Protamine has been used for decades to delay the absorption and prolong the action of insulin and to reverse heparin-induced anticoagulation (1). However, pharmacokinetic variables have not been reported for protamine in any of the common circumstances in which it is prescribed. Indeed, blood concentrations of protamine have not been measured or correlated with efficacy in humans. After heparin, rapid administration of protamine may cause marked systemic arterial hypotension, pulmonary hypertension, or both (2). Adverse drug reactions after protamine are common and typically are underreported (3). Incomplete or nonpersisting reversal of heparin can lead to hemorrhage, transfusion of blood products, and a wide variety of complications (4). However, excess concentrations of protamine have been linked to a long list of adverse effects in vitro, including inhibition of platelet glycoprotein Ib-von Willebrand factor, increased P-selectin expression, block of the calcium release channel (ryanodine receptor type 1), and negative inotropy (4–7). Thus, a more complete understanding of protamine pharmacokinetics and pharmacodynamics could increase the likelihood of successful heparin reversal and could optimize patient safety.
We hypothesized that the rapid (<10 min) protamine dosing technique often used during surgery fails to sustain detectable blood concentrations of protamine beyond 20 min. We assumed that if hemodynamic responses to protamine reflected heparin-protamine interaction rather than responses to protamine per se, then infusion of protamine to subjects who had not received heparin should have minimal hemodynamic consequences (4). To test these hypotheses, we developed a new assay for protamine in plasma and, using this assay, measured protamine blood concentrations, as well as heart rate (HR), arterial blood pressure (BP), and cardiac output (CO), in healthy volunteers who consented to receive a single dose of IV protamine. We used the protamine blood concentration data to study the pharmacokinetics of protamine elimination.
Our study protocol was reviewed and approved by our IRB. Each subject gave his or her written informed consent to participate in our study.
Consenting volunteers (n = 16) allowed us to withdraw 50 mL of venous blood for development of our analytical method. The blood was anticoagulated with EDTA, and the plasma was separated from cellular elements by centrifugation at 11,000 g for 10 min. Protease inhibitors (Boehringer Mannheim, Indianapolis, IN) were added to all blood samples on the basis of pilot experiments demonstrating that protease inhibitors prevented a progressive loss of protamine from samples awaiting analysis. These inhibitors did not interfere with the high-performance liquid chromatography (HPLC) measurement of protamine. Plasma samples were spiked with known amounts of chum salmon protamine (Fujisawa USA, Inc., Deerfield, IL) for calibration. Protamine was extracted from plasma by using solid-phase extraction with Varian (Harbor City, CA) Bond Elut C18/0H cartridges. These columns were conditioned with methanol and equilibrated with water. Plasma samples were centrifuged at 1000 g for 2 min and then applied to the column. The columns were washed with methanol. Protamine was eluted with 1% perchloric acid in acetonitrile. Eluates were evaporated to dryness and reconstituted in acetonitrile before being applied to Zorbax (Chadds Ford, PA) 300SB-C8 4.6 × 250-mm, 5-μm, 300-A columns. The mobile phase A was 0.05% trifluroacetic acid in 5% acetonitrile. The mobile phase B was 0.045% trifluroacetic acid in 25% acetonitrile. The system was equilibrated with mobile phase A and increased to 75% mobile phase B over 22 min. The mobile phase flow was 1 mL/min. The HPLC system consisted of a GPM-2 pump, a variable wavelength detector manufactured by Dionex (Sunnyvale, CA), and a Rheodyne (Cotati, CA) injector. Protamine was monitored at 200 nm. All chemicals were purchased from commercial sources, and all were reagent or HPLC grade.
HPLC confirmed that there were multiple components in the pure chum salmon protamine solutions with which we spiked our plasma samples, and these multiple peaks were also seen in the plasma samples from our volunteers (Fig. 1A). The relative proportions of these peaks remained constant as protamine plasma concentration decreased over time in our volunteers. We used the sum total of these multiple components as our measure of protamine in plasma and in producing our standard curve (Fig. 1B). The extraction yield of free protamine from plasma was 94% ± 4% (mean ± sd, n = 10). Standard calibration curves were linear between 1 and 60 mg/L. The coefficient of variation was 11% (n = 10).
Pilot experiments were also performed to determine whether protamine was sequestered from plasma by blood cells, metabolized in blood cells, or otherwise lost if there were delays in sample preparation (which consisted of adding protease inhibitor and cooling samples to 5°C while awaiting centrifugation) and also to test the consistency of our assay with repeated measurements. Consenting volunteers (n = 5) provided 30 mL of fresh whole blood. Two-milliliter aliquots of this blood were spiked with 20 μg of protamine (Fujisawa USA). These aliquots were maintained at 37°C for 0, 5, 10, or 30 min before they were rapidly cooled to 5°C, and plasma was separated from cells as previously described. Recovery of protamine was nearly the same at baseline (0 min) as at 5, 10, or 30 min. The recovered protamine represented 92% (79%–116%), 93% (82%–108%), and 90% (54%–100%) (median [range]), respectively, of the spiked amount.
Seventeen consenting, fit volunteers between the ages of 21 and 48 yr were studied to model protamine pharmacokinetics. Exclusion criteria included a history of cerebrovascular disease, cardiovascular disease, asthma, renal disease, diabetes, previous exposure to protamine, vasectomy, or pregnancy. The demographic characteristics of these volunteers are provided in Table 1.
The study was conducted in an anesthetic induction room. During the study, the subjects were monitored continuously with electrocardiography (leads II and V5), finger-pulse oximetry, and noninvasive arterial BP measurements (recorded at 5-min intervals). CO was determined with thoracic bioimpedance (model IQ 101; Renaissance Technologies, New Town, PA). Large-bore IV catheters were inserted in both arms to permit dedicated sites for drug administration and blood sampling.
The study volunteers lay supine on a padded hospital gurney. Once the baseline measurements had been repeated and confirmed to be stable (<10% variation between two baseline HRs and mean arterial BPs), the supine volunteers received 0.5 mg/kg protamine over 10 min by use of a syringe pump. Protamine was infused via a port attached to the distal end of the IV catheter, to minimize dead space. In the first seven subjects, blood samples were obtained at baseline and at 10, 11, 13, 15, 20, and 25 min. However, after review of the data from the first 7 subjects, an additional 10 subjects were studied in whom samples were obtained at baseline and at 5, 8, 10, 11, 13, 15, 17.5, 20, 25, and 30 min to better match the observed pattern of protamine’s plasma concentrations. Sample collection began 15 s before the defined sample time and was completed by 10–15 s after the defined sample time. These samples were collected in vacuum tubes containing EDTA and protease inhibitors. All collected samples were immediately cooled to 5°C in a container of ice-water slush. The plasma was separated from cellular elements by centrifugation at 11,000 g for 10 min. Plasma samples were stored at <−30°C while awaiting analysis. A pilot study showed <10% variation in measured concentration when plasma samples were stored for 5–105 days after centrifugation. Plasma samples were subjected to HPLC after solid-phase extraction, as described previously. Hemodynamic measurements were obtained before, during, and after protamine administration.
Concentration-versus-time data were fit to compartment models by using the nonlinear mixed-effects regression techniques of the NONMEM software package (NONMEM Project Group, University of California, San Francisco, CA). We fit both fixed- and random-model parameters by minimizing the maximum likelihood objective function.
Model rate constants (k10, k12, and k21) and the central compartment’s volume of distribution (V1) were estimated directly by the NONMEM program. The model rate constants (per minute) are fractional clearance rates, i.e., the fraction of the drug in a compartment that is cleared by moving into the next compartment.
In addition to the standard compartmental models, one- and two-compartment nonlinear Michaelis-Menten elimination models with and without a parallel compartmental (i.e., constant clearance) elimination and a compartmental model with irreversible binding were tested (8). Michaelis-Menten elimination assumes saturable (e.g., enzymatic) elimination of protamine, resulting in the fractional clearance being a decreasing function of protamine concentration. The rate of Michaelis-Menten elimination is modeled as V = Vmax · Conc/(kM + Conc), where V is the rate of enzymatic elimination, Vmax is the maximum rate of enzymatic elimination (occurs as protamine concentration approaches zero), Conc is concentration, and kM is the Michaelis constant ([E][S]/[ES]; in enzymology, [E] = concentration of an enzyme, [S] = concentration of substrate, and [ES] = concentration of substrate-bound enzyme; in this case, E corresponds to protamine, S to its putative binding site, and ES to bound protamine).
The irreversible binding model had two compartments, with the second compartment as the substrate to which protamine bound irreversibly (i.e., irreversible within our experimental time frame). Thus, initially, protamine elimination occurs through both irreversible binding to the second compartment and first compartment fractional clearance (k10). Then, as the binding sites of the second compartment become saturated, the proportion of protamine cleared through first compartment clearance progressively increases until it becomes exclusive. This irreversible binding model was included to demonstrate alternative saturable clearance models to Michaelis-Menten elimination. The standard one- and two-compartment models were fit by using NONMEM’s Advan 1 and Advan 3 subroutines. The Michaelis-Menten elimination models were fit to the data by using NONMEM’s Advan 10 subroutine. Finally, the two-compartment with irreversible binding models were fit by using Advan 6 with user-supplied differential equations, as described in the Appendix. The interindividual variability among model variables was modeled as log normal in distribution. The intraindividual residual error was fit to the data by using additive, constant coefficient, combined additive and constant coefficient, and power function models.
The Schwarz-Bayesian criterion (SBC) was used to determine which model best fit the data (9). Covariate adjustments were tested in the pharmacokinetic models and included if supported by the SBC. Graphs showing model fits were used to confirm these choices. Absolute performance error (APE) and geometric performance error (GPE) were calculated for each sample concentration, and their median, 75th, 90th, and 95th percentiles were reported. Because compartmental pharmacokinetic models are mathematically exponential in form (mono-, bi-, and triexponential, etc.), we report the 50th, 75th, and 95th GPEs derived from the log scale of the APE. From each drug concentration sample (subject i, sample j), the log-scale APE isMATH
The log-scale APE can then be converted back to the arithmetic scale by taking its antilog. With these calculations, GPE50 is the median GPEij, and GPE75 and GPE95 are the 75th and 95th percentiles of the GPE. The modeling error resulting from an observed concentration that is 5 times the predicted concentration is considered equivalent to the modeling error resulting from an observed concentration that is one fifth of the predicted concentration. A GPE50 of 5 states that 50% of the observed concentrations are between one fifth and 5 times the model-predicted concentration. Likewise, a GPE95 of 2 states that 95% of the observed observations are between one half and twice the model-predicted value.
All analyses were performed with NONMEM (version V, 1.1) and SAS (version 8.0; SAS Institute, Cary, NC), with α < 0.05 considered significant. Demographic variables were compared between men and women by using Student’s t-test. The area under the protamine concentration versus time curves (AUCs) in men versus women was compared by using Wilcoxon’s ranked sum test. The efficiency of pharmacokinetic models was compared by using SBC and performance error estimates (8,10).
The demographic characteristics of our research findings are provided in Table 1. Note that the men were significantly taller and larger (by body surface area) than the women, but that body habitus (body mass index) was similar in men and women.
HR, arterial BP, and CO were unchanged during the administration of protamine (Fig. 2). One volunteer had flushing, anxiety, light-headedness, mild hypertension (154/81 mm Hg), and tachycardia (116 bpm) and demonstrated a macular rash on her neck and arms upon completion of protamine infusion. Her signs and symptoms resolved rapidly after a colleague gave her 355 mL of a cold beverage (Coca-Cola™; The Coca-Cola Co., Atlanta, GA). Her arterial BP and HR were 132/77 mm Hg and 96 bpm, respectively, after her cold drink.
Protamine concentration versus time plots were compared in the first 7 and subsequent 10 experimental subjects. There appeared to be no difference in the shapes of the plots, so both sets of data were combined in all our analyses. Note that in every case, protamine concentrations were less than the limit of detection after 20 min or less (Fig. 3). Note that during the initial infusion, the AUC is concave, which conflicts with the assumptions of conventional compartmental linear pharmacokinetics, in which a convex curve is predicted. As a result, our data fit conventional compartmental (exponential) elimination models poorly.
Protamine concentration versus time data were significantly different between men and women (Table 2). Specifically, weight-adjusted protamine dosing resulted in significantly decreased AUC and significantly greater plasma clearance (CL) and volume distribution at steady state (Vss) in women than in men.
Concentration data were first subjected to a model-independent pharmacokinetic analysis (Table 3). Median (range) values were the following: Vss = 12.3 (6.9, 63.1) L; CL = 2.2 (1.1, 12.1) L/min; and t1/2 = 7.4 (5.9, 9.3) min. Table 4 describes the results of our attempts to fit the data to conventional compartmental models, Michaelis-Menten models, and compartmental models with irreversible binding. Note that among compartmental models, a simple two-compartment model achieved the largest SBC and the smallest objective function. We also fit our data to compartmental models in which k10 and V1 were adjusted for weight and sex. These models performed worse than unadjusted models. We fit our data to compartmental models in which intraindividual error was modeled as a power function or as additive and coefficient of variation. These more complex error models were inferior to the simpler coefficient of variation error structure.
Note also that larger SBCs and smaller objective functions were produced by Michaelis-Menten models for protamine elimination. Table 5 describes the variables of our three best performing pharmacokinetic models, and Figure 4 provides observed/predicted values for these three models.
We report the first measurements of protamine in plasma, as well as the first analysis of protamine pharmacokinetics in humans. The rapid elimination of protamine which we observed could underlie recurrent heparinization after protamine reversal, and we speculate that a single initial dose technique might not maintain effective reversal of heparin anticoagulation in the face of continuing heparin release from stores in the reticuloendothelial system. In the absence of heparin, 16 of 17 subjects had no hemodynamic changes with protamine. The one subject with adverse signs and symptoms seemed to have had histamine release, although we have no specific evidence for this. We also do not know whether our treatment of this volunteer had anything to do with her rapid recovery.
A similar analytical technique was reported for measurement of protamine in blood-free drug formulations (9). We recognize that under most circumstances protamine would largely be complexed with heparin, and a differing analytical method might be required. We also recognize that elimination of protamine/heparin complexes could be faster or slower than elimination of free protamine, and further studies will be necessary to determine whether these theoretical differences are present.
Our data fit conventional compartmental elimination models poorly. Plots of our subjects’ AUC data appeared inconsistent with the usual exponential decay models. Rather than a convex AUC during the 10-min protamine infusion, as would be expected for linear pharmacokinetic compartmental models, the AUCs appeared concave during the 10-min infusion (Fig. 3). We speculate that we could be observing a nonlinear pharmacokinetic process in which the microrate constants are concentration dependent. Our Michaelis-Menten and irreversible saturable binding models are examples of possible concentration-dependent pharmacokinetic models. Although our irreversible saturable binding model best fit the data, we recognize that innumerable alternative concentration-dependent pharmacokinetic models could possibly better fit our data. Nevertheless, identifying the mathematical model that best fits a drug AUC data set may be less important than identifying that a drug that is usually given as a single dose has an unexpectedly short half-life.
We identified significant differences between men and women in AUCs. With AUCs, we make no assumptions about the number of compartments that are present, other than that they follow linear pharmacokinetic form. However, there was no significant association between any compartmental pharmacokinetic variable and sex by use of the one-compartment or two-compartment elimination models. The overall sex difference may be the result of small effects on multiple pharmacokinetic variables that our investigations are underpowered to detect.
As previously noted, the rapid decrease in protamine concentrations suggests that conventional single-dose regimens may not be the optimal technique by which to reverse anticoagulation. Our data suggest that in situations in which patients undergo recurring exposure to small amounts of heparin (e.g., with transfusion of heparinized blood from the cardiopulmonary bypass pump oxygenator or from cell salvage devices), a maintenance infusion of protamine might serve to ensure that sufficient protamine will be present to prevent reheparinization. Obviously this conjecture awaits experimental confirmation. Rapid elimination also offers an explanation for why small additional protamine doses given to patients usually cause no apparent adverse effects despite the numerous adverse laboratory findings linked to excess protamine concentrations (5–7,11–13).
We measured minimal cardiovascular changes associated with protamine administration to volunteers. The lack of serious adverse events in our relatively small number of volunteers does not permit us to comment on the overall safety of excess free protamine concentrations; however, our data are clearly inconsistent with there being frequent, predictable, adverse hemodynamic reactions to free (unbound by heparin) protamine in healthy, unanesthetized patients. Further studies in patients will be needed to determine whether protamine elimination is altered by prior administration of heparin and whether the incidence of adverse reactions to protamine is altered by prior heparinization.
Our study has a number of notable limitations. First and most important, we conducted the study in normal volunteer subjects, not in patients with cardiovascular disease who had received heparin for anticoagulation and who required protamine to reverse the anticoagulation. Second, we cannot identify with certainty an optimal pharmacokinetic model for protamine elimination kinetics. We collected data after a single brief infusion, and, in many cases, the plasma protamine concentrations had decreased to undetectable values within 20 minutes. Thus, despite collecting data from 17 volunteers, we had a limited number of data points with which to perform our analyses. Pharmacokinetic events limited to a <15-minute time frame, as was true in this study, can be heavily influenced by intravascular mixing and early tissue distribution processes. A 15-minute time frame severely limits the number of samples that can be obtained and makes small errors in timing more important than in studies of drugs that are eliminated from plasma more slowly. We sampled venous rather than arterial blood for our assays, and this has several consequences. It takes longer to withdraw blood samples from a peripheral vein than an arterial catheter, and this increased uncertainty about the timing of samples could reduce our ability to define the optimal pharmacokinetic models. In general, arterial sampling is considered more reliable than venous sampling for pharmacokinetic analyses. We chose venous sampling to reduce the risk and discomfort to our volunteers. Although we performed standard curves with every batch of HPLC measurements, we did not test with an internal standard during each subject’s samples. Such an internal standard could identify whether there might be greater day-to-day variation in percentage protamine recovery than was identified in the pilot experiments described in Methods. All these factors may have limited our ability to identify the best pharmacokinetic model. Future studies in patients could use a maintenance infusion of protamine, allowing us to obtain more blood samples with measurable protamine concentrations, possibly permitting us to determine whether protamine follows linear (dose-indepen-dent) kinetics or not.
Our assay technique could potentially permit monitoring of both free and heparin-complexed protamine concentrations. Studies are planned in which our protamine assay technique will be used to identify the optimal protamine infusion technique for heparin reversal in patients.
FIGURETABLE Differential Equations User Supplied to NONMEM to describe the model disposition of protamine
1. Davis SN, Granner DK. Insulin, oral hypoglycemic agents, and the pharmacology of the endocrine pancreas. In: Hardman JG, Limbird LE, eds. Goodman & Gilman’s The pharmacologic basis of therapeutics. 9th ed. New York: McGraw-Hill, 1996: 1487–517.
2. Lowenstein E, Johnston WE, Lappas DG, et al. Catastrophic pulmonary vasoconstriction associated with protamine reversal of heparin. Anesthesiology 1983; 59: 470–3.
3. Kimmel SE, Sekeres MA, Berlin JA, et al. Adverse events after protamine administration in patients undergoing cardiopulmonary bypass: risks and predictors of under-reporting. J Clin Epidemiol 1998; 51: 1–10.
4. Patla V, Comunale ME, Lowenstein E. Heparin neutralization. In: Gravlee GP, Davis RF, Kurusz M, Utley JR, eds. Cardiopulmonary bypass: principles and practice. 2nd ed. Philadelphia: Lippincott Williams & Wilkins, 2000: 473–91.
5. Barstad RM, Stephens RW, Hamers MJ, Sakariassen KS. Protamine sulphate inhibits platelet membrane glycoprotein IB-von Willebrand factor activity. Thromb Haemost 2000; 83: 334–7.
6. Kozek-Langenecker SA, Mohammad SF, Masaki T, et al. The effects of heparin, protamine, and heparinase 1 on platelets in vitro using whole blood flow cytometry. Anesth Analg 2000; 90: 808–12.
7. Koulen P, Ehrlich BE. Reversible block of the calcium release channel/ryanodine receptor by protamine, a heparin antidote. Mol Biol Cell 2000; 11: 2213–9.
8. Bourne DWA. Mathematical modeling of pharmacokinetic data. Lancaster, PA: Technomic Publishing, 1995:6, 10, 27, 38, 40–41.
9. Snycerski A, Dudkiewicz-Wilczynska J, Tautt J. Determination of protamine sulphate in drug formulations using high performance liquid chromatography. J Pharm Biomed Anal 1998; 18: 907–10.
10. Kotz S, Johnson NL. Schwartz criterion. In: Kotz S, Johnson NL, eds. Encyclopedia of statistical sciences. Vol 8. New York: John Wiley, 1988: 289–90.
11. Despotis GJ, Gravlee G, Filos K, Levy J. Anticoagulation monitoring during cardiac surgery: a review of current and emerging technologies. Anesthesiology 1999; 91: 1122–51.
12. Carr JA, Silverman N. The heparin-protamine interaction: a review. J Cardiovasc Surg (Torino) 1999; 40: 659–66.
© 2002 International Anesthesia Research Society
13. Miyashita T, Makajima T, Hayashi Y, Kuro M. Hemostatic effects of low-dose protamine following cardiopulmonary bypass. Am J Hematol 2000; 64: 112–5.