American Journal of Medicine

Volume 120, Issue 10, Pages, 825-918 (October 2007)

1. Obesity, Heart Disease, and Favorable Prognosis—Truth or Paradox? Pages 825-826 Carl J. Lavie, Richard V. Milani and Hector O. Ventura

2. Diagnosis of Acute Pulmonary Embolism: Always Be Vigilant Pages 827-828 Samuel Z. Goldhaber

Review

3. Impact of Diabetes on the Severity of Liver Disease Pages 829-834 Ingrid J. Hickman and Graeme A. Macdonald

Office management: geriatrics

4. Considerations for the Diagnosis and Treatment of Testosterone Deficiency in Elderly Men Pages 835-840 Mohammed Kazi, Stephen A. Geraci and Christian A. Koch

5. Orthostatic Hypotension in the Elderly: Diagnosis and Treatment Pages 841-847 Vishal Gupta and Lewis A. Lipsitz

Diagnostic dilemma

6. Fact or Factitious? Pages 848-850 Stefan Jenni, Beat Gloor, Christoph Stettler and Emanuel R. Christ

7. Over the Speed Limit Pages 851-853 Melissa R. Robinson, Ryan A. Brown, Uma Srivatsa and Ezra A. Amsterdam

Images in dermatology

8. Board Stiff Pages 854-856 Peter Mattei, Julie Templet and Carrie Cusack

ECG image of the month

9. A Heart Aflutter Pages 857-859 Ahmad Khraisat, Sarabjeet Singh, Rohit Arora and Eshraq Al-Jaghbeer

Images in radiology

10. No Simple Sore Throat Pages 860-862 William J. Salyers Jr., Christina Schnose and Adam Zarchan

Clinical research studies

11. Obesity Paradox in Patients with Hypertension and Coronary Artery Disease Pages 863-870 Seth Uretsky, Franz H. Messerli, Sripal Bangalore, Annette Champion, Rhonda M. Cooper-DeHoff, Qian Zhou and Carl J. Pepine

12. Clinical Characteristics of Patients with Acute Pulmonary Embolism: Data from PIOPED II Pages 871-879 Paul D. Stein, Afzal Beemath, Fadi Matta, John G. Weg, Roger D. Yusen, Charles A. Hales, Russell D. Hull, Kenneth V. Leeper Jr., H. Dirk Sostman, Victor F. Tapson, John D. Buckley, Alexander Gottschalk, Lawrence R. Goodman, Thomas W. Wakefied and Pamela K. Woodard

13. Thyroid Function Abnormalities during Amiodarone Therapy for Persistent Atrial Fibrillation Pages 880-885 Elizabeth L. Batcher, X. Charlene Tang, Bramah N. Singh, Steven N. Singh, Domenic J. Reda and Jerome M. Hershman

14. Randomized Trial to Improve Fracture Prevention in Nursing Home Residents Pages 886-892 Cathleen S. Colón-Emeric, Kenneth W. Lyles, Paul House, Deborah A. Levine, Anna P. Schenck, Jeroan Allison, Joel Gorospe, Mary Fermazin, Kristi Oliver, Jeffrey R. Curtis, Norman Weissman, Aiyuan Xie and Kenneth G. Saag

15. Impact of a Fluoroquinolone Restriction Policy in an Elderly Population Pages 893-900 Muhammad Mamdani, David McNeely, Gerald Evans, Janet Hux, Paul Oh, Natalie Forde and John Conly

16. Prevention of Central Venous Catheter-Associated Thrombosis: A Meta-analysis Pages 901.e1-901.e13 Angelia Kirkpatrick, Suman Rathbun, Thomas Whitsett and Gary Raskob

AJM online

Review

17. Diagnostic Evaluation of Mononucleosis-Like Illnesses Pages 911.e1-911.e8 Christopher Hurt and Dominick Tammaro

Clinical research study

18. Incidence and Clinical Spectrum of Thiazide-associated Hypercalcemia Pages 911.e9-911.e15 Robert A. Wermers, Ann E. Kearns, Gregory D. Jenkins and L. Joseph Melton III

Erratum

19. Erratum Page 911

Clinical communications to the editor

20. Marinol-Induced Gynecomastia: A Case Report Page e1 Rebekah C. Allen, Anne Marie Wallace and Melanie Royce

21. Sarcoidosis Manifesting as a Periorbital Purplish Rash Pages e3-e4 Mitsuhito Ota, Daigo Nakazawa and Daisuke Sawamura

Letters

22. Alcohol and Gout Page e5 Hyon K. Choi and Gary Curhan

23. The Reply Page e7 Yuqing Zhang, Tuhina Neogi and David J. Hunter

24. Functional Status in Chronic Obstructive Pulmonary Disease Page e9 Andrea Corsonello, Claudio Pedone and Raffaele Antonelli Incalzi

25. The Reply Page e11 Richard ZuWallack

26. Broadening the Differential Diagnosis from a Different Perspective Page e13 Oscar M. Jolobe

27. Effective Detection of Celiac Disease Using Salivary Anti-transglutaminase Page e15 Annick Ocmant and Françoise Mascart

28. The Reply Page e17 Peter H.R. Green

29. Noninvasive Ventilation in Acute Heart Failure Page e19 Ritesh Agarwal and Rajagopala Srinivas

30. The Reply Page e21 John R. Kapoor and Mark A. Perazella

31. Endothelium-Independent Microvascular Dysfunction in Cardiac Syndrome X Page e23 Pankaj Madan and Ritu Madan

32. The Reply Page e25 Christopher P. Appleton and R. Todd Hurst

APM perspectives

33. Communication Skills: A Call for Teaching to the Test Pages 912-915 Anna Headly

Medical humanities perspectives

34. The Heroic Physician and The Gross Clinic Pages 916-917 Helle Mathiasen

The American Journal of Medicine (2007) 120, 825-826

EDITORIAL Obesity, Heart Disease, and Favorable Prognosis—Truth or Paradox?

The words of truth are always paradoxical. non-purposeful weight loss, which would be associated with a —Lao Tzu poor prognosis, was not assessed. Although the underlying reasons for the obesity paradox in this cohort are not well Despite the fact that obesity has been shown to be an understood, it is possible that obese hypertensives have lower independent risk factor for cardiovascular disease (CVD),1,2 systemic vascular resistance as well as plasma renin activity many studies have demonstrated that obese patients with compared with lean hypertensives, and these hemodynamic established CVD have a better prognosis than do patients parameters and neurohormonal mechanisms might contribute with “ideal” body weight—the so called obesity-paradox. This paradox has been best described for patients with to their better prognosis. advanced systolic heart failure (HF),3,4 including those Although improved outcomes appear to be a consistent where a higher percentage of body fat was the strongest association with increased body weight, one should not predictor of event-free survival.4 In advanced HF, obese conclude that weight reduction is detrimental in over- patients can present earlier because of reduced circulat- weight populations. We previously demonstrated that ing natriuretic peptides, as well as being more severely obese subjects with CHD who were more successful with symptomatic due to non-HF factors, including decondi- weight reduction had considerably greater improvements tioning and restrictive lung disease. Additionally, ca- in most CHD risk factors following rehabilitation com- 2 chexia and non-purposeful weight loss are associated pared with obese subjects who did not lose weight. with a particularly poor prognosis.3,4 Likewise, the par- Likewise, in a study of over 1500 CHD patients, inten- adoxical relationship between obesity and prognosis also tional weight loss from a 6-month diet produced lower 8 is noted in patients with advanced cancers, end-stage incidence of CHD events over 4 years. These studies renal disease, and in the very elderly.3 support purposeful weight reduction in obese CHD pa- In the present issue of The American Journal of tients, despite the “obesity paradox.” Medicine, Uretsky et al5 demonstrate that in a large Finally, we must not lose sight of the fact that obesity population with hypertension and coronary heart disease might be intimately involved in the pathogenesis of hyper- (CHD), overweight and obese patients had decreased risk tension and CHD; therefore, without signiﬁcant weight gain, of major CVD events, particularly mortality, compared many of the overweight and obese patients in the INternational with “normal” weight patients. These data are in agree- VErapamil SR-trandolopril STudy (INVEST) might not have ment with a recent meta-analysis by Romero-Corral et al developed hypertension and/or CHD in the ﬁrst place. As we that demonstrated better CVD outcomes in overweight continue to investigate the obesity paradox in CVD, including and mildly obese CHD patients compared with those CHD, we should remember the old proverb, “Only one thing is with ideal weight and especially compared with under- certain—that is nothing is certain.” weight patients (who generally have the worst prog- Carl J. Lavie, MD nosis in most studies).6 In addition, it also supports the Richard V. Milani, MD results of a recent study from nearly 7000 male non-HF Hector O. Ventura, MD veterans referred for stress testing that demonstrated a Department of Cardiovascular Diseases similar obesity paradox.7 Nevertheless, the manuscript by Ochsner Medical Center Uretsky et al5 is still noteworthy, in that they studied New Orleans, La nearly 23,000 hypertensives with CHD who were vigor- [email protected] ously treated. One of the limitations, as in most studies, is that References 1. Lavie CJ, Milani RV. Obesity and cardiovascular disease: the Hip- Requests for reprints should be addressed to Carl J. Lavie, MD, FACP, pocrates paradox? J Am Coll Cardiol. 2003;42:677-679. FACC, FCCP, Ochsner Medical Center, 1514 Jefferson Highway, New 2. Lavie CJ, Milani RV. Effects of cardiac rehabilitation, exercise training, Orleans, LA 70121-2483. and weight reduction on exercise capacity, coronary risk factors, be-

havioral characteristics, and quality of life in obese coronary patients. 6. Romero-Corral A, Montori VM, Somers VK, et al. Association of Am J Cardiol. 1997;79:397-401. bodyweight with total mortality and with cardiovascular events in cor- 3. Lavie CJ, Mehra MR, Milani RV. Obesity and heart failure prognosis: onary artery disease: a systematic review of cohort studies. Lancet. paradox or reverse epidemiology? Eur Heart J. 2005;26:5-7. 2006;368:666-678. 4. Lavie CJ, Osman AF, Milani RV, Mehra MR. Body composition and 7. McAuley P, Myers J, Abella J, Froelicher V. Body mass, ﬁtness and survival prognosis in chronic systolic heart failure: the obesity paradox. Am J in veteran patients: another obesity paradox? Am J Med. 2007;120:518-524. Cardiol. 2003;91:891-894. 8. Eilat-Adar S, Eldar M, Goldbourt U. Association of intentional changes 5. Uretsky S, Messerli FH, Bangalore S, et al. Obesity paradox in hyper- in body weight with coronary heart disease event rates in overweight tension and coronary artery disease: an analysis from the international subjects who have an additional coronary risk factor. Am J Epidemiol. verapamil SR-trandolapril study. A J Med. (in press). 2005;161;352-358. The American Journal of Medicine (2007) 120, 827-828

EDITORIAL

Diagnosis of Acute Pulmonary Embolism: Always Be Vigilant

Chest CT scanning has heralded unprecedented advances in Thus, the diagnosis of pulmonary embolism remains an the diagnosis of pulmonary embolism. Multiplanar imaging art as well as a science. We must acknowledge that this area permits detection of sub-millimeter thrombi so small that of medicine lacks an entirely satisfactory nomogram or their clinical importance may be uncertain. Can we now rely clinical decision algorithm. The weakness in diagnostic upon these technological achievements to substitute for the approach is at the most crucial decision-making point— art of clinical diagnosis? whether suspicion of pulmonary embolism is justified based Failure to identify pulmonary embolism because the upon clinical presentation. diagnosis is not considered remains the last major prob- PIOPED II captures the clinical characteristics of a lem confronting successful detection of pulmonary em- special type of patient with suspected pulmonary embo- bolism. If the health care provider, patient, and family are lism who agrees to undergo an extensive series of diag- not aware of the risk factors and common clinical pre- nostic tests far beyond the usual combination of D-dimer sentations of pulmonary embolism, the diagnosis will be blood testing and chest CT scan. Patients were excluded overlooked. Many laypersons do not even know what a if they had critical illness, abnormal serum creatinine, pulmonary embolism is. contrast allergy, pregnancy, treatment with long-term an- Unfortunately, the Prospective Investigation of Pulmo- ticoagulants, or an inferior vena caval filter. We should nary Embolism Diagnosis (PIOPED) II informs us that the keep in mind that PIOPED II might not represent the clinical presentation of acute pulmonary embolism remains “typical pulmonary embolism patient” in the “real as elusive as ever.1 The number one symptom was rapid world.” onset dyspnea at rest, usually within seconds to hours. The When I evaluate patients for pulmonary embolism, I second most common symptom was pleuritic chest pain. find the clinical setting and classical risk factors espe- Third was cough, in 43%, though I find cough a rare symp- cially useful, along with the presence of a family history tom in the pulmonary embolism patients whom I evaluate. of pulmonary embolism. I believe that unexplained dys- The most common signs were nonspecific: tachypnea and pnea, usually at rest but sometimes with exertion, is the tachycardia. most useful symptom of pulmonary embolism. Anxiety, Symptoms of pulmonary embolism are often accom- difficulty to define but easy to recognize, has provided for panied by symptoms of deep vein thrombosis such me a useful general sign. I have learned that the absence as calf or thigh pain. This can be a helpful hint. In of tachypnea or tachycardia is quite frequent in young PIOPED II, the presence of calf or thigh pain was the pulmonary embolism patients and in those without prior symptom that most commonly differed between patients cardiopulmonary disease. A useful clue in patients who with and without pulmonary embolism. Similarly, the appear ill is the soft systolic murmur of tricuspid regur- presence of calf or thigh edema, erythema, tenderness, or gitation, heard at the left lower sternal border, with re- a palpable cord was the principal sign differentiating spiratory variation in the murmur’s location within sys- pulmonary embolism from no pulmonary embolism in tole and in the murmur’s intensity. PIOPED II. In summary, be vigilant for the possibility of acute Of special concern in PIOPED II is that 18% of those pulmonary embolism because it can be elusive. This is with pulmonary embolism had a low probability clinical especially important when patients with confirmed pneu- assessment. Even in patients ultimately discovered to monia or congestive heart failure do not improve with have main or lobar pulmonary embolism, 15% had low standard therapy. They may be suffering from concomi- clinical probability. tant pulmonary embolism. Remember that the clinical

0002-9343/$ -see front matter © 2007 Elsevier Inc. All rights reserved. doi:10.1016/j.amjmed.2007.07.019 828 The American Journal of Medicine, Vol 120, No 10, October 2007 setting is of paramount importance. Undertake a thorough Samuel Z. Goldhaber, MD history, including family history, and make careful ob- Cardiovascular Division servations during the physical examination regarding Brigham and Women’s Hospital general appearance and vital signs, including a personally Professor of Medicine Harvard Medical School counted respiratory rate. Pay special attention to the Boston, Mass cardiac and leg examinations. More widespread education of practitioners and the public about the potentially Reference vague presentation of pulmonary embolism will raise 1. Stein PD, Beemath A, Matta F, et al. Clinical characteristics of patients awareness, improve vigilance, and result in fewer missed with acute pulmonary embolism: Data from PIOPED II. Am J Med. cases of this potentially fatal illness. 2007;120:871-879. The American Journal of Medicine (2007) 120, 829-834

REVIEW

Impact of Diabetes on the Severity of Liver Disease Ingrid J. Hickman, PhD,a Graeme A. Macdonald, MBBS, PhDa,b aDiamantina Institute for Cancer, Immunology and Metabolic Medicine, and bDepartment of Gastroenterology and Hepatology, Princess Alexandra Hospital, Brisbane, Queensland, Australia.

ABSTRACT

The prevalence of type 2 diabetes is higher in patients who have liver diseases, such as nonalcoholic fatty liver disease, chronic viral hepatitis, hemochromatosis, alcoholic liver disease, and cirrhosis. The development of diabetes in patients with cirrhosis is well recognized, but evidence is emerging that the development of chronic liver disease and progression to cirrhosis may occur after the diagnosis of diabetes and that diabetes plays a role in the initiation and progression of liver injury. This article provides an overview of the evidence for an increased prevalence of diabetes in a range of liver diseases; the effect of diabetes on the severity of disease; the potential mechanisms whereby coexistent diabetes exacerbates progression of hepatic ﬁbrosis; and the impact of obesity, insulin resistance, and type 2 diabetes on clinical outcomes. © 2007 Elsevier Inc. All rights reserved.

KEYWORDS: Cirrhosis; Fatty liver disease; Insulin resistance; Obesity

The prevalence of type 2 diabetes is increasing with a new insulin resistance and type 2 diabetes include central obe- diagnosis made every 21 seconds.1 Historically, the devel- sity, positive calorie load, physical inactivity, age, and ge- opment of diabetes in patients with cirrhosis is well docu- netic predisposition.7 mented with overt diabetes present in up to 70% of cirrhotic The liver plays a key role in the whole-body response to 2 subjects. However, evidence is emerging that the develop- insulin. In the fasting state the liver releases glucose into the ment of chronic liver disease and progression to cirrhosis circulation. After a meal, as blood glucose increases, insulin may occur after the diagnosis of type 2 diabetes and that is secreted from the pancreas and acts on muscle and fat to diabetes plays a role in the initiation and progression of liver stimulate glucose uptake and on the liver to suppress glu- injury. cose output. In insulin-resistant states, more insulin is required for the same effects.7 PATHOGENESIS OF TYPE 2 DIABETES MELLITUS Type 2 diabetes accounts for 90% to 95% of all diagnosed cases of diabetes mellitus3 and is responsible for the major- PREVALENCE OF TYPE 2 DIABETES IS ity of the health burden attributable to diabetes. Type 2 INCREASED IN CHRONIC LIVER DISEASES diabetes develops from an imbalance between insulin sen- The prevalence of type 2 diabetes is higher in patients who sitivity and insulin secretion.4 The earliest detected abnor- have certain liver diseases. There is a link between the mality in individuals who develop type 2 diabetes is im- presence of type 2 diabetes and the severity of liver injury. pairment in the body’s response to insulin.5,6 This is On analysis of these studies, it is important to remember described as insulin resistance. Risk factors associated with the link between diabetes and cirrhosis, because studies with an increased proportion of cirrhotic patients are more likely to ﬁnd an association between type 2 diabetes and Requests for reprints should be addressed to Ingrid Hickman, PhD, disease severity. The liver diseases associated with type 2 Diamantina Institute for Cancer, Immunology and Metabolic Medicine, diabetes include nonalcoholic fatty liver disease, chronic Princess Alexandra Hospital, Ipswich Road, Woolloongabba Qld 4102, Australia. viral hepatitis, hemochromatosis, alcoholic liver disease, E-mail address: [email protected] and cirrhosis.

Nonalcoholic Fatty Liver Disease association with diabetes is well recognized. Up to 23% of The most prevalent liver disease in developed countries is probands and 13% of family members with previously un- 19 nonalcoholic fatty liver disease, which occurs when there is diagnosed hemochromatosis have diabetes. The pathogen- lipid accumulation within the hepatocytes. This can be as- esis of diabetes in hemochromatosis relates in part to the sociated with hepatocyte injury and inflammation that re- deposition of iron in the pancreas. This iron accumulation sults in hepatic fibrosis and ulti- occurs predominantly in exocrine mately in cirrhosis. This more cells; however, iron-containing severe end of the disease spectrum CLINICAL SIGNIFICANCE granules can be shown to accu- is described as nonalcoholic ste- mulate in islet cells, particularly 20 atohepatitis. It is estimated that up ● Patients with nonalcoholic fatty liver in insulin-secreting beta-cells. to one third of adult Americans disease, hepatitis C, hemochromatosis, Similar findings are present in may have nonalcoholic fatty liver alcoholic liver disease, or cirrhosis have the pancreas of iron-loaded ani- 8 21 disease, whereas the prevalence an increased risk of type 2 diabetes. mals. It might therefore be of nonalcoholic steatohepatitis is expected that hemochromatosis estimated to be 2% to 3%. Nonal- ● Patients with type 2 diabetes and liver would be more prevalent in sub- 22 coholic fatty liver disease is con- disease have an increased risk of severe jects with diabetes. Conte et al sidered the most common cause of complications, such as cirrhosis, liver fail- found an increased prevalence cryptogenic cirrhosis.9 It is asso- ure, and hepatocellular carcinoma. of hemochromatosis in individu- ciated with central obesity, and als with type 2 diabetes. Others virtually all patients have evi- ● Patients with type 2 diabetes should be have detected an increased prev- dence of insulin resistance.10 It is monitored for underlying liver disease. alence of HFE mutations in type 2 diabetes,23 but this has not therefore not surprising that type 2 ● The prevalence of severe chronic liver diabetes is present in 21% to 45% been universal. Dubois-Laforgue disease is expected to increase as obe- 24 of patients with nonalcoholic fatty et al found that although the sity and type 2 diabetes rates increase. liver disease11 and is associated prevalence of HFE mutations witha2to5-fold increased risk for was not increased in type 2 dia- nonalcoholic fatty liver disease.12 betes, type 2 diabetic patients with HFE mutations had greater iron stores and were less 25 Viral Hepatitis likely to be obese. Evidence has accumulated regarding the association between glucose metabolism and chronic hepatitis C virus Alcoholic Liver Disease (HCV) infection. The National Health and Nutrition Exam- Patients with alcoholic liver disease have been shown to be 26 ination Survey identified a 3-fold increased risk of type 2 at increased risk for type 2 diabetes. In a prospective diabetes in subjects who were aged more than 40 years and follow-up study of 8663 men, heavy drinking (Ͼ270 g/wk) had chronic HCV, compared with HCV-negative partici- was associated with a 2-fold increased risk of developing pants.13 Knobler et al14 showed the prevalence of type 2 type 2 diabetes compared with moderate drinking (60-120 27 diabetes to be 33% in noncirrhotic patients with HCV com- g/wk). This association was independent of other risk pared with 5.6% in a control group. The reason for the link factors, such as age, obesity, smoking history, family his- 28 between HCV infection and type 2 diabetes may relate to tory of diabetes, and blood pressure. It is noteworthy that viral effects because specific HCV genotypes (particularly similar studies have not shown the same relationship in 28-30 genotype 1) have been linked to insulin resistance,15 and in women. Excessive alcohol intake may have a direct vitro studies have demonstrated HCV proteins inhibiting effect on the development of type 2 diabetes by decreasing insulin signaling.16 insulin-mediated glucose uptake in the acute situation and In hepatitis B virus (HBV) infection, there is conflicting by damaging pancreatic islet cells with chronic use. evidence as to whether there is an increased prevalence of 14 type 2 diabetes. Knobler et al found that the prevalence of CIRRHOSIS AND DIABETES diabetes was twice as high in noncirrhotic subjects with The term “hepatogenous” diabetes is used to describe the HBV infection compared with controls without liver disease association between cirrhosis and impaired glucose metab- (12% vs 5.6%), although this difference was not statistically 14 olism. Up to 96% of patients with cirrhosis have diabetes or significant. Two other studies failed to show an associa- 2 17,18 impaired glucose tolerance. Hepatogenous diabetes differs tion between type 2 diabetes and HBV. from type 2 diabetes in that there is less association with risk factors such as age, body mass index, and family Hemochromatosis history of diabetes.2 Cirrhosis may contribute to the devel- Hereditary hemochromatosis type 1 is the most prevalent opment of type 2 diabetes through numerous factors. With cause of iron accumulation in populations with Northern the development of portal hypertension, blood shunting European ancestry and is caused by mutations of HFE. Its redirects blood away from hepatocytes and results in re- Hickman and Macdonald Diabetes and Liver Disease 831 duced insulin clearance with peripheral hyperinsulinemia.31 In a study of 1604 alcoholic patients, Naveau et al48 This systemic hyperinsulinemia may contribute to the de- found that being overweight was associated with a greater velopment of insulin resistance through the down-regulation prevalence of alcoholic hepatitis and cirrhosis (60% vs 35% of insulin receptors.32 in lean patients).48 The mechanisms whereby obesity and However, cirrhosis alone does not always induce diabe- insulin resistance may exacerbate liver injury in alcoholic tes, and the cause of liver disease and environmental factors liver disease are still debated. Alterations in cytokine pro- may play a role. Zein et al26 found that the prevalence of duction (leptin, adiponectin, and tumor necrosis factor-␣), diabetes was increased in HCV cirrhosis (25%) and alco- caused by both obesity and alcohol, may work synergisti- holic liver disease (19%) but not in cholestatic liver disease cally to activate hepatic stellate cells, resulting in hepatic 49 (1.3%). fibrosis. This association among diabetes, insulin resistance, and liver disease has implications with respect to treatment. The TYPE 2 DIABETES MELLITUS AND THE SEVERITY peroxisome proliferator-activated receptor-gamma agonists OF LIVER DISEASE pioglitazone and rosiglitazone have shown promise in small There is evidence from a range of liver diseases linking studies of patients with nonalcoholic steatohepatitis, including obesity with insulin resistance and hepatic steatosis, which cirrhotic patients33-35; however, there are concerns about their in turn contribute to liver injury. In nonalcoholic fatty liver use in patients with alanine aminotransferase 2.5 times or disease, a range of studies have consistently identified type more the upper limit of normal or with decompensated 2 diabetes as an independent predictor of fibrosis,39,50-52 cirrhosis.36 Long-term efficacy studies have not been com- faster fibrosis progression,51 and increased mortality.53 This pleted. Similarly, metformin has shown some promise in the relationship is maintained when analysis is restricted to treatment of nonalcoholic steatohepatitis,37 although there noncirrhotic patients.51 Younossi et al54 demonstrated that are concerns about its use in advanced liver disease. patients with type 2 diabetes were at greater risk for the development of adverse outcomes such as cirrhosis or liver- 54 THE LINK BETWEEN OBESITY AND INSULIN related mortality. Scrutiny of the impact of type 2 diabetes in HCV has RESISTANCE AND LIVER INJURY identified a role for insulin resistance and type 2 diabetes in Hepatic steatosis, obesity, and insulin resistance seem to act disease progression. Hyperinsulinemia,42 hyperglycemia,46 38 as cofactors for liver injury in a range of liver diseases. In and insulin resistance55 have all been associated with more patients with nonalcoholic fatty liver disease, older age, severe fibrosis in HCV. Again, an important observation is obesity, and the presence of type 2 diabetes are independent that the onset of insulin resistance occurs early, before the risk factors for more severe fibrosis.39 Obesity-related insu- development of cirrhosis.56 lin resistance plays a clear role in the production of reactive The effect of type 2 diabetes on the histologic severity of oxygen species and altered adipokine production, which in alcoholic liver disease has not been widely studied. Raynard turn leads to up-regulation of proinflammatory cytokines. et al57 showed that obesity and increased fasting glucose Insulin resistance results in enhanced hepatic gluconeogen- levels were associated with increased severity of hepatic esis and impaired hepatic lipid metabolism, which results in fibrosis in alcoholic liver disease, independently of daily hepatic steatosis and liver injury.40 alcohol intake and duration of alcohol abuse.57 Although the Obesity and steatosis are associated with more severe prevalence of type 2 diabetes was not reported in this study, fibrosis in chronic HCV.41 Although steatosis in HCV is the link with elevated serum glucose suggests that type 2 linked to genotype-specific viral effects, steatosis is more diabetes may also be correlated with more severe disease. prevalent and worsened in obese, insulin-resistant patients irrespective of viral genotype.42 Emerging evidence sup- TYPE 2 DIABETES AND COMPLICATIONS OF ports the hypothesis that altered metabolic profiles associ- LIVER CIRRHOSIS ated with insulin resistance contribute to liver injury in Type 2 diabetes seems to be associated with an increased 43 HCV. Insulin resistance is an independent predictor for risk of cirrhosis complications. The Verona Diabetes Study, 15 the rate of fibrosis progression, and elevated fasting insu- a population-based study on more than 7000 subjects with 44,45 46 lin and glucose independently seem to contribute to type 2 diabetes, found an increased risk of death from fibrosis in HCV. chronic liver disease and cirrhosis compared with the gen- Hemochromatosis is another liver disease for which co- eral population (standardized mortality ratio after 5 years of existent obesity and steatosis seem to contribute to liver 2.52, 95% confidence interval [CI], 1.96-3.2).58 In addition, injury.47 In a retrospective study of patients with hemochro- there was an increased risk of mortality from hepatocellular matosis before de-ironing, obesity was independently asso- carcinoma (standardized mortality ratio after 10 years of ciated with the presence of steatosis, which in turn was 1.86, 95% CI, 1.43-2.38).58 Insulin treatment of type 2 associated with more severe fibrosis.47 The prevalence of diabetes, perhaps as a marker of more severe diabetes, was diabetes or degree of insulin resistance was not reported in associated with a particularly high risk of mortality in cir- this study. rhotic patients (relative risk 6.84).59 832 The American Journal of Medicine, Vol 120, No 10, October 2007

Similar observations have been made in smaller cohort chronic HCV are emerging with evidence of genotype- studies. Younossi et al54 found that in nonalcoholic fatty specific interactions with adiponectin.73,74 liver disease, patients with type 2 diabetes had an overall Insulin stimulates the proliferation of hepatic stellate mortality twice that of nondiabetic subjects. After adjust- cells and induces production of collagen, resulting in he- ment for potential confounders, including cirrhosis, the risk patic fibrosis.75 Connective tissue growth factor is produced ratio was 22.83 (95% CI, 2.97-175.03) for liver-related by activated hepatic stellate cells and has been implicated in mortality in those with type 2 diabetes and 3.30 (95% CI, the development and progression of hepatic fibrogenesis. 1.76-6.18) for overall mortality.54 Nishida et al60 also found Hyperglycemia and hyperinsulinemia stimulate connective that the survival rates of cirrhotic patients with type 2 tissue growth factor synthesis in hepatic stellate cell cul- diabetes were significantly lower than those with normal tures, and connective tissue growth factor has been found to glucose tolerance. be overexpressed in liver tissue from patients with nonal- Several studies have demonstrated an increased inci- coholic steatohepatitis.76 dence of diabetes among patients with hepatocellular carcinoma ranging from 2- to 4-fold.11 Type 2 diabetes seems to play an etiologic role in hepatocellular carcinoma cirrhosis SIGNIFICANCE independently of alcohol, viral hepatitis, or demographic Patients with type 2 diabetes seem more likely to have a features,61,62 although the risk of hepatocellular carcinoma range of liver diseases, and patients with liver disease and increases up to 10-fold when viral hepatitis and hazardous diabetes are at risk of severe liver disease, cirrhosis, liver alcohol consumption are combined with type 2 diabetes.63 failure, and hepatocellular carcinoma. This has obvious implications for the clinical management. The increasing incidence of obesity and type 2 diabetes in children means we HOW MIGHT TYPE 2 DIABETES MELLITUS MAKE may see more severe chronic liver disease occurring at LIVER DISEASE WORSE? younger ages. Awareness of type 2 diabetes as a significant The pathogenic mechanisms underlying the relationship be- risk factor for liver injury may improve diagnosis and in- tween type 2 diabetes and chronic liver disease remain to be terventions to minimize the progression of chronic liver elucidated. Generalized peripheral insulin resistance and disease. altered ␤-cell function are usually present. The role of Our understanding of the links between type 2 diabetes increased proinflammatory cytokines, reduction in protec- and the development and progression of chronic liver dis- tive cytokines, hyperinsulinemia and hyperglycemia in the ease has benefited from the mainly retrospective studies activation of hepatic stellate cells, and stimulation of colla- performed to date, but prospective studies are needed to gen production are prime focuses of research in this area. fully elucidate the cause and effect of type 2 diabetes in There seems to be cross-talk between adipose tissue and liver injury. Identifying the mechanisms whereby type 2 other tissues mediated in part by the release of cytokines diabetes increases disease severity could offer new insights from adipose tissue, known collectively as adipokines. Adi- into the treatment of chronic liver disease, including the role pokines, such as leptin and tumor necrosis factor-␣, activate of weight reduction and pharmacologic interventions with inflammatory pathways that may exacerbate liver injury.40 insulin sensitizers. The serum concentration of most adipokines is increased in obesity and type 2 diabetes. An exception to this is adiponectin, a key regulator of insulin sensitivity and tissue CONCLUSIONS inflammation.64 Plasma concentrations of adiponectin and Type 2 diabetes is prevalent in a range of liver diseases, hepatic adiponectin receptor expression65 are reduced in particularly nonalcoholic fatty liver disease, chronic HCV, nonalcoholic fatty liver disease, and it has been hypothe- hemochromatosis, and alcoholic liver disease. Coexistent sized that hypoadiponectinemia may play a role in disease type 2 diabetes seems to be associated with more severe progression. Xu et al66 demonstrated a potent protective liver injury before the onset of cirrhosis and more severe effect of adiponectin in animal models of both alcoholic complications and higher mortality once cirrhosis is estab- liver disease and nonalcoholic fatty liver disease. Adminis- lished. There is evidence that the metabolic disturbances tration of recombinant adiponectin to a mouse model of associated with type 2 diabetes contribute to liver injury, but alcoholic liver disease reduced hepatic steatosis and signif- this relationship is made more complex by the association of icantly reduced hepatic inflammation and serum alanine cirrhosis with hepatogenous diabetes. It is unclear whether aminotransferase. Adiponectin had similar effects in the treatment of coexistent diabetes and improved glycemic ob/ob mouse model of nonalcoholic fatty liver disease.66 To control will benefit chronic liver disease. Clinicians should date, studies in humans of the role of adiponectin in liver be aware that patients with type 2 diabetes may have un- disease have produced conflicting results, with some studies derlying chronic liver disease. In the setting of type 2 reporting a protective effect against hepatic steatosis,67,68 diabetes and cirrhosis, consideration should be given to necroinflammation, and fibrosis,69,70 and others reporting no surveillance for life-threatening complications of liver dis- correlation with disease severity.71,72 Data regarding a link ease, such as hepatocellular carcinoma. A better understand- among adiponectin, steatosis, and disease progression in ing of the factors that modulate liver disease progression is Hickman and Macdonald Diabetes and Liver Disease 833 critical to identify patients who require more aggressive 24. Dubois-Laforgue D, Caillat-Zucman S, Djilali-Saiah I, et al. Mutations monitoring, lifestyle interventions, and pharmacotherapy. in HFE, the hemochromatosis candidate gene, in patients with NIDDM. Diabetes Care. 1998;21:1371-1372. 25. Dubois-Laforgue D, Caillat-Zucman S, Boitard C, Timsit J. Clinical References characteristics of type 2 diabetes in patients with mutations of HFE. 1. The American Diabetes Association. Available at: http://www.diabetes. Diabetes Metab. 2000;26:65-68. org/home.jsp. Accessed July 30, 2007. 26. Zein NN, Abdulkarim AS, Wiesner RH, et al. Prevalence of diabetes 2. Holstein A, Hinze S, Thiessen E, et al. Clinical implications of hep- mellitus in patients with end-stage liver cirrhosis due to hepatitis C, atogenous diabetes in liver cirrhosis. J Gastroenterol Hepatol. 2002; alcohol, or cholestatic disease. J Hepatol. 2000;32:209-217. 17:677-681. 27. Wei M, Gibbons LW, Mitchell TL, et al. Alcohol intake and incidence 3. US Department of Health and Human Services. Trends in the preva- of type 2 diabetes in men. Diabetes Care. 2000;23:18-22. lence and incidence of self-reported diabetes mellitus—United States, 28. Holbrook TL, Barrett-Connor E, Wingard DL. A prospective popula- 1980-1994. MMWR Morb Mortal Wkly Rep. 1997;46:1014-1018. tion-based study of alcohol use and non-insulin-dependent diabetes 4. DeFronzo RA, Bonadonna RC, Ferrannini E. Pathogenesis of NIDDM. mellitus. Am J Epidemiol. 1990;132:902-909. A balanced overview. Diabetes Care. 1992;15:318-368. 29. Stampfer MJ, Colditz GA, Willett WC, et al. A prospective study of 5. Martin BC, Warram JH, Krolewski AS, et al. Role of glucose and moderate alcohol drinking and risk of diabetes in women. Am J insulin resistance in development of type 2 diabetes mellitus. Results Epidemiol. 1988;128:549-558. of a 25-year follow-up study. Lancet. 1992;340:925-929. 30. Balkau B, Randrianjohany A, Papoz L, Eschwege E. Re: A prospective 6. Kahn BB. Type 2 diabetes: when insulin secretion fails to compensate population-based study of alcohol use and non-insulin-dependent di- for insulin resistance. Cell. 1998;92:593-596. abetes mellitus. Am J Epidemiol. 1991;134:1469-1470. 7. Reaven GM. Pathophysiology of insulin resistance in human disease. 31. Deschenes M, Somberg KA. Effect of transjugular intrahepatic porto- Physiol Rev. 1995;75:473-486. systemic shunt (TIPS) on glycemic control in cirrhotic patients with 8. Browning JD, Szczepaniak LS, Dobbins R, et al. Prevalence of hepatic diabetes mellitus. Am J Gastroenterol. 1998;93:483. steatosis in an urban population in the United States: impact of eth- 32. Picardi A, D’Avola D, Gentilucci UV, et al. Diabetes in chronic liver nicity. Hepatology. 2004;40:1387-1395. disease: from old concepts to new evidence. Diabetes Metab Res Rev. 9. Caldwell SH, Oelsner DH, Iezzoni JC, et al. Cryptogenic cirrhosis: 2006;22:274-283. clinical characterization and risk factors for underlying disease. Hepa- 33. Belfort R, Harrison SA, Brown K, et al. A placebo-controlled trial of tology. 1999;29:664-669. pioglitazone in subjects with nonalcoholic steatohepatitis. N Engl 10. Chitturi S, Abeygunasekera S, Farrell G, et al. NASH and insulin J Med. 2006;355:2297-2307. 34. Promrat K, Lutchman G, Uwaifo GI, et al. A pilot study of pioglita- resistance: insulin hypersecretion and specific association with the zone treatment for nonalcoholic steatohepatitis. Hepatology. 2004;39: insulin resistance syndrome. Hepatology. 2002;35:373-379. 188-196. 11. Harrison SA. Liver disease in patients with diabetes mellitus. J Clin 35. Neuschwander-Tetri BA, Brunt EM, Wehmeier KR, et al. Improved Gastroenterol. 2006;40:68-76. nonalcoholic steatohepatitis after 48 weeks of treatment with the 12. Clark JM, Brancati FL, Diehl AM. Nonalcoholic fatty liver disease. PPAR-gamma ligand rosiglitazone. Hepatology. 2003;38:1008-1017. Gastroenterology. 2002;122:1649-1657. 36. Waugh J, Keating GM, Plosker GL, et al. Pioglitazone: a review of its 13. Mehta SH, Brancati FL, Sulkowski MS, et al. Prevalence of type 2 use in type 2 diabetes mellitus. Drugs. 2006;66:85-109. diabetes mellitus among persons with hepatitis C virus infection in the 37. Bugianesi E, Gentilcore E, Manini R, et al. A randomized controlled United States. Ann Intern Med. 2000;133:592-599. trial of metformin versus vitamin E or prescriptive diet in nonalcoholic 14. Knobler H, Schihmanter R, Zifroni A, et al. Increased risk of type 2 fatty liver disease. Am J Gastroenterol. 2005;100:1082-1090. diabetes in noncirrhotic patients with chronic hepatitis C virus infec- 38. Powell EE, Jonsson JR, Clouston AD. Steatosis: co-factor in other tion. Mayo Clin Proc. 2000;75:355-359. liver diseases. Hepatology. 2005;42:5-13. 15. Hui JM, Sud A, Farrell GC, et al. Insulin resistance is associated with 39. Angulo P, Keach JC, Batts KP, Lindor KD. Independent predictors of chronic hepatitis C virus infection and fibrosis progression [corrected]. liver fibrosis in patients with nonalcoholic steatohepatitis. Hepatology. Gastroenterology. 2003;125:1695-1704. 1999;30:1356-1362. 16. Kawaguchi T, Yoshida T, Harada M, et al. Hepatitis C virus down- 40. Roden M. Mechanisms of disease: hepatic steatosis in type 2 diabe- regulates insulin receptor substrates 1 and 2 through up-regulation of tes—pathogenesis and clinical relevance. Nat Clin Pract Endocrinol suppressor of cytokine signaling 3. Am J Pathol. 2004;165:1499-1508. Metab. 2006;2:335-348. 17. Fraser GM, Harman I, Meller N, et al. Diabetes mellitus is associated 41. Hourigan LF, Macdonald GA, Purdie D, et al. Fibrosis in chronic with chronic hepatitis C but not chronic hepatitis B infection. Isr J Med hepatitis C correlates significantly with body mass index and steatosis. Sci. 1996;32:526-530. Hepatology. 1999;29:1215-1219. 18. Papatheodoridis GV, Chrysanthos N, Savvas S, et al. Diabetes mellitus 42. Hickman IJ, Powell EE, Prins JB, et al. In overweight patients with in chronic hepatitis B and C: prevalence and potential association with chronic hepatitis C, circulating insulin is associated with hepatic fi- the extent of liver fibrosis. J Viral Hepatol. 2006;13:303-310. brosis: implications for therapy. J Hepatol. 2003;39:1042-1048. 19. Adams PC, Kertesz AE, Valberg LS. Clinical presentation of hemo- 43. Prati D, Shiffman ML, Diago M, et al. Viral and metabolic factors chromatosis: a changing scene. Am J Med. 1991;90:445-449. influencing alanine aminotransferase activity in patients with chronic 20. Rahier J, Loozen S, Goebbels RM, Abrahem M. The haemochroma- hepatitis C. J Hepatol. 2006;44:679-685. totic human pancreas: a quantitative immunohistochemical and ultra- 44. Hickman I, Powell E, Clouston A, et al. Fibrosis in chronic hepatitis C structural study. Diabetologia. 1987;30:5-12. correlates significantly with circulating insulin levels. J Hepatol. 2002; 21. Iancu TC, Ward RJ, Peters TJ. Ultrastructural changes in the pancreas of 36(Suppl 1):172. carbonyl iron-fed rats. J Pediatr Gastroenterol Nutr. 1990;10:95-101. 45. Maeno T, Okumura A, Ishikawa T, et al. Mechanisms of increased 22. Conte D, Manachino D, Colli A, et al. Prevalence of genetic hemo- insulin resistance in non-cirrhotic patients with chronic hepatitis C chromatosis in a cohort of Italian patients with diabetes mellitus. Ann virus infection. J Gastroenterol Hepatol. 2003;18:1358-1363. Intern Med. 1998;128:370-373. 46. Ratziu V, Munteanu M, Charlotte F, et al. Fibrogenic impact of high 23. Kwan T, Leber B, Ahuja S, et al. Patients with type 2 diabetes have a serum glucose in chronic hepatitis C. J Hepatol. 2003;39:1049-1055. high frequency of the C282Y mutation of the hemochromatosis gene. 47. Powell EE, Ali A, Clouston AD, et al. Steatosis is a cofactor in liver Clin Invest Med. 1998;21:251-257. injury in hemochromatosis. Gastroenterology. 2005;129:1937-1943. 834 The American Journal of Medicine, Vol 120, No 10, October 2007

48. Naveau S, Giraud V, Borotto E, et al. Excess weight risk factor for 64. Whitehead JP, Richards AA, Hickman IJ, et al. Adiponectin—a key alcoholic liver disease. Hepatology. 1997;25:108-111. adipokine in the metabolic syndrome. Diabetes Obes Metab. 2006;8: 49. Diehl AM. Obesity and alcoholic liver disease. Alcohol. 2004;34:81-87. 264-280. 50. de Ledinghen V, Ratziu V, Causse X, et al. Diagnostic and predictive 65. Kaser S, Moschen A, Cayon A, et al. Adiponectin and its receptors in factors of significant liver fibrosis and minimal lesions in patients with non-alcoholic steatohepatitis. Gut. 2005;54:117-121. persistent unexplained elevated transaminases. A prospective multi- 66. Xu A, Wang Y, Keshaw H, et al. The fat-derived hormone adiponectin center study. J Hepatol. 2006;45:592-599. alleviates alcoholic and nonalcoholic fatty liver diseases in mice. 51. Adams LA, Sanderson S, Lindor KD, Angulo P. The histological J Clin Invest. 2003;112:91-100. course of nonalcoholic fatty liver disease: a longitudinal study of 103 67. Mendez-Sanchez N, Chavez-Tapia NC, Villa AR, et al. Adiponectin as patients with sequential liver biopsies. J Hepatol. 2005;42:132-138. a protective factor in hepatic steatosis. World J Gastroenterol. 2005; 52. Wanless IR, Lentz JS. Fatty liver hepatitis (steatohepatitis) and obe- 11:1737-1741. sity: an autopsy study with analysis of risk factors. Hepatology. 1990; 68. Bajaj M, Suraamornkul S, Piper P, et al. Decreased plasma adiponectin 12:1106-1110. concentrations are closely related to hepatic fat content and hepatic 53. Adams LA, Lymp JF, St Sauver J, et al. The natural history of nonalcoholic fatty liver disease: a population-based cohort study. Gas- insulin resistance in pioglitazone-treated type 2 diabetic patients. troenterology. 2005;129:113-121. J Clin Endocrinol Metab. 2004;89:200-206. 54. Younossi ZM, Gramlich T, Matteoni CA, et al. Nonalcoholic fatty 69. Musso G, Gambino R, Biroli G, et al. Hypoadiponectinemia predicts liver disease in patients with type 2 diabetes. Clin Gastroenterol the severity of hepatic fibrosis and pancreatic beta-cell dysfunction in Hepatol. 2004;2:262-265. nondiabetic nonobese patients with nonalcoholic steatohepatitis. Am J 55. Taura N, Ichikawa T, Hamasaki K, et al. Association between liver Gastroenterol. 2005;100:2438-2446. fibrosis and insulin sensitivity in chronic hepatitis C patients. Am J 70. Hui JM, Hodge A, Farrell GC, et al. Beyond insulin resistance in Gastroenterol. 2006;101:2752-2759. NASH: TNF-alpha or adiponectin? Hepatology. 2004;40:46-54. 56. Petit JM, Bour JB, Galland-Jos C, et al. Risk factors for diabetes 71. Bugianesi E, Pagotto U, Manini R, et al. Plasma adiponectin in non- mellitus and early insulin resistance in chronic hepatitis C. J Hepatol. alcoholic fatty liver is related to hepatic insulin resistance and hepatic 2001;35:279-283. fat content, not to liver disease severity. J Clin Endocrinol Metab. 57. Raynard B, Balian A, Fallik D, et al. Risk factors of fibrosis in 2005;90:3498-3504. alcohol-induced liver disease. Hepatology. 2002;35:635-638. 72. Jonsson JR, Moschen AR, Hickman IJ, et al. Adiponectin and its 58. Trombetta M, Spiazzi G, Zoppini G, Muggeo M. Review article: type receptors in patients with chronic hepatitis C. J Hepatol. 2005;43: 2 diabetes and chronic liver disease in the Verona diabetes study. 929-936. Aliment Pharmacol Ther. 2005;22(Suppl 2):24-27. 73. Petit JM, Minello A, Jooste V, et al. Decreased plasma adiponectin 59. de Marco R, Locatelli F, Zoppini G, et al. Cause-specific mortality in type concentrations are closely related to steatosis in hepatitis C virus- 2 diabetes. The Verona Diabetes Study. Diabetes Care. 1999;22:756-761. infected patients. J Clin Endocrinol Metab. 2005;90:2240-2243. 60. Nishida T, Tsuji S, Tsujii M, et al. Oral glucose tolerance test predicts 74. Wang AY, Hickman IJ, Richards AA, et al. High molecular weight prognosis of patients with liver cirrhosis. Am J Gastroenterol. 2006; adiponectin correlates with insulin sensitivity in patients with hepatitis 101:70-75. C genotype 3, but not genotype 1 infection. Am J Gastroenterol. 2005; 61. El-Serag HB, Tran T, Everhart JE. Diabetes increases the risk of chronic liver disease and hepatocellular carcinoma. Gastroenterology. 100:2717-2723. 2004;126:460-468. 75. Svegliati-Baroni G, Ridolfi F, Di Sario A, et al. Insulin and insulin-like 62. Kaczynski J, Hansson G, Wallerstedt S. Diabetes: one of few remark- growth factor-1 stimulate proliferation and type I collagen accumula- able differences in clinicopathologic features between cirrhotic and tion by human hepatic stellate cells: differential effects on signal noncirrhotic Swedes with hepatocellular carcinoma. Dig Dis Sci. 2006; transduction pathways. Hepatology. 1999;29:1743-1751. 51:796-802. 76. Paradis V, Perlemuter G, Bonvoust F, et al. High glucose and hyper- 63. Hassan MM, Hwang LY, Hatten CJ, et al. Risk factors for hepatocel- insulinemia stimulate connective tissue growth factor expression: a lular carcinoma: synergism of alcohol with viral hepatitis and diabetes potential mechanism involved in progression to fibrosis in nonalco- mellitus. Hepatology. 2002;36:1206-1213. holic steatohepatitis. Hepatology. 2001;34:738-744. The American Journal of Medicine (2007) 120, 835-840

OFFICE MANAGEMENT: GERIATRICS Michael W. Rich, MD, Speciality Editor Considerations for the Diagnosis and Treatment of Testosterone Deﬁciency in Elderly Men Mohammed Kazi, MD,a,b Stephen A. Geraci, MD,a,b Christian A. Koch, MD, PhDb aMedical Service, G.V. (Sonny) Montgomery Veterans Affairs Medical Center, Jackson, Miss, bDepartment of Medicine, University of Mississippi Medical Center, Jackson.

ABSTRACT

Increased longevity and population aging will increase the number of men with relative testosterone deficiency, as systemic levels of testosterone decrease by about 1% each year. Androgen deficiency should only be diagnosed in men with definite signs and symptoms, accompanied by low total testosterone levels measured in the morning by a reliable assay. Although clinical trials data are limited, current practice guidelines recommend testosterone replacement therapy for symptomatic men with low testosterone levels to improve bone mineral density, muscle mass and strength, sexual function, and quality of life. Testos- terone replacement is not recommended for all older men with low testosterone levels, and should be avoided in patients with prostate or breast cancer, hyperviscosity, erythrocytosis, untreated obstructive sleep apnea, or severe heart failure. The goal of all available testosterone treatment modalities (intramuscular injections, nongenital patch or gel, bioadhesive buccal and oral testosterone, and pellets) is to achieve serum testosterone levels in the mid-normal range during treatment. Cost varies widely among these preparations and may limit their use. Patients receiving testosterone replacement therapy should be re-evaluated 3 months after testosterone initiation and at least annually thereafter. © 2007 Elsevier Inc. All rights reserved.

KEYWORDS: Elderly; Hypogonadism; Testosterone; Therapy

Male hypogonadism is defined as failure of the testes to clinical practice guidelines developed by the Endocrine produce normal amounts of testosterone, combined with Society.5 signs and symptoms of androgen deficiency. Systemic tes- Serum testosterone levels decrease progressively in ag- tosterone levels fall by about 1% each year in men. There- ing men, but the rate and magnitude of decrease vary con- fore, with increasing longevity and the aging of the popu- siderably. Approximately 1% of healthy young men have lation, the number of older men with testosterone deficiency total serum testosterone levels below 250 ng/dL; in contrast, 1-4 will increase substantially over the next several decades. approximately 20% of healthy men over age 60 years have The goal of this article is to provide the primary care serum testosterone levels below this value.6-8 The Balti- physician with guidelines for diagnosing and treating tes- more Longitudinal Study on Aging reported an average tosterone deficiency in the elderly outpatient population. annual decrease of total serum testosterone of 3.2 ng/dL However, it must be emphasized that treatment of testos- in men older than 53 years (ie, about 1% per year based terone deficiency in older men is controversial due to the on a lower limit of normal of 325 ng/dL).8 According to lack of outcomes data from large scale clinical trials. Rec- the Massachusetts Male Aging Study, sex-hormone-bind- ommendations in this article are therefore based on recent ing globulin (SHBG) increases by 1.2% annually. This is important because most circulating testosterone is bound to SHBG or to albumin. Therefore, free and bioavailable Requests for reprints should be addressed to Prof. Christian A. Koch, testosterone levels decrease with age to a greater degree MD, Director, Division of Endocrinology, University of Mississippi Med- ical Center, 2500 N State Street, Jackson, MS 39216. than is reflected by the total testosterone level. However, E-mail address:[email protected] free and bioavailable testosterone levels can be calcu-

0002-9343/$ -see front matter © 2007 Elsevier Inc. All rights reserved. doi:10.1016/j.amjmed.2007.02.022 836 The American Journal of Medicine, Vol 120, No 10, October 2007 lated from total testosterone, SHBG, and serum albumin Table 1 Testosterone Deficiency in Aging Men concentrations (http://www.issam.ch/freetesto.htm).9 1. Are your erections less strong? 2. Do you have a decrease in libido (sex drive)? ETIOLOGY OF MALE HYPOGONADISM 3. Do you have a lack of energy? Luteinizing hormone (LH) and follicle-stimulating hormone 4. Are you falling asleep after dinner? (FSH) levels decrease with age in healthy men. For this 5. Has there been a recent deterioration in your work reason, the most common cause of androgen deficiency in performance? elderly men is hypogonadotropic hypogonadism (ie, inap- 6. Have you noticed a decreased enjoyment of life? propriately low/normal LH and FSH related to pituitary or 7. Do you have decrease in strength and/or endurance? hypothalamic insufficiency). In addition, Leydig cell func- 8. Have you noted a recent deterioration in your ability tion in the testes decreases with aging and is affected by to play sports? several medications, including glucocorticoids, spironolac- 9. Are you sad and/or grumpy? 10. Have you lost weight? tone, opiates, and ketoconazole. Neuroleptic drugs that Positive test: if the answer is yes to question 1 or 2, or to cause hyperprolactinemia can inhibit the release of gonado- any 3 other questions tropin-releasing hormone, leading to hypogonadotropic Modified from Morley et al:10 Metabolism. 2000;49:1239-1242. hypogonadism.2,5 Medical conditions associated with a high prevalence of low testosterone levels include type 2 diabetes mellitus, end-stage renal disease, osteoporosis or low-trauma frac- Local laboratories usually cannot reliably or accurately ture, infertility, diseases of the sellar region, weight loss due measure free serum testosterone. However, free and bio- to malignancy or human immunodeficiency virus (HIV), available testosterone can be calculated from total testoster- and other less common chronic disorders.5 one, SHBG, and albumin (http://www.issam.ch/freetesto. htm).9 Approximately 2% of serum testosterone is unbound DIAGNOSIS OF MALE HYPOGONADISM or free. Because testosterone can rapidly dissociate from Several questionnaires have been developed to assess albumin, all non-SHBG-bound testosterone is considered symptoms of androgen deficiency (Table 1), but it is essen- bioavailable. For total testosterone, the lower limit of the tial to distinguish this condition from major depressive normal range is considered to be around 315 ng/dL (11 disorder, with which there is substantial symptomatic over- nmol/L); for free testosterone and bioavailable testosterone, lap.6,10,11 Other findings suggestive of testosterone defi- lower limits of normal are around 6.5 ng/dL and 140 ng/dL, ciency in men include hot flashes and diaphoresis, very respectively.2 The “free testosterone index” is obtained by small or shrinking testes (adult testes are usually about 4.5 dividing the total testosterone level (in nanomoles per liter) cm ϫ 2.8 cm), reduced need for shaving, breast discomfort by the SHBG concentration (in nanomoles per liter), but this and gynecomastia, decreased spontaneous erections, re- is not a valid measure of free testosterone in older men. duced sexual desire and activity, reduced muscle bulk and If the initial total testosterone level is low (Ͻ300 ng/dL strength, inability to father children (due to low or zero or 10.4 nmol/L, or below the lower limit of normal for sperm counts), height loss, low bone mineral density, and healthy young men according to the reference range of the low-trauma fractures. Less specific findings include de- local laboratory), the measurement should be repeated, as creased energy, depressed mood, mild anemia, and dimin- 30% of men with an initially low level will have a normal ished physical or work performance.5,12 level upon repeat testing. Conversely, men with true defi- Measuring serum testosterone levels in the general pop- ciency states demonstrate persistently low testosterone lev- ulation to screen for androgen deficiency is not recom- els. In addition, LH and FSH levels should be measured, mended. However, in older men with clinical symptoms and because secondary hypogonadism is a common cause of signs consistent with androgen deficiency, the Endocrine androgen deficiency in older men (Figure). If total testos- Society recommends measuring a mid-morning total serum terone, LH, and FSH are low, measurement of other anterior testosterone level using a reliable assay (Figure). When pituitary hormones (eg, prolactin) and perhaps a magnetic measuring serum testosterone, several factors need to be resonance scan of the pituitary should be considered to considered. Peak testosterone levels occur between 7 and 10 exclude intra-and perisellar lesions. AM. Diet does not significantly affect the serum testosterone A recent cross-sectional analysis in 2 independent pop- level, but a high insulin level (eg, following a high carbo- ulations of healthy men (Belstress study, consisting of 2322 hydrate meal) can lower SHBG. Heavy alcohol consump- men aged 35 to 59 years; and Siblos study, consisting of 358 tion can decrease serum testosterone. On average, smokers men aged 25 to 45 years) showed androgen receptor poly- have total and free testosterone levels 5%-15% higher than glutamine tract polymorphism encoded by a CAG repeat of nonsmokers. SHBG levels are decreased in moderate obe- variable length in exon 1 of the AR gene might play an sity, hypothyroidism, glucocorticoid use, and nephrotic syn- important role in subject variability in serum (free) testos- drome, and increased in hyperthyroidism, anticonvulsant terone in healthy men because of differences in androgen use, cirrhosis, and other conditions (Table 2).5 sensitivity and feedback (LH) setpoint. CAG repeat length Kazi et al Testosterone Deficiency in Elderly Men 837

Figure T ϭ testosterone; bio T ϭ bioavailable testosterone; SFA ϭ seminal fluid analysis; 1° ϭ primary hypogonadism; 2° ϭ secondary hypogonadism. #In some laboratories, the lower limit of the normal testosterone range in healthy young men is approximately 300 ng/dL – this range may vary in different laboratories. One needs to use the lower limit of the range established in one’s reference laboratory. @In some reference laboratories, the lower limit of the normal free testosterone range in healthy young men is approximately 5 ng/dL using equilibrium dialysis or calculated from total testosterone and sex-hormone binding globulin (SHBG) – this may vary in different laboratories. One should use the lower limit of the range established in one’s reference laboratory. Modified from Bhasin S et al:5 J Clin Endocrinol Metab. 2006;91:1995-2010. was positively associated with serum total testosterone in some experts favor treating symptomatic men with testos- both study populations. Increased CAG repeat length was terone levels below 300 ng/dL, whereas others favor a associated with increased free testosterone levels.13 threshold of 200 ng/dL.5 In either case, for men with classical androgen deficiency signs and symptoms accompanied THERAPY by persistently low testosterone levels, testosterone replace- In older men, the testosterone level below which testoster- ment therapy is recommended to improve sexual function, one replacement therapy is recommended is controversial; bone mineral density, and sense of well-being, and to induce and maintain secondary sex characteristics.5,12,14,15 Testos- terone also is suggested for men with low testosterone levels Table 2 Alterations in Sex-Hormone Binding Globulin and erectile dysfunction (ED) after other causes of ED have (SHBG) Concentrations been excluded.12 Short-term adjunctive therapy with testos- Increased SHBG levels terone may be considered in HIV-infected men with low ● Aging testosterone levels and weight loss to improve muscle ● Hyperthyroidism strength and promote weight gain. Testosterone also should ● Liver cirrhosis be considered in men with low testosterone levels who are ● Use of estrogens receiving high dose glucocorticoid therapy in order to help ● Use of anticonvulsants preserve bone mineral density. ● HIV infection Improved sexual activity scores and increased duration Decreased SHBG levels ● and frequency of nocturnal erections result from effective Moderate obesity 5 ● Use of glucocorticoids, progestins, and androgenic steroids testosterone therapy in young hypogonadal men, but data ● Nephrotic syndrome are limited in older men, among whom the association ● Hypothyroidism between androgens and sexual function is more controver- 5,16 Modified from Bhasin S et al:5 J Clin Endocrinol Metab. sial. Dose-dependent increases in hemoglobin concen- 2006;91:1995-2010. tration and in bone mineral density are typically seen with testosterone therapy.17,18 The cardiovascular effects of tes- 838 The American Journal of Medicine, Vol 120, No 10, October 2007 tosterone replacement appear to be neutral or mildly bene- The hematocrit should be measured at baseline, at 3 ficial in young men;15 in older men, there is no convincing months, and annually thereafter. If the hematocrit exceeds evidence that testosterone therapy is either beneficial or 54%, testosterone therapy should be discontinued. Digital harmful to the cardiovascular system.19,20 Testosterone has rectal examination and PSA measurement should be per- minimal effect on serum lipids, whereas the effect on insulin formed before starting testosterone therapy. If the PSA sensitivity is controversial.21 increases above 4 ng/mL, or by more than 1.4 ng/mL within Hormone-dependent malignancies, such as prostate and 12 months of treatment, urological consultation should be breast cancer, may grow faster during testosterone therapy. obtained. Therefore, testosterone replacement is not recommended in Patients on testosterone replacement should be evalu- such patients. It also is not recommended in patients with a ated for adverse drug effects specific to each preparation. palpable prostate nodule, a prostate-specific antigen (PSA) These include excessive erythrocytosis and fluctuations Ͼ 3 ng/mL (pending further urological evaluation), hyper- in mood or libido (injectable testosterone); skin reactions Ͼ viscosity, erythrocytosis (hematocrit 50%), untreated ob- (patch); and alterations in taste and gum irritation (buccal structive sleep apnea, severe benign prostatic hyperplasia, testosterone).5 or decompensated heart failure. Additional factors that may influence the decision to initiate testosterone therapy in older men include functional status and the presence of SUMMARY cognitive dysfunction. For example, a man who is bedrid- Aging in healthy men is associated with a decrease in den is unlikely to benefit from testosterone administered to serum testosterone levels. Clinically significant androgen improve muscle strength. Importantly, the uncertainties deficiency in older men is most often related to pituitary about the benefits and risks of testosterone should be dis- or hypothalamic abnormalities rather than to primary cussed with older men on an individualized basis before testicular failure. The challenge in diagnosing androgen embarking on a course of testosterone therapy. deficiency in the elderly is to link signs and symptoms to The goal of testosterone therapy is to achieve a serum serum testosterone levels. Questionnaires for assessing testosterone level in the mid-normal range during treatment. symptoms of testosterone deficiency in the aging male Several formulations and therapeutic regimens are avail- have been developed, and these instruments also may be able, and selection should be based on patient preference, helpful in evaluating for depression. Measuring serum pharmacokinetics, and cost (Table 3). Testosterone enan- testosterone levels accurately and reliably is problematic, thate or cypionate is administered intramuscularly, weekly and measurement issues must be taken into consideration at a dose of 100 mg, or biweekly at 200 mg. Alternatively, when making a diagnosis of testosterone deficiency in an a 5- or 10-mg nongenital testosterone patch can be applied elderly man. Total serum testosterone should be mea- to the skin (away from pressure areas) each night. Newer sured in the morning in a man with signs and symptoms preparations include testosterone gel applied daily to non- consistent with androgen deficiency. If low, the measure- genital skin (5- to 10-g dose), and bioadhesive testosterone ment should be repeated to confirm the diagnosis, be- tablets (30 mg) applied to the to buccal mucosa every 12 cause there is a 30% “false positive” rate with initial hours. In some countries, oral testosterone undecanoate, screening. LH and FSH also should be measured to dis- injectable testosterone undecanoate, and testosterone pellets tinguish primary from secondary hypogonadism. are available.5 Testosterone doses are generally lower in Although data from large scale clinical trials in elderly older than in younger men because testosterone is metabo- lized more slowly in the elderly. men are lacking, testosterone replacement therapy is rec- After therapy is initiated, the patient should be re-eval- ommended in men with definite symptoms and signs of uated at 3 months and at least annually thereafter. Special androgen deficiency in conjunction with a persistently low attention should be directed to symptoms before and after serum testosterone level. Several testosterone preparations treatment to determine whether there has been a satisfactory are available, and selection of a therapeutic modality should response and to assess for adverse effects. A mid-morning be based on considerations of personal preference, side total serum testosterone level should be obtained, with a effect profile, pharmacokinetics, and cost. In addition, fac- target range of 350-700 ng/mL (12.3-24.5 nmol/L); for tors such as functional status and cognitive impairment may older men, a range of 400-500 ng/dL (14.0-17.5 nmol/L) is influence the decision to treat and the choice of therapy in suggested.5 For injectable testosterone, the serum level older men. Following initiation of treatment, re-evaluation should be measured between injections. For men treated for efficacy and side effects should be performed after 3 with a transdermal testosterone patch, the serum level months of therapy and at least annually thereafter. Finally, should be measured 3 to 12 hours after patch application. In given the anticipated increase in the number of older men patients receiving buccal testosterone tablets, the serum with androgen deficiency, there is a compelling need for level should be measured immediately before application of additional research, including large scale clinical trials, to a fresh system. Patients on testosterone gel may have levels determine the short- and long-term benefits and risks of checked anytime after at least 1 week of therapy.5 testosterone replacement therapy. aie lTsotrn ecec nEdryMen Elderly in Deficiency Testosterone al et Kazi

Table 3 Testosterone Formulations Formulation Replacement Dosage Approximate Cost Advantages Disadvantages T esters 100 mg/wk IM or 200 mg q 2 $70/month Extensive clinical experience IM injection, uncomfortable Fluctuations T enanthate or wk IM (200 mg per 2 Inexpensive in T levels and in libido, energy, cypionate weeks) mood Injectable long-acting T 1000 mg IM, then 1000 mg at Not available in the US Corrects symptoms of androgen deficiency IM injection of a large volume (4 mL), undecanoate in oil 6 wk, and 1000 mg every Requires less frequent administration more uncomfortable 12 wk Nongenital transdermal 5 mg, 1-2 patches every night $192/month (5 mg/d) Physiological circadian T levels, no injection, Skin irritation T patch applied to nonpressure Good adhesion, normal DHT levels areas (thigh, arm) Scrotal T patch 6 mg over 24 h applied daily $131 for 30 patches Symptoms of androgen deficiency are Scrotal skin needs to be shaved. High corrected DHT levels T gels 5-10 gm every AM $205-248/month Physiological T levels, Little skin irritation Potential transfer of T to women and (5 g/d) Dose flexibility, Musk odor children by contact (prevented by washing after 4-6 hr). Costly. High DHT levels Buccal T tablets 30 mg every 12 h $190/month Steady-state T levels, No hand washing Twice daily, gum irritation, Taste bioadhesive alteration, high DHT T pellets 4-6 200 mg pellets implanted $150 per 10 pellets Symptoms of androgen deficiency are Surgical incision needed and pellets may SC corrected extrude T ϭ testosterone; DHT ϭ dihydrotestosterone; IM ϭ intramuscular; SC ϭ subcutaneous. Modified from Bhasin S et al:5 J Clin Endocrinol Metab. 2006;91:1995-2010. 839 840 The American Journal of Medicine, Vol 120, No 10, October 2007

RESOURCES 8. Harman SM, Metter EJ, Tobin JD, et al. Longitudinal effects of aging on serum total and free testosterone levels in healthy men. http://www.hormone.org/public/factsheets.cfm Baltimore Longitudinal study of aging. J Clin Endocrinol Metab. http://www.endo-society.org/quickcontent/clinicalpractice/ 2001;86:724-731. clinical-guidelines/CG_Androgen.cfm 9. Vermeulen A, Verdonck L, Kaufman JM. A critical evaluation of simple methods for the estimation of free testosterone in serum. J Clin http://www.endo-society.org/news/endocrine_news/index. Endocrinol Metab. 1999;84:3666-3672. cfm 10. Morley JE, Charlton E, Patrick P, et al. Validation of a screening http://www.aace.com/pub/guidelines/ questionnaire for androgen deficiency in aging males. Metabolism. http://www.issam.ch/freetesto.htm 2000;49(9):1239-1242. 11. Yoshida NM, Kumano H, Kuboki T. Does the Aging Males’ Symp- toms scale assess major depressive disorder? A pilot study. Maturitas. Note added in proof: The CAG repeat polymorphism in 2006;53(2):171-175. exon 1 of the androgen receptor gene may play a major 12. Mikhail N. Does testosterone have a role in erectile function? Am J Med. 2006;119:373-382. future role in assessing the impact of testosterone therapy 13. Crabbe P, Bogaert V, de Bacquer D, et al. Part of the interindividual and serum testosterone levels. See Zitzmann M. Mecha- variation in serum testosterone levels in healthy men reflects differ- nisms of disease: pharmacogenetics of testosterone ther- ences in androgen sensitivity and feedback setpoint: contribution of the apy in hypogonadal men. Nat Clin Pract Urol. 2007 androgen receptor polyglutamine tract polymorphism. J Clin Endocri- Mar;4(3):161-6. nol Metab. 2007;92:3604-3610. 14. Petak SM, Nankin HR, Spark R, et al. American Association of Clinical Endocrinologists medical guidelines for clinical practice for the evaluation and treatment of hypogonadism in adult male patients— ACKNOWLEDGMENT 2002 update. Endocr Pract. 2002;8(6):439-456. We wish to thank Dr. William C. Nicholas for critical 15. Guay AT, Spark RF, Bansal S, et al. American Association of Clinical Endocrinologists medical guidelines for clinical practice for the eval- review of this manuscript. uation and treatment of male sexual dysfunction: a couple’s problem— 2003 update. Endocr Pract. 2003;9(1):77-95. References 16. Bolona ER, Uraga MV, Haddad RM, et al. Testosterone use in men 1. Liu PY, Iranmanesh A, Nehra AX, et al. Mechanisms of hypoandro- with sexual dysfunction: a systematic review and meta-analysis of genemia in healthy aging men. Endocrinol Metab Clin North Am. randomized placebo-controlled trials. Mayo Clin Proc. 2007;82(1):20- 2005;34:935-955. 28. 2. Kaufman JM, Vermeulen A. The decline of androgen levels in elderly 17. Isidori AM, Giannetta E, Greco EA, et al. Effects of testosterone on men and its clinical and therapeutic implication. Endocr Rev. 2005; body composition, bone metabolism and serum lipid profile in middle- 26(6):833-876. aged men: a meta-analysis. Clin Endocrinol. 2005;63:280-293. 3. Alesci S, Koch CA, Bornstein SR, Pacak K. Adrenal androgen regu- 18. Tracz MJ, Sideras K, Bolona ER, et al. Testosterone use in men and its lation and adrenopause. Endocr Regul. 2001;35(2):95-100. effects on bone health. A systematic review and metaanalysis of 4. Travison TG, Araujo AB, O’Donnel AB, et al. A population-level randomized placebo-controlled trials. J Clin Endocrinol Metab. 2006; decline in serum testosterone levels in American men. J Clin Endo- 91(6):2011-2016. crinol Metab 2007;92:196-202. 19. Shabsigh R, Katz M, Yan G, Makhsida N. Cardiovascular issues in 5. Bhasin S, Cunningham GR, Hayes FJ, et al. Testosterone therapy in hypogonadism and testosterone therapy. Am J Cardiol. 2005; adult men with androgen deficiency syndromes: an endocrine society 96(suppl):67M-72M. clinical practice guideline. J Clin Endocrinol Metab. 2006;91(6):1995- 20. Haddad RM, Kennedy CC, Caples SM, et al. Testosterone and car- 2010. diovascular risk in men: a systematic review and meta-analysis of 6. Liu PY, Swerdloff RS, Wang C. Relative testosterone deficiency in randomized placebo-controlled trials. Mayo Clin Proc. 2007;82(1):29- older men: clinical definition and presentation. Endocrinol Metab Clin 39. North Am. 2005;34:957-972. 21. Kapoor D, Goodwin E, Channer KS, Jones TH. Testosterone replace- 7. Bhasin S, Buckwalter JG. Testosterone supplementation in older men: ment therapy improves insulin resistance, glycaemic control, visceral a rational idea whose time has not yet come. J Androl. 2001;22(5): adiposity and hypercholesterolemia in hypogonadal men with type 2 718-731. diabetes. Eur J Endocrinol. 2006;154(6):899-906. The American Journal of Medicine (2007) 120, 841-847

OFFICE MANAGEMENT: GERIATRICS Michael W. Rich, MD, Speciality Editor Orthostatic Hypotension in the Elderly: Diagnosis and Treatment Vishal Gupta, MD, PhD, Lewis A. Lipsitz, MD Beth Israel Deaconess Medical Center, Hebrew SeniorLife, and Harvard Medical School, Boston, Mass.

ABSTRACT

Orthostatic hypotension is a common problem among elderly patients, associated with significant morbidity and mortality. While acute orthostatic hypotension is usually secondary to medication, fluid or blood loss, or adrenal insufficiency, chronic orthostatic hypotension is frequently due to altered blood pressure regulatory mechanisms and autonomic dysfunction. The diagnostic evaluation requires a comprehensive history including symptoms of autonomic nervous system dysfunction, careful blood pressure measurement at various times of the day and after meals or medications, and laboratory studies. Laboratory investigation and imaging studies should be based upon the initial findings with emphasis on excluding diagnoses of neurodegenerative diseases, amyloidosis, diabetes, anemia, and vitamin deficiency as the cause. Whereas asymptomatic patients usually need no treatment, those with symptoms often benefit from a stepped approach with initial nonpharmacological interventions, including avoidance of potentially hypotensive medications and use of physical counter maneuvers. If these measures prove inadequate and the patient remains persistently symptomatic, various pharmacotherapeutic agents can be added, including fludrocortisone, midodrine, and nonsteroidal anti-inflammatory drugs. The goals of treatment are to improve symptoms and to make the patient as ambulatory as possible rather then trying to achieve arbitrary blood pressure goals. With proper evaluation and management, the occurrence of adverse events, including falls, fracture, functional decline, and myocardial ischemia, can be significantly reduced. © 2007 Elsevier Inc. All rights reserved.

KEYWORDS: Aging; Blood pressure; Elderly; Ofﬁce practice; Orthostatic hypotension

In 1995, the American Academy of Neurology and the Joint EPIDEMIOLOGY Consensus Committee of the American Autonomic Society Previous studies have revealed an increased prevalence of defined orthostatic hypotension as a reduction in systolic orthostatic hypotension with age. In community dwelling blood pressure of at least 20 mm Hg or diastolic blood individuals Ͼ65 years of age, its prevalence is approxi- pressure of at least 10 mm Hg within 3 minutes of assuming Ͼ 1 mately 20%; in those 75 years of age it is as high as 30%. an erect posture. This definition does not account for a fall In frail elderly individuals living in nursing homes, the in blood pressure after 3 minutes or symptoms associated prevalence of orthostatic hypotension is even higher, up to with smaller decreases in blood pressure upon standing. 50% or more. An age-associated increase in supine blood Hence, the significance of any decrease in blood pressure pressure has been implicated as a major determinant of the upon standing should be evaluated according to its associ- development of orthostatic hypotension with aging.2 Ortho- ation with symptoms of dizziness, presyncope, syncope, or static hypotension also is associated with significant mor- falls. bidity at older age. It has been linked to falls, fractures, transient ischemic attacks, syncope, and myocardial infarction. In addition, elderly people with orthostatic hypoten- Requests for reprints should be addressed to Lewis A. Lipsitz, MD, sion are more likely to be physically frail and thus to have Hebrew SeniorLife, Institute for Aging Research, 1200 Centre Street, Boston, MA 02131. decreased functional capacity, a factor that is often over- E-mail address: [email protected] looked during the evaluation of older patients.

to sympathetic activation. Also, an age-related reduction Table 1 Age-Related Changes that Can Affect Normal Blood Pressure Regulation in parasympathetic tone results in less cardioacceleration during the vagal withdrawal that normally occurs with Decreased baroreflex sensitivity standing. Due to reductions in renin, angiotensin, and ␣ Decreased -1-adrenergic vasoconstrictor response to aldosterone with aging, and an elevation in natriuretic sympathetic stimuli peptides, the aged kidney loses some of its ability to Decreased parasympathetic activity Decreased renal salt and water conservation conserve salt and water during periods of fluid restriction Increased vascular stiffness or volume loss, leading to rapid dehydration. In addition, Reduced left ventricular diastolic filling the aged heart becomes stiff and non-compliant, resulting in impaired diastolic filling. This reduces stroke volume when preload is decreased due to standing or volume contraction. PATHOGENESIS Taken together, the reductions in baroreflex-mediated In healthy people, approximately 500 to 1000 milliliters cardioacceleration and vasoconstriction, renal salt and of blood is transferred below the diaphragm upon assum- water conservation, and cardiac filling greatly increase ing an erect posture.3 This leads to decreased venous the risk of hypotension in the elderly. Severe, symptom- return to the heart, reduced ventricular filling, and a atic orthostatic hypotension may develop in the face of transient decrease in cardiac output and blood pressure. any additional stress that lowers blood pressure or im- As a consequence, baroreceptors in the carotid arteries pairs the compensatory response, including certain med- and aorta are activated, resulting in increased sympa- ications, reduced intravascular volume, or other situa- thetic outflow and decreased parasympathetic outflow tions that reduce cardiac preload. from the central nervous system. This compensatory re- flex restores cardiac output and blood pressure by in- ETIOLOGY creasing heart rate and vascular resistance. Causes of orthostatic hypotension can be broadly divided Blood pressure varies directly with heart rate, stroke into acute and chronic (Figure 1). Acute orthostatic hypo- volume, and vascular resistance. Therefore, impairments tension most commonly develops over a relatively short in the response of any of these parameters during postural period of time and is more often symptomatic at the outset. change may result in orthostatic hypotension. As shown Generally, it results from acute conditions such as adrenal in Table 1, aging is associated with a decrease in barore- insufficiency, myocardial ischemia, medication administra- flex sensitivity, which manifests as a diminished heart tion, sepsis, or dehydration. In contrast, chronic orthostatic rate response and ␣-1-adrenegic vasoconstrictor response hypotension develops gradually over a prolonged period of

Figure 1 Etiology of orthostatic hypotension (OH). Gupta and Lipsitz Orthostatic Hypotension in the Elderly 843

Figure 2 Approach to the evaluation of orthostatic hypotension. BMP ϭ basic metabolic profile; CBC ϭ complete blood count; CT ϭ computerized tomography; H&P ϭ history and physical examination; MRI ϭ magnetic resonance imaging; RPR ϭ rapid plasma reagin. time and the patient is usually asymptomatic during the toris, and transient ischemic attacks. In elderly people, dis- initial period. Chronic orthostatic hypotension can be due to turbed speech, visual changes, falls, confusion, and either physiologic or pathologic causes. Physiologic causes impaired cognition are more commonly seen.4 However, the are those attributable to the age-associated changes in blood predictive value of these symptoms in the elderly is poor, pressure regulation described above, as well as the age- due to intake of multiple medications with various side related increase in systolic blood pressure, which further effects and overlapping symptoms arising from comorbid impairs adaptive responses to hypotensive stresses. These conditions. Therefore, careful blood pressure measurements physiologic changes predispose elderly people to symptom- are of critical importance, even in patients with atypical atic hypotension in the face of common everyday stresses, symptoms. such as posture change, meals, new medications, fluid restriction, or any acute illness. Pathologic causes of chronic orthostatic hypotension are secondary to central or periph- EVALUATION eral nervous system diseases that result in autonomic insuf- Our approach to the evaluation of orthostatic hypotension is ficiency (Figure 1). shown in Figure 2. Initial evaluation should include measuring blood pressure and heart rate after the patient has been quietly supine for at least 5 minutes and again after 1 CLINICAL FEATURES minute and 3 minutes of standing. Early morning measure- Orthostatic hypotension may be symptomatic or asymptom- ments, especially after a high carbohydrate meal, are useful atic. However, even in asymptomatic patients it remains a to identify postprandial hypotension. Although postprandial risk for future falls and syncope, and should therefore be hypotension may occur concomitantly with orthostatic hy- minimized as much as possible. Common symptoms at all potension, it is a distinct entity that often occurs while ages include dizziness, light headedness, weakness, syn- sitting after a meal, and may actually resolve upon standing cope, nausea, paracervical pain, low back pain, angina pec- up and walking.5 Detection of orthostatic hypotension may 844 The American Journal of Medicine, Vol 120, No 10, October 2007

Table 2 Additional Clinical Clues and Tests to Order Clinical Presentation Possible Etiology Test to Order Ecchymoses, purpura, macroglossia, numbness, paresthesias, Amyloidosis Rectal biopsy pseudohypertrophy of muscle Diarrhea, vomiting, burns, fever Volume depletion Electrolytes, BUN, Creatinine Gummas, unequal pupils (Argyll Robertson pupil) loss of position Tabes dorsalis RPR, VDRL and vibration senses, history of sexually transmitted disease Early satiety, postprandial fullness, constipation, incontinence, Diabetic neuropathy EKG for deep breath variability, GTT exercise intolerance Chest pain, palpitation, shortness of breath, pedal edema Cardiogenic causes EKG, echocardiogram Reduced sweating, incontinence, constipation, posture Multiple system Autonomic testing difﬁculties, tremors, rigidity atrophy Confusion, cerebellar symptoms, nystagmus, amnesia, Alcoholic neuropathy CBC, random alcohol level confabulation, history of alcohol abuse Smooth beefy red tongue, lemon pallor, recent loss of mental Pernicious anemia CBC, cobalamin level, folate level capacity, paresthesias, ataxia BUN ϭ blood urea nitrogen; CBC ϭ complete blood count; EKG ϭ electrocardiogram; GTT ϭ glucose tolerance test; RPR ϭ rapid plasma reagin; VDRL ϭ venereal disease research laboratory.

require multiple measurements on different days. This can clinical suspicion points towards central nervous system be accomplished with ambulatory blood pressure monitor- pathology. ing, or by loaning the patient an automatic blood pressure Autonomic function testing is helpful when the history monitor with instructions to maintain a diary with record- and physical examination are equivocal, to evaluate the ings of supine and standing blood pressure at different times extent of autonomic involvement, and to monitor the course of the day for several days. Measurements before breakfast, of an autonomic disorder and its response to therapy. Com- after medications, after meals, and before bed are most monly used bedside studies to assess autonomic function are useful. Furthermore, the heart rate response to postural heart rate variation in response to deep breathing (respira- change can provide important clues to the etiology. Minimal tory sinus arrhythmia) and blood pressure response to the cardio-acceleration (Ͻ10 beats per minute) on standing cold pressor test. Heart rate variation during deep breathing from a supine position in the presence of hypotension sug- assesses the function of parasympathetic (vagal) efferents to gests baroreflex impairment, whereas tachycardia (Ͼ20 the heart. Sinus arrhythmia is measured by electrocardiog- beats per minute) indicates volume depletion or orthostatic raphy with the patient lying supine during 1 minute of slow intolerance. Note, however, that lack of tachycardia also and deep breathing with 5 seconds inspiration and 7 seconds may occur in volume-depleted elderly patients due to expiration. In healthy elderly people, the ratio of longest baroreflex impairment. expiratory R-R interval to shortest inspiratory R-R interval Ͼ Once the diagnosis of orthostatic hypotension is estab- is 1.15. Potential confounders that may reduce heart rate lished, a detailed history should be obtained, focusing on variability include medications (beta-blockers, calcium channel blockers, anticholinergic agents), advanced age, the medications (both prescription and nonprescription), vol- patient’s position (sitting vs. supine), and hypocapnia. The ume losses (vomiting, diarrhea, fluid restriction), coexisting cold pressor test evaluates sympathetic innervation of the medical disorders, and autonomic dysfunction. A compre- vasculature. After immersion of one hand in ice cold water hensive physical examination should be performed, seeking at 4°C for 1 minute, a normal response is a systolic blood clinical clues to possible underlying physiological and pressure elevation Ն15 mm Hg and diastolic elevation Ն10 pathological disorders (Table 2). These include signs of mm Hg. Other tests that can be considered include plasma amyloidosis, malignancy, and heart failure. A neurological norepinephrine and vasopressin levels supine and upright to evaluation should include a mental status examination (to distinguish central from peripheral causes of autonomic identify neurodegenerative diseases such as Lewy Body failure. In central causes, supine norepinephrine is normal Dementia), motor testing (Parkinson’s disease or multiple but fails to increase with postural change, and vasopressin is strokes), sensory testing (peripheral neuropathy), and pupil- low. In peripheral causes, supine norepinephrine levels are lary size (Horner’s syndrome). Subsequent laboratory tests low and vasopressin is normal. However, in practice the should be obtained based on the results of these assess- high variability of these levels undermines their utility. ments. These may include hemoglobin and hematocrit levels to evaluate for anemia; blood electrolytes, urea nitrogen, and creatinine to assess for dehydration; a rapid plasma MANAGEMENT reagin (RPR) test for syphilis; and a glucose tolerance test Due to the presence of multiple co-morbid conditions and for diabetes. Brain imaging studies should be ordered if nonspecific signs and symptoms, treatment of orthostatic Gupta and Lipsitz Orthostatic Hypotension in the Elderly 845

straining, and prolonged standing, should be avoided, par- Table 3 Nonpharmacologic Treatment Options for Orthostatic Hypotension ticularly in hot weather. Dorsiflexion of the feet before assuming an upright posture may promote venous return to Withdraw offending medication (either substitution or the heart, accelerate the heart rate, and increase blood pres- discontinuation) sure. Squatting and stooping forward can result in an in- Rise slowly from supine to sitting to standing position crease in blood pressure. In patients who present with symp- Avoid straining, coughing, and prolonged standing in hot weather toms after prolonged standing, simply sitting down can Cross legs while standing often raise the blood pressure. Physical counter-maneuvers Squat, stooping forward like crossing one’s legs while standing and maintaining Raise head of bed 10 to 20 degrees muscle contraction for 30 seconds can increase systemic Small meals and coffee in the morning venous return, thereby causing increased cardiac output and Elastic waist high stocking Increase salt and water intake blood pressure. Waist high compression stockings and ab- Exercise, eg, swimming, recumbent biking, and rowing dominal binders may be helpful. In patients with autonomic failure and supine hypertension, raising the head of the bed by 10 to 20 degrees at night can reduce hypertension, prevent overnight volume loss, and help restore morning hypotension in the elderly is often challenging. Instead of blood pressure upon standing. Liberal intake of salt and aiming to achieve arbitrary blood pressure goals, the treat- water to achieve a 24-hour urine volume of 1.5 to 2 liters ment of orthostatic hypotension should be directed toward may attenuate fluid loss commonly seen in autonomic in- ameliorating symptoms, correcting any underlying cause, sufficiency. In elderly patients with orthostatic hypotension improving the patient’s functional status, and reducing the related to deconditioning, an exercise regimen comprising risk of complications. Broadly, interventions can be divided swimming, recumbent biking, or rowing might lead to dis- into nonpharmacological and pharmacological approaches. appearance of symptoms.

Nonpharmacological Interventions Pharmacological Interventions Generally it is best to start with nonpharmacological inter- Numerous pharmacological agents are available if the pa- ventions and, if this fails, then proceed to drug therapy tient remains symptomatic despite the above measures (Ta- (Table 3). The first management step involves removing any ble 4). One of the most potent agents is fludrocortisone, a medication that could precipitate orthostatic hypotension. synthetic mineralocorticoid, which has a principal mode of Common offending drugs include nitrates, tricyclic antide- action of reducing salt loss and expanding blood volume.7 pressants, neuroleptics, and alpha-blockers (often used for The initial dose is 0.1 mg per day with increments of 0.1 mg urinary frequency or retention). Orthostatic hypotension every week until there is development of trace pedal edema may develop when a patient begins taking an anti-hyperten- or the maximum dose of 1 mg per day is reached. Common sive medication, but it may improve with continued use.6 side effects include hypokalemia, supine hypertension, heart Therefore, it is imperative to start with a low dose and failure, and headache. Elderly patients should be monitored slowly titrate the dose upward. In patients with acute ortho- for fluid overload and hypokalemia. In patients taking static hypotension due to dehydration, fluid replacement higher doses, potassium supplements are usually required. therapy should be initiated. Patients who have had pro- If the patient remains symptomatic, midodrine, an alpha- longed bedrest or inactivity (eg, following hospitalization) agonist with selective vasopressor properties, is often effec- should be instructed to stand up gradually to mitigate ex- tive.8 The starting dose is 2.5 mg 3 times per day, and the cessive pooling of blood in the lower extremities. Activities dose should be titrated upwards in 2.5-mg increments at that decrease venous return to the heart, such as coughing, weekly intervals until a maximum of 10 mg 3 times per day

Table 4 Selected Pharmacologic Agents for Orthostatic Hypotension

Drug Dose Contraindication Common Side Effects Fludrocortisone Initial: 0.1 mg daily Hypersensitivity Supine hypertension, hypokalemia, HF, Max.: 1 mg daily headache Midodrine Initial: 2.5 mg tid Severe OHD, urinary retention, thyrotoxicosis, Supine hypertension, piloerection, Max.: 10 mg tid acute renal failure pruritus, paresthesia Ibuprofen 400-800 mg tid Hypersensitivity to NSAIDs, active bleeding, GI intolerance, bleeding, headache, impaired renal function dizziness, renal insufﬁciency Caffeine 100-250 mg daily Hypersensitivity GI irritation, insomnia, agitation, nervousness Erythropoietin 25-75 U/Kg tiw Uncontrolled hypertension Stroke, myocardial infarction, hypertension GI ϭ gastrointestinal; HF ϭ heart failure; NSAIDs ϭ non-steroidal anti-inﬂammatory drugs; OHD ϭ organic heart disease. 846 The American Journal of Medicine, Vol 120, No 10, October 2007

peripheral ␣-2-adrenergic agonist, clonidine may improve Table 5 Indications for Referral to a Specialist orthostatic hypotension in patients with central nervous sys- Indications for referral to a geriatrician tem causes of autonomic failure, in whom there is little or Multiple comorbid conditions no central sympathetic outﬂow, by promoting peripheral Failure of standard therapy to alleviate symptoms venoconstriction and thereby increasing venous return to the Complications, including recurrent falls, fracture, ␣ functional decline, ischemic events, decreased quality of heart. Yohimbine is a central -2-adrenergic antagonist that life can increase central sympathetic outﬂow in some patients Cognitive decline and confusion with residual sympathetic nervous system efferent output. Frail elderly patient Ͼ70 years old Lack of social support Indications for referral to a cardiologist Uncontrolled supine hypertension despite standard therapy REFERRAL TO A SPECIALIST Advanced coronary artery disease or severe ischemic Major indications for referral to a specialist are listed in symptoms Table 5. In brief, consultation with a geriatrician should Severe left ventricular diastolic or systolic dysfunction be sought for frail elderly patients, those with multiple (ejection fraction Ͻ 30%) comorbid conditions including cognitive decline, failure Recent onset of tachy-/bradyarrhythmia of standard therapy, any symptom-related complication, Indications for referral to a neurologist Specialized diagnostic testing for autonomic failure or lack of social support. In elderly patients requiring Chronic and progressive autonomic failure counseling and reinforcement, referral to a geriatrician

is achieved. For best results, the morning dose should be Table 6 Key Points in Office Management of Orthostatic Hypotension given early and the evening dose no later then 6 PM. Com- bination therapy of fludrocortisone and midodrine using ● Orthostatic hypotension is defined as a reduction in lower doses of both agents (due to synergistic effects) also systolic blood pressure of at least 20 mm Hg or diastolic is beneficial. Adverse effects include supine hypertension, blood pressure of at least 10 mm Hg within 3 minutes of piloerection, pruritus, and paresthesia. Midodrine is contra- assuming an erect posture. However, the significance of indicated in patients with coronary heart disease, heart fail- any decrease in blood pressure upon standing should be ure, urinary retention, thyrotoxicosis, or acute renal failure. evaluated in context with associated symptoms. ● Regardless of whether orthostatic hypotension is Midodrine should be used cautiously in elderly patients who symptomatic or asymptomatic, the elderly patient remains are taking medications that decrease heart rate, such as at significant risk for future falls, fractures, transient beta-blockers, calcium channel blockers, and cardiac ischemic attacks, and myocardial infarction. glycosides. ● Orthostatic hypotension can be acute or chronic. Acute Prostaglandin inhibitors, such as indomethacin and other causes include hypotensive medications, dehydration, and nonsteroidal anti-inflammatory drugs (NSAIDs), can block adrenal insufficiency. Chronic causes can be further sub- the vasodilating effects of prostaglandins and raise the divided into those related to aging or age-related blood blood pressure in some patients with orthostatic hypoten- pressure elevation (physiologic causes) and those due to sion.9 In elderly patients, indomethacin should be avoided central or peripheral autonomic nervous system diseases because of associated confusion, and all NSAIDs should be (pathologic causes). ● used with caution due to gastrointestinal and renal side The diagnostic evaluation of orthostatic hypotension should include a comprehensive history and physical effects. examination, careful blood pressure measurements, and The methylxanthine caffeine, administered in a dose of laboratory studies. 200 mg every morning as 2 cups of brewed coffee or by ● Goals of treatment in the elderly patient include tablet, may attenuate symptoms in some patients. Caffeine ameliorating symptoms, correcting any underlying cause, is an adenosine-receptor blocker that inhibits adenosine- improving the patient’s functional status, and reducing induced vasodilatation by blocking these receptors. To the risk of complications, rather than trying to attain an avoid tolerance and insomnia, caffeine should not be given arbitrary blood pressure goal. more then once in the morning. ● In most cases, treatment of orthostatic hypotension Erythropoietin has been shown to be effective in a sub- begins with nonpharmacological interventions, including group of patients with anemia and autonomic dysfunction.10 withdrawal of offending medications (when feasible), Although the exact mechanism of action is not known, its physical maneuvers, compression stockings, increased intake of salt and water, and regular exercise. effect is probably due to increased red cell mass and blood ● If nonpharmacological measures fail to improve symptoms, volume. The principal disadvantage of this drug is the par- pharmacologic agents should be initiated. Fludrocortisone, enteral route of administration. Serious side effects include midodrine, nonsteroidal anti-inflammatory drugs, caffeine, hypertension, stroke, and myocardial infarction. and erythropoietin have all been used to treat orthostatic Additional pharmacologic agents that may prove useful hypotension due to autonomic failure. in selected patients include clonidine and yohimbine. A Gupta and Lipsitz Orthostatic Hypotension in the Elderly 847 can often prove worthwhile when time constraints limit 3. Lipsitz LA. Orthostatic hypotension in the elderly. N Engl J Med. primary care physician effectiveness. Cardiology consul- 1989;321:952-957. tation is indicated for patients with uncontrolled supine 4. Rutan GH, Hermanson B, Bild DE, et al. Orthostatic hypotension in older adults. The Cardiovascular Health Study. Hypertension. 1992; hypertension despite standard therapy, advanced symp- 19:508-519. tomatic coronary artery disease, severe heart failure, and 5. Oberman AS, Harada RK, Gagnon MM, et al. Effects of postprandial in those with recent onset of tachy- or bradyarrhythmias. walking exercise on meal-related hypotension in frail elderly patients. Referral to a neurologist is suggested primarily for spe- Am J Cardiol. 1999;84:1130-1132. cialized autonomic testing in patients with an unclear 6. Masuo K, Mikami H, Ogihara T, Tuck ML. Changes in frequency of orthostatic hypotension in elderly hypertensive patients under medi- diagnosis or progressive autonomic failure. cations. Am J Hypertens. 1996;9:263-268. Key points in the office management of orthostatic hy- 7. Hussain RM, Mcintosh SJ, Lawson J, Kenny RA. Fludrocortisone in potension in the elderly are outlined in Table 6. the treatment of hypotensive disorders in the elderly. Heart. 1996;76: 507-509. 8. Low PA, Gilden JL, Freeman R, et al. Efficacy of midodrine vs References placebo in neurogenic orthostatic hypotension. A randomized, double- 1. Consensus statement on the definition of orthostatic hypotension, pure blind multicenter study. Midodrine Study Group. JAMA. 1997;277: autonomic failure, and multiple system atrophy. Neurology. 1996;46: 1046-1051. 1470. 9. Kochar MS, Itskovitz HD. Treatment of idiopathic orthostatic hypo- 2. Harris T, Lipsitz LA, Kleinman JC, Cornoni-Huntley J. Postural tension (Shy-Drager syndrome) with indomethacin. Lancet. 1978;1: change in blood pressure associated with age and systolic blood pres- 1011-1014. sure: the National Health and Nutrition Examination Survey II. J 10. Hoeldtke RD, Streetan DHP. Treatment of orthostatic hypotension Gerontol. 1991;46:M159-M163. with erythropoietin. N Engl J Med. 1993;329:611-615. The American Journal of Medicine (2007) 120, 848-850

DIAGNOSTIC DILEMMA: ENDOCRINE Charles M. Wiener, MD, Section Editor Fact or Factitious? Stefan Jenni,a Beat Gloor,b Christoph Stettler,a Emanuel R. Christa aDivision of Endocrinology, Diabetes, and Clinical Nutrition and bDepartment of Visceral and Transplant Surgery, University Hospital of Bern, Switzerland PRESENTATION ASSESSMENT Psychiatric symptoms may mask a physical problem. Fa- The patient appeared fully oriented, with a blood pressure of tigue, stress, disorientation, and lack of concentration lead 105/70 mm Hg and a regular pulse of 96 bpm. Her body to an initial diagnosis of depression for this patient, when, in mass index was 21.7 kg/m2 (height 164 cm, weight 58.4 actuality, her underlying condition was something quite kg). Pulmonary, cardiac, and abdominal examinations were different. Our patient, a 36-year-old nurse, was hospitalized unremarkable, and a fasting test was administered. After in a psychiatric clinic after attempting to commit suicide 13.25 hours of fasting, she exhibited classic signs of hypo- using regular and isophane insulin combined with benzodi- glycemia: neuroglycopenic symptoms (monotone re- azepines. She was found in a hypoglycemic coma with a sponses, troubles with coordination, inadequate response to capillary glucose level of Ͻ 3 mmol/L. Blood taken during a simple commands), a low plasma glucose concentration subsequent hypoglycemic attack in intensive care revealed a (1.13 mmol/L), and quick recovery (within 10 min) upon intravenous administration of glucose. She remained amne- plasma glucose level of 1.8 mmol/L; insulin 11.5 mU/mL; sic for the neuroglycopenic period. Insulin and C-peptide C-peptide 2.5 ng/mL; and a negative sulfonylurea screen. levels were inadequately high, ketone bodies were low, and Over 4 days, 1.750 kg of glucose were needed to maintain the sulfonylurea screen was again negative (Table 1). euglycemia. Abdominal computed tomography (CT) scans revealed 2 She had been in good health until 5 years ago, when she lesions in the pancreatic head (Figure 1). In addition, a began to suffer from fatigue and difficulty concentrating. hypodense structure of 2-cm diameter was found in the Then, about a year ago, she began to experience increas- body of the pancreas. Two of these lesions were confirmed ingly frequent bouts of confusion, fatigue, sweating, and by endoscopic sonography. At this point, the diagnosis of an trouble with orientation and concentration. At that time, a insulinoma in the head or body of the pancreas was pre- workup for convulsive disorder revealed a normal electro- sumed, and the patient was referred for surgery. encephalogram and a normal magnetic resonance (MRI) Intraoperatively, sonography of the pancreatic surface scan of the brain. Since she also felt stressed and over- confirmed a non-pathologic cyst in the left part of the body whelmed in her job, her general practitioner began treating of the pancreas. An abnormal vessel in the corpus (which her for depression. correlated to one of the pancreatic head lesions seen by CT) After the suicide attempt and recovery from coma, she also was identified. Finally, intraoperative sonography and began psychiatric rehabilitation. She was diagnosed with a palpation confirmed the remaining suspect head lesion, severe depressive disorder and started taking venlaflaxine. which had a diameter of 1.2 cm and was only visible in the She nonetheless continued to suffer from recurring attacks CT scan. of fatigue, odd behavior, retrograde amnesia, and hypoglycemia. She was eating frequently, even during the night, but assiduously denied further use of insulin or oral antidiabetic DIAGNOSIS agents. As her symptoms disappeared after eating, she was The patient’s hyperinsulinemic hypoglycemia and reactive referred for endocrine evaluation. depression were caused by a single insulinoma in the head of the pancreas. Insulinomas, which are insulin-producing neuroendocrine tumors of the pancreas, occur in about four Requests for reprints should be addressed to Charles M. Wiener, MD, 1 Johns Hopkins Department of Medicine, 1830 East Monument Street, in 1,000,000 persons. Although most insulinomas are be- Room 9030, Baltimore, MD 21205. nign and single, 10% are multiple; these are often associated E-mail address: [email protected] with multiple endocrine neoplasia 1 syndrome.1

Table 1 Fasting Test

Duration of fasting (hours) 11.5 13.25 Glucose (mmol/l) 1.95 1.13 Insulin (mU/L) 5 6.8 C-peptide (ng/mL) 1.6 2.9 ␥-hydroxy-butyrate (␮mol/L) 44 Sulfonylurea screen (HPLC) negative Neuroglycopenic symptoms ϩϩ Plasma glucose: hexokinase method; insulin: microparticle enzyme immunoassay (Abbott Axsym) (normal values Ͻ 3 mU/mL in hypoglycemia); C-peptide chemiluminescence: DPC Immulite One (normal values Ͻ 0.6 ng/mL in hypoglycemia); sulfonylurea screen: HPLC mass spectroscopy.

During the evaluation of hypoglycemic patients, factitious (self-induced) hypoglycemia must always be considered in the differential diagnosis, as this psychiatric illness often mimics the symptoms of insulinoma.2 Factitious hypoglycemia is particularly likely if psychiatric problems or easy access to glucose-lowering drugs are present. Furthermore, since many patients with factitious disease are healthcare professionals, and most are female, our patient’s symptoms and circumstances initially suggested factitious disease. However, the clear chro- nology of events in the patient’s history (ie, appearance of symptoms of hypoglycemia leading to reactive depression and culminating in a suicide attempt with insulin), made factitious hypoglycemia unlikely. The history was conﬁrmed by the results of the fasting test3 showing neuroglycopenic symptoms in the presence of low glucose concentrations and inadequately raised insulin and C-peptide levels. Exogenous insulin administration was excluded due to the C-peptide concentrations. Furthermore, repeated screenings for sulfonylureas by gas chromatography were negative. (Although it was recently re- Figure 2 Histological detail of the pancreas. A: Hematoxylin & ported that surreptitious use of glinide compounds can cause Eosin staining shows cellular proliferations with a trabecularly hypoglycemia,4 we were prevented from formally excluding oriented epithelial pattern, with only small cellular or nuclear glinide abuse by the unavailability of suitable assays.) atypia, and a solid growth pattern. No inﬁltration, destructive growth, or angioinvasion was found, and there was no evidence of congophilic deposits. B: Immunohistochemical staining with insulin.

In contrast to the common observation that factitious hypoglycemia often mimics insulinoma,2 our case docu- ments that insulinoma can mimic factitious hypoglycemia. Therefore, a standardized workup is warranted for the patient presenting with hypoglycemia, as a differential diagnosis of factitious disease cannot be assumed even in the case of a healthcare professional with a documented history of insulin abuse and psychiatric illness. Paradoxically, our patient suffered from such severe depression that she at- tempted suicide with exactly that substance that was responsible for her underlying condition.

Figure 1 In computed tomography of the pancreas, uptake of MANAGEMENT contrast medium was noted in the early arterial phase, as shown; it The therapeutic strategy for insulinoma consists of resection was hypodense in the other phases (not shown). with minimal damage to the unaffected pancreas. Therefore, 850 The American Journal of Medicine, Vol 120, No 10, October 2007 the preferred procedure is enucleation, but if the tumor lies been preserved. However, due to a mild exocrine insufﬁ- in the cauda, a resection of the pancreatic tail can be per- ciency, pancreatin substitution is necessary. She is success- formed. Because the patient’s tumor was in close proximity fully integrated back into her professional and social life. to the main pancreatic duct and the ductus hepato-chole- dochus, a simple enucleation appeared to be hazardous in References this case, and instead a pylorus-preserving duodeno-pancre- 1. Service FJ, Dale AJ, Elveback LR, Jiang NS. Insulinoma: clinical and atectomy was performed.5 Microscopic and immunohisto- diagnostic features of 60 consecutive cases. Mayo Clin Proc 1976; chemical staining of the resected tumor (Figure 2) con- 51(7):417-29. 2. Ziegler O, Gross P, Kolopp M, et al. Factitious hypoglycemia mimick- ﬁrmed the diagnosis of insulinoma. Thereafter, her blood ing insulinoma. Diabetes Care 1987;10(3):377-8. glucose became normal. 3. Wiesli P, Brandle M, Zapf J, et al.. Assessment of hyperinsulinaemia at The postoperative course was uneventful, and the patient the termination of the prolonged fast. Clin Chim Acta 2004;342(1-2): was transferred to the psychiatric clinic on postoperative 227-31. 4. Hirshberg B, Skarulis MC, Pucino F, et al. Repaglinide-induced facti- day 11. She left the clinic 4 months after her initial admis- tious hypoglycemia. J Clin Endocrinol Metab 2001;86(2):475-7. sion. During nine months of follow-up care, she has always 5. Grant CS. Insulinoma. Best Pract Res Clin Gastroenterol 2005;19(5): been free of hypoglycemia, and her endocrine function has 783-98. The American Journal of Medicine (2007) 120, 851-853

DIAGNOSTIC DILEMMA: CARDIOLOGY Charles M. Wiener, MD, Section Editor Over the Speed Limit Melissa R. Robinson, Ryan A. Brown, Uma Srivatsa, Ezra A. Amsterdam Division of Cardiovascular Medicine, Department of Internal Medicine, University of California, Davis. Ischemic ST segment depression occurs frequently during adenosine, which also normalized his sinus rhythm and supraventricular tachycardia (SVT), whereas elevation of blood pressure and greatly diminished the ST deviation traditional cardiac injury markers in this setting has rarely (Figure 2). The differential diagnosis included atrioventric- been reported. Here, we describe a case of a young man ular nodal reentry tachycardia, atrial flutter with 1:1 atrio- with both ischemic electrocardiographic alterations and ventricular conduction, and orthodromic atrioventricular markedly elevated troponin associated with rapid SVT and reciprocating tachycardia with a concealed accessory path- angiographically normal coronary arteries. way. Because of concern that myocardial ischemia might ensue, the patient was admitted for observation and further evaluation. PRESENTATION A 29-year-old man arrived at the emergency department with palpitations that had begun 90 minutes earlier. The DIAGNOSIS palpitations occurred at rest and were associated with short- Although initially normal, the patient’s serial cardiac injury ness of breath, left arm heaviness, and nausea, but no chest markers rose markedly: peak troponin I was 21.76 ng/dL discomfort. There was no antecedent illness, and the patient (normal, Ͻ0.06), creatine kinase was 1486 mcg/L, and the denied use of recreational drugs or over-the-counter medi- CK-MB fraction was 203.9 mcg/L (normal, 8 mcg/L). An cations. His past medical history included similar episodes ECG showed normal ventricular systolic and diastolic func- that had occurred 1–2 times per month since age 12. These tion, normal chamber dimensions, and no significant valvu- episodes were self-terminating, and evaluation had been lar disease. Cardiac catheterization revealed only minimal limited to electrocardiograms (ECGs), which were normal. luminal irregularities of the coronary arteries. Electrophys- He had no personal history of syncope or seizures and no iology study revealed a concealed left lateral accessory family history of arrhythmia, early coronary artery disease pathway, indicating a diagnosis of orthodromic atrioventric- (CAD), or sudden death. The patient’s only known cardiac ular reciprocating tachycardia. The accessory pathway was risk factor was cigarette smoking. successfully ablated. Data regarding ischemic ECG changes during SVT are ASSESSMENT limited. In a study of 100 middle-aged SVT patients, 89% Ն On physical examination, the patient was pale, diaphoretic, had ST segment depression of 1.0 mm in at least 1 ECG 1 and tachypneic. His blood pressure was 86/44 mm Hg, and lead. The mean ST depression was 2.2 mm per lead, and his pulse was Ͼ200 bpm. Except for an extremely rapid, the magnitude of depression and number of leads with this regular tachycardia, the cardiovascular examination was abnormality were directly related to heart rate. Greater ST unremarkable. The ECG revealed a regular, narrow QRS depression occurred with atrioventricular reciprocating tachycardia at 250 bpm, diffuse ST depression most marked tachycardia than with atrioventricular nodal reentry tachy- Ն in the precordial leads, and right axis deviation (Figure 1). cardia, but even marked ST depression ( 4 mm) did not The patient vomited while in the emergency department correlate with a history of CAD. In fact, in approximately and continued to complain of left arm discomfort. His 50% of the patients, repolarization changes occurred at the symptoms were relieved by intravenous administration of onset of atrioventricular nodal reentry tachycardia during electrophysiology study, and the overwhelming majority (93%) of these patients had no CAD.2 Requests for reprints should be addressed to Ezra A. Amsterdam, MD, Division of Cardiovascular Medicine, Department of Internal Medicine, Few studies have assessed the significance of ST depres- 3,4 4860 Y Street, Suite 2820, Sacramento, CA 95817. sion during SVT in relation to the presence of CAD. E-mail address: [email protected]. However, in a comparison of 39 SVT patients with no prior

Figure 1 Initial ECG showing narrow-complex tachycardia with ischemic-type ST depression. diagnosis of CAD, 21 had Ͼ1.0 mm ST depression during renal insufficiency.8 Remarkably, elevation of injury the arrhythmia, and 18 had no ST changes.5 There were no markers related to SVT has been described in only a few significant differences in age, cardiac risk factors, heart rate, case reports, and increases in troponin I were generally or chest pain during SVT episodes in the 2 groups. Cardiac modest.9,10,11,12 Although our patient was young and stress tests were normal in all 18 who did not exhibit ST exhibited no evidence of structural heart disease, he had depression during SVT, whereas they were “positive” in 7 both marked ST depression and injury marker elevation. of the 21 who did. Coronary angiography confirmed the His episode was prolonged, lasting Ͼ90 minutes, and was presence of CAD (Ն70% stenosis) in all 7 of these patients. associated with hypotension, although the predictive Thus, ST deviation during SVT had positive and negative value of these factors is not established. predictive values of 33 and 100%, respectively, for the Recent studies indicate that most patients who have isch- presence of CAD. emic-type ECG changes and positive markers during SVT ST segment depression has been controversially pro- do not have significant CAD, even in older age groups. A posed as a potential discriminator between atrioventricular potential mechanism for this phenomenon is an SVT-in- reciprocating tachycardia and atrioventricular nodal reentry duced mismatch between myocardial oxygen supply and tachycardia.6,7 It should be noted that this proposal is based demand. Reduced diastolic time would decrease the myo- on studies that excluded patients with a wide QRS complex cardial oxygen supply, and this effect would be com- during SVT, which would predictably lead to repolarization pounded by the increased myocardial oxygen demand re- abnormalities. These findings are also not applicable to lated to tachycardia, which can result in ischemia/injury and patients with atrial fibrillation, who are often older and more troponin release. In fact, the cardiomyopathy of chronic likely to have CAD. tachycardia may occur by a similar mechanism.13 A second Elevation of myocardial injury markers is pivotal in the potential mechanism, microvascular disease, was not spe- diagnosis of acute coronary syndromes but might also occur cifically excluded in this patient, but the risk factors for in numerous conditions unassociated with CAD, including epicardial CAD and for microvascular disease are similar, myocarditis, pericarditis, heart failure, myocardial toxic and current treatment of the latter is largely limited to drugs, shock, pulmonary embolism, strenuous exercise, and addressing these underlying risk factors.

Figure 2 ECG after adenosine. Robinson et al Over the Speed Limit 853

epicardial CAD. As our case illustrates, even marked ST depression and considerable elevation of troponin do not necessarily indicate epicardial CAD. Clinical risk stratiﬁca- tion followed by stress testing in medium- and high-risk patients is a prudent approach to this not uncommon clinical problem. After the successful ablation of the concealed left lateral accessory pathway and quitting smoking, our patient had no further palpitations or arm heaviness over the next 11 months.

References 1. Kim YN, Sousa J, El-Atass R, et al. Magnitude of ST segment depression during paroxysmal supraventricular tachycardia. Am Heart J. 1991;122(5):1486-1487. 2. Lin YJ, Tai CT, Chiang CE, et al. Mechanism of repolarization change during Initiation of supraventricular tachycardia. J Cardiovasc Elec- trophysiol. 2004;15:1233-1237. 3. Petsas AA, Anastassiades LC, Antonopoulos AG. Exercise testing for Figure 3 Algorithm for ischemia evaluation in patients with assessment of the significance of ST segment depression observed during episodes of paroxysmal supraventricular tachycardia. Eur SVT with ST depression. *High-risk patients or those with high- Heart J. 1990;11:974-979. risk features such as syncope might warrant ischemia evaluation. 4. Imrie JR, Yee R, Klein GJ, Sharma AD. Incidence and clinical sig- **Patients are considered high-risk if they have known cardiovas- nificance of ST segment depression in supraventricular tachycardia. cular disease (coronary, peripheral or cerebrovascular disease), Can J Cardiol. 1990;6(8):323-326. diabetes or 2 or more of the following: hypertension, hyperlipid- 5. Güleç S, Ertaþ F, Karaou˘z R, et al. Value of ST-segment depression emia, tobacco use, family history of premature coronary artery during paroxysmal supraventricular tachycardia in the diagnosis of disease or age Ͼ45 years for men and Ͼ55 years for women. coronary artery disease. Am J Card. 1999;83:458-460. 6. Arya A, Kottkamp H, Piorkowski C, et al. Differentiating atrioventricular nodal reentrant tachycardia from tachycardia via concealed Given the prevalence and poor predictive value of isch- accessory pathway. Am J Cardiol. 2005;95(7):875-878. emic-type ST changes in SVT, over-interpretation of such 7. Jaeggi ET, Gilljam T, Bauersfeld U, et al. Electrocardiographic dif- changes would result in invasive evaluation in many pa- ferentiation of typical atrioventricular node reentrant tachycardia from atrioventricular reciprocating tachycardia mediated by concealed ac- tients without CAD. A more practical diagnostic approach is cessory pathway in children. Am J Cardiol. 2003;91:1084-1089. suggested in Figure 3. After appropriate therapy for arrhyth- 8. Jeremias A, Gibson CM. Narrative review: Alternative causes for mia, the decision to measure cardiac injury markers should elevated cardiac troponin levels when acute coronary syndromes are be made based on the clinician’s suspicion of ischemia. It excluded. Ann Intern Med. 2005;142:786-791. should be noted that there is little data concerning patients 9. Redfearn DP, Ratib K, Marshall HJ, Griffith MJ. Supraventricular tachycardia promotes release of troponin I in patients with normal with SVT and chest pain but without ischemic ECG coronary arteries. Intl J Cardiol. 2005;102:521-522 changes. Furthermore, elevated troponin, even in associa- 10. Luna C, Adie MA, Tessler I, Acherman R. Troponin I elevation after tion with significant ST depression, must be interpreted in supraventricular tachycardia in a child with hypertrophic cardiomyop- relation to the patient’s pre-test probability of CAD. athy. Pediatr Cardiol. 2001;22:147-149. 11. Bakshi TK, Choo MKF, Edwards CC, et al. Intern Med J. 2002;32: 520-525. MANAGEMENT 12. Yeo KK, Cruz L, Hong R. Tachycardia-induced elevations in cardiac The benign nature of ST depression in the setting of SVT is troponin in the absence of coronary artery disease. Hawaii Med J. 2006;65(3):86-87. well established. The addition of highly sensitive testing of 13. Spinale FG, Hendreick DA, Crawford FA, et al. Chronic supraven- cardiac injury markers to the initial workup of these pa- tricular tachycardia causes ventricular dysfunction and subendocardial tients, however, can lead to the false diagnosis of significant injury in swine. Am J Physiol. 1991;28:H218-H229. The American Journal of Medicine (2007) 120, 854-856

IMAGES IN DERMATOLOGY Parwathi “Uma” Paniker, MD, Section Editor Board Stiff Peter Mattei, Julie Templet, Carrie Cusack Drexel University College of Medicine, Department of Dermatology, Philadelphia, Penn. The skin is the largest organ in the human body, both in pattern of mixed glomerular and tubular proteinuria consis- terms of surface area and mass, and it performs a multitude tent with diabetic renal disease. of protective, regulatory, and sensory functions. However, it also is subject to a multitude of primary and secondary disorders, many of which have poorly understood etiologies. We DIAGNOSIS describe one such case here. The extensive interstitial mucin deposition in the patient’s skin indicated that her symptoms were caused by one of the cutaneous mucinoses, a group of connective-tissue disor- PRESENTATION ders characterized by the accumulation of mucin in the skin. A 25-year-old African American female came to our clinic The members of this large, heterogeneous group of disor- with a 1-month history of pruritus and tenderness of her ders have a number of different morphologic presentations posterior neck, which she attributed to a “cyst” in that and are easily misdiagnosed. location. She denied systemic symptoms. Her past medical In the present case, the diffuse, woodlike thickening of history was significant for insulin-dependent diabetes mel- the dermis indicated scleredema, an uncommon and ill- litus and a soft-tissue infection of the leg 1 year prior to understood condition. Histologically, scleredema appears as presentation, and she was on insulin, lisinopril, and lova- unaltered epidermis with large, widely spaced, collagen statin. Her family history included insulin-dependent diabe- bundles. With use of special stains or electron microscopy, tes mellitus and eczema. the spaces can be properly visualized as mucin deposits. In the vast majority of cases, the pathology of this uncommon ASSESSMENT disorder remains limited to the skin, although visceral involvement, including cardiomyopathy, has been described Physical examination revealed a firm, nonpitting, hyperpig- where mucin can be seen in the cardiac muscle.1-2 A variety mented plaque with indistinct borders on her posterior neck of clinical situations are associated with scleredema, but beginning at her hairline and extending down the neck and most fall within 1 of 3 subtypes. Each subtype has its own upper back (Figure 1). The remainder of her back was supple, as were the upper extremities. Biopsy of the in- constellation of associated factors and course of illness. volved area revealed a thickened dermis with separation of The most frequently encountered scleredema subtype, collagen bundles. Staining with alcian blue (Figure 2) re- which occurs in non-diabetic young adults, is an acute condi- 3 vealed extensive interstitial mucin deposition throughout tion that may follow a febrile illness (usually streptococcal). the reticular dermis. The skin of the cervical and facial area quickly becomes Her serum glucose was 148 mg/dL, and her total serum hardened and taut, which can result in difficulty opening the protein was 8.5 g/dL (normal 6.0-8.3). Serum protein elec- mouth and swallowing. This form of the disease has a rapid trophoresis revealed no monoclonal gammopathy. A chronic onset and tends to resolve quickly (within months of on- inflammatory response was suggested by an increase in poly- set).3 The second scleredema subtype resembles the first clonal IgG levels (1.8 g/dL; normal 0.6-1.6). The patient also subtype in its symptoms but has a more gradual onset. These had elevated urine protein, and electrophoresis revealed a patients have persistent disease and associated monoclonal gammopathy (most commonly IgG) or multiple myeloma. The gammopathy is commonly undetectable when symp- Requests for reprints should be addressed to Julie T. Templet, MD, toms first arise; in fact, it can be clinically silent for over a Drexel University College of Medicine, Department of Dermatology, 219 N. Broad Street, 4th Floor, Philadelphia, PA 19107. decade after cutaneous presentation, thus requiring periodic E-mail address: [email protected] monitoring.4

The remaining scleredema subtype, known as scleredema diabeticorum, occurs in a distinct milieu. The archetypal patient afflicted by this form of scleredema is the obese male over 40 years of age with uncontrolled insulin- dependent diabetes.5 The posterior neck and back harden, and overlying erythema is sometimes present. The cutaneous changes in scleredema diabeticorum are recalcitrant to treatment, and the course of the disease is unaffected by glycemic management. In the present case, our patient’s condition was best categorized as scleredema diabeticorum, although her youth was atypical for this subtype. Other cases of scleredema that are not readily classified into 1 of the 3 subtype categories also have been documented. Some associations with underlying malignancy, including adenocarcinoma of the gallbladder and insulinoma, have been reported.6-7 Recently, a case of scleredema Figure 2 Skin biopsy stained with Alcian blue magnified 40 was reported in a patient undergoing infliximab therapy. When times shows thickened collagen with mucin deposition. the patient was taken off infliximab, the symptoms resolved, only to return when infliximab therapy was reinitiated.8 increased procollagen synthesis in fibroblasts of affected The differential diagnosis for scleredema includes sclero- skin has been demonstrated.11 Furthermore, the serum of derma, cellulitis, scleromyxedema, and other cutaneous muci- patients with gammopathy-associated scleredema stimulates noses. Scleroderma also causes hardened skin, but unlike normal fibroblasts through an unknown mediator.12 scleredema, it can afflict the extremities and is associated with Raynaud’s phenomenon, telangiectases, and antinu- clear antibodies. Because scleredema associated with dia- MANAGEMENT betes results in erythema, patients may be misdiagnosed Because no standard therapy currently exists for sclere- with cellulitis. Scleromyxedema, a generalized primary cu- dema, treatment for any associated underlying disease (eg, taneous mucinosis that also has associated monoclonal gam- diabetes, malignancy) is the primary goal. However, subse- mopathy, can be differentiated from scleredema by the dif- quent improvement is rare, except in cases of scleredema fuse symmetric eruption of waxy papules. Other cutaneous associated with multiple myeloma in which improvement mucinoses, including reticular erythematous, dysthyroid- after chemotherapy might be seen.4 Some anecdotal suc- otic, and follicular mucinoses, can usually be distinguished cesses have resulted from the many treatments that have from scleredema on a clinical basis. been investigated; however, spontaneous remission also can The precipitant for the aberrant tissue structure of scle- occur. Therapy with psoralen plus UVA (PUVA), oral and redema remains unknown. Proposed theories include sensi- topical corticosteroids, and pulsed-dose cyclosporine may tization to collagen by an infectious agent, autoimmunity, be effective for mildly or moderately severe cases of and non-enzymatic glycosylation of collagen rendering it scleredema. The self-limited cases of the acute, rapid-onset resistant to collagenase activity.5,9 Alternatively, Koga has scleredema subtype that affects young adults often do not suggested that microvascular damage and hypoxia stimulate require therapy, but a trial of topical steroids may be ben- mucin and collagen synthesis.10 Regardless of the etiology, eficial. More aggressive treatment with thalidomide and electron beam radiation should be reserved for those who suffer severe, physically disabling symptoms, such as limited mobility and difficulty swallowing, or for those with visceral involvement. Our patient was initially treated with topical clobetasol ointment 0.05% twice a day to the affected area. A month later, she began treatment with triamcinolone injections (20 mg/cc) into the neck plaque. Since then, she has reported some slight softening of the skin on her neck.

References 1. Erlichman M, Glaser J. Buschke’s scleredema with right-sided heart failure. Echocardiographic and clinical observations. Cardiology. 1983; 70:344-348. 2. Rimon D, Lurie M, Storch S, et al. Cardiomyopathy and multiple myeloma. Complications of scleredema adultorum. Arch Intern Med. Figure 1 Patient’s neck at presentation. 1988;148:551-553. 856 The American Journal of Medicine, Vol 120, No 10, October 2007

3. Venencie PY, Powell FC, Su WP, Perry HO. Scleredema: a review of 9. Vishwanath V, Frank KE, Elmets CA, Dauchot PJ, Monnier VM. thirty-three cases. J Am Acad Dermatol. 1984;11:128-134. Glycation of skin collagen in type I diabetes mellitus. Correlation with 4. Dziadzio M, Anastassiades CP, Hawkins PN, et al. From scleredema long-term complications. Diabetes. 1986;35:916-921. to AL amyloidosis: disease progression or coincidence? Review of the 10. Koga N. Effects of low-density lipoprotein apheresis on coronary and literature. Clin Rheumatol. 2006;25:3-15. carotid atherosclerosis and diabetic scleredema in patients with severe 5. Graff R. Scleredema adultorum. Arch Dermatol. 1968;98:319-320. hypercholesterolemia. Ther Apher. 2001;5:244-251. 6. Manchanda Y, Das S, Sharma VK, Srivastava DN. Scleredema asso- 11. Varga J, Gotta S, Li L, Sollberg S, Di Leonardo M. Scleredema ciated with carcinoma of the gall bladder. Br J Dermatol. 2005;152: adultorum: case report and demonstration of abnormal expression of 1373-1374. extracellular matrix genes in skin fibroblasts in vivo and in vitro. Br J 7. Matsunaga J, Hara M, Tagami H. Scleredema of Buschke associated Dermatol. 1995;132:992-999. with malignant insulinoma. Br J Dermatol. 1992;126:527-528. 12. Ohta A, Uitto J, Oikarinen AI, et al. Paraproteinemia in patients with 8. Ranganathan P. Infliximab-induced scleredema in a patient with rheu- scleredema. Clinical findings and serum effects on skin fibroblasts in matoid arthritis. J Clin Rheumatol. 2005;11:319-322. vitro. J Am Acad Dermatol. 1987;16(1 Pt 1):96-107. The American Journal of Medicine (2007) 120, 857-859

ECG IMAGE OF THE MONTH Julia H. Indik, MD, PhD, Section Editor A Heart Aflutter Ahmad Khraisat, Sarabjeet Singh, Rohit Arora, Eshraq Al-Jaghbeer Section of Cardiology, Department of Medicine, The Chicago Medical School. PRESENTATION the right ventricular outflow tract just inferior to the pul- 1 Most ventricular tachycardias encountered in clinical prac- monic valve. Two phenotypic forms of right ventricular tice occur in patients who have structural heart disease. In outflow tract tachycardia are known to occur, both of which the present case, a 27-year-old female attorney presented occurred in our patient. These forms are nonsustained, re- with palpitations of 1-day duration accompanied by dizzi- petitive, monomorphic ventricular tachycardia2 (Figure 3) ness, shortness of breath, and chest tightness. She had ex- and paroxysmal, exercise-induced, sustained ventricular perienced similar, but briefer, episodes in the previous few tachycardia (Figure 1). Both phenotypes are terminated by months, without fainting. Her past medical history was the administration of adenosine.3 unremarkable, and she was not on any medications. Symptoms of right ventricular outflow tract tachycardia typically arise in the third to fifth decade of life and are usually precipitated by exercise or emotional stress. Most ASSESSMENT patients present with palpitations (80%) or presyncope Lying supine in bed, the patient appeared anxious and had (50%), which are more common than frank syncope. Base- a blood pressure of 115/67 mm Hg and a pulse of 89 bpm. line ECGs recorded during sinus rhythm in patients with The physical examination and laboratory workup were oth- right ventricular outflow tract tachycardia can help distin- erwise unremarkable. guish it from the more serious condition of arrhythmogenic An initial electrocardiogram (ECG) showed a wide- right ventricular cardiomyopathy/dysplasia. The ECG is QRS-complex tachycardia at a rate of 183 bpm with a QRS more often abnormal in the latter condition, most commonly width of approximately 140 msec (Figure 1). The baseline showing persistently inverted T waves in the right precor- ECG was otherwise normal (Figure 2). An echocardiogram dial leads (V1, V2, and V3).2 with attention to the right ventricle was within normal In arrhythmogenic right ventricular cardiomyopathy/dys- limits, and an exercise stress test provoked the same mono- plasia, echocardiography typically shows right ventricular morphic tachycardia observed previously. dilatation with regional wall motion abnormalities.4 Cardiac magnetic resonance imaging (MRI) performed in special- DIAGNOSIS ized centers can detect fatty replacement of the right ventricular myocardium, but this parameter is the least reliable The patient’s monomorphic ventricular tachycardia was and reproducible. Wall motion abnormalities, right ventric- characterized by negative QRS complexes in V1 and V2, ular enlargement, and ectasia of the right ventricular outflow indicating a left bundle-branch block morphology with in- tract are more reliable and reproducible MRI parameters.5 ferior axis (positive complexes in the inferior leads: II, III Wall motion abnormalities can also be found by right ven- and AVF), and an otherwise normal baseline ECG and tricular angiography, which may also reveal aneurysms6 as echocardiogram. These characteristics, the paroxysmal na- areas of increased volume with transversely arranged hy- ture of the tachycardia, and its occurrence in a young female pertrophic trabeculae in the apical region distal to the mod- were suggestive of a right ventricular outflow tract tachy- erator band.7 cardia. In North America, 70% of cases of idiopathic ven- Right ventricular outflow tract tachycardia has been his- tricular tachycardia arise from the right ventricle, chiefly at torically regarded as idiopathic, but this perception is a consequence of reliance on conventional imaging and diag- Requests for reprints should be addressed to Ahmad Khraisat, MD, nostic techniques, such as stress testing, echocardiography, Assistant Clinical Professor of Medicine, Rosalind Franklin University of Medicine and Science/The Chicago Medical School, North Chicago, IL and coronary angiography. These tests usually yield normal 8 60064. results. On the other hand, MRI may show minor abnor- E-mail address: [email protected] malities of the right ventricle in up to 70% of patients,

Figure 1 Twelve-lead ECG showing sustained right ventricular outﬂow tract tachycardia with left bundle-branch block morphology (negative QRS in V1 and V2) and inferior axis (positive QRS in II, III, and AVF).

Figure 2 Baseline 12-lead ECG showing normal sinus rhythm.

Figure 3 ECG showing repetitive monomorphic right ventricular outﬂow tract tachycardia and premature ventricular contractions. Khraisat et al A Heart Aﬂutter 859

Table 1 Clinical and Electrophysiological Differences Between Patients with RVOT* and ARVC/D** Characteristic RVOT ARVD Symptoms and natural history Palpitations, presyncope, with benign Palpitations, syncope, sudden cardiac death, natural history with more serious natural history Family history None Positive for a family member with ARVD or sudden cardiac death Baseline ECG and signal averaged ECG Normal Abnormal in Ͼ60% of cases Transthoracic Echocardigraphy Mostly normal Usually abnormal (especially the right ventricle) Cardiac MRI Mostly Normal or minor abnormalities Mostly abnormal with major abnormalities Right ventricular Angiography Normal Mostly abnormal with major abnormalities Electrophysiological study Demonstrate only a single morphology of Demonstrate more than one morphology of ventricular tachycardia or ectopic beats inducible ventricular tachycardia Due to triggered activity Due to re-entry Medical therapy Responds to adenosine, beta and calcium Does not respond to medical therapy channel blockers Radiofrequency ablation Success in Ͼ90% of cases with low Success in Ͻ40% with high recurrence rate recurrence rate Implantable cardioverter defibrillator Not indicated Indicated with ablation considered adjunct to (ICD) ICD *RVOT: right ventricular outflow tract tachycardia. **ARVC/D: arrhythmogenic right ventricular cardiomyopathy/dysplasia. including focal thinning (fatty replacement), diminished lated successfully. The patient tolerated the procedure well systolic wall thickening, and abnormal wall motion.9 A and was discharged with an event recorder. summary of the clinical and electrophysiological differences between arrhythmogenic right ventricular cardiomy- References opathy/dysplasia and right ventricular outflow tract tachy- 1. Miles WM. Idiopathic ventricular outflow tract tachycardia: Where cardia is detailed in Table 1. does it originate? J Cardiovasc Electrophysiol. 2001;12:536-537. Clinical evidence has implicated several electrophysio- 2. Niroomand F, Carbucicchio C, Tondo C, et al. Electrophysiological logical mechanisms in right ventricular outflow tract tachy- characteristics and outcome in patients with idiopathic right ventricular cardia, including abnormal automaticity and triggered ac- arrhythmia compared with arrhythmogenic right ventricular dysplasia. Heart. 2002;87:41-47. tivity. The most common form of right ventricular outflow 3. Lerman BB. Response of nonreentrant catecholamine-mediated ven- tract tachycardia is related to triggered activity arising from tricular tachycardia to endogenous adenosine and acetylcholine: Evi- delayed afterdepolarization, and is thought to depend on dence for myocardial receptor-mediated effects. Circulation. 1993;87: intracellular calcium overload and cyclic adenosine mono- 382-390. phosphate. Such a ventricular tachycardia is therefore fre- 4. Nava A, Bauce B, Basso C, et al. Clinical profile and long-term follow-up of 37 families with arrhythmogenic right ventricular cardio- quently adenosine sensitive, may terminate with vagal ma- myopathy. J Am Coll Cardiol. 2000;36:2226-2233. 3 neuvers, and is facilitated by catecholamines. 5. Tandri H, Castillo E, Ferrari VA, et al. Magnetic resonance imaging of arrhythmogenic right ventricular dysplasia: sensitivity, specificity, and observer variability of fat detection versus functional analysis of the MANAGEMENT right ventricle. J Am Coll Cardiol. 2006;48:2277-2284. Long-term treatment options for right ventricular outflow 6. Indik JH, Dallas WJ, Ovitt T, et al. Do patients with right ventricular outflow tract ventricular arrhythmias have a normal right ventricular tract tachycardia include medical therapy and radiofre- wall motion? A quantitative analysis compared to normal subjects quency ablation. In view of the fact that right ventricular Cardiology. 2005;104:10-15. outflow tract tachycardia is a benign tachyarrhythmia unre- 7. Daliento L, Rizzoli G, Thiene G, et al. Diagnostic accuracy of right lated to ischemic heart disease or cardiomyopathy, an im- ventriculography in arrhythmogenic right ventricular cardiomyopathy. plantable cardioverter defibrillator is not indicated. Because Am J Cardiol. 1990;66:741-745. 8. Pietras RJ, Lam W, Bauernfeind R, et al. Chronic recurrent right structural heart disease is absent, beta-blockers or calcium ventricular tachycardia in patients without ischemic heart disease: channel blockers can be used. Radiofrequency ablation has Clinical, hemodynamic, and angiographic findings. Am Heart J. 1983; a cure rate of 90%, making it the preferred treatment, given 105:357-366. the relative youth of patients with this arrhythmia.10 9. Globits S, Kreiner G, Frank H, et al. Significance of morphological Our patient was initially treated with beta-blockers. After abnormalities detected by MRI in patients undergoing successful ablation of right ventricular outflow tract tachycardia. Circulation. 1997; treatment options were discussed, she was taken to the 96:2633-2640. electrophysiology laboratory, where mapping confirmed a 10. Lerman BB, Stein KM, Markowitz SM, et al. Ventricular arrhythmias right ventricular outflow origin that was subsequently ab- in normal hearts. Cardiol Clin. 2000;18:265-291. The American Journal of Medicine (2007) 120, 860-862

IMAGES IN RADIOLOGY Michael Bettmann, MD, Section Editor No Simple Sore Throat William J. Salyers, Jr., MD,a Christina Schnose, BA,a Adam Zarchan, MDb aDepartments of Internal Medicine and bDiagnostic Radiology, University of Kansas School of Medicine—Wichita. PRESENTATION DIAGNOSIS The case of a previously healthy 24-year-old woman with a Computed tomography of the neck, with and without con- chief complaint of sore throat provided an opportunity to trast, indicated an abscess adjacent to the right sternoclei- diagnose and treat an unusual entity. Despite outpatient domastoid muscle and another fluid collection alongside the treatment with clarithromycin, the patient’s odynophagia right medial pterygoid muscle (Figure 1). Thrombosis of the worsened, and she developed a fever. She was hospitalized, right internal jugular vein bordered both pockets. A cavitary treated with ceftriaxone, and subsequently discharged home lesion in the right upper lobe of the lung and a peripheral on levofloxacin and prednisone. Her sore throat improved nodule in the left upper lobe also were disclosed. Contrast- over the next week, but approximately 2 weeks after the enhanced computed tomography of the chest showed cavi- initial onset of her illness, she was readmitted for right-sided tary lesions, as well as peripheral nodules in the lungs swelling of the neck, severe pain, and dysphagia. bilaterally, findings consistent with septic pulmonary emboli (Figure 2). The patient’s past medical history was significant only The patient was diagnosed with Lemierre syndrome. for nephrolithiasis. She had no children and no prior history Also known as postanginal sepsis or necrobacillosis, Le- of miscarriages. She was on no chronic medications and mierre syndrome occurs secondary to an oropharyngeal took only ibuprofen as needed for pain. Additionally, she infection and is characterized by thrombophlebitis of the denied tobacco use. Her mother and grandmother each had internal jugular vein with resulting metastatic lesions. Our a history of deep vein thrombosis and pulmonary embolism. patient’s oropharyngeal infection was followed by suppurative jugular venous thrombosis and development of septic pulmonary emboli. Because her family history was signifi- ASSESSMENT cant for deep vein thrombosis and pulmonary embolism, a The patient had a fever of 100.4°F (38.0°C), and she was hypercoagulable workup was performed. She was found to hypotensive, with a blood pressure of 90/62 mm Hg. An have a low-medium positive anticardiolipin IgM titer at 64 examination of her neck revealed an asymmetric mass and MPL U/mL (normal, 0-9 MPL U/mL). right-sided swelling with induration, which altogether mea- Although Lemierre was not the first to discover this sured approximately6x6cm.Theangle of her right disorder, he was the first to write a comprehensive article 1 mandible was tender to palpation, and right-sided anterior describing the characteristic signs. Otherwise healthy 2 cervical lymphadenopathy was present. She was noted to young adults and adolescents are typically affected. Fuso- bacterium necrophorum is the most common pathogen, have trismus. Her lungs were clear to auscultation bilater- 3 ally, and examination of her upper and lower extremities causing up to 81% of cases. Other bacteria isolated from either blood or abscess cultures include Bacteroides species, revealed no erythema, swelling, or pain. Laboratory tests Streptococcus viridans, Peptostreptococcus species, Group disclosed leukocytosis (white blood cell count, 22.6 ϫ 103 B and group C Streptococcus species, Streptococcus pyo- cells/mm3) with 9% bands and an elevated erythrocyte sed- genes, Staphylococcus epidermidis, non-multidrug-resistant imentation rate of 45 mm/hr. methicillin-resistant Staphylococcus aureus, and Enterococ- cus species.2-4 While metastatic lesions can cause septic arthritis and rarely, osteomyelitis, liver abscesses, and soft tissue infections, the lung is the most common site of septic Requests for reprints should be addressed to William J. Salyers, Jr., 3,5 MD, Department of Internal Medicine, University of Kansas School of emboli with involvement in 80-97% of cases. Medicine—Wichita, 1010 N. Kansas Street, Wichita, KS 67214. Radiologic evaluation is a critical piece in solving the E-mail address: [email protected] puzzle of Lemierre syndrome. The important findings can

0002-9343/$ -see front matter © 2007 Elsevier Inc. All rights reserved. doi:10.1016/j.amjmed.2007.07.021 Salyers et al No Simple Sore Throat 861 be demonstrated by ultrasound, computed tomography, and magnetic resonance imaging. Each modality has benefits and drawbacks. Ultrasound reliably visualizes jugular venous thrombosis by demonstrating a noncompressible thrombus and loss of color flow. It does not use ionizing radiation, and it is inexpensive and convenient.6 Although ultrasound cannot visualize the jugular vein behind the clavicle or at the skull base, these limitations are rarely clinically relevant, as the extent of thrombosis does not typically affect clinical management.7 Computed tomography can evaluate the soft tissues surrounding the jugular vein more fully and visualize the septic pulmonary emboli characteristic of Lemierre syndrome. False negative results for jugular venous thrombosis can occur as a result of vascular flow phenomena and the vari- Figure 2 The black arrow indicates a cavitary lesion that is able appearance of the clot over time. However, the related consistent with septic pulmonary emboli. loss of both surrounding soft tissue planes and peripheral enhancement might still suggest the finding.8 Magnetic resonance imaging might be more sensitive for detection of present. Use of anticoagulation is controversial; no studies jugular venous thrombosis due to loss of the normal flow evaluate the efficacy of antibiotics alone or in conjunction void, but it is expensive and time-consuming, making it less with anticoagulation therapy.9 Some anecdotal evidence than ideal for patients who are critically ill.6 As a result, showed anticoagulation in combination with antibiotics computed tomography is the current imaging modality of will resolve symptoms more rapidly and prevent further choice for patients with suspected Lemierre syndrome. metastases.10 Our patient underwent early surgical incision and MANAGEMENT drainage of the abscesses and was treated with intrave- The mainstays of treatment for Lemierre syndrome include nous ceftriaxone and clindamycin. Although blood cultures extended antibiotic therapy with initial coverage for anaer- were negative, cultures from the fluid collection grew Por- obic organisms and surgical drainage if an abscess is phyromonas asaccharolytica and group C Streptococcus species. Porphyromonas species are gram-negative anaerobic pathogens found in oropharyngeal flora. These organisms have been found in deep facial infections, but they have rarely been reported in connection with Lemierre syndrome.11A recent MEDLINE search (keywords: Le- mierre syndrome; postanginal sepsis; necrobacillosis AND Porphyromonas) revealed only 3 reports of a correlation between Lemierre syndrome and Porphyromonas species.12-14 Although Fusobacterium species were not present, previous antibiotic use prior to her presentation might have suppressed the bacteria. Ramirez et al found that patients with Lemierre syndrome who had previously received antibiotics did not grow Fusobacterium necrophorum from blood cultures, including those who had positive blood cultures before initiation of antibiotic therapy.9 Because our patient had a positive anticardiolipin antibody titer, we prescribed enoxaparin and warfarin therapy with a goal international normalized ratio of 2.0-3.0. She responded well to treatment over the following week and was dismissed to follow-up with instructions to continue taking amoxicillin and clavulanate along with anticoagulants. Figure 1 An abscess, as denoted by the white arrow, caused leftward displacement of the trachea with minimal narrowing of the airway. Associated thrombosis of the right internal jugular vein References is marked with a large black arrow; a normal left internal jugular 1. Lemierre A. On certain septicemias due to anaerobic organisms. Lan- vein is designated with a small black arrow. cet. 1936;1:701-703. 862 The American Journal of Medicine, Vol 120, No 10, October 2007

2. Fong SM, Watson M. Lemierre syndrome due to non-multiresistant 9. Ramirez S, Hild TG, Rudolph CN, et al. Increased diagnosis of methicillin-resistant Staphylococcus aureus. J Paediatr Child Health. Lemierre syndrome and other Fusobacterium necrophorum infections 2002;38:305-307. at a Children’s Hospital. Pediatrics. 2003;112:e380. 3. Sinave CP, Hardy GJ, Fardy PW. The Lemierre syndrome: suppurative 10. Lustig LR, Cusick BC, Cheung SW, Lee KC. Lemierre’s syndrome: thrombophlebitis of the internal jugular vein secondary to oropharyn- two cases of postanginal sepsis. Otolaryngol Head Neck Surg. 1995; geal infection. Medicine (Baltimore). 1989;68:85-94. 112:767-772. 4. Wilson P, Tierney L. Lemierre syndrome caused by Streptococcus 11. Brook I. Microbiology and management of deep facial infections and pyogenes. Clin Infect Dis. 2005;41:1208-1209. Lemierre syndrome. ORL J Otorhinolaryngol Relat Spec. 2003;65: 5. Chirinos JA, Lichtstein DM, Garcia J, Tamariz, LJ. The evolution of 117-120. Lemierre syndrome: report of 2 cases and review of the literature. 12. Morizono S, Enjoji M, Sonoda N, et al. Lemierre’s syndrome: Porphyromo- Medicine (Baltimore). 2002;81:458-465. nas asaccharolytica as a putative pathogen. Intern Med. 2005;44:350-353. 6. Screaton NJ, Ravenel JG, Lehner PJ, et al. Lemierre syndrome: forgotten 13. Martínez P, Alcántara MJ, Berbegal J, Serrat C. Lemierre syndrome but not extinct—report of four cases. Radiology. 1999;213:369-374. caused by Porphyromonas endodontalis. Enferm Infecc Microbiol 7. Wing V, Scheible W. Sonography of jugular vein thrombosis. AJR Clin. 1999;17:368-369. Am J Roentgenol. 1983;140:333-336. 14. Vandenbos F, Roth S, Girard-Pipau F, et al. Lemierre syndrome due to 8. Albertyn LE, Alcock MK. Diagnosis of internal jugular vein throm- Porphyromonas spp. in a 21-year-old patient. Rev Med Interne. 2000; bosis. Radiology. 1987;162:505-508. 21:909-910. The American Journal of Medicine (2007) 120, 863-870

CLINICAL RESEARCH STUDY

Obesity Paradox in Patients with Hypertension and Coronary Artery Disease Seth Uretsky, MD,a Franz H. Messerli, MD,a Sripal Bangalore, MD, MHA,a Annette Champion, MBA,b Rhonda M. Cooper-DeHoff, PharmD,c Qian Zhou, PhD,b Carl J. Pepine, MDc aDivision of Cardiology, St Luke’s-Roosevelt Hospital, New York, NY; bAbbott, Abbott Park, Ill; cCardiovascular Medicine, University of Florida, Gainesville.

ABSTRACT

PURPOSE: An obesity paradox, a “paradoxical” decrease in morbidity and mortality with increasing body mass index (BMI), has been shown in patients with heart failure and those undergoing percutaneous coronary intervention. However, whether this phenomenon exists in patients with hypertension and coronary artery disease is not known. METHODS: A total of 22,576 hypertensive patients with coronary artery disease (follow-up 61,835 patient years, mean age 66 Ϯ 9.8 years) were randomized to a verapamil-SR or atenolol strategy. Dose titration and additional drugs (trandolapril and/or hydrochlorothiazide) were added to achieve target blood pressure control according to the Sixth Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure targets. Patients were classified into 5 groups according to baseline BMI: less than 20 kg/m2 (thin), 20 to 25 kg/m2 (normal weight), 25 to 30 kg/m2 (overweight), 30 to 35 kg/m2 (class I obesity), and 35 kg/m2 or more (class II-III obesity). The primary outcome was first occurrence of death, nonfatal myocardial infarction, or nonfatal stroke. RESULTS: With patients of normal weight (BMI 20 to Ͻ25 kg/m2) as the reference group, the risk of primary outcome was lower in the overweight patients (adjusted hazard ratio [HR] 0.77, 95% confidence interval [CI], 0.70-0.86, P Ͻ.001), class I obese patients (adjusted HR 0.68, 95% CI, 0.59-0.78, P Ͻ.001), and class II to III obese patients (adjusted HR 0.76, 95% CI, 0.65-0.88, P Ͻ.001). Class I obese patients had the lowest rate of primary outcome and death despite having smaller blood pressure reduction compared with patients of normal weight at 24 months (Ϫ17.5 Ϯ 21.9 mm Hg/Ϫ9.8 Ϯ 12.4 mm Hg vs Ϫ20.7 Ϯ 23.1 mm Hg /Ϫ10.6 Ϯ 12.5 mm Hg, P Ͻ.001). CONCLUSION: In a population with hypertension and coronary artery disease, overweight and obese patients had a decreased risk of primary outcome compared with patients of normal weight, which was driven primarily by a decreased risk of all-cause mortality. Our results further suggest a protective effect of obesity in patients with known cardiovascular disease in concordance with data in patients with heart failure and those undergoing percutaneous coronary intervention. © 2007 Elsevier Inc. All rights reserved.

KEYWORDS: Beta-blockers; Calcium antagonists; Coronary artery disease; Hypertension; INVEST; Obesity paradox

Obesity has been an established risk factor for increased creased risk of obesity for cardiovascular disease is thought cardiovascular mortality in men and women.1-10 The in- to be mediated in part by the clustering of risk factors such as hypertension, hypercholesterolemia, and diabetes mellitus (metabolic syndrome).11 In an epidemiologic study of This work was presented in part at the 2005 Scientiﬁc Session of the 527,265 men and women, there was an increased risk of American Society of Hypertension, San Francisco, Calif. death associated with excess body weight during midlife.7 Requests for reprints should be addressed to Franz H. Messerli, MD, St Similarly, in a Korean cohort of 1,213,829 men and women, Luke’s-Roosevelt Hospital Center, Roosevelt Division, Department of Car- diology, 1000 10th Avenue, New York, NY 10019. obese people had higher rates of death than people of E-mail address: [email protected]. normal weight.8

Although obesity increases the likelihood of metabolic criteria included patients receiving beta-blockers within 2 syndrome and its consequences, obesity has not been asso- weeks of randomization. ciated with a worse outcome in all patient populations. In This analysis was designed to study the effects of obesity patients with known cardiovascular disease (eg, those with in patients with hypertension and coronary artery disease. heart failure), those undergoing percutaneous coronary in- At each study visit, an extensive cardiovascular examina- terventions, and those with known tion was performed, and body coronary artery disease referred weight and height were recorded for single-photon emission com- CLINICAL SIGNIFICANCE into the electronic web-based data puted tomography, an “obesity capture system. BMI was calcu- paradox,” a paradoxical decrease ● In a hypertensive population with coro- lated by the system as weight in in morbidity and mortality with nary artery disease, overweight or obese kilograms/height in meters squared, increasing body mass index (BMI), stored in the database, and re- 12-18 patients were less likely to die or expe- has been described. In patients rience nonfatal myocardial infarction or corded in the patient’s visit sum- undergoing percutaneous coronary mary, which was printed for the stroke. intervention, obese patients had a patient’s medical record. 5.7% absolute decrease in inci- ● This “obesity paradox” occurred in men dence of death at 1 year compared and women across all age groups, even Statistical Analysis 14 with normal-weight patients. In though blood pressure was better con- Patients were classified into 5 patients with congestive heart fail- trolled in patients of normal weight groups according to their baseline 2 ure there was a 19% relative re- (BMI 20-25 kg/m2). BMI: less than 20 kg/m (thin), 20 duction in the risk of death in to 25 kg/m2 (normal weight), 25 to obese patients compared with nor- ● The reasons for the apparent protective 30 kg/m2 (overweight), 30 to 35 12 mal-weight patients. These find- effect of increased BMI are unclear. kg/m2 (class I obesity), and 35 ings, in contrast with earlier liter- kg/m2 or more (class II-III obe- ature, suggest a protective effect sity). Four patients with presumed associated with obesity in patients erroneous data, defined as BMI with known cardiovascular disease. greater than 100 kg/m2, were excluded. All analyses were However, in patients with hypertension and coronary conducted on the remaining 22,572 patients in the intention- artery disease, the role of BMI in the risk of cardiovascular to-treat population. Primary outcome was the first occur- events is not defined. The objective of the present study was rence of all-cause mortality, nonfatal myocardial infarction, to investigate the effect of obesity on cardiovascular out- and nonfatal stroke. Secondary outcomes were all-cause comes in treated hypertensive patients with known coronary mortality, nonfatal myocardial infarction, and nonfatal artery disease. To study this relationship, the INternational stroke with cardiovascular-related death as an additional VErapamil SR-trandolopril STudy (INVEST) cohort pro- 19 outcome. Patient groups were compared using analysis of vided an ideal population. variance for continuous variables and the chi-square test for METHODS categoric variables. For each outcome, a single unadjusted Cox proportional hazards model was used to compare risk Study Design among the 5 BMI categories (BMI 20-25 kg/m2 as refer- INVEST was a prospective, randomized, international study ence). A stepwise Cox proportional hazard regression of 22,576 patients with hypertension and coronary artery model was used to evaluate the role of BMI on the risk of disease. The inclusion and exclusion criteria, study design, primary outcome. Prespecified covariates forced into the and results have been published elsewhere.19 In brief, eli- model included treatment, age, race (white as reference), gible patients were randomized to a verapamil-SR–based or gender, history of myocardial infarction, and heart failure. an atenolol-based treatment strategy to achieve the Sixth Other baseline covariates were selected for the model on the Joint National Committee on Prevention, Detection, Evalu- basis of a P value of .10 or less. All analyses were per- ation, and Treatment of High Blood Pressure targets formed using standard software (SAS 9.1.3, SAS Institute (Ͻ140/90 mm Hg or Ͻ130/85 mm Hg in those with diabe- Inc, Cary, NC). tes or renal dysfunction).20 Patients were eligible if they were 50 years or older and had essential hypertension and documented coronary artery disease. Coronary artery dis- RESULTS ease was defined as a myocardial infarction 3 months or more before enrollment, coronary angiogram with more Baseline Characteristics than 50% stenosis in at least 1 major coronary artery, angina The mean age was 66 Ϯ 9.8 years, and the mean follow-up pectoris, or evidence of ischemia on at least 2 different was 2.7 years (range, 1 day to 5.4 years) with 61,835 patient modalities of stress tests (electrocardiogram, echocardio- years accumulated. Of the 22,572 patients included in this gram, radionuclide scan) that were consistent. Exclusion analysis, 2.2% were thin, 20.0% were normal weight, 39.9% Uretsky et al Obesity Paradox in Hypertension and CAD 865

Table 1 Baseline Characteristics of Patients According to Baseline Body Mass Index Category Normal Class I Class II-III Thin Weight Overweight Obesity Obesity Baseline Characteristics n ϭ 505 n ϭ 4256 n ϭ 9012 n ϭ 5560 n ϭ 2969 P Value* BMI (kg/m2) 18.5 Ϯ 1.2 23.1 Ϯ 1.3 27.5 Ϯ 1.4 32.1 Ϯ 1.4 39.5 Ϯ 4.6 Ͻ.001 Age (y) 72.0 Ϯ 10.3 69.3 Ϯ 10.3 66.4 Ϯ 9.6 64.4 Ϯ 9.0 62.3 Ϯ 8.5 Ͻ.001 Age Ͼ70 y (%) 54.7 46.9 34.6 26.3 18.2 Ͻ.001 Female (%) 71.1 52.3 47.5 51.6 63.7 Ͻ.001 Diabetes (%)† 12.7 20.3 26.6 33.1 39.7 Ͻ.001 Hypercholesterolemia (%)† 36.2 51.2 56.0 57.9 56.7 Ͻ.001 Renal insufﬁciency (%) 2.4 2.0 1.9 1.5 2.3 .075 Heart failure class I-III (%) 9.5 5.8 4.7 5.2 7.7 Ͻ.001 Current smoker (%) 26.7 15.4 11.7 11.1 10.2 Ͻ.001 Cancer (%) 4.0 4.3 3.4 2.9 2.4 Ͻ.001 Previous MI (%) 35.6 34.8 32.4 30.6 28.4 Ͻ.001 CABG (%) 14.7 16.0 16.4 15.8 13.7 .012 Percutaneous coronary intervention (%) 12.7 13.7 15.4 16.0 14.1 .003 Stroke/TIA (%) 9.7 9.1 6.9 6.6 6.3 Ͻ.001 LVH (%) 22.2 22.5 21.7 21.6 22.4 .711 Angina pectoris (%) 66.3 65.8 65.8 67.0 69.9 Ͻ.001 Peripheral vascular disease (%) 14.9 12.8 11.5 11.5 12.4 .030 Aspirin/antiplatelet therapy (%) 48.5 58.1 58.1 56.3 52.1 Ͻ.001 BMI ϭ body mass index; CABG ϭ coronary artery bypass graft; LVH ϭ left ventricular hypertrophy; MI ϭ myocardial ischemia; TIA ϭ transient ischemic attack. Values expressed as mean Ϯ standard deviation unless otherwise indicated. *Chi-square test for categoric variables and 1-way analysis of variance for numeric variables. †History of or currently taking antidiabetic or lipid-lowering medications.

were overweight, 24.6% had class I obesity, and 13.2% had Blood Pressure Control class II to III obesity. Pertinent baseline characteristics by Class I obese patients had a lower baseline mean systolic BMI class are summarized in Table 1. Compared with blood pressure and higher baseline mean diastolic blood normal-weight patients, overweight and class I to III obese pressure when compared with normal-weight patients (Ta- patients were younger, had a history of diabetes and hyper- ble 2). At 24 months of treatment, class I obese patients had cholesterolemia, and were less likely to be a current smoker a lower reduction in both systolic blood pressure and dia- or to have a history of cancer, previous myocardial infarc- stolic blood pressure when compared with normal-weight tion, stroke/transient ischemic attack (TIA), or peripheral patients (Ϫ17.5 Ϯ 21.9 mm Hg/Ϫ9.8 Ϯ 12.4 mm Hg vs vascular disease. Ϫ20.7 Ϯ 23.1 mm Hg /Ϫ10.6 Ϯ 12.5 mm Hg, P Ͻ.001), and

Table 2 Blood Pressure at Baseline and 24 Months by Baseline Body Mass Index Category

Normal Class I Class II-III Thin Weight Overweight Obesity Obesity n ϭ 301 n ϭ 3094 n ϭ 6358 n ϭ 3950 n ϭ 1986 P Value* Mean blood pressure at baseline (mm Hg) Systolic 154.4 Ϯ 21.9 151.8 Ϯ 19.8 151.0 Ϯ 19.0 149.9 Ϯ 18.7 150.3 Ϯ 19.0 Ͻ.001 Diastolic 85.6 Ϯ 11.5 86.8 Ϯ 12.4 87.6 Ϯ 11.7 87.7 Ϯ 11.5 88.1 Ϯ 11.7 Ͻ.001 Mean change in blood pressure at 24 mo Systolic Ϫ22.9 Ϯ 25.0 Ϫ20.7 Ϯ 23.1 Ϫ19.6 Ϯ 22.3 Ϫ17.5 Ϯ 21.9 Ϫ15.8 Ϯ 21.9 Ͻ.001 Diastolic Ϫ9.8 Ϯ 12.2 Ϫ10.6 Ϯ 12.5 Ϫ10.4 Ϯ 12.3 Ϫ9.8 Ϯ 12.4 Ϫ9.2 Ϯ 12.5 Ͻ.001 Blood pressure control 69.1 73.2 73.3 70.4 63.5 Ͻ.001 (140/90 mm Hg), % JNC VI blood pressure control, % 65.4 67.3 65.6 59.8 51.5 Ͻ.001 BMI ϭ body mass index; JNC ϭ Joint National Committee. *Results from a chi-square test for blood pressure control, 1-way analysis of variance for baseline blood pressure, and analysis of covariance for blood pressure changes with BMI category as the factor and baseline blood pressure as covariate. 866 The American Journal of Medicine, Vol 120, No 10, October 2007

ing 24 months: thin ϭ 2.3/2.3, normal weight ϭ 2.4/2.4, overweight ϭ 2.5/2.6, class I obesity ϭ 2.7/2.7, class II-III obesity ϭ 3.0/3.0). The percentage of patients requiring 3 or more antihypertensive drugs was higher in the higher BMI categories but did not differ between treatment strategies in a given BMI cohort at 24 months (Figure 1). Most patients required 3 or more antihypertensive drugs at 24 months.

Primary and Secondary Outcomes The relationship between the incidence of the primary outcome and BMI followed a J-shaped curve with the lowest Figure 1 Number of antihypertensive drugs according to incidence in the class I obese group. The incidence of the treatment strategy and BMI at 24 months. Ve ϭ verapamil-SR; primary outcome was lower in overweight, class I, and class At ϭ atenolol; BMI ϭ body mass index. II to III obese patients compared with normal-weight patients despite better blood pressure control in the normal- weight patients at 24 months (Table 3 and Figure 2). The a lower percentage of class I obese patients reached the secondary outcome of all-cause mortality and cardiovascu- target blood pressure goal of 140/90 mm Hg and the Sixth lar-related death was lower in overweight, class I, and class Joint National Committee on Prevention, Detection, Evalu- II to III obese patients compared with normal-weight pa- ation, and Treatment of High Blood Pressure target when tients. For the secondary outcomes of nonfatal myocardial compared with normal-weight patients. infarction and nonfatal stroke there was no difference comparing overweight, class I, and class II to III obese patients Medication Use with normal-weight patients. The mean use of study and nonstudy antihypertensive drugs In a multivariate analysis, higher BMI, Asian or Hispanic was higher in the higher BMI groups for both the calcium race (vs white), and hypercholesterolemia were associated channel blocker strategy and atenolol-based strategy (mean with a decreased risk for the primary outcome (Table 4). number of study and nonstudy antihypertensive drugs dur- The following were associated with an increased risk of

Table 3 Primary Outcomes and Secondary Outcomes According to Baseline Body Mass Index Category

Subjects with Total Event BMI Category Events (%) Subjects HR* 95% CI P Value Primary outcome Thin 113 (22.4%) 505 1.74 1.42-2.13 Ͻ.001 Normal weight (reference) 593 (13.1%) 4526 Overweight 858 (9.5%) 9012 0.71 0.64-0.78 Ͻ.001 Class I obesity 445 (8.0%) 5560 0.57 0.51-0.65 Ͻ.001 Class II-III obesity 260 (8.8%) 2969 0.61 0.53-0.71 Ͻ.001 Death Thin 98 (19.4%) 505 1.85 1.49-2.30 Ͻ.001 Normal weight (reference) 484 (10.7%) 4526 Overweight 656 (7.3%) 9012 0.66 0.59-0.74 Ͻ.001 Class I obesity 329 (5.9%) 5560 0.52 0.45-0.59 Ͻ.001 Class II-III obesity 199 (6.7%) 2969 0.57 0.49-0.68 Ͻ.001 Cardiovascular-related death Thin 40 (7.9%) 505 1.56 1.12-2.19 .009 Normal weight (reference) 233 (5.1%) 4526 Overweight 332 (3.7%) 9012 0.70 0.59-0.82 Ͻ.001 Class I obesity 157 (2.8%) 5560 0.51 0.42-0.63 Ͻ.001 Class II-III Obesity 100 (3.4) 2969 0.60 0.48-0.76 Ͻ.001 Nonfatal MI Thin 7 (1.4%) 505 1.12 0.51-2.46 .78 Normal weight (reference) 57 (1.3%) 4526 Overweight 124 (1.4%) 9012 1.07 0.78-1.46 .69 Class I obesity 76 (1.4%) 5560 1.02 0.73-1.44 .89 Class II-III Obesity 40 (1.3%) 2969 0.99 0.66-1.48 .96 Nonfatal stroke Thin 13 (2.6%) 505 1.83 1.01-3.32 .047 Normal weight (reference) 65 (1.4%) 4526 Overweight 106 (1.2%) 9012 0.80 0.59-1.09 .15 Class I obesity 65 (1.2%) 5560 0.77 0.55-1.09 .14 Class II-III obesity 30 (1.0%) 2969 0.65 0.42-1.01 .054 BMI ϭ body mass index; HR ϭ hazard ratio; CI ϭ conﬁdence interval; MI ϭ myocardial infarction. *Unadjusted HRs. Uretsky et al Obesity Paradox in Hypertension and CAD 867

Figure 2 Percentage of patients to reach primary outcome, all- Figure 3 Percentage of men and women to reach primary out- cause mortality, nonfatal stroke, and nonfatal myocardial infarction come according to BMI category. BMI ϭ body mass index. according to BMI category. BMI ϭ body mass index; MI ϭ myocardial infarction. [CI], 0.70-0.86, P Ͻ.001), class I (adjusted HR 0.68, 95% CI, 0.59-0.78, P Ͻ.001), and class II to III obese patients reaching primary outcome: increasing age, thin body, male (adjusted HR 0.76, 95% CI, 0.65-0.88, P Ͻ.001), and higher gender, history of myocardial infarction, congestive heart in thin patients (adjusted HR 1.52, 95% CI, 1.24-1.86, failure, US resident, coronary artery bypass graft or angio- P Ͻ.001). plasty, peripheral vascular disease, renal insufficiency, In each BMI category, men had a higher rate of primary smoking (ever), previous stroke or TIA, and diabetes. outcome than women (Figure 3). In a subgroup analysis of By using normal-weight patients as the reference group, patients with a history of congestive heart failure and per- the risk of primary outcome was lower in the overweight cutaneous coronary intervention, class I obese patients had (adjusted hazard ratio [HR] 0.77, 95% confidence interval a 6.6% and 7.5% lower incidence of the primary outcome, respectively, than normal-weight patients. However, the incidence of primary outcome in class II to III obese patients Table 4 Stepwise Model Risk for Primary Outcome was higher than in class I obese patients, including the subgroups with heart failure and previous percutaneous cor- Variable HR 95% CI P Value onary intervention. Thin (vs normal) 1.52 1.24-1.86 Ͻ.0001 Overweight (vs normal) 0.77 0.70-0.86 Ͻ.0001 Class I obesity (vs normal) 0.68 0.59-0.77 Ͻ.0001 Primary Outcome and Age Class II-III obesity (vs normal) 0.76 0.65-0.88 Ͻ.001 In general, for each BMI category there was an increased Age (10-y increments) 1.63 1.56-1.71 Ͻ.0001 incidence of the primary outcome with age for men and Strategy 0.97 0.89-1.05 .44 women (Figures 4 and 5). For a given age and gender, the Gender (male vs female) 1.12 1.02-1.23 .013 Race (vs white) relationship between BMI and primary outcome followed a Black 1.11 0.98-1.26 .09 J-shaped curve with the lowest incidence of primary out- Hispanic 0.90 0.80-1.00 .049 come in overweight and class I obese patients, although Asian 0.46 0.23-0.93 .03 sample sizes for some of the age groups were small. Other 0.86 0.60-1.25 .44 Previous MI 1.33 1.22-1.46 Ͻ.0001 CHF 1.93 1.70-2.19 Ͻ.0001 DISCUSSION US resident 1.59 1.39-1.81 Ͻ.0001 CABG or percutaneous coronary 1.16 1.06-1.27 .002 This study addressed the effect of BMI on cardiovascular intervention outcomes in a cohort of patients with hypertension and Hypercholesterolemia* 0.81 0.75-0.89 Ͻ.0001 coronary artery disease. Our study is in agreement with Peripheral vascular disease 1.27 1.14-1.42 Ͻ.0001 previous studies that observed an obesity paradox in pa- Renal insufficiency 1.50 1.23-1.82 Ͻ.0001 tients with previous cardiovascular disease. Smoking (ever) 1.40 1.28-1.53 Ͻ.0001 Previous stroke/TIA 1.43 1.27-1.62 Ͻ.0001 Diabetes* 1.79 1.64-1.95 Ͻ.0001 Body Mass Index and Cardiovascular Risk CI ϭ confidence interval; HR ϭ hazard ratio; MI ϭ myocardial in- Epidemiologic studies have described obesity as an inde- farction; CHF ϭ congestive heart failure; CABG ϭ coronary artery by- pendent risk factor for cardiovascular disease and pass graft; TIA ϭ transient ischemic attack. death.1,3-10 In a recent analysis of the Framingham cohort, *History of or currently taking antidiabetic or lipid-lowering overweight subjects were found to have increased risk for medications. developing cardiovascular disease compared with normal- 868 The American Journal of Medicine, Vol 120, No 10, October 2007

Figure 4 Percentage of men to reach primary outcome according to BMI and age. BMI ϭ body mass index. weight patients.11 Manson et al.5 found that excess body This ﬁnding is consistent with the notion of an “obesity weight was directly related to all-cause mortality among the paradox” that has been described in patients with docu- 115,195 women enrolled in the Nurses’ Health Study. Fle- mented cardiac disease (eg, heart failure), patients undergo- gal et al.3 reported obesity to be associated with increased ing percutaneous coronary intervention, and patients with mortality in the participants of the National Health and coronary artery disease referred for single photon emission Nutrition Examination Survey. However, these studies were computed tomography.12-18 In our analysis, overweight, performed in the general population. class I, and class II to III obese patients with a history of congestive heart failure or percutaneous coronary inter- Obesity Paradox in Cardiovascular Disease vention had a lower incidence of the primary outcome than In contrast with these epidemiologic studies, our analysis of normal-weight patients, similar to previous analyses.12-17 In the INVEST cohort suggests that among patients with a a study of 7765 patients with heart failure enrolled in the history of hypertension and coronary artery disease, over- Digitalis Investigation Group trial, overweight and obese weight and class I to III obesity were associated with a patients had a relative risk reduction of 12% and 19%, decreased risk of morbidity and mortality compared with respectively, for all-cause death compared with normal- normal-weight patients, despite less blood pressure control. weight patients.12 This is in concordance with our study that

Figure 5 Percentage of women to reach primary outcome according to BMI and age. BMI ϭ body mass index. Uretsky et al Obesity Paradox in Hypertension and CAD 869 showed overweight and obese patients with a history of better outcome in this population. Fifth, there are conflicting congestive had a relative risk reduction of 20% and 24%, data whether obesity itself confers an increased risk for respectively, for the primary outcome. In a meta-analysis of cardiovascular disease apart from its associated metabolic 7290 patients undergoing percutaneous coronary interven- derangements.11,32 If these risk factors are well managed, tion, overweight and obese patients had a relative risk re- such as hypertension and coronary artery disease, this may duction in death at 1 year of 36% and 27%, respectively.17 negate any increased risk associated with obesity. In the This is in concordance with our study that showed over- INVEST cohort, the blood pressure control was excellent weight and obese patients with a history of percutaneous and may negate the deleterious effects of obesity. Sixth, coronary intervention had a relative risk reduction of 40% BMI itself has been questioned as the optimal measurement and 41%, respectively, for the primary outcome. However, to use for assessing health risk associated with obesity. some reports have questioned the existence of an obesity Other measures, such as waist-to-hip ratio and visceral fat paradox; thus, it remains controversial.21,22 In our study measurement, may be better, and it has been postulated that group, thin patients (BMI Ͻ20 kg/m2) had the highest risk “it is time to throw BMI out.”33,34 In some studies, BMI for the primary outcome, all-cause mortality, cardiovascular predicted mortality in women less well than in men.35 Ashton death, and nonfatal stroke compared with the normal-weight et al.36 found BMI to be a poor discriminator of cardiovascular patients. Although thin patients at baseline were older and heart disease risk in women compared with men despite a had a higher baseline incidence of renal insufficiency, heart worse metabolic profile in those with increased BMI. failure, current smoking, previous myocardial infarction, stroke/TIA, and peripheral vascular disease, they had a Study Limitations higher risk for cardiovascular events in the multivariate This was a post hoc analysis and thus suffers from the model even after accounting for these differences. Whether limitations of such studies. Our conclusions should be con- low BMI itself is a risk factor for or a marker of severity of sidered to be hypothesis generating. The INVEST did not cardiovascular disease is unclear. collect the waist-to-hip ratio data; therefore, we could not compare BMI with waist-to-hip ratio. Although we did find Postulated Mechanisms for the Obesity differences in the primary and secondary outcomes between Paradox in Patients with Cardiovascular BMI groups, the baseline characteristics of the BMI categories Disease were not well matched. Although a stepwise model was used, Several mechanisms could explain this paradox in patients the impact of these baseline differences cannot be ruled out. with cardiovascular disease. First, studies have shown that normal-weight patients have a significantly higher percent- CONCLUSION age of high-risk coronary anatomy (left main disease or In this well-treated hypertensive cohort with coronary artery triple vessel disease) compared with obese patients.23 Cor- disease, increasing BMI was associated with decreased mor- onary artery calcification area measured by electron-beam bidity and mortality when compared with normal-weight computed tomography was found to be significantly greater patients, consistent with an “obesity paradox.” Whether this in overweight patients compared with obese patients.24 relationship is the result of the shortcomings of BMI as a These findings provide a possible anatomic substrate for an risk factor needs to be further elucidated. obesity paradox. Second, leaner patients with heart failure have been shown to have increased levels of tumor necrosis References factor and other inflammatory cytokines compared with 25 1. Calle EE, Thun MJ, Petrelli JM, et al. Body-mass index and mortality obese patients. Adipose tissue has been shown to produce in a prospective cohort of U.S. adults. N Engl J Med. 1999;341:1097- soluble tumor necrosis factor receptor that is thought to 1105. neutralize the deleterious effects of tumor necrosis factor- 2. Batty GD, Shipley M, Jarrett J, et al. Obesity and overweight in alpha on the myocardium, which may explain a protective relation to disease-specific mortality in men with and without existing effect of obesity in patients with heart failure.13,26 Third, coronary heart disease in London: The original Whitehall study. Heart. 2006;92:886-92. Epub 2005 Nov 3. obese patients display a readily identifiable phenotype that 3. Flegal KM, Graubard BI, Williamson DF, Gail MH. Excess deaths is believed to reflect a high risk for cardiovascular disease, associated with underweight, overweight, and obesity. JAMA. 2005; and they may receive or seek treatment earlier in the time 293:1861-1867. course of disease, thereby altering the natural history of 4. Grundy SM, Pasternak R, Greenland P, et al. AHA/ACC scientific their disease when compared with lean patients (lead time statement: Assessment of cardiovascular risk by use of multiple-risk- factor assessment equations: a statement for healthcare professionals bias). Fourth, it has been well documented that the hemo- from the American Heart Association and the American College of dynamics of obesity hypertension are characterized by a Cardiology. J Am Coll Cardiol. 1999;34:1348-1359. high cardiac output, an expanded blood volume, and a lower 5. Manson JE, Willett WC, Stampfer MJ, et al. Body weight and mor- systemic vascular resistance when compared with normal- tality among women. N Engl J Med. 1995;333:677-685. 27-31 6. Hubert HB, Feinleib M, McNamara PM, Castelli WP. Obesity as an weight patients. Because systemic resistance reflects independent risk factor for cardiovascular disease: a 26-year follow-up the severity of hypertensive cardiovascular disease, the of participants in the Framingham Heart Study. Circulation. 1983;67: comparatively low values in obesity may translate into a 968-977. 870 The American Journal of Medicine, Vol 120, No 10, October 2007

7. Adams KF, Schatzkin A, Harris TB, et al. Overweight, obesity, and 21. Domanski MJ, Jablonski KA, Rice MM, et al. Obesity and cardiovas- mortality in a large prospective cohort of persons 50 to 71 years old. cular events in patients with established coronary disease. Eur Heart J. N Engl J Med. 2006;355:763-778. 2006;27:1416-1422. 8. Jee SH, Sull JW, Park J, et al. Body-mass index and mortality in 22. Gruberg L, Mercado N, Milo S, et al. Impact of body mass index on Korean men and women. N Engl J Med. 2006;355:779-787. the outcome of patients with multivessel disease randomized to either 9. Cui R, Iso H, Toyoshima H, et al. Body mass index and mortality from coronary artery bypass grafting or stenting in the ARTS trial: the cardiovascular disease among Japanese men and women: the JACC obesity paradox II? Am J Cardiol. 2005;95:439-444. study. Stroke. 2005;36:1377-1382. 23. Rubinshtein R, Halon DA, Jaffe R, et al. Relation between obesity and 10. Jousilahti P, Tuomilehto J, Vartiainen E, et al. Body weight, cardio- severity of coronary artery disease in patients undergoing coronary vascular risk factors, and coronary mortality. 15-year follow-up of angiography. Am J Cardiol. 2006;97:1277-1280. middle-aged men and women in eastern Finland. Circulation. 1996; 24. Cassidy AE, Bielak LF, Zhou Y, et al. Progression of subclinical 93:1372-1379. coronary atherosclerosis: does obesity make a difference? Circulation. 11. Wilson PW, D’Agostino RB, Sullivan L, et al. Overweight and obesity 2005;111:1877-1882. as determinants of cardiovascular risk: the Framingham experience. Arch Intern Med. 2002;162:1867-1872. 25. Feldman AM, Combes A, Wagner D, et al. The role of tumor necrosis 12. Curtis JP, Selter JG, Wang Y, et al. The obesity paradox: body mass factor in the pathophysiology of heart failure. J Am Coll Cardiol. index and outcomes in patients with heart failure. Arch Intern Med. 2000;35:537-544. 2005;165:55-61. 26. Mohamed-Ali V, Goodrick S, Bulmer K, et al. Production of soluble 13. Horwich TB, Fonarow GC, Hamilton MA, et al. The relationship tumor necrosis factor receptors by human subcutaneous adipose tissue between obesity and mortality in patients with heart failure. J Am Coll in vivo. Am J Physiol. 1999;277(6 Pt 1):E971-E975. Cardiol. 2001;38:789-795. 27. Barrett-Connor E, Khaw KT. Is hypertension more benign when as- 14. Gruberg L, Weissman NJ, Waksman R, et al. The impact of obesity on the sociated with obesity? Circulation. 1985;72:53-60. short-term and long-term outcomes after percutaneous coronary interven- 28. Messerli FH. Cardiovascular adaptations to obesity and arterial hyper- tion: the obesity paradox? J Am Coll Cardiol. 2002;39:578-584. tension: detrimental or beneﬁcial? Int J Cardiol. 1983;3:94-97. 15. Cox N, Resnic FS, Popma JJ, et al. Comparison of the risk of vascular 29. Messerli FH. Cardiopathy of obesity—a not-so-Victorian disease. complications associated with femoral and radial access coronary N Engl J Med. 1986;314:378-380. catheterization procedures in obese versus nonobese patients. Am J 30. Messerli FH, Sundgaard-Riise K, Reisin E, et al. Disparate cardiovas- Cardiol. 2004;94:1174-1177. cular effects of obesity and arterial hypertension. Am J Med. 1983;74: 16. Lavie CJ, Osman AF, Milani RV, Mehra MR. Body composition and 808-812. prognosis in chronic systolic heart failure: the obesity paradox. Am J 31. Perrera GA, Damon A. Height, weight, and their ratio in the acceler- Cardiol. 2003;91:891-894. ated form of hypertension. Arch Intern Med. 1957;100:236-265. 17. Gurm HS, Brennan DM, Booth J, et al. Impact of body mass index on 32. Thomas F, Bean K, Pannier B, et al. Cardiovascular mortality in outcome after percutaneous coronary intervention (the obesity para- overweight subjects: the key role of associated risk factors. Hyperten- dox). Am J Cardiol. 2002;90:42-45. sion. 2005;46:654-659. 18. Kang X, Shaw LJ, Hayes SW, et al. Impact of body mass index on 33. Kragelund C, Omland T. A farewell to body-mass index? Lancet. cardiac mortality in patients with known or suspected coronary artery 2005;366:1589-1591. disease undergoing myocardial perfusion single-photon emission computed tomography. J Am Coll Cardiol. 2006;47:1418-1426. 34. Yusuf S, Hawken S, Ounpuu S, et al. Obesity and the risk of myo- 19. Pepine CJ, Handberg EM, Cooper-DeHoff RM, et al. A calcium cardial infarction in 27,000 participants from 52 countries: a case- antagonist vs a non-calcium antagonist hypertension treatment strategy control study. Lancet. 2005;366:1640-1649. for patients with coronary artery disease. The International Verapamil- 35. Seidell JC, Verschuren WM, van Leer EM, Kromhout D. Overweight, Trandolapril Study (INVEST): a randomized controlled trial. JAMA. underweight, and mortality. A prospective study of 48,287 men and 2003;290:2805-2816. women. Arch Intern Med. 1996;156:958-963. 20. The sixth report of the Joint National Committee on Prevention, 36. Ashton WD, Nanchahal K, Wood DA. Body mass index and metabolic Detection, Evaluation, and Treatment of High Blood Pressure. Arch risk factors for coronary heart disease in women. Eur Heart J. 2001; Intern Med. 1997;157:2413-2446. 22:46-55. The American Journal of Medicine (2007) 120, 871-879

CLINICAL RESEARCH STUDY

Clinical Characteristics of Patients with Acute Pulmonary Embolism: Data from PIOPED II Paul D. Stein, MD,a,b Afzal Beemath, MD,a Fadi Matta, MD,a John G. Weg, MD,c Roger D. Yusen, MD,d Charles A. Hales, MD,e Russell D. Hull, MBBS, MSc,f Kenneth V. Leeper, Jr., MD,g H. Dirk Sostman, MD,h Victor F. Tapson, MD,i John D. Buckley, MD,j Alexander Gottschalk, MD,k Lawrence R. Goodman, MD,l Thomas W. Wakeﬁed, MD,m Pamela K. Woodard, MDn aDepartment of Research, St. Joseph Mercy Oakland Hospital, Pontiac, Mich; bDepartment of Medicine, Wayne State University School of Medicine, Detroit, Mich; cDepartment of Medicine, University of Michigan, Ann Arbor; dDepartment of Medicine, Washington University School of Medicine, St. Louis, Mo; eDepartment of Medicine, Massachusetts General Hospital, and Harvard Medical School, Boston; fDepartment of Medicine, University of Calgary, Calgary, Alberta, Canada; gDepartment of Medicine, Emory University, Atlanta, Ga; hOfﬁce of the Dean, Weill Cornell Medical College and Methodist Hospital, Houston, Tex; iDepartment of Medicine, Duke University, Durham, NC; jDepartment of Medicine, Henry Ford Hospital, Detroit, Mich; kDepartment of Radiology, Michigan State University, East Lansing; lDepartment of Radiology, Medical College of Wisconsin, Milwaukee; mDepartment of Surgery, University of Michigan, Ann Arbor; nDepartment of Radiology, Washington University, St. Louis, Mo.

ABSTRACT

BACKGROUND: Selection of patients for diagnostic tests for acute pulmonary embolism requires recognition of the possibility of pulmonary embolism on the basis of the clinical characteristics. Patients in the Prospective Investigation of Pulmonary Embolism Diagnosis II had a broad spectrum of severity, which permits an evaluation of the subtle characteristics of mild pulmonary embolism and the characteristics of severe pulmonary embolism. METHODS: Data are from the national collaborative study, Prospective Investigation of Pulmonary Embolism Diagnosis II. RESULTS: There may be dyspnea only on exertion. The onset of dyspnea is usually, but not always, rapid. Orthopnea may occur. In patients with pulmonary embolism in the main or lobar pulmonary arteries, dyspnea or tachypnea occurred in 92%, but the largest pulmonary embolism was in the segmental pulmonary arteries in only 65%. In general, signs and symptoms were similar in elderly and younger patients, but dyspnea or tachypnea was less frequent in elderly patients with no previous cardiopulmonary disease. Dyspnea may be absent even in patients with circulatory collapse. Patients with a low-probability objective clinical assessment sometimes had pulmonary embolism, even in proximal vessels. CONCLUSION: Symptoms may be mild, and generally recognized symptoms may be absent, particularly in patients with pulmonary embolism only in the segmental pulmonary branches, but they may be absent even with severe pulmonary embolism. A high or intermediate-probability objective clinical assessment suggests the need for diagnostic studies, but a low-probability objective clinical assessment does not exclude the diagnosis. Maintenance of a high level of suspicion is critical. © 2007 Elsevier Inc. All rights reserved.

KEYWORDS: Clinical diagnosis; Deep venous thrombosis; Pulmonary embolism; Venous thromboembolism

This study was supported by Grants HL63899, HL63928, HL63931, HL063932, HL63940, HL63941, HL63981, HL63982, and HL67453 from the U.S. Department of Health and Human Services, Public Health Services, National Heart, Lung, and Blood Institute, Bethesda, Maryland. Requests for reprints should be addressed to Paul D. Stein, MD, St. Joseph Mercy Oakland, 44405 Woodward Ave., Pontiac, MI 48341-5023. E-mail address: [email protected]

Acute pulmonary embolism in patients with severe or fatal bronchitis, emphysema, chronic obstructive pulmonary dis- pulmonary embolism at autopsy is generally unrecognized ease, current or history of right or left-sided heart failure, antemortem.1-4 Advances in the diagnostic methods for lung cancer, or prior pulmonary embolism. acute pulmonary embolism should affect this high rate of The circulatory collapse syndrome was defined as loss of underdiagnosis.5-8 However, the successful use of diagnos- consciousness or systolic blood pressure of 80 mm Hg tic pathways requires recognition or less. The hemoptysis/pleuritic of patients with possible acute pain syndrome (previously termed pulmonary embolism on the basis CLINICAL SIGNIFICANCE the pulmonary infarction syn- of the clinical characteristics. drome)13 was defined as patients The clinical characteristics of ● In patients with pulmonary embolism only with either hemoptysis or pleuritic patients with acute pulmonary in the segmental pulmonary branches, pain in the absence of circulatory embolism in the first Prospective generally recognized symptoms may be collapse. The uncomplicated dys- Investigation of Pulmonary Em- mild or absent, leading to underdiagnosis. pnea syndrome was defined as dys- bolism Diagnosis (PIOPED), an pnea in the absence of hemoptysis, accuracy study of ventilation- ● Absence of symptoms is most common in pleuritic pain, or circulatory collapse. perfusion scintigraphy, were de- patients with pulmonary embolism only Measurements of arterial blood scribed in all enrolled patients,9 in the segmental pulmonary branches, gases were obtained while the pa- patients with no previous car- but this also occurs in patients with tient was breathing room air. The 10 diopulmonary disease, patients proximal pulmonary embolism. alveolar-arterial (A-a) oxygen dif- grouped according to the present- ference was calculated as follows:15 ing syndromes of pulmonary em- ● A low-probability objective clinical as- bolism,11 and the elderly.12 One of sessment does not exclude pulmonary A-a oxygen difference (mm Hg) the strengths of the data from embolism, even in the proximal pulmo- ϭ Ϫ Ϫ 150 1.25 PACO2 PaO2 PIOPED II is that many patients nary arteries. ϭ with mild pulmonary embolism where PACO2 partial pressure of were included, which permits the carbon dioxide in arterial blood (mm ϭ identification of subtle clinical Hg), and PaO2 partial pressure of characteristics. However, the clinical characteristics of pa- oxygen in arterial blood (mm Hg). tients with pulmonary embolism in PIOPED,9-12 as in PIO- PED II and other investigations of patients enrolled in clinical Statistical Methods trials,13,14 were sufficient to alert attuned physicians to the The chi-square test was used to compare the prevalence of diagnosis and characteristics of patients who met the inclusion clinical features in patients with and without pulmonary em- criteria. Patients who were too ill to participate, who died bolism. Because of the large number of comparisons, P values suddenly, or who were not identified because the clinical find- are underestimates. Comparisons of continuous variable means ings were mild or atypical were not included. With these were made with the 2-tailed Student unpaired t test. constraints in mind, we describe the clinical characteristics of patients enrolled in PIOPED II. RESULTS Acute pulmonary embolism was present in 192 patients, PATIENTS AND METHODS among whom 133 (69%) had no prior cardiopulmonary PIOPED II was a prospective multicenter investigation of disease. Pulmonary embolism was excluded in 632 patients, multidetector computed-tomography angiography alone and among whom 366 (58%) had no prior cardiopulmonary combined with venous phase imaging of the pelvic and disease. thigh veins for the diagnosis of acute pulmonary embolism.5 A composite reference test was used.5 Patients aged 18 years or Syndromes of Pulmonary Embolism more with clinically suspected acute pulmonary embolism The syndrome of hemoptysis or pleuritic pain occurred in were potentially eligible.5 Exclusion criteria included an in- 41% of patients with no prior cardiopulmonary disease and ability to complete tests within 36 hours, critical illness, ven- in 44% of all patients with pulmonary embolism (Table 1). tilatory support, shock, recent myocardial infarction, abnormal The uncomplicated dyspnea syndrome occurred in 36% of serum creatinine, allergy to contrast material, pregnancy, treat- patients with no prior cardiopulmonary disease and in 36% ment with long-term anticoagulants, inferior vena cava filter, of all patients with pulmonary embolism. The circulatory and deep venous thrombosis of the upper extremity. collapse syndrome was uncommon: 8% in patients with no To avoid confusing clinical findings of pulmonary em- prior cardiopulmonary disease and 8% in all enrolled pa- bolism with comorbid conditions, we evaluated patients tients with acute pulmonary embolism. The presentations of with no prior cardiopulmonary disease in addition to eval- 19 patients (14%) with pulmonary embolism and no prior uating all patients. No prior cardiopulmonary disease was cardiopulmonary disease differed from these syndromes. defined as no current asthma, pneumonia, history of chronic The presenting findings in some of these patients were Stein et al Clinical Characteristics of Pulmonary Embolism 873

Table 1 Syndromes of Acute Pulmonary Embolism PE No PE PE No PE No Prior CPD No Prior CPD All Patients All Patients N ϭ 133 N ϭ 366 N ϭ 192 N ϭ 632 n (%) n (%) n (%) n (%) Hemoptysis or pleuritic pain 55 (41) 196 (54)‡ 84 (44) 357 (56)§ Uncomplicated dyspnea 48 (36) 93 (25)‡ 70 (36) 163 (26)§ Circulatory collapse 11 (8) 21 (6) 15 (8) 33 (5) Different presentation 19 (14)* 56 (15) 23 (12)† 79 (13) CPD ϭ cardiopulmonary disease; PE ϭ pulmonary embolism. *Tachypnea or tachycardia and signs or symptoms of DVT in 3; PaO2 Ͻ 80 mm Hg and signs or symptoms of DVT in 1. †Tachypnea or tachycardia and signs or symptoms of DVT in 8; PaO2 Ͻ 80 mm Hg and signs or symptoms of DVT in 1. ‡P Ͻ.025. §P Ͻ.01.

tachypnea, tachycardia, or a PaO2 less than 80 mm Hg with 32% of all patients with pulmonary embolism and 80 mm signs or symptoms of deep venous thrombosis (Table 1). Hg or more in 38% of patients with no prior cardiopulmonary disease. The A-a oxygen difference was 20 mm Hg or Partial Pressure of Oxygen in Arterial Blood less in 32% of all patients with pulmonary embolism and in and Alveolar-Arterial Oxygen Difference 35% of patients with pulmonary embolism and no prior cardiopulmonary disease (Table 2). The partial pressure of oxygen in arterial blood (PaO2) while breathing room air was measured in 74 patients with pulmonary embolism and in 48 patients with pulmonary em- Risk Factors for Pulmonary Embolism bolism and no prior cardiopulmonary disease (Table 2). The Immobilization (bed rest within past month for the most of Ն PaO2 while breathing room air was 80 mm Hg or more in the day for 3 consecutive days) was the most frequent risk

Table 2 Arterial Blood Gases and Alveolar-Arterial Oxygen Difference While Breathing Room Air PE No PE PE No PE No Prior CPD No Prior CPD All Patients All Patients N ϭ 48 N ϭ 88 N ϭ 74 N ϭ 186 n (%) n (%) n (%) n (%)

PaO2 (mm Hg) Յ49 1 (2) 2 (2) 4 (5) 17 (9) 50-59 6 (13) 12 (14) 12 (16) 32 (17) 60-69 15 (31) 14 (16)* 20 (27) 35 (19) 70-79 8 (17) 13 (15) 14 (19) 32 (17) Ն80 18 (38) 47 (53) 24 (32) 70 (38)

PACO2 (mm Hg) Յ35 30 (63) 39 (44)* 42 (57) 65 (35)‡ 36-39 12 (25) 17 (19) 18 (24) 39 (21) Ն40 6 (13) 32 (36) 14 (19) 82 (44)§ pH (units) Ͻ7.35 0 (0) 7 (8)* 0 (0) 13 (7)† 7.35-7.45 29 (60) 60 (68) 41 (55) 131 (70)† Ͼ7.45 19 (40) 21 (24) 33 (45) 42 (23)§

A-a O2 difference (mm Hg) Յ20 17 (35) 44 (50) 24 (32) 70 (38) 21-30 4 (8) 10 (11) 5 (7) 32 (17)* 31-40 11 (23) 13 (15) 18 (24) 30 (16) 41-50 9 (19) 13 (15) 14 (19) 32 (17) 51-60 5 (10) 6 (7) 10 (14) 17 (9) Ն61 2 (4) 2 (2) 3 (4) 5 (3) CPD ϭ cardiopulmonary disease; PE ϭ pulmonary embolism. *P Ͻ.05. †P Ͻ.025. ‡P Ͻ.01. §P Ͻ.001. 874 The American Journal of Medicine, Vol 120, No 10, October 2007

Table 3 Risk Factors PE No PE PE No PE No Prior CPD No Prior CPD All Patients All Patients N ϭ 131-133 N ϭ 361-366 N ϭ 185-192 N ϭ 588-632 n (%) n (%) n (%) n (%) Age 56 Ϯ 16 47 Ϯ 16ʈ 57 Ϯ 17 50 Ϯ 17ʈ Sex (female) 82 (62) 225 (61) 112 (58) 394 (62) Immobilization* 27 (20) 60 (16) 48 (25) 121 (19) Travel Ն 4 h in last month 19 (14) 74 (20) 23 (12) 105 (17) Surgery (Յ3 mo) 30 (23)‡ 53 (14)** 41 (21)§ 83 (13)¶ Malignancy (excluding lung cancer)† 29 (22) 44 (12)# 37 (19) 82 (13)** Thrombophlebitis, ever 11 (8) 10 (3)# 19 (10) 27 (4)# Trauma (Յ3 mo) Lower extremities and pelvis 14 (11) 27 (7) 16 (8) 49 (8) Other 4 (3) 14 (4) 5 (3) 21 (3) Smoke (ever) 55 (42) 156 (43) 90 (47) 321 (51) Ͻ1 pack/d 29 (22) 74 (20) 43 (22) 122 (19) 1-2 packs/d 20 (15) 62 (17) 37 (19) 152 (24 Ͼ2 packs/d 0 (0) 5 (1) 1 (1) 19 (3) Central venous instrumentation (Յ3 mo) 12 (9) 17 (5) 22 (12) 35 (6)# Stroke, paresis, or paralysis 4 (3) 10 (3) 7 (4) 17 (3) Prior pulmonary embolism – – 7 (4) 19 (3) Heart failure – – 10 (5) 60 (10) COPD – – 10 (5) 58 (9) Lung cancer – – 5 (3) 8 (1) CPD ϭ cardiopulmonary disease; PE ϭ pulmonary embolism; COPD ϭ chronic obstructive pulmonary disease. *Bed rest within past month for the most of the day for Ն3 consecutive days. †Actively treated in last 3 months. ‡Among patients with PE and no prior CPD who had surgery as a risk factor (n ϭ 30), the prevalence of heart, abdominal, pelvic, hip/knee-open, hip/knee-replacement, and neurosurgery ranged from 3 to 5. §Among all patients with PE who had surgery as a risk factor (n ϭ 41), 9 had abdominal surgery and heart, pelvic, hip/knee-open, hip/knee- replacement, and neurosurgery ranged from 3 to 5. ʈP Ͻ.0001. ¶P Ͻ.001. #P Ͻ.01. **P Ͻ.05. factor assessed in patients with pulmonary embolism, and with pulmonary embolism (Table 5). The onset was within surgery was the usual cause of immobilization (Table 3). seconds, minutes, or hours in 83% of patients with pulmo- One or more of the assessed risk factors were reported in nary embolism and no prior cardiopulmonary disease and in 92% of patients with pulmonary embolism and no prior 87% of all patients with pulmonary embolism. In some, cardiopulmonary disease. Among all patients with pulmo- however, the onset of dyspnea occurred over days. nary embolism, 94% had 1 or more of the assessed risk Pleuritic chest pain was more frequent than hemoptysis factors. (Table 4). Cough, when present, was usually nonproductive, but purulent sputum and clear sputum also were reported. Symptoms of Pulmonary Embolism Hemoptysis may have been pinkish, blood streaked, or all New dyspnea at rest or on exertion was the most frequent blood. Hemoptysis of pure blood occurred in only 1 patient symptom in patients with pulmonary embolism and no prior with pulmonary embolism and no prior cardiopulmonary cardiopulmonary disease (73%) (Table 4). Dyspnea only on disease, and it was less than 1 teaspoonful. Thigh pain and exertion was observed in 16% of patients with pulmonary swelling were rarely described in the absence of calf pain or embolism and no prior cardiopulmonary disease and in 16% swelling. of all patients with pulmonary embolism (Table 4). Two- pillow orthopnea was often present (Table 4). Orthopnea Signs of Pulmonary Embolism occurred in 37 of 97 patients (38%) with dyspnea who had Tachypnea was present in approximately one half of the pulmonary embolism and no prior cardiopulmonary disease patients with pulmonary embolism (Table 6). Tachycardia and in 11 of 21 patients (52%) with dyspnea only on was present in approximately one fourth. Clinical evidence exertion. of pulmonary hypertension (accentuated pulmonary compo- The onset of dyspnea was rapid (within seconds or min- nent of the second sound), right ventricular pressure over- utes) in 72% of patients with pulmonary embolism and no load or enlargement (right ventricular lift), or elevated right prior cardiopulmonary disease, and in 67% of all patients atrial pressure (jugular venous distension) was shown in Stein et al Clinical Characteristics of Pulmonary Embolism 875

Table 4 Symptoms of Pulmonary Embolism PE No PE PE No PE No Prior CPD No Prior CPD All Patients All Patients N ϭ 127-133 N ϭ 361-366 N ϭ 184-191 N ϭ 622-632 n (%) n (%) n (%) n (%) Dyspnea Dyspnea (rest or exertion) 97 (73) 248 (68) 151 (79) 459 (73) Dyspnea (at rest)# 73 (55) 167 (46) 117 (61) 338 (54) Dyspnea (exertion only)# 21 (16) 73 (20) 31 (16) 111 (18) Orthopnea (Ն2-pillow) 37 (28) 88 (24) 69 (36) 220 (35) Pleuritic pain 58 (44) 207 (57)ٙ 89 (47) 376 (59)ٙ Chest pain (not pleuritic) 25 (19) 80 (22) 33 (17) 130 (21) Cough 45 (34)* 103 (28)** 82 (43)† 248 (39)†† Wheezing 27 (21) 66 (18) 58 (31) 193 (31) Calf or thigh swelling 52 (41) 62 (17)ٙٙ 72 (39) 126 (20)ٙٙ Calf and thigh swelling 9 (7) 14 (4) 15 (8) 35 (6) Calf or thigh pain 56 (44) 83 (23)ٙٙ 78 (42) 156 (25)ٙٙٙ Calf and thigh pain 22 (17) 24 (7)ٙٙ 30 (16) 61 (10)ٙٙٙ CPD ϭ cardiopulmonary disease; PE ϭ pulmonary embolism. *Hemoptysis, PE, no CPD: 2 ϭ slightly pinkish, 4 ϭ blood-streaked, 1 ϭ all blood (Ͻ1 tsp). **Hemoptysis, no PE, no CPD: 1 ϭ slightly pink, 2 ϭ streaked, 7 ϭ all blood (1 patient, too little to quantify; 1 patient, Ͻ1 tsp; 4 patients, 1 tsp to 1⁄2 cup; 1 patient, Ͼ1⁄2 cup). †Hemoptysis, PE, all patients: 3 ϭ slightly pinkish, 6 ϭ blood-streaked, 2 ϭ all blood (Ͻ1 tsp). ††Hemoptysis, No PE, all patients: 7 ϭ slightly pinkish, 9 ϭ blood streaked, 9 ϭ all blood (1 patient, too little to quantify; 3 patients, Ͻ1 tsp; 4 patients, 1 tsp to 1⁄2 cup; 1 patient, Ͼ1⁄2 cup). #Information not available in some. .ٙP Ͻ.01 .ٙٙP Ͻ.001 .ٙٙٙP Ͻ.025

21% of patients with no prior cardiopulmonary disease and Combinations of Signs and Symptoms in 22% of all patients with pulmonary embolism. Lung Either dyspnea or tachypnea was present in 84% of patients examination was abnormal in 29% of patients with pulmo- with pulmonary embolism and no prior cardiopulmonary nary embolism and no prior cardiopulmonary disease and in disease, and in 86% of all patients with pulmonary embo- 37% of all patients with pulmonary embolism. Crackles and lism. Dyspnea, tachypnea, or pleuritic pain was present in decreased breath sounds were the most frequent lung ﬁnd- 92% of patients with pulmonary embolism and no prior ings. Rhonchi and wheezes occurred uncommonly. Signs of cardiopulmonary disease and in 92% of all patients with deep venous thrombosis (edema, erythema, tenderness, or pulmonary embolism. One or more of these signs and symp- palpable cord) in the thigh, in the absence of deep venous toms or signs of deep venous thrombosis were present in thrombosis in the calf, were rare (Table 6). Among all 98% of patients with pulmonary embolism and no prior patients with pulmonary embolism, calf swelling plus pain cardiopulmonary disease, and in 97% of all patients with with palpation of the deep veins occurred in 32%. pulmonary embolism.

Table 5 Rate of Onset of Dyspnea Patients with Patients with Dyspnea and PE Dyspnea and No PE All Patients with All Patients with No Prior CPD No Prior CPD Dyspnea and PE Dyspnea and No PE N ϭ 92 N ϭ 242 N ϭ 143 N ϭ 450 n (%) n (%) n (%) n (%) Seconds 42 (46) 109 (45) 59 (41) 206 (46) Minutes 24 (26) 69 (29) 37 (26) 117 (26) Hours 10 (11) 35 (14) 20 (14) 70 (16) Days 16 (17) 29 (12) 27 (19) 57 (13) CPD ϭ cardiopulmonary disease; DVT ϭ deep vein thrombosis; PE ϭ pulmonary embolism. All differences not signiﬁcant. 876 The American Journal of Medicine, Vol 120, No 10, October 2007

Table 6 Signs of Pulmonary Embolism PE No PE PE No PE No Prior CPD No Prior CPD All Patients All Patients N ϭ 128-132 N ϭ 350-365 N ϭ 184-191 N ϭ 602-629 n (%) n (%) n (%) n (%) General Tachypnea (Ն20/min) 71 (54) 155 (43)# 108 (57) 296 (47)§ Tachycardia (Ͼ100/min) 32 (24) 52 (14)§ 49 (26) 98 (16)§ Diaphoresis 3 (2) 27 (7)** 8 (4) 40 (6) Cyanosis 0 (0) 1 (0.003) 1 (1) 1 (0) Temperature Ͼ 38.5oC(Ͼ101.3oF) 1 (1) 12 (3) 3 (2) 14 (2) Cardiac examination (abnormal) 28 (21) 39 (11)§ 42 (22) 72 (12)ʈ Increased P2† 15 (15) 14 (5)¶ 22 (15) 27 (5)ʈ Right ventricular lift‡ 4 (4) 6 (2) 8 (5) 9 (2)# Jugular venous distension 18 (14) 27 (8)** 25 (13) 50 (8)** Lung examination (abnormal) 38 (29) 94 (26) 70 (37) 227 (36) Rales (crackles) 23 (18) 52 (14) 40 (21) 112 (18) Wheezes 2 (2) 12 (3) 6 (3) 54 (9)# Rhonchi 2 (2) 8 (2) 9 (5) 32 (5) Decreased breath sounds 22 (17) 42 (12) 40 (21) 109 (17) Pleural friction rub 0 (0) 3 (1) 2/ (1) 5 (1) DVT signs†† Calf or thigh 62 (47)* 77 (21)ʈ 90 (47) 146 (23)ʈ Calf and thigh 18 (14) 16 (4)ʈ 23 (12) 30 (5)ʈ CPD ϭ cardiopulmonary disease; PE ϭ pulmonary embolism; P2 ϭ pulmonary component of second sound; DVT ϭ deep venous thrombosis. *Number of patients with PE and no CPD who had 1 or more signs of DVT: edema ϭ 55, erythema ϭ 5, tenderness ϭ 32, palpable cord ϭ 2. †Data in 103 patients with PE and no CPD, 293 with no PE and no CPD, 145 with PE all patients, 512 no PE all patients. ‡Data in 110 patients with PE and no CPD, 301 with no PE and no CPD, 155 with PE all patients, 529 no PE all patients. §P Ͻ.01. ʈP Ͻ.001. ¶P Ͻ.0001. #P Ͻ.025. **P Ͻ.05. ††Edema, erythema, tenderness, or palpable cord.

Patients with Circulatory Collapse embolism in proximal arteries, 94% presented one of the Among patients with circulatory collapse with pulmonary typical syndromes (hemoptysis/pleuritic pain syndrome, un- embolism and no prior cardiopulmonary disease, dyspnea complicated dyspnea syndrome, or circulatory collapse syn- was present in 9 of 11 (82%), dyspnea or tachypnea was drome), whereas in patients with segmental pulmonary em- present in 10 of 11 (91%), and dyspnea, tachypnea, or bolism, only 72% had one of these presentations. The others pleuritic pain was present in 10 of 11 (91%). All 11 patients with segmental emboli had only calf swelling. had dyspnea, tachypnea, pleuritic pain, or signs of deep Dyspnea or tachypnea occurred in 92% of all patients venous thrombosis. with pulmonary embolism in whom the pulmonary embo- Among all patients with circulatory collapse and pulmo- lism was proximal, but in only 65% with segmental pulmonary embolism, dyspnea was present in 13 of 15 (87%), nary embolism. Dyspnea, tachypnea, or pleuritic pain oc- dyspnea or tachypnea was present in of 14 of 15 (93%), and curred in 97% of patients with proximal pulmonary dyspnea, tachypnea, or pleuritic pain was present in 14 of 15 embolism and in 77% of patients with segmental pulmonary (93%). All 15 patients had dyspnea, tachypnea, pleuritic embolism. Dyspnea at rest or during exertion, dyspnea at pain, or signs of deep venous thrombosis. rest, orthopnea, tachypnea, and PACO2 of 35 mm Hg or less were more frequent in patients with proximal pulmonary Clinical Characteristics According to Location embolism, and PACO2 of 40 mm Hg or more was less of Pulmonary Embolism frequent than in patients with segmental pulmonary Among 150 patients with pulmonary embolism in whom embolism. images were classiﬁable, main or lobar (proximal) pulmonary arteries showed pulmonary embolism by computed Signs, Symptoms, and Combinations tomography angiography in 116 (77%). The largest affected According to Age branch was segmental in 32 patients (21%) and subsegmen- Most symptoms and all signs occurred with similar frequen- tal in 2 patients (1%). Among all patients with pulmonary cies in patients aged 70 years or more and younger patients Stein et al Clinical Characteristics of Pulmonary Embolism 877

Table 7 Symptoms in Patients with Pulmonary Embolism Table 9 Signs in Patients with Pulmonary Embolism and and No Preexisting Cardiac or Pulmonary Disease According No Preexisting Cardiac or Pulmonary Disease According to to Age Age

Ն70 Y Ͻ70 Y Ն70 Ͻ70 N ϭ 33-35 N ϭ 93-98 N ϭ 33-35 N ϭ 91-98 n (%) N (%) n (%) n (%) Dyspnea General Dyspnea (rest or exertion) 23 (66) 74 (76) Tachypnea (Ն20 min) 16 (47) 55 (57) Dyspnea (at rest)* 17 (49) 56 (58) Tachycardia (Ͼ100 min) 8 (23) 24 (25) Dyspnea (exertion only)* 5 (14) 16 (16) Diaphoresis 1 (3) 2 (2) Orthopnea (Ն2-pillow) 8 (23) 29 (30) Cyanosis 0 (0) 0 (0) Pleuritic pain 12 (35) 45 (46) Temperature Ͼ 38.5oC(Ͼ101.3oF) 0 (0) 3 (3) Chest pain (not pleuritic) 6 (18) 20 (20) Cardiac examination (any) 8 (23) 20 (21) Cough 10 (29) 35 (36) Increased P2* 3 (12) 12 (16) Wheezing 3 (9) 24 (24) Right ventricular lift† 0 (0) 4 (5) Calf or thigh swelling 9 (26) 43 (44) Jugular venous distension 7 (20) 11 (12) Calf or thigh pain 9 (26) 46 (47)† Lung examination (any) 15 (43) 23 (23)‡ Rales (crackles) 9 (26) 14 (15) *Information not available in some. Wheezes 1 (3) 1 (1) †P Ͻ.05 age Ն 70 years vs Ͻ 70 years. All other differences be- Rhonchi 0 (0) 2 (2) tween age groups are not signiﬁcant. Decreased breath sounds 8 (23) 14 (15) Pleural friction rub 0 (0) 0 (0) *Data in 26 patients Ն 70 years, 77 patients Ͻ 70 years. (Tables 7 to 10). In patients with no prior cardiopulmonary †Data in 30 patients Ն 70 years, 80 patients Ͻ 70 years. disease and in all patients with pulmonary embolism, the ‡P Ͻ.05 age Ն 70 years vs Ͻ 70 years. All other differences be- combination of dyspnea or tachypnea occurred less fre- tween age groups are not signiﬁcant. quently in elderly patients than in younger patients.

Objective Clinical Assessment in Patients patients, and patients who presented with the various syn- with Pulmonary Embolism dromes. Among patients with pulmonary embolism in the The majority of patients with pulmonary embolism (113/ main or lobar pulmonary arteries in whom an objective 176, 64%) had a moderate probability of pulmonary embo- clinical assessment was recorded, 16 of 107 (15%) had a lism by the Wells clinical scoring system.7 The remaining low-probability objective clinical assessment by the Wells 7 patients with pulmonary embolism were equally divided scoring system. with a high-probability clinical assessment (32/176,18%) and a low-probability clinical assessment (31/176,18%). Comparable proportions were found among patients with no Table 10 Signs in All Patients with Pulmonary Embolism prior cardiopulmonary disease, elderly patients, younger Ն70 y Ͻ70 y N ϭ 52-55 N ϭ 130-137 n (%) n (%) Table 8 Symptoms in All Patients with Pulmonary General Embolism According to Age Tachypnea (Ն20 min) 28 (51) 80 (59) Ͼ Ն70 Y Ͻ70 Y Tachycardia ( 100 min) 11 (21) 38 (28) N ϭ 53-55 N ϭ 130-137 Diaphoresis 1 (2) 7 (5) Cyanosis 0 (0) 1 (1) n (%) n (%) Temperature Ͼ 38.5oC 0 (0) 3 (2) Dyspnea (Ͼ101.3oF6) Dyspnea (rest or exertion) 41 (75) 110 (80) Cardiac examination (any) 12 (22) 30 (23) Dyspnea (at rest)* 33 (60) 84 (61) Increased P2* 3 (7) 19 (18) Dyspnea (exertion only)* 7 (13) 24 (18) Right ventricular lift† 2 (4) 6 (5) Orthopnea (Ն2-pillow) 17 (31) 52 (39) Jugular venous distension 10 (19) 15 (11) Pleuritic pain 18 (33) 71 (52)† Lung examination (any) 25 (45) 44 (32) Chest pain (not pleuritic) 7 (13) 26 (19) Rales (crackles) 14 (26) 36 (27) Cough 24 (44) 58 (43) Wheezes 2 (4) 4 (3) Wheezing 13 (25) 45 (33) Rhonchi 3 (6) 6 (4) Calf or thigh swelling 14 (26) 61 (46)† Decreased breath sounds 16 (29) 24 (18) Calf or thigh pain† 15 (28) 62 (46) Pleural friction rub 1 (2) 1 (1) *Information not available in some. All differences between age groups not signiﬁcant. †P Ͻ.025 age Ն 70 years vs Ͻ 70 years. All other differences be- *Data in 42 patients Ն 70 years, 103 patients Ͻ 70 years. tween age groups are not signiﬁcant. †Data in 45 patients Ն 70 years, 110 patients Ͻ 70 years. 878 The American Journal of Medicine, Vol 120, No 10, October 2007

DISCUSSION Signs of deep venous thrombosis in patients with no The data show a broad range of severity of clinical findings prior cardiopulmonary disease were more frequent in in patients with pulmonary embolism. The syndrome of PIOPED II than in PIOPED (47% vs 11%), as were symp- pleuritic pain or hemoptysis, in the absence of circulatory toms of deep venous thrombosis (44% vs 26%). However, collapse, was the most frequent mode of presentation in in PIOPED II the frequency of signs of deep venous throm- PIOPED, occurring in 65% of patients with pulmonary bosis (41%) and symptoms of deep venous thrombosis 10 (39%) were similar to those in the Urokinase Pulmonary embolism and no prior cardiopulmonary disease. The 10,13 present data from PIOPED II showed somewhat fewer pa- Embolism Trial. Dyspnea, tachypnea, pleuritic pain, or signs of deep tients with pleuritic pain or hemoptysis, and more had the venous thrombosis were seen in the majority of patients uncomplicated dyspnea syndrome. Circulatory collapse was with pulmonary embolism in PIOPED9,10 and in the an uncommon mode of presentation in PIOPED10 and present data from PIOPED II. Conversely, in the absence PIOPED II because of selection criteria. Patients with he- of dyspnea, tachypnea, pleuritic pain, or signs of deep moptysis or pleuritic pain syndrome have been shown to venous thrombosis, pulmonary embolism was infre- have less severe pulmonary embolism than patients with quently diagnosed.9,10 uncomplicated dyspnea according to an objective pulmo- 13 The diagnosis of pulmonary embolism among elderly nary angiography scoring system. Patients with circula- patients has been thought to be particularly difficult because tory collapse had the most severe pulmonary embolism the expected signs and symptoms may be absent or ig- based on the angiographic score, but the score was not nored.18-20 This did not seem to be the case in the experi- statistically significantly more than in patients with uncom- ence of PIOPED12 or in the present experience of PIOPED 13 plicated dyspnea. The absence of dyspnea did not exclude II, although among patients with pulmonary embolism the pulmonary embolism, even in patients with circulatory combination of dyspnea or tachypnea was present in fewer collapse.11 elderly patients than younger patients. In the absence of Typically, among patients with acute pulmonary embo- dyspnea or tachypnea among elderly patients in PIOPED, 10 lism, the PaO2 is low. However, acute pulmonary embo- unexplained radiographic abnormalities were important di- 12 lism cannot be excluded on the basis of a normal PaO2. This agnostic clues. When the diagnosis of pulmonary embo- was shown in the present study and in PIOPED,10 in which lism is uncertain, computed tomography angiography can 26% of such patients with acute pulmonary embolism and be performed with the same sensitivity and specificity in no prior cardiopulmonary disease who had measurements of elderly patients as in younger patients,21 although renal the PaO2 while breathing room air had a PaO2 of 80 mm Hg failure was a problem among elderly patients who under- or more. Even among patients with submassive or massive went conventional angiography.12 acute pulmonary embolism in the Urokinase Pulmonary 16 Embolism Trial, 12% had a PaO2 of 80 mm Hg or more. CONCLUSIONS In patients with pulmonary embolism breathing room air, Symptoms may be mild and generally recognized symptoms 17 the A-a oxygen difference closely correlates with the PaO2 may be absent in patients with the largest pulmonary em- and has no greater diagnostic value. bolism in the segmental pulmonary branches, but typical The present data show that patients with pulmonary symptoms may be absent even in patients with large emboli. embolism may have dyspnea only on exertion. Orthopnea A high or intermediate-probability objective clinical assess- also was shown to be a symptom of pulmonary embolism. ment may suggest the need for diagnostic studies, but a Orthopnea occurred in patients with pulmonary embolism low-probability objective clinical assessment was some- who had dyspnea only on exertion and in those with dys- times present, even in patients with proximal pulmonary pnea at rest. Typically, the onset of dyspnea occurred over embolism. Maintenance of a high level of suspicion is seconds, minutes, or hours, but in some it occurred over critical for the identification of patients in whom diagnostic days. Compared with patients who had only segmental tests may be necessary. pulmonary artery pulmonary embolism, patients with proximal pulmonary embolism (main or lobar pulmonary em- ACKNOWLEDGMENT bolism) more often had typical signs, symptoms, and blood Nikunj R. Patel, MD, assisted in analyzing the data. gases. Patients with pulmonary embolism even in the main or lobar pulmonary arteries may have a low-probability References objective clinical assessment. 1. Goldhaber SZ, Hennekens CH, Evans DA, et al. Factors associated In both PIOPED and PIOPED II, pleuritic chest pain was with correct antemortem diagnosis of major pulmonary embolism. more frequent in patients with pulmonary embolism than Am J Med. 1982;73:822-826. hemoptysis.9,10 Hemoptysis, when present, occurred only in 2. Stein PD, Henry JW. Prevalence of acute pulmonary embolism among small amounts. Examination of the lungs was abnormal in a patients in a general hospital and at autopsy. Chest. 1995;108:978-981. 3. Rubenstein I, Murray D, Hoffstein V. Fatal pulmonary emboli in minority (29% with no prior cardiopulmonary disease) of hospitalized patients. An autopsy study. Arch Intern Med. 1988;148: patients with pulmonary embolism. 1425-1426. Stein et al Clinical Characteristics of Pulmonary Embolism 879

4. Attems J, Arbes S, Bohm G, et al. The clinical diagnostic accuracy rate 12. Stein PD, Gottschalk A, Saltzman HA, et al. Diagnosis of acute regarding the immediate cause of death in a hospitalized geriatric pulmonary embolism in the elderly. J Am College Cardiol. 1991;18: population; an autopsy study of 1594 patients. Wien Med Wochenschr. 1452-1457. 2004;154:159-162. 13. Stein PD, Willis PW III, DeMets DL. History and physical examina- 5. Stein PD. Fowler SE, Goodman LR, et al. for the PIOPED II Inves- tion in acute pulmonary embolism in patients without preexisting tigators. Multidetector computed tomography for acute pulmonary cardiac or pulmonary disease. Am J Cardiol. 1981;47:218-223. embolism (PIOPED II). N Engl J Med. 2006;354:2317-2327. 14. Wenger NK, Stein PD, Willis PW III. Massive acute pulmonary 6. Stein PD, Woodard PK, Weg JG, et al. Diagnostic pathways in acute embolism: The deceivingly nonspecific manifestations. JAMA. 1972; pulmonary embolism: recommendations of the PIOPED II investiga- 220:843-844. tors. Am J Med. 2006;119:1048-1055. 15. Guenter CA. Respiratory function of the lungs and blood. In: Guenter 7. Wells PS, Anderson DR, Rodger M, et al. Excluding pulmonary CA, Welch MH, eds. Pulmonary Medicine, 2nd Ed. Philadelphia: JB embolism at the bedside without diagnostic imaging: management of Lippincott; 1982:168. patients with suspected pulmonary embolism presenting to the emer- 16. National Cooperative Study. The Urokinase Pulmonary Embolism gency department by using a simple clinical model and D-dimer. Ann Trial. Circulation. 1973;47/48 (Suppl II):II-81-II-85. Intern Med. 2001;135:98-107. 17. Stein PD, Goldhaber SZ, Henry JW. Alveolar-arterial oxygen gra- 8. Wicki J, Pernerger TV, Junod AF, et al. Assessing clinical probability dient in the assessment of acute pulmonary embolism. Chest. 1995; of pulmonary embolism in the emergency ward: a simple score. Arch 107:139-143. Intern Med. 2001;161:92-97. 18. Taubman LB, Silverstone FA. Autopsy proven pulmonary embolism 9. Stein PD, Saltzman HA, Weg JG. Clinical characteristics of patients among the institutionalized elderly. J Am Geriatr Soc. 1986:34:752- with acute pulmonary embolism. Am J Cardiol. 1991;68:1723-1724. 756. 10. Stein PD, Terrin ML, Hales CA, et al. Clinical, laboratory, roentgen- 19. Morrell MT. The incidence of pulmonary embolism in the elderly. ographic and electrocardiographic findings in patients with acute pul- Geriatrics. 1970;25:138-153. monary embolism and no pre-existing cardiac or pulmonary disease. 20. Busby W, Bayer A, Pathy J. Pulmonary embolism in the elderly. Age Chest. 1991;100:598-603. Ageing. 1988;17:205-209. 11. Stein PD, Henry JW. Clinical characteristics of patients with acute 21. Stein PD, Beemath A, Quinn DA, et al. Usefulness of multidetector pulmonary embolism stratified according to their presenting syn- spiral computed tomography according to age and gender for diagnosis dromes. Chest. 1997;112:974-979. of acute pulmonary embolism. J Cardiol. 2007;99:1303-1305. The American Journal of Medicine (2007) 120, 880-885

CLINICAL RESEARCH STUDY

Thyroid Function Abnormalities during Amiodarone Therapy for Persistent Atrial Fibrillation Elizabeth L. Batcher, MD,a X. Charlene Tang, MD, PhD,b Bramah N. Singh, MD, DSc,a Steven N. Singh, MD,c Domenic J. Reda, PhD,b Jerome M. Hershman, MDa for the SAFE-T Investigators aWest Los Angeles Veterans Affairs Medical Center, Los Angeles, Calif, bEdward Hines, Jr. Veterans Affairs Hospital, Hines, Ill, and cDepartment of Veterans Affairs Medical Center, Washington, DC.

ABSTRACT

BACKGROUND: Many patients receiving amiodarone therapy are male. The long-term risk for amiodarone-induced thyroid dysfunction in these patients has not been systematically and prospectively investigated. The purpose of this study was to determine the extent of amiodarone-induced thyroid dysfunction in a large male cohort. METHODS: This is a substudy of a prospective randomized controlled trial (SAFE-Trial) in which amiodarone, sotalol, and placebo for persistent atrial ﬁbrillation were evaluated. For the purpose of this substudy, sotalol and placebo groups were combined into a control group. Serial thyroid function tests were performed over 1-4.5 years. Of the 665 patients enrolled in the SAFE-Trial, 612 patients were included in this sub-study. RESULTS: Subclinical hypothyroidism, thyroid-stimulating hormone (TSH) level 4.5-10 mU/L, was seen among 25.8% of the amiodarone-treated patients and only 6.6% of controls (P Ͻ.0001). Overt hypothyroidism, TSH level Ͼ10 mU/L, was seen among 5.0% of the amiodarone-treated patients, and only 0.3% of controls (P Ͻ.001). By 6 months, 93.8% of the patients who developed TSH elevations above 10 mU/L on amiodarone had been detected. There was a trend toward a greater proportion of hyperthyroidism, deﬁned as a TSH Ͻ0.35 mU/L, in the amiodarone group compared with the control group (5.3% vs 2.4%, P ϭ .07). CONCLUSIONS: Hypothyroidism developed in 30.8% of older males treated with amiodarone and in only 6.9% of the controls. Hypothyroidism presented at an early stage of therapy. Hyperthyroidism occurred in 5.3% of amiodarone treated patients, and was a subclinical entity in all but 1 case. © 2007 Elsevier Inc. All rights reserved.

KEYWORDS: Amiodarone; Hypothyroidism; Sotalol; Thyroid

Although the unique pharmacologic properties of amioda- dysfunction ranging from elevated serum thyroxine (T4) rone and its varied clinical uses have been known for sev- levels and low triiodothyronine (T3) levels in euthyroid eral decades, its unparalleled effectiveness in maintaining patients1 to overt hypothyroidism or hyperthyroidism. The sinus rhythm has emerged only in recent years. The drug’s primary cause of the thyroid dysfunction is the large major shortcoming has been its propensity to induce disor- iodine load that can cause either hypothyroidism or hy- ders of thyroid function. It causes various forms of thyroid perthyroidism.2,3 Hyperthyroidism can also be caused by a thyroiditis that is iodine-independent, and known as type 2 hyperthyroidism.2,3 Supported by the Cooperative Studies Program of the Department of Studies examining the prevalence of amiodarone-induced Veterans Affairs ofﬁce of Research and Development. thyroid dysfunction vary in regard to geographic location and Requests for reprints should be addressed to Elizabeth Batcher, MD, Department of Medicine, UCLA-Olive View Medical Center, 14445 Olive iodine intake, ratio of women to men, and length of follow-up. View Drive, Sylmar, CA, 91342. Most studies have shown a trend toward more frequent E-mail: [email protected] hypothyroidism in iodine-sufﬁcient areas, such as the

United States, and more frequent thyrotoxicosis in iodine- RESULTS deficient areas, such as Italy.4,5 In general, amiodarone-induced thyroid dysfunction is more common in patients with an Patients underlying thyroid disorder. Rates of thyroid dysfunction have A total of 665 patients with persistent atrial fibrillation were 7 been reported to be 14%-18%, with the highest prevalence enrolled in the SAFE trial. There were 267 patients in the among women and those with anti- amiodarone group, 261 in the so- thyroid antibodies.2-4,6 talol group, and 137 in the placebo group. For this study, 53 patients The data reported in this article CLINICAL SIGNIFICANCE deal with thyroid dysfunction in a were excluded due to the lack of large cohort of male patients treated ● Nearly one third of older men treated with TSH values at either baseline or with amiodarone and compared amiodarone for persistent atrial fibrillation during follow-up. The sotalol and placebo groups were combined to with a control group stemming from developed some degree of hypothyroid- form the control group, as neither a substudy of the Sotalol Amioda- ism, compared with 6.9% of those treated rone Atrial Fibrillation Efficacy placebo nor sotalol would be ex- with sotalol or placebo. 8,9 Trial (SAFE-T).7 The objective of pected to alter thyroid function. the substudy was to determine the ● Amiodarone treatment also was associated Table 1 shows the baseline char- incidence of hypothyroidism, de- with a slight, nonsignificant increase in acteristics of the patients in the fined as thyroid-stimulating hor- hyperthyroidism (5.3% vs 2.4%). amiodarone and control groups. mone (TSH) Ͼ4.5 mU/L, or hyper- Overall, 607 of 612 patients (99%) thyroidism, defined as TSH Ͻ0.35 ● During amiodarone therapy, all patients were male and 5 (0.8%) were fe- mU/L, induced by amiodarone and should be carefully monitored for both male; mean age was 67.0 (Ϯ9.3) compared with sotalol and placebo hypothyroidism and hyperthyroidism. years. Ninety percent of the pa- as the control. tients were white and 5.7% were ● Amiodarone-induced overt hypothyroidism black, without statistical differ- generally warrants levothyroxine treatment. ences in ethnic breakdown be- METHODS tween the groups. There were no Study Design significant differences in the prevalence of diabetes, ischemic heart The details of the study (SAFE-T) design from which the disease, smoking, or baseline TSH level between the groups. data for the current report are derived have been reported Rates of baseline levothyroxine therapy were similar be- previously.7 In brief, patients with persistent atrial fibrilla- tween the control and amiodarone-treated groups, 1.4% tion (AF) were randomly assigned to amiodarone, sotalol, or versus 2.8%, respectively (P ϭ .20). The pairwise compar- placebo after optimal anticoagulation (international normal- isons showed that there were no statistical differences in ized ratio ranging from 2 to 3). After 4 weeks of therapy, baseline TSH between original treatment groups. those not in sinus rhythm (SR) had their AF electrically Sixty-six of 365 patients in the control group received cardioverted. Once SR was achieved, anticoagulation was open-label amiodarone for at least 1 month during fol- continued for at least 8 weeks. Recurrence of AF was low-up and were excluded from the outcome analysis in documented by weekly transtelephonic monitoring. All patients were followed for at least 1 year regardless of their rhythm status. Patients remaining in SR were maintained on their originally assigned treatment. Those who remained in, Table 1 Baseline Patient Characteristics or reverted to, AF after second cardioversion were placed on Amiodarone Control open-label medications with continued follow-up to 1 year. Group Group P Thyroid function was examined at baseline, 3 months, 6 Characteristics (n ϭ 247) (n ϭ 365) Value months, and at every 6 months thereafter by measurement of Age 67.1 Ϯ 9.4 66.9 Ϯ 9.2 .84 serum thyrotropin (TSH) concentration. Free T4 and total Sex T3 levels were not consistently obtained. Male (%) 246 (99.6) 364 (98.9) .65 Body mass index (kg/m2) 31.3 Ϯ 5.8 31.5 Ϯ 5.9 .67 Current smokers (%) 4 (1.6) 3 (0.8) .45 Statistical Analysis Diabetes (%) 62 (25.2) 90 (24.7) .88 Incidence rates of hypothyroidism and hyperthyroidism be- Ischemic HD (%) 64 (25.9) 88 (24.1) .61 tween groups were compared using chi-squared test, or Baseline TSH Fisher’s exact test as appropriate. Logistic regression was (mU/L) (%) Low: Ͻ0.35 5 (2.1) 3 (0.9) .18 used to calculate odds ratios. An alpha value of .05 or less Normal: 0.35 to 4.5 231 (95.0) 334 (93.8) was considered to indicate statistical significance. All High: Ͼ4.5 7 (2.9) 19 (5.3) statistical tests reported were 2-sided. All analyses were Use of levothyroxine (%) 7 (2.8) 5 (1.4) .24 performed using SAS version 8.0 (SAS Institute Inc., HD ϭ heart disease; TSH ϭ thyroid-stimulating hormone. Cary, NC). 882 The American Journal of Medicine, Vol 120, No 10, October 2007

Table 2 Incidence of Hyperthyroidism and Hypothyroidism During Follow-up*

Amiodarone Control Group Group n ϭ 247 n ϭ 299 P TSH Level (mU/L) % (n) % (n) Value Hyperthyroidism (TSH Ͻ0.35) Yes 5.3 (13)† 2.4 (7)† .07 No 94.7 (232) 97.6 (291) Subclinical hypothyroidism (TSH Ͼ4.5-10) Yes 25.8 (62)‡ 6.6 (19)‡ Figure 2 Figure reflects an updated analysis in which 66 pa- No 74.2 (178) 93.4 (271) Ͻ.0001 tients with open-label amiodarone were excluded from the control Overt hypothyroidism Ͼ Ͼ group. Rate of hypothyroidism (TSH 4.5 mU/L) for the amio- (TSH 10) Ͼ Yes 5.0 (12)‡ 0.3 (1)‡ darone and control groups at 3, 6, 12, and 12 months follow-up. No 95.0 (228) 99.7 (289) Ͻ.001 TSH ϭ thyroid-stimulating hormone. *Excluding 66 patients with open-label amiodarone from the con- Incidence of Hypothyroidism and trol group. †Four patients who had low TSH at both baseline and follow-up Hyperthyroidism were excluded from analysis. Table 2 presents the incidence of hypothyroidism and hy- ‡Twenty patients who had elevated TSH at both baseline and perthyroidism for patients in the amiodarone group and the follow-up were excluded from analysis. control group. An elevation in TSH was classified into 2 levels: subclinical hypothyroidism, TSH Ͼ4.5-10 mU/L; and overt hypothyroidism, TSH Ͼ10 mU/L. There was a Table 2, Figure 1, and Figure 2. Those with open-label higher incidence of subclinical hypothyroidism (25.8%) in Ϯ amiodarone tended to be younger (64.8 9.3 years) and had the amiodarone group compared with the control group Ϯ 2 a higher body mass index (33.9 6.1 kg/m ) compared with (6.6%) (P Ͻ.0001). An elevation of TSH Ͼ10 mU/L was ϭ the remaining 299 patients in the control group (P .04, seen among 5.0% of the amiodarone-treated patients and ϭ P .002, respectively). There were no significant differ- only 0.3% of controls (P Ͻ.001), indicating that the patients ences in other demographic and clinical characteristics be- on amiodarone were more likely to develop overt hypothy- tween the patients with open-label amiodarone and the re- roidism than those on placebo or sotalol. As hypothesized, maining patients in the control group. there was no significant difference in the incidence of hypothyroidism between the sotalol- and placebo-treated patients (data not shown). Figure 1 shows the odds ratio of all degrees of hypothyroidism (TSH Ͼ4.5) for all amiodarone-treated patients and each subgroup of these patients in terms of age, ischemic heart disease, diabetes, and duration of AF before the treatment compared with the control group in which the patients remained on sotalol or placebo. Overall, patients on amiodarone had a higher likelihood to develop hypothyroidism compared with those in the control group (odds ratio 4.5, 95% confidence limit 2.8-7.2). The likelihood of developing hypothyroidism in patients on amiodarone was significantly higher than that in patients who were in the control group in all predefined strata. Figure 1 Figure reflects an updated analysis in which 66 pa- There was a trend toward a higher incidence of hyper- tients with open-label amiodarone were excluded from the control thyroidism, defined as a TSH Ͻ0.35 mU/L, in patients on group. Odds ratio (OR) and 95% confidence limits of hypothy- amiodarone compared with those in the control group (5.3% Ͼ roidism (TSH 4.5 mU/L) in the amiodarone group compared vs 2.4%, P ϭ .07) (Table 2). Among the 13 patients with with the control group. Overall, patients on amiodarone had a low TSH levels in the amiodarone group, there were 8 higher odds of hypothyroidism than those who were in the control group (odds ratio 4.5, 95% confidence limit 2.8-7.2). Odds of patients with mild subclinical hyperthyroidism (TSH 0.1- hypothyroidism was significantly higher for the amiodarone group 0.34 mU/L), 4 patients with more significant subclinical than for the control group in each predefined stratum. Twenty hyperthyroidism (TSH Ͻ0.1 mU/L), and 1 patient with patients who had elevated TSH at both baseline and follow-up overt hyperthyroidism requiring immediate treatment with were excluded from analysis. methimazole. Batcher et al Amiodarone and Thyroid Function 883

There were 66 patients in the control group who were rone developed hypothyroidism. Although antibody testing assigned to sotalol or placebo but treated with open-label was not done in our patients, the rate of positive antithyroid amiodarone during follow-up. It was noted that 11 of these 66 antibodies in this predominantly white male cohort is likely patients developed subclinical hypothyroidism and 1 patient to be similar to the 12.9%-14.4% reported among a large developed hyperthyroidism. population of 60-79-year-old white males in the NHANES III (National Health and Nutrition Examination Survey) Time to Elevated TSH study.16 Thus, despite a lower predisposition to thyroid Patients tended to develop hypothyroidism early in the study. dysfunction in men, nearly one third of all patients treated Figure 2 displays the cumulative percentage of the TSH ele- with amiodarone developed some degree of hypothyroid- vations that had been detected by month among the patients ism, compared with 6.9% among the controls (P Ͻ.0001). who developed hypothyroidism. Fifty-eight percent of pa- This incidence of high TSH levels among our control group tients who developed hypothyroidism in the amiodarone is similar to the 8%-12% reported for men and women from group and 50% in the control group were detected at the first 60-79 years of age in the NHANES III study. Because T4 3 months after treatment inception. Hypothyroidism was levels were not consistently obtained during the follow-up detected in approximately 76% of the patients in the amio- period, we cannot determine the rate of overt hypothyroid- darone group and 80% in the control group by 6 months. ism as defined by an elevated TSH level with low FT4. Statistical differences in rates of elevated TSH levels were Instead, TSH levels above 10 mU/L were used as the cutoff not found between the amiodarone and control groups. It for overt hypothyroidism,10 and were found among 5.0% of also was noted that 91.7% of the patients on amiodarone the amiodarone-treated group, compared with 0.3% in the who developed overt TSH elevations (Ͼ10 mU/L) had been control group (P Ͻ.001). detected at 6 months and 88.7% of the milder elevations in Patients who developed elevations in their TSH levels TSH levels (4.5-10 mU/L) had been detected at the end of did so early in both the amiodarone and control groups. At the first year. 6 months, 76% of the cases of hypothyroidism in the amiodarone group and 80% of the cases of hypothyroidism in the Treatment with Levothyroxine control group had been detected. Of the patients on amioda- Patients who developed elevated TSH levels while on ami- rone who developed TSH elevations above 10 mU/L, 92% had odarone were more likely to have been treated with levo- been detected at 6 months. A trend showed that hyperthy- thyroxine than similar patients in the control group. At 6 roidism, defined as TSH Ͻ0.35 mU/L, was more common months and 1 year, 29.7% and 52.7%, respectively, of the among the amiodarone-treated patients. Because hyperthy- patients with elevated TSH levels in the amiodarone group roidism was a subclinical entity in all but 1 patient, and rare were receiving supplemental levothyroxine. In contrast, in absolute numbers, the analysis was directed toward only none of the patients who developed hypothyroidism in the the hypothyroid patients. control group were treated with levothyroxine (P Ͻ.0001). Amiodarone contains 37% iodine by weight. A 200-mg It also was noted that among the amiodarone group a sig- standard daily dose provides about 300 times the usual daily nificantly higher proportion of patients with overt hypothy- iodine intake. With initiation of treatment, the thyroid rap- roidism were treated with levothyroxine than those with idly responds to the iodine load with a reduction in thyroid subclinical hypothyroidism. At 6 months, 25.0% of the hormone synthesis, known as the Wolff-Chaikoff effect. patients with overt hypothyroidism in the amiodarone group Patients with normal thyroid function “escape” from the used levothyroxine, compared with 4.7% of the patients Wolff-Chaikoff effect by reducing thyroid iodide trans- with subclinical hypothyroidism in the same group port.12 Predictable changes of hormone levels that occur ϭ (P .002). Similarly, at 1 year, approximately 43% of pa- within the first week of therapy include a decrease in serum tients in the amiodarone group with overt hypothyroidism T3 levels and increase in serum reverse T3, T4, and TSH were taking levothyroxine, whereas only 9.8% the patients levels due to inhibition of type 1 5=-deiodinase in peripheral with subclinical hypothyroidism were prescribed levothy- tissues and type 2 5=-deiodinase in the pituitary.13 After Ͻ roxine (P .0001). approximately 3 months of amiodarone, the compensatory increase in T4 levels reaches a steady state, and a majority DISCUSSION of patients regain normal TSH levels.1 Through this escape Since amiodarone became available in the United States for mechanism, most patients remain clinically euthyroid while anti-arrhythmic therapy in the 1980s, there have been nu- taking amiodarone.10 However, the large iodine load can merous reports of associated changes in thyroid hormone lead to hypothyroidism in patients who do not escape from metabolism, as well as clinically significant thyroid dys- the Wolff-Chaikoff effect. Additionally, a more recent study function.10-15 Our study shows that amiodarone-induced by Tedelind et al reported that amiodarone and its iodide- hypothyroidism is common among a cohort of older men free analog dronedarone are capable of inhibiting iodide treated for atrial fibrillation as compared with a control uptake by thyroid follicular cells by way of an iodine- group in a blinded study. If subclinical hypothyroidism, TSH independent mechanism that does not involve the sodium- Ͼ4.5-10 mU/L, is included, 31% of the patients on amioda- iodide symporter.14 884 The American Journal of Medicine, Vol 120, No 10, October 2007

Amiodarone-induced hypothyroidism is more common amiodarone, as compared with 2.4% among the control among populations with sufficient iodine intake, such as in group (P ϭ .07). Only 1 of the 13 patients with low TSH the United States. One study comparing an iodine-sufficient levels developed overt hyperthyroidism with clinical fea- area of Massachusetts and an iodine-deficient area of Italy tures of hyperthyroidism and elevated thyroxine level. The found rates of amiodarone-induced hypothyroidism, as de- remaining 12 patients had subclinical hyperthyroidism that fined by elevated TSH levels and low or low-normal thy- did not require antithyroid drugs or cessation of amioda- roxine level, to be 22% in Massachusetts and 5% in Italy.15 rone. Recent prospective and controlled studies have impli- Other risk factors for amiodarone-induced hypothyroidism cated antithyroid antibodies4,6,25 and iodine deficiency in include female sex and the presence of antithyroid antibod- the pathogenesis.15 Our low rate of thyrotoxicosis is not ies.4,6 Antithyroid peroxidase antibodies are more common completely unexpected in an iodine-sufficient area such as among white compared with black and Mexican-American the United States. In one study done in iodine-sufficient individuals; 14.3%, 5.3%, and 10.9 %, respectively.16 Among Massachusetts, 2.0% of patients developed amiodarone-in- a group of 58 patients studied in the Netherlands with an duced thyrotoxicosis compared with 10% in an iodine- 15 overall incidence of amiodarone-induced hypothyroidism of deficient region of Italy. However, the reported preva- 6.9%, female sex and antithyroid antibodies were associated lence of thyrotoxicosis has varied greatly depending on the with relative risks of hypothyroidism of 7.9 and 7.3, respec- group studied. Trip and colleagues followed 58 euthyroid tively. Women with antithyroid antibodies had a combined patients, 72% male, for a mean of 21 months in a moder- 5 relative risk for developing amiodarone-induced hypothy- ately iodine-sufficient area of the Netherlands. The inci- roidism of 13.5.5 dence of amiodarone-induced thyrotoxicosis was 12.1%, Less is known about the development of amiodarone- and occurred at any time during therapy. induced hypothyroidism in men. Albert and colleagues re- In summary, amiodarone-induced hypothyroidism devel- ported the thyroid function tests of a male-predominant oped in 30.8% of older men treated with amiodarone for group on chronic amiodarone therapy followed for a mean chronic atrial fibrillation compared with the control group Ͼ of 27 months.17 Sixty-five percent of the 99 patients were and presented early during therapy. Serum TSH levels 10 mU/L were found in 5.0% of amiodarone-treated men, and male and all were euthyroid at baseline. Hypothyroidism this should warrant treatment with thyroxine. Hyperthyroid- was defined as a TSH level Ͼ20 with a thyroxine level less ism occurred in 5.3% of amiodarone-treated patients, but than the lower limit of normal. In the male subgroup, 30% was a subclinical entity in all but 1 case and not statistically became hypothyroid. different from the rate in the control group. Given the high The laboratory diagnosis of hypothyroidism in patients rate of hypothyroidism among patients taking amioda- on amiodarone therapy has evolved in the past 2 decades. rone, monitoring of thyroid function is recommended at Because of the absence of clinical features of hypothyroid- baseline, 3 months, and every 6 months thereafter during ism and the alteration of thyroid function tests by the block- the therapy. ade of T4 deiodination induced by amiodarone,18 resulting in elevation of serum T4 and free T4, the criterion for diagnosis of hypothyroidism was serum TSH Ͼ14 mU/L.19 References 1. Melmed S, Nademanee K, Reed AW, et al. Hyperthyroxinemia with However, in recent years there has been increasing inter- bradycardia and normal thyrotropin secretion after amiodarone admin- est in the effect of subclinical hypothyroidism on the istration. J Clin Endocrinol Metab. 1981;53:997-1001. cardiovascular system.20-23 Although there is debate about 2. Martino E, Bartalena L, Bogazzi F, Braverman LE. Effects of amio- the benefits of treatment of patients with serum TSH levels darone on the thyroid. Endocr Rev. 2001;22(2):240-254. of 4.5-10 mU/L, there is general agreement that patients 3. Harjai KJ, Licata AA. Effects of amiodarone on thyroid function. Ann Ͼ Intern Med. 1997;126:63-73. with subclinical hypothyroidism and serum TSH levels 10 4. Martino E, Aghini-Lombardi F, Mariotti S, et al. Amiodarone iodine- 24 mU/L will benefit from treatment. It seems reasonable to induced hypothyroidism: risk factors and follow-up in 28 cases. Clin apply this recommendation to patients on amiodarone, and Endocrinol. 1987;26:227-237. in this study, 5.0% of patients on amiodarone warranted 5. Trip MD, Wiersinga W, Plomp TA. Incidence, predictability and pathogenesis of amiodarone induced thyrotoxicosis and hypothyroid- levothyroxine treatment. ism. Am J Med. 1991;91(5):507-511. It is evident from this study that practitioners are hesitant 6. Martino E, Aghini-Lombardi F, Bartalena L, et al. Enhanced suscep- to start levothyroxine therapy in patients on amiodarone tibility to amiodarone-induced hypothyroidism in patients with anuto- despite TSH levels Ͼ10 mU/L, possibly because of concern immune thyroid disease. Arch Intern Med. 1994;23:2722-2726. that the euthyroid state will cause recurrence of the atrial 7. Singh BN, Singh SN, Reda DJ, et al. Amiodarone Atrial Fibrillation Efficacy Trial (SAFE-T) Investigators. Amiodarone versus sotalol for fibrillation. At 1 year, only 43% of patients with this degree atrial fibrillation. Veterans Affairs Cooperative Study. N Engl J Med. of dysfunction on amiodarone were treated with levothy- 2005;352:1861-1872. roxine. With this same level of hypothyroidism, an even 8. Wahlberg P, Wennstrom J, Ekelund P. Control of thyrotoxicosis with lower proportion of patients on sotalol or placebo were sotalol. Ann Clin Res. 1976;8:415-417. 9. Perrild H, Hansen JM, Skovsted L, Christensen LK. Different effects treated with levothyroxine. of propranolol, alprenolol, sotalol, atenolol and metoprolol on se- In regard to amiodarone-induced thyrotoxicosis, 5.3% of rum T3 and serum rT3 in hyperthyroidism. Clin Endocrinol. 1983; patients in this study developed low TSH levels while on 18:139-142. Batcher et al Amiodarone and Thyroid Function 885

10. Nademanee K, Singh BN, Callahan B, et al. Amiodarone, thyroid 18. Hershman JM, Nademanee K, Sugawara M, et al. Thyroxine and hormone indexes and altered thyroid function: long term serial effects triiodothyronine kinetics in cardiac patients taking amiodarone. Acta in patients with cardiac arrhythmias. Am J Cardiol. 1986;58:981-986. Endocrinol. 1986;111:193-199. 11. Kennedy RL, Grifﬁths H, Gray TA. Amiodarone and the thyroid. Clin 19. Nademanee K, Piwonka P, Singh BN, Hershman JM. Amiodarone and Chem. 1989;35:1882-1887. thyroid function. Prog Cardiovasc Dis. 1989;31:427-437. 12. Wiersinga WM, Trip MD. Amiodarone and thyroid hormone metab- 20. Biondi B, Klein I. Hypothyroidism as a risk factor for cardiovascular olism. Postgrad Med J. 1986;62:909-914. disease. Endocrine. 2004;24:1-13. 13. Franklyn JA, Gammage MD, Sheppard MC. Amiodarone and thyroid 21. Rondondi N, Newman AB, Vittinghoff E, et al. Subclinical hypothy- Clin hormone effects on anterior pituitary hormone gene expression. roidism and the risk for heart failure, other cardiovascular events and Endocrinol (Oxf). 1987;27:373-382. death. Arch Intern Med. 2005;165:2460-2466. 14. Tedelind S, Larsson F, Johanson C, et al. Amiodarone inhibits thyroi- 22. Walsh JP, Bremner AP, Bulsara MK, et al. Subclinical thyroid disease dal iodine transport in vitro by a cyclic AMP and iodine independent as a risk factor for cardiovascular disease. Arch Intern Med. 2005;165: mechanism. Endocrinology. 2006;147:2936-2943 (Epub 2006 Mar 9). 15. Martino E, Safran M, Aghini-Lombardi F, et al. Environmental iodine 2467-2472. intake and thyroid dysfunction during chronic amiodarone therapy. 23. Cappola A, Fried L, Arnold A, et al. Thyroid status, cardiovascular Ann Intern Med. 1984;101:28-34. risk and mortality in older adults. JAMA. 2006;295:1033-1041. 16. Hollowell JG, Staehling NW, Flanders D, et al. Serum TSH, T4, and 24. Surks MI, Ortiz E, Daniels GH, et al. Subclinical thyroid disease: thyroid antibodies in the United States population (1988 to 1994): scientiﬁc review and guidelines for diagnosis and management. JAMA. National Health and Nutrition Examination Survey (NHANES III). 2004;291:228-238. J Clin Endocrinol Metab. 2002;87:489-499. 25. Martino E, Macchia E, Aghini-Lombardi F, et al. Is humoral thyroid 17. Albert SG, Alves LE, Rose EP. Thyroid dysfunction during chronic autoimmunity relevant in amiodarone iodine-induced thyrotoxicosis amiodarone therapy. J Am Coll Cardiol. 1987;9:175-183. AIIT? Clin Endocrinol. 1986;24:627-633. The American Journal of Medicine (2007) 120, 886-892

CLINICAL RESEARCH STUDY Randomized Trial to Improve Fracture Prevention in Nursing Home Residents Cathleen S. Colón-Emeric,a,b Kenneth W. Lyles,a,b Paul House,c Deborah A. Levine,d Anna P. Schenck,c Jeroan Allison,d Joel Gorospe,c Mary Fermazin,e Kristi Oliver,d Jeffrey R. Curtis,d Norman Weissman,d Aiyuan Xie,d Kenneth G. Saagd aDuke University Center for Aging and Human Development, Durham, NC; bDurham VA Geriatric Research, Education, and Clinical Center (GRECC), Durham, NC; cCarolinas Center for Medical Excellence, Cary, NC; dCenter for Education and Research on Therapeutics (CERTs) of Musculoskeletal Disorders, University of Alabama at Birmingham, Birmingham, Ala; eHealth Services Advisory Group, Phoenix, Ariz.

ABSTRACT

BACKGROUND: Interventions to improve the fracture prevention in nursing homes are needed. METHODS: Cluster-randomized, single-blind, controlled trial of a multi-modal quality improvement intervention. Nursing homes (n ϭ 67) with Ն10 residents with a diagnosis of osteoporosis or recent hip fracture (n ϭ 606) were randomized to receive an early or delayed intervention consisting of audit and feedback, educational modules, teleconferences, and academic detailing. Medical record abstraction and the Minimum Data Set were used to measure the prescription of osteoporosis therapies before and after the intervention period. Analysis was at the facility-level and Generalized Estimating Equation modeling was used to account for clustering. RESULTS: No significant improvements were observed in any of the quality indicators. The use of osteoporosis pharmacotherapy or hip protectors improved by 8.0% in the intervention group and 0.6% in the control group, but the difference was not statistically significant (P ϭ .72). Participation in the intervention activities was low, but completion of the educational module (odds ratio [OR] 4.8, 95% confidence interval [CI], 1.9-12.0) and direct physician contact by an academic detailer (OR 4.5, 95% CI, 1.1-18.2) were significantly associated with prescription of osteoporosis pharmacotherapy or hip protectors in multivariable models. CONCLUSIONS: Audit-feedback and education interventions were ineffective in improving fracture prevention in the nursing home setting, although results may have been tempered by low participation in the intervention activities. © 2007 Elsevier Inc. All rights reserved.

KEYWORDS: Nursing homes; Osteoporosis; Quality improvement

Fracture prevention in nursing homes is important for several reasons. With the prevalence of low bone mineral 1,2 3 Supported by the Alliance for Better Bone Health, AG-11268, density near 90%, and a high rate of falls, the nursing 3U18HS10389-06S1, K23 AG024787. The analyses upon which this publi- home population has one of the highest fracture rates at 13 cation is based were performed under Contract Number 500-02-AZ02 fractures/100 person-years.1,4 In addition, 60%-75% of pa- AZ0020, funded by the Centers for Medicare & Medicaid Services (CMS), an tients with an acute hip fracture spend time in nursing agency of the U.S. Department of Health and Human Services. The content of facilities.5 This group also has an elevated risk of subse- this publication does not necessarily reﬂect the views or policies of the 6,7 Department of Health and Human Services, nor does mention of trade names, quent fractures. Clinical trials suggest that bisphospho- commercial products, or organizations imply endorsement by the U.S. Gov- nates are safe and effective in nursing home residents.8 ernment. The author assumes full responsibility for the accuracy and com- Although few clinical practice guidelines speciﬁcally ad- pleteness of the ideas presented. Publication No. AZ-8SOW-SS-061206-03. dress the nursing home population, pharmacologic therapies Requests for reprints should be addressed to Cathleen S. Colón-Emeric, MD, MHSc, Box 3003, Duke University Medical Center, Durham, NC 27710. are recommended for most hip fracture patients and selected E-mail address: [email protected] long-term care residents at the highest risk.9

Despite the importance of fracture prevention in the least 4 weeks. In order to obtain a sample of residents most nursing home, previous studies have demonstrated low rates likely to be candidates for the full range of fracture prevention of evaluation and treatment.10-13 We reported that only one therapies, we excluded residents with active cancer, severe third of ambulatory nursing home residents known to have dementia, end-stage renal disease, extensive assistance in osteoporosis or recent hip fracture receive any osteoporosis physical functioning, hospice care, or estimated life expect- pharmacologic therapy or external ancy Յ6 months. Although frac- hip protectors.14 Moreover, there ture prevention may be warranted is substantial variation in the per- CLINICAL SIGNIFICANCE in many of the excluded residents, formance of nursing homes (range including them in our sample of 0%-85% of residents receiving ● In nursing homes, many patients at high would lead to lower rates of re- therapies), suggesting that sub- risk of fracture do not receive the frac- sponse to the intervention because stantial improvements are possible some residents would be deemed 14 ture prevention therapies (bisphospho- in many facilities. nates and hip protectors) indicated by unlikely to benefit from most of Changing provider behavior is, the therapies. clinical practice guidelines. however, a challenging proposition. Studies in other settings have iden- ● In a study of 67 nursing homes, qual- Intervention tified interventions that are some- ity-improvement interventions (audit Nursing home staff in the inter- 15 times effective, such as contact and feedback, educational modules, vention group received: continu- with opinion leaders, audit and ing educational modules on osteo- 16 17 teleconferences, and academic detail- feedback, faxed reminders, and ing) only slightly (and nonsignifi- porosis evaluation and treatment; educational materials to providers cantly) improved adherence to anti- audit and feedback; academic de- and patients.18,19 Little is known fracture practice guidelines. tailing from osteoporosis opinion about which interventions are ef- leaders; case-based teleconference fective in the nursing home. ● Systematic interventions that remove on osteoporosis quality improve- We developed a multi-modal barriers to providing preventive care are ment; and an osteoporosis toolkit. intervention to improve fracture needed. Control nursing homes had no fur- prevention in residents with doc- ther contact until after the study umented osteoporosis or recent was complete, at which time they fracture. received a similar intervention. The intervention period began January 3, 2005 and it ended METHODS May 13, 2005 in Arizona and July 1, 2005 in North This analysis combines data from studies completed with Carolina. the Quality Improvement Organizations in Arizona and Two case-based educational modules for medical staff and North Carolina. The studies were approved by the Duke nursing staff were developed and pilot-tested based on recom- 20 University and University of Alabama at Birmingham mendations from focus groups. The modules could be com- Institutional Review Boards. The Clinical Trial Registry pleted over the Internet or on paper in less than 1 hour, and number is NCT00280943 (http://www.clinicaltrials.gov). 1-hour CME/CEU (continuing medical education/continuing education unit) credit was provided. Recommendations in the Facility Selection and Randomization modules were based on osteoporosis clinical practice guide- 9 The Minimum Data Set, a database of clinical and demo- lines and focused on secondary fracture prevention, including graphic information collected quarterly on all Medicare and assessment for vitamin D deficiency; calcium and vitamin D Medicaid eligible residents, was used to identify nursing supplementation; osteoporosis pharmacotherapy; hip protec- homes in North Carolina and Arizona, with at least 10 tors; and fall prevention. The director of nursing and all med- residents diagnosed with osteoporosis or a hip fracture ical providers received at least 3 fax, e-mail, and telephone within 180 days. Administrators from 67 of 249 eligible reminders to complete the modules. nursing homes agreed to enroll. The nursing homes were Audit and feedback was provided on weeks 3, 8, and 18 to randomized within each state to receive the study interven- all administrators in the intervention group. The blinded report tion immediately (intervention group) or after the follow-up included graphics that compared their facility with other study chart abstraction was completed (control group) using a facilities in the state (Figure). Because of privacy concerns, we random number generator. did not supply the audit and feedback report to medical providers, but asked the administrator to do so. Resident Selection Academic detailing to all medical providers was com- Residents aged 50 years or older with either a history of hip pleted by osteoporosis opinion leaders. At least 3 attempts fracture or a diagnosis of osteoporosis during a 6-month were made to reach each medical provider directly by tele- time period were identified. Residents had to be ambulatory phone, and if unsuccessful, messages about the study were or transfer independently, and have a length of stay of at left. 888 The American Journal of Medicine, Vol 120, No 10, October 2007

Any Pharmaceutical Protection in residents with known Osteoporosis or Fractures Facility characteristics were obtained from public use 1.00 datasets (http://www.medicare.gov/NHCompare). Some 0.90 resident characteristics were obtained from the Minimum 0.80 Data Set. 0.70 0.60 Analysis 0.50 Data were analyzed at the facility level. Nursing homes in 0.42 (75th Percentile) 0.40

0.33 (Median) Arizona and North Carolina were combined for the main 0.30 0.25 (25th Percentile) analysis. The change in the proportion of residents receiving 0.20 therapy between the pre- and postintervention periods was 0.10 compared between the intervention and control nursing 0.00 1 2 3 4 5 6 7 8 9 101112131415161718192021 homes using t tests. All randomized facilities were analyzed Facility regardless of their participation in the study. In order to Figure Sample audit and feedback report provided to facility adjust for characteristics that might impact prescription of administrators, comparing their use of osteoporosis therapies for osteoporosis therapies, multivariable logistic regression high-risk residents compared with other study facilities in their models with generalized estimating equation adjustment for state. repeated binary outcomes were constructed with backwards selection, using all variables with a baseline univariate P Ͻ.20 and variables with high clinical significance (age, Teleconferences were offered 4 times throughout the race, state). The intervention time period by group interac- intervention period. Nursing staff and medical staff were tion term was tested using chi-squared as a test of the impact invited to a case-based presentation on osteoporosis quality of the intervention on prescription of therapies. This method improvement via a faxed and mailed letter. One teleconfer- accounted for the complex nesting of residents within pro- ence was audiotaped, and a compact disc sent to all medical viders and nursing facilities. providers. A prespecified exploratory analysis compared fracture Finally, a “toolkit” of materials was provided to each rates in the intervention period between intervention and administrator. This included posters to hang in the facility control facilities using t test. In order to examine the impact and resident/family brochures. Fall prevention tools and of participation in the intervention components, we con- links to osteoporosis clinical practice guidelines and patient structed Generalized Estimating Equation models using information were provided on the module website. backwards selection, with covariates significant at the Յ.20 level and the intervention participation variables. Data Collection Based on guideline recommendations, quality indicators for Sample Size Considerations osteoporosis evaluation and fracture prevention were mea- Originally, we calculated that enrolling 128 nursing homes 9 The primary outcome variable was prescription of sured. would provide 80% power to detect a 10% difference in the osteoporosis pharmacotherapy or hip protectors. We also change in prescription of any fracture protection between measured a combined outcome variable of prescription of the 2 arms. However, recruitment was challenging, and 67 osteoporosis pharmacotherapy or hip protectors or calcium nursing homes were ultimately enrolled. Post hoc, we de- and vitamin D supplements. We included measures with termined that our sample size provided 80% power to detect and without calcium and vitamin D supplements because a 17% difference, and 98% power to detect a 20% differ- many practice guidelines do not recommend them as mono- ence in the change in proportion of residents receiving any therapy for secondary fracture prevention, and because high fracture protection in the intervention facilities compared baseline prescription rates might introduce a ceiling effect. with the control facilities. Total daily dose of calcium and vitamin D were not assessed because they could come from multiple sources and were difficult to abstract reliably. Facility fracture rates were an RESULTS exploratory outcome measure. Baseline characteristics of the 67 nursing homes and 606 Trained data collectors, blinded to intervention status, residents are shown in Table 1. Intervention residents were abstracted data from the medical record before and after the more likely to be African American, younger, and use tobacco; intervention. Data collectors reviewed prescriptions in a and less likely to have previous fracture or dysphagia. 6-month time period beginning at the first time a fracture or The use of fracture prevention strategies in the preinter- osteoporosis diagnosis was recorded in the Minimum Data vention and postintervention periods is shown in Table 2. Set. The entire medical record was reviewed for selected The primary endpoint of osteoporosis pharmacotherapy or comorbidities, bone mineral density, and laboratory testing. hip protectors improved in intervention homes from 32.6% A random 10% of charts was re-abstracted by a second data to 40.6% (difference 8.0%) and remained unchanged in collector, and inter-rater reliability was maintained at control homes from 38.6% to 39.2%, (difference 0.6%), but Ͼ90%. the differences were not statistically significant (P ϭ .72). Colón-Emeric et al Fracture Prevention in Nursing Home Residents 889

Table 1 Characteristics of the Intervention and Control Nursing Homes and Target Residents during the Control Period Intervention Control P Value* Facility characteristics n ϭ 34 homes n ϭ 33 homes For proﬁt (%) 75.8 82.4 .22 Total number of residents (mean) 110.3 109.3 .62 Rural location (%) 18.8 18.5 .93 Resident characteristics n ϭ 293 n ϭ 313 Women (%) 85.0 87.2 .43 Race/ethnicity (%) Caucasian 81.9 80.8 .73 African American 5.5 2.2 .04 Other 12.6 16.9 .14 Age (mean) 83.0 85.6 Ͻ.001 Insurance status (%) Medicare 30.4 34.5 .28 Medicaid 17.0 13.1 .17 Private 8.5 8.3 .92 Other 79.9 75.7 .22 Previous fracture (%) 16.7 24.6 .02 Resident ambulatory (%) 55.6 62.0 .11 Falls in last 90 days (%) 33.8 37.4 .36 Cognitive impairment (%) 43.3 46.0 .51 Gastroesophogeal reﬂux (%) 39.0 43.8 .23 Ulcer disease or esophagitis (%) 7.5 10.5 .19 Breast cancer (females only) (%) 2.7 4.5 .25 Dysphagia (%) 14.7 20.8 .05 Thromboembolic disease (%) 8.5 8.3 .92 Tobacco use (%) 24.6 16.6 .015 Alcohol abuse (%) 5.1 3.8 .44

The combined endpoint of osteoporosis pharmacotherapy or homes, but baseline treatment rates exceeded 70%, suggest- hip protectors, or calcium and vitamin D supplements ing a possible ceiling effect. Nonsigniﬁcant trends toward showed a 7% improvement in both intervention and control greater improvement in intervention versus control homes

Table 2 Proportion of Residents with Osteoporosis or Recent Hip Fracture Receiving Fracture Prevention Interventions Before and After the Intervention Period in the Intervention and Control Homes

Before After Difference

Intervention Control Intervention Control Intervention Control (n ϭ 33) (n ϭ 34) (n ϭ 33) (n ϭ 34) (n ϭ 33) (n ϭ 34) Diagnostic testing modalities Bone mineral density measurement (%) 1.2 1.1 0.3 0.8 Ϫ0.9 Ϫ0.3 Serum calcium level 58.0 61.2 77.5 73.8 19.5 12.6 25(OH) vitamin D level (%) 0.3 0 1.1 0.5 0.5 0.8 Fracture prevention modalities Calcium (%) 72.5 66.0 82.8 74.5 10.3 8.4 Vitamin D (%) 68.7 57.1* 76.3 70.5 7.7 13.3 Bisphosphonate (%) 17.9 20.2 25.8 20.6 7.9 0.4 Calcitonin (%) 11.5 17.3 9.5 15.6 Ϫ1.9 Ϫ1.7 Raloxifene (%) 2.7 3.7 2.8 2.8 0 -0.9 Teriparatide (%) 0.3 0 0 0 Ϫ0.3 0 HRT in women (%) 2.3 3.8 3.6 2.4 1.3 Ϫ1.4 Testosterone in men (%) 0 0 0 0 0 0 Hip Protectors (%) 2.8 1.3 7.0 3.4 4.2 2.1 Osteoporosis pharmacotherapy or hip protectors (%) 32.6 38.6 40.6 39.2 8.0 0.6 Osteoporosis pharmacotherapy or hip protectors or 79.0 72.8 86.8 80.1 7.7 7.3 calcium and vitamin D (%) HRT ϭ hormone replacement therapy. *Indicates P Ͻ.05 for intervention vs control in the specified time period. 890 The American Journal of Medicine, Vol 120, No 10, October 2007 were seen for most other indicators. There were no significant differences in performance on the quality measures between the states. Multivariable models testing the impact of the intervention on the primary and combined outcome measures were not statistically significant (odds ratio [OR] 1.0, 95% confidence interval [CI], 0.9-1.2, P ϭ .18) even after adjusting for baseline factors that were imbalanced, including bed Absolute Change Any Fracture Prevention (%) size, age, race, sex, previous fracture, insurance status, ambulatory status, gastrointestinal reflux, breast and endome- trial cancer, dysphagia, and tobacco use. Incident fracture rates during the 6-month intervention period were similar Number of Nursing Homes between intervention (4.3%) and control facilities (4.1%), Intervention Homes with Engagement in Activity and were unchanged from the preintervention period. Fall rates were unchanged in the pre- and postintervention periods at 50%-51%. Among the intervention facilities, participation in intervention activities was generally low except for audit feedback, which was provided to 100% of homes (Table 3). Intervention homes with at least 1 participant in the nursing

educational module tended toward greater improvement in Absolute Change Any Fracture Prevention (%) prescription of osteoporosis pharmacotherapy or hip protectors (19.5%) than either control homes (0.6%) or intervention homes without a nursing educational module participant (6.4%, P ϭ .36). Participation in the medical provider educational module showed a similar pattern. In a multiva- Number of Nursing Homes riable Generalized Estimating Equation model including all Intervention Homes with Noin Engagement Activity covariates used above and participation variables, direct physician contact with the academic detailer (OR 4.5, 95% CI, 1.1-18.2, P ϭ .03) and physician completion of the CME module (OR 4.8, 95% CI, 1.9-12.0, P ϭ .001) were significantly associated with prescription of osteoporosis pharmacotherapy or hip protectors compared with control nursing homes and intervention homes without participation. No significant associations were observed when cal- Absolute Change Any Fracture Prevention (%) cium and vitamin D supplements were added to the primary outcome measure, although treatment rates in the follow-up period exceeded 80% for all groups, again suggesting a ceiling effect. 34 0.6 21 9.0 12 6.3 Number of Nursing Homes Group Control DISCUSSION We demonstrated a small, nonsignificant improvement in the prescription of fracture prevention therapies after the implementation of a quality improvement intervention. Al- though our study cohort had high fracture risk given their fall rate (35% fell within 90 days) and previous fracture history (20%), prescription of fracture prevention therapies other than calcium and vitamin D remained low. The magnitude of provider behavior change is consistent with that found in similar osteoporosis quality improvement studies in other settings,21 but is likely insufficient to have an

impact on fracture rates. Change in Prescription of Osteoporosis Pharmacotherapy or Hip Protectors by Level of Nursing Homes’ Participation in the Intervention Activities This study has several strengths. The multi-modal inter-

vention included multiple proven techniques. We targeted directly to at least 1 physician No statistically significant differences*Because in all the intervention 2 homes intervention received subgroups audit compared and with feedback, the the control cell group. is not applicable. not only medical providers, but also the nursing staff and Nursing staffMedical staff 34 34 0.6 0.6 29 26 6.4 6.4 4 7 19.5 13.4 Audit feedbackHomes with academic detailer speaking CME module completed by at least 1: Teleconference participation by at least 1 staff 34 34 0.6 0.6 21 * 8.5 * 12 7.1 33 8.0 Activity administrator, who have considerable influence on decision- Table 3 Colón-Emeric et al Fracture Prevention in Nursing Home Residents 891 making. The study was randomized, outcomes assessment that the goals of residents, administrators, and practitioners blinded, and powered to detect a clinically meaningful dif- are better aligned. ference. Regulations for documentation in nursing homes Third, the osteoporosis guideline recommendations them- make it unlikely that medication therapies were not re- selves may not be optimally suited for the frail nursing home corded in the medical record. We included only residents population. Applying guideline recommendations to resi- with clear indications for osteoporosis therapy and the few- dents with competing comorbidities, surrogate decision- est comorbidities that would limit the applicability of clin- makers, and varying goals of care is challenging; for exam- ical practice guidelines in the nursing home setting. Despite ple, nearly 25% of our sample had peptic ulcer disease, these strengths, there are several potential explanations for esophagitis, or dysphagia that would preclude use of oral the lack of effect. bisphosphonates (Table 1). Most fracture prevention studies First, we had difficulty engaging nursing homes and included postmenopausal women, and generalization to frail providers in our study. Although previous research has older populations also is problematic, although previous surveys have shown that a large majority of nursing home shown that a large majority of Medical Directors and Di- medical directors believe that osteoporosis guidelines are rectors of Nursing believe that fracture prevention is impor- relevant to their patients.22 Guidelines specific to the nurs- tant and effective,22 we successfully recruited only about ing home population that assist providers in determining half of the number of facilities we had anticipated. Admin- residents most likely to benefit are needed.9 istrators most often cited staff turnover, regulatory survey There have been relatively few published randomized demands, and competing clinical projects as reasons for not trials of quality improvement initiatives in the nursing home participating in the study. As a result, our final number of to compare with our results. Effective interventions gener- nursing homes allowed sufficient power to detect a 17% or ally require on-site personnel or extensive provider involve- greater improvement. After enrollment, few providers par- ment in intervention development.23 Other studies have ticipated in the elements of our intervention despite repeated shown a limited effect of group training sessions such as encouragement. The issues of engagement and timing have Quality Improvement Collaboratives24 and highlighted the been shown to be important in behavior change interven- lack of effective quality improvement infrastructure in nurs- tions. It is clear that engagement was critical in this study, ing homes.25 To circumvent some of these issues, system- because those providers we successfully engaged were more atic interventions that remove the responsibility for provid- likely to improve their osteoporosis management compared ing preventive care from individual practitioners are with those who did not participate. Providers who engaged needed. Examples of such interventions might include in the intervention are likely to be more motivated, knowl- standing orders for osteoporosis care that are implemented edgeable, and willing to change. This may partly explain the routinely unless a provider or patient “opts out,” or auto- association between participation and improvement in our mated telephone reminders or letters to patients and families study. Previous studies have suggested that interventions on discharge from the hospital after a fracture. work best when the timing of the intervention is soon after Fracture prevention remains challenging in the nursing the triggering event;18 in our study, an intervention occur- home setting. Further studies are needed to identify effec- ring sooner after a fracture may have been more effective in tive means of changing clinicians’ behavior and testing assisting providers to change their behavior. An important system-wide interventions that could prove more effective goal in future research is to either identify ways to engage than traditional quality improvement approaches. a broader range of providers, or to use systems that do not require individual physician practice change to improve References quality. 1. Chandler J, Zimmerman S, Girman C, et al. Low bone mineral density and risk of fracture in white female nursing home residents. JAMA. Second, there may be barriers to fracture prevention in 2000;284:972-977. the nursing home environment that were not sufficiently 2. Toofanny M, Maddens M, Voytas J, Kowalski D. Low bone mass and addressed by our intervention. In a survey of more than postfall fracture risk among elderly nursing home men. J Am Med Dir 1000 medical directors and directors of nursing, reimburse- Assoc. 2004;5(6):367-370. 3. Tinetti M. Preventing falls in elderly persons. N Engl J Med. 2003; ment issues, short length of stay, and regulatory oversight 348(1):42-49. about the number of medications prescribed were endorsed 4. Cumming R. Nursing home residence and risk of hip fracture. Am J as the primary barriers to adhering to osteoporosis practice Epidemiol. 1996;143:1191-1194. guidelines.22 For example, a nursing home receiving a capi- 5. Brainsky A, Glick H, Lydick E, et al. The economic cost of hip tated reimbursement for the rehabilitation of a hip fracture fractures in community dwelling older adults: a prospective study. J Am Geriatr Soc. 1997;45(3):281-287. patient has little incentive to add an expensive osteoporosis 6. Colón-Emeric C, Kuchibhatla M, Pieper C, et al. The contribution of medication during their stay. Bone density scans are logis- hip fracture to risk of subsequent fractures: data from two longitudinal tically difficult to obtain for frail residents, and Medicare studies. Osteoporos Int. 2003;14(11):879-883. reimburses for testing only every 2 years. These factors 7. Klotzbuecher C, Ross P, Landsman P, et al. Patients with prior fractures have an increased risk of future fractures: a summary of the cannot be easily addressed by an intervention such as ours, literature and statistical analysis. J Bone Miner Res. 2000;15(4):721- and may require systems and health care policy change so 739. 892 The American Journal of Medicine, Vol 120, No 10, October 2007

8. Greenspan S, Schneider D, McClung M, et al. Alendronate improves 17. Majumdar S, Rowe B, Folk D, et al. A controlled trial to increase bone mineral density in elderly women with osteoporosis residing in detection and treatment of osteoporosis in older patients with wrist long-term care facilities. A randomized, double-blind, placebo-con- fracture. Ann Intern Med. 2004;141(5):366-373. trolled trial. Ann Int Med. 2002;136(10):742-746. 18. Gardner M, Brophy R, Demetrakoupoulos D, et al. Interventions to 9. Osteoporosis Clinical Practice Guideline. Columbia, MD: American improve osteoproosis treatment following hip fracture. A prospective, Medical Directors Association, 2003. randomized trial. J Bone Joint Surg Am. 2005;87(1):3-7. 10. Gupta G, Aronow W. Underuse of procedures for diagnosing osteo- 19. Cuddihy M, Amadio P, Gabriel P, et al. A prospective clinical practice porosis and of therapies for osteoporosis in older nursing home resi- intervention to improve osteoprosis management following distal fore- dents. J Am Med Dir Assoc. 2003;4(4):200-202. arm fracture. Osteoporos Int. 2004;15(9):695-700. 11. Jachna C, Shireman T, Whittle J, et al. Differing patterns of antire- 20. Levine DA, Saag KG, Casebeer L, et al. Using a modiﬁed nominal sorptive pharmacotherapy among nursing facility residents and com- group technique to elicit director of nursing input for an osteoporosis munity dwellers. J Am Geriatr Soc. 2005;53(8):1275-1281. intervention. J Am Med Dir Assoc. 2006;7(7):420-425. 12. Kamel H. Underutilization of calcium and vitamin D supplements in 21. Curtis JR, Westfall AO, Stewart E, et al. Challenges in improving the an academic long-term care facility. J Am Med Dir Assoc. 2004;5(2): quality of osteoporosis care for long-term glucocorticoid users: a 98-100. prospective, randomized trial. Arch Intern Med. 2007;167:591-596. 13. Rojas-Fernandez C, Lapane K, MacKnight C, Howard K. Undertreat- 22. Colón-Emeric C, Casebeer L, Saag K, et al. Barriers to providing ment of osteoporosis in residents of nursing homes: population-based osteoporosis care in skilled nursing facilities; perceptions of medical study with use of the Systematic Assessment of Geriatric Drug Use via directors and directors of nursing. J Am Med Dir Assoc. 2004;5:361- Epidemiology (SAGE) Database. Endocr Pract. 2002;8(5):335-342. 366. 14. Colón-Emeric C, Lyles KW, Levine DA, et al. Prevalence and pre- 23. Naughton B, Mylotte J, Ramadan F, et al. Antibiotic use, hospital dictors of osteoporosis treatment in nursing home residents with admissions, and mortality before and after implementing guidelines for known osteoporosis or recent fracture. Osteoporos Int. 2007;18:553- nursing home acquired pneumonia. J Am Geriatr Soc. 2001;49(8): 559. 1020-1024. 15. Solomon D, Morris D, Cheng H, et al. Medication use patterns for 24. Colón-Emeric C, Schenck A, McConnell E, et al. Translating evi- osteoporosis: ab assessment of guidelines, treatment rates, and quality dence-based falls prevention into clinical practice in nursing facilities: improvement interventions. Mayo Clin Proc. 2005;80(2):194-202. results from a state-wide quality improvement collaborative. JAm 16. Charalambous C, Kumar S, Tryfonides M, et al. Management of Geriatr Soc. 2006;54(9):1414-1418. osteoporosis in an orthopaedic department: audit improves practice. Int 25. Lee R, Wendling L. The extent of quality improvement activities in J Clin Pract. 2002;56(8):620-621. nursing homes. Am J Med Qual. 2004;19(6):255-265. The American Journal of Medicine (2007) 120, 893-900

CLINICAL RESEARCH STUDY

Impact of a Fluoroquinolone Restriction Policy in an Elderly Population Muhammad Mamdani, PharmD, MA, MPH,a,b,c David McNeely, MD,c,e Gerald Evans, MD,g Janet Hux, MD, SM,a,c,d Paul Oh, MD,c,f Natalie Forde, MSc,a John Conly, MDh aThe Institute for Clinical Evaluative Sciences; University of Toronto Faculties of bPharmacy and cMedicine; dSunnybrook and Women’s College Health Sciences Centre, Toronto Western Hospital, Toronto, Ontario; eUniversity Health Network and fToronto Rehabilitation Institute, Toronto, Ontario; gQueen’s University Faculty of Health Sciences and Kingston General Hospital, Kingston, Ontario; hUniversity of Calgary Faculty of Medicine and Foothills Medical Centre, Calgary, Alberta.

ABSTRACT

BACKGROUND: In light of growing concerns of bacterial resistance to fluoroquinolones, the province of Ontario instituted a fluoroquinolone restriction policy in March of 2001. The objective of this study was to examine the immediate impact of this policy on the rates of antibiotic prescription use and infectious disease-related hospitalizations among elderly individuals who are dispensed antibiotics. METHODS: An interrupted time series analysis was conducted from January 1, 1994, to March 31, 2002, using administrative health care databases covering more than 1.4 million residents of Ontario, Canada, aged 65 years and older. Population rates of antibiotic use and infectious disease-related hospitalizations within 4 weeks after an antibiotic prescription were examined using interventional autoregressive integrated moving average models. RESULTS: Immediately after the introduction of the fluoroquinolone policy, fluoroquinolone prescription rates decreased to approximately 70% of expected rates (P Ͻ .01). Approximately 30% higher than expected use of sulfonamide (P ϭ .01) and urinary anti-infectives (primarily nitrofurantoin and trimethoprim; P Ͻ .01) were observed within 1 year after policy implementation. No significant changes in the use of any other groups of antibiotics were observed. Although no significant changes in the rates of overall infection-related hospital admissions among antibiotic users were observed, the rate of hospital admission for gastrointestinal infections was 32% lower than expected in the 1 year after the policy change (P Ͻ .01). The hospital admission rate for urinary tract infections was approximately 8% higher than expected (P Ͻ .01). CONCLUSIONS: These findings suggest that formulary restrictions to fluoroquinolones can be implemented effectively to decrease use among an elderly population without adverse impact on hospital admission rates. © 2007 Elsevier Inc. All rights reserved.

Antibiotics are among the most commonly used drugs in the of fluoroquinolones has raised concerns about the appropri- world. However, the potential for overuse has been a grow- ateness of their use (Approximately 40% of fluoroquinolone ing concern globally because of increasing strains of mul- use may be for unapproved indications2). The conse- tiple drug-resistant bacteria.1 As a subset, fluoroquinolones quences of such practices on bacterial resistance are of are the most widely prescribed group of antibiotics in the significant concern, particularly when many other effec- 2 United States, with a broad spectrum of activity and relative and less-expensive alternative antibiotic treatment tively low toxicity profile. However, the growing popularity options are available. Of particular concern is the emer- gence of fluoroquinolone-resistant Streptococcus pneumoniae,3,4 which is one of the most common bacterial Requests for reprints should be addressed to John Conly, MD, Univer- causes of community-acquired pneumonia, meningitis, oti- sity of Calgary Medicine, Room 930, North Tower, Foothills Medical 5-8 Centre 1403-29T, Calgary, Alberta T2N 2T9, Canada. tis media, and sinusitis. In light of clinical and financial E-mail address: [email protected] concerns, several strategies have been proposed to curb

fluoroquinolone use, such as academic detailing,9 formulary Review Board of Sunnybrook and Women’s College Health restriction,10-13 and multidimensional interventions.14 Sciences Centre (Toronto, Ontario, Canada). In the year 2000, Ontario’s Drug Quality and Therapeu- tics Committee (DQTC) undertook a review of antibiotics Data Sources on the Ontario Drug Benefits (ODB) Formulary and recom- The administrative health care databases in Ontario allowed mended criteria to limit the pre- for the assessment of the preva- scribing of fluoroquinolones based lence of prescription drug use and on concerns of growing antibiotic CLINICAL SIGNIFICANCE hospitalizations. The databases in- resistance and misuse. The On- cluded computerized pharmacy tario DQTC review of antibiotics ● Formulary restrictions on fluoroquino- records of the ODB program, noted a particular concern of lone use did not significantly increase which records prescription drugs growing quinolone resistance to the overall rate of hospital admissions in reimbursed by the public drug Acinetobacter species, Pseudomo- an elderly Canadian population in the program for all Ontario residents nas aeruginosa, S. pneumoniae, year after policy implementation. 65 years of age and older. Before Staphylococcus aureus, Neisseria March 1, 2001, all antibiotics were gonorrhoea, Campylobacter jejuni, ● In an elderly population, decreases in listed as General Benefit drugs Salmonella typhimurium DT104, fluoroquinolone use were accompanied and were not subject to any for- and extended spectrum ␤-lacta- by significant decreases in hospital ad- mulary restrictions. As of March mase-producing strains of Esche- missions for gastrointestinal infections 1, 2001, however, restrictions were richia coli and Klebsiella pneu- but slight increases in admissions for imposed on all fluoroquinolones, moniae. The DQTC provides an urinary tract infections. except norfloxacin, making them essential advisory service to the “Limited Use” drugs whose pre- Ontario government through the ● Decreased fluoroquinolone use also was scription required that prescribers assessment of drug products for accompanied by a sizeable (32%) and indicate the reason for use on a government funding and consists significant (P Ͻ .01) decrease in admis- special prescription pad. These of practicing physicians and phar- sions for gastrointestinal infections. policy changes are highlighted in macists who have expertise in a Table 1. wide range of specialties, includ- We obtained hospitalization ing geriatrics, infectious disease, records for all residents of Ontario pharmacology, health economics, epidemiology, and other 65 years of age and older from the Canadian Institute for disciplines. The DQTC mandate also includes the ongoing Heath Information Discharge Abstract Database, which monitoring and evaluation of drug product reimbursement contains a detailed record of all hospital admissions, includ- under the ODB program. ing diagnostic information. The Ontario Registered Persons Although previous studies have demonstrated that for- Database contained basic demographic and vital statistics mulary restriction is an effective means of controlling drug information for each Ontario resident. Each of these data- use,10-13 concerns have been raised regarding the clinical 12 bases bears a unique patient identifier that facilitates deter- consequences of such measures. For example, a retrospec- ministic linkage. The databases used in this study have been tive study has suggested that a restrictive formulary policy 15 shown to be of high quality and are often used for research for fluoroquinolones may result in increased population purposes. rates of hospitalization for pyelonephritis and bronchitis.12 We conducted a population-based study to examine the End Point Ascertainment impact of Ontario’s formulary restriction policy for fluoroquinolones on the rates of overall antibiotic prescription use Quarterly temporal trends for 2 end points were assessed: and infectious disease-related hospitalizations among those antibiotic prescription use rates and hospital admission rates dispensed antibiotics. for infectious disease-related admissions among those dispensed antibiotics. The rate of antibiotic prescription use in METHODS each time interval was determined by dividing the total number of antibiotic prescriptions dispensed by the total Study Design number of individuals alive at the beginning of the interval. We conducted a population-based cross-sectional time se- Antibiotic prescription use rates were expressed as the num- ries analysis from January 1, 1994, to March 31, 2002, using ber of prescriptions per 1000 persons per quarterly interval. administrative health care databases covering the entire The rates of hospitalization for infectious disease among population of 1.4 million residents of Ontario, Canada, aged those dispensed an antibiotic were estimated in each interval 65 years and older. This time period was divided into 33 by dividing the number of infectious disease-related hospi- quarterly intervals. Ontario’s elderly population has univer- talizations within 4 weeks after the dispensing of an outpa- sal access to prescription drugs, hospital care, and physician tient antibiotic prescription by the total number of individ- services. This research study was approved by the Ethics uals dispensed an antibiotic in that interval. Only those Mamdani et al Fluoroquinolone Restriction Policy 895

Table 1 Ontario Drug Benefits Restrictions Imposed on Fluoroquinolone Use as of March 7, 2001 Drug Listing Status Ciprofloxacin Limited use for the treatment of patients with: SST/BJ (gram-negative bacteria): SST/BJ resulting from gram-negative bacteria; severe diabetic foot infection; severe otitis externa; decubitus ulcers GU tract: urinary tract infection/prostatitis/epididymitis caused by Pseudomonas; sexually transmitted diseases COPD with risk: acute bacterial exacerbation of COPD with risk factors; bronchiectasis; pneumonic illness with cystic fibrosis GI: traveler’s diarrhea; enteric fever syndromes Step-down: step-down therapy after parenteral therapy or hospital/emergency department discharge (eg, febrile neutropenia) Exceptional cases of allergy or intolerance to all other appropriate therapies Gatifloxacin Limited use for the treatment of patients with: Levofloxacin CAP with comorbidity: CAP with comorbid illnesses or failure to first-line therapy Moxifloxacin COPD with risk: acute bacterial exacerbation of COPD with risk factors1; bronchiectasis Step-down: step-down therapy after parenteral therapy or hospital/emergency department discharge (eg, febrile neutropenia) Exceptional cases of allergy or intolerance to all other appropriate therapies Ofloxacin Limited use for the treatment of patients with: SST/BJ (gram-negative bacteria): SST/BJ infection resulting from gram-negative bacteria; severe diabetic foot infection GU tract: urinary tract infection/prostatitis/epididymitis; sexually transmitted diseases COPD with risk: Acute bacterial exacerbation of COPD with risk factors1; bronchiectasis GI: traveler’s diarrhea; enteric fever syndromes Step-down: step-down therapy after parenteral therapy or hospital/emergency department discharge Exceptional cases of allergy or intolerance to all other appropriate therapies Norfloxacin General benefit Nalidixic acid Delisted GU ϭ genitourinary; COPD ϭ chronic obstructive pulmonary disease; GI ϭ gastrointestinal; SST ϭ skin and soft tissue; BJ ϭ bones and joints; CAP ϭ community-acquired pneumonia. admitted to a hospital with a most responsible diagnosis Primary and Secondary Analyses of an infectious disease were included for analysis. The primary analysis examined the temporal changes in the Transfer admissions were excluded to avoid double- use of fluoroquinolones and overall hospital admission rates counting. Consistent hospitalization data were available for the infectious disease categories after the introduction of until March 31, 2002, and analyses examining hospital- the revised ODB fluoroquinolone coverage policy. Second- ization rates were therefore limited to this time interval. ary analyses examined temporal changes in each broad Diagnoses were captured in the administrative databases classification of antibiotics as outlined in Table 2,inthe using the International Classification of Diseases, revision total antibiotic prescription use rate, and in each of the 8 9 (ICD-9) coding system until March 31, 2002. Hospital categories of infectious disease-related hospitalizations after diagnoses were subsequently coded using an ICD-10 coding the introduction of the revised ODB fluoroquinolone cov- system. The translation of coding between the ICD-9 and erage policy. We also examined hospitalization rates re- ICD-10 systems is not congruent, and therefore analyses gardless of reason for admission and mortality within the 4 were limited to aforementioned time frames. Hospitaliza- weeks after antibiotic prescription. tion rates were expressed as the number of events per 100,000 persons per quarterly interval. All hospital admis- Statistical Analysis sions assessed were required to have a most responsible To examine changes in the prevalence and rates of antibiotic diagnosis of a relevant infectious disease. Hospital admis- use and hospitalizations over time, we used time series sion for infectious disease was categorized into 8 broad analysis. Time series analysis,16 a collection of techniques groupings: gastrointestinal infections, intra-abdominal in- for modeling autocorrelation in temporally sequenced data, fections, urinary tract infections, upper respiratory tract in- was conducted using exponential smoothing models and fections, pulmonary infections, cellulitis/skin and soft tissue interventional autoregressive integrated moving average infections (with or without abscess), unclassified bacterial models to model interval data from January 1994 to March infections, and miscellaneous infections (sexually transmit- 2002. Interventional autoregressive integrated moving av- ted diseases, central nervous system infections, dentoalve- erage models incorporated a ramp function to estimate the olar infections, osteomyelitis). impact of the fluoroquinolone policy on antibiotic use and 896 The American Journal of Medicine, Vol 120, No 10, October 2007

Table 2 Antibiotic Groupings for Analysis Class of Antibiotic Subgroup Antibiotics Included Quinolones None Ciprofloxacin, gatifloxacin, grepafloxacin, levofloxacin, moxifloxacin, norfloxacin, ofloxacin Urinary Anti-infectives None Fosfomycin, methenamine, nitrofurantoin, pivmecillinam, trimethoprim Sulfonamides and None Sulfamethizole, sulfamethoxazole ϩ trimethoprim, sulfamethoxazole, combinations sulfamethoxazole ϩ phenazopyridine, sulfadiazine Cephalosporins None Cefaclor, cefixime, cefuroxime, cephalexin, cefprozil, cephradine Macrolides Second-generation macrolides Azithromycin, clarithromycin Other macrolides Dirithromycin, erythromycin Penicillins ␤-lactamase resistant Amoxicillin ϩ clavulanic acid Other Amoxicillin, ampicillin, bacampicillin, carbenicillin, cloxacillin, dicloxacillin, flucloxacillin, penicillin, pivampicillin Tetracyclines Doxycycline Doxycycline Other Demeclocycline, minocycline, tetracycline Miscellaneous None Clindamycin, lincomycin, metronidazole, spiramycin, vancomycin

hospital admission rates. Model-derived projections and (n ϭ 210,196) during the third quarter of 2001. The average their 95% confidence intervals for the quarterly intervals population per time interval was approximately 1.40 million from April 2001 to March 2002 were compared with actual elderly persons (standard deviation ϭ 0.07 million), and the use estimates. Stationarity was assessed using the autocor- average number of antibiotic users per time interval was relation function and the augmented Dickey-Fuller test.17 246,559 elderly persons (standard deviation ϭ 22,984). The autocorrelation, partial autocorrelation, and inverse autocorrelation functions were assessed for model parameter Primary Analyses appropriateness and seasonality. The presence of white Before the introduction of the fluoroquinolone policy, fluo- noise was assessed by examining the autocorrelations at 18 roquinolones were the most commonly used group of anti- various lags using the Ljung-Box chi-square statistic. biotics in Ontario at 63.6 prescriptions per 1000 elderly residents per quarter, followed by penicillins (60.0 prescrip- RESULTS tions per 1000 elderly residents per quarter), macrolides Antibiotic use varied by season, with peak use during the (47.9 prescriptions per 1000 elderly residents per quarter), first quarter of each year (ie, January to March) and lowest and cephalosporins (34.3 prescriptions per 1000 elderly use during the third quarter of each year (ie, July to Sep- residents per quarter). Immediately after the introduction of tember). The prevalence ranged from a high of 21.6% of the the fluoroquinolone policy, fluoroquinolone prescription elderly population (n ϭ 283,653) during the first quarter of rates decreased significantly to approximately 70% of ex- 1995 to a low of 14.1% of the elderly population pected rates (P Ͻ .01; Figure 1). However, within 1 year

Figure 1 Overall ﬂuoroquinolone prescription utilization and hospital admission rates for infectious diseases among antibiotic users over time: quarterly rates. Mamdani et al Fluoroquinolone Restriction Policy 897

Figure 2 Fluoroquinolone prescription utilization rate over time: individual fluoroquinolones. after policy implementation, fluoroquinolone prescription quarter vs predicted use ϭ 43.6 prescriptions per 1000 el- use rates started increasing again, but the changes were not derly persons per quarter; P Ͻ .01, Figure 2) and remained significant. Hospitalization rates for infection-related ad- significantly lower than expected rates throughout the missions among antibiotic users before the fluoroquinolone 1-year follow-up period, ciprofloxacin remained the most policy averaged approximately 1313 (standard devia- widely used fluoroquinolone. Significant increases in nor- tion ϭ 135) admissions per 100,000 elderly antibiotic users, floxacin use also were observed after policy implementation ranging from a high of 1629 per 100,000 elderly antibiotic to rates of approximately 65% higher than expected use users during the first quarter of 1998 to a minimum of 1112 (P Ͻ .01; Figure 2). A significant reduction in ofloxacin during the second quarter of 1994. No consistent significant prescription rates also was observed (P Ͻ .01), but this ef- changes in overall hospital admission rates for infection- fect might be of minimal consequence given the low abso- related conditions were observed within 1 year after policy lute use rates. Changes in levofloxacin, moxifloxacin, and implementation (P ϭ .55; Figure 1). gatifloxacin could not be estimated given the short period of time available since their introduction, although their use Secondary Analyses contributed to the increasing total fluoroquinolone use rates Fluoroquinolone Antibiotic Use. Significant changes in over time. the use of specific fluoroquinolones were observed after policy implementation. Although ciprofloxacin use rates Non-Fluoroquinolone Antibiotics. No significant changes decreased significantly in the months after policy imple- in cephalosporin (P ϭ .19; Figure 3), macrolide (P ϭ .16), mentation to approximately 40% of expected use rates (ac- penicillin (P ϭ .16), tetracycline (P ϭ .52), or miscellaneous tual use ϭ 17.1 prescriptions per 1000 elderly persons per (P ϭ .79) antibiotic use rates were observed. Among the

Figure 3 Non-ﬂuoroquinolone antibiotic utilization over time: quarterly rates. 898 The American Journal of Medicine, Vol 120, No 10, October 2007

“other” antibiotics category, significant increases in the use ied, a slight increase in the rate of hospital admission for of sulfonamide (P ϭ .01) and urinary anti-infectives (pri- urinary tract infections was associated with the policy im- marily nitrofurantoin and trimethoprim; P Ͻ .01) after pol- plementation along with a significant reduction in the rate of icy implementation were observed. Observed rates were hospital admission for gastrointestinal infections. approximately 30% higher than expected for each of these Given the broad spectrum of activity of fluoroquinolone groups of antibiotics by the end of the 1-year follow-up antibiotics, they effectively treat both complicated and un- period. complicated urinary tract infections. Fears of increased rates of complicated urinary tract infections with the restriction of Total Antibiotic Use. No significant changes in overall fluoroquinolones have been raised12 because agents with a antibiotic prescription rates were observed after policy im- more limited spectrum of activity would need to be used in ϭ plementation (P .30). their place. Although fluoroquinolones have not tradition- ally been associated with gastrointestinal infections, several Infection-Related Hospitalizations. The most common case-control studies have found significantly increased risks types of infection-related hospital admissions were pulmo- of Clostridium difficile infection with fluoroquinolone nary infections (ϳ760 admissions per 100,000 antibiotic use, particularly the newer broad-spectrum fluoroquinolo- users per quarter), urinary tract infections (ϳ153 admis- nes.20-22 Many of the recently described outbreaks have sions per 100,000 antibiotic users per quarter), and intra- been associated with elderly or nursing home popula- abdominal infections (ϳ124 admissions per 100,000 anti- tions,23-25 and a study from a Canadian center suggested the biotic users per quarter). Although no significant changes in pattern of the infection is changing with an increase of the rates of overall infection-related hospital admissions among proportion of cases with severe and fatal complications.26 antibiotic users were observed, a significant reduction in the rate of hospital admission for gastrointestinal infections was Another recent Canadian study from Quebec found that the observed (P Ͻ .01; Figure 4a). In the year after policy im- use of the newer broad-spectrum fluoroquinolones (moxi- plementation, the hospital admission rate for gastrointesti- floxacin and gatifloxacin), in addition to ciprofloxacin, may nal infections was approximately 32% lower than expected have promoted a major outbreak of severe C. difficile infection in a predominantly elderly population that was as- rates, which translates to approximately 74 fewer hospital 27 admissions for gastrointestinal infections per 100,000 anti- sociated with significant morbidity and mortality. The biotic users. However, an increase in hospital admission for predominant strain of C. difficile associated with this out- urinary tract infection also was observed (P Ͻ .01; Figure break was a single toxigenic clone that was highly resistant 4b). In the year after policy implementation, the hospital to fluoroquinolones. Although we were not able to extract admission rate for urinary tract infections was approxi- data on C. difficile infections, it is interesting to speculate mately 8% higher than expected rates, which translates to whether the findings of significantly fewer gastrointestinal approximately 63 more hospital admissions for urinary in- infections is that less fluoroquinolone use in this elderly fections per 100,000 antibiotic users. No significant changes population may have been associated with less C. difficile in hospital admission rates for pulmonary (P ϭ .55; Figure infection, to which the elderly population is particularly 4c), intra-abdominal (P ϭ .92), cellulitis (P ϭ .63), upper susceptible. We acknowledge that further studies to test this respiratory tract (P ϭ .96), unclassified bacterial (P ϭ .11), hypothesis are necessary. and miscellaneous (P ϭ .82) infections were observed. Several limitations of this study should be noted. First, we used administrative databases to identify and define All-Cause Hospitalization and Mortality. No significant exposure to study drugs and clinical outcomes. We have no differences in all-cause hospitalization (P ϭ .55) or mortal- direct measure of adherence or appropriateness of use. Be- ity (P ϭ .62) rates were observed as a function of the anti- cause antibiotics may be used in varying doses in different biotic restriction policy. regimens, the dose equivalence of various antibiotics could not be adequately examined with these data. Instead, anti- DISCUSSION biotics were examined as they are naturally used under a A significant reduction in fluoroquinolone use after the “real world” context in this population. Second, accurate introduction of a formulary restriction policy aimed at all data on antibiotic resistance could not be reported given the fluoroquinolones (except norfloxacin) was observed in a short follow-up of the study. Third, the generalizability of population 65 years and older. These reductions were ac- our findings to younger patients or settings with different companied by increases in the use of norfloxacin, urinary drug policies over longer durations of follow-up is uncer- anti-infectives, and sulfonamide and combination antibiot- tain. Fourth we did not have the ability to look at longer- ics without any significant increases in macrolide or ceph- term outcomes given the issues with the changes in the alosporin use in the 1-year after policy implementation. coding systems. As mentioned previously, a new diagnostic This is an important observation given that some predicted coding system was implemented in Ontario hospitals as of a reciprocal increase in other classes of antimicrobials April 2002 (ie, Ontario’s hospital systems transitioned from within this population, especially for respiratory tract infec- ICD-9 coding to ICD-10 coding). We were not confident of tions.19 Among users of antibiotics in the population stud- the translation between ICD-9 and ICD-10 coding for all of Mamdani et al Fluoroquinolone Restriction Policy 899

Figure 4 (A) Hospital admisssion rates for gastrointestinal infection-related conditions over time: quarterly rates. (B) Hospital admission rates for urinary tract infection-related conditions over time: quarterly rates. (C) Hospital admission rates for pulmonary infection-related conditions over time: quarterly rates. the hospital codes used in our study and decided to end believe the 1-year follow-up is appropriate for 2 primary study follow-up in March of 2002, just before the coding reasons. First, the impact of the restriction policy on anti- transition. This decision was made to ensure consistency in biotic use was immediate and stabilized within months of coding and maximize the validity of our observations. We implementation. Second, related hospitalizations were lim- 900 The American Journal of Medicine, Vol 120, No 10, October 2007 ited to 4 weeks after the date of antibiotic prescription 10. Mather JL, Bayliff CD, Reider MJ, et al. The impact of formulary dispensing, because the majority of related clinical out- reservations on drug utilization: a controlled trial. Can J Hosp Pharm. comes of antibiotic treatment are expected shortly after their 1994;47:111-116. 11. Martin C, Ofotokun I, Rapp R, et al. Results of an antimicrobial use. However, we cannot comment on the sustainability of control program at a university hospital. Am J Health Syst Pharm. the prescribing trends beyond the 1-year follow-up period. 2005;62:732-738. A final limitation of our study was that we were unable to 12. LeLorier J, Derderian F. Effect of listing ciprofloxacin in provincial identify the specific pathogens involved in the outcomes we formularies on hospitalizations for bronchitis and pyelonephritis. Can assessed. J Clin Pharmacol. 1998;5:133-137. Despite these limitations, the findings of this study sug- 13. Marra F, Patrick DM, White R, et al. Effect of formulary policy gest that formulary restrictions to fluoroquinolones can decisions on antimicrobial drug utilization in British Columbia. J Antimicrob Chemother. 2005;55:95-101. effectively be implemented to decrease their use without 14. Gonzales R, Steiner JF, Lum A, Barrett PH Jr. Decreasing antibiotic significant adverse consequences on population rates of use in ambulatory practice: impact of a multidimensional intervention relevant clinical outcomes. The observations of a significant on the treatment of uncomplicated acute bronchitis in adults. JAMA. reduction in the rate of hospital admissions for gastrointes- 1999;281:1512-1519. tinal infections, which may be a potential beneficial effect, 15. Williams JI, Young W. A summary of studies on the quality of health and a significant increase in hospitalizations for urinary tract care administrative databases in Canada. In: Goel V, Williams JI, infections deserve further study to provide an explanation Anderson GM, et al, eds. Patterns of Health Care in Ontario. The ICES Practice Atlas. 2nd edition. Ottawa: Canadian Medical Associ- for these unexpected findings. ation; 1996:339-345. 16. Pindyck RS, Rubinfeld DL. Econometric Models and Economic Fore- casts, 4th edition. New York: Irwin McGraw-Hill; 1998, chapter 15. ACKNOWLEDGMENTS 17. Dickey DA, Fuller WA. Distribution of the estimators for autoregres- We acknowledge the support and advice of the Drug Pro- sive time series with a unit root. J Am Stat Assoc. 1979:427-431. grams Branch of the Ontario Ministry of Health in conduct- 18. Ljung GM, Box GEP. On a measure of lack of fit in time series models. Biometrika. 1978;65:297-303. ing this study. 19. Mandel J, Boadway T. The Cipro case: a new formulary restriction on the Ontario drug benefit plan. Ontario Medical Review. March 2001: References 16-24. 1. Wenzel RP, Edmond MB. Managing antibiotic resistance. N Engl 20. Muto CA, Pokrywka M, Shutt K, et al. A large outbreak of Clostridium J Med. 2000;343:1961-1963. difficile-associated disease with an unexpected proportion of deaths 2. Linder JA, Huang ES, Steinman MA, et al. Fluoroquinolone prescrib- and colectomies at a teaching hospital following increased fluoroquin- ing in the United States: 1995-2000. Am J Med. 2005;118:259-268. olone use. Infect Control Hosp Epidemiol. 2005;26:273-280. 3. Chen DK, McGeer A, de Azavedo JC, Low DE. Decreased suscepti- 21. McCusker ME, Harris AD, Perencevich E, Roghmann MC. Fluoro- bility of Streptococcus pneumoniae to fluoroquinolones in Canada. quinolone use and Clostridium difficile-associated diarrhea. Emerg Canadian Bacterial Surveillance Network. N Engl J Med. 1999;341: Infect Dis. 2003;9:730-733. 233-239. 22. Yip C, Loeb M, Salama S, et al. Quinolone use as a risk factor for 4. Ferraro MJ. The rise of fluoroquinolone resistance: fact or fiction. nosocomial Clostridium difficile associated diarrhea. Infect Control J Chemother. 2002:14(Suppl 3):31-41. Hosp Epidemiol. 2001;22:572-575. 5. Schuchat A, Robinson K, Wenger JD, et al. Bacterial meningitis in the 23. Gaynes R, Rimland D, Killum E, et al. Outbreak of Clostridium United States in 1995. Active Surveillance Team. N Engl J Med. difficile infection in a long-term care facility: association with gati- 1997;337:970-976. floxacin use. Clin Infect Dis. 2004;38:640-645. 6. Jacobs MR. Increasing importance of antibiotic-resistant Streptococ- 24. Ackermann G, Tang-Feldman YJ, Schaumann R, et al. Antecedent use cus pneumoniae in acute otitis media. Pediatr Infect Dis J. 1996;15: of fluoroquinolones is associated with resistance to moxifloxacin in 940-943. Clostridium difficile. Clin Microbiol Infect. 2003;9:526-530. 7. Maston BJ, Plouffe JF, File TM Jr, et al. Incidence of community- acquired pneumonia requiring hospitalization. Results of a population- 25. Moshkowitz M, Ben Baruch E, Kline Z, et al. Clinical manifestations based active surveillance study in Ohio. The Community-Based Pneu- and outcome of Pseudomembranous colitis in an elderly population in monia Incidence Study Group. Arch Intern Med. 1997;157:1709-1718. Israel. Isr Med Assoc J. 2004;6:201-204. 8. Berg O, Carenfelt C, Kronvall G. Bacteriology of maxillary sinusitis in 26. Pepin J, Valiquette L, Alary ME, et al. Clostridium difficile-associated relation to character of inflammation and prior treatment. Scand J In- diarrhea in a region of Quebec from 1991 to 2003: a changing pattern fect Dis. 1988;20:511-516. of disease severity. CMAJ. 2004;171:466-472. 9. Solomon DH, Van Houten L, Glynn RJ, et al. Academic detailing to 27. Loo V, Poirier L, Miller M, et al. A predominantly clonal multi- improve use of broad-spectrum antibiotics at an academic medical institutional outbreak of Clostridium difficile-associated diarrhea with center. Arch Intern Med. 2001;161:1897-1902. high morbidity and mortality. New Engl J Med. 2005;353:2442-2449. The American Journal of Medicine (2007) 120, 901-910

CLINICAL RESEARCH STUDY

Prevention of Central Venous Catheter-Associated Thrombosis: A Meta-analysis Angelia Kirkpatrick, MD, MPH,a,b Suman Rathbun, MD, MS,a Thomas Whitsett, MD,a Gary Raskob, PhDc aDepartment of Medicine, Cardiovascular Section, University of Oklahoma Health Sciences Center, Oklahoma City; bVeterans Affairs Medical Center, Oklahoma City; cCollege of Public Health, University of Oklahoma Health Sciences Center, Oklahoma City.

ABSTRACT

PURPOSE: Anticoagulant prophylaxis in patients with central venous catheters is controversial. We performed a meta-analysis of randomized controlled trials of anticoagulant prophylaxis in patients with central venous catheters. METHODS: MEDLINE and EMBASE were searched up to May 2006, supplemented by manual searches of conference proceedings and bibliographies. RESULTS: Fifteen trials were included. Unfractionated heparin infusion, oral fixed low-dose vitamin K antagonist, and subcutaneous low-molecular-weight heparin were evaluated. For all catheter-associated deep vein thrombosis (symptomatic and asymptomatic combined), the summary relative risks ranged from 0.31 to 0.73 (all achieved statistical significance). For symptomatic deep vein thrombosis, the summary relative risks ranged from 0.28 to 0.72, but did not achieve statistical significance for any individual regimen. CONCLUSION: Anticoagulant prophylaxis is effective for preventing all catheter-associated deep vein thrombosis in patients with central venous catheters. The effectiveness for preventing symptomatic venous thromboembolism, including pulmonary embolism, remains uncertain. © 2007 Elsevier Inc. All rights reserved.

KEYWORDS: Anticoagulant; Catheter; Meta-analysis; Prophylaxis; Thrombosis; Venous

Central venous catheters are increasingly used in clinical safety of anticoagulant administration for prevention of practice, with a 27% to 67% incidence of catheter-associ- catheter-associated venous thromboembolism. ated deep vein thrombosis,1-3 which is most often asymptomatic. Pulmonary embolism occurs in 15% to 36% of METHODS patients with symptomatic catheter-associated deep vein thrombosis.2,3 Thus, preventive measures for catheter-asso- We searched MEDLINE (1964 to 2006), EMBASE, and ciated venous thromboembolism might be warranted. conference proceedings from The American Society of He- Although studies evaluating thromboembolism prophy- matology (2002 to 2005), The American Society of Clinical laxis in patients with central venous catheters have been Oncology (1999 to 2006), and the International Society of published, some report conﬂicting results, and evidence- Thrombosis and Haemostasis (2001 to 2005) using the fol- based recommendations for clinical practice have varied.4,5 lowing terms: “Thromb:.mp or hypercoag:.mp or occlu: We performed a meta-analysis to clarify the efﬁcacy and .mp”; “anticoag:.mp or anti-vitamin k.mp or thromb: pro- phylax:.mp or heparin:.mp”; “catheter:.mp or cannul:.mp or central lin:.mp or port:.mp or central venous:.mp or vascular access:.mp”; as well as generic and trade names for antico- Requests for reprints should be addressed to Angelia C. Kirkpatrick, agulants and catheters. Limits were set for human subjects MD, MPH, Department of Medicine, Cardiovascular Section, University of Oklahoma Health Sciences Center, 920 Stanton L. Young Blvd., WP 3010, and English language. Authors were contacted for addi- Oklahoma City, OK 73104. tional information if required. Bibliographies were screened E-mail address: [email protected] for relevant articles.

Objective criteria for study inclusion were defined a different population effect sizes.13 Summary relative risks priori. Included were studies that: enrolled adult patients, for each outcome were generated using the fixed effects evaluated systemically administered anticoagulant pro- model, which assumes that all studies estimate the same phylaxis other than flushing of catheters with heparin or population effect size. Analyses were repeated using the saline, evaluated prophylaxis in patients with central- more conservative random effects model, which assumes venous catheters or peripherally the studies estimate different pop- inserted central catheters, objec- ulation effect sizes.13 Statistical tively evaluated patients for deep CLINICAL SIGNIFICANCE significance of summary treatment vein thrombosis, and were ran- effects was determined using a Z domized controlled trials. ● In patients with central venous cathe- test, with a P value Յ.05 consid- 13 Two reviewers (ACK, SWR) ters, anticoagulant prophylaxis reduces ered statistically significant. evaluated each article using pre- risk of all (combined symptomatic and Sensitivity analyses were per- defined criteria that used estab- asymptomatic) catheter-associated deep formed to evaluate the robustness lished methodological standards of our results after removing indi- vein thrombosis. for meta-analyses evaluating ther- vidual studies from the analysis. apy.6-9 A third reviewer (TLW) ● Effective prophylactic treatments in- We repeated the analyses, limiting resolved disagreements by blinded clude low-dose unfractionated heparin the included studies to those that adjudication. infusion, fixed low-dose vitamin-K an- were double-blind, and those pub- The strength and quality of tagonists, and prophylactic doses of lished after the year 2000. Funnel each study was assessed according low-molecular weight heparin. plots were generated to determine to the following: random and con- whether unpublished studies with cealed group allocation; consecu- ● These treatments do not increase risk of nonsignificant results were under- tive enrollment; patient spectrum; major bleeding. represented in our analysis, lead- group similarity at baseline; treating to publication bias that may ment similarity other than the in- ● Whether these treatments prevent symp- yield overly optimistic summary tervention being studied; blinding tomatic deep vein thrombosis, including relative risks.13 Calculations were of patients, clinicians, and out- pulmonary embolism, remains uncertain. performed using Comprehensive comes assessors; a priori defini- Meta-Analysis (Biostat, Inc, Engle- tion of major and minor bleeding wood, NJ). events; and use of an objective test for the evaluation of deep vein thrombosis. DATA SYNTHESIS The frequency of the following outcomes was recorded: Figure 1 shows the process of study selection. The literature all catheter-associated deep vein thrombosis; symptomatic 14-83 catheter-associated deep vein thrombosis; symptomatic pul- search identified 70 studies. Fifty-seven did not meet the predefined eligibility criteria: 24 were reviews,14-37 13 monary embolism; major bleeding; minor bleeding; and 38-50 51,52 death from all causes. Studies were included in the analysis were retrospective, 2 were opinions, 1 was a case series,53 9 were nonconcurrently controlled,54-62 and 8 were of “all catheter-associated deep vein thrombosis” if patients 63-70 71-83 underwent mandatory screening with an objective test for not randomized. Thirteen were randomized and deep vein thrombosis. The analysis of “symptomatic cath- were included. eter-associated deep vein thrombosis” included all studies Our search of conference proceedings yielded 11 abstracts.84-94 Four were retrospective 84-87 and were ex- that reported symptomatic events confirmed by an objective 88,89 diagnostic test. cluded. Two randomized trials were excluded due to incomplete information. Three were also published as full For purposes of data extraction, clinically overt bleeding 90-92 was defined as major if it resulted in a decrease in hemo- manuscripts and were included as such. Two remaining abstracts of randomized controlled trials were globin of 2 g per deciliter or more, required the transfusion 93-94 of 2 or more units of blood, was retroperitoneal, intracranial, included. intraocular, or contributed to death.10 Minor bleeding was defined as clinically overt bleeding not meeting criteria for Study Characteristics major bleeding. Table 1 (available online) shows the characteristics of 15 randomized controlled trials included for data extrac- Statistical Methods tion.71-83,93-94 All studies included only upper-extremity Statistical analyses adhered to established standards for central venous catheters. Ten included only cancer pa- meta-analysis.11-13 We assessed the validity of combining tients,74-76,80-83,93,94 and 5 included only patients receiving results from individual studies using the Q statistic to assess total parenteral nutrition.71-73,78,79 Five different anticoagu- for heterogeneity.13 We considered a P value of .10 or less lant medications were evaluated: low-dose unfractionated statistically significant evidence of heterogeneity, which, if heparin infusion, fixed low-dose vitamin-K antagonists present, would imply that the included studies estimate (warfarin or acenocoumarin), and subcutaneous low-molec- Kirkpatrick et al Catheter-Associated Thrombosis Prevention 903

Figure 3. For low-dose unfractionated heparin infusion,72,77 the summary relative risk was 0.28 (95% CI, 0.05 to 1.69). For ﬁxed low-dose vitamin-K antagonist,74,76,82 the summary relative risk was 0.60 (95% CI, 0.30 to 1.20). For low-molecular-weight heparin,75,81,83 the summary relative risk was 0.69 (95% CI, 0.30 to 1.59). The summary relative risk using any anticoagulant prophylaxis72,74-77,81-83 was 0.59 (95% CI, 0.35 to 0.97). There was no statistical evidence of heterogeneity.

Pulmonary Embolism Five trials reported documentation of symptomatic pulmonary embolism.73,75,79,81,82 The summary relative risk using any anticoagulant prophylaxis for preventing symptomatic pulmonary embolism was 1.96 (95% CI, 0.52 to 7.45) with no statistical evidence of heterogeneity.

Bleeding The results of the analysis for major bleeding are shown in Table 5 and Figure 4. For low-dose unfractionated heparin infusion,72,77,78 the summary relative risk was 0.74 (95% CI, 0.17 to 3.22). For fixed low-dose vitamin-K antagonist,76,82 the summary relative risk was 0.24 (95% CI, 0.03 to 2.13). For low-molecular-weight heparin,75,81,83,93 the Figure 1 Flow diagram of the trial selection process. summary relative risk was 0.66 (95% CI, 0.12 to 3.68). The summary relative risk of major bleeding with any anticoagulant prophylaxis 72,75-78,81-83,93 was 0.54 (95% CI, 0.20 to ular-weight heparins (dalteparin, enoxaparin, and nadropa- 1.42). There was no statistical evidence of heterogeneity. rin).71-74,76-79,82,94 One study compared nadroparin 2850 Seven studies reported minor bleeding events.72,75,76,78,81-83 units daily with warfarin 1 mg daily,80 and one compared For low-dose unfractionated heparin infusion,72,78 the sum- dalteparin 5000 units daily with acenocoumarin 1 mg mary relative risk was 3.74 (95% CI, 0.36 to 38.31). For daily.94 fixed low-dose vitamin-K antagonist,76,82 the summary relative risk was 1.76 (95% CI, 0.49 to 6.40). For low-molec- Study Quality ular-weight heparin,75,81,83 the relative risk was 1.32 (95% Table 2 (available online) summarizes the measures of qual- CI, 0.87 to 2.01). The summary relative risk of minor ity assessed for each study. bleeding with any anticoagulant prophylaxis72,75,76,78,81-83 was 1.40 (95% CI, 0.94 to 2.07). There was no statistical All Deep Vein Thrombosis evidence of heterogeneity. The results of analysis of all catheter-associated deep vein thrombosis are shown in Table 3 and Figure 2. For low- All-Cause Mortality dose unfractionated heparin infusion,71-73,77-79 the sum- Seven trials reported documentation of all-cause mortal- mary relative risk was 0.31 (95% CI, 0.13 to 0.71). For ity.74-77,81-83 For the single trial using low-dose unfraction- fixed low-dose vitamin-K antagonist prophylaxis,74,94 the ated heparin infusion,77 the relative risk was 0.97 (95% CI, summary relative risk was 0.37 (95% CI, 0.26 to 0.52), 0.14 to 6.67). For fixed low-dose vitamin-K antago- and for low-molecular-weight heparin,75,81,83,93,94 the nist,74,76,82 the summary relative risk was 0.95 (95% CI, summary relative risk was 0.72 (95% CI, 0.57 to 0.90). 0.62 to 1.46). For low-molecular-weight heparin,75,81,83 the No analysis demonstrated statistical evidence of hetero- summary relative risk was 1.57 (95% CI, 0.54 to 4.58). The geneity. For studies comparing low-molecular-weight summary relative risk of mortality using any anticoagulant heparin to fixed low-dose vitamin-K antagonist80,94 the prophylaxis74-77,81-83 was 1.03 (95% CI, 0.70 to 1.52). summary relative risk was 1.88 for low-molecular-weight There was no statistical evidence of heterogeneity. heparin (95% CI, 1.28 to 2.75) with no statistical evidence of heterogeneity. Sensitivity Analyses For low-dose unfractionated heparin infusion, the summary Symptomatic Deep Vein Thrombosis relative risk for symptomatic catheter-associated deep vein The results of the analysis for symptomatic catheter-asso- thrombosis in patients receiving parenteral nutrition in- ciated deep vein thrombosis are shown in Table 4 and creased from 0.28 to 1.00 (95% CI, 0.02 to 48.09) after 904 The American Journal of Medicine, Vol 120, No 10, October 2007

Table 3 Relative Risk of All Catheter-Associated Deep Vein Thrombosis with Anticoagulant Prophylaxis

Prophylaxis No Prophylaxis Citation n/n (%) n/n (%) Relative Risk (95% Conﬁdence Interval) P Value Fixed low-dose heparin Abdelkeﬁ et al, 200477 1/65 (1.5) 8/63 (12.7) 0.1212 (0.0156-0.9409) .0136 Brismar et al, 198278 0/23 (0) 4/26 (15.4) 0.1250 (0.0071-2.2036) .0804 Fabri et al, 198271 2/24 (8.3) 7/22 (31.8) 0.2619 (0.0608-1.1290) .0449 Fabri et al, 198472 0/20 (0) 0/20 (0) 1.0000 (0.0208-48.0855) 1.0000 Macoviak et al, 198473 2/17 (11.8) 1/20 (5.0) 2.3529 (0.2331-23.7458) .4525 Ruggiero et al, 198379 0/17 (0) 0/17 (0) 1.0000 (0.0210-47.7081) 1.0000 Fixed 5/166 (3.0) 20/168 (11.9) 0.3053 (0.1320-0.7061) .0055 Random 5/166 (3.0) 20/168 (11.9) 0.3450 (0.1357-0.8771) .0254 Low-molecular-weight heparin Conte et al, 200393 4/38 (10.5) 3/20 (15.0) 0.7018 (0.1738-2.8332) .6191 DeCicco et al, 200694 46/114 (40.4) 63/114 (55.3) 0.7302 (0.5531-0.9638) .0242 Karthaus et al, 200681 20/294 (6.8) 11/145 (7.6) 0.8967 (0.4416-1.8209) .7631 Monreal et al, 199675 1/16 (6.3) 8/13 (61.5) 0.1016 (0.0145-0.7108) .0014 Verso et al, 200583 22/155 (14.2) 28/155 (18.1) 0.7857 (0.4708-1.3112) .3542 Fixed 93/617 (15.1) 113/447 (25.3) 0.7162 (0.5689-0.9018) .0045 Random 93/617 (15.1) 113/447 (25.3) 0.7345 (0.5645-0.9557) .0216 Fixed low-dose vitamin K antagonist Bern et al, 199074 4/42 (9.5) 15/40 (37.5) 0.2540 (0.0921-0.7004) .0027 DeCicco et al, 200694 26/120 (21.7) 63/114 (55.3) 0.3921 (0.2686-0.5723) Ͻ.0001 Fixed 30/162 (18.5) 78/154 (50.6) 0.3655 (0.2561-0.5217) Ͻ.0001 Random 30/162 (18.5) 78/154 (50.6) 0.3718 (0.2609-0.5300) Ͻ.0001 Fixed combined (13) 128/945 (13.5) 211/769 (27.4) 0.5485 (0.4549-0.6613) Ͻ.0001 Random combined (13) 128/945 (13.5) 211/769 (27.4) 0.5288 (0.3717-0.7522) .0004

removing the study in cancer patients by Abdelkefi et al.77 significance (relative risk, 0.42; 95% CI, 0.16 to 1.06). For A similar analysis for all catheter-associated deep vein fixed low-dose vitamin-K antagonist, the summary relative thrombosis did not materially alter the summary relative risk for symptomatic catheter-associated deep vein throm- risk; however, the estimate no longer reached statistical bosis increased from 0.60 to 1.28 (95% CI, 0.45 to 3.61)

Figure 2 Relative risk of all catheter-associated deep vein thrombosis with anticoagulant prophylaxis. Kirkpatrick et al Catheter-Associated Thrombosis Prevention 905

Table 4 Relative Risk of Symptomatic Catheter-Associated Deep Vein Thrombosis with Anticoagulant Prophylaxis

Prophylaxis No Prophylaxis Citation n/n (%) n/n (%) Relative Risk (95% Conﬁdence Interval) P Value Fixed low-dose heparin Abdelkeﬁ et al, 200477 1/65 (1.5) 5/68 (7.4) 0.2092 (0.0251-1.7430) .1063 Fabri et al, 198472 0/20 (0) 0/20 (0) 1.0000 (0.0208-48.0855) 1.0000 Fixed 1/85 (1.2) 5/88 (5.7) 0.2826 (0.0473-1.6886) .1659 Random 1/85 (1.2) 5/88 (5.7) 0.3001 (0.0467-1.9269) .2046 Low-molecular-weight heparin Karthaus et al, 200681 10/293 (3.4) 5/145 (3.5) 0.9898 (0.3447-2.8423) .9847 Monreal et al, 199675 0/16 (0) 0/13 (0) 0.8235 (0.0174-38.9166) .9213 Verso et al, 200583 2/191 (1.1) 6/194 (3.1) 0.3386 (0.0692-1.6566) .1595 Fixed 12/500 (2.4) 11/352 (3.1) 0.6890 (0.2995-1.5850) .3808 Random 12/500 (2.4) 11/352 (3.1) 0.7177 (0.3047-1.6905) .4480 Fixed low-dose vitamin K antagonist Bern et al, 199074 4/42 (9.5) 13/40 (32.5) 0.2930 (0.1042-0.8238) .0103 Couban et al, 200582 6/130 (4.6) 5/125 (4.0) 1.1538 (0.3613-3.6849) .8089 Heaton et al, 200276 2/45 (4.4) 1/43 (2.3) 1.9111 (0.1798-20.3173) .5840 Fixed 12/217 (5.5) 19/208 (9.1) 0.6039 (0.3044-1.1983) .1492 Random 12/217 (5.5) 19/208 (9.1) 0.6039 (0.2250-2.1395) .5247 Fixed combined (8) 25/802 (3.1) 35/648 (5.4) 0.5879 (0.3549-0.9740) .0392 Random combined (8) 25/802 (3.1) 35/648 (5.4) 0.6113 (0.3584-1.0426) .0708

after removing the study by Bern et al.74 For low-molecular- bosis, limiting the analysis to studies that were double- weight heparin, the summary relative risk for symptomatic blind81-83 increased the summary relative risk for fixed low- catheter-associated deep vein thrombosis increased from 0.69 dose vitamin-K antagonist from 0.60 to 1.15 (95% CI, 0.36 to to 0.98 (95% CI, 0.35 to 2.70) after removing the study by 3.68) and increased the summary relative risk for any antico- Verso et al.83 No other summary estimates of relative risk were agulant prophylaxis from 0.59 to 0.82 (95% CI, 0.41 to 1.62). materially altered by removal of individual studies except that Limiting the analysis to studies published after the year some estimates failed to reach statistical significance. 200077,81,83,93,94 did not materially alter any of the summary Limiting the analysis of all catheter-associated deep vein relative risks for all catheter-associated deep vein thrombo- thrombosis to studies that were double-blind73,81,83 in- sis. A similar analysis for symptomatic catheter-associated creased the summary relative risk for low-dose unfraction- deep vein thrombosis76,77,81-83 increased the summary related heparin infusion from 0.31 to 2.35 (95% CI, 0.23 to ative risk using fixed low-dose vitamin-K antagonist from 23.75). For symptomatic catheter-associated deep vein throm- 0.60 to 1.28 (95% CI, 0.45 to 3.61). Further subgroup

Figure 3 Relative risk of symptomatic catheter-associated deep vein thrombosis with anticoagulant prophylaxis. 906 The American Journal of Medicine, Vol 120, No 10, October 2007

Table 5 Relative Risk of Major Bleeding with Anticoagulant Prophylaxis

Prophylaxis No Prophylaxis Citation n/n (%) n/n (%) Relative Risk (95% Conﬁdence Interval) P Value Fixed low-dose heparin Abdelkeﬁ et al, 200477 2/55 (3.6) 3/53 (5.7) 0.6424 (0.1118-3.6931) .6168 Brismar et al, 198278 0/23 (0) 0/26 (0) 1.1250 (0.0232-54.5411) .9525 Fabri et al, 198472 0/20 (0) 0/20 (0) 1.0000 (0.0208-48.0855) 1.0000 Fixed 2/98 (2.0) 3/99 (3.0) 0.7432 (0.1718-3.2157) .6913 Random 2/98 (2.0) 3/99 (3.0) 0.7427 (0.1700-3.2445) .6925 Low-molecular-weight heparin Conte et al, 200393 0/38 (0) 0/20 (0) 0.5385 (0.0111-26.1756) .7511 Karthaus et al, 200681 1/285 (0.4) 0/140 (0.7) 0.4912 (0.0310-7.7957) .6069 Monreal et al, 199675 0/17 (0) 0/15 (0) 0.8889 (0.0187-42.2610) .9523 Verso et al, 200583 0/189 (0) 0/193 (0) 1.0211 (0.0204-51.1956) .9917 Fixed 1/529 (0.2) 1/368 (0.3) 0.6582 (0.1177-3.6802) .6339 Random 1/529 (0.2) 1/368 (0.3) 0.6526 (0.1143-3.7265) .6312 Fixed low-dose vitamin K antagonist Couban et al, 200582 0/130 (0) 3/125 (2.4) 0.1374 (0.0072-2.6333) .1208 Heaton et al, 200276 0/45 (0) 0/43 (0) 0.9565 (0.0194-47.1638) .9822 Fixed 0/175 (0) 3/168 (1.8) 0.2400 (0.0270-2.1320) .2003 Random 0/175 (0) 3/168 (1.8) 0.2788 (0.0265-2.9347) .2875 Fixed combined (9) 3/802 (0.4) 7/635 (1.1) 0.5356 (0.2025-1.4163) .2082 Random combined (9) 3/802 (0.4) 7/635 (1.1) 0.5923 (0.2146-1.6350) .3121

analyses did not materially alter any other summary esti- were distributed on either side of the summary estimate, mates of relative risk; however, some estimates no longer suggesting that studies with significant and nonsignificant reached statistical significance. results were equally represented in our analysis. The summary estimate of the relative risk of symptomatic catheter-associated deep vein thrombosis for any anticoagulant no longer achieved statistical significance using DISCUSSION the more conservative random-effects model (relative risk, Our results indicate that all of the anticoagulant regimens 0.61; 95% CI, 0.36 to 1.04). Using the random-effects evaluated are effective for preventing all catheter-associated model did not alter the results for any other outcomes. deep vein thrombosis with relative risk reductions of 27% to Funnel plots of effect size versus precision show no 70% (absolute risk reduction 9% to 32%) (Table 3). The evidence of publication bias. A similar number of studies results also provide suggestive evidence that anticoagulant

Figure 4 Relative risk of major bleeding with anticoagulant prophylaxis. Kirkpatrick et al Catheter-Associated Thrombosis Prevention 907 prophylaxis is effective for preventing symptomatic cathe- meaningful benefit. Additional randomized controlled trials ter-associated deep vein thrombosis, based on the pooled evaluating these outcomes are needed. analysis across anticoagulant regimens (Table 4). The ef- A recently completed randomized trial89 compared fixed fectiveness of any individual regimen for preventing symp- low-dose warfarin (1 mg) with either adjusted-dose warfarin tomatic catheter-associated deep vein thrombosis or pulmo- (international normalized ratio 1.5 to 2.0) or no prophylaxis nary embolism remains uncertain due to the small number in cancer patients receiving chemotherapy via central ve- of symptomatic events in each subgroup. Confidence inter- nous catheters. Because the outcome data were reported for vals for the summary relative risks were broad, however, only 90% of patients, this trial was not included in our and do not exclude the possibility of a potentially clinically analysis. The results suggest that adjusted-dose warfarin important benefit. was more effective than fixed-dose warfarin, reducing the The clinical importance of asymptomatic catheter-asso- incidence of symptomatic thrombosis from 7% to 3%; how- ciated deep vein thrombosis is incompletely under- ever, major bleeding increased from 2% to 4%. The ob- stood,94,95 although catheter-associated deep vein thrombo- served rates of symptomatic thrombosis for fixed-dose war- sis has been associated with clinically important pulmonary farin or no prophylaxis were similar (5% and 6%, embolism,2,3,96 including fatal embolism. Anticoagulant respectively, odds ratio 0.94; 95% CI, 0.52 to 1.72). The prophylaxis, therefore, has the potential to prevent serious 95% confidence interval does not exclude the possibility of thromboembolic events. a clinically important benefit of fixed low-dose warfarin for Our analysis does not support the view that the mate- preventing symptomatic thrombotic events. Because clini- rials and structure of central venous catheters have im- cians selected which patients entered into the randomized proved recently, leading to lower rates of catheter-asso- comparison of either fixed-dose warfarin versus no prophy- ciated thrombosis. Limiting our analysis to studies laxis, or to the comparison of fixed versus adjusted-dose published after the year 2000 did not change the reduc- warfarin, the results may be biased against fixed low-dose tion in relative risk of all catheter-associated deep vein warfarin if clinicians were effective at selecting patients at thrombosis using any anticoagulant regimen, and the lower risk for thromboembolism for comparison with no observed rates of all deep vein thrombosis without pro- prophylaxis, or patients at higher risk of thromboembolism phylaxis in these studies ranged from 7.6% to 55.3%. A for comparison against adjusted-dose warfarin. similar analysis for symptomatic catheter-associated Three previous systematic reviews have evaluated the deep vein thrombosis increased the summary relative risk efficacy of anticoagulant prophylaxis for prevention of cath- in only one subgroup of studies evaluating fixed low-dose eter-associated deep vein thrombosis.14,15,30 None of these vitamin-K antagonists. Due to smaller combined sample reviews included 3 randomized controlled trials77,81,94 that sizes, 95% confidence intervals for this estimate were broad were included in our analysis. and do not exclude the potential for a clinically meaningful Limitations of our meta-analysis include the potential for benefit. publication bias. Although our funnel plots did not suggest Our analysis found that the absolute risk of major bleed- publication bias, the low number of trials makes definitive ing was low for each of the regimens (Table 5). The pooled interpretation difficult. Our search strategy was comprehen- analysis did not detect an increase in major bleeding asso- sive, including multiple databases, abstracts presented at ciated with anticoagulant prophylaxis. The absolute in- national meetings, and manual searches for articles not crease in the incidences of major bleeding with anticoagu- retrieved by database searches. The overall limited number lant prophylaxis compared with no prophylaxis is unlikely of trials retrieved, and their small sample sizes, limited our to be greater than 0.9% for any regimen. ability to precisely quantitate the efficacy of individual Analysis of 2 randomized trials80,94 comparing fixed anticoagulant prophylaxis regimens for reducing symptom- low-dose vitamin-K antagonist to low-molecular-weight atic venous thromboembolism. In addition, tests of hetero- heparin indicates that low-molecular-weight heparin is less geneity and sensitivity analyses were limited by the small effective for preventing all catheter-associated deep vein number of trials. Finally, only 3 trials evaluating symptom- thrombosis than fixed low-dose vitamin-K antagonist pro- atic outcomes were double-blinded with concealment of phylaxis (summary relative risk 1.9; 95% CI, 1.3 to 2.7). randomization.80-82 These limitations all support the need This finding is contradictory to results of recent clinical for additional high-quality randomized controlled trials. trials showing superior efficacy of low-molecular-weight In conclusion, anticoagulant prophylaxis is effective for heparin compared with vitamin-K antagonists for treatment preventing all catheter-associated deep vein thrombosis. of established venous thromboembolism in cancer pa- This finding is driven primarily by a reduction in asymp- tients,97,98 and requires further evaluation. tomatic thrombi detected by mandatory screening for deep We did not detect a statistically significant reduction in vein thrombosis. The effectiveness of individual anticoag- either pulmonary embolism or all-cause mortality with an- ulant regimens for preventing symptomatic venous throm- ticoagulant prophylaxis. Due to the low frequency of these boembolism, including pulmonary embolism, remains un- outcomes, 95% confidence intervals for summary relative certain. Based on the available evidence, the risk-benefit of risks were broad and did not exclude the potential for anticoagulant prophylaxis for preventing clinically impor- 908 The American Journal of Medicine, Vol 120, No 10, October 2007 tant venous thromboembolic events is uncertain. With in- 21. von Depka Prondzinski M, Karthaus M, Ganser A, Barthels M. Anti- creasing central venous catheter use, there is a need for coagulant prophylaxis and therapy in patients with cancer. Antibiot large rigorous randomized trials to resolve this issue. Chemother. 2000;50:149-158. 22. Otten HM, Prins MH, Smorenburg SM, Hutten BA. Risk assessment Currently, the decision to use anticoagulant prophylaxis and prophylaxis of venous thromboembolism in non-surgical patients: remains a clinical judgment based on the estimated risks cancer as a risk factor. Haemostasis. 2000;30(Suppl 2):72-76. of thromboembolism and major bleeding in the individ- 23. Slagle DC, Gates RH Jr. Unusual case of central vein thrombosis and ual patient. sepsis. Am J Med. 1986;81(2):351-354. 24. Krzywda EA. Predisposing factors, prevention, and management of central venous catheter occlusions. J Intraven Nurs. 1999;22(6 Suppl): References S11-S17. 1. Verso M, Agnelli G. Venous thromboembolism associated with long- 25. Murphy LM, Lipman TO. Central venous catheter care in parenteral term use of central venous catheters in cancer patients. J Clin Oncol. nutrition: a review. JPEN J Parenter Enteral Nutr. 1987;11(2):190-201. 2003;21:3665-3675. 26. Prandoni P, Piccioli A. Venous thromboembolism and cancer: a two- 2. Prandoni P, Polistena P, Bernardi E, et al. Upper-extremity deep vein way clinical association. Front Biosci. 1997;2:e12-e20. thrombosis: risk factors, diagnosis, and complications. Arch Intern 27. Levine MN. Prevention of thrombotic disorders in cancer patients Med. 1997;157:57-62. undergoing chemotherapy. Thromb Haemost. 1997;78(1):133-136. 3. Monreal M, Raventos A, Lerma R, et al. Pulmonary embolism in 28. Bona RD. Thrombotic complications of central venous catheters in patients with upper extremity DVT associated to venous central cancer patients. Semin Thromb Hemost. 1999;25(2):147-155. lines—a prospective study. Thromb Haemost. 1994;72(4):548-550. 29. Levine M. Is antithrombotic prophylaxis required in cancer patients 4. Geerts WH, Heit JA, Clagett GP, et al. Prevention of venous throm- with central venous catheters? Still an open question J Thromb Hae- boembolism. Chest. 2001(Suppl);119(1):132S-175S. most. 2006;4(1):16-18. 5. Geerts WH, Pineo GF, Heit JA, et al. Prevention of venous thrombo- 30. Cunningham MS, White B, Hollywood D, O’Donnell J. Primary embolism. Chest. 2004(Suppl);126(3):338S-400S. thromboprophylaxis for cancer patients with central venous cathe- 6. Mulrow CD, Cook DJ, Davidoff F. Systematic reviews: critical links ters—a reappraisal of the evidence. Br J Cancer. 2006;94:189-194. in the great chain of evidence. Ann Intern Med. 1997;126(5):389-391. 31. Burton A. Catheter-related clotting not reduced by dalteparin. Lancet 7. McAlister FA, Clark HD, van Walraven C, et al. The medical review Oncol. 2006;7(3):201. article revisited: has the science improved? Ann Intern Med. 1999; 32. Moia M, Cortelezzi A, Falanga A. Catheter-related thrombosis in 131(12):947-951. hematologic patients. Rev Clin Exp Hematol. 2004;8(1):E5. 8. Guyatt GH, Sackett DL, Cook DJ. Users’ guides to the medical 33. Huisman MV. Is antithrombotic prophylaxis required in cancer pa- literature. II. How to use an article about therapy or prevention. A. Are tients with central venous catheters? Yes for special patient groups. J the results of the study valid? Evidence-Based Medicine Working Thromb Haemost. 2006;4:10-13. Group. JAMA. 1993;270(21):2598-2601. 34. Agnelli G, Verso M. Is antithrombotic prophylaxis required in cancer 9. Guyatt GH, Sackett DL, Cook DJ. Users’ guides to the medical patients with central venous catheters? No. J Thromb Haemost. 2006; literature. II. How to use an article about therapy or prevention. B. 4:14-15. What were the results and will they help me in caring for my 35. Laporte S, Decousus H, Mismetti P. Thrombosis prophylaxis in cancer patients? Evidence-Based Medicine Working Group. JAMA. 1994; patients with a central venous catheter. Arch Intern Med. 2004;164:459. 271(1):59-63. 36. Van Rooden CJ, Tesselaar MET, Osanto FR, Rosendaal FR, Huisman 10. Graafsma YP, Prins MH, Lensing AW, de Haan RJ, Huisman MV, MV. Deep vein thrombosis associated with central venous catheter—a Buller HR. Bleeding classification in clinical trials: observer variabil- review. J Thromb Haemost. 2005;3:2409-2419. ity and clinical relevance. Thromb Haemost. 1997;78(4):1189-1192. 37. Agnelli G, Verso M. Therapy insight: venous-catheter-related throm- 11. Lau J, Ioannidis JPA, Schmid CH. Quantitative synthesis in systematic bosis in cancer patients. Nat Clin Pract Oncol. 2006;3(2):214-222. reviews. Ann Intern Med. 1997;127:820-826. 38. Cortelezzia A, Fracchiolla NS, Maisonneuve P, et al. Central venous 12. Sacks HS, Berrier J, Reitman D, Ancona-Berk VA, Chalmers TC. catheter-related complications in patients with hematological malig- Meta-analyses of randomized controlled trials. N Engl J Med. 1987; nancies: a retrospective analysis of risk factors and prophylactic mea- 316(8):450-455. sures. Leuk Lymphoma. 2003;44(9):1495-1501. 13. Sutton AJ, Abrams KR, Jones DR, Sheldon TA, Song F. Methods for 39. Bodner LJ, Nosher JL, Patel KM, et al. Peripheral venous access ports: Meta-analysis in Medical Research. New York, NY: John Wiley & outcomes analysis in 109 patients. Cardiovasc Intervent Radiol. 2000; Sons, Inc.; 2000. 23(3):187-193. 14. Randolph AG, Cook DJ, Gonzales CA, Andrew M. Benefit of heparin 40. Carr KM, Rabinowitz I. Physician compliance with warfarin prophy- in central venous and pulmonary artery catheters: a meta-analysis of laxis for central venous catheters in patients with solid tumors. J Clin randomized controlled trials. Chest. 1998;113(1):165-171. Oncol. 2000;18(21):3665-3667. 15. Klerk CP, Smorenburg SM, Buller HR. Thrombosis prophylaxis in 41. Lagro SW, Verdonck LF, Borel Rinkes IH, Dekker AW. No effect of patient populations with a central venous catheter: a systematic review. nadroparin prophylaxis in the prevention of central venous catheter Arch Intern Med. 2003;163(16):1913-1921. (CVC)-associated thrombosis in bone marrow transplant recipients. 16. Kakkar AK, Williamson RC. Prevention of venous thromboembolism Bone Marrow Transplant. 2000;26(10):1103-1106. in cancer using low-molecular-weight heparins. Haemostasis. 1997; 42. Kaufman JA, Salamipour H, Geller SC, Rivitz SM, Waltman AC. 27(Suppl 1):32-37. Long-term outcomes of radiologically placed arm ports. Radiology. 17. Kakkar AK, Williamson RC. Thromboprophylaxis in the cancer pa- 1996;201(3):725-730. tient. Haemostasis. 1998;28(Suppl 3):61-65. 43. Imperial J, Bistrian BR, Bothe A Jr, Bern M, Blackburn GL. Limita- 18. Letai A, Kuter DJ. Cancer, coagulation, and anticoagulation. Oncolo- tion of central vein thrombosis in total parenteral nutrition by contin- gist. 1999;4(6):443-449. uous infusion of low-dose heparin. J Am Coll Nutr. 1983;2(1):63-73. 19. Brismar B, Nystrom B. Thrombophlebitis and septicemia—complica- 44. Walters MB, Stanger HA, Rotem CE. Complications with percutane- tions related to intravascular devices and their prophylaxis. A review. ous central venous catheters. JAMA. 1972;220(11):1455-1457. Acta Chir Scand Suppl. 1986;530:73-77. 45. Kuriakose P, Colon-Otero G, Paz-Fumagalli R. Risk of deep venous 20. Denny DF Jr. Placement and management of long-term central thrombosis associated with chest versus arm central venous subcuta- venous access catheters and ports. AJR Am J Roentgenol. 1993; neous port catheters: a 5-year single-institution retrospective study. J 161(2):385-393. Vasc Interv Radiol. 2002;13(2 Pt 1):179-184. Kirkpatrick et al Catheter-Associated Thrombosis Prevention 909

46. Minassian VA, Sood AK, Lowe P, Sorosky JI, Al-Jurf AS, Buller RE. 66. Lokich JJ, Bothe A Jr, Benotti P, Moore C. Complications and man- Longterm central venous access in gynecologic cancer patients. JAm agement of implanted venous access catheters. J Clin Oncol. 1985; Coll Surg. 2000;191(4):403-409. 3(5):710-717. 47. Duerksen DR, Ahmad A, Doweiko J, Bistrian BR, Mascioli EA. Risk 67. Timsit JF, Farkas JC, Boyer JM, et al. Central vein catheter-related of symptomatic central venous thrombotic complications in AIDS thrombosis in intensive care patients: incidence, risks factors, and patients receiving home parenteral nutrition. JPEN J Parenter Enteral relationship with catheter-related sepsis. Chest. 1998;114(1):207-213. Nutr. 1996;20(4):302-305. 68. Chastre J, Cornud F, Bouchama A, Viau F, Benacerraf R, Gibert C. 48. Eastman ME, Khorsand M, Maki DG, et al. Central venous device- Thrombosis as a complication of pulmonary-artery catheterization via related infection and thrombosis in patients treated with moderate dose the internal jugular vein: prospective evaluation by phlebography. continuous-infusion interleukin-2. Cancer. 2001;91(4):806-814. N Engl J Med. 1982;306(5):278-281. 49. Veerabagu MP, Tuttle-Newhall J, Maliakkal R, Champagne C, Mas- 69. Bern MM, Bothe A Jr, Bistrian B, Champagne CD, Keane MS, cioli EA. Warfarin and reduced central venous thrombosis in home Blackburn GL. Prophylaxis against central vein thrombosis with low- total parenteral nutrition patients. Nutrition. 1995;11(2):142-144. dose warfarin. Surgery. 1986;99(2):216-221. 70. Levine M, Hirsh J, Gent M, et al. Double-blind randomised trial of a 50. Magnagoli M, Masci G, Castagna L, et al. Prophylaxis of central very-low-dose warfarin for prevention of thromboembolism in stage venous catheter-related thrombosis with minidose warfarin in patients IV breast cancer. Lancet. 1994;343(8902):886-889. treated with high-dose chemotherapy and peripheral-blood stem-cell 71. Fabri PJ, Mirtallo JM, Ruberg RL, et al. Incidence and prevention of transplantation: retrospective analysis of 228 cancer patients. Am J thrombosis of the subclavian vein during total parenteral nutrition. Hematol. 2006;81(1):1-4. Surg Gynecol Obstet. 1982;155(2):238-240. 51. Daniell HW. Heparin prevention of catheter thrombosis. JAMA. 1971; 72. Fabri PJ, Mirtallo JM, Ebbert ML, Kudsk KA, Powell C, Ruberg RL. 216(7):1200. Clinical effect of nonthrombotic total parenteral nutrition catheters. 52. Thromboembolic complications and central venous catheters. Lancet. JPEN J Parenter Enteral Nutr. 1984;8(6):705-707. 1982;2(8304):936-937. 73. Macoviak JA, Melnik G, McLean G, et al. The effect of low-dose 53. Mailloux RJ, DeLegge MH, Kirby DF. Pulmonary embolism as a heparin on the prevention of venous thrombosis in patients receiving complication of long-term total parenteral nutrition. JPEN J Parenter short-term parenteral nutrition. Curr Surg. 1984;41(2):98-100. Enteral Nutr. 1993;17(6):578-882. 74. Bern MM, Lokich JJ, Wallach SR, et al. Very low doses of warfarin 54. Loughlin V, Lalsingh R. Pulmonary artery flotation catheters and can prevent thrombosis in central venous catheters. A randomized microemboli. Br J Parenter Ther. 1986;122-123. prospective trial. Ann Intern Med. 1990;112(6):423-428. 55. Wu X, Studer W, Skarvan K, Seeberger MD. High incidence of 75. Monreal M, Alastrue A, Rull M, et al. Upper extremity deep venous intravenous thrombi after short-term central venous catheterization of thrombosis in cancer patients with venous access devices—prophy- the internal jugular vein. J Clin Anesth. 1999;11(6):482-485. laxis with a low molecular weight heparin (Fragmin). Thromb Hae- 56. Mainwaring CJ, Naylor E, Jerwood S, Hall V, Page A, James CM. most. 1996;75(2):251-253. Superior venocaval obstruction secondary to central venous catheter- 76. Heaton DC, Han DY, Inder A. Minidose (1 mg) warfarin as pro- related thromboses in two patients with metastatic colorectal carci- phylaxis for central vein catheter thrombosis. Intern Med J. 2002; noma receiving weekly 5-flurouracil. Should adjusted-dose warfarin 32(3):84-88. be used as thromboprophylaxis? Blood Coagul Fibrinolysis. 2002; 77. Abdelkefi A, Ben Othman T, Kammoun L, et al. Prevention of central 13(8):749-753. venous line-related thrombosis by continuous infusion of low-dose 57. Koksoy C, Kuzu A, Erden I, Akkaya A. The risk factors in central unfractionated heparin, in patients with haemato-oncological disease. venous catheter-related thrombosis. AustNZJSurg. 1995;65(11): A randomized controlled trial. Thromb Haemost. 2004;92(3):654-661. 796-798. 78. Brismar B, Hardstedt C, Jacobson S, Kager L, Malmborg AS. Reduc- 58. Boraks P, Seale J, Price J, et al. Prevention of central venous catheter tion of catheter-associated thrombosis in parenteral nutrition by intra- associated thrombosis using minidose warfarin in patients with haema- venous heparin therapy. Arch Surg. 1982;117(9):1196-1199. tological malignancies. Br J Haematol. 1998;101:483-486. 79. Ruggiero RP, Aisenstein TJ. Central catheter fibrin sleeve—heparin 59. Masci G, Magagnoli M, Zucali PA, et al. Minidose warfarin prophy- effect. JPEN J Parenter Enteral Nutr. 1983;7(3):270-273. laxis for catheter-associated thrombosis in cancer patients: can it be 80. Mismetti P, Mille D, Laporte S, et al. Low-molecular-weight heparin safely associated with fluorouracil-based chemotherapy? J Clin Oncol. (nadroparin) and very low doses of warfarin in the prevention of upper 2003;21(4):736-739. extremity thrombosis in cancer patients with indwelling long-term central venous catheters: a pilot randomized trial. Haematologica. 60. Lersch C, Eckel F, Sader R, et al. Initial experience with Healthport 2003;88(1):67-73. miniMax and other peripheral arm ports in patients with advanced 81. Karthaus M, Kretzschmar A, Kroning H, et al. Dalteparin for preven- gastrointestinal malignancy. Oncology. 1999;57(4):269-275. tion of catheter-related complications in cancer patients with central 61. Lersch C, Kotowa W, Fung S, Janssen D. Prophylaxis of port system- venous catheters: final results of a double-blind, placebo-controlled associated thromboses in advanced oncology patients using heparin phase III trial. Ann Oncol. 2006;17:289-296. flushing. J Cancer Res Clin Oncol. 2004;130(4):235-241. 82. Couban S, Goodyear M, Burnell M, et al. Randomized placebo- 62. Zanon C, Bortolini M, Chiappino I. Low-molecular-weight heparin controlled study of low-dose warfarin for the prevention of central and calcium heparin in thrombosis prophylaxis in patients with per- venous catheter-associated thrombosis in patients with cancer. J Clin cutaneous arterial and venous ports for colorectal liver metastases. Oncol. 2005;23(18):4063-4069. Tumori. 2005;91(6):477-480. 83. Verso M, Agnelli G, Bertoglio S, et al. Enoxaparin for the prevention 63. Ratcliffe M, Broadfoot C, Davidson M, Kelly KF, Greaves M. Throm- of venous thromboembolism associated with central vein catheter: a bosis, markers of thrombotic risk, indwelling central venous catheters double-blind, placebo-controlled randomized study in cancer patients. and antithrombotic prophylaxis using low-dose warfarin in subjects J Clin Oncol. 2005;23(18):4057-4062. with malignant disease. Clin Lab Haematol. 1999;21(5):353-357. 84. Magagnoli M, Masci G, Castagna L, et al. Retrospective analysis of 64. Bozzetti F, Scarpa D, Terno G, et al. Subclavian venous thrombosis central venous catheter-related thrombosis in 427 cancer patients pro- due to indwelling catheters: a prospective study on 52 patients. JPEN phylaxed with low-dose warfarin [abstract]. J Clin Oncol. 2004; J Parenter Enteral Nutr. 1983;7(6):560-562. 22(suppl):754S, abstract 8102. 65. Haire WD, Edney JA, Landmark JD, Kessinger A. Thrombotic com- 85. Nguyen NT, Kassis J. Thrombotic and infectious complications related plications of subclavian apheresis catheters in cancer patients: preven- to central venous catheters in patients with or without cancer [ab- tion with heparin infusion. J Clin Apher. 1990;5(4):188-191. stract]. Blood. 2003;102(Suppl):abstract 4158. 910 The American Journal of Medicine, Vol 120, No 10, October 2007

86. Tesselaar MET, Ouwerkerk J, Rosendaal FR, Osanto S. Prophylaxis 92. Agnelli G, Verso M, Bertoglio S, et al. A double-blind placebo- with low molecular weight heparin reduces the risk for catheter-related controlled randomized study on the efﬁcacy and safety of enoxaparin venous thrombosis in cancer patients with centrally but not peripher- for the prevention of upper limb deep vein thrombosis in cancer ally inserted central venous catheters for administration of chemo- patients with central vein catheter. J Clin Oncol. 2004;22(Suppl): therapy [abstract]. Suppl Thromb Haemost. 2001;abstract P1501. abstract 8021:734s. Available at http://www.cartesian-secure.com/isth2001/iAbstract/ 93. Conte GF, Aravena PC, Fardella PD, et al. Prophylaxis of venous html/absP1501.html. Accessed August 14, 2007. thrombosis (VT) associated with central venous catheter (CVC) with 87. Iniesta P, Garcia T, Ayala F, et al. Dalteparin is effective as prophy- low molecular weight heparin (LMWH) in hematologic malignancies laxis of central venous catheter-related thrombosis in cancer patients [abstract]. Blood. 2003;102(Suppl 11):122b. [abstract]. Proc Am Soc Clin Oncol. 2003;22(Suppl):abstract 3151: 94. DeCicco M, Matovic M, Pacenzia R, et al. Short-term acenocumarine 784. (A) or dalteparin (D) for the prevention of central venous catheter- related thrombosis (CVCrT) in cancer patients. A randomized con- 88. Park K, Oh SY, Kim WS, et al. Randomized phase III trial of very low trolled study based on serial venographies [abstract]. J Clin Oncol. dose warfarin to prevent catheter-associated thrombosis [abstract]. Rev 2006;24(18S):abstr 8549:480. Prog Proc Am Soc Clin Oncol. 1999;18(Suppl):abstract 2330:603a. 95. Hull RD, Raskob GE, Pineo GF, eds. Venous Thromboembolism: An 89. Young AM, Begum G, Billingham LJ, et al. WARP-A multicenter Evidence-Based Atlas. New York, NY: Futura Publishing Company, prospective randomized controlled trial of thrombosis prophylaxis Inc.; 1996:309. with warfarin in cancer patients with central venous catheters. J Clin 96. Verso M, Agnelli G. Venous thromboembolism associated with long- Oncol. 2005;23(Suppl):730S, abstr LBA8004. term use of central venous catheters in cancer patients. J Clin Oncol. 90. Reichardt P, Kretzschmar A, Biakhov M, et al. A phase III double- 2003;21(19):3665-3675. blind, placebo-controlled study evaluating the efﬁcacy and safety of 97. Lee AY, Levine MN, Baker RI, et al. Low-molecular-weight heparin daily low-molecular-weight heparin (dalteparin sodium, Fragmin) in versus a coumarin for the prevention of recurrent venous thrmoboem- preventing catheter-related complications in cancer patients with cen- bolism in patients with cancer. N Engl J Med. 2003;349:146-153. tral venous catheters [abstract]. J Clin Oncol. 2002;21(Suppl):703a, 98. Hull R, Pineo GF, Mah A. A randomized trial evaluating long-term abstract 1474. low-molecular-weight heparin therapy for three months vs intravenous 91. Couban S, Goodyear M, Burnell M, et al. A randomized double-blind heparin followed by warfarin sodium in patients with current cancer placebo-controlled study of low dose warfarin for the prevention of (abstract). Thromb Haemost. 2003;(Suppl):P1373a. Available at http:// symptomatic central venous catheter associated thrombosis in patients www.blackwellpublishing.com/isth2003/abstract.asp?idϭ9558. with cancer [abstract]. Blood. 2002;100(Suppl):703a. Accessed August 14, 2007. ikarc ta ahtrAscae hobssPrevention Thrombosis Catheter-Associated al et Kirkpatrick Table 1 Study Characteristics

Vein Length Sample Catheter Where Duration Onset Duration Prophylaxis of Author/Year Size Drug Dose Route Frequency Type Inserted Catheterization Prophylaxis Prophylaxis Continuous Follow-up Fabri et al, 46 Heparin 3000 USP units/L IV With TPN Central Subclavian Not clear After Not clear Not Clear Not Clear 198271 line insertion Fabri et al, 40 Heparin 3000 units/L IV With TPN Central Subclavian 22.1 Ϯ 3.2 After Not clear Yes 22.1 Ϯ 3.2 198472 line days insertion days Brismar et al, 49 Heparin 5000 U IV Every 6 PICC External 7-94 days At 7-94 days No 7-94 days 198278 hours jugular (mean 25 insertion (mean (mean days) 25 days) 25 days) Ruggiero et al, 34 Heparin 1000 U/L IV With TPN Central Subclavian Ͼ14 days in Not clear 7-43 days Yes Ͼ14 days 198379 line 18 patients (mean in 18 18 days) patients Macoviak et al, 37 Heparin 1 unit/mL IV With TPN Central Subclavian At least 4 After At least 4 No 4 Weeks 198473 (2 L) line weeks insertion weeks Abdelkeﬁ et al, 108 Heparin 100 IU/Kg/D IV Continuous Central Subclavian 8-81 days Not clear 8-81 days Yes 8 Weeks 200477 infusion line after removal Bern et al, 121 Warfarin 1 mg PO Daily Tunneled Subclavian Not stated Before 90 days or No 90 days 199074 insertion DVT Heaton et al, 98 Warfarin 1 mg PO Daily Tunneled Subclavian 30 Lines At Mean 41 No 90 days 200276 removed insertion days before 90 days Couban et al, 255 Warfarin 1 mg PO Daily Tunneled, Subclavian, Median 25 After Median 8 No 3 months 200582 PICC, IJ, weeks insertion weeks after port basilic, line cephalic removed Monreal et al, 29 Dalteparin 2500 units SC Daily Tunneled Subclavian Not stated Before 90 days or Yes 90 days 199675 insertion DVT Conte et al, 58 Dalteparin 5000 IU SC Daily Central Subclavian Ͼ1 week Not clear 7 days Yes 7 days 200393 line &IJ Verso et al, 310 Enoxaparin 40 mg SC Daily Tunneled, Subclavian, 42 Ϯ 2 days Before 42 Ϯ 2 Yes 3 months 200583 PICC IJ and insertion days EJ Karthaus et al, 425 Dalteparin 5000 IU SC Daily Not clear Not clear At least 12 Within 5 16 weeks Yes 16 weeks 200681 weeks days of insertion Mismetti et al, 59 Nadroparin vs 2850 units vs SC and Daily Tunneled Subclavian 90 days except Before 90 Ϯ 5 For ﬁrst 90 6 months 200380 Warfarin 1mg PO and IJ for 12 caths insertion days or days, yes Ͻ90 days DVT DeCicco et al, 450 Dalteparin vs 5000 units vs SC and Daily Not clear Not clear Not clear Before 8 days Not clear Not clear 200694 Acenocumarine 1mgvsNo PO insertion vs Control Prophylaxis 910.e1 USP ϭ United States Pharmacopeial Convention; L ϭ liter; U ϭ units; mL ϭ milliliter; IU ϭ international units; Kg ϭ kilogram; D ϭ day; mg ϭ milligram; IV ϭ intravenously; PO ϭ orally; SC ϭ subcutaneously; TNP ϭ total parenteral nutrition; PICC ϭ peripherally inserted central catheter; IJ ϭ internal jugular; EJ ϭ external jugular; DVT ϭ deep venous thrombosis. 910.e2 The American Journal of Medicine, Vol 120, No 10, October 2007

Table 2 Study Quality

Consecutive Randomization Similar at Treated Author/Year Drug Enrollment Concealed Baseline Similarly Patient Spectrum Fabri et al, 198271 Heparin Not clear Not clear Yes Not clear Parenteral nutrition Fabri et al, 198472 Heparin Yes Not clear Not clear Not clear Parenteral nutrition Brismar et al, 198278 Heparin Not clear Not clear Yes Not clear Parenteral nutrition Ruggiero et al, 198379 Heparin Yes Not clear Not clear Not clear Parenteral nutrition Macoviak et al, 198473 Heparin Not clear Yes Yes Not clear Parenteral nutrition Abdelkeﬁ et al, 200477 Heparin Yes Not clear Yes Yes Hematologic cancer or hematologic disorders, age Ͻ60 Bern et al, 199074 Warfarin Yes No Not clear Yes Cancer Heaton et al, 200276 Warfarin No Not clear Yes No Hematologic cancer

Couban et al, 200582 Warfarin Not clear Yes Yes Yes Cancer

Monreal et al, 199675 Dalteparin Yes Not clear Yes Not clear Cancer Conte et al, 200393 Dalteparin Not clear No Not clear Not clear Hematologic cancer

Verso et al, 200583 Enoxaparin Yes Yes Yes Yes Cancer Karthaus et al, 200681 Dalteparin Not clear Yes Not clear Not clear Cancer

Mismetti et al, 200380 Nadroparin vs warfarin Yes Yes No Yes Cancer DeCicco et al, 200694 Dalteparin vs Acenocumarine Yes Not clear Not clear Not clear Cancer vs control The American Journal of Medicine (2007) 120, 911

AJM Online

mimic mononucleosis, but fail to generate EBV’s archetypal triad of AJM Online is a special section of The Amer- clinical ﬁndings. A systematic approach to the diagnosis of mononucleosis- ican Journal of Medicine, which provides like illnesses ensures that conditions warranting speciﬁc therapy are dis- additional original research and reviews via the tinguished from others requiring only supportive care.

Internet. Every month, new peer-reviewed arti- CLINICAL RESEARCH STUDY cles covering important medical advances are Incidence and Clinical Spectrum of Thiazide-associated published in AJM Online. These articles are Hypercalcemia indexed in Index Medicus, ScienceDirect, and Wermers RA,a Kearns AE,a Jenkins GD,b Melton LJ III.a,c aDivision of Endocrinology, Department of Medicine, and bDivisions of MEDLINE, among other services. Abstracts Biostatistics and cEpidemiology, Department of Health Sciences from AJM Online submissions appear in every Research, Mayo Clinic College of Medicine, Rochester, Minn. issue of The American Journal of Medicine in PURPOSE: The study determines the incidence of thiazide-associated hypercalcemia and clarifies its clinical features and natural history. this section. The complete original articles are METHODS: In a population-based descriptive study, Olmsted County, Minn, available only online. residents with thiazide-associated hypercalcemia were identified through the Rochester Epidemiology Project and the Mayo Clinic Laboratory Information System. Changes in incidence rates were evaluated by Poisson regression. RESULTS: Seventy-two Olmsted County residents (68 women and 4 men; REVIEW mean age, 64 years) with thiazide-associated hypercalcemia first recog- Diagnostic Evaluation of Mononucleosis-Like Illnesses nized in 1992 to 2001 were identified. The overall annual age- and sex- Hurt C,a Tammaro D.b adjusted (to 2000 US whites) incidence was 7.7 (95% confidence interval aDepartment of Medicine, Division of Infectious Diseases, University of [CI], 5.9-9.5) per 100,000. There was an increase in incidence after 1996, North Carolina at Chapel Hill, School of Medicine, Chapel Hill, NC; peaking at 16.3 (95% CI, 8.3-24.3) per 100,000 in 1998. The highest rate bDepartment of Medicine, Division of General Internal Medicine, was 55.3 per 100,000 in 70- to 79-year-old women. Hypercalcemia was Brown Medical School, Providence, RI. identified a mean of 6 Ϯ 7 years after thiazide initiation, and the average Clinicians face a diagnostic challenge when a patient with the classic fever, highest serum calcium was 10.7 Ϯ 0.3 mg/dL with serum parathyroid pharyngitis, and lymphadenopathy triad of infectious mononucleosis has a hormone (obtained in 53 patients) of 4.8 Ϯ 2.7 pmol/L. Of 33 patients who negative “spot” heterophile antibody test. This screening test, although discontinued the thiazide, 21 (64%) had persistent hypercalcemia. Patients commonly considered sensitive for the presence of Epstein-Barr virus subsequently diagnosed with primary hyperparathyroidism had the highest (EBV) infection, may be negative early after infection. A growing number average serum calcium and parathyroid hormone levels of 11.0 Ϯ 0.3 of pathogens have been reported to cause heterophile-negative mononu- mg/dL and 6.3 Ϯ 2.4 pmol/L, respectively. cleosis-like illnesses, including cytomegalovirus (CMV), human herpesvi- CONCLUSION: The persistence of hypercalcemia in patients discontinuing rus 6 (HHV-6), human immunodeficiency virus (HIV), adenovirus, herpes thiazides, and similarities in the clinical spectrum, suggest that underlying simplex virus (HSV), Streptococcus pyogenes, and Toxoplasma gondii. primary hyperparathyroidism is common in patients who develop hyper- Other infectious and noninfectious disorders also may present in ways that calcemia while taking thiazide diuretics.

ERRATUM

Babbott SF, Beasley BW, Hinchey KT, Blotzer JW, The APM Perspectives article in the August issue of the Holmboe ES. The predictive validity of the Internal Journal was inadvertently published without the author’s Medicine In-Training Examination. Am J Med. 2007; corrections. The corrected version now appears online at 120:735-740. http://www.amjmed.com.

To access AJM Online articles and the complete text of this month’s Green Journal, go to www.amjmed.com/current

Table 2 Continued Patients Clinicians Outcomes Assessors Method Thrombus Major Bleeding Minor Bleeding All Patients Blinded Blinded Blinded Detection Deﬁned Deﬁned Accounted For Not clear Not clear Not clear Venogram all patients Not clear Not clear Yes Not clear Not clear Not clear Venogram all patients Not clear Not clear Yes No No Not clear Venogram all patients Not clear Yes Yes Not clear Not clear Yes Venogram all patients Not clear Not clear Yes Yes Yes Yes Venogram all patients Not clear Not clear Yes Not clear Not clear Yes Ultrasound all patients Yes No Yes

No No Yes Venogram all patients Not clear Not clear Yes No No Not clear Venogram symptomatic Not clear Not clear Yes patients Yes Yes Yes Ultrasound or venogram Yes Yes Yes symptomatic Not clear Not clear Yes Venogram all patients Not clear Not clear Yes No No Yes Echo-Doppler all Not clear Not clear Yes patients Yes Yes Yes Venogram all patients Yes Yes Yes Yes Yes Not clear Ultrasound and Yes No Yes venogram all No No Yes Venogram all patients Yes No Yes Not clear Not clear Not clear Venogram all patients Not clear Not clear Not clear The American Journal of Medicine (2007) 120, 911.e1-911.e8

REVIEW

Diagnostic Evaluation of Mononucleosis-Like Illnesses Christopher Hurt, MD,a Dominick Tammaro, MDb aDepartment of Medicine, Division of Infectious Diseases, University of North Carolina at Chapel Hill, School of Medicine, Chapel Hill, NC; bDepartment of Medicine, Division of General Internal Medicine, Brown Medical School, Providence, RI.

ABSTRACT

Clinicians face a diagnostic challenge when a patient with the classic fever, pharyngitis, and lymphadenopathy triad of infectious mononucleosis has a negative “spot” heterophile antibody test. This screening test, although commonly considered sensitive for the presence of Epstein-Barr virus (EBV) infection, may be negative early after infection. A growing number of pathogens have been reported to cause heterophile- negative mononucleosis-like illnesses, including cytomegalovirus (CMV), human herpesvirus 6 (HHV-6), human immunodeficiency virus (HIV), adenovirus, herpes simplex virus (HSV), Streptococcus pyogenes, and Toxoplasma gondii. Other infectious and noninfectious disorders also may present in ways that mimic mononucleosis, but fail to generate EBV’s archetypal triad of clinical findings. A systematic approach to the diagnosis of mononucleosis-like illnesses ensures that conditions warranting specific therapy are distinguished from others requiring only supportive care. © 2007 Elsevier Inc. All rights reserved.

KEYWORDS: Acute retroviral syndrome; Cytomegalovirus; Epstein-Barr virus; Human herpesvirus 6; Human immunodeﬁciency virus; Infectious mononucleosis; Mononucleosis-like illness; Toxoplasmosis

A 26-year-old graduate student presents with a 2-day his- the peripheral blood smears of 6 college students presenting tory of fever, headache, and sore throat. She denies any with glandular fever revealed striking similarities: an abso- rhinorrhea, cough, or sick contacts. Physical examination lute lymphocytosis, with atypically abundant cytoplasm in reveals slight tachycardia with normal temperature and many mononuclear cells. In 1932, Paul and Bunnell discov- blood pressure. Diffuse erythema of the pharynx is noted ered that serum from patients with IM caused sheep eryth- without tonsillar exudates. Her lungs are clear bilaterally. rocytes to agglutinate, and their so-called “heterophile” an- Rapid pharyngeal testing for group A streptococcal antigen tibody test became the basis for serologic diagnosis of is negative, and supportive care is advised. She returns infectious mononucleosis.2 several days later with a fever of 38.9°C, persistent pharyn- When a laboratory worker infected with the newly dis- geal erythema, and scattered tender anterior cervical lymph covered Epstein-Barr virus (EBV) in 1968 developed clin- nodes. The tip of the spleen is palpable. A heterophile ical symptoms of IM and heterophile antibodies,3 the cause antibody test for Epstein-Barr virus-induced infectious of the disease was finally identified. EBV accounts for mononucleosis is negative. How should you proceed? approximately 9 of every 10 clinical presentations suggestive of IM, and 25% to 30% of adolescents and adults up to 4 INFECTIOUS MONONUCLEOSIS age 30 years with primary EBV infection will fall ill. In contrast, childhood infection is generally subclinical. Although the clinical triad of pharyngitis, fever, and lymphadenopathy was first described in 1889 as “glandular fever,” Within industrialized societies, lower socioeconomic status groups are infected with EBV at younger ages than affluent it was not until 1920 that the first formal definition of 5 infectious mononucleosis (IM) was made.1 Examination of groups; whites in the United States are 30 times more likely than blacks to develop IM.6 More than 90% of adults worldwide who are seropositive for EBV have lifetime latent Requests for reprints should be addressed to Dominick Tammaro, MD, Rhode Island Hospital, Jane Brown Ground, Suite 0100, 593 Eddy Street, viral infection of their B lymphocytes and persistent viral Providence, RI 02903. shedding into saliva—the most probable source for transmis- E-mail address: [email protected] sion.7 The diagnosis of “infectious mononucleosis” is reserved

0002-9343/$ -see front matter © 2007 Elsevier Inc. All rights reserved. doi:10.1016/j.amjmed.2006.12.011 911.e2 The American Journal of Medicine, Vol 120, No 10, October 2007 for the syndrome caused by EBV, and similar presentations of those of sheep.16 Development of a slide-based test using caused by other processes should be referred to as “mononu- equine erythrocytes resulted in the “spot” test.17 cleosis-like illnesses” (MLI). Of the adolescents and adults who develop clinical IM, up to 85% have detectable HetAb.18 The antibodies develop Clinical Presentation within the first 7 days after the onset of symptoms, peak In IM, the subacute onset of phar- between 2 and 5 weeks into ill- yngitis is accompanied by moder- ness, and can be detected at low ate-to-high fevers (Ն37.5°C) and CLINICAL SIGNIFICANCE levels up to 12 months later. The generalized lymphadenopathy.8 heterophile test may be falsely Up to 25% of patients have pete- ● The diagnosis of “infectious mononucle- negative in up to 25% of patients chiae of the palate at least tran- osis” describes the syndrome of fever, in the first week of symptoms, siently, and the majority have pharyngitis, and lymphadenopathy, and when antibody levels are below pharyngeal erythema noted on ex- is specific for illness caused by Epstein- the limit of detection of the as- 8,9 say.19 Although heterophile test- amination. An evaluation of 70 Barr virus (EBV). different clinical signs and symp- ing in the pediatric population toms of IM showed that only 4 ● Of patients with EBV infection, 10% will may miss 50% to 75% of acute occurred statistically more often be persistently heterophile-negative. EBV infections, it remains an ex- in patients with a positive hetero- cellent test for adolescents and ● phile antibody test: petechiae of Cytomegalovirus and human herpesvirus adults, with the capability to de- the palate, and adenopathy in the 6 are the most common non-EBV causes tect between 71% and 90% of 20 inguinal, axillary, and posterior of mononucleosis-like illness. cases. Nearly 1 in 10 adults with 9 true IM will be persistently hetero- auricular lymph node groups. ● Acute HIV infection is important to con- phile-negative, but can be diag- Among patients over age 40 years sider also because conventional diag- presenting with IM, cervical lymph- nosed by detection of IgM anti- nostic methods will be negative until adenopathy is observed with a bodies against the viral capsid 21 much lower incidence, whereas detectable antibodies develop. antigen (VCA) of EBV. Many hepatomegaly and jaundice are of these patients are at the ex- more common.10 tremes of age. Lymphadenopathy in IM is typically symmetric, moder- Because of the excellent speci- ately tender, and tends to peak during the first week of ficity of current heterophile tests for IM, a positive result is symptoms. Mild-to-moderate tonsillar enlargement is com- generally considered definitive for the diagnosis of acute mon, frequently with grayish exudates. In general, urticarial EBV infection. However, reports of EBV-negative, hetero- and maculopapular rashes are rare except among those pa- phile-positive patients presenting with symptomatic, acute tients given beta-lactam antibacterials erroneously, 90% of infection from human immunodeficiency virus, type 1 22 whom go on to develop a rash.11 (HIV-1) are important to bear in mind. A palpably enlarged spleen may be present in as many as 63% of patients.12 In a study of 29 patients hospitalized on HETEROPHILE-NEGATIVE MONONUCLEOSIS-LIKE an otolaryngological service for severe IM, all were found ILLNESSES to have splenomegaly ultrasonographically, but only 17% 13 Heterophile-negative conditions with a clinical presentation had a palpable spleen on physical examination. Sponta- similar to IM (Table 1) can be grouped into 3 principal neous atraumatic splenic rupture is an exceedingly rare 14 categories: non-EBV viral etiologies, bacterial infections, complication of IM. and protozoal causes. Although some literature discusses systemic disorders such as sarcoidosis and malignancies Diagnosis of IM: The Heterophile like Hodgkin’s disease as causes of MLI (Table 2), their Antibody Test inclusion is based mostly on the presence of a particular The Paul-Bunnell heterophile antibody (HetAb) is actually a finding, such as atypical lymphocytosis or adenopathy, heterogeneous group of mostly IgM-class immunoglobulins rather than the classic triad of IM’s physical findings—and generated in response to acute EBV infection. Immunologic they thus fall outside the scope of this review. studies suggest that the Paul-Bunnell antigen is actually a complex glycoprotein structure on the surface of EBV- Viral Causes infected cells.15 Structurally similar epitopes on nonhuman Cytomegalovirus. Cytomegalovirus (CMV) causes an es- erythrocytes cross-react with HetAb, forming the basis of timated 7% of MLI cases.23 A herpesvirus relative of EBV, the red cell agglutination test. Absorbing other nonhetero- CMV establishes latent infection in a substantial portion of phile antibodies from patient serum with guinea pig kidney the general population and may reactivate with immune cells improves the specificity of these assays,2 with even compromise.24 Adolescents and adults in close contact with greater gains seen when horse erythrocytes are used instead children under age 2 years, including daycare workers and Hurt and Tammaro Mononucleosis-Like Illnesses 911.e3

Table 1 Characteristics of Infectious Mononucleosis and Mononucleosis-Like Illnesses Estimated Proportion of MLI Diagnostic Test(s) Agent Associated Condition(s) Presentations* Distinguishing Features for Acute Infection Epstein-Barr Virus Infectious 50%-90% Tender inguinal, axillary, or posterior Heterophile (“spot”) (EBV) mononucleosis auricular LAD test Petechiae of palate EBV anti-VCA IgM, Tonsillar enlargement IgG Splenomegaly Adolescents and adults up to age 30 Higher socioeconomic status in childhood Human Herpesvirus Roseola infantum 9% Bilateral, nontender, anterior and Anti-HHV-6 IgM and 6 (HHV-6) (Exanthem subitum) posterior LAD lasting up to 3 IgG months HHV-6 PCR Cytomegalovirus Mononucleosis-like 5%-7% Anicteric hepatitis Anti-CMV IgM (CMV) illness Prolonged fevers Spin amplified urine Mild cervical LAD culture for CMV, Contact with children, especially with pp65 antigen younger than age 2 years detection CMV PCR Herpes Simplex Virus, Herpes labialis 6% Gingivostomatitis, tonsillar exudates Slide-based DFA Type 1 (HSV-1) Profound odynophagia Viral throat culture Group A, ␤-hemolytic Pharyngitis 3%-4% Abrupt onset of sore throat RADT Streptococcus Rheumatic fever Tonsillopharyngeal erythema Bacterial throat pyogenes (GABHS) Tender, enlarged anterior cervical LAD culture Absence of hepatomegaly or splenomegaly Winter and early spring peak incidence Toxoplasma gondii Toxoplasmosis Յ3% Small, symmetric, nontender LAD Anti-Toxoplasma IgM History of ingesting undercooked Anti-Toxoplasma IgG meat ELISA and/or Exposure to cats or cat droppings avidity assay Human Acute retroviral Յ2% Abrupt onset of symptoms, lasting up ELISA with Western Immunodeficiency syndrome (ARS) to 2 weeks blot Virus, Type 1 AIDS Painful mucocutaneous ulcerations on HIV-1 PVL (HIV-1) oral mucosa, penis, or anus Nontender axillary, cervical, and occipital LAD between 7 and 14 days Nonpruritic, macular or maculopapular exanthem generalizing from face, chest to extremities—including palms and soles Intravenous drug use, unprotected sexual intercourse, or other HIV exposure risks Adenovirus Nonspecific upper Յ1% Clinically similar to GABHS EIA respiratory symptoms Conjunctivitis may accompany Viral culture of Pharyngo-conjunctival pharyngitis conjunctivae or fever throat Pneumonia Shell vial culture of throat or nasopharyngeal secretions AIDS ϭ acquired immune deficiency syndrome; DFA ϭ direct fluorescent antibody; EIA ϭ enzyme immunoassay; ELISA ϭ enzyme-linked immunosorbent assay; LAD ϭ lymphadenopathy; PCR ϭ polymerase chain reaction; PVL ϭ plasma viral load; RADT ϭ rapid antigen detection test; VCA ϭ viral capsid antigen. *Data from: 23, 47, 52, 60. 911.e4 The American Journal of Medicine, Vol 120, No 10, October 2007

5% of healthy adults have circulating anti-HHV-6 IgM at Table 2 Diseases with Presentations Suggestive of 32 Infectious Mononucleosis any time, detection is not always diagnostic for acute infection. Comparison of acute and convalescent sera dem- Connective tissue disorders onstrating an increase in titers is compelling evidence, but Sarcoidosis unhelpful during the acute illness. Culture remains the ref- Systemic lupus erythematosus Malignancies erence standard for diagnosis, although PCR-assisted detec- Hodgkin’s disease tion of viral DNA in whole blood in the absence of detect- Non-Hodgkin lymphoma able anti-HHV-6 antibodies is both highly sensitive and Infections specific for primary infection.33 Bartonella henselae (cat-scratch disease) Corynebacterium diphtheriae (diphtheria) Enteroviruses (coxsackieviruses, ECHO viruses) Human Immunodeficiency Virus, Type 1. The acute ret- Francisella tularensis (oropharyngeal tularemia) roviral syndrome (ARS) of symptomatic early HIV-1 infec- Hepatitis A virus tion was first described as a MLI in 1985.34 Approximately Hepatitis B virus 90% of patients develop ARS within 6 months of acquiring Mycobacterium tuberculosis (tuberculous adenitis) HIV,35 and many are ill enough to seek medical attention.35 Rubella virus (German measles) Drug reactions Symptoms develop abruptly after an average incubation Carbamazepine time of 2 to 4 weeks and may include sore throat, myalgias, Minocycline arthralgias, headache, malaise, and nausea.36 Fever may be Phenytoin as high as 40°C and accompanies pharyngitis and nontender lymphadenopathy of the axillary, cervical, and occipital nodes.37 Mucocutaneous ulceration may be seen in primary schoolteachers, are at higher risk of acute CMV infection. HIV-1 infection, with well-demarcated, painful, shallow Although primary infection is usually asymptomatic, CMV ulcers of the oral mucosa, penis, or anus.38 A nonpruritic, can produce a MLI difficult to distinguish clinically from maculopapular rash is common in ARS. Developing 48 to IM. Sore throat, fatigue, and malaise are prominent in both, 72 hours after the onset of fever and lasting up to a week, although the degree of lymphadenopathy, pharyngeal ery- the exanthem erupts on the face and upper chest before thema, and splenomegaly is generally less with CMV.25 spreading to the extremities, including the palms and Nonspecific rashes also may be seen. soles.38 Unlike IM, elevated transaminases are frequent in CMV- Standard enzyme-linked immunosorbent assays (ELISAs) induced MLI, occurring in up to 92% of cases.26 Although detect the presence of HIV-specific antibodies from clinical this sometimes causes confusion with more typical forms of specimens. Serum is incubated in wells of a microtiter plate viral hepatitis, the increase in transaminase levels rarely containing immobilized HIV antigens, allowing any anti- exceeds 5-fold above normal—in sharp contrast to the in- bodies present in the serum to bind to their corresponding creases as high as 100-fold seen with classic hepatitis vi- antigens. A second, assay-specific, enzyme-conjugated ruses. Assays for anti-CMV IgM antibodies, generally pos- immunoglobulin is then added, which attaches to any plate- itive during acute infection, have been replaced as the bound patient antibodies. The enzyme’s activity is mea- diagnostic test of choice by antigen detection assays. In the sured, serving as a proxy for the amount of original anti- most useful of these, monoclonal antibodies are used to HIV antibody present in the patient’s serum. Typically, detect pp65, a component of the shell surrounding the virus’ anti-HIV antibodies do not reach a detectable level for about nucleoprotein core—either directly from clinical specimens 2 weeks after infection, so ELISAs therefore cannot be or in shell vial cultures of CMV.27 Antigenemia assays and relied upon to diagnose ARS.39 commercially available polymerase chain reaction (PCR)- Because initial, unchecked replication of HIV-1 in a new based techniques have proven their utility in diagnosing host leads to high levels of viremia, HIV antigen assays CMV disease among post-transplant, immunocompromised were used to detect acute infection before the advent of patients,28 and may have a role for immunocompetent ones widespread plasma viral load (PVL) testing.40 One antigen as well.29 in particular, a structural protein of the viral capsid named p24, proved particularly useful. However, with inferior sen- Human Herpesvirus 6. Lesser known than EBV or CMV, sitivity to PVL and false-negative results in almost 25% of human herpesvirus 6 (HHV-6) causes a generally mild but patients with ARS,41 p24 antigen testing has fallen out of prolonged febrile MLI among adults, characterized primar- favor. Although not yet licensed by the Food and Drug ily by nontender cervical lymphadenopathy.30 HHV-6 is Administration (FDA) for the diagnosis of ARS, reverse responsible for a classic childhood exanthem: roseola infan- transcriptase polymerase chain reaction (RT-PCR) PVL tum (also called exanthem subitum or “sixth disease”). testing appears to be highly sensitive and specific for this Similar to its herpesvirus cousins, HHV-6 usually produces purpose.41 False-positive RT-PCR results have been re- latent infection early in life, with the highest seroconversion ported at a rate of about 2% to 3%,41 and are suggested in rates between 6 and 8 months of age.31 IgM titers increase those patients with less than 2000 copies of HIV-1 RNA per to detectable levels within days after infection, but because cubic centimeter of blood (copies/cc). If ARS is strongly Hurt and Tammaro Mononucleosis-Like Illnesses 911.e5 suspected and the PVL result is Ͻ10,000 copies/cc, the test 90% to 95% if properly collected.54 Although rapid antigen should be repeated.42 detection tests (RADTs) are not as sensitive as throat culture, their specificity for GABHS significantly increases the Adenovirus. A common cause of self-limited childhood number of patients treated appropriately with antibiotics.55 respiratory tract infections, adenovirus is often more aggres- Because of the low incidence of GABHS pharyngitis among sive among adults. Spread by aerosols or fecal-oral trans- adults, current recommendations suggest that a confirma- mission, the virus is hearty and can survive for long periods tory throat culture is not necessary if the RADT is outside of the host. Pharyngitis and coryza are common negative.56 presentations of infection, often accompanied by fever and 43 cervical lymphadenopathy. When conjunctivitis is present Protozoal Causes as well, the findings mark one of the classic syndromes of Toxoplasma gondii. Toxoplasmosis is the main protozoal adenoviral infection, pharyngoconjunctival fever—large cause of MLI. The life cycle of Toxoplasma gondii can only outbreaks of which have been associated with public swim- be completed through sexual replication in the feline intes- ming pools. Adults may develop tracheobronchitis or a mild tinal tract; the host cat sheds oocysts in its feces.57 Shortly atypical pneumonia, although manifestations are often more 44 after ingestion by other animals, oocysts transform into severe among immunosuppressed patients. Enzyme im- freely motile tachyzoites that invade gut epithelium and munoassay (EIA) and PCR-based rapid diagnostic methods 45 disseminate. Tachyzoites tend to localize in brain and mus- are available, but the reference standard remains isolation cle tissue, encyst, and lay dormant for the life of the host. In of the virus in culture from nasopharyngeal or oropharyn- most of the world, ingestion of undercooked meat contain- geal secretions. ing T. gondii cysts appears to be the major vector for transmission.57 Herpes Simplex Virus, Type 1. Although the “cold sore” Immunocompetent patients with primary T. gondii infec- of herpes simplex virus, type 1 (HSV-1) is thought to be its tion are often asymptomatic, but nontender cervical or major clinical manifestation, herpes labialis actually repre- occipital lymphadenopathy is sometimes seen.58 Consti- sents reactivation disease. Pharyngitis, tonsillar exudates, tutional symptoms are mild. Maculopapular rashes, pharyn- and gingivostomatitis are the most frequent manifestations 46 gitis, and hepatosplenomegaly also occur, but much less of primary herpetic infection. A study of over 600 college frequently. Toxoplasmosis is generally self-limited, resolv- students demonstrated HSV-1 to be the cause of pharyngitis 47 ing spontaneously over several months. Diagnosis of acute in almost 6% of cases. Although fever and odynophagia infection in pregnancy is particularly important, as toxo- are present for 3 to 8 days, cervical lymphadenopathy may plasmosis may cause damage to the developing fetal ner- continue for several weeks. Serologic techniques require vous system.58 comparison of acute and convalescent sera, and have a Because anti-toxoplasma IgM antibodies can persist for limited role in diagnosing acute infection. Rapid detection years after infection, their presence alone cannot be used to of HSV is possible with various ELISA and PCR-based 48 diagnose primary infection. The same is true for anti-toxo- methods. From studies of genital ulcerative disease, PCR plasma IgG antibodies, which appear within 2 weeks of has proven to be both faster and more sensitive than tradi- 58 49 primary infection and remain detectable for life. Acute tional viral culture. versus chronic infection may be distinguished by IgG “avid- Bacterial Causes ity” testing, based on the finding that prolonged immunologic exposure to the organism results in the production of Streptococcus pyogenes. Group A ␤-hemolytic Streptococ- anti-toxoplasma IgG antibodies with progressively stronger cus pyogenes (GABHS) is the most frequent bacterial cause binding to (or avidity for) toxoplasmal antigens. Thus, in a of acute pharyngitis.50 Most cases of “strep throat” occur patient with a positive IgM, weaker binding of IgG in an in the winter or early spring months in temperate cli- 59 avidity assay is suggestive of more recent infection. mates. Among all adults presenting with sore throat, GABHS accounts for up to 10% of cases.51 Streptococcal illness is more likely among patients who have significant APPROACH TO DIAGNOSIS contact with school-aged children, especially those be- Given the array of conditions mimicking infectious mono- tween 5 and 15 years of age. In 2 large studies of patients nucleosis (Table 1), a systematic approach to heterophile- evaluated for MLI, rates of GABHS-associated pharyn- negative mononucleosis-like illness is essential. Before em- gitis were Ͻ5%.52,53 barking on any laboratory assessment, a comprehensive Streptococcal pharyngitis presents with the abrupt onset history should be obtained from the patient, including past of fever and intense odynophagia. Physical examination medical problems, family history, contact with pets or with generally reveals hyperemia of the pharynx, with or without any sick persons, sexual history, and any recent travel. exudates. Erythema and edema of the uvula and soft palate Although physical examination may reveal only nonspecific may be seen, occasionally with petechiae. Anterior cervical findings, the discovery of characteristic features of some lymph nodes may become enlarged and tender. Throat cul- diseases—such as mucocutaneous ulceration in acute HIV-1 ture remains the diagnostic standard, with a sensitivity of infection—can prove invaluable. 911.e6 The American Journal of Medicine, Vol 120, No 10, October 2007

Figure Diagnostic algorithm for guidance in evaluation of MLI. CMV ϭ cytomegalovirus; EBV ϭ Epstein-Barr virus; HHV-6 ϭ human herpesvirus 6; IM ϭ infectious mononucleosis; LAD ϭ lymphadenopathy; MLI ϭ mononucleosis-like illness; VCA ϭ viral capsid antigen; WBC ϭ white blood cell. *Consider possibility of false-positive heterophile test due to HIV-1 before ﬁnalizing diagnosis. Adapted from Tsaparas YF et al, with permission from Archives of Pathology & Laboratory Medicine. Copyright 2006. College of American Pathologists.

An algorithm to guide the laboratory diagnosis of IM and SUMMARY heterophile-negative MLI is presented in the Figure, When a patient presenting with pharyngitis, lymphadenop- adapted from one published previously.60 Initial screening athy, and fever has negative results on both HetAb and for a clinical picture consistent with IM should include EBV-specific serologic tests, the clinician is faced with a heterophile antibody testing. If positive, this is highly sug- diagnostic challenge. Consideration must be given to the gestive of EBV-induced IM, but does not rule out the many potential causes of heterophile-negative mononucle- possibility of other infections, including HIV-1.22 If nega- osis-like illness, with confirmatory testing driven by a care- tive, a complete blood count (CBC) with automated differ- ful appraisal of the patient’s clinical course, history of ential may be helpful. Marked lymphocytosis (over 50% exposures and risks factors, and physical examination. of all leukocytes) with atypical cells comprising at least 10% of all leukocytes constitutes Hoagland’s criteria for 8 atypical lymphocytosis, suggesting heterophile-negative ACKNOWLEDGMENTS EBV-induced IM. Specific serologies for antibodies against We thank Penelope Dennehy, MD, Staci Fischer, MD, and EBV’s capsid (VCA) should be sent for confirmation. If the Edward Wing, MD for their thoughtful review of the anti-VCA IgM and IgG assays are negative, request sero- manuscript. logic testing for the 2 other main viral etiologies of MLI: CMV and HHV-6. Negative results should prompt a reas- References sessment of the patient’s symptoms and history, with 1. Sprunt TP, Evans FA. Mononucleosis leukocytosis in reaction to acute thought given to other less common diagnoses and appro- infections (infectious mononucleosis). Johns Hopkins Hosp Bull. priate testing. 1920;31:410. Hurt and Tammaro Mononucleosis-Like Illnesses 911.e7

2. Davidsohn I. Serologic diagnosis of infectious mononucleosis. JAMA. 27. Shuster EA, Beneke JS, Tegtmeier GE, et al. Monoclonal antibody for 1937;108:289-295. rapid laboratory detection of cytomegalovirus infections: characteriza- 3. Henle G, Henle W, Diehl V. Relation of Burkitt’s tumor associated tion and diagnostic application. Mayo Clin Proc. 1985;60:577-585. herpestype virus to infectious mononucleosis. Proc Natl Acad SciUSA. 28. Tendero DT. Laboratory diagnosis of cytomegalovirus (CMV) infec- 1968;59:94-101. tions in immunodepressed patients, mainly in patients with AIDS. Clin 4. Hallee TJ, Evans AS, Niederman JC, Brooks CM, Voegtly H. Infec- Lab. 2001;47:169-183. tious mononucleosis at the United States Military Academy. A pro- 29. Lesprit P, Scieux C, Lemann M, Carbonelle E, Modai J, Molina JM. spective study of a single class over four years. Yale J Biol Med. Use of the cytomegalovirus (CMV) antigenemia assay for the rapid 1974;47:182-195. diagnosis of primary CMV infection in hospitalized adults. Clin Infect 5. Nye FJ. Social class and infectious mononucleosis. J Hyg (Lond). Mar Dis. 1998;26:646-650. 1973;71:145-149. 30. Niederman JC, Liu CR, Kaplan MH, Brown NA. Clinical and sero- 6. Heath CW Jr, Brodsky AL, Potolsky AI. Infectious mononucleosis in logical features of human herpesvirus-6 infection in three adults. a general population. Am J Epidemiol. 1972;95:46-52. Lancet. 1988;2:817-819. 7. Babcock GJ, Decker LL, Volk M, Thorley-Lawson DA. EBV persis- 31. Hall CB, Long CE, Schnabel KC, et al. Human herpesvirus-6 infection tence in memory B cells in vivo. Immunity. 1998;9:395-404. in children. A prospective study of complications and reactivation. 8. Hoagland RJ. The clinical manifestations of infectious mononucleosis: N Engl J Med. 1994;331:432-438. a report of two hundred cases. Am J Med Sci. 1960;240:55-63. 32. Suga S, Yoshikawa T, Asano Y, et al. IgM neutralizing antibody 9. Aronson MD, Komaroff AL, Pass TM, Ervin CT, Branch WT. Het- responses to human herpesvirus-6 in patients with exanthem subitum erophil antibody in adults with sore throat: frequency and clinical or organ transplantation. Microbiol Immunol. 1992;36:495-506. presentation. Ann Intern Med. 1982;96:505-508. 33. Chiu SS, Cheung CY, Tse CY, Peiris M. Early diagnosis of primary 10. Axelrod P, Finestone AJ. Infectious mononucleosis in older adults. Am human herpesvirus 6 infection in childhood: serology, polymerase Fam Physician. 1990;42:1599-1606. chain reaction, and virus load. J Infect Dis. 1998;178:1250-1256. 11. Pullen H, Wright N, Murdoch JM. Hypersensitivity reactions to antibac- 34. Cooper DA, Gold J, Maclean P, et al. Acute AIDS retrovirus infection. terial drugs in infectious mononucleosis. Lancet. 1967;2:1176-1178. Definition of a clinical illness associated with seroconversion. Lancet. 12. Joncas J, Chiasson JP, Turcotte J, Quennec P. Studies on infectious 1985;1:537-540. mononucleosis. 3. Clinical data, serologic and epidemiologic findings. 35. Tindall B, Barker S, Donovan B, et al. Characterization of the acute Can Med Assoc J. 1968;98:848-854. clinical illness associated with human immunodeficiency virus infec- 13. Dommerby H, Stangerup SE, Stangerup M, Hancke S. Hepatospleno- tion. Arch Intern Med. 1988;148:945-949. megaly in infectious mononucleosis, assessed by ultrasonic scanning. 36. Gaines H, von Sydow M, Pehrson PO, Lundbegh P. Clinical picture of J Laryngol Otol. 1986;100:573-579. primary HIV infection presenting as a glandular-fever-like illness. 14. Farley DR, Zietlow SP, Bannon MP, Farnell MB. Spontaneous rupture BMJ. 1988;297:1363-1368. of the spleen due to infectious mononucleosis. Mayo Clin Proc. 1992; 37. Quinn TC. Acute primary HIV infection. JAMA. 1997;278:58-62. 67:846-853. 38. Lapins J, Gaines H, Lindback S, Lidbrink P, Emtestam L. Skin and 15. Golaszewska E, Kurowska E, Duk M, Koscielak J. Paul-Bunnell mucosal characteristics of symptomatic primary HIV-1 infection. antigen and a possible mechanism of formation of heterophile anti- AIDS Patient Care STDS. 1997;11:67-70. bodies in patients with infectious mononucleosis. Acta Biochim Pol. 39. Hurtado R, Rosenberg ES. Before seroconversion: the clinical and 2003;50:1205-1211. scientific events of acute HIV infection and the immunologic benefits 16. Evans AS, Niederman JC, Cenabre LC, West B, Richards VA. A of treatment. AIDS Clin Care. 2001;13:1-3. prospective evaluation of heterophile and Epstein-Barr virus-specific 40. von Sydow M, Gaines H, Sonnerborg A, Forsgren M, Pehrson PO, IgM antibody tests in clinical and subclinical infectious mononucleo- Strannegard O. Antigen detection in primary HIV infection. Br Med J sis: specificity and sensitivity of the tests and persistence of antibody. (Clin Res Ed). 1988;296:238-240. J Infect Dis. 1975;132:546-554. 41. Daar ES, Little S, Pitt J, et al. Diagnosis of primary HIV-1 infection. 17. Lee CL, Davidsohn I, Panczyszyn O. Horse agglutinins in infectious Los Angeles County Primary HIV Infection Recruitment Network. mononucleosis. II. The spot test. Am J Clin Pathol. 1968;49:12-18. Ann Intern Med. 2001;134:25-29. 18. Nikoskelainen J, Leikola J, Klemola E. IgM antibodies specific for 42. Rich JD, Merriman NA, Mylonakis E, et al. Misdiagnosis of HIV Epstein-Barr virus in infectious mononucleosis without heterophil infection by HIV-1 plasma viral load testing: a case series. Ann Intern antibodies. Br Med J. 1974;4:72-75. Med. 1999;130:37-39. 19. Hoagland RJ. Infectious mononucleosis. Prim Care. 1975;2:295-307. 43. Melon S, Mendez S, Iglesias B, et al. Involvement of adenovirus in 20. Linderholm M, Boman J, Juto P, Linde A. Comparative evaluation of clinical mononucleosis-like syndromes in young children. Eur J Clin nine kits for rapid diagnosis of infectious mononucleosis and Epstein- Microbiol Infect Dis. 2005;24:314-318. Barr virus-specific serology. J Clin Microbiol. 1994;32:259-261. 44. Zahradnik JM. Adenovirus pneumonia. Semin Respir Infect. 1987;2: 21. Henle W, Henle GE, Horwitz CA. Epstein-Barr virus specific diag- 104-111. nostic tests in infectious mononucleosis. Hum Pathol. 1974;5:551-565. 45. Raty R, Kleemola M, Melen K, Stenvik M, Julkunen I. Efficacy of 22. Vidrih JA, Walensky RP, Sax PE, Freedberg KA. Positive Epstein- PCR and other diagnostic methods for the detection of respiratory Barr virus heterophile antibody tests in patients with primary human adenoviral infections. J Med Virol. 1999;59:66-72. immunodeficiency virus infection. Am J Med. 2001;111:192-194. 46. Glezen WP, Fernald GW, Lohr JA. Acute respiratory disease of 23. Evans AS. Infectious mononucleosis and related syndromes. Am J Med university students with special reference to the etiologic role of Sci. 1978;276:325-339. Herpesvirus hominis. Am J Epidemiol. 1975;101:111-121. 24. Crumpacker CS. Cytomegalovirus. In: Mandell GL, Bennett JE, Dolin 47. McMillan JA, Weiner LB, Higgins AM, Lamparella VJ. Pharyngitis R, eds. Mandell, Douglas, and Bennett’s Principles and Practice of associated with herpes simplex virus in college students. Pediatr Infect Infectious Diseases, 5th ed. Philadelphia, PA: Churchill Livingstone; Dis J. 1993;12:280-284. 2000:1586-1599. 48. Cleator GM, Klapper PE. Herpes simplex. In: Zuckerman AJ, Ba- 25. Jordan MC, Rousseau W, Stewart JA, Noble GR, Chin TD. Sponta- natvala JE, Pattison JR, Griffiths PD, Schoub BD, eds. Principles and neous cytomegalovirus mononucleosis. Clinical and laboratory obser- Practice of Clinical Virology, 5th ed. New York, NY: John Wiley and vations in nine cases. Ann Intern Med. 1973;79:153-160. Sons; 2004:27-54. 26. Horwitz CA, Henle W, Henle G. Diagnostic aspects of the cytomeg- 49. Bruisten SM, Cairo I, Fennema H, et al. Diagnosing genital ulcer alovirus mononucleosis syndrome in previously healthy persons. Post- disease in a clinic for sexually transmitted diseases in Amsterdam, The grad Med. 1979;66:153-158. Netherlands. J Clin Microbiol. 2001;39:601-605. 911.e8 The American Journal of Medicine, Vol 120, No 10, October 2007

50. Bisno AL. Acute pharyngitis: etiology and diagnosis. Pediatrics. 1996; 56. Bisno AL, Gerber MA, Gwaltney JM Jr, Kaplan EL, Schwartz RH. 97:949-954. Practice guidelines for the diagnosis and management of group A 51. Bisno AL. Acute pharyngitis. N Engl J Med. 2001;344:205-211. streptococcal pharyngitis. Infectious Diseases Society of America. 52. Chretien JH, Esswein JG. How frequent is bacterial superinfection of Clin Infect Dis. 2002;35:113-125. the pharynx in infectious mononucleosis? Observations on incidence, 57. Tenter AM, Heckeroth AR, Weiss LM. Toxoplasma gondii: from recognition, and management with antibiotics Clin Pediatr (Phila). animals to humans. Int J Parasitol. 2000;30:1217-1258. 1976;15:424-427. 58. Montoya JG, Liesenfeld O. Toxoplasmosis. Lancet. 2004;363:1965-1976. 53. Merriam SC, Keeling RP. Beta-hemolytic streptococcal pharyngitis: 59. Suzuki LA, Rocha RJ, Rossi CL. Evaluation of serological markers for uncommon in infectious mononucleosis. South Med J. 1983;76: the immunodiagnosis of acute acquired toxoplasmosis. J Med Micro- 575-576. biol. 2001;50:62-70. 54. Gerber MA. Comparison of throat cultures and rapid strep tests for 60. Tsaparas YF, Brigden ML, Mathias R, Thomas E, Raboud J, Doyle diagnosis of streptococcal pharyngitis. Pediatr Infect Dis J. 1989;8: PW. Proportion positive for Epstein-Barr virus, cytomegalovirus, hu- 820-824. man herpesvirus 6, Toxoplasma, and human immunodeﬁciency virus 55. Lieu TA, Fleisher GR, Schwartz JS. Clinical evaluation of a latex types 1 and 2 in heterophile-negative patients with an absolute lym- agglutination test for streptococcal pharyngitis: performance and im- phocytosis or an instrument-generated atypical lymphocyte ﬂag. Arch pact on treatment rates. Pediatr Infect Dis J. 1988;7:847-854. Pathol Lab Med. 2000;124:1324-1330. The American Journal of Medicine (2007) 120, 911.e9-911.e15

CLINICAL RESEARCH STUDY

Incidence and Clinical Spectrum of Thiazide-associated Hypercalcemia

Robert A. Wermers, MD,a Ann E. Kearns, MD,a Gregory D. Jenkins, MS,b L. Joseph Melton, III, MDa,c aDivision of Endocrinology, Department of Medicine, and bDivisions of Biostatistics and cEpidemiology, Department of Health Sciences Research, Mayo Clinic College of Medicine, Rochester, Minn.

ABSTRACT

PURPOSE: The study determines the incidence of thiazide-associated hypercalcemia and clarifies its clinical features and natural history. METHODS: In a population-based descriptive study, Olmsted County, Minn, residents with thiazide- associated hypercalcemia were identified through the Rochester Epidemiology Project and the Mayo Clinic Laboratory Information System. Changes in incidence rates were evaluated by Poisson regression. RESULTS: Seventy-two Olmsted County residents (68 women and 4 men; mean age, 64 years) with thiazide-associated hypercalcemia first recognized in 1992 to 2001 were identified. The overall annual age- and sex-adjusted (to 2000 US whites) incidence was 7.7 (95% confidence interval [CI], 5.9-9.5) per 100,000. There was an increase in incidence after 1996, peaking at 16.3 (95% CI, 8.3-24.3) per 100,000 in 1998. The highest rate was 55.3 per 100,000 in 70- to 79-year-old women. Hypercalcemia was identified a mean of 6 Ϯ 7 years after thiazide initiation, and the average highest serum calcium was 10.7 Ϯ 0.3 mg/dL with serum parathyroid hormone (obtained in 53 patients) of 4.8 Ϯ 2.7 pmol/L. Of 33 patients who discontinued the thiazide, 21 (64%) had persistent hypercalcemia. Patients subsequently diagnosed with primary hyperparathyroidism had the highest average serum calcium and parathyroid hormone levels of 11.0 Ϯ 0.3 mg/dL and 6.3 Ϯ 2.4 pmol/L, respectively. CONCLUSION: The persistence of hypercalcemia in patients discontinuing thiazides, and similarities in the clinical spectrum, suggest that underlying primary hyperparathyroidism is common in patients who develop hypercalcemia while taking thiazide diuretics. © 2007 Elsevier Inc. All rights reserved.

KEYWORDS: Epidemiology; Hypercalcemia; Hyperparathyroidism; Incidence; Thiazide diuretic; Trends

Hypercalcemia associated with thiazide diuretic use is a zides have several metabolic effects that may contribute to well-recognized clinical entity.1 Mean 24-hour plasma cal- increased calcium levels. A reduction in urine calcium ex- cium concentrations are increased with thiazide use, but cretion is the most likely explanation,3-6 but a metabolic mean 24-hour plasma parathyroid hormone levels remain alkalosis associated with diuretic use also could cause an unchanged in subjects with normal baseline parathyroid elevation in total serum calcium through a pH-dependent hormone levels and no evidence of hypercalciuria.2 Thia- increase in protein-bound calcium. Although plasma 7 1,25(OH)2D levels are unchanged, increased intestinal calcium absorption in response to thiazide treatment has been This project was supported in part by an ENHANCE award from the Department of Internal Medicine, Mayo Clinic Rochester, and by a grant noted and could also contribute to an increase in serum 8,9 (RO1 AR 30582) from the National Institutes of Health, U.S. Public Health calcium. A ﬁnal factor that may lead to the development Service. of hypercalcemia is hemoconcentration associated with Dr. Wermers served as a clinical investigator for Eli Lilly & Company. diuresis.10 Requests for reprints should be addressed to Robert A. Wermers, MD, Despite the well-known effect of thiazides increasing Division of Endocrinology, Mayo Clinic, 200 First Street SW, Rochester, MN 55905. serum calcium levels, the incidence of thiazide-associated E-mail address: [email protected] hypercalcemia has never been reported. Furthermore, prior

0002-9343/$ -see front matter © 2007 Elsevier Inc. All rights reserved. doi:10.1016/j.amjmed.2006.07.044 911.e10 The American Journal of Medicine, Vol 120, No 10, October 2007 prevalence studies were performed before the introduction of between 1992 and 2001, were identified directly from automated serum calcium measurements. The only available Mayo’s Laboratory Information System. data are from the early 1970s in Sweden where, in a health Patients were accepted as having thiazide-associated hy- screen, 1034 of 15,903 persons between 20 and 63 years of age percalcemia if they met the following criteria: sustained (66% women) were taking thiazides; among the thiazide- hypercalcemia (serum calcium Ͼ 10.1 mg/dL) documented treated subjects, 20 (1.9%) were on 2 or more measurements with found to have hypercalcemia.11 The concomitant thiazide diuretic use prevalence of hypercalcemia in this CLINICAL SIGNIFICANCE for which no other cause (eg, group was greater than that in the pathologically or biochemically entire population (0.6%). Stenstrom ● The annual incidence of thiazide-associ- proven primary hyperparathyroid- 12 and Heedman estimated the over- ated hypercalcemia was 7.7 per 100,000, ism before initiation of the thia- all prevalence of hypercalcemia to with an increasing trend after 1996. zide diuretic, malignancy, family be 0.1% to 0.2%, with a prevalence history of familial benign hyper- of 0.4% in thiazide-treated subjects. ● Thiazide-associated hypercalcemia is of- calcemia, creatinine level Ͼ 2 mg/ The primary aim of this study ten discovered several years after thia- dL, or lithium therapy) was iden- was to determine the incidence of zide initiation. tified. The methods used at the thiazide-associated hypercalcemia Mayo Clinic to measure serum in the era of routine serum cal- ● The persistence of hypercalcemia follow- calcium levels changed over time; cium measurement. In addition, ing discontinuation of thiazides sug- however, the normal range (8.9- we sought to clarify the clinical gests that underlying primary hyperpara- 10.1 mg/dL) remained the same characteristics and natural history thyroidism is common in patients who because the instrumentation was of thiazide-associated hypercalce- develop hypercalcemia while taking thi- calibrated not against the manu- mia in the general population. azide diuretics. facturer’s standard but against Given the demonstrated benefits atomic absorption spectrophotom- of thiazide diuretics13 and their etry (according to certified refer- status as preferred first-step agents ences from the National Bureau of in the treatment of hypertension,14 we also wanted to pro- Standards). The date of diagnosis of thiazide-induced hy- vide timely clinical guidance in the evaluation and manage- percalcemia was the date of the first elevated calcium level ment of this disorder. while taking a thiazide diuretic. Patients with thiazide-associated hypercalcemia who subsequently had biochemi- METHODS cally proven primary hyperparathyroidism must have met the inclusion criteria outlined in our previous studies16,17 3 Population-based research is feasible in Olmsted County, months or more after thiazide discontinuation.1 Serum intact Minn, because medical care is essentially self-contained parathyroid hormone was measured by a 2-site immuno- within this community of approximately 124,000 residents. chemiluminometric assay (normal range: 1.0-5.2 pmol/L). The area has relatively few providers, and 90% of the To qualify for the study, patients must have been residents population is white. Most endocrinologic care is provided of Olmsted County on the date the first elevation in serum by the Mayo Clinic, which has maintained a common med- calcium was discovered. In addition, to ensure comparabil- ical record with its 2 hospitals (St Marys and Rochester ity with the earlier study of primary hyperparathyroid- Methodist) for more than 100 years. The diagnoses and ism,16,17 we also established residency in Olmsted County surgical procedures recorded in these records are indexed, for at least 1 year before the diagnosis of thiazide-associated as are the medical records of other providers who serve the local population.15 By using this unique record system (the hypercalcemia. Rochester Epidemiology Project), we identified all Olmsted For each case identified, the complete (inpatient and County residents who were diagnosed with hypercalcemia outpatient) medical record in the community was reviewed between 1992 and 2001. We also identified all patients with by 1 of the investigators (R.A.W.). Mayo Clinic records primary hyperparathyroidism as described elsewhere.16 The contain the details of every inpatient hospitalization at its 2 following diagnostic rubrics were searched through 2003 (to hospitals, every outpatient office or clinic visit, all emer- allow patients in the process of being examined to be in- gency department and nursing home care, and all labora- cluded in the diagnostic index): hyperparathyroidism (Inter- tory, radiographic, and pathology reports, including autop- national Classification of Diseases, Ninth Revision [ICD-9] sies. This information was supplemented by that available code, 252.0), parathyroid adenoma (ICD-9 code, 227.1), from the other providers of care to local residents, most osteitis fibrosa cystica (ICD 9-code, 588.8), malignant hy- notably the Olmsted Medical Center.15 percalcemia/ectopic hormone secretion (ICD-9 code, Incidence rates were calculated as of the date of the 259.3), and hypercalcemia not otherwise specified (ICD-9 initial elevated serum calcium level after initiation of the code, 275.4). In addition, all Olmsted County residents with thiazide. The denominator age- and sex-specific person- serum calcium levels exceeding 10.1 mg/dL at least twice, years for the entire population of Olmsted County were Worthley et al Thiazide-associated Hypercalcemia 911.e11

Table 1 Incidence of Thiazide-associated Hypercalcemia Among Olmsted County, Minn, Residents, 1992 to 2001, by Gender and Age Group

Women Men Both Sexes

Age Group (y) n Rate* (95% CI) n Rate* (95% CI) n Rate* (95% CI) Ͻ3000 00 00 30-39 1 1.0 0 0 1 0.5 40-49 5 5.5 1 1.2 6 3.4 50-59 21 35.4 1 1.8 22 19.0 60-69 18 45.5 1 2.8 19 25.1 70-79 17 55.3 1 4.4 18 33.8 80-89 5 25.0 0 0 5 17.0 Ն90 1 18.1 0 0 1 14.6 Total 13.5 (10.3-16.8)† 0.9 (0.0-1.8)† 7.7 (5.9-9.5)‡ CI ϭ conﬁdence interval. *Incidence per 100,000 person-years. †Incidence per 100,000 person-years directly age-adjusted to the US white population in 2000. ‡Incidence per 100,000 person-years directly age- and sex-adjusted to the US white population in 2000.

estimated from decennial census data with interpolation initiation of thiazide treatment (range, 14 days to 27 between census years.18 Standard errors and 95% confi- years). Serum parathyroid hormone was measured in 53 dence intervals were calculated for the rates, assuming that patients after the identification of hypercalcemia and av- they follow a Poisson distribution. Incidence rates were eraged 4.8 Ϯ 2.7 pmol/L (range, 0.4-13 pmol/L). The directly age- and/or age- and sex-adjusted to the population serum parathyroid hormone was less than 2.1 pmol/L in 7 structure of white persons in the United States in 2000. subjects (13.2%), 2.2 to 5.2 pmol/L in 29 subjects (54.7%), Poisson regression was used to compare male and female and more than 5.2 pmol/L in 17 subjects (32.1%). None of incidence rates adjusted for age. The cumulative probability the patients with a suppressed parathyroid hormone level of primary hyperparathyroidism diagnosis after thiazide dis- had an identifiable secondary cause of their hypercalcemia, continuation was calculated using the Kaplan-Meier 19 and the onset of hypercalcemia in all of these patients was method. more than 1 year after thiazide initiation. Patients without a serum parathyroid hormone measurement had less severe RESULTS hypercalcemia (10.46 Ϯ 0.19) but were otherwise similar to We identified 72 Olmsted County residents with thiazide- the overall cohort. The most common reason for thiazide associated hypercalcemia (68 women and 4 men) during the initiation was hypertension (94%); edema (3%) and hyper- 10-year study period, 1992 to 2001. The overall annual age- calciuria/nephrolithiasis (3%) were the other rationales for and sex-adjusted incidence was 7.7 (95% confidence inter- its use. val [CI], 5.9-9.5) per 100,000 (Table 1). The incidence of thiazide-associated hypercalcemia increased after 1996, with a peak incidence of 16.3 (95% CI, 8.3-24.3) per Table 2 Trends in the Incidence of Thiazide-associated 100,000 in 1998 (Table 2). The age-adjusted incidence was Hypercalcemia Among Olmsted County, Minn, Residents, higher in women (13.5 per 100,000 person-years; 95% CI, 1992 to 2001, by Year 10.3-16.8) than in men (0.9 per 100,000 person-years; 95% CI, 0.0-1.8; P Ͻ.0001). The highest incidence was 55.3 per Year n Rate* (95% CI) 100,000 person-years in 70- to 79-year-old women. 1992 4 4.9 (0.1-9.8) The mean age at diagnosis was 64 Ϯ 11 years, and 1993 3 3.6 (0-7.8) women comprised the majority of cases (94%) (Table 3). 1994 6 7.1 (1.4-12.8) Laboratory results indicated mild hypercalcemia, with a 1995 4 4.6 (0.1-9.0) Ϯ 1996 3 3.0 (0.0-6.5) mean highest serum calcium on thiazides of 10.7 0.3 1997 8 8.6 (2.6-14.6) mg/dL (range, 10.2-11.5 mg/dL). Before initiation of 1998 16 16.3 (8.3-24.3) thiazides, the mean serum calcium was 9.7 Ϯ 0.4 mg/dL 1999 12 12.3 (5.3-19.2) as measured in 57 subjects. Of those without a serum 2000 6 5.8 (1.1-10.5) calcium level before thiazide initiation, 10 patients (67%) 2001 10 9.5 (3.5-15.4) had a normal serum calcium measurement before the CI ϭ confidence interval. detection of hypercalcemia. The elevated serum calcium *Incidence per 100,000 person-years directly age- and sex-adjusted to the US white population in 2000. level was identified on average 6.0 Ϯ 7.2 years after the 911.e12 The American Journal of Medicine, Vol 120, No 10, October 2007

Table 3 Clinical and Laboratory Spectrum of Thiazide-associated Hypercalcemia Among Olmsted County, Minn, Residents, 1992 to 2001, Overall and Subset Later Found to Have Primary Hyperparathyroidism

All Primary Patients Hyperparathyroidism Mean Ϯ SD, Subset Mean Ϯ SD, Characteristic or n (%) or n (%) Female gender 68 (94.4) 19 (95.0) Age at onset of hypercalcemia, y 63.9 Ϯ 11.3 66.2 Ϯ 11.2 Serum calcium before thiazide use, mg/dL 9.7 Ϯ 0.4 9.7 Ϯ 0.5 Maximum serum calcium on thiazides, mg/dL 10.7 Ϯ 0.3 11.0 Ϯ 0.4 Serum parathyroid hormone, pmol/L 4.8 Ϯ 2.7 6.3 Ϯ 4.4 Years from thiazide start to hypercalcemia 6.0 Ϯ 7.2 7.3 Ϯ 8.5 Reason for thiazide use Hypertension 68 (94.4) 20 (100) Edema 2 (2.8) 0 (0) Hypercalciuria/nephrolithiasis 2 (2.8) 0 (0) SD ϭ standard deviation.

The thiazide was abruptly discontinued in 33 patients (in Thirty-seven subjects (51%) continued the thiazide and 12.5% [95% CI, 4.5-19.8] within 1 year, 25.1% [95% CI, did not have parathyroid surgery; 2 subjects had parathyroid 14.3-34.4] within 3 years, and 34.6% [95% CI, 22.2-45.1] surgery without discontinuing the thiazide. Serum calcium within 5 years after initiation of treatment) and was then levels were measured in all patients after meeting the inclu- reinitiated in 2 patients (Figure 1). Serum calcium stayed sion criteria for thiazide-associated hypercalcemia. Their persistently normal in 7 patients discontinuing thiazides, last measured mean serum calcium level was 9.9 Ϯ 0.4 with a mean last serum calcium measurement 5 Ϯ 4.0 years mg/dL (range, 8.8-11.1 mg/dL) an average of 5.7 Ϯ 2.1 after the thiazide was discontinued. Twenty-one patients years (range, 2.6-11.1 years) after detection of hypercal- (64%) continued to have hypercalcemia despite discontinu- cemia, with continued hypercalcemia on the last mea- ing the thiazide, 18 of whom were formally diagnosed with surement in 11 subjects (30%). In the 26 patients with a primary hyperparathyroidism (39.7 % [95% CI, 19.2- last measured serum calcium level in the normal range, 5 55.2] within 1 year, 48.3% [95% CI, 25.8-68.4] within 3 (19%) had a single measurement, 7 (27%) had persis- years, and 65.5% [95% CI, 32.5-83.7] within 5 years tently normal calcium levels, and 14 (54%) had intermit- since discontinuation) (Figure 2). In patients with content hypercalcemia. When compared with those in whom tinued hypercalcemia after discontinuing the thiazide, 5 the thiazide was discontinued, patients remaining on thia- patients had initial normalization of their serum calcium zides had lower mean serum calcium and parathyroid level, followed by recurrent hypercalcemia an average of 3.1 Ϯ 1.2 years later. Five patients who discontinued thia- hormone levels (Table 4). zides did not have their serum calcium level measured Primary hyperparathyroidism was diagnosed in 20 pa- after discontinuation. tients with thiazide-associated hypercalcemia. Of the patients with primary hyperparathyroidism, 10 (50%) had pathologic confirmation, 7 (35%) had an inappropriate parathyroid hormone in the setting of hypercalcemia, and 3 (15%) had persistent hypercalcemia for 1 year or more after discontinuing the thiazide. In patients diagnosed with primary hyperparathyroidism, 18 (90%) were asymptomatic. The 2 patients with symptomatic disease had kidney stones. The mean age at onset of hypercalcemia in the patients diagnosed with primary hyperparathyroidism was 66 years, and 19 (95%) were women (Table 3). The mean maximum serum calcium in this subset of patients was 11.0 Ϯ 0.3 mg/dL, with a mean parathyroid hormone level of 6.3 Ϯ 2.4 pmol/L (range, 3.4-12.0 pmol/L) at the time closest to the diagnosis of primary hyperparathyroidism. The mean time to identification of hypercalcemia after initiation of the Figure 1 Management and diagnostic profile of thiazide-associated thiazide in patients diagnosed with primary hyperparathy- hypercalcemia in Olmsted County, Minn, residents, 1992-2001. roidism was 7.3 years. Worthley et al Thiazide-associated Hypercalcemia 911.e13

is often necessary to discontinue the thiazide for up to 3 months,1 because hypercalcemia has been reported to resolve once thiazides are discontinued. However, Chris- tensson and colleagues11 reported that hypercalcemia resolved in only 5 of 20 subjects 1 to 3 months after thiazide discontinuation and that 14 of the 15 subjects with persistent hypercalcemia had pathologically confirmed primary hyperparathyroidism. This is consistent with our finding that 64% of community patients identified with thiazide-associated hypercalcemia remained hypercalce- mic after treatment was discontinued, suggesting another underlying cause, for example, primary hyperparathyroidism. Figure 2 Diagnosis of primary hyperparathyroidism (solid line) with 95% confidence intervals (dashed lines) after discontinuation Indeed, the clinical characteristics of thiazide-associated of the thiazide diuretic among Olmsted County, Minn, residents hypercalcemia are similar to those seen for primary hyper- 16,17 diagnosed with thiazide-associated hypercalcemia in 1992-2001. parathyroidism in this population. The average age was 64 years and 94% were females in the thiazide cohort, and the average age was 56 years and 69% were women in the DISCUSSION primary hyperparathyroidism cohort from the same peri- In this study, the first to estimate the incidence of thiazide- od.17 The observed gender discrepancy suggests that associated hypercalcemia, we found an overall age- and postmenopausal women are predisposed to developing sex-adjusted incidence of 7.7 per 100,000 person-years in hypercalcemia while taking thiazides, but an additional our community. In general, the incidence increased after factor may be that more women than men are taking 1996, with a peak of 16.3 per 100,000 in 1998. One poten- thiazides for the treatment of hypertension.21 The mean tial explanation for this trend could be the increased use of maximum serum calcium level (10.7 mg/dL) in our subjects thiazides since the ALLHAT trial.20 However, because of was identical to that seen in primary hyperparathyroidism.17 significant changes in our prescription ordering system dur- Furthermore, in subjects who remained on thiazides, the last ing the period of interest, we were unable to confidently serum calcium measurement was 9.9 mg/dL, suggesting a quantify thiazide prescription trends at our institution. nonprogressive process as is often seen in mild primary Factors other than thiazides may be important in the hyperparathyroidism.22 Consistent with previous literature, development of hypercalcemia. Patients with a clear complications were limited to 2 patients, both of whom cause for hypercalcemia were excluded from our results. were later diagnosed with primary hyperparathyroidism.16 Thus, the primary diagnostic considerations in this set- Although patients diagnosed with primary hyperparathy- ting would be primary hyperparathyroidism or the thia- roidism appeared to be similar to the overall group of zide itself. To distinguish between these 2 conditions, it patients with thiazide-associated hypercalcemia, on av-

Table 4 Clinical and Laboratory Spectrum of Thiazide-Associated Hypercalcemia Among Olmsted County, Minn, Residents, 1992 to 2001, Comparing Subsets Continuing and Discontinuing Thiazide

Thiazide Continuation Without Thiazide Discontinuation Parathyroid Surgery Mean Ϯ Characteristic Mean Ϯ SD, or n (%) SD, or n (%) Female gender 30 (90.9) 36 (97.3) Age at onset of hypercalcemia, y 67.48 Ϯ 10.24 60.76 Ϯ 11.03 Serum calcium before thiazide use, mg/dL 9.64 Ϯ 0.37 9.71 Ϯ 0.37 Maximum serum calcium on thiazides, mg/dL 10.80 Ϯ 0.38 10.56 Ϯ 0.20 Serum parathyroid hormone, pmol/L 5.38 Ϯ 2.82 3.66 Ϯ 1.83 Range, pmol/L 1.8-13 0.4-7.6 Not measured 5 14 Parathyroid hormone Յ 2.1 pmol/L 2 (7.1) 5 (21.7) Parathyroid hormone 2.2-5.2 pmol/L 16 (57.1) 13 (56.5) Parathyroid hormone Ͼ 5.2 pmol/L 10 (35.7) 5 (21.7) Years from thiazide start to hypercalcemia 7.92 Ϯ 8.40 4.67 Ϯ 6.01 Reason for thiazide use Hypertension 31 35 Edema 2 0 Hypercalciuria/nephrolithiasis 0 2 SD ϭ standard deviation. 911.e14 The American Journal of Medicine, Vol 120, No 10, October 2007 erage, they had higher maximum serum calcium (11.0 vs zides do not worsen hypercalcemia in patients who de- 10.7 mg/dL) and parathyroid hormone (6.3 vs 4.8 velop mild primary hyperparathyroidism.29 However, pmol/L) levels. It is also likely that several of the subjects there are case reports of severe hypercalcemia in patients remaining on thiazides with unsuppressed parathyroid hor- with primary hyperparathyroidism who are initiated on mone levels or continued hypercalcemia had underlying thiazide diuretics.30 This study did not address the issue primary hyperparathyroidism. of starting thiazides in patients with known primary The specific relationship between thiazide use and pri- hyperparathyroidism. mary hyperparathyroidism is uncertain. Work from Pickle- man and colleagues23 suggested that thiazide use induced CONCLUSION enlargement of the parathyroid glands. Alternatively, the The overall age- and sex-adjusted annual incidence of thiazide- association may simply represent a chance association be- associated hypercalcemia in Olmsted County from 1992 to tween 2 conditions (hypertension and primary hyperpara- 2001 was estimated at 7.7 per 100,000 with an increasing trend thyroidism) that are more prevalent with increasing age. In since 1996 for unclear reasons. The typical patients were addition, the renal tubular Na-Cl cotransporter has been women with mild, uncomplicated, and nonprogressive hyper- suggested as a link between hypertension and calcium ho- calcemia that was discovered approximately 6 years after thi- 24 meostasis. Specifically, higher urinary calcium excretion azide initiation. Excluding readily identifiable causes of hyper- and higher parathyroid hormone levels have been reported calcemia, we estimate that approximately two thirds of the 24 in patients with essential hypertension. Finally, thiazides patients with thiazide-associated hypercalcemia have underly- might unmask mild or normocalcemic primary hyperpara- ing primary hyperparathyroidism. Because there is increasing 25 thyroidism. Conversely, thiazide use theoretically could use of thiazides as first-line antihypertensive agents, combined 3-6 reduce parathyroid gland stimulation through renal and with an aging population at increased risk for parathyroid intestinal8,9 mechanisms and delay the development of pri- disease, a further increase in the incidence of thiazide-associ- mary hyperparathyroidism.26 ated hypercalcemia might be expected. It will be important to Our study has several limitations attributable to its de- monitor long-term trends in this condition and to consider sign. We did not have prospective measurements of serum prospective studies to further characterize its long-term natural calcium in all patients taking thiazides, and serum calcium history. was not measured at specific time points before or after thiazide initiation. Also, we did not have routine measurements of serum albumin or a formal assessment of dietary ACKNOWLEDGMENT calcium and vitamin D intake, all of which may influence The authors thank Mrs Mary Roberts for assistance in pre- serum calcium measurements. In addition, the population of paring the article. Olmsted County is primarily white, limiting the application of study results to more ethnically diverse populations. References Nonetheless, the results would be representative of typical 1. Silverberg S, Bilezikian JP. Clinical course of primary hyperparathy- patients encountered in clinical practice. Furthermore, our roidism. In: Bilezikian JP, Marcus R, Levine MA, eds. The Parathy- roids, Basic and Clinical Concepts, 2nd ed. San Diego: Academic unique record system allowed comprehensive follow-up of Press; 2001:387-398. all community medical care. A comparable prospective 2. Rejnmark L, Vestergaard P, Heickendorff L, Andreasen F, Mosekilde study to identify thiazide-associated hypercalcemia would L. Loop diuretics alter the diurnal rhythm of endogenous parathyroid require an estimated 3600 study subjects followed for at hormone secretion. A randomized-controlled study on the effects of least 6 years. loop- and thiazide-diuretics on the diurnal rhythms of calcitropic hormones and biochemical bone markers in postmenopausal women. The appropriate clinical management of patients with thia- Eur J Clin Invest. 2001;31:764-772. zide-associated hypercalcemia is poorly delineated. After con- 3. Coe FL, Canterbury J, Reiss E. Hyperparathyroidism in idiopathic firmation of sustained hypercalcemia, a serum parathyroid hor- hypercalciuria: primary or secondary? Trans Assoc Am Physicians. mone determination is useful in narrowing the diagnostic 1971;84:152-161. 4. Brickman AS, Massry SG, Colburn JW. Changes in serum and urinary considerations. Although thiazides will often be discontinued calcium during treatment with hydrochlorothiazide: study on mecha- to determine whether they are the cause of the hypercalce- nisms. J Clin Invest. 1972;51:945-954. mia, hypercalcemia has been described as mild and nonpro- 5. Middler S, Pak CY, Murad F, Bartter FC. Thiazide diuretics and gressive, and often consists of a temporary elevation shortly calcium metabolism. Metabolism. 1973;22:139-146. after the onset of therapy.12,27 Our results are consistent 6. Jorgensen FS, Brunner S. The long-term effect of bendroflumethiazide on renal calcium and magnesium excretion and stone formation in with mild, asymptomatic, and nonprogressive disease. In patients with recurring renal stones. Scand J Urol Nephrol. 1974;8: patients without an easily identifiable cause of hypercalce- 128-131. mia who are taking thiazides and have an unsuppressed 7. Riis B, Christiansen C. Actions of thiazide on vitamin D metabolism: parathyroid hormone level, a reasonable strategy may be to a controlled therapeutic trial in normal women early in the postmeno- pause. Metabolism. 1985;34:421-424. follow guidelines for asymptomatic primary hyperpara- 28 8. Jorgensen FS, Transbol I. The effect of bendroflumethiazide on the thyroidism. Although exacerbation of hypercalcemia is intestinal absorption of calcium in normocalcaemic renal stone formers a concern if thiazides are continued, it seems that thia- and in hyperparathyroidism. Acta Med Scand. 1974;195:33-36. Worthley et al Thiazide-associated Hypercalcemia 911.e15

9. Bazzini C, Vezzoli V, Sironi C, et al. Thiazide-sensitive NaCl-cotrans- 19. Turnbull BW. The empirical distribution function with arbitrarily porter in the intestine: possible role of hydrochlorothiazide in the grouped, censored and truncated data. J Royal Stat Soc, Series B intestinal Ca2ϩ uptake. J Biol Chem. 2005;280:19902-19910. (Methodological). 1976;38:290-295. 10. Heath H III. Postural and venous stasis-induced changes in total 20. Austin PC, Mamdani MM, Tu K, Zwarenstein M. Changes in pre- calcium. Mayo Clin Proc. 2005;80:1101. scribing patterns following publication of the ALLHAT Trial. JAMA. 11. Christensson T, Hellstrom K, Wengle B. Hypercalcemia and primary 2004;291:44-45. hyperparathyroidism. Prevalence in patients receiving thiazides as 21. Weiss R, Buckley K, Clifford T. Changing patterns of initial drug detected in a health screen. Arch Intern Med. 1977;137:1138-1142. therapy for the treatment of hypertension in a Medicaid population, 12. Stenstrom G, Heedman P-A. Clinical findings in patients with hyper- 1997-2000. Clin Ther. 2002;24:1451-1462. calcemia. A final investigation based on biochemical screening. Acta 22. Silverberg SJ, Shane E, Jacobs TP, Siris E, Bilezikian JP. A 10-year Med Scand. 1974;195:473-477. prospective study of primary hyperparathyroidism with or without 13. The ALLHAT Officers and Coordinators for the ALLHAT Collabo- parathyroid surgery. N Engl J Med. 1999;341:1249-1255. rative Research Group. Major outcomes in high-risk hypertensive 23. Pickleman JR, Straus FH, Forland M, Paloyan E. Thiazide induced Metabolism. patients randomized to angiotensin-converting enzyme inhibitor or parathyroid stimulation. 1969;18:867-873. 24. McCarron DA, Pingree PA, Rubin RJ, Gaucher SM, Molitich M, calcium channel blocker vs diuretic: the Antihypertensive and Lipid- Krutzik S. Enhanced parathyroid function in essential hypertension: a Lowering Treatment to Prevent Heart Attack Trial (ALLHAT). JAMA. homeostatic response to a urinary calcium leak. Hypertension. 1980; 2002;288:2981-2997. 2:162-168. 14. Chobanian AV, Bakris GL, Black HR, et al. Seventh Report of the 25. Silverberg SJ, Bilezikian JP. “Incipient” primary hyperparathyroidism: Joint National Committee on Prevention, Detection, Evaluation, and a “forme fruste” of an old disease. J Clin Endocrinol Metab. 2003;88: Treatment of High Blood Pressure. Hypertension. 2003;42:1206-1252. 5348-5352. 15. Melton LJ III. History of the Rochester Epidemiology Project. Mayo 26. Arnold A, Brown MF, Urena P, Gaz RD, Sarfati E, Drueke TB. Clin Proc. 1996;71:266-274. Monoclonality of parathyroid tumors in chronic renal failure and in 16. Wermers RA, Khosla S, Atkinson EJ, Hodgson SF, O’Fallon WM, primary parathyroid hyperplasia. J Clin Invest. 1995;95:2047-2053. Melton LJ III. The rise and fall of primary hyperparathyroidism: a 27. Duarte CG, Winnacker JL, Becker KL, Pace A. Thiazide-induced population-based study in Rochester, Minnesota, 1965-1992. Ann In- hypercalcemia. NEngl J Med. 1971;284:828-830. tern Med. 1997;126:433-440. 28. Bilezikian JP, Potts JT Jr, Fuleihan GH, et al. Summary statement from 17. Wermers RA, Khosla S, Atkinson EJ, et al. Incidence of primary a workshop on asymptomatic primary hyperparathyroidism: a perspec- hyperparathyroidism in Rochester, Minnesota, 1993-2001: an update tive for the 21st century. J Bone Miner Res. 2002;17(Suppl 2):N2-N11. on the changing epidemiology of the disease. J Bone Miner Res. 29. Farquhar CW, Spathis GS, Barron JL, Levin GE. Failure of thiazide 2006;21:171-177. diuretics to increase plasma calcium in mild primary hyperparathy- 18. Bergstralh EJ, Offord KP, Chu CP, Beard CM, O’Fallon WM, Melton roidism. Postgrad Med J. 1990;66:714-716. LJ III. Calculating Incidence, Prevalence and Mortality Rates for 30. Strong P, Jewell S, Rinker J, Hoch D, Crapo L. Thiazide therapy and Olmsted County, Minnesota: An Update. Rochester, MN: Mayo Clin- severe hypercalcemia in a patient with hyperparathyroidism. West ic; 1992. Technical Report Series, No. 49. J Med. 1991;154:338-340. The American Journal of Medicine (2007) 120, e1

CLINICAL COMMUNICATION TO THE EDITOR

Marinol-Induced Gynecomastia: A Case Report relationship between marijuana use and gynecomastia has been documented. Marijuana has been found to be associated with To the Editor: hyperprolactinemia and gynecomastia.3 However, another study of the relationship between gynecomastia and cannabis The relationship between marijuana use and gyneco- smoking among men showed no association between the two.4 mastia remains controversial. The association was first pub- The concentration of ␦-9-tetrahydrocannabinol consumed by lished by Harmon and Aliapoulios.1 Several publications patients in previous studies was not quantified. This makes it confirming this relationship followed. Patients using mari- difficult to know the dosage at which patients using dronabinol juana were noted to have both gynecomastia and an increase (Marinol) may experience gynecomastia. Two further studies in the human chorionic gonadotropin levels. The discontin- published conflicting results concerning the effect of marijuana uation of marijuana use resulted in resolution of the gyneco- use on testosterone levels in men. One study found a dose- mastia and normalization of the human chorionic gonado- related effect of marijuana on plasma levels of testosterone. tropin levels.2 The present report describes a patient with However, they were unable to confirm the relationship be- gynecomastia induced by dronabinol (Marinol, Solvay 5 Pharmaceuticals Inc, Marietta, Ga). tween cannabis and gynecomastia. The other study found no relationship between marijuana use and testosterone levels in men.6 These studies demonstrate the need for further research CASE REPORT to elucidate the role of ␦-9-tetrahydrocannabinol in The patient presented is a 48-year-old man with a 2-month gynecomastia. history of a right retroareolar breast mass. The mass appeared When presented with a patient with a breast mass and a approximately 30 days after he began taking 5 mg of dronabi- history of dronabinol (Marinol) use, physicians should be nol (Marinol) per day for recurrent, severe nausea secondary to aware of the possible diagnosis of gynecomastia. an extensive gastrointestinal history. On physical examination, a b,d the mass was mobile and tender to palpation without concern- Rebekah C. Allen, MS, Anne Marie Wallace, MD, c,d ing lymphadenopathy. There was no hepatosplenomegaly on Melanie Royce, MD, PhD aSchool of Medicine, abdominal examination, and no testicular masses. Mammo- b gram and ultrasound of the mass revealed a small, dense focus Department of Surgery, cDepartment of Internal Medicine, Division of of parenchyma immediately retroareolar without suspicious Hematology/Oncology, microcalcifications. Fine-needle aspiration of the mass demon- dCancer Research and Treatment Center, strated cellularity consistent with gynecomastia. Serum levels University of New Mexico, Albuquerque of testosterone, prolactin, and thyroxine were within normal limits, as well as liver function test results. Pertinent to his doi:10.1016/j.amjmed.2006.02.003 medical history, the patient had a benign contralateral breast mass consistent with gynecomastia removed nearly 20 years References ago. This followed a history of testicular cancer treated with 1. Harmon J, Aliapoulios MA. Gynecomastia in marijuana users. N Engl unilateral orchiectomy. J Med. 1972;287(18):936. 2. Garnick MB. Spurious rise in human chorionic gonadotropin induced by marihuana in patients with testicular cancer. N Engl J Med. 1980; DISCUSSION 303(20):1177. This case report demonstrates an observed relationship be- 3. Olusi SO. Hyperprolactinemia in patients with suspected cannabis- induced gynaecomastia. Lancet. 1980;1(8162):255. tween gynecomastia and dronabinol (Marinol) use. This rela- 4. Cates W Jr, Pope JN. Gynecomastia and cannabis smoking. A nonas- tionship has not been mentioned in the literature; however, the sociation among US Army soldiers. Am J Surg. 1977;134(5):613-615. 5. Kolodny RC, Masters WH, Kolodner RM, Toro G. Depression of Requests for reprints should be addressed to Anne Marie Wallace, MD, plasma testosterone levels after chronic intensive marihuana use. N Engl University of New Mexico, Cancer Research and Treatment Center, De- J Med. 1974;290(16):872-874. partment of Surgery, 1 University of New Mexico/CRTC MSC 08 4630, 6. Mendelson JH, Kuehnle J, Ellingboe J, Babor TF. Plasma testosterone Albuquerque, NM 87131-0001. levels before, during, and after chronic marihuana smoking. N Engl E-mail address: [email protected] J Med. 1974;291(20):1051-1055.

CLINICAL COMMUNICATION TO THE EDITOR

Sarcoidosis Manifesting as a Periorbital Purplish Rash

To the Editor: Sarcoidosis is a multisystemic non-caseating granuloma- tous disease that occurs in people of all races and ages. Cutaneous involvement in sarcoidosis is often present at the onset of the disease and the frequency is approximately 25%.1

CASE REPORT A previously healthy 66-year-old Japanese man presented with a 6-month history of pruritic puffy erythematous eyelids. He was diagnosed with allergic blepharitis by an oph- thalmologist and subsequently a dermatologist, and treated with antihistamines and nonsteroidal anti-inflammatory drugs. Physical examination showed a periorbital purplish Figure Periorbital purplish erythema with notable swelling of the upper eyelids. Inset: Non-caseating granulomas composed of erythema with a notable bilateral swelling of his upper epithelioid cells in the dermis. eyelids. A skin biopsy specimen taken from the upper eyelid revealed that there were multiple non-caseating granulomas DISCUSSION composed of epithelioid cells in the dermis (Figure). A histological diagnosis of sarcoidosis was made. He had no We believe the initial diagnosis of allergic blepharitis is not unsuitable for a clinical diagnosis in the present erythema. other skin lesions or superficial lymphadenopathy. A com- The accompanying pruritus supported this diagnosis. An- puted tomography examination of the chest showed right- other important clinical differential diagnosis might be a sided paratracheal lymphadenopathy with small nodules in heliotrope rash. Persistent violaceous erythema with edema both lower lobes of the lung. Gallium scanning test showed of the periorbital skin strongly suggests a heliotrope rash, an abnormal uptake of the bilateral hilar lymph nodes. which is a pathognomonic cutaneous feature of dermato- These results supported the notion that he was suffering myosistis.2 However, search of the literature has failed to from sarcoidosis. Laboratory data including serum angio- disclose examples of cutaneous sarcoidosis presenting as tensin-converting enzyme and calcium were within normal periorbital erythema. The present rash was, thus, an unusual limits. Pulmonary function studies were normal. An oph- manifestation of sarcoidosis. thalmologic consultation determined that he had no evi- Cutaneous sarcoidosis has been classically classified into dence of uveitis or presence of sarcoidal nodules in the the following groups: maculopapular eruptions, plaques, orbit. His periorbital skin lesions soon improved after top- subcutaneous nodules, infiltration of old scars, and lupus ical treatment with potent corticosteroids. His pulmonary pernio. Subcutaneous sarcoid nodules tend to spontaneously and nodal lesions have remained static, and he is seen remit, but lupus pernio is more persistent. Clinical investi- regularly. gations, including chest radiography, gallium scanning test, pulmonary function tests, and routine blood analysis— particularly serum angiotensin-converting enzyme and calcium level—should be undertaken in all patients with cuta- Requests for reprints should be addressed to Mitsuhito Ota, MD, Depart- ment of Dermatology, Tomakomai City General Hospital, Honkocho 1, neous sarcoidal lesions, even in cases with minimal skin Tomakomai 053-8567, Japan. involvement. None of these findings are specific for sar- E-mail address: [email protected]. coidosis, and the most important single criterion for the

0002-9343/$ -see front matter © 2007 Elsevier Inc. All rights reserved. e4 The American Journal of Medicine, Vol 120, No 10, October 2007 diagnosis is the finding of typical granulomas histologically.1 Mitsuhito Ota However, these investigations may confirm the diagnosis of Department of Dermatology sarcoidosis and may be helpful in the assessment for sys- Tomakomai City General Hospital temic involvements and monitoring the activity of the dis- Tomakomai, Japan ease.1,3 Cutaneous sarcoidal lesions most effectively im- Daigo Nakazawa prove with topical or oral corticosteroid therapies, and Internal Medicine methotrexate has been recommended for use as a cortico- Tomakomai City General Hospital steroid-sparing agent in severe cases.3 Tomakomai, Japan It has been reported that physicians often have encoun- Daisuke Sawamura tered difficulties in making a diagnosis of sarcoidosis, so Department of Dermatology that the diagnosis of sarcoidosis was delayed from the onset Hokkaido University Graduate School of Medicine of symptoms. Conversely, sarcoidosis is more easily diag- Sapporo, Japan nosed when the skin manifestations are the first symptoms doi:10.1016/j.amjmed.2006.03.017 noticed by the patient or after physical examination.4 How- ever, this does not necessarily mean that it is easy to rec- References ognize a skin lesion as typical sarcoidosis. Cutaneous in- 1. Newman LS, Rose CS, Maier LA. Sarcoidosis. N Engl J Med. 1997; volvement can manifest itself in an unusual form, as in the 336:1224-1234. present case. This case reminds clinicians that if an eruption 2. Callen JP. Dermatomyositis. Lancet. 2000;355:53-57. looks suspicious, an early skin biopsy should be performed. 3. Mana J, Marcoval J, Graells J, et al. Cutaneous involvement in sarcoidosis. Relationship to systemic disease. Arch Dermatol. 1997;133:882-888. This step could make the correct diagnosis of sarcoidosis 4. Judson MA, Thompson BW, Rabin DL, et al. The diagnostic pathway significantly quicker. to sarcoidosis. Chest. 2003;123:406-412. The American Journal of Medicine (2007) 120, e5

LETTER

Alcohol and Gout levels among gout patients. Rather, moderate wine drinking (perhaps not beer or spirits) may be allowed to help achieve 5 To the Editor: other important health benefits, as above. Future studies specifically powered for the effect among gout patients Zhang and colleagues recommend that “subjects with gout would help clarify the issue further. Finally, as in any should avoid drinking alcoholic beverages entirely, despite the dietary/lifestyle guideline, this recommendation should salutary effects of light-to-moderate alcohol intake on other 1 carefully take into account other related benefits and risks to diseases.” In the general population, the overall health benefits maximize the overall health outcomes on a long-term as Յ of moderate drinking (1-2 drinks/day for men and 1 drink/ well as a short-term basis.5 day for women) likely outweigh the risks, as permitted by dietary guidelines from the United States Department of Ag- Hyon K. Choi, MD, DrPH riculture (www.healthierus.gov/dietaryguidelines)2 and the Rheumatology Division American Heart Association,3 and a recent food pyramid (http:// Arthritis Research Centre of Canada www.hsph.harvard.edu/nutritionsource/pyramids.html).4,5 More Department of Medicine than 60 prospective studies have consistently indicated that mod- Vancouver General Hospital University of British Columbia erate alcoholic consumption is associated with a 25% to 3,5 Vancouver, Canada 40% reduced risk for coronary heart disease (CHD). Also, Channing Laboratory prospective studies suggest a similar protective effect Boston, Mass. against other cardiovascular diseases and deaths.3 These benefits are particularly relevant to middle-aged men,3,5 the Gary Curhan, MD, ScD demographic in whom gout occurs most often. Furthermore, Channing Laboratory the benefits may be more relevant to gout patients, who Boston, Mass. often suffer cardiovascular comorbidities5 and are at an Renal Division Department of Medicine increased risk for CHD (http://www.heart.org/presenter. ϭ 6 Brigham and Women’s Hospital jhtml?identifier 3041796). Harvard Medical SchoolBoston, Mass. We note that the reported risk estimates in Zhang et al’s study were derived from a relatively small number of pa- doi:10.1016/j.amjmed.2006.09.025 tients who experienced gout attack(s) during the study (n Յ144).1 Furthermore, the observed associations are open References to potentially important recall bias given the fact that the 1. Zhang Y, Woods R, Chaisson CE, et al. Alcohol consumption as a alcohol-gout link is well-known. Nevertheless, at the rele- trigger of recurrent gout attacks. Am J Med. 2006;119:800 e13-e18. vant moderate intake level (1-2 drinks/day), no individual 2. United States Department of Agriculture Dietary guidelines for Amer- icans. Available at http://www.healthierus.gov/dietaryguidelines. Ac- alcoholic beverages suggested an increased risk of gout cessed September 7, 2007. (odds ratio [OR] 1.1 for beer, 1.2 for wine, and 0.8 for 3. Goldberg IJ, Mosca L, Piano MR, Fisher EA. AHA Science Advisory: spirits, all P-values Ͼ.05).1 Furthermore, the OR for wine, wine and your heart. Circulation. 2001;103:472-475. in particular, remained similar even at the level of 3-4 4. Harvard School of Public Health. Food pyramids. Available at http:// drinks/day (OR 0.9), although ORs for beer and spirits www.hsph.harvard.edu/nutritionsource/pyramids.html. Accessed Sep- 1 tember 7, 2007. increased (1.6 and 2.0, respectively). These data agree with 5. Choi HK. Diet, alcohol, and gout: how do we advise patients given the null association with moderate wine drinking (and pos- recent developments? Curr Rheumatol Rep. 2005;7:220-226. itive associations with beer and spirits) found in our pro- 6. American Heart Association. Gout and the risk of acute myocardial spective study of incident gout based on 730 cases in 47,150 infarction. Available at http://www.heart.org/presenter.jhtml?identifierϭ men.7 This is further supported by the null association 3041796. Accessed September 7, 2007. 8 7. Choi HK, Atkinson K, Karlson EW, Willett WC, Curhan G. Alcohol between moderate wine drinking and serum urate levels. intake and risk of incident gout in men. Lancet. 2004;363:1277-1281. Thus, the available data do not support a strict recom- 8. Choi HK, Curhan G. Beer, liquor, wine, and serum uric acid level. mendation to forbid all alcohol intake at light-to-moderate Arthritis Rheum. 2004:1023-1029.

LETTER

The Reply: maintenance of optimal body weight, also play important roles in reducing the risk of cardiovascular and many other In their letter to the Editor, Choi et al voiced their diseases.5 Furthermore, for those who are at higher risk of concern regarding our recommendation that “subjects with cardiovascular disease, chemoprevention, such as statin use, gout should avoid drinking alcoholic beverages entirely, 5 despite the salutary effects of light-to-moderate alcohol in- may also provide a significant clinical benefit. Thus, for take on other diseases.” They commented that no individual subjects with gout, various measures other than alcohol are alcoholic beverages at the relevant moderate intake levels available to reduce their risk of cardiovascular disease with- (1-2 drinks/day) in our study suggested an increased risk of out necessarily increasing their risk for painful gout attacks. gout attacks. They suggest that (among patients with gout) Compared with the totality of evidence on an association moderate wine drinking (perhaps not beer or spirits) may be between total amount of alcohol consumption and gout allowed to help achieve other important health benefits. attacks, to date, there are a paucity of data suggesting that In their letter, Choi et al did not comment on the strong risk of either incident or recurrent gout attack vary accord- dose-response relationship between the total amount of al- ing to specific alcoholic beverages. Additional studies with cohol (wine, beer, and spirits) and risk of incident and adequate statistical power are needed to test this hypothesis. recurrent gout attacks. We reported that the odds ratios for Thus, to reduce the risk of recurrent gout attacks, the best recurrent gout attacks were 1.4, 1.6, 2.7, and 3.1 for con- advice for the patients with gout is to avoid alcohol intake.6 sumption of 1 to 2, 3 to 4, 5 to 6, and 7 or more alcoholic drinks over 24 hours before a gout attack, respectively, Yuqing Zhang, DSc compared with no alcohol consumption (P for trend Ͻ .023).1 Tuhina Neogi, MD Choi et al also reported that the relative risks of incident David J. Hunter, MD, PhD gout attack were 1.32, 1.49, 1.96, and 2.53 for subjects Clinical Epidemiology Research and Training Unit consuming 10.0 to 14.9, 15.0 to 29.9, 30.0 to 49.9, and Ն50 g Department of Medicine Boston University School of Medicine per day of alcohol, respectively, compared with those who Boston, Mass did not drink alcohol (P for trend Ͻ .001) among participants in the Health Professionals Follow-up Study.2 Both doi:10.1016/j.amjmed.2006.11.021 studies demonstrated that light-to-moderate alcohol consumption increases the risk for incident and recurrent gout References attacks, and the magnitude of effect was similar. 1. Zhang YQ, Woods R, Chaisson CE, et al. Alcohol consumption and the Many patients with gout also have or are at higher risk of risk for recurrent gout attacks. Am J Med. 2006;119:e13-18. other diseases, including cardiovascular disease.3 Although 2. Choi HK, Atkinson K, Karlson EW, et al. Alcohol intake and risk of randomized clinical trial data are lacking, numerous epide- incident gout in men: a prospective study. Lancet. 2004;363:1277-1281. miologic studies have shown that regular moderate alcohol 3. Rott KT, Agudelo CA. Gout. JAMA. 2003;289:2857-2860. 4. Sesso HD. Alcohol and cardiovascular health: recent findings. Am J consumption, irrespective of beverage type, is associated Cardiovasc Drugs. 2001;1:167-172. with a lower risk of an initial or of recurrent cardiovascular 5. Available at: http://www.nhlbi.nih.gov/health/public/heart/index.htm. 4 disease. Similarly, other healthy lifestyle factors, such as 6. Terkeltaub RA. Clinical practice. Gout. New Engl J Med. 2003;23: abstinence from smoking, regular physical exercise, and 1647-1655.

LETTER

Functional Status in Chronic Obstructive Therefore, a brief clinical evaluation of functional status in Pulmonary Disease patients with COPD should take into account not only the ability to perform physically demanding activities, such as To the Editor: showering or leaving the house, but also the ability to perform cognitively demanding IADL. In a recent article, Reardon and coworkers1 reviewed the Andrea Corsonello, MD impact of exertional dyspnea on functional status and qual- Fondazione San Raffaele ity of life in patients with chronic obstructive pulmonary Cittadella della Carità disease (COPD). They also reviewed the methods to mea- Taranto, Italy sure functional status and quality of life, and proposed a Istituto Nazionale di Ricovero e Cura per Anziani brief evaluation based on a simple scoring of overall dys- Cosenza, Italy pnea together with an unstructured functional status evalu- Claudio Pedone, MD ation covering 1 or more activity of daily living, such as Cattedra di Geriatria showering, leaving the house, or a particular social activity. Università Campus BioMedico By using main components analysis, we recently re- Rome, Italy ported that elderly hospitalized patients with COPD have a distinctive clustering of basic and instrumental activities of Raffaele Antonelli Incalzi, MD Fondazione San Raffaele daily living (IADL) that are clearly different from those Cittadella della Carità observed in elderly patients with other chronic conditions, Taranto, Italy such as diabetes mellitus and congestive heart failure. The Cattedra di Geriatria hierarchy of IADL that characterizes COPD reveals 2 fac- Università Campus BioMedico tors: the expression of IADL related to outdoor mobility and Rome, Italy selected, highly demanding IADL requiring both physical doi:10.1016/j.amjmed.2006.09.023 and mental capabilities, such as managing money, taking medicines, and traveling. Both factors are associated with a References longer hospital stay, that is, a greater need of care, whereas 1. Reardon JZ, Lareau SC, ZuWallack R. Functional status and quality of only the latter is associated with a greater prevalence of life in chronic obstructive pulmonary disease. Am J Med. 2006; 2 cognitive impairment. 119(10A):S32-S37. Although difﬁculty in moving outdoors is a distinctive 2. Antonelli Incalzi R, Corsonello A, Pedone C, et al. Construct Validity and established feature of COPD-related disability,1,3 de- of Activities of Daily Living Scale. A clue to distinguish the disabling pendency in IADL requiring good cognitive function would effects of COPD and congestive heart failure. Chest. 2005;127:830-838. 3. Isohao R, Puolijoki H, Huhti E, et al. Chronic obstructive pulmonary remain concealed using the brief clinical assessment pro- disease and self-maintaining functions in the elderly: a population-based 1 posed by Reardon and coworkers. For example, depen- study. Scand J Prim Health Care. 1995;13:122-127. dency in taking medicine because of memory impairment 4. Antonelli Incalzi R, Gemma A, Marra C, et al. Verbal memory impair- may have a relevant impact on the burden of disease. In- ment in COPD: its mechanisms and clinical relevance. Chest. 1997; 112:1506-1513. deed, compliance with both drugs and oxygen therapy is 5. Fuso L, Antonelli Incalzi R, Pistelli R, et al. Predicting mortality of poor in patients with COPD with cognitive impairment, and patients hospitalized for acutely exacerbated chronic obstructive pul- poor compliance is a risk factor for acute exacerbations.4,5 monary disease. Am J Med. 1995;98:272-277.

LETTER

The Reply: with chronic obstructive pulmonary disease. Our recommendation to the clinician of including a few brief questions Our review focused on the importance of dyspnea and on physical activity limitation in addition to dyspnea eval- activity limitation in the quality of life of individuals with uation was not meant to be exhaustive. chronic obstructive pulmonary disease. In particular, we emphasized the inverse relationship between exertional dys- Richard ZuWallack, MD pnea and physical activity limitation in this disease. Cor- St Francis Hospital sonello and colleagues describe their research indicating Pulmonary Hartford, Conn that impairments in cognitive function are also potentially important in the functional status of hospitalized patients doi:10.1016/j.amjmed.2006.10.012

LETTER

Broadening the Differential Diagnosis from a small bowel cancer and non-Hodgkin lymphoma, respec- Different Perspective tively) in those with the association than in those without the association.9 To the Editor: Oscar M. Jolobe, MRCP(UK) (retired geriatrician) Although not included in the disease associations men- Manchester Medical Society Manchester, United Kingdom tioned by the author,1 celiac disease may be associated with inflammatory bowel disease (IBD) to a degree that is per- doi:10.1016/j.amjmed.2006.09.026 haps greater than can be ascribed to pure chance.2 In the event of such an association, there is the potential for co- References existing celiac disease to be overlooked or for one of the 1. Rajendra A, Perepletchikov A, Kopelman RI. Broadening the differen- presenting symptoms, such as diarrhea, to be misattributed tial diagnosis. Am J Med. 2006;119:410-412. to celiac disease or IBD. Misattribution also can occur when 2. Yang A, Chen Y, Scherl E, et al. Inflammatory bowel disease in patients with celiac disease. Inflamm Bowel Dis. 2005;11:528-532. the presenting symptom is abdominal pain, given the fact 3. Cooke T, Peeney ALP, Hawkins CF. Symptoms, signs, and diagnostic that this is a symptom common to both celiac disease and features of idiopathic steatorrhoea. Q J Med. 1953;XXII:59-77. 3,4 Crohn’s disease, and, for the same reason, in the event of 4. Glickman RM. Inflammatory bowel disease: ulcerative colitis and a presentation characterized by hematinic deficiencies.1,5-7 Crohn’s disease. In: Fauci AS, Braunwald E, Isselbacher KJ, et al., eds. Supporting evidence for the association of celiac disease Harrison’s Principles of Internal Medicine, 14th Edition, Chapter 286. New York, St Louis, San Francisco: McGraw-Hill; 1998;1633-1645. and IBD comes from a study that compared the prevalence 5. Pennazio M, Santucci R, Rondonotti E, et al. Outcome of patients with of IBD in a cohort of 455 patients with celiac disease with obscure gastrointestinal bleeding after capsule endoscopy: report of 100 the prevalence of IBD in the US population. This yielded an consecutive cases. Gastroenterology. 2004;126:643-653. age- and sex-adjusted prevalence ratio of 8.49 (95% confi- 6. Chowers Y, Sela B-A, Holland R, et al. Increased levels of homocys- dence interval, 3.53-20.42) for Crohn’s disease and 3.56 teine in patients with Crohn’s disease are related to folate levels. Am J 2 Gastroenterol. 2000;96:3498-3502. (95% confidence interval, 1.48-8.56) for ulcerative colitis. 7. Dickey W. Low serum B12 is common in celiac disease and is not due Among first-degree relatives of patients with celiac disease, to autoimmune gastritis. Eur J Gastroenterol Hepatol. 2002;14:425- there also seems to be an increased risk of IBD; the relative 427. risk of ulcerative colitis, in particular, is 5 times higher in 8. Shah A, Mayberry JF, Williams G, et al. Epidemiological survey of these patients than for the general population.8 celiac disease and inflammatory bowel disease in first degree relatives of coeliac patients. Q J Med. 1953;74(275):283-288. The association also has prognostic implications, as 9. Peters U, Askling J, Gridley G, et al. Causes of death in patients with shown in a study in which the coexistence of celiac disease celiac disease in a population-based Swedish Cohort. Arch Intern Med. and IBD conferred a higher mortality risk (attributable to 2003;163:1566-1572.

LETTER

Effective Detection of Celiac Disease Using salivary samples was determined by comparing the results Salivary Anti-transglutaminase for subjects with CD with those obtained for healthy controls and was fixed at 4 U/mL. We tested the saliva collected To the Editor: from 20 untreated patients with CD, 38 healthy control subjects, and 22 patients with clinical symptoms compatible Celiac disease (CD) was previously considered to be a with a diagnosis of CD but who finally did not meet the rare pediatric disease caused by gluten intolerance and pre- criteria for diagnosis (disease controls). Receiver operating senting most often with diarrhea, steatorrhea, and weight curve analysis of these data gives a sensitivity of 90% and loss. Recently, however, different trends in the presentation a specificity of 96.7% for the diagnosis of untreated CD. of CD have been reported. Most cases are diagnosed in These data indicate that measuring salivary anti-TG immu- 1 adults presenting no diarrhea, but rather other various clin- noglobulin-A is a valuable, noninvasive test that could be ical disorders, such as iron-deficiency anemia, osteoporosis, used for CD screening on a large scale in paucisymptomatic neurologic or psychiatric symptoms, abdominal pain, and populations. Any suspicion of CD based on this test should even constipation. In the general population, several sero- be confirmed by the gold standard test, the demonstration of logic screening studies have demonstrated a prevalence of a typical villous atrophy of the duodenal mucosa. approximately 0.5% to 1%. Detection of serum anti-human transglutaminase (TG) Annick Ocmant, Ph immunoglobulin-A antibodies is recognized as the most Françoise Mascart, MD, PhD reliable serologic screening test.2 For less-invasive screen- Clinique d’Immunobiologie ing tests, several attempts have been made to detect these Hôpital Erasme antibodies in saliva.3,4 Unfortunately, the only successful Université Libre de Bruxelles test cannot be used on a large scale in most clinical diag- Brussels, Belgium nostic laboratories because it is a fluid-phase radioimmuno- doi:10.1016/j.amjmed.2006.09.029 assay using radio-labeled TG.3 We report the results obtained with a commercial solid-phase enzyme-linked immunosorbent assay test (Celikey, Sweden Diagnostics, References Freiburg, Germany) demonstrating an excellent discrimina- 1. Rampertab SD, Pooran N, Brar P, et al. Trends in the presentation of celiac disease. Am J Med. 2006;119:314-355. tion between saliva from subjects with and without CD. 2. Reeves GE, Squance ML, Duggan AE, et al. Diagnostic accuracy of Saliva samples were collected with an Omnisal device (Sa- coeliac serological tests: a prospective study. Eur J Gastroenterol liva Diagnostic System, London, UK) and treated (20 min- Hepatol. 2006;18:493-501. utes at room temperature) with N-acetyl-cysteine to achieve 3. Bonamico M, Ferri M, Nenna R, et al. Tissue transglutaminase auto- a final concentration of 0.63 mg N-acetyl-cysteine per mil- antibody detection in human saliva: a powerful method for celiac disease screening. J Pediatr. 2004;144:632-636. liliter of saliva. The samples were then tested without fur- 4. Baldas V, Tommasini A, Santon D, et al. Testing for anti-human ther dilution with the Celikey following the manufacturer’s transglutaminase antibodies in saliva is not useful for diagnosis of celiac instructions for serum samples. The positivity threshold for disease. Clin Chem. 2004;50:216-219.

LETTER

The Reply: Columbia University College of Physicians and Surgeons New York The development of new, alternate, noninvasive methods of detecting celiac disease-associated antibodies is exciting, doi:10.1016/j.amjmed.2007.04.019 especially in view of the increasing realization that celiac disease is common, affecting approximately 1% of the pop- References 1 ulation, and may be extremely diverse in its manifestations. 1. Alaedini A, Green PH. Narrative review: celiac disease: understanding The gold standard of diagnosis, however, remains the small a complex autoimmune disorder. Ann Intern Med. 2005;142:289-298. intestinal biopsy taken at endoscopy.2 Previous studies have 2. Green PH, Rostami K, Marsh MN. Diagnosis of coeliac disease. Best shown that both fecal and salivary antibody estimation are Pract Res Clin Gastroenterol. 2005;19:389-400. neither sensitive nor specific enough to be used for celiac 3. Halblaub JM, Renno J, Kempf A, et al. Comparison of different salivary 3,4 and fecal antibodies for the diagnosis of celiac disease. Clin Lab. disease diagnosis. The currently proposed test may be an 2004;50:551-557. improvement that requires greater validation. The recent 4. Kappler M, Krauss-Etschmann S, Diehl V, et al. Detection of secretory development of rapid finger-stick blood tests that use tissue IgA antibodies against gliadin and human tissue transglutaminase in transglutaminase testing for in-office results may prove of stool to screen for coeliac disease in children: validation study. BMJ. value and be more accepted by medical practitioners.5 2006;332:213-214. 5. Raivio T, Kaukinen K, Nemes E, et al. Self transglutaminase-based Peter H. R. Green, MD rapid coeliac disease antibody detection by a lateral flow method. Celiac Disease Center at Columbia University Aliment Pharmacol Ther. 2006;24:147-154.

LETTER

Noninvasive Ventilation in Acute Heart However, therapy with NIV in AHF is probably benefi- Failure cial only in patients with systolic dysfunction. In patients with predominantly diastolic dysfunction who require a To the Editor: relatively high filling pressure, the effects of positive pressure therapy can compromise venous return, resulting fur- We read with interest the article “Diagnostic and Ther- ther in hypotension and prerenal azotemia.7 apeutic Approach to Acute Decompensated Heart Failure,” Finally, which subgroup of patients should receive NIV? wherein the authors discuss the management of acute de- NIV is likely to benefit patients with severe AHF non- compensated heart failure.1 However, we were surprised to response to conventional medical therapy, presentation with see no mention of the role of noninvasive ventilation (NIV) apHϽ7.25, or patients who are not candidates for intuba- in acute heart failure (AHF). NIV is the application of tion either because of a previous directive or as a result of mechanical ventilation without the use of an endotracheal poor prognosis due to underlying disease. airway and has been utilized for diverse forms of respiratory failure.2 The application of NIV to standard medical treat- Ritesh Agarwal, MD, DM ment of patients with acute respiratory failure not only Rajagopala Srinivas, MD prevents endotracheal intubation and its attendant compli- Department of Pulmonary Medicine cations, but also can decrease mortality in selected patients.3 Postgraduate Institute of Medical Education and Research As early as 1936, continuous positive airway pressure Chandigarh, India (CPAP) has been shown to be an effective therapy for doi:10.1016/j.amjmed.2007.02.024 AHF.4 NIV augments the inspiratory flow, increases the tidal volume, and unloads the inspiratory muscles. It improves References alveolar ventilation, re-expands flooded alveoli and pre- 1. Kapoor JR, Perazella MA. Diagnostic and therapeutic approach to acute decompensated heart failure. Am J Med. 2007;120:121-127. vents microatelectasis. By virtue of the above mechanisms, 2. Brochard L. Noninvasive ventilation for acute respiratory failure. 5 it decreases respiratory rate, and the work of breathing. The JAMA. 2002;288:932-935. effective filling and emptying of the heart is determined in 3. Ram FSF, Picot J, Lightowler J, Wedzicha JA. Non-invasive positive part by cardiac transmural pressure (PTM), the pressure pressure ventilation for treatment of respiratory failure due to exacer- difference between the inside of the heart and the intratho- bations of chronic obstructive pulmonary disease. Cochrane Database racic pressure. In patients with AHF, the amplitude of in- Syst Rev. 2004;3:CD004104. 5 4. Poulton EP, Oxon DM. Left-sided heart failure with pulmonary edema: spiratory swings is greater and results in higher PTM. Dur- its treatment with the “pulmonary plus pressure machine”. Lancet. ing systole, NIV increases the intrathoracic pressure and 1936;2:981-983. reduces venous return, thus decreasing the right and left 5. Agarwal R, Aggarwal AN, Gupta D, Jindal SK. Non-invasive ventila- ventricular preload; in diastole, NIV increases the pericar- tion in acute cardiogenic pulmonary oedema. Postgrad Med J. 2005; 5 81:637-643. dial pressure, reduces PTM, and thus decreases afterload. NIV also causes a decrease in the heart rate secondary to 6. Peter JV, Moran JL, Phillips-Hughes J, Graham P, Bersten AD. Effect lung inflation and resultant increased parasympathetic of non-invasive positive pressure ventilation (NIPPV) on mortality in 5 patients with acute cardiogenic pulmonary oedema: a meta-analysis. tone. Recent evidence suggests that the use of CPAP in Lancet. 2006;367:1155-1163. patients with AHF decreases intubation rate and im- 7. Agarwal R, Gupta D. What is the role of noninvasive ventilation in proves survival.5,6 diastolic heart failure? Intensive Care Med. 2005;31:1451.

LETTER

The Reply: vasive positive pressure ventilation. We do appreciate, however, the in-depth discussion by the authors of this important We thank Agarwal and Srinivas for their comments on therapy in patients with acute decompensated heart failure. our article, “Diagnostic and Therapeutic Approach to Acute Decompensated Heart Failure,” published in The American John R. Kapoor, MD, PhD 1 Journal of Medicine. The authors state that they were Section of Cardiology surprised to see no mention on the role of noninvasive Stanford University ventilation in acute heart failure. Unfortunately, the authors Stanford, Calif missed our brief mention of this therapy in our manuscript. In the ﬁrst paragraph of the “Therapeutic Interventions” Mark A. Perazella, MD section on page 123, we do, in fact, discuss this aspect of Section of Nephrology Yale University treatment as noted by the following sentences “Supplemen- New Haven, Conn tal oxygen, noninvasive positive pressure ventilation, and mechanical ventilation are considered for hypoxemic and doi:10.1016/j.amjmed.2007.03.019 hypercarbic patients. Positive pressure ventilation decreases pulmonary edema, because increases in intratho- Reference racic pressure decrease venous return.” Due to space limits, 1. Kapoor JR, Perazella MA. Diagnostic and therapeutic approach to acute we could not provide a more detailed description of nonin- decompensated heart failure. Am J Med. 2007;120:121-127.

LETTER

Endothelium-Independent Microvascular thelial dysfunction, endothelium-independent mechanisms Dysfunction in Cardiac Syndrome X such as impairment of smooth muscle relaxation also contribute to the pathogenesis of CSX. To the Editor: Pankaj Madan, MD I read with interest the article by Hurst and colleagues.1 Baylor College of Medicine Although there is evidence that endothelial dysfunction is Houston, Tex one of the pathologic mechanisms in the causation of car- Ritu Madan, MBBS diac syndrome X (CSX), endothelium-independent mecha- Safdarjung Hospital nisms also play a major role in the pathogenesis of cardiac Delhi, India syndrome. It seems that the reviewers have overlooked endothelial-independent microvascular dysfunction in their doi:10.1016/j.amjmed.2006.07.041 review. Many studies have demonstrated the impaired endothe- References lium-independent coronary microvascular dilatation in pa- 1. Hurst T, Olson TH, Olson LE, Appleton CP. Cardiac syndrome X and tients with CSX using different endothelium-independent endothelial dysfunction: new concepts in prognosis and treatment. Am J Med. 2006;119:560-566. coronary vasodilators (dipyridamole, papaverine, and aden- 2. Opherk D, Zebe H, Weihe E, et al. Reduced coronary dilator capacity 2 osine). Opherk et al showed that coronary blood ﬂow and ultrastructural changes of the myocardium in patients with angina increased 3.8-fold in controls but only 2-fold in patients pectoris but normal coronary arteriograms. Circulation. 1981; 63;817- with CSX in response to administration of dipyridamole. 825. Bottcher et al3 also reported similar results. Chauhan et al4 3. Bottcher M, Botker HE, Sonne H, et al. Endothelium-dependent and independent perfusion reserve and the effect of L-arginine on myocar- reported a reduced augmentation of coronary blood ﬂow in dial perfusion in patient with syndrome X. Circulation. 1999; 99:1795- a CSX group (185% Ϯ 74% vs 411% Ϯ 59%, P Ͻ .0001) 1801. after administration of papaverine. Other investigators using 4. Chauhan A, Mullins PA, Taylor G, et al. Both endothelium-dependent cardiac magnetic resonance also demonstrated that adeno- and endothelium-independent function is impaired in patients with sine administration did not increase myocardial perfusion angina pectoris and normal coronary angiograms. Eur Heart J. 1997; 18:60-68. index in the endocardium of patients with CSX, whereas it 5. Panting JR, Gatehouse PD, Yang GZ, et al. Abnormal subendocardial 5 showed an increase in healthy controls. perfusion in cardiac syndrome X detected by cardiovascular magnetic These studies clearly indicate that in addition to endo- resonance imaging. N Engl J Med. 2002;346:1948-1953.

LETTER

The Reply: function as a mechanism with clinical relevance in developing a treatment plan for these individuals. The hope is that We appreciate the interest of Madan and Madan in our future research will uncover effective therapy based on article on Cardiac Syndrome X (CSX). endothelium independent microvascular function. We could not agree more that CSX is a complex disorder with a multifactorial mechanism. We also agree that there is Christopher P. Appleton, MD evidence to support an endothelium independent component R. Todd Hurst, MD to this syndrome as they have cited and as manifest with a Mayo Clinic Graduate School of Medicine high frequency of chest pain during dipyridamole infusion Division of Cardiovascular Diseases in our clinical experience. While we did not want to deem- Mayo Clinic Arizona Scottsdale phasize the importance of other mechanisms, the focus of our article was endothelium dependent microvascular dys- doi:10.1016/j.amjmed.2007.07.016

0002-9343/$ -see front matter © 2007 Elsevier Inc. All rights reserved. APM Perspectives The Association of Professors of Medicine (APM) is the national organization of departments of internal medicine at the US medical schools and numerous afﬁliated teaching hospitals as represented by chairs and appointed leaders. As the ofﬁcial sponsor of The American Journal of Medicine, the association invites authors to publish commentaries on issues concerning academic internal medicine.

APM Perspectives For the latest information about departments of internal medicine, please visit APM’s website at www.im.org/APM. Communication Skills: A Call for Teaching to the Test Anna Headly, MD, MFA Director of Undergraduate Medical Education, Internal Medicine, UMDNJ/Robert Wood Johnson Medical School, Camden, NJ.

Educational authorities have been calling for improved “Demonstrate effective listening by hearing and under- physician-patient communication for decades, yet stu- standing in a way that the patient feels heard and dents and residents seem to feel less prepared than ever understood.”9 While these goals are admirable, they are for difficult situations with patients, and patients are not particularly useful when a student interacts with a becoming less satisfied with physician communication patient. What students typically ask when given such skills.1,2 The Association of American Medical Col- vague guidelines is, “But what should I say?” leges (AAMC) made communication central to its Effective physician-patient communication is cen- 3 Medical Schools Outcomes Project, requiring medical tral to patient care, and should therefore be a major students to pass a clinical skills examination as part of emphasis of medical education. Yet teaching commu- the United States Medical Licensing Examination nication skills has historically been relegated to a sub- (USMLE) Step 2. The clinical skills examination in- component of other curricula or limited to the basics of 4 cludes components of communication, but there is a medical interviewing.10,11 The result is that most US paucity of data on the best methods to teach such medical graduates report little training (and less com- 5,6 skills. A recent large and costly intervention using a fort) in end-of-life communication skills; most resi- general communication skills curriculum resulted in dents lack competence in delivering bad news and are 7 only a 5% improvement in student scores. Most of the unskilled in the use of interpreters; and physician-in- published approaches to the teaching of communication training scores of physician-patient communication in suffer from being vague, teacher-dependent, and non- the primary care setting declined from 1996 to reproducible. For instance, the Kalamazoo Consensus 1999.1,2,12,13 Not only is patient satisfaction affected by Statement on the essential elements of medical com- poor communication skills, but there is a significant munication in medical encounters delineates 7 tasks, correlation between effective physician-patient com- including “build the doctor-patient relationship” and munication and improved health outcomes as well as “understand the patient’s perspective.”8 More recently, reductions in malpractice claims.14,15 the Accreditation Council for Graduate Medical Edu- Educators have shied away from the idea of “teach- cation (ACGME) expanded its recommendations for ing to the test” when it comes to communication skills, communication competencies to include language such yet this reluctance leaves physicians-in-training ill- as “Be ‘present,’ paying attention to the patient” and equipped when they are, in fact, tested.7,16 This is especially true for “higher order” communication skills; Requests for reprints should be addressed to Anna Headly, MD, educators tend to approach emotionally charged discus- MFA, Director of Undergraduate Medical Education, Internal Med- sions as very different from “routine” medical inter- icine, UMDNJ/Robert Wood Johnson Medical School, 401 Haddon Avenue, Room 242, Camden, NJ 08103. viewing. For example, multiple sources give remark- E-mail address: [email protected]. ably similar lists of specific questions to ask patients

0002-9343/$ -see front matter © 2007 The Association of Professors of Medicine. All rights reserved. doi:10.1016/j.amjmed.2007.06.024 Headly Communication Skills 913 about chest pain (location, duration, timing, exacerbat- lend themselves to relatively simple scripts. The tasks ing and alleviating factors, quality, specific associated should be specific and concrete, such as in the suggested symptoms, etc), but very little advice is available on how sample scripts provided in the Table.26,27 Although exact to tell patients the diagnosis. When physicians-in-training statements made in each particular scenario will differ— enter into difficult discussions, they are forced to do so obtaining informed consent for an HIV test is different unrehearsed, which is a disservice than obtaining it for a high- to both the physician-in-training risk surgical procedure— and the patient (and imparts the PERSPECTIVES VIEWPOINTS the specific tasks them- false message that “routine” inter- selves (eg, explaining the viewing is not emotionally difficult ● Effective physician-patient communica- diagnosis or the problem for physicians or patients). tion is central to patient care. and explaining the proposed Physician-in-training confidence in procedure) are fairly uni- their general social communication ● Communication skills should be a major form. Educators will dis- skills might actually hamper perfor- emphasis of medical education. agree about the most impor- mance in clinical interactions,17 tant scenarios to teach or the ● Communication “scripts” are a useful which should not be surprising, as a appropriateness of individ- person who is good at small talk in tool in medical education curricula. ual tasks, and the suggested social settings may believe that con- scripts in the Table are in- necting with patients is no different. tended to provoke discus- Yet if physicians-in-training never practice saying “I’m sion and new suggestions. Nevertheless, the underlying afraid I have some bad news,” they are likely to find hypothesis—that communication skills should be taught themselves at a loss when the time comes to relay bad by giving physicians-in-training explicit step-by-step news to a patient. Some small studies have shown that skills rather than attempting to impart abstract concepts or practicing communication skills tailored to specific sit- attitudes—is not dependent on which particular tasks or uations can improve performance,13,18-20 and a random- situations are addressed. Some critics may argue that this ized controlled trial in which residents were given step- approach risks creating automatons that mouth rigid by-step instruction and practice in interviewing for scripts and ignore all the subtleties present in a patient specific situations demonstrated improvement in resi- interaction. On the contrary, giving students tools for and dents’ abilities to conduct the medical interview.21 The practice in making the kinds of statements that are known General Internal Medicine Generalist Educational to be effective in specific situations will help students Leadership (GIMGEL) Group has published practical unlock the mysteries of difficult patient interactions and recommendations on clinical teaching of psychosocial go on to build unique physician-patient relationships. aspects of patient care, which includes some specific If communication tasks were taught in the same way questions addressing topics such as health literacy (“A as other skills in medicine, using concrete algorithms lot of people have trouble reading things they get from that students can adapt and build on as they mature in the doctor because of all the medical words. Is it hard their clinical abilities, it could markedly change current for you to read the things you get here?”).22 Specific educational practice. For instance, master clinicians statements made by clinicians enhance the satisfaction would have a formal structure in which to teach inter- of critically ill patients’ families (eg, assurances that the personal skills. Such experts have developed commu- patient will be comfortable and will not suffer).23 The nication skills through experience and mentoring; they next step is the development of a comprehensive cur- all have “scripts” to use in certain situations and it is riculum based on clear statements such as these. Stu- likely that their scripts share many common ele- dents need concrete, situation-specific tasks from which ments.28,29 Several of the articles in the Annals of they can build communication skills. Internal Medicine series, “Words that Make a Differ- The topic of physician-patient communication is ence,” include brief scripts,27,30-33 such as, “I am hop- broad, but it can be broken down into specific teachable ing for the best. I think that, at the same time, we need components. For example, breaking bad news and con- to prepare for the worst.”30 If more of these scripts ducting end-of-life discussions are important situations could be compiled, they would be a tremendous re- in which young physicians typically feel the most ill- source for physicians-in-training. These scripts need prepared.12,24 Communicating with patients who have not be only written—video scenarios of clinicians in limited English proficiency is a rapidly growing prob- action could help bring the scripts to life. lem. Discussing risks and benefits of procedures and The current state of knowledge in medical commu- obtaining informed consent has been emphasized by the nication education lags far behind that of other educa- US Agency on Healthcare Research and Quality as vital tional interventions. There is a notable lack of scholarly to patient autonomy and safety.25 These topics, while assessment of medical communication instruction. beyond the basics of the initial medical interview, still Medical educators insist on evidence-based patient 914 The American Journal of Medicine, Vol 120, No 10, October 2007

Table Sample Scripts for Specific Patient Encounters26,27 Giving bad news: Sit down facing the patient, if possible; do not stand over patient or sit on bed. Mention the purpose of the meeting: “I’m going to go over the test results with you now.” Set the stage: “I’m afraid I have some bad news.” Give the news unequivocally: “The test showed that you do have ______.” Stay silent while the patient absorbs the news. Touch the patient (on the hand, arm, or shoulder only) if it seems appropriate. Find out what the patient knows: “Tell me what you know about this illness.” Show empathy: “I wish I had better news.” Make a plan for the next step: “Here’s what’s going to happen next.” Informed consent: Explain the diagnosis or the problem: “You’ve been having chest pain, and we don’t know if this pain is coming from your heart.” Explain the proposed procedure: “We’d like to do something called a heart catheterization . . .” Explain the risks of the procedure: “We do everything as carefully as possible, but even when we do everything right, things can go wrong . . .” Explain the benefits of the procedure: “The reason I’m recommending that you have this done is because it will tell us if your heart is okay, or if you have a blockage somewhere that needs surgery or another procedure.” Make sure the patient understands that it is up to him or her to decide: “I’m recommending this, but you are in charge, and you don’t have to do this if you don’t want to. There are other things we can do if you don’t want this.” Ensure understanding: “Can you tell me in your own words what I’ve just explained about this procedure?” Working with an interpreter: Determine whether an interpreter is needed: “How is your English?” Do not use a family member unless the patient insists and clearly understands that another interpreter can be obtained. Do not use a minor child. Speak in the first person to the patient: “Tell me what brings you here.” Look at the patient, not the interpreter or the phone. Speak in short phrases. Redirect the interpreter if necessary: “Please translate everything the patient is saying.”

care, but they also should insist on evidence-based New guidelines are mandating that medical educa- education. A great deal of curriculum development is tion incorporate ideals of patient-centered care, eth- occurring in the country, but it is rarely subjected to ics, professionalism, and humanism, and communi- controlled evaluation, meaning that there is almost no cation skills are central to all of these ideas. To date, evidence upon which to select elements of a curricu- no method for achieving these goals has been de- lum. Adding elements to a curriculum comes at an scribed. Although it is admittedly hard to teach appro- expense in terms of both time and money,34 so educa- priate behavior and harder still to measure it,44 reaching tors need to know what works and what does not. a consensus on the framework of appropriate commu- However, for rigorous evaluation of communication nication would provide medical educators with a com- teaching methods to be possible, a generally agreed- mon starting point and provide students with the skills upon set of critical elements of communication in dif- to hit the right notes with patients. ferent settings needs to exist. Once a communications curriculum is constructed, methods are already in place for teaching and evaluating it. Role-playing is an effective technique for ACKNOWLEDGMENT teaching communication skills to medical stu- The author thanks Michael Picchioni, MD, and Joshua dents35-38 as are standardized patient workshops.39 P. Metlay, MD, for their helpful comments on earlier Watching clinicians at work is an invaluable learning drafts of this essay. experience and while it is unfortunately not feasible for every physician-in-training to be apprenticed to an expert communicator, the use of digital teaching References formats shows promise.40 Objective structured clin- 1. Ury WA, Berkman CS, Weber CM, et al. Assessing medical ical examinations (OSCEs) are a widely accepted students’ training in end-of-life communication: A survey of examination technique for clinical competence, in- interns at one urban teaching hospital. Acad Med. 2003; 41-43 78(5):530-537. cluding competence in communication. More- 2. Murphy J, Chang H, Montgomery JE, et al. The quality of over, OSCEs are similar in format to the clinical physician-patient relationships. Patients’ experiences 1996-1999. skills examination portion of USMLE Step 2.4 J Fam Pract. 2001;50(2):123-129. Headly Communication Skills 915

3. Association of American Medical Colleges. Contemporary Is- 25. Pizzi L, Goldfarb N, Nash D. Making health care safer: a critical sues in Medicine: Communications in Medicine. Report III. Med- analysis of patient safety practices. AHRQ Evidence Report/ ical School Objectives Project, Washington, DC: 1999. Technology Assessment. 2001;43:546-554. 4. United States Medical Licensing Examination. USMLE Bulletin 26. Gordon G. Giving bad news. In: Behavioral Medicine in Primary of Information. Online. http://www.usmle.org/bulletin/default. Care (2nd Edition). Feldman MD, Christensen JF (eds). New htm. Accessed June 26, 2007. York, NY: Lange Medical Books, 2003. 5. Buyck D, Lang F. Teaching medical communication skills: A 27. Quill TE, Arnold RM, Platt F. “I wish things were different”: call for greater uniformity. Fam Med. 2002;34(5):337-343. Expressing wishes in response to loss, futility, and unrealistic 6. Gorter S, Rethans JJ, Scherpbier A, et al. Developing case- hopes. Ann Intern Med. 2001;135(7):551-555. specific checklists for standardized-patient-based assessments in 28. Klitzman R. Improving education on doctor-patient relationships internal medicine: A review of the literature. Acad Med. 2000; and communication: Lessons from doctors who become patients. 75(11):1130-1137. Acad Med. 2006;81(5):447-453. 7. Yedidia MJ, Gillespie CC, Kachur E, et al. Effect of communi- 29. Sargeant J, Mann K, Sinclair D, et al. Learning in practice: cations training on medical student performance. JAMA. 2003; Experiences and perceptions of high-scoring physicians. Acad 290(9):1157-1165. Med. 2006;81(7):655-660. 8. Makoul G. Essential elements of communication in medical 30. Back AL, Arnold RM, Quill TE. Hope for the best, and prepare encounters: The Kalamazoo consensus statement. Acad Med. for the worst. Ann Intern Med. 2003;138(5):439-443. 2001;76(4):390-393. 31. Back AL, Arnold RM, Tulsky JA, et al. On saying goodbye: 9. Rider EA, Keefer CH. Communication skills competencies: Def- Acknowledging the end of the patient-physician relationship initions and a teaching toolbox. Med Educ. 2006;40(7):624-629. with patients who are near death. Ann Intern Med. 2005;142(8): 10. Brotherton SE, Rockey PH, Etzel SI. US graduate medical edu- 682-685. cation, 2002-2003. JAMA. 2003;290(9):1197-1202. 32. Baker LH, O’Connell D, Platt FW. “What else?” Setting the 11. Brody B. The script. N Engl J Med. 2006;355(10):979-981. agenda for the clinical interview Ann Intern Med. 2005;143(10): 12. Eggly S, Afonso N, Rojas G, et al. An assessment of residents’ 766-770. competence in the delivery of bad news to patients. Acad Med. 33. Levinson W, Cohen MS, Brady D, Duffy FD. To change or not 1997;72(5):397-399. to change: “Sounds like you have a dilemma”. Ann Intern Med. 13. Lee KC, Winickoff JP, Kim MK, et al. Resident physicians’ use 2001;135(5):386-391. of professional and nonprofessional interpreters: a national sur- 34. Kelly M, Murphy A. An evaluation of the cost of designing, vey. JAMA. 2006;296(9):1050-1053. delivering and assessing an undergraduate communication skills 14. Stewart MA. Effective physician-patient communication and module. Med Teach. 2004;26(7):610-614. health outcomes: A review. CMAJ. 1995;152(9):1423-1433. 35. Benbassat J, Baumal R. A step-wise role playing approach for 15. Levinson W. Doctor-patient communication and medical malprac- teaching patient counseling skills to medical students. Patient tice: Implications for pediatricians. Pediatr Ann. 1997;26(3):186- Educ Couns. 2002;46(2):147-152. 193. 36. Brady D, Schultz L, Spell N, Branch WT Jr. Iterative method for 16. Hanna M, Fins JJ. Viewpoint: Power and communication: Why learning skills as an efficient outpatient teacher. Am J Med Sci. simulation training ought to be complemented by experiential 2002;323(3):124-129. and humanist learning. Acad Med. 2006;81(3):265-270. 17. Humphris GM. Communication skills knowledge, understanding 37. Rosenbaum ME, Kreiter C. Teaching delivery of bad news using and OSCE performance in medical trainees: A multivariate pro- experiential sessions with standardized patients. Teach Learn spective study using structural equation modelling. Med Educ. Med. 2002;14(3):144-149. 2002;36(9):842-852. 38. Skelton JR, Matthews PM. Teaching sexual history taking to 18. Colletti L, Gruppen L, Barclay M, Stern D. Teaching students to health care professionals in primary care. Med Educ. 2001;35(6): break bad news. Am J Surg. 2001;182(1):20-23. 603-608. 19. Henwood PG, Altmaier EM. Evaluating the effectiveness of 39. Haist SA, Griffith IC, Hoellein AR, et al. Improving students’ communication skills training: a review of research. Clin Per- sexual history inquiry and HIV counseling with an interactive form Qual Health Care. 1996;4(3):154-158. workshop using standardized patients. J Gen Intern Med. 2004; 20. Siegal HA, Cole PA, Li L, Eddy MF. Can a brief clinical 19(5 Pt 2):549-553. practicum influence physicians’ communications with patients 40. Solomon DJ, Ferenchick GS, Laird-Fick HS, Kavanaugh K. A about alcohol and drug problems? Results of a long-term follow- randomized trial comparing digital and live lecture formats. BMC up. Teach Learn Med. 2000;12(2):72-77. Med Educ. 2004;4(27). 21. Smith RC, Lyles JS, Mettler J, et al. The effectiveness of inten- 41. Vu NV, Barrows HS. Use of standardized patients in clinical sive training for residents in interviewing. A randomized, con- assessments: Recent developments and measurement findings. trolled study. Ann Intern Med. 1998;128(2):118-126. Educational Researcher. 1994;23(3). 22. Kern DE, Branch WT, Jr, Jackson JL, et al. Teaching the psy- 42. Hodges B, Turnbull J, Cohen R, Bienenstock A, Norman G. chosocial aspects of care in the clinical setting: Practical recom- Evaluating communication skills in the OSCE format: Reliability mendations. Acad Med. 2005;80(1):8-20. and generalizability. Med Educ. 1996;30(1):38-43. 23. Stapleton RD, Engelberg RA, Wenrich MD, et al. Clinician 43. Yudkowsky R, Downing SM, Sandlow LJ. Developing an insti- statements and family satisfaction with family conferences in the tution-based assessment of resident communication and interper- intensive care unit. Crit Care Med. 2006;34(6):1679-1685. sonal skills. Acad Med. 2006;81(12):1115-1122. 24. Fraser H, Kutner J, Pfeifer M. Senior medical students’ percep- 44. Moorhead R, Winefield H. Teaching counseling skills to fourth- tions of the adequacy of education on end-of-life issues. J Palliat year medical students: A dilemma concerning goals. Fam Pract. Med. 2001;4(3):337-343. 1991;8(4):343-346. The American Journal of Medicine (2007) 120, 916-917

MEDICAL HUMANITIES PERSPECTIVES Helle Mathiasen, Cand Mag., PhD, Section Editor The Heroic Physician and The Gross Clinic Helle Mathiasen, Cand Mag., PhD Program in Medical Humanities, University of Arizona College of Medicine, Tucson. I never knew of but one artist, and this is Tom Eakins, who could resist the temptation to see what they think ought to be rather than what is. —Walt Whitman “One painting, many visitors – ‘The Gross Clinic’ is a hit in its new home. Eakins’s masterpiece went on display at the Art Museum. It was worth the fuss, they said.”1 This headline in the Philadelphia Inquirer, January 6, 2007, announces the arrival and display of an important medical painting at the Philadelphia Museum of Art. In 1875, Dr. Gross’s students commissioned Philadel- phia-born painter Thomas Eakins (1844-1916) to create a portrait of the distinguished surgeon Dr. Samuel Gross (1805-1884). Eakins chose to show his teacher at work in the operating theater at Jefferson Medical College. The venerable Dr. Gross was about to retire, and his students wanted to honor him. They bought the work from Eakins for $200 and donated it to Jefferson Medical College. One hundred thirty-two years later, in January 2007, as Jefferson Medical College was getting ready to put it up for sale, Christie’s valued the painting at $68 million.2 After an intensive local and national fundraising campaign, the Pennsylvania Academy of the Fine Arts and the Philadel- phia Museum of Art purchased Eakins’s most controversial painting in joint ownership. Although too avant-garde for Figure Portrait of Dr. Samuel D. Gross (The Gross Clinic)by 19th-century Philadelphia, Eakins’s representation of the Thomas Eakins. Philadelphia Museum of Art: Gift of the Alumni blood and nudity of surgical practice in The Gross Clinic Association to Jefferson Medical College in 1878 and purchased has been praised as “. . . the greatest single painting in the by the Pennsylvania Academy of the Fine Arts and the Philadel- history of American art” (p 190).3 phia Museum of Art in 2007 with the generous support of more At 30 years of age, Eakins wrote to a friend, “I have just than 3400 donors. got a new picture blocked in and it is far better than any- thing I have ever done” (p 181).3 Eakins decided to submit his portrait of Dr. Gross as an entry for the Philadelphia ing for its representation of a surgeon’s bloody hand and Centennial Exhibition of 1876. However, the Centennial scalpel, claiming it was unsuitable for ladies and children to Selection Committee judged the work unﬁt for display in look at. The patient’s nude, lower body prominently dis- the Main Exhibition Hall. It was hung in a side gallery played was another source of outrage. This censure caused among army medical exhibits. Critics condemned the paint- the artist great disappointment. In the painting (Figure), we see Dr. Gross in the amphi- Requests for reprints should be addressed to Helle Mathiasen, Cand. theater at Jefferson Medical College performing an opera- Mag, PhD. tion for which he was famous: removing infected bone from E-mail address: [email protected] a young patient with osteomyelitis of the femur. Earlier

0002-9343/$ -see front matter © 2007 Elsevier Inc. All rights reserved. doi:10.1016/j.amjmed.2007.04.021 Mathiasen The Heroic Physician and The Gross Clinic 917 treatments of osteomyelitis often had resulted in amputa- Arts (p 20).3 He attended anatomical classes, drawing tion, but Dr. Gross had perfected a surgical procedure in- classes, and Dr. Gross’s surgical clinics. While an art stu- volving anesthesia. The painting’s composition is a triangle, dent in Paris from 1866 to 1869, Eakins dissected and its apex Dr. Gross’s head. The surgeon holds a bloody witnessed surgical clinics at Paris Hospitals and the École scalpel between his right thumb and forefinger; he turns to de Medicine. During this period, he took drawing and paint- the right, away from the patient, lecturing to students and ing classes where live nude models were used. Returning to faculty in the audience. On his left is the bed with the young Philadelphia, he continued anatomy and dissection classes, patient reclining on his right side, his buttocks and sock-clad later telling an interviewer: “No one dissects to quicken his feet toward the viewer. In the foreground, we see a table eye for, or his delight in, beauty. He dissects simply to covered with surgical instruments. The sole woman in the increase his knowledge of how beautiful objects are put painting is dressed in black and sits below Dr. Gross’s right together to the end that he may be able to imitate them” hand. Possibly the patient’s mother, she lifts her arms and (p 55).4 Eakins’s fascination with the body profited his art, curls her fingers in front of her face in a protective gesture. as evidenced in The Gross Clinic, but it hindered his career The 5 men assisting the surgeon have been identified by advancement. name. The anesthesiologist presses a chloroform-soaked cloth Eakins became a popular teacher at the Pennsylvania onto the patient’s face; another physician is scraping the pa- Academy of the Fine Arts, where he launched a number of tient’s femur; a third pulls the retractor holding the wound curricular modifications, including teaching his students to open, and a fourth holds the patient’s feet. The fifth assistant is work from live nude models. Because of the difficulty of invisible. We glimpse his thumb pulling the other retractor finding models, especially women, he asked some of his beneath Dr. Gross’s left elbow. The person writing in the front students to pose nude and did so himself. In addition, he row is the Medical College recorder. Eakins has painted him- painted and photographed some of his students in the nude self drawing or writing to the viewer’s right. In the back- in the studio and outdoors. The Pennsylvania Academy ground, the artist has placed dark figures of spectators, some of Board regulations prohibited students from studying nude whom have been identified. models in mixed classes. As a consequence of his rebellion The artist emphasizes the drama of the operating theater against these rules, Eakins was forced to resign his faculty by his use of chiaroscuro, a technique made famous by Rembrandt in his painting “The Anatomy Lesson of Dr. position. Tulp.” Most of the picture is dark, but significant light falls Although his personal life continued to be unconven- on Dr. Gross’s head and hand, the sheet, and the patient’s tional, none of the artist’s work after The Gross Clinic lower body. Eakins illuminates the surgeon’s high forehead, caused similar controversy. Eakins became an esteemed creating an aura of his gray hair. The light symbolizes the painter of athletes; he painted portraits of people he ad- doctor’s remarkable intellect. The heroic physician stands mired, including 25 Philadelphia physicians. His contem- erect, master of himself and his task. The cringing woman porary and fellow American master Walt Whitman recog- on his right may embody the uneducated layperson who nized Thomas Eakins’s true genius, as art historians do 4 reacts emotionally to seeing a child’s blood and exposed, today (p 144). nude body. Eakins’s wife, the painter Susan Macdowell Eakins, said References that her husband contemplated becoming a physician or an 1. Salisbury S. One painting, many visitors. Philadelphia Inquirer 2007 artist. He began studying anatomy, physiology, and dissec- Jan 6. tion with physicians already as a student at Central High 2. http://www.philamuseum.org. Accessed March 26, 2007. 3. Kirkpatrick SD. The Revenge of Thomas Eakins. New Haven, CT: Yale School in Philadelphia. He enrolled as a medical student at University Press; 2006. Jefferson Medical College in 1864, and continued his study 4. Johns E. Thomas Eakins, the Heroism of Modern Life. Princeton, NJ: of the human body at the Pennsylvania Academy of Fine Princeton University Press; 1983.