Vitamin K antagonists (VKAs) such as warfarin have been the backbone of oral anticoagulation in clinical practice since the middle of the last century. Despite their efficacy, VKAs have well‐recognized limitations that have led to their underutilization in patients who would otherwise be candidates for oral anticoagulation.14 These limitations include a narrow therapeutic window and significant intra‐ and interindividual variability in dose requirements as well as numerous drugdrug and drugfood interactions.59 Therefore, VKAs require close laboratory monitoring to prevent excessive or under‐anticoagulation, and maintaining therapeutic anticoagulation with VKAs remains a challenging task in many patients.2 It has been shown that 30%50% of international normalized ratio (INR) results fall outside of the targeted therapeutic range.10, 11 Consequently, it is not surprising that warfarin is a common cause of medication‐related emergency room visits.12 Despite many fruitless years of searching for better alternatives, VKAs have remained the mainstay of oral anticoagulation for more than 60 years.8

An ideal anticoagulant would be orally administered, effective, safe, exhibit a predictable pharmacokinetic profile and a low potential for drug or dietary interactions, and therefore would not require routine laboratory monitoring.2, 5, 13 Other desirable characteristics would include a rapid onset of action to decrease or eliminate the need for bridging therapy, and rapid reversibility with or without an antidote.8, 13 To date, no oral anticoagulant has been developed that possesses all of these desired characteristics. Dabigatran etexilate (Pradaxa, Boehringer Ingelheim Pharmaceuticals, Inc.) has recently become the first oral anticoagulant to be available for wide clinical use since the 1950s.14 In the following sections, we provide an overview of dabigatran etexilate, with a special focus on issues that are pertinent to hospitalists and the hospitalized patient.

PHARMACOLOGY OF DABIGATRAN ETEXILATE

Pharmacokinetics and Pharmacodynamics of Dabigatran Etexilate

A comparison of the pharmacokinetic (PK) and pharmacodynamic (PD) properties of dabigatran etexilate (dabigatran) and warfarin are presented in Table 1. Dabigatran etexilate (referred to from this point as dabigatran) is a prodrug of dabigatran, which blocks the terminal coagulation cascade by binding to the active site of thrombin and selectively inhibiting this critical serine protease in a dose‐dependent and reversible fashion.15 Thrombin plays a central role in blood coagulation by converting fibrinogen to fibrin, amplifying its own generation by feedback activation of factors V, VIII, and XI, and by activating platelets (Figure 1).16 Dabigatran is a direct thrombin inhibitor that acts independently of anti‐thrombin to inhibit both free and clot‐bound thrombin.17, 18 The bioavailability of dabigatran after oral intake is low (6%7%).1923 After absorption, the prodrug is rapidly converted by plasma and hepatic esterases to the active drug dabigatran, but it is not metabolized by the CYP‐450 system, therefore reducing the potential for drugdrug interactions.8, 2328 The long half‐life of dabigatran allows for once or twice daily dosing.21, 24 The PK profile of dabigatran is predictable, with minimal inter‐ and intraindividual variation.21, 22

Figure 1

Table 1.Comparison of Warfarin and Dabigatran

	Warfarin	Dabigatran
Abbreviations: CYP‐450, cytochrome P‐450. INR, international normalized ratio; P‐gp, P‐glycoprotein. Modified from reference 9.5, 6, 8, 16, 19, 2022, 24, 28, 30, 36, 4244
Mechanism of action	Reduces functional levels of vitamin Kdependent factors II, VII, IX, and X by inhibiting vitamin K epoxide reductase	Binds to active site of thrombin (factor IIa) and reversibly inhibits free and clot‐bound thrombin
Prodrug	No	Yes
Bioavailability	>90%95%	6%7%
Protein binding	99%	35%
Time to reach peak plasma levels	7296 hr	23 hr
Half‐life	3644 hr	1217 hr
Routine coagulation monitoring	Required, but frequency varies based on clinical situation	No requirement for routine monitoring
Schedule	INR‐adjusted, usually once daily	Fixed dose, once or twice daily
Metabolism	CYP‐450 hepatic microsomal enzymes, especially CYP2C9, CYP1A2, and CYP3A4	Esterase‐catalyzed hydrolysis in plasma or liver after intestinal P‐gp transport
Clearance	Almost entirely hepatic	80% unchanged renally (after an intravenous dose), 20% hepatic after conjugation
Drug interactions	Drugs that affect CYP‐450 hepatic microsomal enzymes and those that displace warfarin from plasma proteins	P‐gp inhibitors (CYP‐450 system not involved)
Antidote	Yes (vitamin K and plasma products)	No

Dabigatran is packaged in capsules that are hygroscopic. Therefore, the capsules should be stored in the original container with the cap tightly closed. Exposure of dabigatran capsules to air for prolonged periods outside the original container can result in deterioration of the active compound and reduced efficacy.27, 28 Dabigatran capsules contain tartaric acid which is necessary to facilitate dissolution of the medication in the gastrointestinal tract for optimal absorption.2 Breaking the capsules or removing the drug from the capsule can result in increased exposure. Therefore, dabigatran capsules should be taken intact, and patients should be instructed that dabigatran capsules should not be broken, chewed, or opened before administration.28 Alternative anticoagulants should be used if patients cannot swallow the capsule intact for any reason (eg, intubated patients).

EFFICACY OF DABIGATRAN

In this section, we provide a brief review of the major phase 3 trials that evaluated dabigatran for different indications (see references 13, 16, and 24 for recent detailed reviews of the clinical trials of dabigatran).

SAFETY OF DABIGATRAN

Aside from the bleeding risks discussed earlier, the most commonly reported side effect of dabigatran was dyspepsia. Dyspepsia occurred twice as frequently in patients taking dabigatran versus warfarin in the RE‐LY trial (11.5% vs 5.8%).35 One possible explanation for the higher incidence of dyspepsia is the tartaric acid component in dabigatran capsules.2 In the RE‐LY study, myocardial infarction occurred more commonly in the dabigatran arms (0.72% with 110 mg and 0.74% with 150 mg) than the warfarin arm (0.53%, P = 0.07 and 0.048, respectively).24, 35 It has been postulated that this observation could be related to a greater efficacy of warfarin for the prevention of myocardial infarction rather than an adverse effect of dabigatran.2 There was no increase in acute coronary syndrome rates noted with dabigatran in the other phase 3 trials.3234, 38, 40 No increased risk of elevated liver function test has been noted with dabigatran, but long‐term data are unavailable.32, 34, 35, 38

MANAGEMENT OF SPECIAL SITUATIONS THAT MAY ARISE IN THE USE OF DABIGATRAN

CONCLUSIONS

Dabigatran is a novel, oral direct thrombin inhibitor that exhibits several advantages over warfarin. The predictable pharmacokinetic profile and minimal food and drug interactions of dabigatran allow for a fixed‐dosing regimen and obviate the need for routine laboratory monitoring. However, this apparent advantage is also a disadvantage. The lack of a reliable method to monitor dabigatran makes it more difficult to assess compliance, measure the impact of drug interactions, evaluate for toxicity, and determine bona fide therapeutic failure versus noncompliance in the event of breakthrough thromboembolism.28, 42 Other limitations of dabigatran include the lack of an antidote and the dependence on normal renal function for elimination, with the potential for drug accumulation and toxicity with renal impairment. The noninferiority design of the clinical trials that evaluated dabigatran, the absence of long‐term safety and efficacy data, and issues related to the cost effectiveness of dabigatran should be considered when prescribing this agent. More studies are needed to assess dabigatran in special patient populations (eg, the elderly, patients with renal and hepatic impairment, pediatric and pregnant patients) and to better understand dabigatrandrug interactions.

As more novel oral anticoagulant agents, such as factor Xa inhibitors, become available for clinical use, comparative studies will need to be performed to better define the role of each agent for specific indications. In the future, it might be possible to tailor the choice of the oral anticoagulant to the individual patient not only on the basis of the clinical indication but also the specific patient characteristics and possible drug interactions. For example, rivaroxaban (Xarelto) is an oral direct factor Xa that was recently approved in the United States for VTE thromboprophylaxis following orthopedic surgery and in patients with non‐valvular atrial fibrillation.2 Similar to dabigatran, rivaroxaban exhibits predictable PK and PD that allow fixed once or twice daily dosing and obviate the need for routine monitoring of its anticoagulant effects.2, 16 Unlike dabigatran, rivaroxaban is an active drug and not a prodrug, and has a significantly higher bioavailability than dabigatran (>80% vs 6%).16 In addition, the levels of rivaroxaban can be affected by drugs that interfere with both P‐gp and the hepatic CYP‐450 system, compared with dabigatran, which is affected only by drugs that affect P‐gp.8, 16

Vitamin K antagonists (VKAs) such as warfarin have been the backbone of oral anticoagulation in clinical practice since the middle of the last century. Despite their efficacy, VKAs have well‐recognized limitations that have led to their underutilization in patients who would otherwise be candidates for oral anticoagulation.14 These limitations include a narrow therapeutic window and significant intra‐ and interindividual variability in dose requirements as well as numerous drugdrug and drugfood interactions.59 Therefore, VKAs require close laboratory monitoring to prevent excessive or under‐anticoagulation, and maintaining therapeutic anticoagulation with VKAs remains a challenging task in many patients.2 It has been shown that 30%50% of international normalized ratio (INR) results fall outside of the targeted therapeutic range.10, 11 Consequently, it is not surprising that warfarin is a common cause of medication‐related emergency room visits.12 Despite many fruitless years of searching for better alternatives, VKAs have remained the mainstay of oral anticoagulation for more than 60 years.8

An ideal anticoagulant would be orally administered, effective, safe, exhibit a predictable pharmacokinetic profile and a low potential for drug or dietary interactions, and therefore would not require routine laboratory monitoring.2, 5, 13 Other desirable characteristics would include a rapid onset of action to decrease or eliminate the need for bridging therapy, and rapid reversibility with or without an antidote.8, 13 To date, no oral anticoagulant has been developed that possesses all of these desired characteristics. Dabigatran etexilate (Pradaxa, Boehringer Ingelheim Pharmaceuticals, Inc.) has recently become the first oral anticoagulant to be available for wide clinical use since the 1950s.14 In the following sections, we provide an overview of dabigatran etexilate, with a special focus on issues that are pertinent to hospitalists and the hospitalized patient.

PHARMACOLOGY OF DABIGATRAN ETEXILATE

Pharmacokinetics and Pharmacodynamics of Dabigatran Etexilate

A comparison of the pharmacokinetic (PK) and pharmacodynamic (PD) properties of dabigatran etexilate (dabigatran) and warfarin are presented in Table 1. Dabigatran etexilate (referred to from this point as dabigatran) is a prodrug of dabigatran, which blocks the terminal coagulation cascade by binding to the active site of thrombin and selectively inhibiting this critical serine protease in a dose‐dependent and reversible fashion.15 Thrombin plays a central role in blood coagulation by converting fibrinogen to fibrin, amplifying its own generation by feedback activation of factors V, VIII, and XI, and by activating platelets (Figure 1).16 Dabigatran is a direct thrombin inhibitor that acts independently of anti‐thrombin to inhibit both free and clot‐bound thrombin.17, 18 The bioavailability of dabigatran after oral intake is low (6%7%).1923 After absorption, the prodrug is rapidly converted by plasma and hepatic esterases to the active drug dabigatran, but it is not metabolized by the CYP‐450 system, therefore reducing the potential for drugdrug interactions.8, 2328 The long half‐life of dabigatran allows for once or twice daily dosing.21, 24 The PK profile of dabigatran is predictable, with minimal inter‐ and intraindividual variation.21, 22

Figure 1

Table 1.Comparison of Warfarin and Dabigatran

	Warfarin	Dabigatran
Abbreviations: CYP‐450, cytochrome P‐450. INR, international normalized ratio; P‐gp, P‐glycoprotein. Modified from reference 9.5, 6, 8, 16, 19, 2022, 24, 28, 30, 36, 4244
Mechanism of action	Reduces functional levels of vitamin Kdependent factors II, VII, IX, and X by inhibiting vitamin K epoxide reductase	Binds to active site of thrombin (factor IIa) and reversibly inhibits free and clot‐bound thrombin
Prodrug	No	Yes
Bioavailability	>90%95%	6%7%
Protein binding	99%	35%
Time to reach peak plasma levels	7296 hr	23 hr
Half‐life	3644 hr	1217 hr
Routine coagulation monitoring	Required, but frequency varies based on clinical situation	No requirement for routine monitoring
Schedule	INR‐adjusted, usually once daily	Fixed dose, once or twice daily
Metabolism	CYP‐450 hepatic microsomal enzymes, especially CYP2C9, CYP1A2, and CYP3A4	Esterase‐catalyzed hydrolysis in plasma or liver after intestinal P‐gp transport
Clearance	Almost entirely hepatic	80% unchanged renally (after an intravenous dose), 20% hepatic after conjugation
Drug interactions	Drugs that affect CYP‐450 hepatic microsomal enzymes and those that displace warfarin from plasma proteins	P‐gp inhibitors (CYP‐450 system not involved)
Antidote	Yes (vitamin K and plasma products)	No

Dabigatran is packaged in capsules that are hygroscopic. Therefore, the capsules should be stored in the original container with the cap tightly closed. Exposure of dabigatran capsules to air for prolonged periods outside the original container can result in deterioration of the active compound and reduced efficacy.27, 28 Dabigatran capsules contain tartaric acid which is necessary to facilitate dissolution of the medication in the gastrointestinal tract for optimal absorption.2 Breaking the capsules or removing the drug from the capsule can result in increased exposure. Therefore, dabigatran capsules should be taken intact, and patients should be instructed that dabigatran capsules should not be broken, chewed, or opened before administration.28 Alternative anticoagulants should be used if patients cannot swallow the capsule intact for any reason (eg, intubated patients).

EFFICACY OF DABIGATRAN

In this section, we provide a brief review of the major phase 3 trials that evaluated dabigatran for different indications (see references 13, 16, and 24 for recent detailed reviews of the clinical trials of dabigatran).

SAFETY OF DABIGATRAN

Aside from the bleeding risks discussed earlier, the most commonly reported side effect of dabigatran was dyspepsia. Dyspepsia occurred twice as frequently in patients taking dabigatran versus warfarin in the RE‐LY trial (11.5% vs 5.8%).35 One possible explanation for the higher incidence of dyspepsia is the tartaric acid component in dabigatran capsules.2 In the RE‐LY study, myocardial infarction occurred more commonly in the dabigatran arms (0.72% with 110 mg and 0.74% with 150 mg) than the warfarin arm (0.53%, P = 0.07 and 0.048, respectively).24, 35 It has been postulated that this observation could be related to a greater efficacy of warfarin for the prevention of myocardial infarction rather than an adverse effect of dabigatran.2 There was no increase in acute coronary syndrome rates noted with dabigatran in the other phase 3 trials.3234, 38, 40 No increased risk of elevated liver function test has been noted with dabigatran, but long‐term data are unavailable.32, 34, 35, 38

MANAGEMENT OF SPECIAL SITUATIONS THAT MAY ARISE IN THE USE OF DABIGATRAN

CONCLUSIONS

Dabigatran is a novel, oral direct thrombin inhibitor that exhibits several advantages over warfarin. The predictable pharmacokinetic profile and minimal food and drug interactions of dabigatran allow for a fixed‐dosing regimen and obviate the need for routine laboratory monitoring. However, this apparent advantage is also a disadvantage. The lack of a reliable method to monitor dabigatran makes it more difficult to assess compliance, measure the impact of drug interactions, evaluate for toxicity, and determine bona fide therapeutic failure versus noncompliance in the event of breakthrough thromboembolism.28, 42 Other limitations of dabigatran include the lack of an antidote and the dependence on normal renal function for elimination, with the potential for drug accumulation and toxicity with renal impairment. The noninferiority design of the clinical trials that evaluated dabigatran, the absence of long‐term safety and efficacy data, and issues related to the cost effectiveness of dabigatran should be considered when prescribing this agent. More studies are needed to assess dabigatran in special patient populations (eg, the elderly, patients with renal and hepatic impairment, pediatric and pregnant patients) and to better understand dabigatrandrug interactions.

As more novel oral anticoagulant agents, such as factor Xa inhibitors, become available for clinical use, comparative studies will need to be performed to better define the role of each agent for specific indications. In the future, it might be possible to tailor the choice of the oral anticoagulant to the individual patient not only on the basis of the clinical indication but also the specific patient characteristics and possible drug interactions. For example, rivaroxaban (Xarelto) is an oral direct factor Xa that was recently approved in the United States for VTE thromboprophylaxis following orthopedic surgery and in patients with non‐valvular atrial fibrillation.2 Similar to dabigatran, rivaroxaban exhibits predictable PK and PD that allow fixed once or twice daily dosing and obviate the need for routine monitoring of its anticoagulant effects.2, 16 Unlike dabigatran, rivaroxaban is an active drug and not a prodrug, and has a significantly higher bioavailability than dabigatran (>80% vs 6%).16 In addition, the levels of rivaroxaban can be affected by drugs that interfere with both P‐gp and the hepatic CYP‐450 system, compared with dabigatran, which is affected only by drugs that affect P‐gp.8, 16

Vitamin K antagonists (VKAs) such as warfarin have been the backbone of oral anticoagulation in clinical practice since the middle of the last century. Despite their efficacy, VKAs have well‐recognized limitations that have led to their underutilization in patients who would otherwise be candidates for oral anticoagulation.14 These limitations include a narrow therapeutic window and significant intra‐ and interindividual variability in dose requirements as well as numerous drugdrug and drugfood interactions.59 Therefore, VKAs require close laboratory monitoring to prevent excessive or under‐anticoagulation, and maintaining therapeutic anticoagulation with VKAs remains a challenging task in many patients.2 It has been shown that 30%50% of international normalized ratio (INR) results fall outside of the targeted therapeutic range.10, 11 Consequently, it is not surprising that warfarin is a common cause of medication‐related emergency room visits.12 Despite many fruitless years of searching for better alternatives, VKAs have remained the mainstay of oral anticoagulation for more than 60 years.8

An ideal anticoagulant would be orally administered, effective, safe, exhibit a predictable pharmacokinetic profile and a low potential for drug or dietary interactions, and therefore would not require routine laboratory monitoring.2, 5, 13 Other desirable characteristics would include a rapid onset of action to decrease or eliminate the need for bridging therapy, and rapid reversibility with or without an antidote.8, 13 To date, no oral anticoagulant has been developed that possesses all of these desired characteristics. Dabigatran etexilate (Pradaxa, Boehringer Ingelheim Pharmaceuticals, Inc.) has recently become the first oral anticoagulant to be available for wide clinical use since the 1950s.14 In the following sections, we provide an overview of dabigatran etexilate, with a special focus on issues that are pertinent to hospitalists and the hospitalized patient.

PHARMACOLOGY OF DABIGATRAN ETEXILATE

Pharmacokinetics and Pharmacodynamics of Dabigatran Etexilate

A comparison of the pharmacokinetic (PK) and pharmacodynamic (PD) properties of dabigatran etexilate (dabigatran) and warfarin are presented in Table 1. Dabigatran etexilate (referred to from this point as dabigatran) is a prodrug of dabigatran, which blocks the terminal coagulation cascade by binding to the active site of thrombin and selectively inhibiting this critical serine protease in a dose‐dependent and reversible fashion.15 Thrombin plays a central role in blood coagulation by converting fibrinogen to fibrin, amplifying its own generation by feedback activation of factors V, VIII, and XI, and by activating platelets (Figure 1).16 Dabigatran is a direct thrombin inhibitor that acts independently of anti‐thrombin to inhibit both free and clot‐bound thrombin.17, 18 The bioavailability of dabigatran after oral intake is low (6%7%).1923 After absorption, the prodrug is rapidly converted by plasma and hepatic esterases to the active drug dabigatran, but it is not metabolized by the CYP‐450 system, therefore reducing the potential for drugdrug interactions.8, 2328 The long half‐life of dabigatran allows for once or twice daily dosing.21, 24 The PK profile of dabigatran is predictable, with minimal inter‐ and intraindividual variation.21, 22

Figure 1

Table 1.Comparison of Warfarin and Dabigatran

	Warfarin	Dabigatran
Abbreviations: CYP‐450, cytochrome P‐450. INR, international normalized ratio; P‐gp, P‐glycoprotein. Modified from reference 9.5, 6, 8, 16, 19, 2022, 24, 28, 30, 36, 4244
Mechanism of action	Reduces functional levels of vitamin Kdependent factors II, VII, IX, and X by inhibiting vitamin K epoxide reductase	Binds to active site of thrombin (factor IIa) and reversibly inhibits free and clot‐bound thrombin
Prodrug	No	Yes
Bioavailability	>90%95%	6%7%
Protein binding	99%	35%
Time to reach peak plasma levels	7296 hr	23 hr
Half‐life	3644 hr	1217 hr
Routine coagulation monitoring	Required, but frequency varies based on clinical situation	No requirement for routine monitoring
Schedule	INR‐adjusted, usually once daily	Fixed dose, once or twice daily
Metabolism	CYP‐450 hepatic microsomal enzymes, especially CYP2C9, CYP1A2, and CYP3A4	Esterase‐catalyzed hydrolysis in plasma or liver after intestinal P‐gp transport
Clearance	Almost entirely hepatic	80% unchanged renally (after an intravenous dose), 20% hepatic after conjugation
Drug interactions	Drugs that affect CYP‐450 hepatic microsomal enzymes and those that displace warfarin from plasma proteins	P‐gp inhibitors (CYP‐450 system not involved)
Antidote	Yes (vitamin K and plasma products)	No

Dabigatran is packaged in capsules that are hygroscopic. Therefore, the capsules should be stored in the original container with the cap tightly closed. Exposure of dabigatran capsules to air for prolonged periods outside the original container can result in deterioration of the active compound and reduced efficacy.27, 28 Dabigatran capsules contain tartaric acid which is necessary to facilitate dissolution of the medication in the gastrointestinal tract for optimal absorption.2 Breaking the capsules or removing the drug from the capsule can result in increased exposure. Therefore, dabigatran capsules should be taken intact, and patients should be instructed that dabigatran capsules should not be broken, chewed, or opened before administration.28 Alternative anticoagulants should be used if patients cannot swallow the capsule intact for any reason (eg, intubated patients).

EFFICACY OF DABIGATRAN

In this section, we provide a brief review of the major phase 3 trials that evaluated dabigatran for different indications (see references 13, 16, and 24 for recent detailed reviews of the clinical trials of dabigatran).

SAFETY OF DABIGATRAN

Aside from the bleeding risks discussed earlier, the most commonly reported side effect of dabigatran was dyspepsia. Dyspepsia occurred twice as frequently in patients taking dabigatran versus warfarin in the RE‐LY trial (11.5% vs 5.8%).35 One possible explanation for the higher incidence of dyspepsia is the tartaric acid component in dabigatran capsules.2 In the RE‐LY study, myocardial infarction occurred more commonly in the dabigatran arms (0.72% with 110 mg and 0.74% with 150 mg) than the warfarin arm (0.53%, P = 0.07 and 0.048, respectively).24, 35 It has been postulated that this observation could be related to a greater efficacy of warfarin for the prevention of myocardial infarction rather than an adverse effect of dabigatran.2 There was no increase in acute coronary syndrome rates noted with dabigatran in the other phase 3 trials.3234, 38, 40 No increased risk of elevated liver function test has been noted with dabigatran, but long‐term data are unavailable.32, 34, 35, 38

MANAGEMENT OF SPECIAL SITUATIONS THAT MAY ARISE IN THE USE OF DABIGATRAN

CONCLUSIONS

Dabigatran is a novel, oral direct thrombin inhibitor that exhibits several advantages over warfarin. The predictable pharmacokinetic profile and minimal food and drug interactions of dabigatran allow for a fixed‐dosing regimen and obviate the need for routine laboratory monitoring. However, this apparent advantage is also a disadvantage. The lack of a reliable method to monitor dabigatran makes it more difficult to assess compliance, measure the impact of drug interactions, evaluate for toxicity, and determine bona fide therapeutic failure versus noncompliance in the event of breakthrough thromboembolism.28, 42 Other limitations of dabigatran include the lack of an antidote and the dependence on normal renal function for elimination, with the potential for drug accumulation and toxicity with renal impairment. The noninferiority design of the clinical trials that evaluated dabigatran, the absence of long‐term safety and efficacy data, and issues related to the cost effectiveness of dabigatran should be considered when prescribing this agent. More studies are needed to assess dabigatran in special patient populations (eg, the elderly, patients with renal and hepatic impairment, pediatric and pregnant patients) and to better understand dabigatrandrug interactions.

As more novel oral anticoagulant agents, such as factor Xa inhibitors, become available for clinical use, comparative studies will need to be performed to better define the role of each agent for specific indications. In the future, it might be possible to tailor the choice of the oral anticoagulant to the individual patient not only on the basis of the clinical indication but also the specific patient characteristics and possible drug interactions. For example, rivaroxaban (Xarelto) is an oral direct factor Xa that was recently approved in the United States for VTE thromboprophylaxis following orthopedic surgery and in patients with non‐valvular atrial fibrillation.2 Similar to dabigatran, rivaroxaban exhibits predictable PK and PD that allow fixed once or twice daily dosing and obviate the need for routine monitoring of its anticoagulant effects.2, 16 Unlike dabigatran, rivaroxaban is an active drug and not a prodrug, and has a significantly higher bioavailability than dabigatran (>80% vs 6%).16 In addition, the levels of rivaroxaban can be affected by drugs that interfere with both P‐gp and the hepatic CYP‐450 system, compared with dabigatran, which is affected only by drugs that affect P‐gp.8, 16

The patient with a markedly high serum triglyceride (TG) level poses an interesting challenge for hospitalists. Hypertriglyceridemia (HTG) is defined as a fasting plasma TG level that is above the 95^th percentile for age and sex.1 TG levels are commonly classified into categories according to Adult Treatment Panel III guidelines with desirable levels <150 mg/dL (1.7 mmol/L), borderline levels 150199, high levels 200499 mg/dL, and very high levels >500 mg/dL (5.6 mmol/L).2 A TG level exceeding an arbitrary threshold of >1000 mg/dL (11.3 mmol/L) is referred to as severe HTG. The Lipid Research Clinics Program Prevalence Study found that 1.79 per 10,000 outpatients (<0.02%) had TG levels > 2000 mg/dL.3 Chylomicronemia syndrome occurs when severe HTG is accompanied by 1 or more of the following: symptoms of abdominal pain or acute pancreatitis or physical examination findings such as eruptive xanthomas or lipemia retinalis. There is no TG level above which pancreatitis invariably occurs, making the decision to hospitalize difficult. The goal of this review is to discuss the causes of severe HTG; the clinical assessment, including criteria for hospitalization; and the available treatment options for this infrequent but serious condition. We begin with a clinical case of severe HTG.

CASE PRESENTATION

A 47‐year‐old woman with a history of chronic myelogenous leukemia was admitted to the hospital with a serum triglyceride level of 17,393 mg/dL. Two years prior to admission, she underwent allogenic stem cell transplantation for chronic myelogenous leukemia and has since remained in remission. Six months prior to admission, severe diarrhea from intestinal graft‐versus‐host disease required the use of total parenteral nutrition (TPN) and immunosuppressive therapy consisting of prednisone 20 mg/day, mycophenolate mofetil 250 mg thrice daily, and sirolimus 0.3 mg/day. During treatment with steroids, she developed diabetes mellitus requiring insulin, with a subsequent hemoglobin A1c level of 7.7% (normal, <7%). The serum TG level prior to transplantation was unknown but was 343 mg/dL prior to TPN initiation. One month prior to admission, the diarrhea resolved and TPN was stopped. The TG level was 7463 mg/dL 1 week prior to admission, and despite use of fenofibrate, it rose to 17,393 mg/dL. The patient denied abdominal pain, and did not have abdominal tenderness or eruptive xanthomas. She denied a family history of dyslipidemia or recent medication changes. Given the extreme TG elevation, the lack of response to outpatient treatment and the concern for developing acute pancreatitis, the patient was admitted to the hospital for inpatient TG‐lowering treatment.

Upon admission, serum lipase and amylase were within normal limits, but the blood glucose level was 243 mg/dL. Insulin infusion and oral fenofibrate 145 mg/day was started, and the patient was kept non per os (NPO). Six hours later, despite insulin infusion, the TG level rose to 26,250 mg/dL. Therapeutic plasma exchange (TPE) was performed on 2 consecutive days with a resultant decrease in TG level to 530 mg/dL. The patient was later discharged home on fenofibrate and omega‐3 ethyl esters, her same immunosuppressive and insulin regimen, and instructions for a very low‐fat diet. In the next 3 months, her serum TG level did not rise above 530 mg/dL. The cause of our patient's extreme TG elevation was likely a combination of genetic factors exacerbated by immunosuppressive and glucocorticoid therapy.

This case featured dramatic elevations in serum TG levels that the managing doctors believed merited a hospital admission. Management of patients with severe HTG first requires an understanding of TG metabolism.

ETIOLOGY

Serum TGs produced by the liver are carried by very low‐density lipoproteins (VLDLs), whereas TGs derived from dietary fat are carried by chylomicrons. Both chylomicrons and VLDLs are hydrolyzed by the same enzymelipoprotein lipase (LPL). TGs are hydrolyzed into fatty acids for uptake by muscle and adipose tissue, whereas remnants of VLDL and chylomicrons are removed by the liver. More details on TG pathophysiology may be found in a recent review.4 When LPL is saturated with VLDL, ingestion of a fatty meal may cause chylomicrons to linger in circulation for days instead of hours. Asking the laboratory to spin down the blood of a patient with severe HTG and keep the test tube upright at 4C may reveal a large creamy supernatant layer demonstrating chylomicronemia.

A fasting TG level drawn 12 hours after the last meal reflects hepatic TG production. Although a nonfasting TG level may reflect postprandial chylomicrons, values above 1000 mg/dL strongly suggest true HTG, particularly in the setting of acute pancreatitis. Treatment should not be delayed to obtain a fasting TG level.

HTG may result from increased VLDL production, reduced VLDL/chylomicron clearance, or more likely a combination of the two. The causes of these metabolic derangements are classified as primary (genetic) or secondary (acquired) (Table 1). In adult patients, HTG is usually the result of a combination of primary and secondary causes. A study of 123 patients with TG levels >2000 mg/dL found that all patients had a primary metabolic defect and 110/123 had a coexistent secondary cause.3 An underlying genetic lipoprotein metabolism derangement is often clinically silent until coupled with a secondary cause of HTG that together raise TG levels high enough to cause the chylomicronemia syndrome.

The most common primary cause of HTG in adults is familial HTG, an autosomal dominant condition with a population prevalence ranging from 1%2% to 5%10% and age‐dependent penetrance.5, 6 Other genetic causes are much rarer, such as LPL deficiency (1 in 1 million patients), apolipoprotein‐CII, and other mutations resulting in impaired binding to LPL.79 Primary causes of HTG are often listed as Fredrickson phenotypes (Table 1). Recent genome‐wide association studies reveal a complex polygenic basis to the Fredrickson categories and suggest additional undefined genes or nongenetic factors may significantly contribute to the final phenotype.10 Diagnosis of familial lipid disorders requires an accurate family history that may be difficult to obtain.

Secondary causes of HTG can be categorized using a four‐D mnemonic: Diseases, Diet, Disorder of Metabolism, and Drugs.11 The most common condition associated with HTG is obesity.6 The mechanism between obesity and HTG is complex and likely involves an increase in fatty acid flux from adipose tissue to other tissues and insulin resistance.12 Asking whether a patient's current weight is close to the heaviest lifetime weight is a clue to diagnosing obesity‐driven HTG. One case series of hypertriglyceridemic pancreatitis found that diabetes or excessive alcohol intake account for the majority of secondary causes of HTG.13 The cause of HTG among patients with diabetes is multifactorial: insulin deficiency reduces LPL levels (insulin is required for synthesis of LPL), whereas insulin resistance attenuates the ability of insulin to decrease hepatic cholesterol synthesis and thus increases hepatic secretion of VLDL.14 Alcohol impairs lipolysis and increases VLDL production that can lead to severe HTG, particularly in those patients with an underlying functional deficiency in LPL. Other secondary etiologies of severe HTG may be elicited through careful attention to medical and medication history.

CLINICAL ASSESSMENT

Table 2 proposes a reasonable initial assessment of the history and physical and laboratory tests in patients with severe HTG.

Distinguishing physical examination findings may arise when serum TG levels exceed 1000 mg/dL. Eruptive xanthoma form when large amounts of TG are sequestered in cutaneous histiocytes, resulting in small yellow‐orange papules with an erythematous base. This finding is seen in a minority of HTG patients and may be missed altogether without a careful examination of the extensor surfaces of arms, legs, back, and buttocks.15 Effective TG‐lowering treatment will result in resolution of these xanthomas. An ophthalmologic examination may reveal lipemia retinalis, a condition that occurs when retinal vessels appear white from lipemic serum and contrast against a pale salmon‐colored retina. Although a dramatic finding, these changes do not result in vision impairment. Hepatomegaly from fatty infiltration of the liver occurs frequently,16 and diffuse lymphadenopathy may also be found.17

Laboratory tests are also essential in the clinical assessment of severe HTG. Important tests include thyroid‐stimulating hormone, creatinine, serum urea nitrogen, and a urinalysis. A hemoglobin A1c test provides information on the level of glycemic control and is now recognized by the American Diabetes Association to diagnose diabetes mellitus.18 Liver function tests commonly reveal a transaminase elevation from underlying steatohepatitis and also provide a baseline value prior to initiating any lipid‐lowering medications. Additional diagnostic tests may be useful in selected patients: HIV testing, serum protein electrophoresis, and urine protein electrophoresis to help diagnose paraproteinemias such as multiple myeloma, and an antinuclear antibody and double‐stranded DNA for systemic lupus erythematosus.

Severe HTG may interfere with the result of 2 commonly obtained laboratory tests. The sodium concentration can be falsely low (pseudohyponatremia) due to the high levels of TG displacing sodium containing water from the plasma.19 Due to interference by plasma lipids, amylase levels may be near normal in up to 50% of patients with hypertriglyceridemic pancreatitis at the time of admission.20 Thus, if the suspicion for pancreatitis is high, it is reasonable to proceed to imaging if amylase or lipase levels are not confirmatory. Abdominal imaging with computed tomography or magnetic resonance imaging may be used to diagnose acute pancreatitis.

Behind excessive alcohol consumption and gallstone disease, HTG is the third leading cause of pancreatitis, accounting for up to 10% of cases in the general population.21 The exact mechanism by which HTG causes pancreatitis is unclear. One theory is that elevated plasma TG levels are hydrolyzed in the pancreas to cause an increase in local free fatty acids, which in turn may cause inflammation and overt pancreatitis.22 Another theory proposes that elevated levels of chylomicrons lead to plasma hyperviscosity, which causes ischemia and local acidosis in pancreatic capillaries.23 Whatever the cause, it is unclear why only some patients with severe HTG develop acute pancreatitis. One study of 129 patients with severe HTG found mean serum TG levels to be higher in patients with acute pancreatitis than in those without (4470 versus 2450 mg/dL), suggesting the threshold to develop acute pancreatitis is higher than previously thought.24 Without a firm TG threshold above which patients develop pancreatitis, the decision to hospitalize can be difficult.

WHEN TO HOSPITALIZE?

The choice of whether to hospitalize a patient with severe HTG is first based on the presence or absence of abdominal pain and/or acute pancreatitis. Figure 1 diagrams a suggested admission and treatment algorithm. If abdominal pain is present, the patient should be hospitalized and assessed for possible triggers with prompt initiation of pharmacologic treatment. In the absence of abdominal pain, the decision to admit the patient with severe HTG requires clinical judgment. In these cases, prompt consultation with a physician experienced in the management of lipid disorders is recommended. In our experience, admission is usually driven by factors such as (1) severe hyperglycemia requiring inpatient insulin therapy; (2) severe HTG at or near a level where pancreatitis has occurred in the past in a patient for whom adherence is suspect (mindful of the great variability at the levels where patients develop pancreatitis); (3) unremitting triggers of severe HTG such as ongoing use of essential medications also known to exacerbate HTG (such as some forms of chemotherapy) or pregnancy in the third trimester. TG levels rise continuously throughout pregnancy and peak during the third trimester, when hypertriglyceridemic pancreatitis most often occurs. Asymptomatic patients with severe HTG not requiring hospitalization need close outpatient follow‐up to prevent the onset of chylomicronemia syndrome.

Figure 1

INPATIENT MANAGEMENT

No professional recommendations exist regarding a standardized treatment plan for severe HTG. The treatment regimen is first based on the presence or absence of symptoms. Treatment of hypertriglyceridemic pancreatitis should target a serum TG level <1000 mg/dL and resolution of abdominal pain. The initial goal for asymptomatic patients is a TG level <1000 mg/dL, as this level represents a significant reduction in the risk of developing chylomicronemia syndrome. In either case, the first 2 components of the treatment regimen are dietary changes and oral medications.

CONCLUSIONS

As the prevalence of obesity and diabetes continues to rise, so too does the clinical importance of proper management of severe HTG. Recognizing chylomicronemia syndromeone of the most dramatic consequences of lipid disordersand the underlying primary and secondary causes of HTG is required before starting treatment. Patients with severe HTG may require hospitalization for immediate reduction in TG levels and relief of abdominal pain, if present. Treatment involves modifying secondary causes, if possible, and eliminating dietary fat intake. Although use of medications such as an oral fibrate, omega‐3 fatty acids, and insulin are routine, the use of a more invasive procedure such as TPE should be considered on a case‐by‐case basis and may be limited by availability. Upon hospital discharge, careful follow‐up should promote lifestyle changes and medication adherence to prevent recurrence of severe HTG.

The patient with a markedly high serum triglyceride (TG) level poses an interesting challenge for hospitalists. Hypertriglyceridemia (HTG) is defined as a fasting plasma TG level that is above the 95^th percentile for age and sex.1 TG levels are commonly classified into categories according to Adult Treatment Panel III guidelines with desirable levels <150 mg/dL (1.7 mmol/L), borderline levels 150199, high levels 200499 mg/dL, and very high levels >500 mg/dL (5.6 mmol/L).2 A TG level exceeding an arbitrary threshold of >1000 mg/dL (11.3 mmol/L) is referred to as severe HTG. The Lipid Research Clinics Program Prevalence Study found that 1.79 per 10,000 outpatients (<0.02%) had TG levels > 2000 mg/dL.3 Chylomicronemia syndrome occurs when severe HTG is accompanied by 1 or more of the following: symptoms of abdominal pain or acute pancreatitis or physical examination findings such as eruptive xanthomas or lipemia retinalis. There is no TG level above which pancreatitis invariably occurs, making the decision to hospitalize difficult. The goal of this review is to discuss the causes of severe HTG; the clinical assessment, including criteria for hospitalization; and the available treatment options for this infrequent but serious condition. We begin with a clinical case of severe HTG.

CASE PRESENTATION

A 47‐year‐old woman with a history of chronic myelogenous leukemia was admitted to the hospital with a serum triglyceride level of 17,393 mg/dL. Two years prior to admission, she underwent allogenic stem cell transplantation for chronic myelogenous leukemia and has since remained in remission. Six months prior to admission, severe diarrhea from intestinal graft‐versus‐host disease required the use of total parenteral nutrition (TPN) and immunosuppressive therapy consisting of prednisone 20 mg/day, mycophenolate mofetil 250 mg thrice daily, and sirolimus 0.3 mg/day. During treatment with steroids, she developed diabetes mellitus requiring insulin, with a subsequent hemoglobin A1c level of 7.7% (normal, <7%). The serum TG level prior to transplantation was unknown but was 343 mg/dL prior to TPN initiation. One month prior to admission, the diarrhea resolved and TPN was stopped. The TG level was 7463 mg/dL 1 week prior to admission, and despite use of fenofibrate, it rose to 17,393 mg/dL. The patient denied abdominal pain, and did not have abdominal tenderness or eruptive xanthomas. She denied a family history of dyslipidemia or recent medication changes. Given the extreme TG elevation, the lack of response to outpatient treatment and the concern for developing acute pancreatitis, the patient was admitted to the hospital for inpatient TG‐lowering treatment.

Upon admission, serum lipase and amylase were within normal limits, but the blood glucose level was 243 mg/dL. Insulin infusion and oral fenofibrate 145 mg/day was started, and the patient was kept non per os (NPO). Six hours later, despite insulin infusion, the TG level rose to 26,250 mg/dL. Therapeutic plasma exchange (TPE) was performed on 2 consecutive days with a resultant decrease in TG level to 530 mg/dL. The patient was later discharged home on fenofibrate and omega‐3 ethyl esters, her same immunosuppressive and insulin regimen, and instructions for a very low‐fat diet. In the next 3 months, her serum TG level did not rise above 530 mg/dL. The cause of our patient's extreme TG elevation was likely a combination of genetic factors exacerbated by immunosuppressive and glucocorticoid therapy.

This case featured dramatic elevations in serum TG levels that the managing doctors believed merited a hospital admission. Management of patients with severe HTG first requires an understanding of TG metabolism.

ETIOLOGY

Serum TGs produced by the liver are carried by very low‐density lipoproteins (VLDLs), whereas TGs derived from dietary fat are carried by chylomicrons. Both chylomicrons and VLDLs are hydrolyzed by the same enzymelipoprotein lipase (LPL). TGs are hydrolyzed into fatty acids for uptake by muscle and adipose tissue, whereas remnants of VLDL and chylomicrons are removed by the liver. More details on TG pathophysiology may be found in a recent review.4 When LPL is saturated with VLDL, ingestion of a fatty meal may cause chylomicrons to linger in circulation for days instead of hours. Asking the laboratory to spin down the blood of a patient with severe HTG and keep the test tube upright at 4C may reveal a large creamy supernatant layer demonstrating chylomicronemia.

A fasting TG level drawn 12 hours after the last meal reflects hepatic TG production. Although a nonfasting TG level may reflect postprandial chylomicrons, values above 1000 mg/dL strongly suggest true HTG, particularly in the setting of acute pancreatitis. Treatment should not be delayed to obtain a fasting TG level.

HTG may result from increased VLDL production, reduced VLDL/chylomicron clearance, or more likely a combination of the two. The causes of these metabolic derangements are classified as primary (genetic) or secondary (acquired) (Table 1). In adult patients, HTG is usually the result of a combination of primary and secondary causes. A study of 123 patients with TG levels >2000 mg/dL found that all patients had a primary metabolic defect and 110/123 had a coexistent secondary cause.3 An underlying genetic lipoprotein metabolism derangement is often clinically silent until coupled with a secondary cause of HTG that together raise TG levels high enough to cause the chylomicronemia syndrome.

The most common primary cause of HTG in adults is familial HTG, an autosomal dominant condition with a population prevalence ranging from 1%2% to 5%10% and age‐dependent penetrance.5, 6 Other genetic causes are much rarer, such as LPL deficiency (1 in 1 million patients), apolipoprotein‐CII, and other mutations resulting in impaired binding to LPL.79 Primary causes of HTG are often listed as Fredrickson phenotypes (Table 1). Recent genome‐wide association studies reveal a complex polygenic basis to the Fredrickson categories and suggest additional undefined genes or nongenetic factors may significantly contribute to the final phenotype.10 Diagnosis of familial lipid disorders requires an accurate family history that may be difficult to obtain.

Secondary causes of HTG can be categorized using a four‐D mnemonic: Diseases, Diet, Disorder of Metabolism, and Drugs.11 The most common condition associated with HTG is obesity.6 The mechanism between obesity and HTG is complex and likely involves an increase in fatty acid flux from adipose tissue to other tissues and insulin resistance.12 Asking whether a patient's current weight is close to the heaviest lifetime weight is a clue to diagnosing obesity‐driven HTG. One case series of hypertriglyceridemic pancreatitis found that diabetes or excessive alcohol intake account for the majority of secondary causes of HTG.13 The cause of HTG among patients with diabetes is multifactorial: insulin deficiency reduces LPL levels (insulin is required for synthesis of LPL), whereas insulin resistance attenuates the ability of insulin to decrease hepatic cholesterol synthesis and thus increases hepatic secretion of VLDL.14 Alcohol impairs lipolysis and increases VLDL production that can lead to severe HTG, particularly in those patients with an underlying functional deficiency in LPL. Other secondary etiologies of severe HTG may be elicited through careful attention to medical and medication history.

CLINICAL ASSESSMENT

Table 2 proposes a reasonable initial assessment of the history and physical and laboratory tests in patients with severe HTG.

Distinguishing physical examination findings may arise when serum TG levels exceed 1000 mg/dL. Eruptive xanthoma form when large amounts of TG are sequestered in cutaneous histiocytes, resulting in small yellow‐orange papules with an erythematous base. This finding is seen in a minority of HTG patients and may be missed altogether without a careful examination of the extensor surfaces of arms, legs, back, and buttocks.15 Effective TG‐lowering treatment will result in resolution of these xanthomas. An ophthalmologic examination may reveal lipemia retinalis, a condition that occurs when retinal vessels appear white from lipemic serum and contrast against a pale salmon‐colored retina. Although a dramatic finding, these changes do not result in vision impairment. Hepatomegaly from fatty infiltration of the liver occurs frequently,16 and diffuse lymphadenopathy may also be found.17

Laboratory tests are also essential in the clinical assessment of severe HTG. Important tests include thyroid‐stimulating hormone, creatinine, serum urea nitrogen, and a urinalysis. A hemoglobin A1c test provides information on the level of glycemic control and is now recognized by the American Diabetes Association to diagnose diabetes mellitus.18 Liver function tests commonly reveal a transaminase elevation from underlying steatohepatitis and also provide a baseline value prior to initiating any lipid‐lowering medications. Additional diagnostic tests may be useful in selected patients: HIV testing, serum protein electrophoresis, and urine protein electrophoresis to help diagnose paraproteinemias such as multiple myeloma, and an antinuclear antibody and double‐stranded DNA for systemic lupus erythematosus.

Severe HTG may interfere with the result of 2 commonly obtained laboratory tests. The sodium concentration can be falsely low (pseudohyponatremia) due to the high levels of TG displacing sodium containing water from the plasma.19 Due to interference by plasma lipids, amylase levels may be near normal in up to 50% of patients with hypertriglyceridemic pancreatitis at the time of admission.20 Thus, if the suspicion for pancreatitis is high, it is reasonable to proceed to imaging if amylase or lipase levels are not confirmatory. Abdominal imaging with computed tomography or magnetic resonance imaging may be used to diagnose acute pancreatitis.

Behind excessive alcohol consumption and gallstone disease, HTG is the third leading cause of pancreatitis, accounting for up to 10% of cases in the general population.21 The exact mechanism by which HTG causes pancreatitis is unclear. One theory is that elevated plasma TG levels are hydrolyzed in the pancreas to cause an increase in local free fatty acids, which in turn may cause inflammation and overt pancreatitis.22 Another theory proposes that elevated levels of chylomicrons lead to plasma hyperviscosity, which causes ischemia and local acidosis in pancreatic capillaries.23 Whatever the cause, it is unclear why only some patients with severe HTG develop acute pancreatitis. One study of 129 patients with severe HTG found mean serum TG levels to be higher in patients with acute pancreatitis than in those without (4470 versus 2450 mg/dL), suggesting the threshold to develop acute pancreatitis is higher than previously thought.24 Without a firm TG threshold above which patients develop pancreatitis, the decision to hospitalize can be difficult.

WHEN TO HOSPITALIZE?

The choice of whether to hospitalize a patient with severe HTG is first based on the presence or absence of abdominal pain and/or acute pancreatitis. Figure 1 diagrams a suggested admission and treatment algorithm. If abdominal pain is present, the patient should be hospitalized and assessed for possible triggers with prompt initiation of pharmacologic treatment. In the absence of abdominal pain, the decision to admit the patient with severe HTG requires clinical judgment. In these cases, prompt consultation with a physician experienced in the management of lipid disorders is recommended. In our experience, admission is usually driven by factors such as (1) severe hyperglycemia requiring inpatient insulin therapy; (2) severe HTG at or near a level where pancreatitis has occurred in the past in a patient for whom adherence is suspect (mindful of the great variability at the levels where patients develop pancreatitis); (3) unremitting triggers of severe HTG such as ongoing use of essential medications also known to exacerbate HTG (such as some forms of chemotherapy) or pregnancy in the third trimester. TG levels rise continuously throughout pregnancy and peak during the third trimester, when hypertriglyceridemic pancreatitis most often occurs. Asymptomatic patients with severe HTG not requiring hospitalization need close outpatient follow‐up to prevent the onset of chylomicronemia syndrome.

Figure 1

INPATIENT MANAGEMENT

No professional recommendations exist regarding a standardized treatment plan for severe HTG. The treatment regimen is first based on the presence or absence of symptoms. Treatment of hypertriglyceridemic pancreatitis should target a serum TG level <1000 mg/dL and resolution of abdominal pain. The initial goal for asymptomatic patients is a TG level <1000 mg/dL, as this level represents a significant reduction in the risk of developing chylomicronemia syndrome. In either case, the first 2 components of the treatment regimen are dietary changes and oral medications.

CONCLUSIONS

As the prevalence of obesity and diabetes continues to rise, so too does the clinical importance of proper management of severe HTG. Recognizing chylomicronemia syndromeone of the most dramatic consequences of lipid disordersand the underlying primary and secondary causes of HTG is required before starting treatment. Patients with severe HTG may require hospitalization for immediate reduction in TG levels and relief of abdominal pain, if present. Treatment involves modifying secondary causes, if possible, and eliminating dietary fat intake. Although use of medications such as an oral fibrate, omega‐3 fatty acids, and insulin are routine, the use of a more invasive procedure such as TPE should be considered on a case‐by‐case basis and may be limited by availability. Upon hospital discharge, careful follow‐up should promote lifestyle changes and medication adherence to prevent recurrence of severe HTG.

The patient with a markedly high serum triglyceride (TG) level poses an interesting challenge for hospitalists. Hypertriglyceridemia (HTG) is defined as a fasting plasma TG level that is above the 95^th percentile for age and sex.1 TG levels are commonly classified into categories according to Adult Treatment Panel III guidelines with desirable levels <150 mg/dL (1.7 mmol/L), borderline levels 150199, high levels 200499 mg/dL, and very high levels >500 mg/dL (5.6 mmol/L).2 A TG level exceeding an arbitrary threshold of >1000 mg/dL (11.3 mmol/L) is referred to as severe HTG. The Lipid Research Clinics Program Prevalence Study found that 1.79 per 10,000 outpatients (<0.02%) had TG levels > 2000 mg/dL.3 Chylomicronemia syndrome occurs when severe HTG is accompanied by 1 or more of the following: symptoms of abdominal pain or acute pancreatitis or physical examination findings such as eruptive xanthomas or lipemia retinalis. There is no TG level above which pancreatitis invariably occurs, making the decision to hospitalize difficult. The goal of this review is to discuss the causes of severe HTG; the clinical assessment, including criteria for hospitalization; and the available treatment options for this infrequent but serious condition. We begin with a clinical case of severe HTG.

CASE PRESENTATION

A 47‐year‐old woman with a history of chronic myelogenous leukemia was admitted to the hospital with a serum triglyceride level of 17,393 mg/dL. Two years prior to admission, she underwent allogenic stem cell transplantation for chronic myelogenous leukemia and has since remained in remission. Six months prior to admission, severe diarrhea from intestinal graft‐versus‐host disease required the use of total parenteral nutrition (TPN) and immunosuppressive therapy consisting of prednisone 20 mg/day, mycophenolate mofetil 250 mg thrice daily, and sirolimus 0.3 mg/day. During treatment with steroids, she developed diabetes mellitus requiring insulin, with a subsequent hemoglobin A1c level of 7.7% (normal, <7%). The serum TG level prior to transplantation was unknown but was 343 mg/dL prior to TPN initiation. One month prior to admission, the diarrhea resolved and TPN was stopped. The TG level was 7463 mg/dL 1 week prior to admission, and despite use of fenofibrate, it rose to 17,393 mg/dL. The patient denied abdominal pain, and did not have abdominal tenderness or eruptive xanthomas. She denied a family history of dyslipidemia or recent medication changes. Given the extreme TG elevation, the lack of response to outpatient treatment and the concern for developing acute pancreatitis, the patient was admitted to the hospital for inpatient TG‐lowering treatment.

Upon admission, serum lipase and amylase were within normal limits, but the blood glucose level was 243 mg/dL. Insulin infusion and oral fenofibrate 145 mg/day was started, and the patient was kept non per os (NPO). Six hours later, despite insulin infusion, the TG level rose to 26,250 mg/dL. Therapeutic plasma exchange (TPE) was performed on 2 consecutive days with a resultant decrease in TG level to 530 mg/dL. The patient was later discharged home on fenofibrate and omega‐3 ethyl esters, her same immunosuppressive and insulin regimen, and instructions for a very low‐fat diet. In the next 3 months, her serum TG level did not rise above 530 mg/dL. The cause of our patient's extreme TG elevation was likely a combination of genetic factors exacerbated by immunosuppressive and glucocorticoid therapy.

This case featured dramatic elevations in serum TG levels that the managing doctors believed merited a hospital admission. Management of patients with severe HTG first requires an understanding of TG metabolism.

ETIOLOGY

Serum TGs produced by the liver are carried by very low‐density lipoproteins (VLDLs), whereas TGs derived from dietary fat are carried by chylomicrons. Both chylomicrons and VLDLs are hydrolyzed by the same enzymelipoprotein lipase (LPL). TGs are hydrolyzed into fatty acids for uptake by muscle and adipose tissue, whereas remnants of VLDL and chylomicrons are removed by the liver. More details on TG pathophysiology may be found in a recent review.4 When LPL is saturated with VLDL, ingestion of a fatty meal may cause chylomicrons to linger in circulation for days instead of hours. Asking the laboratory to spin down the blood of a patient with severe HTG and keep the test tube upright at 4C may reveal a large creamy supernatant layer demonstrating chylomicronemia.

A fasting TG level drawn 12 hours after the last meal reflects hepatic TG production. Although a nonfasting TG level may reflect postprandial chylomicrons, values above 1000 mg/dL strongly suggest true HTG, particularly in the setting of acute pancreatitis. Treatment should not be delayed to obtain a fasting TG level.

HTG may result from increased VLDL production, reduced VLDL/chylomicron clearance, or more likely a combination of the two. The causes of these metabolic derangements are classified as primary (genetic) or secondary (acquired) (Table 1). In adult patients, HTG is usually the result of a combination of primary and secondary causes. A study of 123 patients with TG levels >2000 mg/dL found that all patients had a primary metabolic defect and 110/123 had a coexistent secondary cause.3 An underlying genetic lipoprotein metabolism derangement is often clinically silent until coupled with a secondary cause of HTG that together raise TG levels high enough to cause the chylomicronemia syndrome.

The most common primary cause of HTG in adults is familial HTG, an autosomal dominant condition with a population prevalence ranging from 1%2% to 5%10% and age‐dependent penetrance.5, 6 Other genetic causes are much rarer, such as LPL deficiency (1 in 1 million patients), apolipoprotein‐CII, and other mutations resulting in impaired binding to LPL.79 Primary causes of HTG are often listed as Fredrickson phenotypes (Table 1). Recent genome‐wide association studies reveal a complex polygenic basis to the Fredrickson categories and suggest additional undefined genes or nongenetic factors may significantly contribute to the final phenotype.10 Diagnosis of familial lipid disorders requires an accurate family history that may be difficult to obtain.

Secondary causes of HTG can be categorized using a four‐D mnemonic: Diseases, Diet, Disorder of Metabolism, and Drugs.11 The most common condition associated with HTG is obesity.6 The mechanism between obesity and HTG is complex and likely involves an increase in fatty acid flux from adipose tissue to other tissues and insulin resistance.12 Asking whether a patient's current weight is close to the heaviest lifetime weight is a clue to diagnosing obesity‐driven HTG. One case series of hypertriglyceridemic pancreatitis found that diabetes or excessive alcohol intake account for the majority of secondary causes of HTG.13 The cause of HTG among patients with diabetes is multifactorial: insulin deficiency reduces LPL levels (insulin is required for synthesis of LPL), whereas insulin resistance attenuates the ability of insulin to decrease hepatic cholesterol synthesis and thus increases hepatic secretion of VLDL.14 Alcohol impairs lipolysis and increases VLDL production that can lead to severe HTG, particularly in those patients with an underlying functional deficiency in LPL. Other secondary etiologies of severe HTG may be elicited through careful attention to medical and medication history.

CLINICAL ASSESSMENT

Table 2 proposes a reasonable initial assessment of the history and physical and laboratory tests in patients with severe HTG.

Distinguishing physical examination findings may arise when serum TG levels exceed 1000 mg/dL. Eruptive xanthoma form when large amounts of TG are sequestered in cutaneous histiocytes, resulting in small yellow‐orange papules with an erythematous base. This finding is seen in a minority of HTG patients and may be missed altogether without a careful examination of the extensor surfaces of arms, legs, back, and buttocks.15 Effective TG‐lowering treatment will result in resolution of these xanthomas. An ophthalmologic examination may reveal lipemia retinalis, a condition that occurs when retinal vessels appear white from lipemic serum and contrast against a pale salmon‐colored retina. Although a dramatic finding, these changes do not result in vision impairment. Hepatomegaly from fatty infiltration of the liver occurs frequently,16 and diffuse lymphadenopathy may also be found.17

Laboratory tests are also essential in the clinical assessment of severe HTG. Important tests include thyroid‐stimulating hormone, creatinine, serum urea nitrogen, and a urinalysis. A hemoglobin A1c test provides information on the level of glycemic control and is now recognized by the American Diabetes Association to diagnose diabetes mellitus.18 Liver function tests commonly reveal a transaminase elevation from underlying steatohepatitis and also provide a baseline value prior to initiating any lipid‐lowering medications. Additional diagnostic tests may be useful in selected patients: HIV testing, serum protein electrophoresis, and urine protein electrophoresis to help diagnose paraproteinemias such as multiple myeloma, and an antinuclear antibody and double‐stranded DNA for systemic lupus erythematosus.

Severe HTG may interfere with the result of 2 commonly obtained laboratory tests. The sodium concentration can be falsely low (pseudohyponatremia) due to the high levels of TG displacing sodium containing water from the plasma.19 Due to interference by plasma lipids, amylase levels may be near normal in up to 50% of patients with hypertriglyceridemic pancreatitis at the time of admission.20 Thus, if the suspicion for pancreatitis is high, it is reasonable to proceed to imaging if amylase or lipase levels are not confirmatory. Abdominal imaging with computed tomography or magnetic resonance imaging may be used to diagnose acute pancreatitis.

Behind excessive alcohol consumption and gallstone disease, HTG is the third leading cause of pancreatitis, accounting for up to 10% of cases in the general population.21 The exact mechanism by which HTG causes pancreatitis is unclear. One theory is that elevated plasma TG levels are hydrolyzed in the pancreas to cause an increase in local free fatty acids, which in turn may cause inflammation and overt pancreatitis.22 Another theory proposes that elevated levels of chylomicrons lead to plasma hyperviscosity, which causes ischemia and local acidosis in pancreatic capillaries.23 Whatever the cause, it is unclear why only some patients with severe HTG develop acute pancreatitis. One study of 129 patients with severe HTG found mean serum TG levels to be higher in patients with acute pancreatitis than in those without (4470 versus 2450 mg/dL), suggesting the threshold to develop acute pancreatitis is higher than previously thought.24 Without a firm TG threshold above which patients develop pancreatitis, the decision to hospitalize can be difficult.

WHEN TO HOSPITALIZE?

The choice of whether to hospitalize a patient with severe HTG is first based on the presence or absence of abdominal pain and/or acute pancreatitis. Figure 1 diagrams a suggested admission and treatment algorithm. If abdominal pain is present, the patient should be hospitalized and assessed for possible triggers with prompt initiation of pharmacologic treatment. In the absence of abdominal pain, the decision to admit the patient with severe HTG requires clinical judgment. In these cases, prompt consultation with a physician experienced in the management of lipid disorders is recommended. In our experience, admission is usually driven by factors such as (1) severe hyperglycemia requiring inpatient insulin therapy; (2) severe HTG at or near a level where pancreatitis has occurred in the past in a patient for whom adherence is suspect (mindful of the great variability at the levels where patients develop pancreatitis); (3) unremitting triggers of severe HTG such as ongoing use of essential medications also known to exacerbate HTG (such as some forms of chemotherapy) or pregnancy in the third trimester. TG levels rise continuously throughout pregnancy and peak during the third trimester, when hypertriglyceridemic pancreatitis most often occurs. Asymptomatic patients with severe HTG not requiring hospitalization need close outpatient follow‐up to prevent the onset of chylomicronemia syndrome.

Figure 1

INPATIENT MANAGEMENT

No professional recommendations exist regarding a standardized treatment plan for severe HTG. The treatment regimen is first based on the presence or absence of symptoms. Treatment of hypertriglyceridemic pancreatitis should target a serum TG level <1000 mg/dL and resolution of abdominal pain. The initial goal for asymptomatic patients is a TG level <1000 mg/dL, as this level represents a significant reduction in the risk of developing chylomicronemia syndrome. In either case, the first 2 components of the treatment regimen are dietary changes and oral medications.

CONCLUSIONS

As the prevalence of obesity and diabetes continues to rise, so too does the clinical importance of proper management of severe HTG. Recognizing chylomicronemia syndromeone of the most dramatic consequences of lipid disordersand the underlying primary and secondary causes of HTG is required before starting treatment. Patients with severe HTG may require hospitalization for immediate reduction in TG levels and relief of abdominal pain, if present. Treatment involves modifying secondary causes, if possible, and eliminating dietary fat intake. Although use of medications such as an oral fibrate, omega‐3 fatty acids, and insulin are routine, the use of a more invasive procedure such as TPE should be considered on a case‐by‐case basis and may be limited by availability. Upon hospital discharge, careful follow‐up should promote lifestyle changes and medication adherence to prevent recurrence of severe HTG.

Every health professional bears responsibility at some point for helping dying patients and their survivors face death spiritually and emotionally. But, because most Americans die in hospitals, that responsibility falls disproportionately to hospitalists and other hospital‐based health professionals.

Personal beliefs about what happens at death surely influence whether patients welcome or dread it, and whether survivors remember it with relief or regret.1 We believe, therefore, that competent, compassionate end‐of‐life care requires hospitalists and other health professionals who attend dying patients to address such beliefs. But people do not readily volunteer them, health professionals rarely elicit them, and little research describes them.

We, therefore, performed a large exploratory study to begin characterizing patients' beliefs about what happens at the time of death. We assumed that culturethe values a group uses to interpret shared experiences and transmits across generations2, 3influences those beliefs.1, 410 We reasoned that, because death is a universal human experience, every culture must address its meaning.5 Prior studies showing ethnic cultural differences over advance care planning, life support, and other aspects of dying further supported our assumption.3, 8, 1012

Our interview study revealed occasional beliefs that may characterize Americans in general, some beliefs that may characterize only certain ethnic groups or genders, and many beliefs that may characterize only particular individuals.

METHODS

We constructed a semistructured interview, based on topics and questions from the end‐of‐life literature and our own encounters with dying patients and their survivors. The interview schedule covered topics such as the right time to die, what happens at death, and the afterlife. We pretested all questions before using them in interviews.

Study participants were older inpatients from the 3 largest American ethnic groupsMexican Americans (MAs), Euro‐Americans (EAs), and African Americans (AAs),13 as identified by a validated ethnic algorithm.14 We reasoned that age, current serious (though not necessarily terminal) illness, and having experienced the deaths of others had already prompted these older inpatients to think about death.11

Admission logs from 2 San Antonio, Texas, hospitals identified all patients, aged 50 to 79, who were admitted over a 9‐month period for any of 10 common internal medicine diagnoses. From these logs we selected interviewees by purposive sampling, a nonstatistical technique that ensured adequate participant numbers by ethnic group and gender.15, 16 We invited patients to interview only after their primary physicians gave permission.

Sixty of 65 participants who began interviews completed them, and 58 of the 60 could be classified into 1 of the 3 ethnic groups.14 These 58, who produced saturation for all themes mentioned by more than 5% of participants, constituted our analysis sample. Participants included 26 MAs (14 men, 12 women), 18 EAs (7 men, 11 women), and 14 AAs (7 men, 7 women). The most prevalent admitting diagnoses were congestive heart failure (19 participants), angina (17 participants), and pneumonia and chronic obstructive pulmonary disease (5 participants each). The 3 ethnic group samples had similar mean ages but differed in other ways (Table 1). MAs were typically Roman Catholic, educated through grade 7, and married; EAs were divided between Roman Catholic and Protestant, educated through grade 12, and mostly unmarried; and AAs were nearly all Protestant, educated through grade 11, and mostly unmarried. The genders within each ethnic group sample were similar by age, religion, and education (data not shown). AA men and women were also similar by marital status. However, MAs and EAs had more men than women who were married, and more women than men who were widowed.

Two trained, bilingual women1 MA and 1 EA, not specifically matched to participants by ethnic groupused the schedule of questions to interview participants. The interviews usually took place 3 days after admission, involved one‐on‐one engagement in participants' hospital rooms, were audiotaped, and lasted roughly 90 minutes. Most questions were open‐ended, allowing participants to express beliefs in their own words. For example, the open‐ended question, What do you think happens at the end of a person's life? introduced the topic covered here. To help focus responses, interviewers asked participants early on to name the closest person to them to have died. Interviewers then encouraged participants to describe their beliefs specifically in terms of that person's death. (We assumed the closeness of the relationships had kept those deaths vivid for participants even years afterward.) Most participants responded by referring to that person, but a few described their own death‐like experiences or their general beliefs about death. Interviewers probed as necessary to clarify responses.

Participants interviewed in Spanish or English as they preferred. Two MAs interviewed entirely in Spanish, 10 MAs interviewed partly in Spanish and partly in English, and all other participants interviewed entirely in English. Bilingual typists transcribed the audiotapes, translating any Spanish into English. Two bilingual experts independently confirmed the accuracy of the translations.

The coders who content‐analyzed responses varied by ethnic group, gender, and professional training. They conducted their analysis in 4 steps, each involving initial, independent, blinded reviews by 2 coders; comparison of interpretations; and consensual resolution of disagreements. First, 2 coders deleted any comments irrelevant to death or dying. Second, the same coders assembled the remaining passages by topic, such as beliefs about what happens at the time of death. Third, 1 original coder and a senior investigator naive to the responses identified themes within each topic. Fourth, that original coder and either of 2 new coders determined for each interview the presence or absence of each theme. Theme presence required agreement between the original coder and the new coder or, when they differed, agreement between one of these coders and an independent adjudicator. Lastly, the 3 authors aggregated themes into meaningful categories by consensus, and checked these categories for trustworthiness against participants' original comments.

We report the results for each theme primarily as the percentages of participants within these ethnic group or gender samples who mentioned the theme. Though content analyses are usually reported qualitatively, percentages have 2 advantages here.17 First, readers can see the percentages and judge for themselves the important similarities, differences, and patterns in the data. Second, percentages enable researchers to formulate their own questions for further study, say, questions based on the largest percentages or largest percentage differences in the data. We also report representative quotes to illustrate the depth and richness of participant responses.

The study complied with all institutional review board requirements.

RESULTS

Most participants in all 3 ethnic group samples named a parent as the closest person to them to have died (Table 2). Among these participants, MAs named their mothers overwhelmingly, but EAs and AAs named their mothers and fathers nearly equally. Other participants named siblings, children, other relatives, or friends. Of 13 widowed participants, only 4 named their spouses.

Thirty‐nine participants20 MAs, 13 EAs, and 6 AAsdescribed the closest person's time of death as either momentary (typically less than a minute) or prolonged (longer than a few minutes). More EAs described it as momentary than as prolonged (50% vs 22%). One EA woman said, We were right outside [the hospital room when my father suffered his cardiac arrest]. We knew, when the alarm went off on the heart monitor, it was the last time we'd see him alive. In contrast, MAs and AAs split roughly equally between describing death as momentary or as prolonged (MAs: 35% and 42%, respectively; AAs: 21% for both).

The ethnic group samples also differed about harm from treatments the closest person had received when dying. Of participants who specified such treatments, disproportionately more AAs (4 of 6) than MAs (0 of 5) or EAs (1 of 8) said those treatments had caused the person to suffer at the time of death. Recalling the prolonged resuscitation efforts on his father, one AA man said that the doctors were trying to keep him alive I said, Don't put his body through that.

Many participants went on to describe their beliefs about what happens at the time of death, about the physiologic signs that define that time, and about the senses that persist after death.

DISCUSSION

Beliefs about what happens at death surely affect the whole dying experience1 and may help guide end‐of‐life care. Yet for all the research on dying, the health professions literatures contain virtually no studies describing such beliefs.19 This exploratory study begins the descriptive process.

The results suggest a taxonomyhowever provisionalfor those beliefs (Table 6). Occasional beliefs, such as the one that death separates the dead from the living, may be held by many Americans and thereby characterize American society in general. Other beliefs may characterize only particular American ethnic groups or genders. Ethnically based beliefs may include, for MAs, the belief that death occurs when an external force, specifically God or Jesus, takes the dead person away; for EAs, the beliefs that death occurs in less than a minute and that physiologic signs define the time of death; and, for AAs, the belief that cooling of the body never defines the time of death. A gender‐based belief may be the belief of EA women that some senses persist after death. Still other beliefs may be idiosyncratic, varying among individuals without a demographic pattern. Idiosyncratic beliefs may include which particular physiologic sign defines the time of death.

The reader must consider these results in light of the study's assumptions, weaknesses, and strengths. Two assumptions were key: that participants expressed themselves fully (despite the difficulty of articulating such beliefs), and that the content analysts interpreted them accurately. Study weaknesses included possible conditioning of responses through prior interview questions, incomplete responsiveness about certain themes, nongeneralizability beyond these sample groups due to the purposive sampling, and educational and marital status differences as possible alternative explanations for the results. Important study strengths included the clinically important topic; the ill, older participants who had already faced death for themselves or others; the pretested, bilingual interview schedule; the open‐ended questions allowing participants to express beliefs in their own words; and the rigorous content analysis.

While indicating directions for future research, our results also provide several useful lessons for current end‐of‐life care. First, hospitalists and other health professionals who attend the dying must not assume they can accurately predict the beliefs of patients or survivors about what happens at the time of death. Many beliefs that participants expressed here neither we nor the health professionals to whom we have presented the results could have imagined beforehand. Health professionals simply cannot expect patients and survivors to hold the same beliefs as they. Furthermore, while some beliefs may characterize certain demographic groups in general, large idiosyncratic variation within groups compromises many demographically based predictions of particular individuals' beliefs. Thus, health professionals can probably learn such beliefs only by eliciting them individual by individual.18 Admittedly awkward, the necessary inquiries6, 20, 21 demand courage and patience, but, when done well, may help prepare all for the death.

Second, these inquiries require health professionals to cultivate techniques for eliciting people's beliefs sensitively and accurately.18 For example, health professionals might ask questions using the same terms for death (such as goes, passes, or rests) that the patient or survivors use. This technique encourages open expression of beliefs by assuring people that health professionals are listening carefully. Still, professionals must guard against misunderstandings that such terms may createparticularly by giving false hope or ignoring sad realities.

Third, armed with knowledge of patient and survivor beliefs, health professionals should tailor perimortem care accordingly. For example, because many people suffer pangs of separation at a death, health professionals should address those feelings expressly.22 Doing so may promote closure for the grieving.20, 2325 Health professionals should also attend to beliefs such as those about the signs and duration of the time of death. As this study showed, different people may time death by different physiologic signs. The patient with a warm body but no heartbeat may be simultaneously alive to someone who sees cooling of the body as defining the time of death, and dead to someone else who sees cessation of heartbeat as doing so.18, 26 Unfortunately, certain perimortem procedures, such as harvesting organs for transplantation, moving the body to the morgue, or performing an autopsy, require declaring death unambiguously at one specific time. When differences may exist over the signs or duration of death, the best way to avoid agonizing arguments or decisional paralysis when death occurs is to articulate any differences beforehand and to resolve them with a mutually acceptable management plan.27 Health professionals should also honor survivor beliefs about sentience of the dead. Unlike differences over signs and duration of death, differences over sentience of the dead may often be accommodated without definitive resolution. For example, although health professionals should treat any body respectfully, they must make special efforts to handle the dead body gently in case survivors believe it can still feel pain.

CONCLUSION

A patient's death is a climactic event for patient, survivors, and health professionals alike. It warrants careful management. Personal beliefs about what happens at death surely affect how patients anticipate it and what survivors remember of it. Compassionate perimortem care, therefore, must address those beliefs.28

Yet demographic characteristics such as ethnic group or gender offer only limited clues to such beliefs, making health professionals elicit them directly.4, 6, 9, 21 Discussions with patients and survivors about these beliefs are bound to create emotional discomfort for everybody, but health professionals may dispel much of it by demonstrating a sincere commitment to personalized end‐of‐life care, showing respect for the beliefs of patients and survivors,7, 21, 23 and accommodating those beliefs whenever possible.2, 28 The result may be the best of all possible outcomessensitive, respectful, and compassionate care for patients and survivors, and rewarding caregiving experiences for health professionals.20, 24, 25

Every health professional bears responsibility at some point for helping dying patients and their survivors face death spiritually and emotionally. But, because most Americans die in hospitals, that responsibility falls disproportionately to hospitalists and other hospital‐based health professionals.

Personal beliefs about what happens at death surely influence whether patients welcome or dread it, and whether survivors remember it with relief or regret.1 We believe, therefore, that competent, compassionate end‐of‐life care requires hospitalists and other health professionals who attend dying patients to address such beliefs. But people do not readily volunteer them, health professionals rarely elicit them, and little research describes them.

We, therefore, performed a large exploratory study to begin characterizing patients' beliefs about what happens at the time of death. We assumed that culturethe values a group uses to interpret shared experiences and transmits across generations2, 3influences those beliefs.1, 410 We reasoned that, because death is a universal human experience, every culture must address its meaning.5 Prior studies showing ethnic cultural differences over advance care planning, life support, and other aspects of dying further supported our assumption.3, 8, 1012

Our interview study revealed occasional beliefs that may characterize Americans in general, some beliefs that may characterize only certain ethnic groups or genders, and many beliefs that may characterize only particular individuals.

METHODS

We constructed a semistructured interview, based on topics and questions from the end‐of‐life literature and our own encounters with dying patients and their survivors. The interview schedule covered topics such as the right time to die, what happens at death, and the afterlife. We pretested all questions before using them in interviews.

Study participants were older inpatients from the 3 largest American ethnic groupsMexican Americans (MAs), Euro‐Americans (EAs), and African Americans (AAs),13 as identified by a validated ethnic algorithm.14 We reasoned that age, current serious (though not necessarily terminal) illness, and having experienced the deaths of others had already prompted these older inpatients to think about death.11

Admission logs from 2 San Antonio, Texas, hospitals identified all patients, aged 50 to 79, who were admitted over a 9‐month period for any of 10 common internal medicine diagnoses. From these logs we selected interviewees by purposive sampling, a nonstatistical technique that ensured adequate participant numbers by ethnic group and gender.15, 16 We invited patients to interview only after their primary physicians gave permission.

Sixty of 65 participants who began interviews completed them, and 58 of the 60 could be classified into 1 of the 3 ethnic groups.14 These 58, who produced saturation for all themes mentioned by more than 5% of participants, constituted our analysis sample. Participants included 26 MAs (14 men, 12 women), 18 EAs (7 men, 11 women), and 14 AAs (7 men, 7 women). The most prevalent admitting diagnoses were congestive heart failure (19 participants), angina (17 participants), and pneumonia and chronic obstructive pulmonary disease (5 participants each). The 3 ethnic group samples had similar mean ages but differed in other ways (Table 1). MAs were typically Roman Catholic, educated through grade 7, and married; EAs were divided between Roman Catholic and Protestant, educated through grade 12, and mostly unmarried; and AAs were nearly all Protestant, educated through grade 11, and mostly unmarried. The genders within each ethnic group sample were similar by age, religion, and education (data not shown). AA men and women were also similar by marital status. However, MAs and EAs had more men than women who were married, and more women than men who were widowed.

Two trained, bilingual women1 MA and 1 EA, not specifically matched to participants by ethnic groupused the schedule of questions to interview participants. The interviews usually took place 3 days after admission, involved one‐on‐one engagement in participants' hospital rooms, were audiotaped, and lasted roughly 90 minutes. Most questions were open‐ended, allowing participants to express beliefs in their own words. For example, the open‐ended question, What do you think happens at the end of a person's life? introduced the topic covered here. To help focus responses, interviewers asked participants early on to name the closest person to them to have died. Interviewers then encouraged participants to describe their beliefs specifically in terms of that person's death. (We assumed the closeness of the relationships had kept those deaths vivid for participants even years afterward.) Most participants responded by referring to that person, but a few described their own death‐like experiences or their general beliefs about death. Interviewers probed as necessary to clarify responses.

Participants interviewed in Spanish or English as they preferred. Two MAs interviewed entirely in Spanish, 10 MAs interviewed partly in Spanish and partly in English, and all other participants interviewed entirely in English. Bilingual typists transcribed the audiotapes, translating any Spanish into English. Two bilingual experts independently confirmed the accuracy of the translations.

The coders who content‐analyzed responses varied by ethnic group, gender, and professional training. They conducted their analysis in 4 steps, each involving initial, independent, blinded reviews by 2 coders; comparison of interpretations; and consensual resolution of disagreements. First, 2 coders deleted any comments irrelevant to death or dying. Second, the same coders assembled the remaining passages by topic, such as beliefs about what happens at the time of death. Third, 1 original coder and a senior investigator naive to the responses identified themes within each topic. Fourth, that original coder and either of 2 new coders determined for each interview the presence or absence of each theme. Theme presence required agreement between the original coder and the new coder or, when they differed, agreement between one of these coders and an independent adjudicator. Lastly, the 3 authors aggregated themes into meaningful categories by consensus, and checked these categories for trustworthiness against participants' original comments.

We report the results for each theme primarily as the percentages of participants within these ethnic group or gender samples who mentioned the theme. Though content analyses are usually reported qualitatively, percentages have 2 advantages here.17 First, readers can see the percentages and judge for themselves the important similarities, differences, and patterns in the data. Second, percentages enable researchers to formulate their own questions for further study, say, questions based on the largest percentages or largest percentage differences in the data. We also report representative quotes to illustrate the depth and richness of participant responses.

The study complied with all institutional review board requirements.

RESULTS

Most participants in all 3 ethnic group samples named a parent as the closest person to them to have died (Table 2). Among these participants, MAs named their mothers overwhelmingly, but EAs and AAs named their mothers and fathers nearly equally. Other participants named siblings, children, other relatives, or friends. Of 13 widowed participants, only 4 named their spouses.

Thirty‐nine participants20 MAs, 13 EAs, and 6 AAsdescribed the closest person's time of death as either momentary (typically less than a minute) or prolonged (longer than a few minutes). More EAs described it as momentary than as prolonged (50% vs 22%). One EA woman said, We were right outside [the hospital room when my father suffered his cardiac arrest]. We knew, when the alarm went off on the heart monitor, it was the last time we'd see him alive. In contrast, MAs and AAs split roughly equally between describing death as momentary or as prolonged (MAs: 35% and 42%, respectively; AAs: 21% for both).

The ethnic group samples also differed about harm from treatments the closest person had received when dying. Of participants who specified such treatments, disproportionately more AAs (4 of 6) than MAs (0 of 5) or EAs (1 of 8) said those treatments had caused the person to suffer at the time of death. Recalling the prolonged resuscitation efforts on his father, one AA man said that the doctors were trying to keep him alive I said, Don't put his body through that.

Many participants went on to describe their beliefs about what happens at the time of death, about the physiologic signs that define that time, and about the senses that persist after death.

DISCUSSION

Beliefs about what happens at death surely affect the whole dying experience1 and may help guide end‐of‐life care. Yet for all the research on dying, the health professions literatures contain virtually no studies describing such beliefs.19 This exploratory study begins the descriptive process.

The results suggest a taxonomyhowever provisionalfor those beliefs (Table 6). Occasional beliefs, such as the one that death separates the dead from the living, may be held by many Americans and thereby characterize American society in general. Other beliefs may characterize only particular American ethnic groups or genders. Ethnically based beliefs may include, for MAs, the belief that death occurs when an external force, specifically God or Jesus, takes the dead person away; for EAs, the beliefs that death occurs in less than a minute and that physiologic signs define the time of death; and, for AAs, the belief that cooling of the body never defines the time of death. A gender‐based belief may be the belief of EA women that some senses persist after death. Still other beliefs may be idiosyncratic, varying among individuals without a demographic pattern. Idiosyncratic beliefs may include which particular physiologic sign defines the time of death.

The reader must consider these results in light of the study's assumptions, weaknesses, and strengths. Two assumptions were key: that participants expressed themselves fully (despite the difficulty of articulating such beliefs), and that the content analysts interpreted them accurately. Study weaknesses included possible conditioning of responses through prior interview questions, incomplete responsiveness about certain themes, nongeneralizability beyond these sample groups due to the purposive sampling, and educational and marital status differences as possible alternative explanations for the results. Important study strengths included the clinically important topic; the ill, older participants who had already faced death for themselves or others; the pretested, bilingual interview schedule; the open‐ended questions allowing participants to express beliefs in their own words; and the rigorous content analysis.

While indicating directions for future research, our results also provide several useful lessons for current end‐of‐life care. First, hospitalists and other health professionals who attend the dying must not assume they can accurately predict the beliefs of patients or survivors about what happens at the time of death. Many beliefs that participants expressed here neither we nor the health professionals to whom we have presented the results could have imagined beforehand. Health professionals simply cannot expect patients and survivors to hold the same beliefs as they. Furthermore, while some beliefs may characterize certain demographic groups in general, large idiosyncratic variation within groups compromises many demographically based predictions of particular individuals' beliefs. Thus, health professionals can probably learn such beliefs only by eliciting them individual by individual.18 Admittedly awkward, the necessary inquiries6, 20, 21 demand courage and patience, but, when done well, may help prepare all for the death.

Second, these inquiries require health professionals to cultivate techniques for eliciting people's beliefs sensitively and accurately.18 For example, health professionals might ask questions using the same terms for death (such as goes, passes, or rests) that the patient or survivors use. This technique encourages open expression of beliefs by assuring people that health professionals are listening carefully. Still, professionals must guard against misunderstandings that such terms may createparticularly by giving false hope or ignoring sad realities.

Third, armed with knowledge of patient and survivor beliefs, health professionals should tailor perimortem care accordingly. For example, because many people suffer pangs of separation at a death, health professionals should address those feelings expressly.22 Doing so may promote closure for the grieving.20, 2325 Health professionals should also attend to beliefs such as those about the signs and duration of the time of death. As this study showed, different people may time death by different physiologic signs. The patient with a warm body but no heartbeat may be simultaneously alive to someone who sees cooling of the body as defining the time of death, and dead to someone else who sees cessation of heartbeat as doing so.18, 26 Unfortunately, certain perimortem procedures, such as harvesting organs for transplantation, moving the body to the morgue, or performing an autopsy, require declaring death unambiguously at one specific time. When differences may exist over the signs or duration of death, the best way to avoid agonizing arguments or decisional paralysis when death occurs is to articulate any differences beforehand and to resolve them with a mutually acceptable management plan.27 Health professionals should also honor survivor beliefs about sentience of the dead. Unlike differences over signs and duration of death, differences over sentience of the dead may often be accommodated without definitive resolution. For example, although health professionals should treat any body respectfully, they must make special efforts to handle the dead body gently in case survivors believe it can still feel pain.

CONCLUSION

A patient's death is a climactic event for patient, survivors, and health professionals alike. It warrants careful management. Personal beliefs about what happens at death surely affect how patients anticipate it and what survivors remember of it. Compassionate perimortem care, therefore, must address those beliefs.28

Yet demographic characteristics such as ethnic group or gender offer only limited clues to such beliefs, making health professionals elicit them directly.4, 6, 9, 21 Discussions with patients and survivors about these beliefs are bound to create emotional discomfort for everybody, but health professionals may dispel much of it by demonstrating a sincere commitment to personalized end‐of‐life care, showing respect for the beliefs of patients and survivors,7, 21, 23 and accommodating those beliefs whenever possible.2, 28 The result may be the best of all possible outcomessensitive, respectful, and compassionate care for patients and survivors, and rewarding caregiving experiences for health professionals.20, 24, 25

Every health professional bears responsibility at some point for helping dying patients and their survivors face death spiritually and emotionally. But, because most Americans die in hospitals, that responsibility falls disproportionately to hospitalists and other hospital‐based health professionals.

Personal beliefs about what happens at death surely influence whether patients welcome or dread it, and whether survivors remember it with relief or regret.1 We believe, therefore, that competent, compassionate end‐of‐life care requires hospitalists and other health professionals who attend dying patients to address such beliefs. But people do not readily volunteer them, health professionals rarely elicit them, and little research describes them.

We, therefore, performed a large exploratory study to begin characterizing patients' beliefs about what happens at the time of death. We assumed that culturethe values a group uses to interpret shared experiences and transmits across generations2, 3influences those beliefs.1, 410 We reasoned that, because death is a universal human experience, every culture must address its meaning.5 Prior studies showing ethnic cultural differences over advance care planning, life support, and other aspects of dying further supported our assumption.3, 8, 1012

Our interview study revealed occasional beliefs that may characterize Americans in general, some beliefs that may characterize only certain ethnic groups or genders, and many beliefs that may characterize only particular individuals.

METHODS

We constructed a semistructured interview, based on topics and questions from the end‐of‐life literature and our own encounters with dying patients and their survivors. The interview schedule covered topics such as the right time to die, what happens at death, and the afterlife. We pretested all questions before using them in interviews.

Study participants were older inpatients from the 3 largest American ethnic groupsMexican Americans (MAs), Euro‐Americans (EAs), and African Americans (AAs),13 as identified by a validated ethnic algorithm.14 We reasoned that age, current serious (though not necessarily terminal) illness, and having experienced the deaths of others had already prompted these older inpatients to think about death.11

Admission logs from 2 San Antonio, Texas, hospitals identified all patients, aged 50 to 79, who were admitted over a 9‐month period for any of 10 common internal medicine diagnoses. From these logs we selected interviewees by purposive sampling, a nonstatistical technique that ensured adequate participant numbers by ethnic group and gender.15, 16 We invited patients to interview only after their primary physicians gave permission.

Sixty of 65 participants who began interviews completed them, and 58 of the 60 could be classified into 1 of the 3 ethnic groups.14 These 58, who produced saturation for all themes mentioned by more than 5% of participants, constituted our analysis sample. Participants included 26 MAs (14 men, 12 women), 18 EAs (7 men, 11 women), and 14 AAs (7 men, 7 women). The most prevalent admitting diagnoses were congestive heart failure (19 participants), angina (17 participants), and pneumonia and chronic obstructive pulmonary disease (5 participants each). The 3 ethnic group samples had similar mean ages but differed in other ways (Table 1). MAs were typically Roman Catholic, educated through grade 7, and married; EAs were divided between Roman Catholic and Protestant, educated through grade 12, and mostly unmarried; and AAs were nearly all Protestant, educated through grade 11, and mostly unmarried. The genders within each ethnic group sample were similar by age, religion, and education (data not shown). AA men and women were also similar by marital status. However, MAs and EAs had more men than women who were married, and more women than men who were widowed.

Two trained, bilingual women1 MA and 1 EA, not specifically matched to participants by ethnic groupused the schedule of questions to interview participants. The interviews usually took place 3 days after admission, involved one‐on‐one engagement in participants' hospital rooms, were audiotaped, and lasted roughly 90 minutes. Most questions were open‐ended, allowing participants to express beliefs in their own words. For example, the open‐ended question, What do you think happens at the end of a person's life? introduced the topic covered here. To help focus responses, interviewers asked participants early on to name the closest person to them to have died. Interviewers then encouraged participants to describe their beliefs specifically in terms of that person's death. (We assumed the closeness of the relationships had kept those deaths vivid for participants even years afterward.) Most participants responded by referring to that person, but a few described their own death‐like experiences or their general beliefs about death. Interviewers probed as necessary to clarify responses.

Participants interviewed in Spanish or English as they preferred. Two MAs interviewed entirely in Spanish, 10 MAs interviewed partly in Spanish and partly in English, and all other participants interviewed entirely in English. Bilingual typists transcribed the audiotapes, translating any Spanish into English. Two bilingual experts independently confirmed the accuracy of the translations.

The coders who content‐analyzed responses varied by ethnic group, gender, and professional training. They conducted their analysis in 4 steps, each involving initial, independent, blinded reviews by 2 coders; comparison of interpretations; and consensual resolution of disagreements. First, 2 coders deleted any comments irrelevant to death or dying. Second, the same coders assembled the remaining passages by topic, such as beliefs about what happens at the time of death. Third, 1 original coder and a senior investigator naive to the responses identified themes within each topic. Fourth, that original coder and either of 2 new coders determined for each interview the presence or absence of each theme. Theme presence required agreement between the original coder and the new coder or, when they differed, agreement between one of these coders and an independent adjudicator. Lastly, the 3 authors aggregated themes into meaningful categories by consensus, and checked these categories for trustworthiness against participants' original comments.

We report the results for each theme primarily as the percentages of participants within these ethnic group or gender samples who mentioned the theme. Though content analyses are usually reported qualitatively, percentages have 2 advantages here.17 First, readers can see the percentages and judge for themselves the important similarities, differences, and patterns in the data. Second, percentages enable researchers to formulate their own questions for further study, say, questions based on the largest percentages or largest percentage differences in the data. We also report representative quotes to illustrate the depth and richness of participant responses.

The study complied with all institutional review board requirements.

RESULTS

Most participants in all 3 ethnic group samples named a parent as the closest person to them to have died (Table 2). Among these participants, MAs named their mothers overwhelmingly, but EAs and AAs named their mothers and fathers nearly equally. Other participants named siblings, children, other relatives, or friends. Of 13 widowed participants, only 4 named their spouses.

Thirty‐nine participants20 MAs, 13 EAs, and 6 AAsdescribed the closest person's time of death as either momentary (typically less than a minute) or prolonged (longer than a few minutes). More EAs described it as momentary than as prolonged (50% vs 22%). One EA woman said, We were right outside [the hospital room when my father suffered his cardiac arrest]. We knew, when the alarm went off on the heart monitor, it was the last time we'd see him alive. In contrast, MAs and AAs split roughly equally between describing death as momentary or as prolonged (MAs: 35% and 42%, respectively; AAs: 21% for both).

The ethnic group samples also differed about harm from treatments the closest person had received when dying. Of participants who specified such treatments, disproportionately more AAs (4 of 6) than MAs (0 of 5) or EAs (1 of 8) said those treatments had caused the person to suffer at the time of death. Recalling the prolonged resuscitation efforts on his father, one AA man said that the doctors were trying to keep him alive I said, Don't put his body through that.

Many participants went on to describe their beliefs about what happens at the time of death, about the physiologic signs that define that time, and about the senses that persist after death.

DISCUSSION

Beliefs about what happens at death surely affect the whole dying experience1 and may help guide end‐of‐life care. Yet for all the research on dying, the health professions literatures contain virtually no studies describing such beliefs.19 This exploratory study begins the descriptive process.

The results suggest a taxonomyhowever provisionalfor those beliefs (Table 6). Occasional beliefs, such as the one that death separates the dead from the living, may be held by many Americans and thereby characterize American society in general. Other beliefs may characterize only particular American ethnic groups or genders. Ethnically based beliefs may include, for MAs, the belief that death occurs when an external force, specifically God or Jesus, takes the dead person away; for EAs, the beliefs that death occurs in less than a minute and that physiologic signs define the time of death; and, for AAs, the belief that cooling of the body never defines the time of death. A gender‐based belief may be the belief of EA women that some senses persist after death. Still other beliefs may be idiosyncratic, varying among individuals without a demographic pattern. Idiosyncratic beliefs may include which particular physiologic sign defines the time of death.

The reader must consider these results in light of the study's assumptions, weaknesses, and strengths. Two assumptions were key: that participants expressed themselves fully (despite the difficulty of articulating such beliefs), and that the content analysts interpreted them accurately. Study weaknesses included possible conditioning of responses through prior interview questions, incomplete responsiveness about certain themes, nongeneralizability beyond these sample groups due to the purposive sampling, and educational and marital status differences as possible alternative explanations for the results. Important study strengths included the clinically important topic; the ill, older participants who had already faced death for themselves or others; the pretested, bilingual interview schedule; the open‐ended questions allowing participants to express beliefs in their own words; and the rigorous content analysis.

While indicating directions for future research, our results also provide several useful lessons for current end‐of‐life care. First, hospitalists and other health professionals who attend the dying must not assume they can accurately predict the beliefs of patients or survivors about what happens at the time of death. Many beliefs that participants expressed here neither we nor the health professionals to whom we have presented the results could have imagined beforehand. Health professionals simply cannot expect patients and survivors to hold the same beliefs as they. Furthermore, while some beliefs may characterize certain demographic groups in general, large idiosyncratic variation within groups compromises many demographically based predictions of particular individuals' beliefs. Thus, health professionals can probably learn such beliefs only by eliciting them individual by individual.18 Admittedly awkward, the necessary inquiries6, 20, 21 demand courage and patience, but, when done well, may help prepare all for the death.

Second, these inquiries require health professionals to cultivate techniques for eliciting people's beliefs sensitively and accurately.18 For example, health professionals might ask questions using the same terms for death (such as goes, passes, or rests) that the patient or survivors use. This technique encourages open expression of beliefs by assuring people that health professionals are listening carefully. Still, professionals must guard against misunderstandings that such terms may createparticularly by giving false hope or ignoring sad realities.

Third, armed with knowledge of patient and survivor beliefs, health professionals should tailor perimortem care accordingly. For example, because many people suffer pangs of separation at a death, health professionals should address those feelings expressly.22 Doing so may promote closure for the grieving.20, 2325 Health professionals should also attend to beliefs such as those about the signs and duration of the time of death. As this study showed, different people may time death by different physiologic signs. The patient with a warm body but no heartbeat may be simultaneously alive to someone who sees cooling of the body as defining the time of death, and dead to someone else who sees cessation of heartbeat as doing so.18, 26 Unfortunately, certain perimortem procedures, such as harvesting organs for transplantation, moving the body to the morgue, or performing an autopsy, require declaring death unambiguously at one specific time. When differences may exist over the signs or duration of death, the best way to avoid agonizing arguments or decisional paralysis when death occurs is to articulate any differences beforehand and to resolve them with a mutually acceptable management plan.27 Health professionals should also honor survivor beliefs about sentience of the dead. Unlike differences over signs and duration of death, differences over sentience of the dead may often be accommodated without definitive resolution. For example, although health professionals should treat any body respectfully, they must make special efforts to handle the dead body gently in case survivors believe it can still feel pain.

CONCLUSION

A patient's death is a climactic event for patient, survivors, and health professionals alike. It warrants careful management. Personal beliefs about what happens at death surely affect how patients anticipate it and what survivors remember of it. Compassionate perimortem care, therefore, must address those beliefs.28

Yet demographic characteristics such as ethnic group or gender offer only limited clues to such beliefs, making health professionals elicit them directly.4, 6, 9, 21 Discussions with patients and survivors about these beliefs are bound to create emotional discomfort for everybody, but health professionals may dispel much of it by demonstrating a sincere commitment to personalized end‐of‐life care, showing respect for the beliefs of patients and survivors,7, 21, 23 and accommodating those beliefs whenever possible.2, 28 The result may be the best of all possible outcomessensitive, respectful, and compassionate care for patients and survivors, and rewarding caregiving experiences for health professionals.20, 24, 25