Rivaroxaban Delivery and Reversal at a Venous Flow Rate
Objective—Rivaroxaban is an oral anticoagulant that directly targets both free factor Xa and factor Xa in complex with its protein cofactor, factor Va, in the prothrombinase complex. It is approved in the United States for the prophylaxis of deep vein thrombosis and stroke in patients with atrial fibrillation; however, it also carries a black box warning regarding the risk of thrombosis after discontinuation of treatment. The purpose of this study was to determine the degree to which rivaroxaban, over a range of physiologically relevant free plasma concentrations, inhibits preassembled prothrombinase at a typical venous shear rate (100 s−1) and to determine the dynamics of rivaroxaban washout.
Methods and Results—Prothrombinase was assembled on phospholipid-coated glass capillaries. Its activity was characterized with respect to the activation of prothrombin (mean plasma concentration, 1.4 μmol/L) in the absence and presence of rivaroxaban (2, 5, and 10 nmol/L). The degree of inactivation of preassembled prothrombinase is sensitive to the solution-phase rivaroxaban concentration; however, prothrombinase unmasking upon removal of rivaroxaban is concentration independent.
Conclusion—The model system presented suggests that when rivaroxaban plasma concentrations decrease after cessation of therapy, there will be an unmasking of thrombus-associated prothrombinase that may be related to the reported rebound phenomena.
Rivaroxaban is a direct oral factor (f) Xa inhibitor that shows strong correlations between dosing regimens and pharmacokinetic responses, unlike the classical therapies—warfarin1 and unfractionated heparin.2 Thus, rivaroxaban therapy is not expected to require routine monitoring of anticoagulation status to ensure a therapeutic response in patients. Rivaroxaban plasma concentrations can vary over 1 to 2 orders of magnitude during a typical dosing cycle; Table 1 presents the maximal (Cmax) and minimal (Cmin) plasma concentrations of rivaroxaban during a 10-mg once-daily dosing cycle for the 5th, 50th, and 95th percentiles of patients as described by Mueck et al.3 Only 5% to 10% of rivaroxaban in the blood is unbound, whereas the rest is associated with plasma proteins other than fXa.4,5 In in vitro models of tissue factor–initiated coagulation, the unbound fraction seems to be the functionally available form of rivaroxaban.6 Cessation of rivaroxaban treatment without administration of another anticoagulant may lead to an increased risk of stroke and other thrombotic complications, leading to a black box warning from the United States Food and Drug Administration with respect to nonvalvular atrial fibrillation.7
Unlike the fXa inhibitor fondaparinux (whose function is linked to the inhibitor antithrombin), rivaroxaban can efficiently inhibit both free fXa (half-maximal [inhibitory concentration]50, 0.4 nmol/L) and fXa associated with its protein cofactor fVa forming the prothrombinase complex ([inhibitory concentration]50, 2.1 nmol/L), the major source of prothrombin (fII) activation.8,9 However, this inhibition of fXa species, unlike that potentiated by fondaparinux, is reversible. Thus, rivaroxaban suppression of fXa catalysis depends on the maintenance of the equilibrium between protease-associated and solution-phase rivaroxaban. Therefore, given the reversible nature of rivaroxaban inhibition of its target protease and the potential pathological complications associated with stopping rivaroxaban therapy, we elected to study the delivery of rivaroxaban to and removal from preassembled prothrombinase at a typical venous shear rate (100 s−1) using a prothrombinase flow reactor.10–12
Although flow has been demonstrated to be an important mediator of coagulation,10,11,13–15 a description of the actions of direct fXa and fIIa inhibitors under flow conditions has been lacking. Burghaus et al16 have recently published a computational model of rivaroxaban action that incorporates flow-mediated effects; however, empirical evaluations have not yet been reported. The aim of the current study was to identify the kinetics of rivaroxaban inhibition and removal from prothrombinase under venous flow conditions, as a function of available rivaroxaban concentrations during a typical dosing regimen.
Our findings show that rivaroxaban exhibits a dose-dependent response in its ability to inhibit preassembled prothrombinase, an analog to thrombus-associated prothrombinase in vivo, at concentrations of rivaroxaban reflecting the 50th percentile of patients receiving a 10-mg once-daily dose. The data suggest that rivaroxaban removal is directly related to the degree of prothrombinase inhibition by rivaroxaban. To explain these observations, we introduce a model for rivaroxaban removal from prothrombinase, which mimics the flow conditions. In this model, we calculate the removal of rivaroxaban as occurring in discrete steps in which dissociated rivaroxaban is fully removed from the system and no longer able to interact with preassembled prothrombinase. This model simplifies the description of rivaroxaban removal in the flow system described herein; however, it does not capture the dynamic equilibrium between free and complexed rivaroxaban, hence we term this description as a stepwise quasi-equilibrium model.
Synthetic phospholipid vesicles were prepared from 75% dioleoyl phosphatidylcholine and 25% dioleoyl phosphatidylserine from Avanti Polar Lipids (Alabaster, AL) as previously described.17 fII, fX, and factor V (fV) were purified as described previously18,19 or received as a gift from Haematologic Technologies (Essex Junction, VT). Spectrozyme TH and hirudin were purchased from American Diagnostica (Stamford, CT). The fluorogenic substrate SN-7 was synthesized by and received as a gift from Dr Saulius Butenas.20,21 Rivaroxaban was purchased from Alsachim (Illkirch Graffenstaden, France) and was determined to be equivalent to that produced by Bayer with regard to its absorption spectrum and inhibition of fXa. fV was activated to fVa by incubating fV (1 μmol/L) with fIIa (10 nmol/L) in HEPES-buffered saline (HBS; 20 mmol/L HEPES, 150 mmol/L NaCl, pH 7.4) containing 0.1% polyethylene glycol 8000 (PEG) and 2 mmol/L CaCl2 (Ca2+) for 20 minutes at 37°C before stopping further reactions with the addition of hirudin (12 nmol/L). The concentration of fVa was verified by its activity in fV-deficient plasma in a prothrombin time clotting assay using the TriniClot PT Excel S reagent from Trinity Biotech (Bray, Ireland).
Rivaroxaban Inhibition of Preassembled Prothrombinase Under Flow
Supported phospholipid bilayers were assembled on borosilicate glass capillaries (2 mm×0.2 mm×5 cm) from VitroCom (Mountain Lakes, NJ) as described previously, with minor modifications.10,11 Briefly, plasma-cleaned capillaries were incubated for 1 hour with 100 μmol/L phosphatidylcholine-phosphatidylserine vesicles in HBS containing 2 mmol/L Ca2+ at 4°C to assemble a supported phospholipid bilayer,22,23 rinsed 3× via capillary action with HBS containing 2 mmol/L Ca2+, and stored in the same buffer at 4°C for up to 48 hours. Prothrombinase was assembled in the capillaries by transferring fVa (20 nmol/L) and fXa (0.2 nmol/L) in HBS containing 0.1% PEG and 2 mmol/L Ca2+ into the capillary for 15 minutes and was used immediately.
All experiments were conducted at 37°C. The flow apparatus was used as previously described, with 2 model 22 syringe pumps from Harvard Apparatus (Holliston, MA) and a manual 4-way valve (GE Healthcare) to switch between different fluid mixtures.10,11 After prothrombinase assembly, capillaries were rinsed with ≈200 to 300 μL HBS containing 0.1% PEG and 2 mmol/L Ca2+ at 100 s−1 (0.085 mL/min). To establish the thrombin-generating capacity of each capillary, a mean plasma concentration of fII (1.4 μmol/L) was flowed over the capillary at 100 s−1 for ≈10 minutes. Next, a reaction mixture containing fII (1.4 μmol/L) and rivaroxaban (2, 5, or 10 nmol/L) was flowed over the capillary for ≈10 minutes. Finally, rivaroxaban in the flowing reaction mixture was discontinued, and the capillary was supplied with only fII (1.4 μmol/L) for the final ≈5 minutes. The reacted effluent was collected dropwise into HBS containing 0.1% PEG and 20 mmol/L EDTA to stop further Ca2+-dependent reactions and was assayed with Spectrozyme TH (200 μmol/L) to determine the concentration of active thrombin species.10,11
Determination of Prothrombinase Density Within Capillaries
After each experiment, capillaries were rinsed for 5 minutes at 250 s−1 with HBS containing 0.1% PEG and 2 mmol/L Ca2+. The capillaries were then soaked for 15 minutes in HBS containing 0.1% PEG and 20 mmol/L EDTA (200 μL) to extract fXa from the capillary, which is limiting to prothrombinase assembly and reflects prothrombinase levels within each capillary. The fXa levels were then assayed using a fluorogenic assay similar to that described previously10: the capillary extract was diluted 1:4 in HBS containing 0.1% PEG, 20 mmol/L EDTA, and 50 nmol/L hirudin to a total volume of 200 μL and was incubated at 37°C for 5 minutes. The SN-7 (50 μmol/L) substrate was then added, and its hydrolysis was monitored using a Fluoromax-2 fluorometer from Jobin-Yvon-Spex (Edison, NJ) for 5 minutes at 37°C (wavelength [λ]excitation, 350 nm; λemission, 470 nm; λcutoff-filter, 450 nm). fXa levels were determined by comparison to a standard curve.
Initial generation of active species from each capillary was fitted to Equation 1 using the Graphpad Prism (v 5.02) software package to establish each capillary’s maximal thrombin-generating potential:(1)
where [P] is the concentration of product at any given time (T) or under final steady-state (ss) conditions, T0 is the turning point of the function, and τ characterizes the rate of change upon approach to the final plateau. The percentage inhibition of fIIa activity after 10 minutes of rivaroxaban exposure, as well as the percentage of maximal thrombin-generating potential after cessation of rivaroxaban for 5 minutes, is characterized as the average of last 3 recorded data points under each set of conditions relative to [P]ss. Differences in each data set as a function of rivaroxaban concentration were analyzed for statistical significance using a one-way ANOVA followed by Tukey’s test with the GraphPad Prism software package.
Dissociation of Rivaroxaban From Prothrombinase Under Flow
An apparent dissociation constant (Kd) of rivaroxaban for prothrombinase at a venous shear rate of 100 s−1 was calculated by fitting Equation 2 to the data of the percentage of inhibited prothrombinase as a function of the concentration of rivaroxaban supplied to the system:(2)
where I is the percentage inhibition of prothrombinase and [R] is the concentration of rivaroxaban supplied to the system.
Model for Rivaroxaban Reversal
For the following treatment, we assume that rivaroxaban inhibits prothrombinase via a rapidly reversible on-off mechanism:(3)
To model the removal of rivaroxaban from preassembled prothrombinase under venous flow conditions, we use a stepwise quasi-equilibrium model. We use our measured Kd for rivaroxaban dissociation from prothrombinase at a shear rate of 100 s−1 to model the concentrations of free and complexed species:(4)
where R·fXa is the rivaroxaban-fXa (prothrombinase) complex and f denotes a free species. We assume that the total concentration of fXa present within the capillary is equal to 0.2 nmol/L, introduced during prothrombinase assembly; furthermore, we do not correct for diffusion because rivaroxaban should be able to readily diffuse throughout the capillary in the average capillary passage time (14.1 s) at a shear rate of 100 s−1. The concentration of free fXa can be expressed as the total fXa concentration minus that which is associated with rivaroxaban, where t denotes total:(5)
We also assume that the free concentration of rivaroxaban is reset to 0 after each equilibration of the system, such that after each iteration, the new free rivaroxaban concentration can be expressed as the free fXa concentration minus any free fXa from previous iterations (φ). Under these conditions, we can express the dissociation constant for rivaroxaban from prothrombinase as follows:(6)
We then iteratively solve Equation 6 until no fXa is associated with rivaroxaban (≈3 to 4 iterations).
Finally, we calculate the dissociation rate constant (koff) of rivaroxaban from prothrombinase by measuring the rate of prothrombinase recovery and assuming first-order dissociation as described below:(7)
and from this, we estimate the association rate constant (kon) for prothrombinase inhibition under flow using the apparent Kd.
Prothrombinase Density in Capillaries
Fourteen capillaries used in this study were assayed for their prothrombinase levels: 4.2 ± 2.7 fmoles or 1.9 ± 1.3 fmoles cm−2 (mean±SD). These levels are consistent with the 4 fmoles of nominal prothrombinase that were introduced into each capillary. Furthermore, these levels and SDs of fXa recovery are similar to those observed in our previous study of a bovine prothrombinase system under flow.10
Description of Typical Thrombin Generation Curves
Typical thrombin generation curves showing rivaroxaban delivery and washout are shown in Figure 1. The initial thrombin generation curves in the absence of rivaroxaban show 3 distinct regions: an initial region of no thrombin activity in which the rinsing buffer is leaving the capillary, a rising slope in which the fII-containing solution is entering the capillary, and a steady-state plateau region that is characteristic of the maximal thrombin-generating capacity of each capillary which was determined using Equation 1. Some variability in the thrombin-generating capacity of each capillary was observed, as can be noted in the 3 representative curves shown in Figure 1. The average maximal thrombin generation was 47 ± 14 nmol/L (mean±SD; n=14). These results also demonstrate good reproducibility between capillaries, despite the relatively large SD in the measured recovery levels of fXa. Upon the introduction of rivaroxaban to the system, there is a sharp decrease in the thrombin-generating capacity of each capillary that reaches an apparent steady state within 10 minutes of perfusing the fII and rivaroxaban mixture. Finally, when rivaroxaban is removed from the reaction mixture, a recovery of thrombin-generating capacity in each capillary is observed.
Inhibition of Preassembled Prothrombinase as a Function of Rivaroxaban Concentration
Figure 2 presents the percentage decreases in thrombin-generating capacity with respect to 2, 5, and 10 nmol/L rivaroxaban after ≈10 minutes. The absolute differences in the inhibition of prothrombinase vary significantly with respect to the concentration of rivaroxaban supplied to the system in a dose-dependent manner, with the inhibition of prothrombinase at 2 nmol/L rivaroxaban being significantly different from that at both 5-nmol/L (P<0.01) and 10-nmol/L (P<0.001) rivaroxaban. Table 2 presents a breakdown of the apparent equilibrium distributions of free and rivaroxaban-associated prothrombinase concentrations for each rivaroxaban concentration. By fitting the prothrombinase inhibition data to Equation 2, we determined that the apparent Kd for rivaroxaban binding to prothrombinase at a shear rate of 100 s−1 is 1.2 ± 0.1 nmol/L (mean±SD). This result is in good agreement with the reported [inhibitory concentration]50 of rivaroxaban for prothrombinase of 2.1 nmol/L and for clot-associated fXa of 0.7 nmol/L in closed (nonflow) systems.8,9
Recovery of Thrombin-Generating Capacity on Cessation of Rivaroxaban Supply
The rates of the reappearance of thrombin-generating potential after rivaroxaban were no longer supplied to the catalytically active capillary, with the fII perfusion seemed to vary with the steady-state levels of inhibition achieved by each rivaroxaban concentration (Figure 2). Rates of prothrombinase unmasking, reflecting the entire 5-minute observation interval, are presented in Table 2, with the fastest recovery rate observed when the initial level of inhibition is greatest. These rate data, reflecting a contribution of dissociation and flow removal, were used to estimate the rate constant controlling the dissociation of rivaroxaban from the rivaroxaban-inhibited prothrombinase complex. An average dissociation rate constant of rivaroxaban dissociation from prothrombinase at 100 s−1 of 1.8×10−3 s−1 (t1/2 ≈6 minutes) was calculated, which is comparable with the reported rivaroxaban dissociation from fXa in nonflow systems of 5×10−3 s−1.8,9 An average association rate constant of rivaroxaban to prothrombinase at 100 s−1 was estimated as 1.5×106 mmol/L−1 s−1, which is approximately an order of magnitude lower than 1.7×107 mmol/L−1 s−1 reported for rivaroxaban association with free fXa under nonflow conditions.8,9 Dissociation rates and constants for each rivaroxaban concentration studied are shown in Table 2.
Figure 3 compares the percentage recovery of maximal thrombin-generating capacity after 5 minutes of stopping rivaroxaban perfusion for each of the concentrations studied. It shows that ≈65% of the initial thrombin-generating capacity of each capillary was recovered. Ideally, assuming a first-order process, after 1 hypothetical half-life, prothrombinase levels of 69%, 59%, and 56% for 2, 5, and 10 nmol/L rivaroxaban, respectively, would be predicted, given the levels of active prothrombinase measured in the presence of rivaroxaban. Collectively, these data are consistent with the dissociation of rivaroxaban from prothrombinase at 100 s−1 being a first-order process with a half-life in the range of 6 minutes.
Modeling Rivaroxaban Removal From Prothrombinase Under Flow
Using the mathematical model described in the Methods section, we model the removal of rivaroxaban from prothrombinase assuming that the average percentage of prothrombinase inhibited by rivaroxaban reflected the decreased thrombin activity reported in Figure 2. As illustrated in Figure 4, under all 3 starting conditions, ≈90% of the prothrombinase-associated fXa is free after 1 iteration of our quasi-equilibrium model; furthermore, after the first iteration of the model, the percentage of free prothrombinase increased for 2, 5, and 10 nmol/L rivaroxaban-treated capillaries by 54%, 71.5%, and 78.5%, respectively, indicating that the rate of rivaroxaban removal from prothrombinase-associated fXa is directly related to the degree of its inhibition (Table 2).
The results of our study indicate that rivaroxaban in a flowing solution inhibits surface-localized prothrombinase in a dose-dependent manner over a range of free rivaroxaban plasma concentrations that reflect the mean 50th percentile of patients receiving a dose of 10-mg once-daily dose. This finding is consistent with both closed-system studies,8,9 as well as the flow modeling work of Burghaus et al.16 A mouse arterial thrombosis injury model by Wagner et al24 showed similar dose-dependent responses and suggests that observations made in the current study at a venous shear rate (100 s−1) may be applicable under higher shear conditions
What is more interesting is our finding that the rate of rivaroxaban removal from membrane-localized prothrombinase at a shear rate of 100 s−1 is directly related to the percentage inhibition of prothrombinase at the time at which rivaroxaban therapy is discontinued. As shown with our quasi-equilibrium dissociation model, this result is a direct consequence of flow-providing vehicle for the removal downstream of dissociating rivaroxaban, as the non-rivaroxaban–containing solution is introduced to sites of preassembled prothrombinase.
We believe that our model of rivaroxaban inhibition of preassembled prothrombinase under venous flow conditions may be directly compared with clot-associated prothrombinase at a site of venous thrombosis. One drawback of our model is that only a single level of preassembled prothrombinase is used. We speculate, however, that increasing the density of functional prothrombinase within the fulminating clot would result in a higher concentration of rivaroxaban required to drive the level of functional prothrombinase below the clotting threshold. Another drawback of our model is that it does not take into account the presence of the pool of rivaroxaban associated with other plasma proteins, nor the structural features of actual thrombi. That being said, we believe that our data suggest that on cessation of rivaroxaban treatment, the consequent decreased plasma availability of rivaroxaban will result in an ongoing release of the drug from thrombus-associated prothrombinase. Although the apparent rebound in thrombin generation described herein occurs much more rapidly (t1/2 ≈ 6 minutes) than the half-life of rivaroxaban in vivo (≈7–11 hours), we presume that the described rebound phenomenon will still occur in vivo as the equilibrium shifts from complexed to free rivaroxaban pools.25
An exact model of the equilibrium distribution of rivaroxaban species in vivo is difficult to construct because direct measurements of free and bound rivaroxaban species across a standard dosing cycle are yet to be reported, although pharmacokinetic data on total plasma levels of rivaroxaban based on organic extractions of patient plasma preparations collected across the dosing cycle have been reported.4 Reports on the association of rivaroxaban with the cellular components of blood are not available, yet secondary analyses determining free and bound distributions in plasma have reported an average of ≈10% free distribution when assessed in a small group of individuals at one time (Cmax).4 However, direct measurements addressing the issue of whether this equilibrium distribution is constant across a standard dosing cycle are not available. Albumin has been reported to bind rivaroxaban,5 but the kinetic constants controlling the equilibrium between free and bound rivaroxaban have not been reported, nor are there studies investigating rivaroxaban binding to other proteins. Predictive models of rivaroxaban pharmacokinetics have been reported in which both the uptake of orally administered rivaroxaban into the blood and its elimination are defined as first-order processes.26 A first-order elimination process is consistent with a mechanistic scenario in which ≈90% of rivaroxaban associated with plasma proteins is bound to high-abundance proteins (>10 μmol/L; eg, albumin, IgG), with relatively low affinity (Kd> μmol/L). In this case, replenishment of the free rivaroxaban pool is most likely a uniform rather than a stepwise process, with free plasma levels decreasing uniformly as a constant fraction of the total pool when rivaroxaban is eliminated from the blood. Additional studies would be required to determine whether other sources of bound rivaroxaban with alternative equilibrium dynamics are contributors to the pharmacodynamic availability of rivaroxaban at different times in a dosing cycle. Therefore, given the lack of pharmacokinetic data regarding the equilibrium distribution of free and bound rivaroxaban species, we believe that our current model system provides a reasonable model for the removal of rivaroxaban from thrombus-associated prothrombinase in vivo.
Our group and others have previously demonstrated that prothrombinase complex produced during tissue factor-initiated blood coagulation persists and is capable of reinitiating coagulation without input from extrinsic tenase activity and there are significant amounts of clot-associated fXa to support these reactions.27–30 Furthermore, our group has also demonstrated that concentrations of rivaroxaban that prolong tissue factor-initiated whole blood clot times 2- to 3-fold are not sufficient to block thrombin generation by clot-associated prothrombinase.6 This conclusion is supported by evidence that another reversible direct fXa inhibitor (ZK-807834) prevents reocclusion in a dose-dependent manner in a canine coronary thrombosis model,31 thus implying that when thrombus-associated fXa/prothrombinase is not inhibited by a direct fXa inhibitor, such as rivaroxaban, there is the potential for a thrombotic event in the absence of a de novo tissue factor-initiated event. Furthermore, as the prothrombinase complex is 300 000× more efficient than fXa in activating prothrombin, we therefore speculate that it is this loss of inhibition of clot-associated fXa/prothrombinase that leads to prothrombotic rebound phenomena.32,33
Given this difference in rivaroxaban’s effectiveness, the data from the current study suggest that when anticoagulated individuals discontinue rivaroxaban therapy without the introduction of a secondary anticoagulant, the potential for a resurgence in prothrombotic activity due to loss of prothrombinase/fXa inhibition at thrombotic sites will be greater at any point in time than the likelihood of a de novo tissue factor-initiated event. Although the half-life of rivaroxaban in vivo is considerably longer than the ≈6-minute half-life observed in our model system,25 we propose that as rivaroxaban plasma levels decrease so will rivaroxaban be washed away from thrombus-associated prothrombinase, which along with the equilibrium processes leads to the reactivation of small amounts of preassembled prothrombinase that are sufficient to lead to a procoagulant response as previously demonstrated by our laboratory.27,28
We thank Vincentios Zachery and Matthew Gissel for their technical assistance.
Sources of Funding
This work was supported by the National Institute of Health Grant P01 HL46703 (project 1) to Dr Mann. Dr Haynes is supported by a National Institute of Health Thrombosis and Hemostasis Training Grant 5T32HL007594 to Dr Mann. The laboratory has previously received funding from Johnson and Johnson.
Dr Mann has served as a Consultant to Daiichi-Sankyo, Merck, Baxter, GTI, Alnylam, T2 Biosystems. He is the Chairman of the Board of Haematologic Technologies, Inc. The other authors have no conflicts to report.
- Received July 10, 2012.
- Accepted September 14, 2012.
- © 2012 American Heart Association, Inc.
- Ageno W,
- Gallus AS,
- Wittkowsky A,
- Crowther M,
- Hylek EM,
- Palareti G
- Mueck W,
- Borris LC,
- Dahl OE,
- Haas S,
- Huisman MV,
- Kakkar AK,
- Kälebo P,
- Muelhofer E,
- Misselwitz F,
- Eriksson BI
- Perzborn E,
- Roehrig S,
- Straub A,
- Kubitza D,
- Mueck W,
- Laux V
- Haynes LM,
- Dubief YC,
- Mann KG
- Hathcock JJ
- Okorie UM,
- Denney WS,
- Chatterjee MS,
- Neeves KB,
- Diamond SL
- Higgins DL,
- Mann KG
- Katzmann JA,
- Nesheim ME,
- Hibbard LS,
- Mann KG
- Bajaj SP,
- Mann KG
- Wagner NM,
- Dressel T,
- Schäfer K,
- Konstantinides S
- Orfeo T,
- Butenas S,
- Brummel-Ziedins KE,
- Mann KG
- Orfeo T,
- Brummel-Ziedins KE,
- Gissel M,
- Butenas S,
- Mann KG
- Hérault JP,
- Bernat A,
- Pflieger AM,
- Lormeau JC,
- Herbert JM
- Abendschein DR,
- Baum PK,
- Verhallen P,
- Eisenberg PR,
- Sullivan ME,
- Light DR
- Nesheim ME,
- Taswell JB,
- Mann KG