–20 and over 40 h post-dose) were less informative. There was no apparent difference in the visual predictive check plots stratified by sampling frequency (results not shown). The plot of NPDE vs. the individual predictions did not reveal any discrepancies in model fit (data not shown). The plot of NPDE vs. age is shown in Figure 2.

Table 2.

Ganciclovir population pharmacokinetic parameter estimates and bootstrap results

Fixed Effects Oarameter	Final estimate	Relative SE (%)*	95% CI	Bootstrap estimate (median)	Bootstrap 95% CI
CL/F (l h−1)	14.5	12.7	10.9, 18.1	14.7	11.9, 18.0
CLcr exponent	0.492	15.7	0.340, 0.644	0.490	0.347, 0.635
V2/F (l)	87.5	38.6	21.3, 154	84.3	53.1, 139
Q/F (l h−1)	4.80	46.7	0.410, 9.19	4.64	1.88, 21.3
V3/F (l)	42.6	37.6	11.2, 74.0	47.1	24.9, 177
Ka	3.0 (fixed)	–	–	3.0 (fixed)	–
Lag time	0.5 (fixed)	–	–	0.5 (fixed)	–
Random effects parameter	Final estimate	Relative SE (%)**		Bootstrap estimate (median)	Bootstrap 95% CI
Variability of CL/F	33.5%	11		33.2%	26.2%, 40.4%
Residual variability	32.7%	6.6		31.9%	27.9%, 36.3%

^*Relative SE (%) = 100% × SE/estimate.

^**Relative SE (%) = 100% × SE/estimate*0.5 for parameters reported as a CV%.

Figure 1.

Visual predictive check plot for full model in all subjects (n = 95). Open circles, observations; solid black line, median of the observations; dashed grey lines, 10^th and 90^th percentiles of the observations; light grey bars, 95% prediction intervals for the simulated 10^th and 90^th percentiles; medium grey bars, 95% prediction interval for the simulated median; dark grey bars, overlap between the 95% prediction intervals for the simulated 10^th and 90^th percentiles and the median

Figure 2.

Normalized prediction distribution errors (NPDE) vs. age

Derived ganciclovir metrics: exposure and half-life

The median (range) post hoc estimated individual ganciclovir AUC(0,∞) was 37.4 (11.1–161) μg ml⁻¹ h. The median (range) AUC(0,∞) among subjects taking valganciclovir 900 mg every 24 h (n = 9) or 450 mg every 24 h (n = 4) was 57.4 (30.9–162) μg ml⁻¹ h and 34.3 (11.1–70.3) μg ml⁻¹ h, respectively. The median (range) post hoc estimated individual terminal half-life for ganciclovir was 9.2 (4.2–24) h. The median terminal half-life was comparable for subjects receiving either valganciclovir 900 mg or 450 mg every 24 h (data not shown). The median (range) terminal half-lives for children and adults were 7.0 (4.2–11) h and 10 (7.1–24) h, respectively. The median (range) terminal half-lives for kidney, liver and lung transplant recipients were 9.4 (5.2–24) h, 9.5 (4.2–18) h, and 8.2 (7.7–14) h, respectively. The resulting median (range) AUC(0,∞) among kidney, liver, and lung transplant recipients was 38.6 (11.9–161) μg ml⁻¹ h, 26.4 (11.1–70.3) μg ml⁻¹ h and 56.5 (39.5–71.9) μg ml⁻¹ h, respectively. All lung transplant recipients received valganciclovir at a dose of 900 mg every 24 h resulting in a higher median AUC(0,∞) whereas the majority of adult kidney and liver transplant recipients received 450 mg every 24 h.

Discussion

This is the first population analysis describing ganciclovir pharmacokinetics after oral administration of valganciclovir prophylaxis in both paediatric and adult solid organ transplant recipients. A unique aspect of the study population was the inclusion of lung transplant recipients for whom there are limited ganciclovir pharmacokinetic data. In addition, individuals with normal renal function receiving valganciclovir at doses of either 900 mg or 450 mg every 24 h were included.

The data were well described by a two compartment structural pharmacokinetic model with lag time and first order absorption and elimination. This is consistent with previous population analyses in either paediatric or adult solid organ transplant recipients [16,18,31–34]. In a pilot study, we found that a one compartment model best described ganciclovir pharmacokinetics in eight paediatric kidney and liver transplant patients taking valganciclovir for the prevention of EBV disease [35]. This difference in structural models is likely due to the small sample size and limited concentration–time data available for analysis in the earlier study. In order to obtain a stable base model in the current study, weight adjusted creatinine clearance on CL/F and weight-based allometric scaling of CL/F, V₂/F, Q/F and V₃/F were included as covariates. All previous population pharmacokinetic studies identified creatinine clearance as an important covariate on CL/F [16,18,31–34]. This is expected given that ganciclovir is renally eliminated [1]. Body weight was also identified in earlier studies as a significant covariate on clearance and volume of distribution terms [16,18,31–34]. In our study weight was centred at 70 kg and standard fixed allometric exponents of 0.75 for clearance terms and 1.0 for volume distribution terms were applied. In a recent meta-analysis of 121 population pharmacokinetic models across drug classes, weight-based allometry with a fixed exponent of 0.75 was found to be the most common pre-defined relationship on clearance and was usually applied in study populations that included paediatrics or neonates [36]. When applying allometry to clearance there is controversy over whether or not exponents should be fixed to 0.75 or estimated independently for the drug of interest [37–39]. We considered this point given the age and weight ranges in our population and attempted to estimate the exponent. However, the model was over-parameterized and successful estimation and covariance steps could not be achieved. Perrottet et al. reported that ganciclovir clearance in adults was also influenced by gender and graft type with women having 23% slower clearance than men and with heart and lung or liver transplant recipients having 40% and 13% slower clearance, respectively, compared with kidney transplant recipients [18]. In three previous reports gender also influenced central volume of distribution with lower values in female adult solid organ transplant patients as compared with males [16,18,31]. This finding was attributed primarily to weight differences between the genders [31]. In our study, gender and organ transplant type were not significant during likelihood ratio testing. This was a clinical observational study that employed a sparse sampling study design. Therefore, the limitations of the data may have contributed to our inability to fit more complex models and identify other significant covariate relationships.

In general, ganciclovir population pharmacokinetic parameter estimates were in agreement with values reported in the literature. However, our population volume of distribution estimates were larger compared with other studies [16,18,31–34]. This could be due in part to the range of weights represented in the study as a result of including both children and adults. In addition, the sparse sampling design likely resulted in data with limited information about ganciclovir's distribution parameters under a two compartment model. An estimate of inter-individual variability was obtained for apparent oral clearance. Interindividual variability is expected for the other pharmacokinetic parameters as well. However, we were unable to estimate this variability given the limited information on ganciclovir's distribution kinetics as a result of the study's sampling design. Over 30% of the variability in clearance remains unexplained, which is reasonable for this pharmacokinetic parameter. The estimate of residual unexplained variability was approximately 33%, which is high and can be attributed to the clinical nature of the study design, the reliance on patient self-report for valganciclovir dosing histories, and the dynamic health status of the patients including routine post-transplantation drug therapy changes that could not be tested in the covariate model. Epsilon-shrinkage was less than 20% indicating no significant overfitting of the data. Eta-shrinkage was less than 5% for CL/F indicating that the individual estimates of this parameter are informative. The stratified bootstrap analysis confirmed the final model's robustness with 92% of successful estimation and covariance steps achieved during the 1000 bootstrap runs. The results for the visual predictive check plot implied that the final model and parameter estimates adequately described the observed data. In addition, the NPDE suggested that there were no discrepancies in model fit according to age.

The median post hoc terminal half-life for ganciclovir was 9 h. This is similar to values reported previously (i.e. 3–8 h) [1]. Median post hoc terminal half-lives were similar between children and adults and across transplant types. The median post hoc estimated individual ganciclovir AUC(0,∞) for subjects receiving valganciclovir 900 mg every 24 h was 57.4 μg ml⁻¹ h. This is consistent with the proposed range of AUC values (i.e. 40 to 50 μg ml⁻¹ h) needed to suppress CMV viraemia and lower the risk of post-transplant CMV disease [40]. All lung transplant recipients received valganciclovir 900 mg every 24 h and achieved a median AUC(0,∞) of 56.5 μg ml⁻¹ h. However most adult kidney and liver transplant recipients received valganciclovir 450 mg every 24 h, which is reflected in the lower median AUC(0,∞) values (i.e. 38.6 μg ml⁻¹ h and 26.4 μg ml⁻¹ h, respectively) observed in these subgroups. In general, the median AUC(0,∞) for subjects receiving valganciclovir 450 mg every 24 h was approximately 50% lower than following standard dosing and may not provide the same protection against CMV disease. This is speculative and virologic outcomes data are needed to evaluate this hypothesis. Importantly, ganciclovir's antiviral activity is actually due to the intracellular ganciclovir triphosphate moiety, which is difficult to measure clinically. Data on the correlation between plasma ganciclovir concentrations and intracellular ganciclovir triphosphate concentrations are lacking. However, as discussed above, efficacy data for CMV disease prevention has been linked to ganciclovir plasma AUC [40].

Most of the ganciclovir concentrations achieved in our study were above the reported in vitro IC₅₀ values for EBV. This is a reasonable comparison given that ganciclovir plasma protein binding is only 1 to 2% [1]. In addition, ganciclovir exposure targets for EBV disease prevention are unknown. Virologic outcome data on the effectiveness of valganciclovir EBV prophylaxis are limited. Recently, Hocker et al. prospectively compared the incidence of EBV viremia in a cohort of paediatric kidney transplant recipients on oral ganciclovir or valganciclovir prophylaxis (n = 20) with a similar control cohort not on antiviral prophylaxis (n = 8) [41]. Prophylaxis with oral ganciclovir or valganciclovir was associated with a significantly lower incidence of primary EBV infection (P < 0.0001) and a lower EBV viral load (P < 0.0001) in the first year post-transplant [41]. Ganciclovir pharmacokinetic data were not reported in this study.

The strengths of our study were its prospective nature and liberal inclusion criteria thus allowing a large number of covariates across a relatively wide range to be evaluated during the model development process. We also demonstrated that concentration–time data from children and adults could be analyzed simultaneously using allometric scaling of pharmacokinetic parameters. In addition, ganciclovir plasma pharmacokinetics following valganciclovir administration are similar among kidney, liver and lung transplant recipients. One limitation to this study was the sparse sampling study design. A more robust design might have improved population and individual estimates of ganciclovir's distribution kinetics under a two compartment model and allowed us to detect important covariate relationships present in this complex patient population. However, it was not possible to sample subjects intensively or require that samples be drawn within specific post-dose time windows because this study was conducted in conjunction with regularly scheduled post-transplant clinic and laboratory visits.

In conclusion, ganciclovir population pharmacokinetics in paediatric and adult solid organ transplant recipients receiving valganciclovir prophylaxis were well described by a two compartment structural model. Allometric scaling of pharmacokinetic parameters allowed the simultaneous analysis of concentration–time data from children and adults. The sparse sampling study design resulted in limited information about covariate–parameter relationships and interindividual variability in ganciclovir pharmacokinetics. Median ganciclovir exposure following valganciclovir 450 mg every 24 h was 50% lower than that achieved after standard dosing and may be suboptimal in some individuals. Ganciclovir pharmacokinetic data need to be evaluated along with clinically relevant virologic outcomes for CMV and EBV.

Competing Interests

All authors have completed the Unified Competing Interest form at http://www.icmje.org/coi_disclosure.pdf (http://www.icmje.org/coi_disclosure.pdf) (available on request from the corresponding author). H.E.V had support from the National Institutes of Health and the University of Minnesota Graduate School for the submitted work. There are no financial relationships with any other organizations that might have an interest in the submitted work in the previous 3 years and no other relationships or activities that could appear to have influenced the submitted work.