Pharmacokinetics of Dopamine in Healthy Male Subjects

This study was reviewed and approved by the Institutional Review Board of Wake Forest University Baptist Medical Center. Informed consent was obtained from nine healthy, nonobese male volunteers between the ages of 23 and 45 years who were compensated for their time. Healthy, young-adult, male volunteers were selected as the study population to minimize demographic variability between subjects. Specifically, our goal was to minimize the effects of sex hormones, extremes of age and weight, and organ dysfunction. We planned to study 10 subjects, but 1 subject unexpectedly withdrew, and we were unable to schedule a replacement while we had research space and staff to complete the study.

An intravenous catheter was placed in each volunteer’s dominant arm, and an arterial catheter was placed in the opposite radial artery. Monitors consisted of electrocardiogram, pulse oximeter, and noninvasive blood pressure cuff. Serum electrolyte and creatinine concentrations were measured for each volunteer. Once heart rate and blood pressure had remained unchanged for a period of 10 min, a continuous infusion of dopamine at 10 μg · kg⁻¹· min⁻¹was begun and continued for 10 min. After discontinuing the infusion, a 30-min washout period was observed. At the conclusion of this washout period, dopamine infusion was restarted at 3 μg · kg⁻¹· min⁻¹and continued for 90 min. Another 30-min washout period was observed after the 3-μg · kg⁻¹· min⁻¹dopamine infusion was discontinued. Heart rate and blood pressure were monitored during each phase of the study.

Table 1details the timing of 32 blood samples obtained from each subject. Evacuated blood collection tubes were used to withdraw blood, with 3 ml of blood removed 5 s before the stated collection times, and a 3-ml sample withdrawn beginning precisely at the listed times. Blood samples drawn from the catheter were immediately placed on ice. During each washout phase and immediately after completion of each phase of the study (i.e. , within 20 min for each sample), samples were centrifuged, the pellet was discarded, and the plasma was removed and frozen at −70°C. Plasma samples were thawed, and dopamine concentrations were analyzed after alumina extraction by high-performance liquid chromatogaphy and electrochemical detection as previously described. 2Absolute detection limit was 0.8 ng/l. Interval reference standards were mixed randomly with subjects’ samples to confirm the ongoing accuracy and precision of the assays; interassay coefficient of variation at 10 ng/l was 4.6%.

Concentration-versus -time data were fit to several variations of one-, two-, and three-compartment models with and without absorption lags and rates. Population models were fit using nonlinear mixed-effects regression techniques with the NONMEM software package (NONMEM Project Group, University of California, San Francisco, CA). NONMEM fits models by minimizing the −2 × log likelihood objective function as it simultaneously estimates the model’s pharmacokinetic parameters, their interindividual variance, and the intraindividual residual variability (the mixed-effects approach). This objective function minimization was achieved using NONMEM’s Laplacian method, an algorithm using second-order derivatives. In addition to the aforementioned population modeling, NONMEM was used to fit models to each individual patient separately (the interindividual variance parameters being excluded). The parameter estimates from each individual were then combined using geometric means (the two-stage approach). In an attempt to maximize the precision of our models, parameter covariates of age, sex, weight, and serum creatinine were included in some of the models tested. Models were judged best that simultaneously minimized both the −2 × log likelihood objective function and number of model parameters, as indicated by the Schwartz-Bayesian criterion. Selection was further aided by the graphic examination of the model fits and residual plots. Model microrate constants, k¹⁰, k¹², k^21,k¹³, and k³¹, and the central compartment’s volume of distribution, V¹, were estimated directly by the NONMEM program. Clearances (Cl) and the remaining compartment volumes of distribution were calculated from the rate constants as follows:

Parameter subscripts refer to the model’s compartment number. Double subscripts refer to flow from one compartment to the next (e.g. , k¹²is the rate constant describing drug movement from compartment 1 to compartment 2). Compartment 0 is outside the body (i.e. , k¹⁰is the rate constant describing drug elimination). Exponential coefficients and half-lives were calculated using standard equations.

For two-compartment models, steady-state concentrations (C^ss) during continuous infusions were calculated as follows 3:

where k⁰is the infusion rate constant, α is the initial exponential rate constant from biexponential disposition model, and β is the terminal exponential rate constant from biexponential disposition model.

The possible effects of sex, age, weight, and creatinine clearance on C^sswere tested using linear and loglinear regression. The mixed-effects interpatient variability of the rate constants and V¹were modeled lognormal in distribution. Models of intraindividual residual variability fit by NONMEM included the constant coefficient of variability model, the combined additive and constant coefficient of variability model, and the power function model.

Simple measures of model performance were calculated, including absolute performance error (APE). APE calculations were made for each individual, modified from procedures described by Coetzee et al. , 4as follows:

where obs is the observed (measured) concentration; pred is the predicted concentration from the model; and the subscripts i and j refer to the jth time of the ith individual.

With these calculations, APE⁵⁰is the median APE^ijover the entire range of i and j, and APE⁹⁵is the 95th percentile APE^ij. The APE⁵⁰and APE⁹⁵are based on the absolute values and represent the ranges within which 50% or 95% of the APE occur.

Repeated measured analysis of variance tested for heart rate changes between baseline, the 10-μg · kg⁻¹· min⁻¹infusion peak, the first washout, the 3-μg · kg⁻¹· min⁻¹infusion peak, and the second washout period. This analysis of variance was performed using the Proc Mixed subroutines of SAS version 6.12 (SAS Institute, Inc., Cary, NC).

The demographics of our nine subjects are listed in table 2. The body mass index [weight in kilograms/(height in meters)²] ranged from 25.6 to 35.4 with a mean ± SD of 29.2 ± 3.0. Dopamine concentrations for all subjects and time points are illustrated in figure 1. There was a total of nine missing data points because of sampling errors. One of these errors occurred during the initial baseline phase, and another (different subject) occurred during the final washout phase. There were no other technical errors during the study or during measurement of plasma concentrations.

Initial basal dopamine plasma concentrations varied over a wide range from 3 to 8,023 ng/l. These concentrations were not significantly different from those measured at the end of the two subsequent washout periods (fig. 2, top), and there was no correlation between basal dopamine concentrations and weight. After initiation of the 10-μg · kg⁻¹· min⁻¹infusion (10 min), plasma concentrations increased rapidly, with individual plasma concentrations for the 7- and 10-min samples ranging from 5,616 to 201,513 ng/l. After termination of the 10-min bolus infusions, dopamine concentrations decreased rapidly, ranging from undetectable to 1,191 ng/l. Dopamine concentrations during the 90-min infusion at 3 μg · kg⁻¹· min⁻¹appeared to reach a steady-state plateau rapidly, with individual measurements of plasma concentrations varying between 367 to 83,745 ng/l. After termination of the infusion, dopamine plasma concentrations decreased precipitously back to baseline levels. Figure 2(top) demonstrates the concentrations obtained during the final two measurements (7 and 10 min) during the 10-μg · kg⁻¹· min⁻¹infusion and the final three measurements (30, 60, and 90 min) during the 3-μg · kg⁻¹· min⁻¹infusions, as well as the interceding baseline values. Of note, the subjects with the highest baseline concentrations did not have the highest concentrations during the drug infusions, and many of the washout concentrations were lower than the original baseline concentrations for the same subjects.

The mixed-effects population models best fit the data to a two-compartment model with an absorption lag of 0.5 min and V¹adjusted for body weight. This best model characterized the interindividual variability as lognormal in distribution and modeled the intraindividual residual error using a constant coefficient of variability model (constant coefficient of variability, residual error =ς¹× predicted outcome). Selection criteria for a representative four of our tested models are listed in table 3. This table is not inclusive of all the mixed-effects models tested but demonstrates the process of building what ultimately became the best fit model (model D). This best fit model performs best using the Schwartz-Bayesian criterion.

The best model is characterized by its extremely rapid microrate constants (k¹⁰= 1.10, k¹²= 0.244, and k²¹= 0.07), resulting in an initial half-life, t^1/2,α= 0.5 min, a systemic clearance (Cl¹) equal to 110% of the central compartment’s volume of distribution (V¹) per minute, and a terminal half-life, t^1/2,β= 12.3 min (table 4). These rapid rates agree with the immediate increase and decrease of dopamine concentrations before and after infusion (fig. 1). Dopamine C^ss(in ng · l⁻¹· kg⁻¹) during any prolonged continuous infusion can be estimated from and our model as:

where k⁰is the infusion rate (in μg · kg⁻¹· min⁻¹), and 1,000 corrects the dose units (μg) to concentration units (ng).

This equation estimates C^ssfor 3- and 10-μg · kg⁻¹· min⁻¹infusions at 11,300 and 37,700 ng · kg⁻¹· l⁻¹, respectively. These concentrations at steady state (C^ss) are asymptotes; C^ssis only reached as time approaches infinity. However, based on simulations using our model, dopamine concentrations during continuous infusion will reach 88.9% of C^ssat 10 min, 90% of C^ssat 12.3 min, and 95% of C^ssat 24.8 min. This best model uses a weight-adjusted infusion rate, but there were no significant effects of patient age, sex, or serum creatinine on C^ss.

Our best model outperforms all other models using the Schwartz-Bayesian criterion as well as two other commonly used indices of selection criteria, the −2 · log [likelihood] and the Akaike Information Criterion. Using this best model, the APE⁵⁰and APE⁹⁵were 59% and 514%, respectively. These performance measures can be interpreted as 50% of our measured concentrations differ from predicted concentration by no more than 59%, whereas 95% of measured concentrations differ from model predictions by no more than 514% of the predicted concentrations. Graphical representation of our best fit pharmacokinetic model is included in figure 1as the shaded curve.

The best fit two-stage modeling approach resulted in a one-compartment model with no lag time (k¹⁰= 0.44 min⁻¹and V¹= 55.5 l) and had an overall APE⁵⁰of 81% and APE⁹⁵> 100,000%. The two-stage model fits were far inferior to our best mixed-effects models and thus are not mentioned further.

Table 5demonstrates the performance characteristics of our best mixed-effects fit model by individual subject. This table demonstrates that for five of the nine subjects, the curve fit very well (APE⁹⁵< 100%), whereas four subjects varied from the predicted values by considerable amounts (APE⁹⁵from 194% to 648%). Additional parameter adjustments for the demographic covariates of age, weight, or serum creatinine could not account for this subject-to-subject variability.

Figure 2(bottom) displays the heart rates measured during the initial baseline; 8 min into the 10-μg · kg⁻¹· min⁻¹infusion; 25 min into the first washout period; 45 min into the 3-μg · kg⁻¹· min⁻¹infusion; and at the end of the final washout period. Heart rate was monitored as a safety issue rather than a therapeutic end point. Repeated measures analysis of variance found no differences between the heart rates for the three baselines. Mean heart rate increased significantly with the infusion of 10 μg · kg⁻¹· min⁻¹but did not increase above baseline with the prolonged infusion of 3 μg · kg⁻¹· min⁻¹.

Most studies of drug pharmacokinetics assume a single compartment into which the drug is administered, commonly represented by plasma volume, and that there is no other source of drug within the body. The elimination of drug from this single compartment is dependent on clearance of the drug out of the plasma either by excretion from the body or through terminal metabolism. Our study modeled the pharmacokinetics of dopamine in healthy male subjects receiving clinically relevant doses of dopamine infusions at 10 and 3 μg · kg⁻¹· min⁻¹using compartmental models. Compartmental pharmacokinetic models are based on the assumption that the concentration at any time is directly proportional to the infusion dosages. This assumption of linearity would suggest that for a given dosing scheme, tripling the amount of drug infused will triple the expected drug concentrations at any given sampling time.

Rarely do drugs actually follow such a simplified kinetic scheme; most drugs are better described by a more complex exponential model, thus accounting for alterations in drug concentration that result from redistribution of drug into a second “peripheral” compartment from which it can distribute back into the plasma “central” compartment. Multiexponential models describing the pharmacokinetics of drugs with more than a single compartment have been described for a number of medications, including narcotics, inotropic medications, benzodiazepines, and other anesthetic drugs. 4–11Defining the pharmacokinetic parameters for a drug allows for more precise determination of dosing to accommodate changes in drug distribution associated with physiologic parameters such as age, gender, and body weight, as well as changes brought about by disease states and organ dysfunction. 5,7,12–15Many of these pharmacokinetic studies have raised doubt about using body weight as the primary adjustment for dosing drugs, a conflict recently highlighted by Bouillon and Shafer. 16 

Dopamine is known to increase cardiac output, which, in turn, may affect its own clearance. Thus, rates of clearance and other pharmacokinetic parameters may change as plasma dopamine concentrations change. Such dose and concentration-dependent changes in pharmacokinetic parameters would violate pharmacokinetic linearity and may account for our model predictions (fig. 1) overestimating plasma concentrations during the 10-min infusion. Nonlinearity may also explain why “renal dose” dopamine (3.0 μg · kg⁻¹· min⁻¹) infusions result in plasma concentrations equivalent to plasma concentrations achieved with 10-μg · kg⁻¹· min⁻¹infusions. Any concentration-dependent changes in the pharmacokinetic parameters would violate the assumption of linearity and question the validity of compartmental models.

Regardless of nonlinear pharmacokinetics and other possible model misspecifications, our two-compartment model predictions do, for the most part, follow the median observed dopamine concentrations (fig. 1). Individual measurements of plasma dopamine concentrations do, however, vary as much as 75-fold, despite the use of identical weight-based infusion rates. The random plasma concentration peaks observed at lower concentrations during the postinfusion sample times may reflect experimental error or system instability caused by the influences of endogenous dopamine concentrations. Such large variability in plasma concentrations in a homogeneous population of healthy male subjects casts doubt on the reliability of using any pharmacokinetic model to predict plasma concentrations.

Study designs for determining drug pharmacokinetics involve starting and stopping the drug and measuring concentrations frequently as the changes in plasma concentration occur. Ideally, the drug could be administered in doses less than, equivalent to, and greater than standard clinical practices to define the upper and lower limits of plasma concentrations and pharmacokinetic variables. Our design began with a group of healthy male volunteers to limit demographic differences and to use a moderately high dose (10 μg · kg⁻¹· min⁻¹) but to limit the duration of this infusion to avoid unpleasant side effects encountered in previous studies we conducted with the drug. Thus, 10 min was chosen for the higher dose, then a prolonged infusion with the better-tolerated 3-μg · kg⁻¹· min⁻¹dose. We prospectively chose not to randomize this order because we felt a prolonged infusion was more likely to cause accumulation of drug, and a longer washout phase would be required. Given the rapidity of washout of the drug, this assumption seems to have been incorrect, and dose randomization could have been performed. Our design leaves open the potential that sequential dosing effects have influenced our results.

Previous studies have measured endogenous catecholamine levels under different physiologic conditions. Regitz et al.  17demonstrated large variability in dopamine (and other catecholamines) in healthy control subjects. Schwartz et al.  18demonstrated variations in dopamine concentrations at rest and during angina. Zaloga et al.  19reported a wide range of epinephrine concentrations after coronary artery bypass grafting. In this same study, dopamine concentrations were also measured but were not reported because of the tremendous variability in the measured concentrations (personal communication from the authors, September 1998). These wide variations are confirmed by the baseline values we measured, and the large intrapatient variability is confirmed by the fact that the concentrations measured during the washout phases were lower than baseline concentrations for some of our subjects. It remains unclear how much of the discrepancies in plasma concentrations are cause by endogenous dopamine, initial distribution, compartmental differentiation between subjects, genetic differences in catecholamine metabolism, or dopamine-induced changes in metabolism of the drug. Regardless of the cause of these intrapatient and interpatient differences, our data demonstrate that plasma dopamine concentrations in subjects receiving “renal dose” dopamine at 3 μg · kg⁻¹· min⁻¹may actually exceed plasma concentrations in other subjects receiving 10-μg · kg⁻¹· min⁻¹infusions. This phenomenon occurred in our study, with four subjects having peak plasma concentrations during the 10-μg · kg⁻¹· min⁻¹infusions that were lower than maximal concentrations achieved by four other subjects during the 3-μg · kg⁻¹· min⁻¹infusions.

Additional studies have attempted to define pharmacokinetic variables for dopamine. Le Corre et al.  20determined that the volume of distribution and clearance of dopamine varied according to the dose of drug infused. This finding suggested nonlinear, dose-dependent pharmacokinetics for the drug and possible receptor or metabolic “saturation.” The reported steady-state volumes of distribution were 0.78 ± 0.22 l/kg and 1.58 ± 0.62 l/kg for doses of 3 and 6 μg · kg⁻¹· min⁻¹, respectively, and these values were similar whether a one- or two-compartment model was used. However, this study was conducted using a small number of patients within the intensive care unit, with varying degrees of organ dysfunction, age, gender, and associated diseases. In a study by Ratge et al. , 14dopamine metabolites were measured using three male and three female volunteers receiving 5 μg · kg⁻¹· min⁻¹of the drug. This study used a single-compartment model to determine the volume of distribution as 2.1 ± 0.5 l/kg. It is unclear if gender had an influence on dopamine pharmacokinetics in this study, but each of these studies demonstrated considerable variability in dopamine concentrations.

The 0.5-min lag time was the most stable model parameter, with a coefficient of variability that was < 1%. The determination of 0.5-min lag time is consistent with the experimental methods; the infusion of drug was intravenous, whereas the sampling of blood was from the arterial catheter. Thus, a short (0.5-min) delay to allow for mixing of samples in the blood stream and circulation of drug to the sampling site is reasonable. A lag is commonly not required for pharmacokinetic modeling when elimination of drug is slower, where the delay is inconsequential. Dopamine concentrations, however, increase and decrease very rapidly, and thus a lag accounts for the rapid shift from low to higher concentrations.

Our study represents the “best case scenario” with little demographic variation between subjects. It is reasonable to assume that had we studied subjects with a wider range of body mass, we would have found even greater variability in pharmacokinetics, as has been demonstrated recently by Egan et al.  21using the opioid remifentanil. Similarly, hormonal influences, age, concomitant administration of medications, and associated disease processes commonly alter the pharmacokinetics of drugs; therefore, extrapolating our predicted concentrations to a population of critically ill patients should be performed with extreme caution. Our study design also does not eliminate the possibility of a drug sequencing effect by starting with a high dose of drug followed by infusion of a lower dose. The likelihood of such a “priming” effect is unlikely given the fact that plasma concentrations decrease back to very low levels (at or below the initial baseline values) between the two drug administrations, and our washout between infusions was of sufficient duration to allow for complete elimination of the drug from the system. In retrospect, however, randomizing the order of infusions would have been appropriate.

There are no published reports that associate the physiologic effects of dopamine with defined plasma concentrations or that determine whether selective stimulation of dopaminergic, β, and α receptors is better accounted for by specific plasma concentrations or by specific infusion rates. Analysis of variance testing showed that heart rate was increased by the higher infusion rate in our study. Nevertheless, we caution that we did not design this study to test for a relationship between dopamine concentrations and clinical effects and can only speculate about the relationship between dopamine concentrations and other clinical variables, such as cardiac output, urine volume, or organ function.

In conclusion, despite designing our study with a homogeneous population of healthy male subjects and using specific weight-based dosing, the plasma concentrations of dopamine we measured varied widely. Although we have been able to construct a pharmacokinetic model of dopamine, the APE for this model remains high. Our data suggest that dopamine dosing based on body weight does not yield consistent, predictable plasma concentrations, even in a population of healthy male subjects.

The authors thank James C. Eisenach, M.D., for his assistance with performing the high-performance liquid chromatogaphy assays, and Judy Bennett, R.N., Lynne Harris, R.N., and Kendra Murphy, R.N., for their assistance with completing this study.