Development of a Pharmacokinetic Model of Mitotane: Toward Personalized Dosing in Adrenocortical Carcinoma
Thomas M. A. Kerkhofs, MD,* Luc J. J. Derijks, PhD, f Hester Ettaieb, MD,* Jan den Hartigh, PhD,¿ Kees Neef, PhD,§| Hans Gelderblom, PhD, | Henk-Jan Guchelaar, PhD, and Harm R. Haak, PhD ***
Background: Mitotane is the drug of choice in medical treatment of adrenocortical carcinoma. The antineoplastic effect seems to be correlated with a minimum plasma level of 14 mg/L, but plasma concentration build-up is in general slow due to the long elimination half-life. Consequently, the therapeutic effect sets in after weeks or even months. The objective of this study was to develop a pharma- cokinetic model that enables clinicians to adjust dosing based on a target drug exposure, which facilitates personalized therapy.
Methods: Data on dosing and plasma level measurements per- formed throughout mitotane therapy were retrospectively collected in a population of 29 patients from 2 hospitals. A population pharmacokinetic model was constructed based on data from 20 patients using iterative 2-stage Bayesian fitting (MWPharm). The model was validated in an independent sample of 9 patients.
Results: The concentration-time data were best described by a 3-compartment model. The model estimated mitotane clearance at 0.94 ± 0.37 L/h and a volume of distribution in the steady state at 161 ± 68 L/kg of lean body mass. The mean prediction error was 14% ± 13%.
Conclusions: A pharmacokinetic model was developed, which characterized mitotane by slow clearance and large volume of distribution. The model seems to be able to predict mitotane levels in individual patients with an error margin of 14%. The model enables one to adapt dosing based on individual plasma level measurements in prospective setting, which improves the accuracy of the prediction. We expect that individualization of mitotane dosing leads to anticipated and more rapid attainment of the therapeutic levels and potentially to improved clinical management of mitotane treatment.
From the Departments of *Internal Medicine; }Clinical Pharmacology, Máxima Medical Center, Eindhoven; ¿ Department of Clinical Pharmacy and Toxi- cology, Leiden University Medical Center; §Department of Clinical Phar- macy and Toxicology; CAPHRI, School for Public Health and Primary Care, Maastricht University Medical Center+; |Department of Clinical Oncology, Leiden University Medical Center; and ** Division of General Internal Medicine, Department of Internal Medicine, Maastricht Univer- sity Medical Center+, the Netherlands.
The authors declare no conflict of interest.
Correspondence: Thomas M. A. Kerkhofs, MD, Department of Internal Medicine, Máxima Medical Center, Ds. Th. Fliednerstraat 1, 5631 BM Eindhoven, the Netherlands (e-mail: t.Kerkhofs@mmc.nl).
Copyright @ 2014 Wolters Kluwer Health, Inc. All rights reserved.
Key Words: adrenocortical carcinoma, mitotane, pharmacokinetics (Ther Drug Monit 2015;37:58-65)
INTRODUCTION
Mitotane is the drug of choice in patients with adrenocortical carcinoma. It can be combined with cytotoxic chemotherapy in patients with extensive and/or rapidly pro- gressive disease and is recommended as an adjuvant therapy in patients at a high risk of recurrence.1-4 In fact, withholding mitotane should only be considered after radical resection of the primary tumor and ki-67 index <10%.4,5
Little is known about the pharmacodynamic and phar- macokinetic properties of mitotane.6 The antineoplastic effect seems to be correlated with mitotane plasma levels: several studies demonstrated an objective response in patients whose plasma levels were >14 mg/L, whereas lower levels were associated with a lower efficacy.7-9 The time to reach steady- state plasma concentrations takes months due to its long elim- ination half-life. In most patients, a steady state is reached after 3 months of treatment with a daily dose of about 6.0 g.10,11 Due to the slow build-up of mitotane plasma levels, the therapeutic effect sets in after several weeks or months of treatment. This complicates the timing of therapy evaluation and the decision to include cytotoxic chemotherapy to the treatment.
Dosing regimens are based on clinical experience and adjusted according to plasma concentration and tolerability. The inability to predict mitotane levels may lead to relative underdosing and a prolonged build-up phase in some patients, whereas others unexpectedly demonstrate high plasma levels early in therapy, causing increased toxicity. A recent study on mitotane initiation in a population of patients with advanced adrenocortical carcinoma describes the relationship between plasma levels and 2 different dosing regimens during the first 12 weeks of treatment.12 Despite a confounding effect of con- comitant chemotherapy and the fact that a steady state was not reached during the study, a trend toward higher plasma levels using a higher initial dose was described.
Research studies have indicated that individual genetic differences are significantly associated with mitotane levels achieved after 3 and 6 months of treatment.13 Also, a previous study demonstrated a semilogarithmic relationship between mitotane concentration in plasma and adipose tissue.14 This
| Total (n = 29) | Development Group (MMC Patients) (n = 20) | Validation Group (LUMC Patients) (n = 9) | P | |
|---|---|---|---|---|
| Age, median (range), yrs | 53 (20-76) | 54 (20-76) | 47 (33-72) | 0.741 |
| Gender, [n (%)] | ||||
| Male | 13 | 8 (40%) | 5 (56%) | 0.688 |
| Female | 16 | 12 (60%) | 4 (44%) | |
| BMI, median (range), kg/m2 | 23.8 (18.1-40.8) | 25.1 (18.1-40.8) | 23.0 (19.4-31.2) | 0.322 |
| Treatment setting, [n (%)] | ||||
| Adjuvant | 10 | 7 (35%) | 3 (33%) | 0.890 |
| Monotherapy | 8 | 5 (25%) | 3 (33%) | |
| Combined with chemo | 11 | 8 (40%) | 3 (33%) | |
| Time on mitotane, median (range), mos | 11 (1-54) | 11 (2-54) | 11 (1-24) | 0.587 |
| Total dose administered, median (range), g | 1491 (121-3685) | 1704 (210-3685) | 1140 (121-2928) | 0.157 |
| Number of measurements, median (range) | 14 (4-28) | 15 (5-28) | 11 (4-23) | 0.202 |
| Time to reach 14 mg/L, median (range), d | 116 (21-321) | 116 (50-212) | 137 (21-321) | 0.923 |
| Dose needed to reach 14 mg/L, median (range), g | 626 (168-1621) | 628 (214-1621) | 691 (168-1097) | 0.815 |
P value for comparison between groups (Mann-Whitney U test).
suggests that individual body fat percentages influence the time to reach therapeutic mitotane plasma levels.
To our knowledge, there are no studies describing mitotane plasma levels and maintenance doses in long-term users. Also, there are no studies describing the influence of basic patient characteristics, such as gender, age, and body mass index (BMI), on mitotane pharmacokinetics.
The aim of this study was to develop a pharmacoki- netic model of mitotane taking into account clinical patient characteristics. This model should enable clinicians to adjust dosing based on a target drug exposure, which facilitates personalized therapy.
METHODS
Patients
This retrospective study was carried out among patients from the Máxima Medical Center (MMC) and Leiden Univer- sity Medical Center (LUMC), who were treated with mitotane between 2002 and 2012. To be eligible for this study, dosing information from the start of the therapy had to be available. All dose modifications and all plasma level measurements were extracted from the medical records. All measurements were routinely performed throughout the course of mitotane therapy. Further, the following patient characteristics were extracted: age, gender, weight (baseline), height, and the setting in which mitotane was administered (adjuvant therapy, monotherapy, or in combination with cytotoxic chemotherapy). All mitotane plasma concentrations were determined by gas liquid chro- matography at the Department of Clinical Pharmacy and Tox- icology in the LUMC.15 In the Netherlands, anonymous use
Copyright @ 2014 Wolters Kluwer Health, Inc. All rights reserved.
of retrospective clinical data is permitted without explicit informed consent from the patient.
Modeling
The model was created based on the medication history of 20 patients from MMC. Data were analyzed using MW-Pharm Computer Aided Therapeutic Drug Monitoring version 3.81 (Mediware, Zuidhorn, The Netherlands).16 This program is able to calculate 1-, 2-, and 3-compartmental models using iterative 2-stage Bayesian fitting.17,18 Before the actual modeling starts, rough estimates of the model parameters and their SD are entered. In the first stage of modeling, the software calculates individual pharmacokinetic parameters ±SD that best fit the actual plasma measure- ments in every given patient. The previously entered esti- mates are used as the starting point. In the second stage, all individual values are pooled resulting in (1) a population mean for every parameter, (2) a covariance matrix with val- ues for interparameter associations, and (3) an estimate of the residual SD. Then, the cycle is repeated using the newly found population values as the starting point. The calcula- tion is finished when the new population values and the residual SD are similar (0.001% difference) to the values from the previous cycle.
Drug clearance was normalized according to the equa- tion kel = kelm + kelr. CLcr, where kel is the total elimination rate constant, kelm is the metabolic elimination rate constant, kelr is the renal elimination rate constant, and Cler is the cre- atinine clearance. The model includes the covariates age, weight, height, gender, body surface area (BSA), and lean body mass (LBM). Influence of the covariates on the phar- macokinetic parameters was calculated by performing
30-
Mitotane plasma level (mg/L)
20
10-
0
0
200
400
600
800
A
Time (days)
30
Mitotane plasma level (mg/L)
20.
10.
0
0
200
400
600
800
1000
B
Time (days)
30-
Mitotane plasma level (mg/L)
20
10-
0
0
200
400
600
800
C
Time (days)
| Model Parameter | Estimate |
|---|---|
| V1 (mean ± SD), L/kg·LBM | 1.216 ± 0.511 |
| kelm (mean ± SD), hr-1 | 0.014 ± 0.001 |
| k12 (mean ± SD), hr-1 | 0.150 ± 0.010 |
| k21 (mean ± SD), hr-1 | 0.002 ± 0.0001 |
| k13 (mean ± SD), hr-1 | 0.023 ± 0.001 |
| k31 (mean ± SD), hr-1 | 0.0004 ± 0.0001 |
| ka(po) (mean ± SD), hr-1 | 0.005 ± 0.005 |
| F(po) (mean ± SD) | 0.30 |
Estimates are accurate to 4 decimals and truncated to 3 decimals. The fourth decimal is displayed if the value would otherwise have been zero.
ka(po), Absorption rate constant (oral administration); kelm, elimination rate constant; k12, rate constant from compartment 1 to compartment 2; k13, rate constant from com- partment 1 to compartment 3; k21, rate constant from compartment 2 to compartment 1; k31, rate constant from compartment 3 to compartment 1; Fpo, bioavailability (oral administration); V1, volume of distribution in the central compartment.
regression analysis of each individual parameter against each patient covariate and was displayed using the correlation coefficient (r). The modeling approach based on rate con- stants combines both size- and function-related parameters, which potentially complicates the evaluation of correlations. However, this approach resulted in a better fit to the data than a model in which clearance was individualized according to body weight, although there was no difference in the strength of correlations between both models.
For the matter of simplification, twice and thrice daily doses were cumulated and registered as a single daily dose at a fixed time point (08:00 AM). Based on the extremely long half-life of mitotane, we do not expect this to be of significant influence on the outcome. Also, all plasma measurements were presumed to have been performed at fixed time points (08:00 AM). It is a stan- dard procedure in both hospitals to perform mitotane sampling in the morning at least 12 hours after the last dose. The estimate for bioavailability was fixed in model calculations. Based on the existing literature, this value was set at 30%.19
Validation was performed by comparing predicted mitotane plasma levels with observed levels using 3 different methods: (1) by fitting the model to individual patients (n = 9) from an independent group who were trea- ted in another hospital (LUMC), (2) by fitting the model to individual patients (n = 20) from the same development group, and (3) by fitting the model to individual patients (n = 100) from a simulated population that was generated using the Monte Carlo method. This latter technique randomly assigned individual pharmacokinetic parameters ±SD to virtual patients based on a “standard” patient (male, age 55 years, height 1.75 m, and weight 75 kg) with a medication history of 2 years, monthly plasma measurements during the first year and bimonthly measurements during the second year.
Definitions
The prediction error was quantified by calculating the proportional difference of the predicted value relative to the observed value as follows:
Copyright @ 2014 Wolters Kluwer Health, Inc. All rights reserved.
Prediction error
V(
predicted - observed
2
×100). .
observed
BMI was calculated as (weight/height2). LBM was calcu- lated as 50.0 + 0.9 (height - 152) in males and 45.5 + 0.9 (height - 152) in females.2º BSA was calculated as (weight0.425 . height0.725.0.007184).21 The volume of distribu- tion in the steady state (Vss) follows from the equation
Vss = V1 x (1 +412 + K13)
k12
k21
k31
RESULTS
Patient Characteristics
Data on mitotane treatment of patients from the devel- opment group and the LUMC validation group are specified in Table 1. There were no significant differences in age, gender, BMI, and the setting in which mitotane was administered. Figure 1A-C displays plasma concentration-time curves both from real populations and the virtual Monte Carlo population. There were 302 plasma level measurements in the development group and 112 measurements in the validation group. All patients except for 2 (1 from every group) reached the thera- peutic level during mitotane treatment. Cumulative dose was 2640 g for simulated patients in the Monte Carlo group.
Model Characteristics
The concentration-time data were best described by a 3- compartment model. Parameters of the model are summarized in Table 2. The mean volume of distribution in the central compartment (V1) was 1.216 ± 0.511 L/kg.LBM and Vss was 161 ± 68 L/kg. LBM. Analysis of covariates showed that weight is significantly correlated with V1 (r = - 0.505), k21 (r = 0.478), and ka(po) (r =- 0.528); height with k12 (r =- 0.487) and k21 (r = 0.607); gender with k12 (r = 0.555); BSA with kelm (r = 0.446), V1 (r = - 0.576), k21 (r = 0.626), and ka(po) (r= - 0.509); LBM with V1 (r = - 0.469), k12 (r = - 0.600), and k21 (r = 0.688).
Prediction and Treatment Simulation
Analysis of observed and predicted values in the validation group resulted in a mean prediction error of 14% ± 13%. Figure 2A displays a goodness-of-fit plot and a plot of weighted residuals versus time of the validation group. Analysis of the development group resulted in a mean pre- diction error of 25% + 39%, goodness-of-fit and weighted residuals are displayed in Figure 2B. In the Monte Carlo population, the mean prediction error was 6% ± 6%.
Three individual examples from the validation group of good, average and poor fits are given in Figure 3A-C. Mean prediction error in a patient with a good fit was 11% ± 11%; in a patient with an average fit, it was 14% ± 14%; and in a patient with a poor fit, it was 19% + 16%.
Bayesian fitting of the model to individual patients in the LUMC validation group resulted in a mean clearance of
30
6-
Observed mitotane plasma concentration (mg/L)
4.
20-
€
2.
Weighted residuals
0
20€
400
600
800
10-
-2.
-4.
0
0
10
20
30
-6
A
Predicted mitotane plasma concentration (mg/L)
Time (days)
30
6-
Observed mitotane plasma concentration (mg/L)
4
20
2.
Weighted residuals
0
200
400
600
…
800
1000
10-
-2.
-4.
0
0
10
20
30
-6
B
Predicted mitotane plasma concentration (mg/L)
Time (days)
0.94 ± 0.37 L/h. Mean clearance was 1.12 ± 0.41 L/h among patients in the MMC development group.
DISCUSSION
This study was aimed at developing a pharmacokinetic model of mitotane using data from patients on long-term treatment. The data were best described by a 3-compartment model. The model enables individualization of mitotane dosing in a prospective setting based on plasma-level measurements.
Two previous studies prospectively investigated mitotane pharmacokinetics during the first 3 months of treatment.12,22 In the first study, a high dose strategy was deployed in 22 patients, median cumulative dose administered within 3 months was 405 g (range, 157-546 g).22 Ten out of 22 patients (45%) reached the therapeutic level of 14 mg/L within 3 months. In the second study, a low-dose strategy was prospectively compared with a high-dose strategy in 40 patients.12 Median cumulative doses in both regimens were 495 g (range, 163-593 g, high dose) and 242 g (range, 97-541 g, low dose), respectively. Fourteen out
Copyright @ 2014 Wolters Kluwer Health, Inc. All rights reserved.
20
Mitotane plasma level (mg/L)
15.
10.
5
0
0
200
400
600
800
A
Time (days)
20
Mitotane plasma level (mg/L)
15-
10
5-
0
0
100
200
300
400
B
Time (days)
20
Mitotane plasma level (mg/L)
15-
10-
5-
0
0
50
100
150
200
C
Time (days)
of 32 patients who completed the study (44%) reached the therapeutic level within 3 months: 10/20 from the high dose regimen and 4/12 from the low-dose regimen. These results illustrate that plasma concentration build-up is highly variable among patients. The strength of the proposed pharmacokinetic model is adaptability to individual patients in daily clinical practice by taking into account individual plasma level meas- urements. This learning aspect improves the accuracy of indi- vidual predictions and facilitates tailored dose adjustments to reach and maintain the therapeutic level within a certain period of time. We expect that this leads to early dose increments in patients who require high mitotane doses, which speeds up the attainment of the therapeutic plasma level. Also, patients who display a rapid increase in mitotane plasma level can be iden- tified. Timely dose reductions in these patients are expected to prevent an overshoot to potentially toxic levels. In patients with advanced disease, cytotoxic chemotherapy is usually added in the case of disease progression and/or failure of mitotane therapy. Timing of chemotherapy initiation can be complicated if therapeutic levels of mitotane have not been reached yet, because a therapeutic effect could still be expected and chemo- therapy might not be necessary at that time. We expect that improved prediction of mitotane levels leads to rapid and antic- ipated attainment of 14 mg/L and therefore to adequately timed and more accurate therapy evaluation.
The model performed well in predicting mitotane levels in a validation group from another hospital. Distribution of observed versus predicted values indicates a slight tendency toward underestimation of plasma levels in or above the therapeutic range, predominantly in the development group. In a prospective setting, this could lead to dose increments where continuation would be appropriate or dose continuation where reduction would be appropriate. However, this is assumed to be less harmful to patients than the opposite, which could result in subtherapeutic levels. The plasma concentration-time data contained several outliers, that is, surprisingly high mitotane levels during the maintenance phase of therapy. The model could not adequately account for these values because mitotane plasma levels usually increase gradually over the course of weeks. In most cases, a satisfying clinical explanation could not be found. We hypothesize that the outliers were caused by medication in- teractions or undocumented weight loss triggering redistribu- tion of mitotane to the central compartment.
Previous research indicates that mitotane is a highly lipophilic drug.19 The estimated volume of distribution is relatively high, and the estimated clearance is relatively low. These findings are compatible with the lipophilicity of the drug and suggest that the plasma concentration is deter- mined by distribution processes rather than elimination pro- cesses. Previous observations of fatty tissue concentrations that were approximately 200-fold higher than plasma concen- trations are in agreement with this hypothesis.14 Analysis of covariates showed considerably weak correlations (r = - 0.50) between weight (both total weight and LBM) and V1. Other covariates (age, gender, height, and BSA) displayed correla- tions in the same order of magnitude. We hypothesize that factors not accounted for in the model, particularly differences in genetic constitution, are of importance to explain the residual
variation of mitotane pharmacokinetics in the population. This assumption is supported by recent research suggesting that a polymorphism in the gene coding for the CYP2B6 enzyme is associated with higher plasma concentrations after 3 months of treatment.13
We acknowledge the limitations of our study. We propose a pharmacokinetic model based on retrospective data. The predictive value will have to be confirmed in a pro- spective setting. In both hospitals, the standard operating procedure for mitotane level measurement is to assess trough levels. Due to the retrospective nature of the data, we cannot guarantee that all levels were trough levels. Selection bias might be of influence, as all patients (except 2) had reached the therapeutic level during treatment. Because there were no patients with liver or kidney failure, we could not assess the influence of these factors on mitotane pharmacokinetics. Finally, interindividual dietary differences could have influ- enced the model through drug absorption.
CONCLUSIONS
A 3-compartment pharmacokinetic model of mitotane was developed, which estimated mean Vss at 161 ± 68 L/kg.LBM and mean mitotane clearance at 0.94 + 0.37 L/h. The model seems to be able to predict mitotane levels in individual patients with reasonable accuracy considering an error margin of 14% ± 13% in an independent validation group. Residual variance in population pharmacokinetic parameters remains to be eluci- dated. Future research should be aimed at exploring the genetic factors involved in mitotane pharmacokinetics.
REFERENCES
1. Fassnacht M, Allolio B. Clinical management of adrenocortical carci- noma. Best Pract Res Clin Endocrinol Metab. 2009;23:273-289.
2. Baudin E, Leboulleux S, Al Ghuzlan A, et al. Therapeutic management of advanced adrenocortical carcinoma: what do we know in 2011? Horm Cancer. 2011;2:363-371.
3. Fassnacht M, Terzolo M, Allolio B, et al. Combination chemotherapy in advanced adrenocortical carcinoma. N Engl J Med. 2012;366:2189-2197.
4. Terzolo M, Angeli A, Fassnacht M, et al. Adjuvant mitotane treatment for adrenocortical carcinoma. N Engl J Med. 2007;356:2372-2380.
5. Berruti A, Baudin E, Gelderblom H, et al. Adrenal cancer: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann Oncol. 2012;23:vii131-vii138.
6. Hahner S, Fassnacht M. Mitotane for adrenocortical carcinoma treatment. Curr Opin Investig Drugs. 2005;6:386-394.
7. Haak HR, Hermans J, van de Velde CJ, et al. Optimal treatment of adrenocortical carcinoma with mitotane: results in a consecutive series of 96 patients. Br J Cancer. 1994;69:947-951.
8. Baudin E, Pellegriti G, Bonnay M, et al. Impact of monitoring plasma 1,1-dichlorodiphenyldichloroethane (o,p’DDD) levels on the treatment of patients with adrenocortical carcinoma. Cancer. 2001;92:1385-1392.
9. van Slooten H, Moolenaar AJ, van Seters AP, et al. The treatment of adrenocortical carcinoma with o,p’-DDD: prognostic implications of serum level monitoring. Eur J Cancer Clin Oncol. 1984;20:47-53.
10. Faggiano A, Leboulleux S, Young J, et al. Rapidly progressing high o, p’DDD doses shorten the time required to reach the therapeutic threshold with an acceptable tolerance: preliminary results. Clin Endocrinol (Oxf). 2006;64:110-113.
11. Terzolo M, Pia A, Berruti A, et al. Low-dose monitored mitotane treatment achieves the therapeutic range with manageable side effects in patients with adrenocortical cancer. J Clin Endocrinol Metab. 2000;85:2234-2238.
12. Kerkhofs T, Baudin E, Terzolo M, et al. Comparison of two mitotane starting dose regimens in patients with advanced adrenocortical carci- noma. J Clin Endocrinol Metab. 2013;98:4759-4767.
Copyright @ 2014 Wolters Kluwer Health, Inc. All rights reserved.
13. D’Avolio A, De Francia S, Basile V, et al. Influence of the CYP2B6 polymorphism on the pharmacokinetics of mitotane. Pharmacogenet Genomics. 2013;23:293-300.
14. von Slooten H, van Seters AP, Smeenk D, et al. O,p’-DDD (mitotane) levels in plasma and tissues during chemotherapy and at autopsy. Cancer Chemother Pharmacol. 1982;9:85-88.
15. Hermsen IG, den Hartigh J, Haak HR. Mitotane serum level analysis; good agreement between two different assays. Clin Endocrinol (Oxf). 2010;73:271-272.
16. Proost JH, Meijer DK. MW/Pharm, an integrated software package for drug dosage regimen calculation and therapeutic drug monitoring. Com- put Biol Med. 1992;22:155-163.
17. Proost JH, Eleveld DJ. Performance of an iterative two-stage bayesian technique for population pharmacokinetic analysis of rich data sets. Pharm Res. 2006;23:2748-2759.
18. Proost JH, Schiere S, Eleveld DJ, et al. Simultaneous versus sequential pharmacokinetic-pharmacodynamic population analysis using an itera- tive two-stage Bayesian technique. Biopharm Drug Dispos. 2007;28: 455-473.
19. Moy RH. Studies of the pharmacology of o,p’DDD in man. J Lab Clin Med. 1961;58:296-304.
20. Chennavasin P, Brater DC. Aminoglycoside dosage adjustment in renal fail- ure: a hand-held calculator program. Eur J Clin Pharmacol. 1982;22:91-94.
21. Du Bois D, Du Bois EF. A formula to estimate the approximate surface area if height and weight be known. 1916. Nutrition. 1989;5:303-311; discussion 312-313.
22. Mauclere-Denost S, Leboulleux S, Borget I, et al. High-dose mitotane strategy in adrenocortical carcinoma: prospective analysis of plasma mi- totane measurement during the first 3 months of follow-up. Eur J Endo- crinol. 2012;166:261-268.