Milrinone is a phosphodiesterase-3 (PDE-3) inhibitor that improves cardiac function in adults [1,2], children , and infants . The volume of distribution, protein binding, metabolism, and clearance of drugs that influence the plasma concentration of a drug are influenced by age . A study of amrinone kinetics in infants and children after open heart surgery (OHS) showed age-related differences . Although the pharmacokinetics of milrinone after OHS have been reported in adults [7,8], they remain undetermined in infants and children.
Traditionally, pharmacokinetic parameters have been estimated by detailed studies in a limited number of subjects . Individual pharmacokinetic parameters are then used to calculate population means. However, if the intraindividual variation is large (because only a few samples could be obtained), then the pharmacokinetic parameter estimate is imprecise, and dosing might be inaccurate. In addition, interindividual variation in parameters also affects dosing. Nonlinear mixed effects modeling (NONMEM) is a pharmacokinetic strategy for estimating population parameters using sparse data collected from a large patient population . Pharmacokinetic studies in pediatric patients are often constrained by their smaller blood volume and reduced sampling volumes. Hence, NONMEM would be an ideal pharmacokinetic method in this population.
Patients often require inotropic medications to be weaned from cardiopulmonary bypass (CPB) . When administered during CPB, 20% of amrinone binds to the CPB circuit, affecting drug dosing . Whether milrinone binds to CPB circuitry is not known.
This prospective, open-label, dose-escalating study of milrinone in infants and children after OHS was designed to describe the pharmacokinetics and side effects of milrinone. Pharmacokinetic parameters were estimated using both traditional and NONMEM analysis. In addition, we evaluated the binding of milrinone to the CPB circuit.
Our protocol was approved by the institutional review board, and written parental consent was obtained. Patients aged 1 mo to 13 yr undergoing OHS between August 1994 and March 1996 were enrolled in the study. Criteria for enrollment included a clinical indication for postoperative inotropic support and platelet counts >100,000 cells/mm-3 on admission to the intensive care unit (ICU).
Patients were divided into two groups to receive milrinone as follows. In the small dose group (n = 11), patients received an initial 25-micro g/kg bolus over 5 min and a milrinone infusion of 0.25 micro g [center dot] kg-1 [center dot] min-1 was started at the end of the first dose. Thirty minutes later, a second 25-micro g/kg bolus was given, and the infusion was increased to 0.5 micro g [center dot] kg-1 [center dot] min-1. In the large dose group (n = 8), patients received a 50-micro g/kg bolus administered over 10 min, and an infusion of 0.5 micro g [center dot] kg-1 [center dot] min-1 was started at the end of the bolus. A 25-micro g/kg bolus was given 30 min after the first bolus, and the final rate of infusion was increased to 0.75 micro g [center dot] kg-1 [center dot] min-1. In addition, patients in both groups received a third 25-micro g/kg milrinone bolus based on their clinical response, but their infusion rates were unchanged.
Blood was sampled for the plasma milrinone concentration 30 min after each loading dose. To obtain the milrinone concentration at steady state, blood samples were also drawn after 22 and 24 h of continuous milrinone infusion. For washout kinetics, samples were drawn 5, 10, 20, and 60 min and 3, 5, and 7 h after discontinuing the infusion. The samples were stored at -70[degree sign]C, and milrinone levels were measured using high-pressure liquid chromatography as previously reported for amrinone . The within-run and between-run coefficient of variation of the milrinone assay was <or=to5% at the concentrations measured.
After successful termination of CPB, seven identical CPB circuits were used from patients who had not received milrinone or amrinone. The circuit components and methods were as reported previously . These circuits contained waste blood that was circulated at 2.5 L/min. The weight of blood in the circuit was determined by weighing the circuit empty and when filled with blood. The volume of blood in the circuit was calculated from the weight of blood using a specific gravity of 1.04. A control blood sample was obtained, milrinone was injected proximal to the oxygenator into the circuit, and samples were collected distal to the oxygenator at 1, 10, and 20 min. The milrinone dose used (0.3 mg/L of prime volume) was calculated to yield a plasma concentration of 300 ng/mL, which is within the reported adult therapeutic range (100-400 ng/mL) . The samples were stored and analyzed for milrinone as described above.
Evaluation of blood samples for platelet count, liver enzymes (alanine aminotransferase, aspartate aminotransferase, gamma-glutamyl transferase), and kidney function tests (blood urea nitrogen and creatinine) was performed before milrinone loading and after milrinone infusion. During milrinone infusion, platelet counts were obtained every 12 h.
During the study period, serial platelet counts were prospectively collected in patients who were not receiving milrinone and who were admitted to the ICU after OHS. Platelet counts were recorded from their arrival in ICU (Postoperative Day [POD] 0) until discharge from ICU or POD 3, whichever was later. The duration of CPB and inotrope requirements were noted. For the purposes of the study, thrombocytopenia was defined as a platelet count <or=to100,000 cells/mm-3. In the perioperative period after OHS, thrombocytopenia, either alone or in conjunction with clinical evidence of bleeding, is a trigger for platelet transfusion in our institution.
Data are reported as mean +/- SD. Continuous data were analyzed by using a multivariate analysis of variance adjusted for milrinone dose. To test the effects of milrinone on serial platelet counts, data were analyzed by using a multivariate analysis of variance adjusted for dose and duration of milrinone infusion. Categorical data were analyzed by the using chi squared and chi squared for trend tests; significance was set at P <or=to 0.05.
Plasma milrinone concentration versus time data were analyzed by using traditional pharmacokinetic methods and population data analysis (NONMEM). Initially, individual plasma milrinone concentration and time data were analyzed by using traditional pharmacokinetic methods. Population analysis was then performed to determine the influence of patient characteristics on the pharmacokinetic parameters obtained.
The milrinone plasma concentrations obtained after the infusion were fitted to both a one- and a two-compartment model using Scientist (Micromath, Inc., Salt Lake City, UT) to obtain the terminal elimination rate constant (beta). Steady-state milrinone plasma concentrations obtained at 22 and 24 h on infusion were averaged to obtain a mean steady-state milrinone concentration. Plasma clearance (CL) was determined as the ratio of infusion rate to mean steady-state milrinone concentration. The volume of distribution (V beta) was calculated as the ratio of CL and beta. The terminal elimination half-life (T1/2 beta) was determined as 0.693/beta.
The milrinone plasma concentrations obtained after loading doses, at steady state, and during washout were used in the population analysis. In total, 217 observed values were available. The population analysis was performed on a Sun Workstation using the NONMEM program version IV level 1.0 with double precision. NONMEM provides estimates and variances of the population parameters and determines the influence of patient factors on the parameters . One- and two-compartment open models without any covariants were initially compared using NONMEM subroutines ADVAN1 TRANS2 and ADVAN3 TRANS4, respectively, assuming an exponential error model. The two-compartment model described by NONMEM has four parameters: clearance (CL), central or initial volume (Vc), peripheral volume (Vp), and intercompartmental clearance (Q). Model building was performed by the three-step method described by Mandema et al. : an initial choice of a pharmacokinetic model, evaluation of the influence of patient covariants, and the choice of an error model.
The patient covariants considered in model building were age, gender, body weight, body surface area (BSA), presence or absence of Down syndrome, duration of CPB, and duration of milrinone infusion. The influence of patient covariates on milrinone disposition was evaluated using changes in the objective function computed by NONMEM. The objective function is equal to minus twice the log likelihood of the data. The difference in the objective function between full and reduced hierarchical models has a chi squared distribution with the degrees of freedom given by the reduction in the number of parameters . The likelihood ratio test was used to compare models that were not subsets of each other. Significant differences in the model were assumed at P <or=to 0.005 to allow for the influence of multiple statistical testing. A change in objective function of 4 is required to reach a significance level of P < 0.05 in the case of related models . When comparing nonhierarchical models, a change of 10.5 is required. After determination of the full model, parameters were deleted one at a time to determine whether covariates showed a significant contribution in the final model. Finally, the appropriate statistical error model was determined by comparing the additive versus proportional error models. The statistical error model accounts for the interindividual variation in the population parameters and the proportional error model for intraindividual variation. Correlations among patient parameters (age, BSA, body weight) were determined by linear regression analysis using the Spearman's coefficient of rank correlation to determine statistical significance.
Nineteen patients were entered in the study (Table 1). There were 12 infants (aged <1 yr) and 7 children (aged >or=to1 yr and <13 yr).
In all but one patient, the milrinone concentrations after discontinuing the infusion (washout) fit the two-compartment model as determined by the coefficient of variation of the estimated parameters. The traditional pharmacokinetic parameter estimates are shown in Table 2. At steady state, both the small and large dose groups had plasma concentrations that were >or=to100 ng/mL, a level considered therapeutic in adults. Age (infants versus children) did not significantly affect V beta; however, CL was significantly lower in infants (Table 3). We compared our data in infants and children with those reported in adults after OHS . V beta in infants was significantly larger compared with adults, and both infants and children had significantly faster CL than adults (Table 3).
The results of NONMEM model building are summarized in Table 4. There was a large interpatient variation in CL, V beta, and T1/2 beta when patient covariants were not included. The two-compartment model described the data significantly better than the onecompartment model (-148 change in objective function). The estimates of the population means (intersubject coefficients of variation) for the four parameters without covariants were CL 2.18 L/h (140%), Vc 0.0102 L (237%), Vp 6.25 L (141%), Q 3.54 L/h (118%). The addition of either body weight or BSA to the equations significantly influenced both CL and Vc (Table 4). Age alone was a significant factor, but because both body weight and BSA correlated highly with age (r = 0.914 and 0.947, respectively; P < 0.0001), deletion of age from the model did not alter the goodness of fit. Inclusion of body weight in Vp decreased the coefficient of variation but did not alter the objective function when deleted from the model. Neither the duration of infusion nor the dose of infusion significantly altered the CL and or the T1/2 beta. Likewise, the addition of serum creatinine or gender did not significantly influence any of the estimated parameters. Final models are given in Table 5.
Mean plasma milrinone concentrations in the CPB circuit were 343 +/- 63 ng/mL at 1 min, 320 +/- 99 ng/mL at 10 min, and 324 +/- 82 ng/mL at 20 min (P >or=to 0.05).
At enrollment, all patients had platelet counts >100,000 cells/mm-3. Over time, the platelet count decreased significantly (P <or=to 0.05); however, this was not different between dose groups (Table 6). Of the 19, 11 patients (58%) developed thrombocytopenia (defined as platelet count <or=to100,000 cells/mm-3) during milrinone infusion. At 24 h of infusion, 21% were thrombocytopenic, at 36 h, 47% were, and at 48 h, 42% were. At 60 h, only 4 patients (2 in each group) were receiving milrinone; 0 were thrombocytopenic. Of the 19 patients, 2 (11%; 1 in each group), both infants, required a platelet transfusion while receiving milrinone.
Over the study period, 128 patients not receiving milrinone were prospectively evaluated for thrombocytopenia. The overall incidence of thrombocytopenia in this group was 25% (PODs 0-3), which was a lower incidence than in patients receiving milrinone (P <or=to 0.05). chi squared for trend analysis showed a significantly increased risk of thrombocytopenia from POD 0 to POD 3 in all patients after OHS; additionally, the use of amrinone significantly increased the risk of thrombocytopenia. When patients receiving milrinone were compared with those receiving amrinone at 48 h, patients receiving milrinone had a greater risk of thrombocytopenia (P <or=to 0.02, chi squared).
Of the 19 patients, 2 (11%) developed arrythmias during milrinone administration, 1 in each dose group. An infant in the small dose group who underwent a tetralogy of Fallot repair was in nodal rhythm on arrival in ICU, then developed junctional ectopic tachycardia at a rate of 160-170 bpm. This occurred 8-10 h after starting milrinone. Milrinone was discontinued at 19 h. The plasma concentration just before discontinuing the infusion was 178 ng/mL. The second infant, who underwent atrioventricular canal repair, presented to the ICU in junctional rhythm with a heart rate of 180-190 bpm. The patient later developed atrioventricular dissociation diagnosed as junctional ectopic tachycardia, and the milrinone infusion was stopped at 16 h. The plasma level 5 min after discontinuing the infusion had a milrinone concentration of 345 ng/mL.
Milrinone infusion did not significantly alter blood urea nitrogen or serum creatinine levels. The liver enzymes (aspartate amino transferase, alanine aminotransferase, and gamma-glutamyl transferase) were also not significantly changed.
In this study of infants and children given milrinone infusion after OHS, a two-compartment model best describes the pharmacokinetics. Both the CL and V beta obtained by individual pharmacokinetic and population (NONMEM) analysis provided comparable results (Table 7). The V beta was similar in infants (aged <1 year) and children (aged 1-13 years), but the CL was significantly lower in infants. These results have dosage implications for the two age groups.
NONMEM analysis demonstrated a strong correlation between body weight (or BSA) and the CL and V beta of milrinone in children. The large intersubject variability in both CL and V beta was reduced significantly when the parameters were adjusted for body weight or BSA. The steady-state plasma concentrations of milrinone correlate inversely with the plasma CL of milrinone. Therefore, the required maintenance infusion of milrinone correlates directly with plasma CL. Including BSA or body weight in the CL model resulted in a decrease in the intersubject variability from 140% to approximately 9% or 11%, respectively (Table 5).
Determination of the total loading dose of milrinone is based on the V beta (Vc + Vp). Because milrinone was initiated with a combination of multiple IV boluses and an infusion, it was not possible to determine Vc by using traditional pharmacokinetic methods; however, we were able to do so by using NONMEM analysis. By including body weight in the volume model, intersubject variability was reduced from 237% to 16% (Table 5). Therefore, the final CL and V beta models allow a more accurate determination of both the loading and maintenance doses required to obtain a desired plasma level (Table 8).
The initial V beta in infants is significantly larger than the 0.3 L/kg described in adults after OHS (Table 3) . The CL in both infants and children was also significantly greater than the 2 mL [center dot] kg-1 [center dot] min-1 reported in adults . The T1/2 beta was not significantly longer than that described in adults after OHS . This, once again, highlights the fact that compared with adults, pediatric patients often require larger doses (both bolus and infusion) to achieve comparable plasma levels.
We demonstrated a 58% incidence of thrombocytopenia while patients were receiving milrinone. However, only two patients (11%) required platelet transfusion. In a similar study with amrinone, 43% had thrombocytopenia and one (8%) was transfused with platelets .
The overall incidence of thrombocytopenia after OHS (patients not receiving milrinone) in our institution is 25%, and the risk increases with the duration of ICU stay. In addition, the use of amrinone increases the risk of thrombocytopenia. Over the study period, patients receiving milrinone as part of their inotropic support after OHS had a greater risk of thrombocytopenia at 48 hours than those receiving amrinone. This may be related to different patient populations, e.g., younger and sicker patients in the milrinone study. Although in this study CPB duration did not correlate with thrombocytopenia, other CPB-associated factors alter platelet function and survival . Hence, in this group of patients, the effect of PDE-3 inhibitors on platelets is not easily separated from that of CPB itself. However, our data confirm the increased risk of thrombocytopenia associated with the use of PDE-3 inhibitors. Therefore, we recommend monitoring serial platelet counts while the patient is receiving milrinone.
In our group of patients, we observed arrythmias that could be related to the surgery itself in 2 of 19 patients. High plasma levels may have contributed to the ongoing arrhythmia in 1 of our patients. Although a significant increase in heart rate with milrinone in neonates has been observed , no correlation of arrhythmias with plasma milrinone concentration has been reported. Unlike milrinone, the heart rate either decreased  or showed a small but insignificant increase  after amrinone loading.
In contrast to amrinone , milrinone does not seem to bind to CPB circuitry. The differences in up-take by CPB circuit between amrinone and milrinone probably relate to differences in the physio-chemical properties between the two drugs, because the two experiments used very similar methods.
In conclusion, milrinone has a larger V beta and a faster CL in infants and children compared with published data in adults. These findings have dosage implications for the pediatric patient. Thrombocytopenia associated with the use of milrinone is dependent on the duration of infusion, and serial platelet counts should be monitored. Milrinone does not bind to CPB circuitry.
The authors acknowledge the help of Mary-Kay Nespeca, RN, in data collection; of Renee Oakes, CCP, in studying milrinone binding to CPB circuitry; to D. G. Hall, MD, and F. M. Lupinetti, MD, for allowing their patients to be studied; and to the intensive care nurses for their patience and support during the study.
1. Feneck RO. Intravenous milrinone following cardiac surgery: effects of bolus infusion followed by variable dose maintenance infusion. J Cardiothorac Anesth 1992;6:554-62.
2. Wright EM, Skoyles J, Sherry M. Milrinone in the treatment of low cardiac output states following cardiac surgery. Eur J Anesthesiol 1992;5(Suppl):21-6.
3. Barton P, Garcia J, Koutali A, et al. Hemodynamic effects of IV milrinone lactate in pediatric patients with septic shock. Chest 1996;109:1302-12.
4. Chang AC, Atz AM, Wernovsky G, et al. Milrinone: systemic and pulmonary hemodynamic effects in neonates after cardiac surgery. Crit Care Med 1995;23:1907-14.
5. Besunder JB, Reed MD, Blumer JL. Principles of drug biodisposition in the neonate: a critical evaluation of the pharmacokinetic-pharmacodynamic interface-Part II. Clin Pharmacokinet 1988;14:261-86.
6. Lawless S, Burckart G, Diven W, et al. Amrinone in neonates and infants after cardiac surgery. Crit Care Med 1989;17:751-4.
7. Bailey JM, Levy JH, Kikura M, et al. Pharmacokinetics of intravenous milrinone in patients undergoing cardiac surgery. Anesthesiology 1994;81:616-22.
8. Butterworth JF, Hines RL, Royster RL, James RL. A pharmacokinetic and pharmacodynamic evaluation of milrinone in adults undergoing cardiac surgery. Anesth Analg 1995;81:783-92.
9. Ludden TM. Population pharmacokinetics. J Clin Pharmacol 1988;28:1059-63.
10. Sheiner LB, Rosenberg B, Marathe VV. Estimation of population characteristics of pharmacokinetic parameters from routine clinical data. J Pharmacokinet Biopharm 1977;5:445-79.
11. Hines R. Clinical applications of amrinone. J Cardiothorac Anesth 1989;3(6 Suppl 2):24-32.
12. Williams GD, Sorensen GK, Oakes R, et al. Amrinone loading during cardiopulmonary bypass in neonates, infants, and children. J Cardiothorac Anesth 1995;9:278-82.
13. Lynn AM, Sorensen GK, Williams GD, et al. Hemodynamic effects of amrinone and colloid administration in children following cardiac surgery. J Cardiothorac Anesth 1993;7:560-5.
14. Das PA, Skoyles JR, Peacock JE, et al. Disposition of milrinone in patients after cardiac surgery. Br J Anesth 1994;72:426-9.
15. Mandema JW, Verotta D, Sheiner LB. Building population pharmacokinetic-pharmacodynamic models. I. Models for covariate effects. J Pharmacokinet Biopharm 1992;20:511-28.
16. Kestin AS, Valeri CR, Khuri SF, et al. The platelet function defect of cardio-pulmonary bypass. Blood 1993;82:107-17.
© 1998 International Anesthesia Research Society
17. Sorensen GK, Ramamoorthy C, Lynn AM, et al. Hemodynamic effects of amrinone in children after Fontan surgery. Anesth Analg 1996;82:241-6.