How I approach pharmacological thromboprophylaxis in children Free

Recent epidemiologic evidence suggest that the incidence of hospital-acquired venous thromboembolism (VTE) among children in the United States has continued to rise over the past 20 years.¹ Although pediatric risk assessment models have been published, including research using single institutional and multi-institutional retrospective data sets, putative prothrombotic risk factors and retrospectively derived risk models still require prospective validation.^2-7 VTE prevention in children is a critical need that requires the identification of high-risk subpopulations and settings, as well as the investigation of safety and efficacy of thromboprophylaxis (TP) interventions (including pharmacological TP) in these subpopulations and settings, in order to safely prevent VTE among those children at highest VTE risk.

To date, pediatric trial–derived evidence regarding the efficacy and safety of pharmacologic TP has focused on specific pediatric subpopulations (ie, those with hematologic malignancy, congenital heart disease, COVID-19–related illness, and central venous catheterization [CVC]) but have not addressed pharmacological TP in other common pediatric clinical presentations. This, along with a perception of a heightened risk of major bleeding that accompanies pharmacological TP^8-10 and the absence of pediatric evidence related to the equivalence or added benefit of mechanical TP as compared with pharmacological TP alone, contribute to the substantial variation observed across inpatient health care settings regarding pediatric VTE preventive strategies.^11-14 This manuscript will consider 4 exemplar case–based scenarios in common clinical experience; succinctly review the literature regarding pediatric VTE risk factors (Figure 1) and pharmacological TP trial findings (Table 1); and present a pragmatic approach to pharmacological TP in each scenario, noting preferences for drug selection, dosing, duration, and laboratory monitoring, as applicable.

Patient 1 is a 26-month-old boy with hypoplastic left heart syndrome with dextrocardia who is undergoing postoperative care after an extracardiac fenestrated Fontan operation. He had previously undergone a modified Blalock-Taussig-Thomas shunt placement at age 4 days, followed by a bidirectional Glenn procedure at age 4 months. There is no family history of hypercoagulability or thromboembolism, although he has a history of a prior lower extremity VTE after his Glenn procedure with a workup that revealed heterozygosity for the prothrombin gene (factor 2 G20210A) variant. The child was extubated after the procedure, has multiple intracardiac lines, a percutaneous CVC for hemodynamic monitoring, has required minimal perioperative blood transfusions, and is recovering in the pediatric cardiac intensive care unit. His end-organ function appears to be normal, with no present evidence of hepatic impairment or renal dysfunction. Current laboratory data are within normal limits, although his preoperative hemoglobin was 18 g/dL, and his current blood urea nitrogen is slightly elevated and he has a negative fluid balance, likely due to diuretic use in the postoperative setting.

Patient 2 is a 16-year-old girl who is hospitalized for multisystem inflammatory syndrome in children (MIS-C) in the general pediatric ward 2 weeks after an initial presentation with mild COVID-19 respiratory infection. She has severe obesity but no other past medical history. Her home medication list includes oral contraception. Clinical symptoms include febrile state for >4 days, myalgia, abdominal discomfort, and supporting laboratory data including a D-dimer of 3.9 mg/L fibrinogen equivalent units, C-reactive protein of 4.4 mg/L, and platelet count of 85 × 10⁹/L. The child is receiving systemic corticosteroids and antipyretics, and is awaiting echocardiography and other screening laboratory data while receiving supportive care for MIS-C.

Patient 3 is a 14-year-old girl with spastic quadriplegia, intractable epilepsy, and chronic respiratory failure with tracheostomy and mechanical ventilator dependence who was admitted from the emergency department to the pediatric intensive care unit for acute on chronic respiratory failure secondary to bacterial pneumonia, dehydration, and fluid-responsive shock. A percutaneous internal jugular venous CVC and femoral arterial catheter were placed to optimize hemodynamic monitoring, fluid balance, and ventilatory strategy. Over the first several days of hospitalization, the child has received broad spectrum antibiotics and required escalated respiratory support with high-frequency oscillation. As such, she is now receiving neuromuscular blockade to improve ventilatory-patient synchrony and is completely immobilized. The clinical team had ordered mechanical TP after initiation of neuromuscular blockade with intermittent pneumatic compression devices applied to the lower extremities.

Patient 4 is a 3-year-old girl with newly-diagnosed acute lymphoblastic leukemia (ALL) who just underwent a tunneled subclavian venous CVC (ie, port) placed by general surgery for protocolized induction chemotherapy with vincristine, prednisone, doxorubicin, and calaspargase pegol. Routine laboratory testing reveals a hemoglobin of 7.7 g/L, leukocyte count of 44.9 × 10⁹/L, and platelet count of 55 × 10⁹/L.

The application of pharmacological TP among hospitalized children is complicated by a perception that prothrombotic risk significantly varies by unique risk factors across diverse pediatric subpopulations. However, in recent years, numerous published studies have considered several pediatric subgroups, including children who are critically ill in the intensive care unit, hospitalized children who are noncritically ill in the general pediatric ward, and those hospitalized with congenital or acquired heart disease, yielding comparable results across hospitalized populations and settings in regard to identified VTE risk factors and retrospectively derived VTE risk prediction model performance.

Morrison et al¹⁶ and Atchison et al⁷ summarized data regarding prothrombotic risk among hospitalized children who are noncritically ill and found considerable concordance across various study methodologies regarding independently associated hospital-acquired VTE risk factors including the presence of CVC, concurrent infection, and extended length of stay (LOS). Among children who are critically ill, Jaffray et al from the Children’s Healthcare Advancements in Thrombosis Consortium derived a hospital-acquired VTE risk prediction model from a retrospective multi-institutional consortium–derived data set that identified acute immobility (measured by Braden Q values), intensive care unit LOS, and the presence of CVC as independent risk factors.² Similarly, a narrative review by Sochet et al along with other multi-institutional and single-center reports, found that the presence of a CVC, extended LOS, concurrent infection, and impaired mobility are salient VTE risk factors among children who are critically ill.^6,19,20,31 Among those hospitalized for management of congenital and acquired heart disease, studies seeking to identify the incidence and risk factors for hospital-acquired VTE likewise identified the presence of a CVC and extended LOS to be associated with hospital-acquired VTE irrespective of perioperative and anatomic variation that may infer prothrombotic risk.^17,32,33 Although multicenter collaborative initiatives supported by the Children’s Healthcare Advancements in Thrombosis Consortium are underway to prospectively validate pediatric risk prediction models for hospital-acquired VTE for all hospitalized children, the consistency regarding risk factors across pediatric subpopulations as described herein infer that a core set of prothrombotic risk factors are consistently relevant in pediatrics and could assist in the approach to the prescription of pharmacological TP. Guidelines for the application of pharmacological TP in children that address universal TP and distinct at-risk subpopulations and settings for VTE are presently lacking, although efforts in this regard are underway through the American Society of Hematology. This is, in part, because of sparce trial-derived evidence for the efficacy and safety of pharmacological TP in pediatrics. Existing trials, reviewed hereafter, have focused on subpopulations including those undergoing cardiothoracic surgery, children who are critically ill with CVC, those with hematologic malignancy, and those hospitalized for COVID-19–related illness.

Hospitalized children, particularly those critically ill after major surgery or trauma, are known to be at elevated risk of International Society on Thrombosis and Haemostasis (ISTH)-defined bleeding. White et al, in a single-center cohort study of >400 children who are critically ill without bleeding observed the first 24-hours of hospitalization, estimated the rate of clinically relevant bleeding during hospitalization to be 9.1%.⁸ In a prospective cohort study of children who are critically ill, Greenway et al estimated the frequency of clinically relevant bleeding of 10% with most attributable to surgical site bleeding or those related to recent trauma.¹⁰ Children hospitalized after cardiothoracic surgery including cases requiring cardiopulmonary bypass are at elevated and anticipated risk of postoperative bleeding.³⁴ Two meta-analyses have estimated the risk of clinically relevant bleeding to be between 0.6% and 2.3% among hospitalized children receiving anticoagulant TP.^35,36 Taken in sum, the approach to primary VTE prevention must consider the net clinical benefit of pharmacological TP, a balance between the potential risk reduction in VTE and the risk increase in clinically relevant bleeding attributable to anticoagulation.

With this in mind, we review the trial-derived safety and efficacy findings for pharmacological TP in populations relating to the cases described earlier. Then, we describe how we approach pharmacological TP for each scenario, for the clinical intention of TE prevention. As appropriate, we also refer to evidence on dosing, duration, and considerations for laboratory monitoring.

Three recent phase 2 and phase 3 multicenter trials have been conducted evaluating direct oral anticoagulants (DOACs) as pharmacological TP for children with congenital and acquired heart disease in the outpatient setting. None of the trials was powered to definitively test hypotheses on comparative efficacy or safety, but each provided descriptive analyses of the frequencies of TE and bleeding in patients at risk for TE treated with DOACs or standard-of-care comparator antithrombotic agents. UNIVERSE was a multicenter, open-label, randomized clinical trial (RCT) that evaluated rivaroxaban in children with single-ventricle physiology who underwent a Fontan procedure, conducted via a pharmacokinetics/pharmacodynamics safety phase followed by a randomized phase comparing rivaroxaban with aspirin.²⁵ Rates of bleeding and thrombotic events were qualitatively similar for rivaroxaban and aspirin. ENNOBLE-ATE was an multicenter, open-label RCT that compared edoxaban with standard care (ie, low-molecular-weight heparin [LMWH] and vitamin K antagonists) among children with congenital and acquired cardiac disease and noted no thromboembolic events in the edoxaban group and qualitatively similar cumulative incidences of major bleeding.²⁶ SAXOPHONE was a multicenter, open-label RCT that investigated apixaban vs standard care in a similar study population to that of ENNOBLE-ATE and observed no thromboembolic events in either group and qualitatively similar cumulative incidences of major bleeding.²⁸ Although these trials were not powered for comparative efficacy and safety and are limited, in part, by study population heterogeneity, they represent a growing body of evidence in support of DOACs as additional options to conventional antithrombotic agents (ie, warfarin, LMWH, and aspirin) for pharmacological TP in at-risk pediatric populations with cardiac disease.

This case describes a patient with post-Fontan congenital heart disease who has ongoing prothrombotic risk factors, a history of VTE, and underlying mild inherited thrombophilia (heterozygosity for factor 2 G20210A). In the case of post-Fontan congenital heart disease patients who do not have a history of VTE, a fenestrated Fontan, or strong risk factors for TE (ie, central venous or arterial catheters, infection, etc), we generally prescribe aspirin monotherapy as our agent of choice given corresponding data from phase 3 trials and observational cohort studies. For those who do have a history of VTE, a fenestrated Fontan, or strong TE risk factors, we generally use anticoagulant monotherapy for pharmacological TP. We consider using unfractionated heparin or bivalirudin in the immediate postoperative setting and transitioning to a longer-acting agent (eg, LMWH) once a child is at lower perceived risk of immediate postoperative bleeding. For long-term TE prevention, once this child with history of Fontan and prior VTE is beyond the acute postoperative period, we consider using a DOAC, presuming end-organ function (ie, no evidence of hepatic impairment or renal dysfunction) remains stable. Our approach to duration of pharmacological TP is in children with congenital and acquired cardiac disease is affected by the clinical intention and the type and duration of prothrombotic risk factors. For example, for primary TE prevention in the presence of acute low cardiac output syndrome, acute infection, and temporary CVC among patients with biventricular physiology, the duration of anticoagulation for pharmacological TP extends until the aforementioned risk factor is no longer present. By contrast, in single-ventricle physiology or 1 in which there is chronically turbulent blood flow with a prior history of VTE, long-term pharmacologic TP may be prescribed. In general, we avoid dual anticoagulant/antiplatelet therapy because of the risk of increased bleeding events. In the cases of DOACs, we do not have preference for a specific DOAC to be used for pharmacological TP in children with congenital and acquired cardiac diseases. As reviewed earlier, several DOACs have been studied in multicenter trials in cardiac populations (including various proportions of patients with Kawasaki disease, following a Fontan procedure, and other “at-risk” populations for TE), and none was definitively powered to test efficacy or safety relative to conventional anticoagulants. Additionally, our approach includes pragmatic considerations for real-world practice, including route of delivery (eg, “nothing by mouth” status), product formulation (eg, apixaban only comes in tablets, and rivaroxaban has a liquid formulation), and drug availability (eg, on-formulary for inpatients, approved by insurance for outpatients, etc). For dosing of DOACs as secondary prophylaxis in cases with a history of VTE, we generally use half-therapeutic dosing in children in whom we consider prothrombotic risk to be moderate and full therapeutic dosing among those with high prothrombotic risk. If children in the former group are hospitalized, we recommend escalating secondary pharmacological TP dosing to full therapeutic dose in the absence of increased bleeding risk. Regarding laboratory monitoring for this case and for the case discussions to follow, we follow laboratory monitoring assays in the “prophylactic” range as published by consensus-based recommendations for pediatric antithrombotic therapy, or when none provided therein, in a range just below the lower limit of the “therapeutic” range. For example, we generally target 0.1 to <0.35 anti-Xa U/mL for unfractionated heparin, 0.2 to <0.5 anti-Xa U/mL for LMWH, and an activated partial thromboplastin time ratio on treatment to baseline of 1.2 to <1.5 for bivalirudin.

For children who are critically ill without congenital heart disease, pharmacological TP trials have focused on risk reduction for children hospitalized for COVID-19–related illness and those at risk for CVC-related deep venous thrombosis (DVT). With the onset of the COVID-19 pandemic and the associated recognition that hospitalized children were experiencing heightened rates of VTE,³⁷ consensus-based pediatric recommendations for pharmacological TP were published by the ISTH, which recommended pharmacological TP for children hospitalized for COVID-19 with elevated D-dimer levels or other superimposed risk factors.³⁸ These guidelines issued a call to action for rapid activation of clinical trials that would assess the dosing safety and efficacy of pharmacological TP strategies for primary TE prevention in children hospitalized with COVID-19. In response, the COVAC-TP trial, an investigator-initiated, multicenter, open-label, phase 2 trial, was designed and rapidly launched by Sochet et al in 2020.²⁷ Twice daily subcutaneous enoxaparin throughout hospitalization as pharmacological TP for children aged <18 years hospitalized for COVID-19–related illness including MIS-C was administered. No child had major bleeding, and 92% achieved prophylactic anti-XA activity levels (ie, the primary end point). However, hospital-acquired VTE occurred in 5% of the study population, with both instances representing CVC-related DVT among those with primary COVID-19 infection, begging the question of dose requirements for pharmacological TP in this population.

For children hospitalized for COVID-19 in the context of an elevated D-dimer, CVC, personal or family history of young-onset VTE or known thrombophilia, and otherwise acceptable bleeding risk we use and recommend pharmacological TP with LMWH, citing the evidence of COVAC-TP trial, and acknowledging that the trial was not designed to test for superiority or noninferiority relative to standard care (no pharmacological TP), given that COVAC-TP was a single-arm phase 2 trial. In this instance, we prescribe enoxaparin 0.5 mg/kg per dose twice daily (maximum: 30 mg per dose) targeting LMWH anti-Xa levels as described earlier in case 1. Combined with other risk factors for VTE present in this clinical scenario (ie, chronic exposure to oral contraceptives, prolonged hospital LOS, and obesity) among children who are noncritically ill as described by Morrison et al,¹⁶ our concern for hospital-acquired VTE risk in this scenario was heightened. Similar to the other scenarios provided, the duration of pharmacologic TP is prescribed at minimum throughout the period of acute hyperinflammatory symptoms associated with MIS-C (ie, fevers, evidence of organ dysfunction, and persistently elevated D-dimer levels [>5 the upper limit of normal]). Current guidelines issued by the ISTH recommend against the use of pharmacological TP in children determined to have asymptomatic COVID-19 infection in the absence of additional risk factors. At our institution, we recommend pharmacological TP throughout hospitalization, and as reported by Goldenberg et al³⁸ and Rajput et al,³⁹ when strong prothrombotic risk factors persist beyond the duration of hospitalization (eg, the combination of obesity and estrogen-containing oral contraception as described in the case, or an indwelling CVC, and persistently elevated D-dimer) for up to 30 days after discharge in consultation and close follow-up with a pediatric hematologist.⁴⁰ More recent epidemiologic data suggest the rate of VTE during and after COVID-19–related illness may be declining and vary by clinical presentation, viral strain, and study definitions.^41,42 

Faustino et al, in collaboration with the Pediatric Acute Lung Injury and Sepsis Consortium, conducted a Bayesian phase 2b RCT (the CRETE study) of enoxaparin vs standard care among children hospitalized in the pediatric intensive care unit without critical congenital heart disease after the placement of a percutaneous femoral or internal jugular CVC.²⁴ Of note, clinically relevant CVC-associated DVT occurred in 3.7% of the enoxaparin arm and 29.2% of the usual-care arm, with only 1 instance of major bleeding observed. The PROTEKT trial, although not specific to children who are critically ill, was a multicenter open-label RCT of reviparin vs standard care (no pharmacological TP) for the prevention of CVC-associated DVT in hospitalized children.²² Among 172 children (50% with cancer and 22.5% with congenital heart disease) who completed the trial (78 randomized to reviparin and 93 to standard of care without anticoagulation), no major bleeding was observed in the reviparin group, and difference was discerned in the cumulative incidence of DVT as measured by surveillance venograms. However, the CRETE study was not powered to definitively evaluate the efficacy and safety of pharmacological TP for pediatric CVC-related DVT prevention relative to the standard of care (no pharmacological TP), and the PROTEKT study did not reach its definitively powered sample size for same. The COVAC-TP findings²⁷ and a secondary analysis of the CRETE study⁴³ suggest that TP dosing of LMWH, as determined by achievement of plasma anti-Xa activity levels between 0.2 and 0.49 IU/mL, may be insufficient. Additional data on pharmacodynamically driven dose requirements for inpatient pharmacological TP (eg, age-based dose requirements to achieve targeted anti-Xa activity levels for LMWHs) are needed, particularly among children who are critically ill, as are phase 3 RCTs definitively powered to assess comparative efficacy and safety of pharmacological TP vs standard care (no pharmacological TP).

Because the case describes a child who is critically ill with multiple prothrombotic risk factors (eg, CVC, a prolonged LOS, complete immobility with neuromuscular blockade, and concurrent infection), we recommend pharmacological TP with LMWH in the absence of significant renal impairment, active bleeding, a known history of inherited or acquired bleeding disorder, or laboratory evidence of coagulopathy (eg, thrombocytopenia and hypofibrinogenemia) for the prevention of VTE and recurrent VTE.⁴⁴ Should an emergent or impending surgical procedure arise, we choose unfractionated heparin or bivalirudin in lieu of LMWH, for more rapid extinction of anticoagulant effect. Although the risk of hospital-acquired VTE among pediatric patients who are chronically immobilized, as described in the vignette, is unclear, the acute impairment of mobility should raise concern for prothrombotic risk and is an area of research interest. The use of mechanical TP would not dissuade us from prescribing pharmacological TP in this context of a children who are critically ill with several prothrombotic risk factors, given that pharmacological TP (whether alone or in combination with mechanical TP) was recently shown to be independently associated with a reduced odds of hospital-acquired VTE in a multicenter analysis of the Pediatric Hospital Information Systems database by Betensky et al.⁴⁵ While we await the completion and analysis of the ongoing phase 3 CRETE RCT of LMWH vs no pharmacological TP for primary prevention of CVC-related DVT in children who are medically critically ill, the presence of a “high-VTE-risk” profile as defined by Arlikar et al (ie, presence of a CVC, comorbid infection, and anticipated LOS of ≥4 days⁶) supports our decision to prescribe pharmacological TP in this context. In regard to the duration of pharmacological TP, in this case, we typically continue LMWH until the sooner of intensive care unit discharge or CVC is removal and completion of antimicrobial therapy for infection. If a child who is critically ill is hospitalized with a prior VTE history, we prescribe secondary pharmacological TP for hospital-acquired VTE prevention if the prior VTE was hospital-acquired or if the patient has a known underlying thrombophilia state and/or positive first-degree family history of early-onset (aged <50 years) VTE using the anticoagulants already described herein targeting laboratory levels as described.

Pediatric trials have also evaluated pharmacological TP in the context of hospitalized children undergoing induction chemotherapy for ALL and/or lymphoma (ALL/L). The THROMBOTECT trial, an investigator-initiated multicenter, open-label, phase 3 RCT comparing LMWH, antithrombin, and unfractionated heparin as pharmacological TP for primary VTE prevention in pediatric patients with ALL from day 8 to day 33 of induction chemotherapy.²³ Of the 949 participants randomized, 95% had a CVC, and treatment groups were well balanced regarding prognostic indicators of disease outcome. Cumulative incidence of VTE (not centrally adjudicated) within 3 months of enrollment was 8% for unfractionated heparin, 3.5% for enoxaparin, and 1.9% for antithrombin concentrate; the reduced cumulative incidences of VTE in the latter 2 arms, when compared with the former, were statistically significant. The cumulative incidence of bleeding events of all severities (major, clinically relevant nonmajor, and minor) within 3 months of enrollment was 1% among patients treated with unfractionated heparin, 0.5% of enoxaparin-treated patients, and 0.9% of antithrombin-treated patients. The recently reported PREVAPIX-ALL trial, a multicenter open-label phase 3 RCT, studied apixaban as pharmacological TP in comparison with standard care (ie, no pharmacological TP) among children with newly-diagnosed ALL/L undergoing induction chemotherapy with a newly placed CVC.³⁰ Of the 438 children who completed study procedures, the cumulative incidence of adjudicated VTE events or VTE-related death at ∼1 month follow-up was 12% in the apixaban arm and 18% in the comparator arm (no pharmacological TP), with a relative risk reduction that was not statistically significant. Two major bleeding events occurred in each group, without detectable differences in other adverse events.

In pediatric patients with ALL/L in induction who have a CVC and known thrombophilia or a personal or family history of young-onset VTE, and who have low bleeding risk, we strongly consider using pharmacological TP with a DOAC or LMWH (goal LMWH anti-Xa level). Based on the quality of the available evidence, we do not have a preference for either agent. By contrast, given the equivocal findings and/or questionable net clinical benefit in otherwise unselected children meeting the characteristics of those studied in the aforementioned trials, we do not routinely use primary or secondary pharmacological TP in pediatric patients with ALL/L in induction in the absence of a personal family history of young-onset thrombosis or known thrombophilia. As in the prior case of the child who is medically critical ill, if a child with ALL/L in induction has a prior VTE history, we prescribe secondary pharmacological TP for hospital-acquired VTE using the risk-screening and anticoagulants, as described earlier.

The risk of TE is increased in several pediatric subpopulations, some of which have been studied in multicenter RCTs of pharmacological TP completed over the past decade. These include children with congenital and acquired cardiac diseases who have prothrombotic risk factors, as well as patients with cancer, particularly hematological malignancies, during induction therapy with a CVC. Although evidence-based guidelines are mostly lacking for pharmacological TP in children, efforts are underway for guideline development. In the meantime, approaches can be considered based on clinical trial evidence, despite the important limitations of each of the trials published. It is becoming increasingly apparent that approaches to pharmacological TP need to be risk-stratified in regard to the risks of both VTE and clinically relevant bleeding. Key unanswered questions include (1) comparative efficacy and safety of pharmacological TP relative to standard care in numerous subpopulations and settings in which TE risk is heightened; (2) the optimal intensity and duration of pharmacological TP in each subpopulation and setting (including the question of continued pharmacological TP at home, for hospitalized children with persistent prothrombotic risk factors at hospital discharge); and (3) the role of mechanical TP. We call ourselves and our colleagues to action in designing and completing definitive RCTs, some of which are underway or have been recently developed, to address these critical knowledge gaps in pediatric care.

Contribution: All authors designed and wrote the manuscript.

Conflict-of-interest disclosure: A.K. is on a steering committee for a Boehringer-Ingelheim Pharmaceuticals: Pradaxa Pellet 1160-0309 Study (in kind support). N.A.G. receives or has recently received consultancy fees from Bayer, Johnson & Johnson, and the University of Colorado–affiliated Academic Research Organization Colorado Prevention Center Clinical Research for roles in clinical trial planning or oversight committees in pharmaceutical industry–sponsored pediatric multicenter clinical trials of antithrombotic agents; he receives salary support from the National Heart, Lung, and Blood Institute, National Institutes of Health (1K24HL166791); he is a member of the Pedi-ATLAS Group; his employer receives salary support on his behalf from Boehringer-Ingelheim for data coordinating center leadership for a pediatric antithrombotic multicenter prospective observational study; and he is a member of the American Society of Hematology subcommittee on clinical trials. A.A.S. receives support from the National Heart, Lung, and Blood Institute, National Institutes of Health (1K23HL177270).

Correspondence: Anthony A. Sochet, Johns Hopkins All Children’s Hospital, Anesthesia and Critical Care Medicine, Research and Education Building, Room 3220, 600 5th St S, St. Petersburg, FL 33701; email: sochet@jhmi.edu.