Issue. 1 March 2004 (Vol. 116, Issue 5)

Transcription

1

2 Issue 1 March 2004 (Vol. 116, Issue 5) Clinical study Diagnosing pulmonary embolism in outpatients with clinical assessment, D-Dimer measurement, venous ultrasound, and helical computed tomography: a multicenter management study Perrier A, Roy PM, Aujesky D, Chagnon I, Howarth N, Gourdier AL, Leftheriotis G, Barghouth G, Cornuz J, Hayoz D, Bounameaux H pages Abstract Full Text PDF (120 KB) Clinical study A multivariate model for predicting mortality in patients with heart failure and systolic dysfunction Brophy JM, Dagenais GR, McSherry F, Williford W, Yusuf S pages Abstract Full Text PDF (98 KB) Clinical study Heart failure in rheumatoid arthritis: rates, predictors, and the effect of anti tumor necrosis factor therapy Wolfe F, Michaud K pages Abstract Full Text PDF (111 KB) Clinical study Regional cerebral hypoperfusion in patients with celiac disease Addolorato G, Giuda DD, Rossi GD, Valenza V, Domenicali M, Caputo F, Gasbarrini A, Capristo E, Gasbarrini G pages Abstract Full Text PDF (168 KB) Clinical study

3 Cerebrospinal fluid interleukin 8 concentrations and the subsequent development of postherpetic neuralgia Kotani N, Kudo R, Sakurai Y, Sawamura D, Sessler DI, Okada H, Nakayama H, Yamagata T, Yasujima M, Matsuki A pages Abstract Full Text PDF (136 KB) Clinical study Cost-effectiveness of inhaled corticosteroids for chronic obstructive pulmonary disease according to disease severity Sin DD, Golmohammadi K, Jacobs P pages Abstract Full Text PDF (88 KB) Review An approach to the diagnosis and treatment of cryofibrinogenemia Amdo TD, Welker JA pages Abstract Full Text PDF (160 KB) Review Leukotriene receptor antagonists for allergic rhinitis: a systematic review and meta-analysis Wilson AM, O'Byrne PM, Parameswaran K pages Abstract Full Text PDF (132 KB) Brief clinical observation Radioiodine ablation of the thyroid to allow the reintroduction of amiodarone treatment in patients with a prior history of Amiodarone-Induced thyrotoxicosis Hermida JS, Jarry G, Tcheng E, Moullart V, Arlot S, Rey JL, Delonca J, Schvartz C pages Full Text PDF (63 KB) Images of Osler Cases from the Osler Medical Service at Johns Hopkins University Schopick E, Cavanaugh K pages Full Text PDF (165 KB) Editorial Lessening the diagnostic uncertainty in patients with suspected pulmonary embolism Hull RD, Ghali WA, Pineo GF pages Full Text PDF (42 KB) Correspondence

4 Long-term outcome of patients treated with hem atopoietic growth factors for idiosyncratic druginduced agranulocytosis Andrès E, Noel E, Maloisel F page 354 Full Text PDF (42 KB) Correspondence Hypothyroidism and pulmonary arterial hypertension Ghamra ZW, Dweik RA, Arroliga AC pages Full Text PDF (55 KB) Correspondence Longitudinal myelitis in a pregnant patient with sle Moranne O, Hachulla E, Valat AS, Sotoares G, Pagniez D, Boulanger E pages Full Text PDF (96 KB) Correspondence Identification of drug-drug interactions with personal digital assistant based software Robinson RL, Burk MS pages Full Text PDF (53 KB) Correspondence Oral vitamin b12 therapy in vitamin b12 deficiency Roth M, Orija I page 358 Full Text PDF (44 KB) APM views Using local solutions to address a national issue: ASP's recommendations to ensure the vitality of clinical research Crist TB, Barr WG, Linas SL, Zuckerman KS pages Full Text PDF (76 KB)

5 CLINICIAL STUDIES Diagnosing Pulmonary Embolism in Outpatients with Clinical Assessment, D-Dimer Measurement, Venous Ultrasound, and Helical Computed Tomography: A Multicenter Management Study Arnaud Perrier, MD, Pierre-Marie Roy, MD, Drahomir Aujesky, MD, Isabelle Chagnon, MD, Nigel Howarth, MD, Anne-Laurence Gourdier, MD, Georges Leftheriotis, MD, Ghassan Barghouth, MD, Jacques Cornuz, MD, MPH, Daniel Hayoz, MD, Henri Bounameaux, MD PURPOSE: To evaluate a diagnostic strategy for pulmonary embolism that combined clinical assessment, plasma D-dimer measurement, lower limb venous ultrasonography, and helical computed tomography (CT). METHODS: A cohort of 965 consecutive patients presenting to the emergency departments of three general and teaching hospitals with clinically suspected pulmonary embolism underwent sequential noninvasive testing. Clinical probability was assessed by a prediction rule combined with implicit judgment. All patients were followed for 3 months. RESULTS: A normal D-dimer level ( 500 g/l by a rapid enzyme-linked immunosorbent assay) ruled out venous thromboembolism in 280 patients (29%), and finding a deep vein thrombosis by ultrasonography established the diagnosis in 92 patients (9.5%). Helical CT was required in only 593 patients (61%) and showed pulmonary embolism in 124 patients (12.8%). Pulmonary embolism was considered ruled out in the 450 patients (46.6%) with a negative ultrasound and CT scan and a low-to-intermediate clinical probability. The 8 patients with a negative ultrasound and CT scan despite a high clinical probability proceeded to pulmonary angiography (positive: 2; negative: 6). Helical CT was inconclusive in 11 patients (pulmonary embolism: 4; no pulmonary embolism: 7). The overall prevalence of pulmonary embolism was 23%. Patients classified as not having pulmonary embolism were not anticoagulated during follow-up and had a 3-month thromboembolic risk of 1.0% (95% confidence interval: 0.5% to 2.1%). CONCLUSION: A noninvasive diagnostic strategy combining clinical assessment, D-dimer measurement, ultrasonography, and helical CT yielded a diagnosis in 99% of outpatients suspected of pulmonary embolism, and appeared to be safe, provided that CT was combined with ultrasonography to rule out the disease. Am J Med. 2004;116: by Excerpta Medica Inc. In recent years, extensive research has been devoted to developing noninvasive (1 4) and cost-effective (5,6) diagnostic strategies for pulmonary embolism, considering that angiography is costly and invasive. From Medical Clinic 1 (AP), and Divisions of Angiology and Hemostasis (IC, HB) and Radiodiagnosis (NH), Geneva University Hospital, Geneva, Switzerland; Emergency Department (PMR), Department of Radiology (ALG), and Vascular Investigations Department (GL), Angers University Hospital, Angers, France; and Department of Medicine (DA, JC), Departments of Radiology and Nuclear Medicine (GB), Institute of Social and Preventive Medicine (JC), and Division of Hypertension and Vascular Medicine (DH), University Hospital, Lausanne, Switzerland. This study was supported by a grant ( ) from the Swiss National Research Foundation; grants 97/4-T10 and 00/4-T9 from the Royal College of Physicians and Surgeons, Canada; a grant from La Fondation Québécoise pour le Progrès de la Médecine Interne and Les Internistes et Rhumatologues Associés de l Hôpital du Sacré-Cœur, Montreal, Canada; and grant 2001/021 from the Direction of Clinical Research of the Angers University Hospital. Requests for reprints should be addressed to Arnaud Perrier, MD, Medical Clinic 1, Geneva University Hospital, 24 rue Micheli-du-Crest, 1211 Geneva 14, Switzerland, or Arnaud.Perrier@hcuge.ch. Manuscript submitted June 2, 2003, and accepted in revised form September 11, Moreover, even patients discharged without anticoagulant treatment based on a normal pulmonary angiogram have an approximately 1% 3-month risk of thromboembolic events (7,8). Plasma D-dimer, a degradation product of crosslinked fibrin, has been studied extensively as a first-line test to rule out pulmonary embolism (1,9 11), but its use is not endorsed universally. Lower limb venous ultrasonography (12) has a low diagnostic yield when performed in patients with a nondiagnostic ventilation/perfusion (V/Q) lung scan or a negative helical computed tomography (CT) scan (13). Nevertheless, when used before other imaging modalities, it shows a proximal deep vein thrombosis in about 10% of patients suspected of pulmonary embolism (1,14). The sensitivity of single-detector helical CT may be as low as 70% (15,16). Onthe other hand, 15% of patients with clinical symptoms of pulmonary embolism and a negative helical CT scan have a deep vein thrombosis, and combining ultrasonography with CT may reduce the overall rate of false-negative results (16). Finally, the stratification of patients according 2004 by Excerpta Medica Inc /04/$ see front matter 291 All rights reserved. doi: /j.amjmed

6 Diagnosing Pulmonary Embolism in Outpatients/Perrier et al to the clinical probability of pulmonary embolism, whether assessed implicitly (1,17) or by a prediction rule (18,19), may reduce the requirement for invasive tests and resource use (1,2). Therefore, we designed this prospective outcome study to evaluate the efficacy and safety of a diagnostic strategy combining clinical probability assessment, D-dimer measurement, lower limb ultrasonography, and helical CT in a multicenter cohort of unselected emergency department patients. METHODS Patients Data were collected from October 1, 2000, to June 30, 2002, at three medical centers that serve as general or teaching hospitals (Geneva University Hospital, Geneva, Switzerland; University Hospital, Lausanne, Switzerland; and Angers University Hospital, Angers, France). The study was approved by the Ethics Committee of the Department of Medicine, Geneva University Hospital; the Ethics Committee of the Lausanne Medical School; and the Comité Consultatif de Protection des Personnes dans la Recherche Biomédicale des Pays de la Loire in Angers. Consecutive patients presenting to the emergency department were eligible if there was suspicion of pulmonary embolism, defined as acute onset of new or worsening shortness of breath or chest pain without another obvious etiology. Among the 1290 screened patients, 258 patients (20.0%) were excluded according to the following predefined criteria: ongoing anticoagulant treatment for reasons other than venous thromboembolism (n 43); contraindication to CT scan (known allergy to iodine contrast agents or at risk of allergic reaction) (n 36); creatinine clearance below 30 ml/min as calculated by the Cockroft formula (n 53) (20); informed consent impossible due to cognitive impairment (n 24); patient refusal (n 57); suspected massive pulmonary embolism with shock (n 10); pregnancy (n 9); estimated survival less than 3 months (n 8); follow-up not possible (n 11); and other reasons (n 7). Another 67 patients (5.2%) were excluded because of protocol violations: diagnostic test not performed (D-dimer measurement: n 1; ultrasonography: n 21; helical CT: n 17; pulmonary angiography: n 17); and final diagnosis established by criteria different from those required by the study (n 11). The final study sample consisted of 965 patients (75%). An ancillary study comparing the Geneva prediction rule (19) and the Wells rule (18) for assessing the clinical probability of pulmonary embolism in a subset of the study sample has been published (21). Table 1. Prediction Rule for Evaluating the Clinical Probability of Pulmonary Embolism Variable Score* Previous pulmonary embolism or 2 deep vein thrombosis Heart rate 100 beats per minute 1 Recent surgery 3 Age (years) PaCO kpa (36 mm Hg) kpa ( mm Hg) 1 PaO kpa (48.7 mm Hg) kpa ( mm Hg) kpa ( mm Hg) kpa ( mm Hg) 1 Chest radiograph Platelike atelectasis 1 Elevated hemidiaphragm 1 * Clinical probability: low, 0 to 4 points; intermediate, 5 to 8 points; high, 9 or more points. From Wicki et al (19). PaCO 2 partial pressure of carbon dioxide, arterial; PaO 2 partial pressure of oxygen, arterial. Study Design The study was designed as a prospective management trial with a 3-month follow-up. Patients underwent clinical evaluation in the emergency department by the physicians in charge based on a previously published clinical prediction rule (19) prior to any other test (Table 1). Physicians filled out a standardized data collection form that also recorded risk factors for venous thromboembolism; symptoms and signs frequently encountered in pulmonary embolism, including symptoms or signs of deep vein thrombosis; a description of the electrocardiogram and chest radiograph; and the likelihood of an alternative diagnosis. As described previously (21), the physicians could override the prediction rule by implicit probability assessment in case of disagreement, in either direction. Hence, the final classification on which the diagnostic workup was based was an aggregate of the score and implicit clinical judgment. Sequential noninvasive tests were then performed (Figure). The initial test was plasma D-dimer measurement by enzyme-linked immunosorbent assay (ELISA) ruling out pulmonary embolism when below the cutoff of 500 g/l. Patients with D-dimer levels 500 g/l proceeded to lower limb compression ultrasonography. An ultrasound showing a deep vein thrombosis warranted anticoagulant treatment without further testing. Patients with a normal ultrasound underwent helical CT, and patients in whom the scan showed pulmonary embolism 292 March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume 116

7 Diagnosing Pulmonary Embolism in Outpatients/Perrier et al Figure. Flow chart summarizing the diagnostic process in the study. Several patients who were categorized as not having pulmonary embolism by the study criteria were anticoagulated during follow-up for reasons other than venous thromboembolism. The number of patients who were not anticoagulated at any time during follow-up is indicated in the Figure under the caption no Rx. Those numbers were used to calculate the 3-month thromboembolic risk. Ranges in square brackets indicate 95% confidence intervals. CT computed tomography; DVT deep vein thrombosis; ELISA enzyme-linked immunosorbent assay; PE pulmonary embolism; Rx treatment. were treated accordingly. Patients were then stratified according to the clinical probability of pulmonary embolism as determined before D-dimer measurement and ultrasound were performed. Patients with a low or intermediate clinical probability and a negative CT scan and ultrasound were not treated with anticoagulants (unless another indication for anticoagulant treatment was present) and did not undergo further testing. Those with a high clinical probability of pulmonary embolism underwent pulmonary angiography. A small proportion of patients had an inconclusive CT scan due to technical reasons (e.g., insufficient contrast enhancement of the pulmonary arteries or motion artifacts). Ventilation/perfusion scintigraphy or pulmonary angiography was performed to reach a definite diagnosis in that subgroup. Diagnostic Studies Plasma D-dimer levels (rapid ELISA assay, Vidas DD; BioMérieux, Marcy l Etoile, France) was assayed with an automated quantitative analyzer (22). Real-time, lower limb, B-mode venous compression ultrasonography of the common femoral and popliteal veins was performed within 24 hours in all patients. The criterion for diagnosing deep vein thrombosis was incomplete compressibility of the vein (23). The protocols for performing helical CT varied among centers and over time. The following features were common to all centers and CT examinations throughout the study: pulmonary arteries were evaluated up to and including the segmental vessels from the level of the aortic arch to the lowest hemidiaphragm. Patients were examined during suspended inspiration or shallow breathing, depending on the degree of dyspnea. Each vessel was scored for the presence or absence of a clot, including subsegmental vessels, when visualized. A clot was considered to be present if contrast material outlined an intraluminal defect or if a vessel was totally occluded by lowattenuation material on at least two adjacent slices. In Geneva and Angers, a single-detector CT was used for most of the study period (16) (Geneva: 81% of patients [348/430]; Angers: 80% of patients [227/251]). The acquisition parameters for single-detector CT were a total volume of 120 to 140 ml of nonionic contrast material injected with a power injector at 3 to 5 ml/s; imaging 12 to 15 seconds after initiation of the contrast material injection; scans performed at 3 mm per section with a pitch of 1.6 to 2.0, 120 kv, 200 mas, requiring 1.0 second per rotation; and images reconstructed at 2- to 3-mm intervals. Multidetector CT (24) was used in all 251 patients from Lausanne and in the remaining patients from Geneva and Angers. The acquisition parameters for multidetector CT were a total volume of 100 to 120 ml of nonionic contrast material injected with a power injector at 3 to 5 ml/s; imaging 9 to 20 seconds after initiation of the contrast material injection; scans performed at 1 to 1.3 mm per section with a pitch of 1.25 to 1.75, 120 kv, 115 to 260 mas; and images reconstructed at 0.6- to 0.8-mm intervals. For obese patients, slice thickness was sometimes increased to 2.5 mm. The technique for performing and interpreting lung scan and pulmonary angiography has been described elsewhere (14,25). Outcomes The outcome measurement for efficacy was the proportion of patients in whom a definite diagnosis could be made by the diagnostic strategy without an angiogram. Safety was assessed by the 3-month risk of thromboembolism in patients classified as not having pulmonary embolism and not anticoagulated at any time during follow-up except during the diagnostic workup. Diagnoses of venous thromboembolic events were established with usual criteria (deep venous thrombosis: abnormal ultrasonography; pulmonary embolism: high-probability V/Q scan, or helical CT or angiogram). Deaths were adjudicated as definitely caused by pulmonary embolism, March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume

8 Diagnosing Pulmonary Embolism in Outpatients/Perrier et al definitely unrelated to pulmonary embolism, or possibly due to pulmonary embolism. A 3-month thromboembolic risk of 4% (upper limit of the 95% confidence interval) was considered acceptable, as in similar outcome studies (1 3,26). Episodes of major bleeding (bleeding requiring transfusion, retroperitoneal, joint or cerebral hemorrhage) were also recorded. Three independent experts adjudicated outcome events. 3-Month Follow-up Patients were followed by their family physicians and interviewed by telephone by one of the study coordinators at the end of the follow-up period. In 5% of patients, follow-up information was obtained by mail. The family physician was contacted whenever a possible event was disclosed by the interim history, and charts were reviewed if a patient was readmitted to a hospital for any cause. Statistical Analysis Characteristics of patients were compared by the chisquared test with Fisher correction for small numbers for categorical variables and by the Mann-Whitney test for continuous variables (Statview 5.0 software for Windows; Abacus Concepts, Inc., Berkeley, California). The 95% confidence intervals for the incidence of thromboembolic and bleeding events during follow-up were calculated from the binomial distribution by means of Confidence Interval Analysis software (Trevor Bryant, University of Southampton, United Kingdom). RESULTS The 965 patients who met the inclusion criteria were younger (61 19 years vs years), had fewer comorbid conditions ( vs ), and a lower prevalence of previously diagnosed heart failure (10% [95/965] vs. 18% [59/325]) as compared with patients who were excluded (Table 2). Nevertheless, the prevalence of pulmonary embolism was similar in both included and excluded patients (23% [n 222] vs. 21% [n 69]). Diagnosis of Venous Thromboembolism Only 5 (6.8%) of the 74 patients with a high clinical probability of pulmonary embolism had normal D-dimer levels ( 500 g/l), compared with 31% (275/891) of patients with a low or intermediate clinical probability (P ) (Table 3). D-dimer and ultrasonography established a definite diagnosis in 38.5% of the entire cohort. Helical CT scans were positive for pulmonary embolism in 124 patients (12.8%). The most proximal levels of the detected clots were the main pulmonary arteries in 37 patients (30%), the lobar arteries in 41 (33%), and the segmental vessels in 44 (35%). Only 2 patients had multiple subsegmental pulmonary emboli. No patient had an isolated subsegmental clot. Ultrasound and CT scans were negative in 458 patients, of whom clinical probability was low or intermediate in 450 who were considered as not having pulmonary embolism and not required to undergo further testing. Eight patients (0.8%) had a high clinical probability and underwent angiography, which showed pulmonary emboli in 2 patients (bilateral subsegmental embolus in 1 patient and bilateral lobar embolus in the other patient). CT scans could not be interpreted for technical reasons in 11 (1.9%) of the 593 patients who underwent helical CT. Pulmonary embolism was ruled out in 7 of these patients (normal pulmonary angiogram in 2 patients; normal V/Q lung scan in 4; combination of a low-probability V/Q scan, a low clinical probability, and a negative venous ultrasound in 1) and was established by a high-probability V/Q scan in the remaining 4 patients. Assessment of Clinical Probability of Pulmonary Embolism The prediction rule for pulmonary embolism could not be calculated in 194 patients (20%) because arterial blood gas analysis was either not performed (n 137) or performed in patients receiving supplemental oxygen (n 57). Clinical probability was assessed by implicit judgment in those patients. In the remaining 771 patients, physicians adopted the score assessment in 592 patients and there was disagreement in 179 patients (23%). The clinical probability as assessed by the score was increased in 126 patients (70% of disagreements) and decreased in 53 patients. Pulmonary embolism was diagnosed in 34 of the 522 patients in the low-probability group (7%; 95% confidence interval [CI]: 5% to 9%), 125 of the 369 patients in the intermediate-probability group (34%; 95% CI: 29% to 39%), and 63 of the 74 patients in the highprobability group (85%; 95% CI: 75% to 92%). Follow-up Three patients were lost to follow-up. The two patients who were classified as not having pulmonary embolism were aged 21 and 35 years, were in good health, and had a low clinical probability of pulmonary embolism. One had a D-dimer level of 123 g/l and was discharged with a diagnosis of bronchitis. Pulmonary embolism was ruled out in the other patient by a negative ultrasound and CT scan; this patient was discharged with a diagnosis of bronchopneumonia and was alive at the end of the follow-up period. Among the 743 patients in whom venous thromboembolism was considered to be absent according to the diagnostic strategy, 58 underwent anticoagulation during follow-up for reasons other than venous thromboembolism, mainly atrial fibrillation. In the 685 remaining patients, 7 had an acute venous thromboembolic event during the 3-month follow-up (1.0%; 95% CI: 0.5% to 2.1%). All thromboembolic events occurred in the 406 patients with a low-to-intermediate clinical probability of pulmonary embolism, a negative ultrasound, and a neg- 294 March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume 116

9 Diagnosing Pulmonary Embolism in Outpatients/Perrier et al Table 2. Characteristics of the Study Sample (n 965) Characteristic Number (%), Mean SD, or Median (Interquartile Range) Pulmonary embolism 222 (23) Age (years) Female sex 562 (58) Risk factors Family history of deep vein thrombosis or pulmonary embolism 102 (11) Previous deep vein thrombosis 129 (13) Previous pulmonary embolism 89 (9) Known heart failure 95 (10) Paralysis 29 (3) Chronic obstructive pulmonary disease 99 (10) Cancer 89 (9) Surgery within 1 month 57 (6) Plaster cast within 12 weeks 13 (1) Immobilization* 165 (17) Oral contraceptives 69 (7) Hormone replacement therapy 53 (6) Clinical presentation Chest pain 678 (70) Syncope 68 (7) Dyspnea 634 (66) Clinical presentation Symptoms of deep vein thrombosis 197 (20) Hemoptysis 43 (5) Elevated jugular venous pressure 108 (11) Signs of deep vein thrombosis 174 (18) Signs of chronic venous insufficiency 199 (21) Pleural effusion 137 (14) Elevated hemidiaphragm 114 (12) Band atelectasis 88 (9) Heart rate (beats per minute) Systolic blood pressure (mm Hg) Temperature ( C) Respiratory rate (breaths per minute) 20 (8) PaO 2 (mm Hg) PaCO 2 (mm Hg) 36 6 * Bed rest 48 hours or travel 6 hours within 1 month. Unilateral lower limb edema and pain on palpation of the deep leg veins. Varicose veins, ochre dermatitis, or venous ulcers. PaCO 2 partial pressure of carbon dioxide, arterial; PaO 2 partial pressure of oxygen, arterial. ative helical CT scan, yielding a 3-month thromboembolic risk of 1.7% (95% CI: 0.8% to 3.5%), compared with 0% (95% CI: 0% to 1.4%) in the 268 patients with normal D-dimer levels. Two of these patients died of a possible pulmonary embolism (Table 4); the remaining 5 patients had nonfatal pulmonary embolism (n 3) or distal deep vein thrombosis (n 2). Nine (4.1%) of the patients with initial pulmonary embolism had a recurrent thromboembolic event (Table 4). Major bleeding was observed in 7 of the 222 patients who underwent anticoagulation for pulmonary embolism (3.2%; 95% CI: 1.5% to 6.4%), of whom 2 died (risk of fatal hemorrhage: 0.9%; 95% CI: 0.2% to 3.2%) (Table 4). Overall 3-month mortality was 7.7% (95% CI: 4.8% to 11.9%) in patients with pulmonary embolism and 2.7% (95% CI: 1.7% to 4.1%) in those in whom the condition was ruled out. DISCUSSION We found our diagnostic algorithm to be safe and effective in outpatients. The overall 3-month risk of thromboembolism in patients classified as having no pulmonary embolism by the study algorithm was 1.0% (95% CI: 0.5% to 2.1%), which is in keeping with that found in March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume

10 Diagnosing Pulmonary Embolism in Outpatients/Perrier et al Table 3. Diagnostic Criteria According to the Clinical Probability of Pulmonary Embolism in the Study Low (n 522) Clinical Probability of Pulmonary Embolism Intermediate (n 369) High (n 74) Number (%) Pulmonary embolism* DVT shown by ultrasound 10 (1.9) 46 (12.5) 36 (48.6) Positive CT scan 22 (4.2) 78 (21.1) 24 (32.4) Positive angiogram 2 (2.7) CT scan inconclusive 2 (0.4) 1 (0.3) 1 (1.4) No pulmonary embolism* D-dimer level 500 g/l 238 (45.6) 37 (10.0) 5 (6.8) Negative ultrasound and CT scan 245 (46.9) 205 (55.6) Normal angiogram 6 (8.1) CT scan inconclusive 5 (1.0) 2 (0.5) *Asdefined by the study criteria. Pulmonary embolism diagnosed by high-probability lung scan. Including low or intermediate clinical probability. Pulmonary embolism ruled out by a normal pulmonary angiogram (n 2); a normal V/Q lung scan (n 1); or a combination of a low clinical probability, a low-probability V/Q scan, and a negative ultrasound (n 1). CT computed tomography; DVT deep vein thrombosis; V/Q ventilation/perfusion. other similar management studies (1,2,27 29) (Table 5) and with the risk in patients who were not anticoagulated based on a normal pulmonary angiogram (6). Admittedly, all thromboembolic events during follow-up occurred in low- or intermediate-probability patients who had a negative venous ultrasound and a negative CT scan, yielding a risk of 1.7% (95% CI: 0.8% to 3.5%) in that subgroup. However, that rate is still acceptable and similar to that found in a French management study (29) in which the same diagnostic criterion was applied (Table 5). Pulmonary angiography was necessary in only 10 patients (1.0%). Adding the 17 patients who were excluded because the angiogram required by the study protocol was not performed, the frequency of angiography would still be only 2.7% in our cohort. Furthermore, performing D-dimer measurements as the first-line test followed by venous ultrasonography in patients with abnormal D- dimer levels reduced the number of CT scans required to reach a definite diagnosis (61.5% vs. 100% in other strategies based on CT [28,29]; Table 5), which may result in Table 4. Venous Thromboembolic Events and Major Bleeding Events during the 3-Month Follow-up (Including Patients without Pulmonary Embolism Who Were Anticoagulated during Follow-up) Events Pulmonary Embolism (n 222) No Pulmonary Embolism (n 743) Number (%) Thromboembolic events Fatal pulmonary embolism 4 0 Possible fatal pulmonary embolism 1 2 Nonfatal pulmonary embolism 0 3 Deep vein thrombosis (distal) 4 2 Total 9 (4.1) 7 (0.9) Major bleeding events Fatal 2 2* Nonfatal 5 5 Total 7 (3.2) 7 (0.9) * No anticoagulant treatment (cancer in 1 patient; multiple organ failure in 1 patient). Anticoagulant treatment for causes other than venous thromboembolism in 3 patients; spontaneous bleeding in 2 patients. 296 March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume 116

11 March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume Table 5. Comparison of Various Diagnostic Strategies for Suspected Pulmonary Embolism Evaluated in Management Studies Angiography Alone First Author (Reference) Musset (29) Van Strijen (28) Wells (2) Perrier (1) Present Study Number of tests performed per 1000 patients* D-dimer Venous ultrasound Helical CT V/Q lung scan Angiography Costs of diagnostic tests ($) 510, , , , , ,000 3-month nonfatal risk of 1.7% (1.0% 2.7%) 1.8% (0.8% 3.3%) 0.8% (0.3% 2.3%) 0.6% (0.2% 1.4%) 0.9 (0.2% 2.7%) 1.0% (0.5% 2.1%) thromboembolism (95% confidence interval) 3-month risk of fatal thromboembolism (95% confidence interval) 0.3% (0.02% 0.7%) 1.0% (0.4% 2.3%) 0.3% (0% 1.5%) 0% (0% 0.5%) 0% (0% 1.2%) 0.3% (0.1% 1.1%) * The figures are extracted from the original studies and applied to a hypothetical cohort of 1000 patients. In patients with an inconclusive ultrasound or CT scan, or those with a high clinical probability of pulmonary embolism and a negative ultrasound and CT scan. In patients with an inconclusive CT scan. Costs calculated for a hypothetical cohort of 1000 patients using U.S. costs extracted from reference 30: D-dimer: $12; ultrasound: $69; helical CT: $135; V/Q lung scan: $683; angiography: $510. In patients classified as without pulmonary embolism and not anticoagulated during the 3-month follow-up. CT computed tomography; V/Q ventilation/perfusion. Diagnosing Pulmonary Embolism in Outpatients/Perrier et al

12 Diagnosing Pulmonary Embolism in Outpatients/Perrier et al substantial cost savings, as shown by a recent cost-effectiveness analysis (6). Our study also confirms the accuracy of clinical probability assessment by the prediction rule with possible override by clinical judgment (21). The limitations of the prediction rule (19) have been described (21). Indeed, the score could be obtained for only 80% of patients, and clinicians overrode the score assessment using implicit clinical judgment in 23% of the patients in whom the score was available. We believe that combining implicit judgment with the score makes clinical sense, since prediction rules do not take into account rare but potentially important individual characteristics. Moreover, combining the two methods increased physician acceptance of the rule. The final classification of patients an aggregate of score and implicit judgment was not entirely standardized, but it had a fair accuracy and was probably more standardized than would have been using implicit assessment alone. The study provides additional evidence on the safety and efficacy of a highly sensitive ELISA D-dimer assay as the first-line test for ruling out pulmonary embolism in outpatients. Indeed, none of the 268 patients with D- dimer levels 500 g/l and who were not anticoagulated had a thromboembolic event during follow-up. Pooling these results with those from a previous study that used the same assay (1), the 3-month risk of thromboembolism in patients who were not anticoagulated based on the results of this particular ELISA test was 0% (95% CI: 0% to 0.9%; 0/423). However, those results cannot be extrapolated to other less sensitive assays or to the inpatient setting. We chose to perform venous ultrasonography before thoracic imaging, which revealed unequivocal signs of proximal deep vein thrombosis in 9.5% of patients, whereas in a recent Dutch study in which ultrasonography was carried out after a normal helical CT scan, only two cases of thrombosis were found among the 248 patients (0.8%) (28). Venous ultrasonography also served to compensate for the limited sensitivity of single-slice helical CT. Indeed, previous studies (16,29) showed that 15% of patients with clinical symptoms of pulmonary embolism and a negative helical CT scan have a deep vein thrombosis. Hence, our results should not be interpreted as a validation of a negative CT scan to rule out pulmonary embolism. Furthermore, pending results of outcome studies using multidetector CT (24), CT should be combined with venous ultrasonography to exclude venous thromboembolism safely. Moreover, even if multidetector CT was proven to be safe as a single imaging modality for suspected pulmonary embolism, such a strategy would be considerably more resource intensive than a strategy that combined CT and other less expensive noninvasive tests (Table 5). Generalization of our findings might be a concern since 25% of eligible patients were excluded. However, predefined exclusion criteria largely accounted for the differences between included and excluded patients. The mix of single- and multidetector CT scans is also a limitation, but was inevitable because of the continuous evolution of radiological technology. However, four of the five pulmonary emboli diagnosed during follow-up in patients without pulmonary embolism on initial workup occurred in the 183 patients evaluated by multidetector CT; this included the two deaths that were possibly attributable to pulmonary embolism. Hence, it is unlikely that the 3-month risk of thromboembolism would have been different had we used only multidetector machines. In conclusion, our study shows that a diagnostic strategy combining clinical probability assessment, D-dimer measurement, venous ultrasonography, and helical CT is safe and effective in outpatients suspected of pulmonary embolism. However, this series was not designed to test the hypothesis that single-slice helical CT alone is safe for ruling out pulmonary embolism, and venous ultrasonography should imperatively be combined with CT to compensate for its limited sensitivity pending the results of ongoing studies of multidetector technology. ACKNOWLEDGMENT We gratefully acknowledge the skilled assistance of Catherine Ternisien, MD, Catherine Ridereau-Zins, MD, Alain Dauver, MD, Amel Benoit, MD, Alain Furber, MD, Philippe Geslin, MD, Gilles Berrut, MD, Thierry Urban, MD, and Alain Delhumeau, MD, in Angers; of Philippe de Moerloose, MD, Paul Cirafici, MD, Thierry Merminod, MD, Marc Righini, MD, Dominique Didier, MD, and Pascal Bachmann, in Geneva; and of all the physicians in the Departments of Emergency, Radiology, Angiology, and Nuclear Medicine in Angers, Lausanne, and Geneva. We would also like to thank Sylvain de Lucia, MD, in Geneva; Mathilde Marchand, MD, in Angers; and Dominique Gillis, MD, in Lausanne; for their help in data collection. Finally, we would like to thank all the residents who rotated in the emergency departments of the three centers for their invaluable collaboration throughout the study. REFERENCES 1. Perrier A, Desmarais S, Miron MJ, et al. Non-invasive diagnosis of venous thromboembolism in outpatients. Lancet. 1999;353: Wells PS, Anderson DR, Rodger M, et al. Excluding pulmonary embolism at the bedside without diagnostic imaging: management of patients with suspected pulmonary embolism presenting to the emergency department by using a simple clinical model and D- dimer. Ann Intern Med. 2001;135: Wells PS, Ginsberg JS, Anderson DR, et al. Use of a clinical model for safe management of patients with suspected pulmonary embolism. Ann Intern Med. 1998;129: Perrier A, Bounameaux H. Cost-effective diagnosis of deep vein thrombosis and pulmonary embolism. Thromb Haemost. 2001;86: March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume 116

13 Diagnosing Pulmonary Embolism in Outpatients/Perrier et al 5. Paterson DI, Schwartzman K. Strategies incorporating spiral CT for the diagnosis of acute pulmonary embolism: a cost-effectiveness analysis. Chest. 2001;119: Perrier A, Nendaz MR, Sarasin FP, Howarth N, Bounameaux H. Cost-effectiveness of diagnostic strategies for suspected pulmonary embolism including helical computed tomography. Am J Respir Crit Care Med. 2003;167: Stein PD, Athanasoulis C, Alavi A, et al. Complications and validity of pulmonary angiography in acute pulmonary embolism. Circulation. 1992;85: van Beek EJ, Brouwerst EM, Song B, Stein PD, Oudkerk M. Clinical validity of a normal pulmonary angiogram in patients with suspected pulmonary embolism a critical review. Clin Radiol. 2001; 56: Bates SM, Grand Maison A, Johnston M, Naguit I, Kovacs MJ, Ginsberg JS. A latex D-dimer reliably excludes venous thromboembolism. Arch Intern Med. 2001;161: Ginsberg JS, Wells PS, Kearon C, et al. Sensitivity and specificity of a rapid whole-blood assay for D-dimer in the diagnosis of pulmonary embolism. Ann Intern Med. 1998;129: Perrier A, Desmarais S, Goehring C, et al. D-dimer testing for suspected pulmonary embolism in outpatients. Am J Respir Crit Care Med. 1997;156: Kearon C, Ginsberg JS, Hirsh J. The role of venous ultrasonography in the diagnosis of suspected deep venous thrombosis and pulmonary embolism. Ann Intern Med. 1998;129: Turkstra F, Kuijer PMM, van Beek EJR, Brandjes DPM, ten Cate JW, Büller HR. Diagnostic utility of ultrasonography of leg veins in patients suspected of having pulmonary embolism. Ann Intern Med. 1997;126: Perrier A, Bounameaux H, Morabia A, et al. Diagnosis of pulmonary embolism by a decision analysis-based strategy including clinical probability, D-dimer levels, and ultrasonography: a management study. Arch Intern Med. 1996;156: Rathbun SW, Raskob GE, Whitsett TL. Sensitivity and specificity of helical computed tomography in the diagnosis of pulmonary embolism: a systematic review. Ann Intern Med. 2000;132: Perrier A, Howarth N, Didier D, et al. Performances of helical computed tomography in unselected outpatients with suspected pulmonary embolism. Ann Intern Med. 2001;135: The PIOPED Investigators. Value of the ventilation-perfusion scan in acute pulmonary embolism. JAMA. 1990;263: Wells PS, Anderson DR, Rodger M, et al. Derivation of a simple clinical model to categorize patients probability of pulmonary embolism: increasing the model s utility with the SimpliRED D- dimer. Thromb Haemost. 2000;83: Wicki J, Perneger TV, Junod A, Bounameaux H, Perrier A. Assessing clinical probability of pulmonary embolism in the emergency ward: a simple score. Arch Intern Med. 2001;161: Cockcroft DW, Gault MH. Prediction of creatinine clearance from serum creatinine. Nephron. 1976;16: Chagnon I, Bounameaux H, Aujesky D, et al. Comparison of two clinical prediction rules and implicit assessment for suspected pulmonary embolism. Am J Med. 2002;113: de Moerloose P, Desmarais S, Bounameaux H, et al. Contribution of a new, rapid, individual and quantitative automated D-dimer ELISA to exclude pulmonary embolism. Thromb Haemost. 1996;75: Lensing AWA, Prandoni P, Brandjes D, et al. Detection of deepvein thrombosis by real-time B-mode ultrasonography. N Engl J Med. 1989;320: Ghaye B, Szapiro D, Mastora I, et al. Peripheral pulmonary arteries: how far in the lung does multi-detector row spiral CT allow analysis? Radiology. 2001;219: Bounameaux H, Cirafici P, de Moerloose P, et al. Measurement of D-dimer in plasma as diagnostic aid in suspected pulmonary embolism. Lancet. 1991;337: Hull RD, Raskob GE, Ginsberg JS, et al. A noninvasive strategy for the treatment of patients with suspected pulmonary embolism. Arch Intern Med. 1994;154: Hull RD, Raskob GE, Coates G, Panju AA. Clinical validity of a normal perfusion lung scan in patients with suspected pulmonary embolism. Chest. 1990;97: Van Strijen MJ, De Monye W, Schiereck J, et al. Single-detector helical computed tomography as the primary diagnostic test in suspected pulmonary embolism: a multicenter clinical management study of 510 patients. Ann Intern Med. 2003;138: Musset D, Parent F, Meyer G, et al. Diagnostic strategy for patients with suspected pulmonary embolism: a prospective multicentre outcome study. Lancet. 2002;360: van Erkel AR, van den Hout WB, Pattynama PM. International differences in health care costs in Europe and the United States: do these affect the cost-effectiveness of diagnostic strategies for pulmonary embolism? Eur Radiol. 1999;9: March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume

14 A Multivariate Model for Predicting Mortality in Patients with Heart Failure and Systolic Dysfunction James M. Brophy, MD, PhD, Gilles R. Dagenais, MD, Frances McSherry, MS, William Williford, PhD, Salim Yusuf, DPhil BACKGROUND: Heart failure is a leading cause of morbidity and mortality, but there are no reliable models based on readily available clinical variables to predict outcomes in patients taking angiotensin-converting enzyme (ACE) inhibitors. METHODS: A multivariate statistical model to predict mortality was developed in a random sample (n 4277 patients [67%]) of the 6422 patients enrolled in the Digitalis Investigation Group trial who had a depressed ejection fraction ( 45%), were in sinus rhythm, and were taking ACE inhibitors. The model was then validated in the remaining 2145 patients. RESULTS: Total mortality in the derivation sample was 11.2% (n 480) at 12 months and 29.9% (n 1277) at 36 months. Lower ejection fraction, worse renal function, cardiomegaly, worse functional class, signs or symptoms of heart failure, lower blood pressure, and lower body mass index were associated with reduced 12-month survival. This model provided good predictions of mortality in the verification sample. The same variables, along with age and the baseline use of nitrates, were also predictive of 36-month mortality. CONCLUSION: Routine clinical variables can be used to predict short- and long-term mortality in patients with heart failure and systolic dysfunction who are treated with ACE inhibitors. Am J Med. 2004;116: by Excerpta Medica Inc. Heart failure is a leading cause of morbidity and mortality, and its incidence and prevalence are increasing (1 3). Angiotensin-converting enzyme (ACE) inhibitors improve outcomes in these patients (4) and are recommended as first-line treatment (5). However, the clinical correlates of unfavorable outcomes among patients with heart failure who are treated with ACE inhibitors have not been well defined. The development of risk models to predict outcomes and select treatments has been recognized as a research priority (6). The Digitalis Group study (7,8) was the largest heart failure trial, and data collected in the study can be used to identify both short- and long-term determinants of heart failure mortality and morbidity. Almost all patients in this trial which included substantial numbers of women and the elderly were treated with ACE inhibitors. From this database, we built predictive models for short- and long-term mortality for patients with heart failure and depressed systolic function. From the Division of Cardiology and Clinical Epidemiology (JMB), McGill University Health Center, Montréal, Quebec, Canada; Institut Universitaire de Cardiologie et Pneumologie de Laval (GRD), Ste. Foy, Quebec, Canada; VA Medical Center (FM, WW), Perry Point, Maryland; and Division of Cardiology (SY), McMaster University, Hamilton, Ontario, Canada. Dr. Brophy is a funded researcher from Le Fonds de la Recherche en Santé du Québec. Dr. Yusuf holds a Senior Scientist Award of the Canadian Institutes of Health Research and is an Endowed Chair of the Heart and Stroke Foundation of Ontario. Requests for reprints should be addressed to James Brophy, MD, PhD, Division of Cardiology, McGill University Health Center, Royal Victoria Hospital, 687 Pine Street West, Montréal, Quebec H3A 1A1, Canada, or jbroph@po-box.mcgill.ca. Manuscript submitted November 28, 2001, and accepted in revised form September 5, METHODS Study Sample The trial recruited 7788 patients from February 1991 through August 1993 in over 300 North American centers. The design and principal outcomes of this trial, which compared digoxin with placebo in patients with heart failure and normal sinus rhythm, have been published (7,8). The diagnosis of heart failure was based on current or past symptoms and signs, or radiological evidence of pulmonary congestion. Patients were excluded from the trial if they had a serum creatinine level 3.0 mg/dl, age 21 years, unstable coronary syndromes, cor pulmonale, complex congenital heart diseases, or recognizable noncardiac causes of heart failure. This analysis considers only the 6800 patients with sinus rhythm and left ventricular ejection fractions 0.45%, of whom 6422 were being treated with ACE inhibitors. Baseline data were limited to historical, physical examination (recorded by experienced physicians), and routine laboratory results. The ejection fraction was measured using angiographic, radionuclide, or echocardiographic techniques as decided by individual investigators. To facilitate the interpretation of the final results, some continuous variables, including age, have been categorized. Signs and symptoms of heart failure (including rales, elevated jugular venous pressure, peripheral edema, dyspnea at rest, exertional dyspnea, limitation of activity, S 3 gallop, and radiologic evidence of pulmonary congestion) were collated into a single categorical variable (levels 0/1, 2/3, 4/5, 6 or greater). Patients had regular visits at 4-month intervals. The mean duration of follow-up was 37 months (range, 28 to 58 months). Mortality was determined by chart review or by Excerpta Medica Inc /04/$ see front matter All rights reserved. doi: /j.amjmed

15 Predicting Mortality in Heart Failure/Brophy et al family interviews. Vital status was unknown for only 1.4% of patients; even for these patients, information was available until the time of censoring. Model Selection and Validation Our objective was to determine the clinical correlates of total mortality at 12 and 36 months. Univariate analysis was performed on a total of 28 baseline variables, and treatment allocation (digoxin vs. placebo). The following 18 variables were then included in multivariate models: sex, race, age group, angina, diabetes, etiology, cardiothoracic ratio, duration of heart failure, New York Heart Association (NYHA) functional class, left ventricular ejection fraction, systolic and diastolic blood pressure, body mass index, serum creatinine level, nitrate use, heart rate, treatment assignment, and a composite variable of clinical signs and symptoms. Ten prespecified interaction terms involving age, sex, diabetes, etiology, renal function, and digoxin treatment (age * etiology, age * treatment, sex * etiology, sex * creatinine, sex * treatment, diabetes * etiology, diabetes * creatinine, diabetes * treatment, treatment * etiology, treatment * creatinine) were also entered into the models. Because stepwise approaches often have difficulty distinguishing between competing models, ignore the uncertainty involved in multivariate model selection, and may overfit models, we used a Bayesian approach that averaged over the set of best models according to posterior model probability. This was performed with the Bayesian Information Criterion (9) using S-Plus (Insightful Corporation, Seattle, Washington). This technique was applied to a random two-third selection of the cohort (derivation sample). This model selection process has optimal properties for predicting events in future patients with similar characteristics. Nevertheless, we validated the model in the remaining one third of the cohort. For each patient in this validation set, we calculated the probability of survival using Cox proportional hazards models (SAS Institute, Cary, North Carolina) with the variables identified from the derivation set. We ranked the predicted values into deciles of mortality, and then compared the observed and predicted mortality for each decile. RESULTS Patients in the derivation and validation samples had similar characteristics (Table 1). The mean ( SD) age in the derivation sample was years. Over half were known to have heart failure for at least 12 months. Virtually all had signs and symptoms of heart failure at randomization, and only a minority was in NYHA functional class I. The mean ejection fraction was 28% 9%, and the mean creatinine level was mg/dl. Multivariate Correlations with Outcomes Total mortality in the derivation sample was 11.2% (n 480) at 12 months and 29.9% (n 1277) at 36 months. Increasing age, worse functional class, and more signs or symptoms of heart failure were independently associated with increased 36-month mortality (Table 2), as were lower body mass index and systolic blood pressure. Lower ejection fraction, worse renal function, cardiomegaly, and use of nitrates were also associated with greater long-term mortality. Patients with diabetes and an ischemic etiology for their heart failure (n 914) were also at increased risk. In general, the same variables were predictive of short-term mortality, except for age and nitrate use; diastolic blood pressure replaced systolic blood pressure. As had been reported, digoxin treatment was not associated with short- or long-term mortality. Model Validation Models based on these variables were able to predict observed survival across different risk strata in the validation sample (Figures 1 and 2). For example, the lowest risk decile had a predicted 12-month mortality of 4.0% (vs. an observed mortality of 3.8%), which increased to 11.8% (vs. an observed mortality of 10.7%) at 36 months. In general, the models performed less well for the highest decile, although even here the results were reasonable (e.g., an observed mortality of 30.5% vs. a predicted mortality of 35.1% at 12 months). An example is provided in the Appendix. DISCUSSION More than 50 demographic, clinical, biochemical, hemodynamic, electrophysiologic, and echocardiographic factors have been correlated with outcomes in heart failure patients; neurohormonal markers have perhaps the strongest association (10). However, most patients are diagnosed, treated, and followed in general medical practices and do not always have access to these markers. Therefore, we concentrated on predictors that are readily available, and demonstrated that they can be used to predict short- and long-term mortality across a wide spectrum of disease severity. The variables have good face validity and have generally been shown to be associated with adverse outcomes in other studies. Our model confirms the prognostic value of age (11), impaired left ventricular function (12), cardiomegaly (12), and renal insufficiency (13), which have been identified as predictors of adverse outcome in smaller studies that often antedated the systematic use of ACE inhibitors. Although there is a very weak correlation between cardiothoracic ratio and ejection fraction in this sample (14),we found that both cardiac size and function were independent predictors of mortality. Signs and symptoms of heart failure were common at March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume

16 Predicting Mortality in Heart Failure/Brophy et al Table 1. Baseline Characteristics of the Patients in the Derivation and Validation Samples, and among Patients in the Derivation Sample Who Died during 36-Month Follow-up Derivation Sample (n 4277) Validation Sample (n 2145) Died during 36-Month Follow-up (n 1277) Characteristic Number (%) or Mean SD Male sex 3319 (77.6) 1670 (77.9) 1016 (79.6)* Race White 3637 (85.0) 1845 (86.0) 1086 (85.1) Black 521 (12.2) 232 (10.8) 160 (12.5) Other 119 (2.8) 68 (3.2) 31 (2.4) Age (years) (11.9) 225 (10.5) 103 (8.1)* 50 to (21.0) 443 (20.7) 198 (15.5) 60 to (36.5) 801 (37.3) 478 (37.4) (30.5) 676 (31.5) 498 (39.0) Angina 1126 (26.3) 562 (26.2) 365 (28.6)* Diabetes 1214 (28.4) 621 (29.0) 438 (34.3)* Hypertension 1954 (45.7) 968 (45.1) 589 (46.1) Ischemic etiology 2990 (70.1) 1526 (71.3) 902 (70.9) Nitrate use 1815 (42.4) 903 (42.1) 636 (49.8)* NYHA class I 582 (13.6) 272 (12.7) 117 (9.2)* II 2271 (53.2) 1195 (55.8) 584 (45.8) III/IV 1420 (33.2) 676 (31.5) 575 (45.0) Duration of heart failure (months) (12.4) 247 (11.6) 160 (12.6)* 3 to (34.1) 701 (32.7) 392 (30.8) 13 to (15.6) 358 (16.7) 188 (14.8) (37.9) 835 (39.0) 533 (41.8) Digoxin treatment 2145 (50.2) 1052 (49.0) 651 (51.0) Physical examination Body mass index (kg/m 2 ) (25.5) 534 (24.9) 400 (31.3)* 23.7 to (25.0) 513 (23.9) 314 (24.6) 26.5 to (24.9) 533 (24.9) 296 (23.2) (24.6) 565 (26.3) 267 (20.9) Systolic blood pressure (mm Hg) (29.4) 609 (28.4) 445 (34.8)* 111 to (19.9) 416 (19.4) 264 (20.7) 122 to (23.6) 515 (24.0) 271 (21.2) (27.1) 604 (28.2) 297 (23.3) Diastolic blood pressure (mm Hg) (24.5) 522 (24.4) 374 (29.3)* 70 to (25.7) 549 (25.6) 340 (26.6) 75 to (27.9) 607 (28.3) 330 (25.8) (21.9) 465 (21.7) 233 (18.3) No. of signs or symptoms of heart failure 0/1 136 (3.2) 67 (3.1) 25 (2.0)* 2/3 656 (15.4) 351 (16.4) 137 (10.8) 4/ (24.9) 534 (24.9) 278 (21.8) (55.5) 1191 (55.6) 833 (65.4) Cardiothoracic ratio (61.6) 1329 (62.0) 898 (70.3)* Ejection fraction (%) * Serum creatinine (mg/dl) * Heart rate (beats per minute) * * P 0.05 by comparison with those who lived. Represents the baseline reference group for the multivariate analysis presented in Table 2. Rales, elevated jugular venous pressure, peripheral edema, dyspnea at rest or on exertion, limitation of activity, S 3 gallop, and radiologic evidence of pulmonary congestion. NYHA New York Heart Association. 302 March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume 116

17 Predicting Mortality in Heart Failure/Brophy et al Table 2. Independent Predictors of Short- and Long-term Mortality Variable* Mortality at 12 Months Mortality at 36 Months Hazard Ratio (95% Confidence Interval) Age 1.20 ( ) Ejection fraction (per 10% decrease) 1.34 ( ) 1.34 ( ) NYHA class 1.44 ( ) 1.29 ( ) Cardiothoracic ratio 50% 1.60 ( ) 1.34 ( ) Clinical signs or symptoms 1.21 ( ) 1.14 ( ) Serum creatinine (per mg/dl increase) 1.85 ( ) 1.73 ( ) Body mass index 1.18 ( ) 1.09 ( ) Diastolic blood pressure 1.17 ( ) Systolic blood pressure 1.11 ( ) Nitrates 1.18 ( ) Diabetes* ischemic etiology 1.46 ( ) 1.43 ( ) * See Table 1 for the reference groups; hazard ratios are per one-group increase for age, NYHA class, signs or symptoms, and per one-group decrease for body mass index and blood pressure. Applicable for baseline ejection fractions between 3% and 45%. Applicable for baseline creatinine values between 0.3 and 3.0 mg/dl. NYHA New York Heart Association. baseline; when combined into a single score, they were strongly associated with mortality. For example, six or more signs or symptoms increased mortality by 50% at 36 months compared with patients who had one or no sign or symptom. These findings are in general agreement with a recent analysis (15), which found that an elevated jugular venous pressure and third heart sound were associated with repeat hospitalizations, but had a borderline significant association with total mortality. Other investigators have found that an ischemic etiology for heart failure is associated with decreased survival (16), but this has not been a consistent finding (17). We found a strong interaction between etiology and diabetes: an ischemic etiology was associated with reduced survival among patients with diabetes, whereas etiology was not associated with mortality in patients without diabetes. This confirms previous findings (18). Finally, nitrate use was a predictor of long-term, but not short-term, mortality, perhaps because it is a proxy for more severe heart failure that requires two vasodilators or because it indicates more severe angina. Our predictive models have several strengths. They are based on a large, relatively unselected, and contemporary cohort of patients being treated with ACE inhibitors, including a substantial proportion of women. The large number of deaths permitted us to estimate the effects of several clinical variables simultaneously. One potential limitation of this study is that all of the subjects were participating in a clinical trial and it is unknown how representative they are of those in routine clinical practice. However, it seems unlikely that there would be substantial bias in patient selection in a trial of almost 7000 patients from 300 clinical sites. Our results may not apply to patients who are not in normal sinus rhythm, or who have severe renal insufficiency. In addition, we did not have neurohormonal, echocardiographic, and hemodynamic data. Our goal, however, was to develop a model that is clinically relevant by including variables that are available to all physicians who care for heart failure patients. Furthermore, the incremental value of these more sophisticated measurements beyond what can be estimated from clinical variables has not been confirmed. Figure 1. Observed and predicted survival at 12 months. P 0.14 in a goodness-of-fit test. Figure 2. Observed and predicted survival at 36 months. P 0.01 in a goodness-of-fit test. March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume

18 Predicting Mortality in Heart Failure/Brophy et al Prognosis in heart failure is a function of both patient and physician characteristics (19), but we did not have detailed information about the physicians who participated in this trial. We also did not have information on the use of beta-blocker or spironolactone therapy. In conclusion, we have shown that it is possible to risk stratify patients with heart failure using easily available clinical measurements. High-risk patients may benefit from more intensive multidisciplinary follow-up (20). Predictive models may also facilitate the interpretation of studies of quality of care. The identification of new prognostic variables should evaluate their incremental predictive value beyond that supplied by clinical models. This model should also be evaluated among patients being treated with beta-blockers, which have become an essential part of treatment (21) since this study was completed. REFERENCES 1. American Heart Association. Heart Disease and Stroke Statistics 2003 Update. Dallas, Texas: American Heart Association. Available at: Accessed October 27, Gheorghiade M, Bonow RO. Chronic heart failure in the United States: a manifestation of coronary artery disease. Circulation. 1998; 97: Graves EJ. National Hospital Discharge Survey: Annual Summary, National Health Survey. 1995;121:1 63. Vital and Health Statistics Series, No Garg R, Yusuf S, for the Collaborative Group on ACE Inhibitor Trials. Overview of randomized trials of angiotensin-converting enzyme inhibitors on mortality and morbidity in patients with heart failure. JAMA. 1995;273: ACC/AHA Task Force on Practice Guidelines. Evaluation and management of heart failure. J Am Coll Cardiol. 1995;26: Krumholz HM, Baker DW, Ashton CM, et al. Evaluating quality of care for patients with heart failure. Circulation. 2000;101:e122 e The Digitalis Investigation Group. Rationale, design, implementation, and baseline characteristics of patients in the DIG trial: a large, simple, long-term trial to evaluate the effect of digitalis on mortality in heart failure. Control Clin Trials. 1996;17: The Digitalis Investigation Group. The effect of digoxin on mortality and morbidity in patients with heart failure. The Digitalis Investigation Group. N Engl J Med. 1997;336: Raftery AE. Bayesian model selection in social research (with discussion by Andrew Gelman, Donald B. Rubin, and Robert M. Hauser). In: Marsden PV, ed. Sociological Methodology. Oxford, United Kingdom: Blackwell; 1995: Eichhorn EJ. Prognosis determination in heart failure. Am J Med. 2001;110(suppl 7A):14S 36S. 11. Ho KK, Anderson KM, Kannel WB, Grossman W, Levy D. Survival after the onset of congestive heart failure in Framingham Heart Study subjects. Circulation. 1993;88: Cohn JN. Prognosis in congestive heart failure. J Card Fail. 1996; 2(suppl):S225 S Dries DL, Exner DV, Domanski MJ, Greenberg B, Stevenson LW. The prognostic implications of renal insufficiency in asymptomatic and symptomatic patients with left ventricular systolic dysfunction. J Am Coll Cardiol. 2000;35: Philbin EF, Garg R, Danisa K, et al. The relationship between cardiothoracic ratio and left ventricular ejection fraction in congestive heart failure. Digitalis Investigation Group. Arch Intern Med. 1998; 158: Drazner MH, Rame JE, Stevenson LW, Dries DL. Prognostic importance of elevated jugular venous pressure and a third heart sound in patients with heart failure. N Engl J Med. 2001;345: Bart BA, Shaw LK, McCants CB Jr, et al. Clinical determinants of mortality in patients with angiographically diagnosed ischemic or nonischemic cardiomyopathy. J Am Coll Cardiol. 1997;30: Gavazzi A, Berzuini C, Campana C, et al. Value of right ventricular ejection fraction in predicting short-term prognosis of patients with severe chronic heart failure. J Heart Lung Transplant. 1997;16: Dries DL, Sweitzer NK, Drazner MH, Stevenson LW, Gersh BJ. Prognostic impact of diabetes mellitus in patients with heart failure according to the etiology of left ventricular systolic dysfunction. J Am Coll Cardiol. 2001;38: Auerbach AD, Hamel MB, Davis RB, et al. Resource use and survival of patients hospitalized with congestive heart failure: differences in care by specialty of the attending physician. SUPPORT Investigators. Study to Understand Prognoses and Preferences for Outcomes and Risks of Treatments. Ann Intern Med. 2000;132: McAlister FA, Lawson FM, Teo KK, Armstrong PW. A systematic review of randomized trials of disease management programs in heart failure. Am J Med. 2001;110: Brophy JM, Joseph L, Rouleau JL. Beta-blockers in congestive heart failure. A Bayesian meta-analysis. Ann Intern Med. 2001;134: APPENDIX The following example shows how to estimate 36-month mortality. Consider a 74-year-old nondiabetic patient, with an ischemic cardiomyopathy, ejection fraction (EF) of 25%, New York Heart Association (NYHA) class III, a creatinine (CR) level of 1.5 mg/dl, a cardiothoracic (CT) ratio 0.5, and a clinical (CLIN) score of 4 (rales, increased jugular venous pressure, rest and exertional dyspnea). The patient is not taking nitrates, and is in the reference group for body mass index and blood pressure. This results in a risk sum as follows: AGE *AGE ln RR AGE *age group 0.182* EF *EF ln RR EF *EF 0.30* / NYHA *NYHA ln RR NHYA *NYHA 0.255* CT *CT ln RR CT *CT 0.293* cong *CLIN ln RR CLIN *CLIN 0.131* Cr *Cr ln RR CR *Cr 0.552* Patient sum 3.04 The average 36-month average survival, S o (t), is 70.1% and the mean sum is Then the patient s predicted 36-month survival is exp( ) or 72%. Survival at 12 months can be estimated based on the overall average 12-month survival of 88.8% and mean sum of March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume 116

19 Heart Failure in Rheumatoid Arthritis: Rates, Predictors, and the Effect of Anti Tumor Necrosis Factor Therapy Frederick Wolfe, MD, Kaleb Michaud, MS PURPOSE: We sought to determine the frequency of heart failure in patients with rheumatoid arthritis, and to determine its predictors, particularly the use of anti tumor necrosis factor (TNF) therapy. METHODS: Rheumatoid arthritis (n 13,171) and osteoarthritis (n 2568) patients were studied during a 2-year period ending in June The diagnosis of heart failure was based on self-report or review of medical records. Propensity scores were used to adjust for the risk of anti-tnf (infliximab and etanercept) prescription. RESULTS: Heart failure was more common among patients with rheumatoid arthritis (3.9% [n 461]) than in those with osteoarthritis (2.3% [n 87]), after adjusting for differences in demographic characteristics. Patients with rheumatoid arthritis had similar risk factors for heart failure (e.g., hypertension, prior myocardial infarction, diabetes, advanced age) as persons in population-based studies. Heart failure was significantly (P 0.05) less common in anti-tnf treated patients (3.1% [180/ 5832]) than in the remaining patients (3.8% [281/7339]), even after adjusting for baseline differences. In the absence of preexisting cardiovascular disease, the risk of heart failure was low (0.4% [24/6251]) and was not related to anti-tnf therapy. CONCLUSION: Our results suggest that rheumatoid arthritis increases the risk of heart failure, which may be ameliorated by anti-tnf therapies. Am J Med. 2004;116: by Excerpta Medica Inc. From the Arthritis Research Center Foundation (FW, KM) and the University of Kansas School of Medicine (FW), Wichita, Kansas. The infliximab registry is support by a grant from Centocor, Inc., Malvern, Pennsylvania. Requests for reprints should be addressed to Frederick Wolfe, MD, National Data Bank for Rheumatic Diseases, Arthritis Research Center Foundation, 1035 N. Emporia, Suite 230, Wichita, Kansas 67214, or fwolfe@arthritis-research.org. Manuscript submitted January 14, 2003, and accepted in revised form September 9, There is now substantial evidence that rheumatoid arthritis is associated with increased cardiovascular morbidity and mortality (1 12). Cardiovascular manifestations of rheumatoid arthritis include pericarditis, myocardial infarction, and heart failure (13), perhaps including diastolic dysfunction (14). Although the risk of myocardial infarction appears to be increased, little is known about the risk of heart failure in rheumatoid arthritis. Heart failure is of special interest because the failing heart produces tumor necrosis factor (TNF), but the normal heart does not (15). Although the effects of circulating TNF- on cardiovascular function are uncertain (15 18), data from murine heart failure models support the theory that blockade of circulating TNF may ameliorate ventricular dysfunction (19,20). However, clinical trials of TNF blockade in patients with advanced heart failure have shown little benefit or even harm (17,21,22). It is not known whether anti-tnf therapy, which is being used increasingly in the treatment of rheumatoid arthritis, affects the risk of heart failure in these patients. The purpose of this report was to determine the prevalence of heart failure in patients with rheumatoid arthritis, in comparison with patients with osteoarthritis; to determine the factors associated with heart failure; and to study the effects of anti-tnf therapy on the risk of heart failure. METHODS Subjects were participants in the National Data Bank for Rheumatic Diseases study of the outcomes of arthritis. Patients are recruited for this ongoing study from the practices of U.S. rheumatologists (23 25), and are followed with semiannual questionnaires. Approximately 8% of patients decline to participate per year. This report includes 13,171 rheumatoid arthritis patients (including 3862 who were enrolled as part of an infliximab safety registry) who completed 35,064 biannual questionnaires during consecutive 6-month assessment periods ending in June In addition, data from 2568 patients with osteoarthritis of the hip or knee who were enrolled similarly were also analyzed for comparison. At each assessment, demographic and clinic variables are obtained (26 35). Patients report all comorbid conditions, medications, and side effects of treatment. We obtained a history of pre-existing cardiovascular illness, current and previous hypertension, myocardial infarction, and other cardiovascular conditions. We asked specifically, During the last 6 months were you diagnosed or treated for heart failure? We also requested and reviewed records for all hospitalizations. Patients reporting heart failure were interviewed by the research staff using a standardized, written protocol. A diagnosis of heart failure was considered valid if the patient provided data indicat by Excerpta Medica Inc /04/$ see front matter 305 All rights reserved. doi: /j.amjmed

20 Heart Failure in Rheumatoid Arthritis/Wolfe and Michaud Table 1. Characteristics of the Patients with Rheumatoid Arthritis (N 13,171) Characteristic Number (%) or Mean SD Age (years) Male sex 3015 (23) Married 9039 (69) Nonhispanic white 12,308 (94) At least high school graduate 11,712 (89) Income ($) 46,000 28,800 Disease duration (years) Lifetime comorbidity score (scale, 0 to 11) Body mass index (kg/m 2 ) Use of cardiovascular medication(s) 6797 (52) History of hypertension 6190 (47) Any cardiovascular history 6849 (52) Treatment Methotrexate 7386 (56) Prednisone 5160 (39) Infliximab 4307 (33) Hydroxychloroquine 2590 (20) Leflunomide 2316 (18) Etanercept 1680 (13) Sulfasalazine 761 (6) No DMARD or anti-tnf agent* 4307 (14) Severity and status (scale) Health Assessment Questionnaire disability index (0 3) Rheumatoid Arthritis Disease Activity Index (0 10) Pain (0 10) Global severity (0 10) Fatigue (0 10) Depression (0 10) SF-36 physical component score (0 100) SF-6 utility index (0 1) EuroQol-5D utility (0 1) * DMARDs include methotrexate, sulfasalazine, hydroxychloroquine, leflunomide, oral and injectable gold, azathioprine, and cyclosporine. DMARD disease-modifying antirheumatic drug; SF short form; TNF tumor necrosis factor. ing that a physician diagnosed heart failure, or if the diagnosis was supported by medical records or physician contact. Reports of heart failure were found to be valid in more than 90% of cases. Incident cases of heart failure were defined as those occurring in persons without a history of cardiovascular disease, including the use of cardiovascular medications. Statistical Analyses Each study observation relates to events that occurred in the previous 6-month period; patients had between one and four such observations. Of these, one observation was chosen for study. For patients with heart failure, we selected the first observation when heart failure was present. For patients without heart failure, a random observation was chosen. Propensity scores were used to adjust for nonrandom assignment to treatment (36 38). Variables used in constructing the propensity score were Health Assessment Figure. Frequency of heart failure in 13,171 rheumatoid arthritis patients stratified by age and sex. Bars represent 95% confidence intervals. Numbers for each age group represent the numerator/ denominator by sex. Rates increase with age (P 0.001) and are always greater in men than in women (P 0.001). 306 March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume 116

21 Heart Failure in Rheumatoid Arthritis/Wolfe and Michaud Table 2. Univariate Correlates of Heart Failure in Rheumatoid Arthritis Patients Variable All Heart Failure (n 461 Cases) Incident Cases of Heart Failure Only (n 42 Cases) Odds Ratio (95% Confidence Interval) Hypertension (ever) 2.6 ( ) Myocardial infarction (last 6 months) 16.1 ( ) Myocardial infarction (ever) 6.6 ( ) Diabetes 3.4 ( ) 2.1 ( ) Body mass index per 10-unit increase 1.3 ( ) 1.6 ( ) Body mass index 25 kg/m ( ) 1.4 ( ) Male sex 1.8 ( ) 1.7 ( ) Current or past smoker 1.6 ( ) 1.8 ( ) Disability index per unit increase (lagged) 2.1 ( ) 2.8 ( ) Disability index 0 (lagged) 4.2 ( ) NA* Age (per 10-year increase) 1.8 ( ) 1.6 ( ) Education High school graduate Reference Reference 0 8 years 1.7 ( ) 1.5 ( ) 8 11 years 1.6 ( ) 1.5 ( ) years 0.7 ( ) 0.4 ( ) 16 years 0.7 ( ) 0.5 ( ) * All incident cases of heart failure had a disability index 0 at previous observation. Questionnaire disability index, pain, global severity, prednisone use, age, age squared, and sex. Two sets of potential covariates were used. In one analytic model, variables were lagged from data of the previous observation for each patient; thus, the covariates were measured before the heart failure outcome (or the random observation in those without heart failure). For approximately one third of patients, lagged variables were not available, and variables from the current observation were used based on the assumption that disability index, pain, and global severity had similar effects regardless of whether they came before or after the event. However, to assess the validity of this assumption, analyses were also conducted after excluding patients without lagged variables. Because results were very similar regardless of method, we report results from the first model including all participants. Adjustment to the demographic characteristics of the general population was made using Monte Carlo simulations with 1000 repetitions (39). This method was also used to adjust the frequency of heart failure in osteoarthritis patients. All statistical analyses were performed using Stata (College Station, Texas), version 7.0 (40). This study was approved by the Via Christi Regional Medical Center Institutional Review Board, Wichita, Kansas. RESULTS The majority of the patients with rheumatoid arthritis were white women (Table 1); about half had a history of cardiovascular disease. About one third were treated with infliximab, reflecting the inclusion of patients in the infliximab safety registry. By comparison, patients with osteoarthritis were older (mean [ SD] age, years), less likely to be male (18% [n 482]), and slightly more likely to have graduated from high school (92% [n 2350]) than were those with rheumatoid arthritis. There were 461 cases of heart failure among the 13,171 patients with rheumatoid arthritis, including 42 incident cases of which 24 occurred in patients without previous cardiovascular disease history. The frequency of heart failure in the rheumatoid arthritis cohort was 5.2% in men and 3.0% in women, for an overall risk of 3.5%. Increased age was associated with a greater frequency of heart failure (Figure). Of the 2568 patients with osteoarthritis, 87 (3.4%) had heart failure. After adjusting for demographic characteristics, the risk of heart failure was 3.9% (95% confidence interval [CI]: 3.4% to 4.3%) in patients with rheumatoid arthritis compared with 2.3% (95% CI: 1.6% to 3.3%) in those with osteoarthritis. In general, the factors associated with the presence of heart failure among patients with rheumatoid arthritis were similar to those in the general population (Table 2). Several measures of disease, including the disability index, pain, and global severity, were also associated with the prevalence and incidence of heart failure (Table 2). Effects of Anti-TNF Therapy There were significant differences between nearly all of the characteristics of patients who were and were not March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume

22 Heart Failure in Rheumatoid Arthritis/Wolfe and Michaud Table 3. Characteristics of Patients by Type of Treatment Variable Anti-TNF* (N 5832) Infliximab (N 4152) Etanercept (N 1525) No Anti-TNF (N 7339) P Value Number (%) or Mean SD Age (years) Male sex 1283 (22) 955 (23) 310 (20) 1725 (24) 0.06 Nonhispanic white 5540 (95) 3998 (96) 1408 (92) 6759 (92) Married 3966 (68) 2703 (65) 1127 (74) 5101 (70) 0.01 High school graduate 5097 (87) 3558 (85) 1400 (92) 6605 (90) Total income ($) 46,800 29,100 43,300 28,100 53,000 29,800 45,500 28, Disease duration (years) Lifetime comorbidity score (0 11) Health Assessment Questionnaire disability index (0 3) Rheumatoid Arthritis Disease Activity Index (0 10) Pain (0 10) Global severity (0 10) Fatigue (0 10) Depression (0 10) Physical component score (SF-36) SF-6 utility index (0 1) EuroQol-5D utility index (0 1) Treatment Methotrexate 3907 (67) 3135 (76) 669 (44) 3479 (47) Prednisone 2729 (47) 2151 (49) 593 (39) 2429 (33) Hydroxychloroquine 816 (14) 561 (14) 239 (16) 1769 (24) Leflunomide 793 (14) 544 (13) 223 (15) 1519 (21) Sulfasalazine 280 (5) 195 (5) 79 (5) 84 (7) Etanercept 1680 (29) (100) 0 Azathioprine 152 (3) 112 (3) 29 (2) 132 (2) Injectable gold 58 (1) 42 (1) 15 (1) 161 (2) Minocycline 58 (1) 33 (1) 23 (2) 139 (2) Anakinra 41 (1) 25 (1) 11 (1) 44 (1) 0.52 Cyclosporine 17 (0) 17 (0) 0 29 (0) 0.22 Auranofin 6 (0) 4 (0) 3 (0) 15 (0) Penicillamine 6 (0) 8 (0) 0 37 (1) 0.27 Cyclophosphamide (0) 0.20 * Patients that received both etanercept and infliximab are included here. Comparing anti-tnf with no anti-tnf. SF Short Form; TNF tumor necrosis factor. treated with anti-tnf therapy (Table 3). There were also differences between patients treated with infliximab versus etanercept (Table 3), suggesting that infliximabtreated patients had worse clinical status and slightly lower socioeconomic status. Among all cases of heart failure (both prevalent and incident), patients receiving anti-tnf therapy were less likely to have heart failure than were those not receiving anti-tnf treatment. These differences became even greater after covariate adjustment: the frequency of heart failure was 1.2% lower in anti-tnf treated patients in adjusted analyses. There were no significant differences in the risk of incident heart failure by use of anti-tnf therapy, although the numbers of cases were small (Table 4). Of patients receiving infliximab, 47.8% (2057/4307) had a history of cardiovascular disease compared with 54.4% (4820/8864) of those not receiving infliximab. We therefore performed sensitivity analyses by adding prior cardiovascular history to the models. For all cases of heart failure, the adjusted frequency of heart failure was 2.8% in those treated with anti-tnf therapies, versus 3.9% in the remaining patients (P 0.03). Among incident cases of heart failure, the risks were 0.2% in both groups (P 0.68). When limited to observations prior to the U.S. Food and Drug Administration s warning of increased 308 March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume 116

23 Heart Failure in Rheumatoid Arthritis/Wolfe and Michaud Table 4. Adjusted and Unadjusted Rates of Heart Failure by Treatment* Outcome All Anti-TNF (n 5832) Infliximab (n 4152) Percentage (N) Etanercept (n 1525) No Anti-TNF (n 7339) All heart failure Unadjusted 3.1 (180) 3.2 (134) 2.6 (40) 3.8 (281) Adjusted to 3.9 Difference from no anti-tnf 1.2% ( 1.9% to 0.5%) 1.4% ( 2.2% to 0.6%) 0.5% ( 1.5% to 0.4%) (95% confidence interval) Incident heart failure (n 3003) (n 2162) (n 768) (n 3248) Unadjusted 0.4 (12) 0.5 (10) 0.3 (2) 0.4 (12) Adjusted to 0.3 Difference from no anti-tnf (95% confidence interval) 0.1% ( 0.1% to 0%) 0.1% ( 0.1% to 0%) 0% (0% to 0.1%) * All heart failure refers to heart failure occurring in patients with or without a history of previous cardiovascular disease, including use of cardiovascular medications. Incident heart failure refers to heart failure occurring in persons without any history of cardiovascular disease, including use of cardiovascular medications. Adjusted rates were based on use of a propensity score. Includes patients that took both infliximab and etanercept Adjusted rates vary because propensity scores varied in the three comparison groups. TNF tumor necrosis factor. heart failure with anti-tnf use, the adjusted frequency of heart failure was 3.6% (n 1555) in those treated with anti-tnf therapies versus 4.3% (n 4034) in the remaining patients (P 0.16). We also examined incident cases of heart failure occurring in patients under the age of 50 years. No increase was found (0/1569 patients using anti-tnf therapy vs. 3/1401 not using anti-tnf therapy). DISCUSSION Our results suggest that at least as compared with patients who had osteoarthritis the prevalence of heart failure is increased among patients with rheumatoid arthritis. However, we found no increase in heart failure among patients receiving anti-tnf therapy. Because of recent concerns about the risk of anti-tnf therapy among patients without a history of cardiovascular disease (41), we also looked at patients 50 years of age and found no evidence of harmful effects. The overall increase in heart failure among rheumatoid arthritis patients may be related to increased inflammatory activity, perhaps leading to premature arteriosclerosis. Potential explanations include endothelial dysfunction and injury (42 44), lipid abnormalities and atherogenic lipoprotein factors (45 47), adhesion molecules (48), and proinflammatory cytokines, including interleukin 1 and TNF (17,18,49). Indeed, clinical markers of inflammation have been associated with cardiovascular mortality and morbidity in these patients (50). There is growing recognition that TNF- is involved in the inflammatory process in rheumatoid arthritis and cardiac disease, and that anti-tnf therapy may be effective (15,16,18,42). Several trials have evaluated the effects of anti-tnf therapy in patients with severe heart failure. However, the Research Into Etanercept: Cytokine Antagonism in Ventricular Dysfunction (RECOVER) (N 900), Randomized Etanercept North American Strategy to Study Antagonism of Cytokines (RENAISSANCE) (N 900), and Randomized Etanercept World Wide Evaluation (RE- NEWAL) (N 1500) trials of etanercept as treatment for heart failure were stopped due to the lack of efficacy (21,22,51). In addition, there was a nonstatistically significant increase in mortality among patients with heart failure in the Anti-TNF Alpha Therapy Against CHF (ATTACH) (N 150) trial who received infliximab therapy (10 mg/kg) when compared with placebo (21,51). These data raise concerns about the use of infliximab or etanercept in patients with symptomatic heart failure (17). Because most of our results were based on prevalent cases of heart failure, it is possible that infliximab was less likely to be given to patients with known heart failure, particularly in view of the U.S. Food and Drug Administration s warning. Indeed, fewer patients receiving infliximab had a history of cardiovascular disease than those not receiving infliximab. When we adjusted for previous history, however, we continued to find no increase in heart failure in patients treated with anti-tnf agents. In summary, heart failure is more common in rheumatoid arthritis patients than in those with osteoarthritis. Rheumatoid arthritis activity and severity measures were associated with heart failure, whereas patients receiving anti-tnf therapy were less likely to have heart failure. March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume

24 Heart Failure in Rheumatoid Arthritis/Wolfe and Michaud REFERENCES 1. Monson RR, Hall AP. Mortality among arthritics. J Chronic Dis. 1976;29: Mutru O, Koota K, Isomäki HA. Causes of death in autopsied RA patients. Scand J Rheumatol. 1976;5: Koota K, Isomäki HA, Mutru O. Death rate and causes of death in RA patients during a period of five years. Scand J Rheumatol. 1977; 6: Lewis P, Hazleman BL, Hanka R, Roberts S. Cause of death in patients with rheumatoid arthritis with particular reference to azathioprine. Ann Rheum Dis. 1980;39: Pincus T. Is mortality increased in rheumatoid arthritis? J Musculo Med. 1988;5: Wolfe F, Mitchell DM, Sibley JT, et al. The mortality of rheumatoid arthritis. Arthritis Rheum. 1994;37: Wallberg-Jonsson S, Ohman ML, Dahlqvist SR. Cardiovascular morbidity and mortality in patients with seropositive rheumatoid arthritis in Northern Sweden. J Rheumatol. 1997;24: Goodson NJ, Wiles NJ, Lunt M, et al. Mortality in early inflammatory polyarthritis: cardiovascular mortality is increased in seropositive patients. Arthritis Rheum. 2002;46: Wolfe F, Freundlich B, Straus WL. Increase in cardiovascular and cerebrovascular disease prevalence in rheumatoid arthritis. J Rheumatol. 2003;30: DeMaria AN. Relative risk of cardiovascular events in patients with rheumatoid arthritis. Am J Cardiol. 2002;89:33D 38D. 11. Goodson N. Coronary artery disease and rheumatoid arthritis. Curr Opin Rheumatol. 2002;14: del Rincon ID, Williams K, Stern MP, et al. High incidence of cardiovascular events in a rheumatoid arthritis cohort not explained by traditional cardiac risk factors. Arthritis Rheum. 2001;44: Khan AH, Spodick DH. Rheumatoid heart disease. Semin Arthritis Rheum. 1972;1: Mustonen J, Laakso M, Hirvonen T, et al. Abnormalities in left ventricular diastolic function in male patients with rheumatoid arthritis without clinically evident cardiovascular disease. Eur J Clin Invest. 1993;23: Feldman AM, Combes A, Wagner D, et al. The role of tumor necrosis factor in the pathophysiology of heart failure. J Am Coll Cardiol. 2000;35: Hurlimann D, Forster A, Noll G, et al. Anti-tumor necrosis factoralpha treatment improves endothelial function in patients with rheumatoid arthritis. Circulation. 2002;106: Mann DL. Inflammatory mediators and the failing heart: past, present, and the foreseeable future. Circ Res. 2002;91: Dinarello CA, Pomerantz BJ. Proinflammatory cytokines in heart disease. Blood Purif. 2001;19: Kubota T, Bounoutas GS, Miyagishima M, et al. Soluble tumor necrosis factor receptor abrogates myocardial inflammation but not hypertrophy in cytokine-induced cardiomyopathy. Circulation. 2000;101: Kadokami T, Frye C, Lemster B, et al. Anti-tumor necrosis factoralpha antibody limits heart failure in a transgenic model. Circulation. 2001;104: Hughes S. Infliximab harmful in CHF final results of ATTACH. Heartwire. June 12, Available at: Accessed January 1, Wood S. RENEWAL trial: no improvement in CHF with etanercept. Heartwire. June 11, Available at: theheart.org. Accessed January 1, Wolfe F, Anderson J, Burke TA, et al. Gastroprotective therapy and risk of gastrointestinal ulcers: risk reduction by COX-2 therapy. J Rheumatol. 2002;29: Wolfe F, Flowers N, Burke TA, et al. Increase in lifetime adverse drug reactions, service utilization, and disease severity among patients who will start COX-2 specific inhibitors: quantitative assessment of channeling bias and confounding by indication in 6689 patients with rheumatoid arthritis and osteoarthritis. J Rheumatol. 2002;29: Michaud K, Messer J, Choi HK, Wolfe F. Direct medical costs and their predictors in persons with rheumatoid arthritis: a 3 year study of 7,527 patients. Arthritis Rheum. 2003;48: Fries JF, Spitz PW, Young DY. The dimensions of health outcomes: the health assessment questionnaire, disability and pain scales. J Rheumatol. 1982;9: Fries JF, Spitz PW, Kraines RG, Holman HR. Measurement of patient outcome in arthritis. Arthritis Rheum. 1980;23: Hawley DJ, Wolfe F. Depression is not more common in rheumatoid arthritis: a 10 year longitudinal study of 6,608 rheumatic disease patients. J Rheumatol. 1993;20: Hawley DJ, Wolfe F. Anxiety and depression in patients with rheumatoid arthritis: a prospective study of 400 patients. J Rheumatol. 1988;15: Ware JE, Sherbourne CD. The MOS 36-Item Short-Form Health Survey (SF-36). 1. Conceptual framework and item selection. Med Care. 1992;30: Hurst NP, Kind P, Ruta D, et al. Measuring health-related quality of life in rheumatoid arthritis: validity, responsiveness and reliability of EuroQol (EQ-5D). Br J Rheumatol. 1997;36: Wolfe F, Hawley DJ. Measurement of the quality of life in rheumatic disorders using the EuroQol. Br J Rheumatol. 1997;36: Brazier J, Roberts J, Deverill M. The estimation of a preferencebased measure of health from the SF-36. J Health Econ. 2002;21: Brazier J, Usherwood T, Harper R, Thomas K. Deriving a preference-based single index from the UK SF-36 Health Survey. J Clin Epidemiol. 1998;51: Wolfe F, Hawley DJ, Wilson K. The prevalence and meaning of fatigue in rheumatic disease. J Rheumatol. 1996;23: Rubin DB. Estimating causal effects from large data sets using propensity scores. Ann Intern Med. 1997;127: Rubin DB. Estimation of nonrandomized treatment comparisons using subclassification on propensity scores. In: Abel U, Koch A, eds. Nonrandomized Comparative Clinical Studies: Proceedings of the International Conference on Nonrandomized Comparative Clinical Studies in Heidelberg. Heidelberg, Germany: Symposium Publishing; Becker SO, Ichino A. Estimation of average treatment effects based on propensity scores. Stata J. 2002;2: CLARIFY: Software for Interpreting and Presenting Statistical Results. Version 2.0. Cambridge, Massachusetts: Harvard University; Stata Statistical Software: Release 7.0. College Station, Texas: Stata Corporation; Kwon HJ, Cote TR, Cuffe MS, et al. Case reports of heart failure after therapy with a tumor necrosis factor antagonist. Ann Intern Med. 2003;138: Bacon PA, Raza K, Banks MJ, et al. The role of endothelial cell dysfunction in the cardiovascular mortality of RA. Int Rev Immunol. 2002;21: Bergholm R, Leirisalo-Repo M, Vehkavaara S, et al. Impaired responsiveness to NO in newly diagnosed patients with rheumatoid arthritis. Arterioscler Thromb Vasc Biol. 2002;22: Park YB, Ahn CW, Choi HK, et al. Atherosclerosis in rheumatoid arthritis: morphologic evidence obtained by carotid ultrasound. Arthritis Rheum. 2002;46: Park YB, Lee SK, Lee WK, et al. Lipid profiles in untreated patients with rheumatoid arthritis. J Rheumatol. 1999;26: March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume 116

25 Heart Failure in Rheumatoid Arthritis/Wolfe and Michaud 46. Kavanaugh A. Lipid profiles in patients with rheumatoid arthritis. Ann Rheum Dis. 1998;57: Hurt-Camejo E, Paredes S, Masana L, et al. Elevated levels of small, low-density lipoprotein with high affinity for arterial matrix components in patients with rheumatoid arthritis: possible contribution of phospholipase A2 to this atherogenic profile. Arthritis Rheum. 2001;44: Wallberg-Jonsson S, Cvetkovic JT, Sundqvist KG, et al. Activation of the immune system and inflammatory activity in relation to markers of atherothrombotic disease and atherosclerosis in rheumatoid arthritis. J Rheumatol. 2002;29: Bacon PA, Townend JN. Nails in the coffin: increasing evidence for the role of rheumatic disease in the cardiovascular mortality of rheumatoid arthritis. Arthritis Rheum. 2001;44: Wallberg-Jonsson S, Johansson H, Ohman ML, Rantapaa-Dahlqvist S. Extent of inflammation predicts cardiovascular disease and overall mortality in seropositive rheumatoid arthritis. A retrospective cohort study from disease onset. J Rheumatol. 1999;26: Anker SD, Coats AJ. How to RECOVER from RENAISSANCE? The significance of the results of RECOVER, RENAISSANCE, RENEWAL and ATTACH. Int J Cardiol. 2002;86: March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume

26 Regional Cerebral Hypoperfusion in Patients with Celiac Disease Giovanni Addolorato, MD, Daniela Di Giuda, MD, Giuseppe De Rossi, MD, Venanzio Valenza, MD, Marco Domenicali, MD, Fabio Caputo, MD, Antonio Gasbarrini, MD, Esmeralda Capristo, MD, Giovanni Gasbarrini, MD BACKGROUND: Neurological and psychiatric disorders occur in approximately 10% of patients with celiac disease. Although some of these alterations respond to a gluten-free diet, the etiology of these abnormalities is uncertain. Because of a case report that cerebral hypoperfusion in a celiac patient resolved after a gluten-free diet, we studied brain perfusion changes in untreated celiac patients, treated celiac patients, and healthy controls. METHODS: A total of 15 untreated celiac patients without conditions affecting brain perfusion were enrolled; none had neurological or psychiatric disorders other than anxiety or depression. We also studied 15 celiac patients who were on a gluten-free diet for almost 1 year, and 24 healthy volunteers of similar sex and age. All subjects underwent cerebral single photon emission computed tomography examination. RESULTS: Of the 15 untreated celiac patients, 11 (73%) had at least one hypoperfused brain region, compared with only 1 (7%) of the 15 celiac patients on a gluten-free diet and none of the controls (P 0.01). Cerebral perfusion was significantly lower (P 0.05) in untreated celiac patients, compared with healthy controls, in 7 of 26 brain regions. No significant differences in cerebral perfusion were found between celiac patients on a gluten-free diet and healthy controls. CONCLUSION: There is evidence of regional cerebral blood flow alteration in untreated celiac patients. Am J Med. 2004; 116: by Excerpta Medica Inc. Celiac disease is an autoimmune gluten-dependent enteropathy characterized by subtotal or total atrophy of intestinal villi that improves after withdrawal of gluten from the diet (1). Celiac disease is relatively common, affecting 1 of every 120 to 300 persons in Europe and the United States (2). Its classic symptoms include diarrhea, steatorrhea, abdominal distension, flatulence, and weight loss (1 3). Subclinical, or silent forms, with extraintestinal manifestations have been described (4 6). Several neurological, psychiatric, and affective disorders occur in celiac patients. In particular, cerebellar ataxia (7,8), peripheral neuropathies (9), epileptic seizures and intracranial calcifications (10), isolated brain vasculitis (11), brain atrophy and dementia (12), myoclonus and posterior column demyelization (13), schizophrenia (14), depressive syndrome (15 17), and anxiety disorders (15 17) have been reported. These complications occur in approximately 8% to 10% of celiac patients (18). Some of these alterations, such as cerebellar ataxia, From the Institutes of Internal Medicine (GA, EC, GG) and Pathology (AG), and Department of Nuclear Medicine (DDG, GDR, VV), Catholic University, Rome, Italy; and Institute of Internal Medicine, Cardioangiology, and Hepatology (MD, FC), University of Bologna, Bologna, Italy. The study was partially supported by grants from the Associazione Ricerca in Medicina, Bologna-Rome, Italy. Requests for reprints should be addressed to Giovanni Addolorato, MD, Institute of Internal Medicine, Catholic University, Gemelli Hospital, L.go A. Gemelli 8, Rome, Italy, or g.addolorato@rm. unicatt.it. Manuscript submitted March 20, 2003, and accepted in revised form September 26, respond to a gluten-free diet (19), whereas others, such as dementia, usually progress despite gluten restriction (20). The etiology and pathogenesis of these neurological, psychiatric, and affective disorders is uncertain (20). A case of brain perfusion abnormalities, assessed by single photon emission computed tomography (SPECT), has been reported in a celiac patient with schizophrenia. Regression of both cerebral hypoperfusion and schizophrenic symptoms was observed after 6 months of a gluten-free diet (14). Given that case report, the aim of the present study was to investigate brain perfusion changes in untreated celiac patients, comparing them with treated celiac patients and healthy subjects. METHODS Patients and Controls The study protocol complied fully with the guidelines of the Ethics Committee of the Università Cattolica del Sacro Cuore in Rome, Italy; all subjects provided informed consent. A group of 19 newly diagnosed adult patients affected by the classic form of celiac disease was enrolled consecutively in the study during a 1-year period. Four of these patients refused to undergo SPECT examination because they suffered from claustrophobia; the remaining 15 patients (9 women and 6 men; mean [ SD] age, 37 9 years) were studied. We excluded patients with endocrine disorders, abuse of alcohol or other substances, consumption of psychotropic drugs, and secondary causes of villous atrophy. The diagnosis of celiac disease was based on the presence of antigliadin and an by Excerpta Medica Inc /04/$ see front matter All rights reserved. doi: /j.amjmed

27 Brain Hypoperfusion in Celiac Disease/Addolorato et al tiendomysium antibodies, and was confirmed by histological evidence of subtotal or total duodenal villous atrophy and increased intraepithelial lymphocytes and crypt hyperplasia (1,2). No patients had neurological or psychiatric disorders other than anxiety or depression, or had a history of head trauma. The control groups included 15 treated celiac patients (9 women and 6 men; mean age, 30 8 years) on a gluten-free diet for almost 1 year (range, 1 to 10 years) and 24 healthy volunteers (15 women and 9 men; mean age, years) of similar sex and age and recruited from the hospital staff, after authorization from the Ethics Committee. These controls had no systemic or metabolic disease, neurological disorder, psychiatric illness, affective disorders, abuse of alcohol or other drugs, or history of head trauma. Treated celiac patients were symptom free: follow-up intestinal biopsy specimens had shown restoration of normal mucosal morphology, laboratory variables, and serum immunoglobulin (Ig) A and IgG antigliadin titers had returned to normal or decreased considerably. State and trait anxiety, and current depression, were assessed using the State and Trait Anxiety Inventory test and the modified version of Zung self-rating depression scale (16). A high level of state anxiety was defined as a score 40; a high level of trait anxiety was defined as a score 40; and depression was defined as a score 38. SPECT Imaging All subjects underwent brain SPECT examination. Technetium-99m-hexa-methyl-propylene-amine-oxime ( 99m Tc- HMPAO) was used as the SPECT tracer. Potassium perchlorate (200 mg) was administered orally to reduce thyroid and salivary gland uptake of tracer activity. 99m Tc-HMPAO (Ceretec; Nycomed-Amersham, Little Chalfont, Bucks, United Kingdom) was prepared according to the manufacturer s instructions; within 15 minutes, 740 MBq of 99m Tc-HMPAO was injected into an arm vein. Subjects were required to lie quietly in a dimly lit room for approximately 10 minutes before and after tracer injection. All SPECT scans were performed with a four-head brain-dedicated tomograph (CER.TO 96; SELO, Milan, Italy), which has higher resolution and sensitivity than standard rotating gamma cameras (21). Imaging was started 10 minutes after injection for a total acquisition time of 30 minutes. The head of the subject was settled on a flexible headrest and inserted in the gantry. Low-energy, high-resolution parallel-hole collimators with a system resolution of 7.5 mm full width half maximum at the center of a transverse slice were used. Up to 28 transaxial, coronal, and sagittal slices were reconstructed using Hanning filtered back-projections (cutoff frequency, 0.5 cycle/cm) and Chang s first order method of attenuation correction (coefficient, 0.12 cm -1 ). Reconstructed slices (nominal thickness, 7.5 mm) were oriented to the three spatial planes using a software program (SELO) that aligned the sagittal sections to the fronto-occipital line. To yield a satisfactory visualization of the hippocampal areas, each study was also realigned to a plane at 35 from the fronto-occipital line (22). Reconstructed slices were displayed on a color monitor and reviewed by two nuclear medicine specialists experienced in interpreting brain SPECT studies; they were blinded to the clinical characteristics of the patients. A semiquantitative analysis was performed on consecutive nonsummed transaxial slices. Using software (SELO), 23 anatomic regions of interest were selected (23) and adapted to the size of the examined brain areas. For each hemisphere, we selected three regions of interest (superior frontal, central, and parietal regions) as defined on three consecutive upper slices; six regions of interest (anterior frontal, middle frontal, sylvian, temporo-parietooccipital, occipital, thalamic regions) on five successive middle slices; and three regions of interest (anterior temporal region, cerebellar lobes, and pons) on three consecutive lower sections (Figure 1). Figure 1. Predetermined regions of interest (20). Selected brain regions are shown on the left of each diagram. 1: anterior temporal region; 2: pons; 3: cerebellar lobe; 4: anterior frontal region; 5: middle frontal region; 6: sylvian area; 7: temporo-parieto-occipital region; 8: occipital region; 9: thalamus area; 10: superior frontal region; 11: central region; 12: parietal region. March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume

28 Brain Hypoperfusion in Celiac Disease/Addolorato et al Figure 2. Selected transaxial slice at the basal ganglia level in an untreated celiac patient. The continuous color scale (256 colors) on the left of the image represents the distribution of radioactivity counts throughout the cortical and subcortical regions, normalized to maximum counts in the brain: the top of the color scale (white) corresponds to the highest tracer uptake, whereas the bottom (blue) represents the lowest tracer uptake. Image shows decreased tracer uptake (hypoperfusion) in anterior and middle frontal areas of both hemispheres (double arrows). Frontal regions appear thin and less red than the surrounding brain areas, as well as corresponding regions in healthy controls. L left hemisphere; R right hemisphere. Figure 3. Selected transaxial slice above the basal ganglia level in an untreated celiac patient. The explanation for the color scale is reported in the legend of Figure 2. Circumscribed areas of decreased tracer uptake (hypoperfusion) can be seen in bilateral superior frontal regions (double white arrows). These areas appear less red. Mild hypoperfusion is also present in the left parietal region (double yellow arrows), which appear thin in comparison with surrounding cortical regions or the equivalent area of the contralateral hemisphere, as well as corresponding regions in healthy controls. L left hemisphere; R right hemisphere. Figure 4. Selected transaxial slice at the basal ganglia level in an untreated celiac patient. The explanation for the color scale is reported in the legend of Figure 2. Image shows reduced tracer uptake (hypoperfusion) in anterior, middle frontal, and sylvian areas of the left hemisphere (double arrows). These regions appear thin and yellow in comparison with equivalent areas of the contralateral hemisphere, as well as corresponding regions in healthy controls. L left hemisphere; R right hemisphere. In addition, a square 36-pixel region of interest (one pixel 3.25 mm) in the anterior cingulated cortex and two 60-pixel regions of interest in the right and left hippocampal regions were drawn manually in three consecutive transaxial tomograms, based on a stereotaxic brain atlas (24). Thus, a total of 26 regions of interest were obtained for each subject. The mean counts per pixel of each selected region of interest were expressed as a percentage of those of a 60% isocontour whole brain region of interest that was drawn on all summed sagittal slices. The reliability of a whole brain reference region in the semiquantitative assessment of regional cerebral blood flow has been demonstrated (25). The regional values of each celiac patient were compared with the mean values ( 2 SD) obtained in the normal controls. Regional percentage values were considered abnormal if they were more than 2 SD below or above the normal mean value, according to the European Association of Nuclear Medicine procedure guidelines for brain perfusion SPECT using 99m Tc-labelled radiopharmaceuticals (26). Statistical Analysis All regional cerebral blood flows showed a normal distribution (P 0.20), as evaluated by the Kolmogorov-Smirnov test. Flows were compared among the three groups with multivariate analysis of variance (Hotelling s T squared). The Scheffé test was applied to account for multiple comparisons. Dichotomous variables were analyzed with Yates corrected chi-squared, or Fisher exact test, as appropriate. Data analysis was performed using SPSS 8.0 (Chicago, Illinois). Two-tailed P values 0.05 were considered statistically significant. 314 March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume 116

29 Brain Hypoperfusion in Celiac Disease/Addolorato et al Table. Cerebral Perfusion, by Region, in Celiac Patients (Untreated and Treated) and Healthy Control Subjects Brain Region Untreated Celiac Patients (n 15) Treated Celiac Patients (n 15) Healthy Controls (n 24) Untreated vs. Healthy Controls P Value Untreated vs. Treated Treated vs. Healthy Controls Mean SD* Pons Right anterior temporal Right anterior frontal Right middle frontal Right sylvian Right temporo-parieto-occipital Right occipital Right thalamus Right superior frontal Right central Right parietal Right cerebellar lobe Right hippocampal Anterior cingulated Left anterior temporal Left anterior frontal Left middle frontal Left sylvian Left temporo-parieto-occipital Left occipital Left thalamus Left superior frontal Left central Left parietal Left cerebellar lobe Left hippocampal * As measured by SPECT imaging. See Methods for details. A P value 0.05 was considered statistically significant. SPECT single photon emission computed tomography. RESULTS No cerebral perfusion abnormalities were found in the healthy control subjects. Tracer distribution was homogeneous throughout the grey matter, with the greatest uptake in the cerebellum and occipital lobes. Of the 15 untreated celiac patients, 11 (73%) had at least one hypoperfused brain region (Figures 2 to 4) (median value per subject: 4; range, 1 to 10), while only 1 (7%) of 15 celiac patients on a gluten-free diet had hypoperfusion. No hyperperfusion values were found in any celiac patients. An overall multivariate test showed a significant difference in cerebral perfusions among the three groups of subjects (P 0.01). Considering each single region (Table), untreated celiac patients had significantly lower cerebral perfusion than controls in seven of the 26 regions. There were no significant differences in cerebral perfusion between untreated patients and those on a glutenfree diet, or between celiac patients on a gluten-free diet and healthy controls (Table). High levels of state anxiety were more common in untreated celiac patients (73% [11/15]) than in healthy controls (21% [5/24]; P 0.003); treated celiac patients were intermediate (33% [5/15]). The prevalence of trait anxiety was similar in the three groups (untreated celiac: 27% [4/15], treated celiac: 20% [3/15], and healthy controls: 17% [4/24]). Depression was more common in untreated celiac patients (67% [10/15]) and in treated celiac patients (60% [9/15]) than in healthy controls (17% [4/24]; P 0.01 for both comparisons). When we compared the 11 untreated celiac patients who had high levels of state anxiety or depression (or both) with the remaining 4 patients, we found that cerebral hypoperfusion was lower in March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume

30 Brain Hypoperfusion in Celiac Disease/Addolorato et al the anxious or depressed patients in five of the 26 regions of interest. DISCUSSION We found that regional cerebral blood flow abnormalities were more common in untreated celiac patients than in healthy controls. We did not see a difference in cerebral perfusion between treated celiac patients and healthy controls, perhaps reflecting a beneficial effect of a glutenfree diet on these alterations, confirming our previous case report (14). Little is known about cerebral blood flow changes in celiac disease. A SPECT study in a small number of celiac patients with epilepsy and cerebral calcifications reported occipital hypoperfusion extending to the temporoparietal or to frontal cortices in most of these patients (27). However, information about the type of celiac disease and its treatment was not provided; moreover, all of these celiac patients had neurological disorders. Our results, obtained in celiac patients without neurological illness or psychiatric disorders other than anxiety or depression, showed a regional hypoperfusion in the frontal and parietal lobes of untreated patients. Perfusion defects were noted in the superior and anterior areas of the frontal cortex with involvement of the adjacent anterior cingulated cortex. Similar changes have been reported in patients suffering from psychiatric disorders (28). Prefrontal cortex hypoactivity, in the left hemisphere in particular, may be a marker for depression (29); this hypoactivity often returns to normal after successful treatment. Affective disorders, such as depression and anxious-neurotic behavior, were common in our untreated celiac patients, and cerebral hypoperfusion was lower in the anxious/depressed, untreated celiac patients. We did observe that anxiety, but not depression, was less common in treated celiac patients than in untreated patients. This observation supports a previous observation of an association between anxiety disorders and frontal cortex hypoperfusion (30). Future studies in a larger group of subjects, including controls with anxiety or depression, are needed to confirm this hypothesis. Similar blood flow alterations have been reported in untreated patients with anorexia nervosa; perfusion returned to normal after clinical remission (31,32). Similarly, in a recent study (33), anorectic patients had hypometabolism of glucose in some cortical regions, especially the frontal and parietal areas. A possible association between migraine and celiac disease has been reported (34,35). We recently found that 4.4% (4/90) of migraine patients had celiac disease, compared with 0.4% (1/236) in the healthy control group. All migraine patients with celiac disease had regional cerebral blood flow alteration evaluated by SPECT. Six months of a gluten-free diet improved the frequency, duration, and intensity of migraine, and the SPECT abnormalities (35). Blood flow abnormalities have also been reported in other anatomical regions of untreated celiac patients. For example, Arienti et al (36) showed variations in blood flow velocity in the mesenteric circulation of untreated celiac patients. These hemodynamic changes were linked to disease activity, and resolved after a gluten-free diet. In conclusion, regional cerebral blood flow alterations were present in untreated celiac disease, but not in patients on a gluten-free diet. The explanation for these observations is uncertain. Cerebral hypoperfusion may be related to intestinal hyperemia. Since celiac disease is sometimes associated with autoimmune disorders (2), it is possible that cerebral hypoperfusion depends on autoimmune or immune complex related endothelial inflammation, perhaps involving antigliadin antibodies or an unidentified neurotoxic antibody (8,37). Further longitudinal studies are needed to confirm our findings and to clarify the possible mechanisms that are involved. ACKNOWLEDGMENT We express our gratitude to the patients and controls who underwent SPECT examination for making the study possible. REFERENCES 1. Corazza GR, Gasbarrini G. Coeliac disease in adults. Baillieres Clin Gastroenterol. 1995;9: Farrel RJ, Kelly CP. Celiac sprue. N Engl J Med. 2002;346: Capristo E, Mingrone G, Addolorato G, et al. Differences in metabolic variables between adult coeliac patients at diagnosis and patients on a gluten-free diet. Scand J Gastroenterol. 1997;32: Catassi C, Ratsch IM, Fabiani E, et al. Coeliac disease in the year 2000: exploring the iceberg. Lancet. 1994;343: Corazza GR, Andreani ML, Venturo N, et al. Celiac disease and alopecia areata. Report of a new association. Gastroenterology. 1995; 109: Volta U, De Franceschi L, Lari F, et al. Coeliac disease hidden by cryptogenic hypertransaminasaemia. Lancet. 1998;352: Ward ME, Murphy JT, Greenberg GR. Celiac disease and spinocerebellar degeneration with normal vitamin E status. Neurology. 1985;35: Hadjivassiliou M, Gibson A, Davies-Jones GA, et al. Does cryptic gluten sensitivity play a part in neurological illness? Lancet. 1996; 347: Kaplan JG, Pack D, Horoupian D, et al. Distal axonopathy associated with chronic gluten enteropathy: a treatable disorder. Neurology. 1988;38: Gobbi G, Bouquet F, Greco L, et al. Coeliac disease, epilepsy, and cerebral calcifications. The Italian Working Group on Coeliac Disease and Epilepsy. Lancet. 1992;340: Rush PJ, Inman R, Bernstein M, et al. Isolated vasculitis of the central nervous system in a patient with celiac disease. Am J Med. 1986;81: Collin P, Pirttila T, Nurmikko T, et al. Celiac disease, brain atrophy, and dementia. Neurology. 1991;41: March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume 116

31 Brain Hypoperfusion in Celiac Disease/Addolorato et al 13. Kinney HC, Burger PC, Hurwitz BJ, et al. Degeneration of the central nervous system associated with celiac disease. J Neurol Sci. 1982; 53: De Santis A, Addolorato G, Romito A, et al. Schizophrenic symptoms and SPECT abnormalities in a coeliac patient: regression after a gluten-free diet. J Intern Med. 1997;242: Addolorato G, Stefanini GF, Capristo E, et al. Anxiety and depression in adult untreated celiac subjects and in patients affected by inflammatory bowel disease: a personality trait or a reactive illness? Hepatogastroenterology. 1996;43: Addolorato G, Capristo E, Ghittoni G, et al. Anxiety but not depression decreases in coeliac patients after one-year gluten-free diet: a longitudinal study. Scand J Gastroenterol. 2001;36: Hallert C, Astrom J. Psychic disturbances in adult coeliac disease. II. Psychological findings. Scand J Gastroenterol. 1982;17: Finelli PF, McEntee WJ, Ambler M, Kestenbaum D. Adult celiac disease presenting as cerebellar syndrome. Neurology. 1980;30: Beversdorf D, Moses P, Reeves A, Dunn J. A man with weight loss, ataxia, and confusion for 3 months. Lancet. 1996;347: Wills A. Neurological complication of enteric disease. Gut. 1996;39: De Rossi G. A new cerebral tomograph: technical features and preliminary experience. Radiol Diagn. 1992;33: Duncan R, Patterson J, Roberts R, et al. Ictal/postictal SPECT in the pre-surgical localization of complex partial seizures. J Neurol Neurosurg Psychiatr. 1993;56: Tran Dinh Y, Mamo H, Cervoni J, et al. Disturbances in the cerebral perfusion of human immune deficiency virus-1 seropositive asymptomatic subjects: a quantitative tomography study of 18 cases. J Nucl Med. 1990;31: Talairach J, Zilkha G, Tournoux P, et al. Atlas d Anatomie Stereotactique du Telencephale. Paris, France: Masson; Ottaviani F, Di Girolamo S, Briglia G, et al. Tonotopic organization of human auditory cortex analyzed by SPET. Audiology. 1997;36: Tatsch K, Asenbaum S, Bartenstein P, et al. European Association of Nuclear Medicine procedure guidelines for brain perfusion SPECT using 99m Tc-labelled radiopharmaceuticals. Eur J Nucl Med. 2002; 29:BP36 BP Messa C, D Amico A, Lucignani G, Fazio F. Cerebral blood flow in patients with coeliac disease, epilepsy and cerebral calcifications: a SPECT study. In: Gobbi G, Bianchini G, Naccarato S, Andermann F, eds. Epilepsy and Other Neurological Disorders in Coeliac Disease. London, Paris, Rome: John Libbey & Company Ltd; 1997: O Connel RA. SPECT brain imaging in psychiatric disorders: current clinical status. In: Grünwald F, Kasper S, Biersack H-J, Möller H-J, eds. Brain SPECT Imaging in Psychiatry. Berlin, New York: W. De Gruiter; 1995: Grasby PM, Bench C. Neuroimaging in mood disorders. Curr Opin Psychiatry. 1997;10: Davidson RJ, Abercrombie H, Nitschke JB, Putnam K. Regional brain function, emotion and disorders of emotion. Curr Opin Neurobiol. 1999;9: Kuruoglu AC, Kapucu Ö, Atasever T, et al. Technetium-99m- HMPAO brain SPECT in anorexia nervosa. J Nucl Med. 1998;39: Addolorato G, Taranto C, Capristo E, Gasbarrini G. A case of marked cerebellar atrophy in a woman with anorexia nervosa and cerebral atrophy and a review of the literature. Int J Eat Disord. 1998;24: Delvenne V, Lostra F, Goldman S, et al. Brain hypometabolism of glucose in anorexia nervosa: a PET scan study. Biol Psychiatry. 1995; 37: Hadjivassiliou M, Grunewald RA, Lawden M, et al. Headache and CNS white matter abnormalities associated with gluten sensitivity. Neurology. 2001;56: Gabrielli M, Cremonini F, Fiore G, et al. Association between migraine and coeliac disease. Am J Gastroenterol. 2003;98: Arienti V, Califano C, Brusco G, et al. Doppler ultrasonographic evaluation of splanchnic blood flow in coeliac disease. Gut. 1996; 39: Volta U, De Giorgio R, Petrolini N, et al. Clinical findings and anti-neuronal antibodies in coeliac disease with neurological disorders. Scand J Gastroenterol. 2002;37: March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume

32 Cerebrospinal Fluid Interleukin 8 Concentrations and the Subsequent Development of Postherpetic Neuralgia Naoki Kotani, MD, PhD, Ryoko Kudo, Yutaka Sakurai, MD, PhD, Daisuke Sawamura, MD, PhD, Daniel I. Sessler, MD, Hiroshi Okada, MD, PhD, Hiroto Nakayama, MD, PhD, Toshifumi Yamagata, Minoru Yasujima, MD, PhD, Akitomo Matsuki, MD, PhD PURPOSE: Other than age, the risk factors for postherpetic neuralgia are not well established. We studied whether the concentration of interleukin 8 in the cerebrospinal fluid is associated with the risk of postherpetic neuralgia. METHODS: We enrolled 170 patients more than 50 years old who had a typical painful and nontrigeminal herpetic rash. Patients were treated with acyclovir; no corticosteroids were given. Cerebrospinal fluid was taken for analysis of interleukin 8 during and at full crusting of the herpetic rash. Age, sex, comorbid conditions, prodromal pain, localization and severity of herpetic rash, number of skin lesions, and degree of pain were recorded. We used multivariate logistic regression modeling to identify significant predictive factors. Receiver operating characteristic (ROC) curves were evaluated to determine the contribution of each factor. RESULTS: Six months after healing, 31 patients (18%) had postherpetic neuralgia; 27 patients still had it after 1 year. Only three variables age (odds ratio [OR] 2.7 per 10-year increase; 95% confidence interval [CI]: 1.2 to 6.2), acute pain (OR 1.8 per unit increase in visual analog scale; 95% CI: 1.2 to 2.8), and interleukin 8 concentration in the cerebrospinal fluid at full crusting of the herpetic rash (OR 1.6 per 20- g/l increase; 95% CI: 1.3 to 2.0) were significant predictors of postherpetic neuralgia at 1 year. Interleukin 8 concentration also had the highest area under the ROC curve at these evaluation points (P 0.001). CONCLUSION: Our results suggest that interleukin 8 concentration in the cerebrospinal fluid at full crusting of herpetic rash may be useful for identifying patients who are likely to develop intractable postherpetic neuralgia. Am J Med. 2004;116: by Excerpta Medica Inc. Postherpetic neuralgia is characterized as continuous, lancinating pain and allodynia that persists for years, possibly for life. Identifying reliable predictors of who is likely to develop postherpetic neuralgia might allow physicians to advise patients on what to expect after the rash heals and how to detect early warning From the Outcomes Research Group (NK, DIS), Department of Anesthesiology (NK), Yamagata University, Yamagata, Japan; Departments of Anesthesiology and Laboratory Medicine (RK, MY, AM), University of Hirosaki, Hirosaki, Japan; Department of Hygiene (YS), National Defense Medical College, Tokorozawa, Japan; Department of Dermatology (DS), Hokkaido University, Sapporo, Japan; Department of Anesthesiology (DIS), University of Louisville, Louisville, Kentucky; Department of Anesthesiology (HO, HN), Iwate Medical University, Morioka, Japan; and Tohoku Chemical Co. Ltd (TY), Hirosaki, Japan. This work was supported by Grants-in-aid for Scientific Research B and C from the Ministry of Education and Science, Tokyo, Japan; Grant-in-aid for Cancer Research No from the Ministry of Health, Labor, and Welfare, Tokyo, Japan; grants GM58273 and GM from the National Institutes of Health, Bethesda, Maryland; the Joseph Drown Foundation, Los Angeles, California; and the Commonwealth of Kentucky Research Challenge Trust Fund, Louisville, Kentucky. Requests for reprints should be addressed to Naoki Kotani, MD, PhD, Department of Anesthesiology, Yamagata University School of Medicine, Yamagata , Japan, or kotani@med.id.yamagatau.ac.jp. Manuscript submitted February 24, 2003, and accepted in revised form October 20, signs (1). Physicians can also plan early and effective treatments for the chronic pain associated with postherpetic neuralgia. Advanced age is the only established risk factor for postherpetic neuralgia; others include female sex, prodromal pain, localization and severity of the herpetic rash, number of skin lesions, and degree of herpetic pain (2 6). However, the accuracy of these predictors varies among studies, and none seems especially helpful for planning early and effective prevention of postherpetic neuralgia. Other possibilities for predicting postherpetic neuralgia include biochemical markers in the cerebrospinal fluid. Although herpes zoster rarely leads to clinical complications in the central nervous system, histopathologic (7,8) and magnetic resonance studies (9), as well as cerebrospinal fluid analyses (9), indicate that the condition is associated with serious inflammatory and destructive changes in the spinal cord and its surrounding tissue. For example, interleukin 8 concentration in the cerebrospinal fluid of patients with postherpetic neuralgia is increased markedly (10,11). Furthermore, its concentration correlates with the degree and duration of postherpetic neuralgia (11). We therefore tested the hypothesis that cerebrospinal fluid interleukin 8 concentration predicts intractable postherpetic neuralgia better than do other reported factors by Excerpta Medica Inc /04/$ see front matter All rights reserved. doi: /j.amjmed

33 Interleukin 8 for Postherpetic Neuralgia/Kotani et al METHODS Patients Eligible patients were at least 50 years old and had a typical painful vesicular rash that was consistent with herpes zoster and had been present for less than 96 hours. Patients were referred to the participating investigators from university-affiliated hospitals. The diagnosis of herpes zoster was confirmed by detection of herpes zoster virus specific antibody. We excluded patients with any of the following conditions: painless zoster rash; pain in regions innervated by the trigeminal nerve; disseminated herpes zoster, defined by more than 10 discrete papules or vesicles outside primary dermatomes; recent immunosuppressive treatments, including steroids; or serious neurologic disorders, including metastases to the central nervous system. We also measured cerebrospinal fluid protein concentration, cell counts, and interleukin 8 concentration obtained from 50 relatively healthy (i.e., American Society of Anesthesiologists class 1 or 2) patients aged 50 years who underwent minor surgery (e.g., inguinal hernioplasty, hemorrhoidectomy) under spinal anesthesia. These control patients did not have current herpes zoster or postherpetic neuralgia. Other exclusion criteria were the same as for the patients with herpes zoster. In addition, patients with inflammatory disorders such as acute appendicitis were excluded. These patients had a mean ( SD) age of years; 29 were men. Data from these patients have been reported previously (11). The aim of the study and the nature of the tests were explained to the participants in accordance with the Helsinki Declaration. The Institutional Review Board of the University of Hirosaki and Iwate Medical University approved our protocol, and written informed consent was obtained from all participating patients. Acute and Chronic Treatment Acute herpes zoster was treated with acyclovir (500 mg intravenously for 5 days, followed by 800 mg orally five times daily for 1 week on an inpatient basis). Corticosteroids were not given, but diclofenac was prescribed for all patients. Patients who received only minimal relief with diclofenac were given buprenorphine. A few days after healing of herpes zoster (full crusting of the herpetic rash) and a second sampling of cerebrospinal fluid, patients were discharged. On an outpatient basis, we treated patients conventionally until pain subsided, analgesics were no longer required, patients were able to return to normal daily activity, or 1 year elapsed. Conventional treatments included antidepressants, anticonvulsants, topical agents, epidural nerve block, and intravenous ketamine, alone or in combination. These treatments were discontinued 1 week before each pain evaluation (see below). Furthermore, patients were not allowed to take analgesics during the 24 hours before the pain evaluation. Resolution of pain was defined as the last occurrence of any pain attributed to herpes zoster. Outcome Ascertainment Our study period extended from the initial visit to a physician to 1 year after healing of the herpetic rash. After enrollment, patients were evaluated daily until the herpetic rash had healed completely. At the patient s first visit to a physician, we recorded age, sex, comorbid diseases, prodromal signs and symptoms, localization and severity of the herpetic rash, number of skin lesions, and degree of pain. Comorbid disorders included diabetes mellitus, malignancy, and immune disorders such as acquired immune deficiency syndrome (AIDS) and autoimmune disease. Localization of the herpetic rash was recorded in four categories: cervical, thoracic, lumbar, or sacral. Because sacral localization increases the risk of postherpetic neuralgia (2,4), patients were grouped in two categories: high risk (sacral) and low risk (cervical, thoracic, and lumbar). The severity of the herpetic rash in the dermatomes was graded as none (0 points), mild (1 point), moderate (2 points), or severe (3 points) (12). The number of skin lesions was counted and graded as none (0 points), mild ( 25 lesions; 1 point), moderate (25 to 50 lesions; 2 points), or severe ( 50 lesions; 3 points) (13). Complications including meningitis, motor neuropathy, and skin superinfection that developed before or concurrently with postherpetic neuralgia were recorded. The degree of pain was evaluated on a 10-cm visual analog scale (0 cm no pain, 10 cm the worst imaginable pain). Twenty-four hours after analgesics were discontinued, we examined patients and recorded their current pain at 2 months, 6 months, and 1 year after healing of the herpetic rash. If pain persisted 2 months after full crusting of the herpetic rash, we considered that as herpes zoster associated pain. We diagnosed intractable postherpetic neuralgia when patients had chronic pain lasting 6 months or more. Sampling and Assay of Cerebrospinal Fluid To sample cerebrospinal fluid, we inserted 26-gauge pencil-point spinal needles into the lumbar intrathecal space. The first sample of cerebrospinal fluid was taken after confirmation of the diagnosis and before administration of acyclovir. We took a second sample at the time of full crusting of the herpetic rash. We also sampled cerebrospinal fluid from 50 patients who had neither herpetic nor neurologic disorders (controls). Assays of cerebrospinal fluid were performed as described (10,11). Briefly, the number of cells and protein concentration were measured. Samples were then centrifuged at 500 G for removal of cellular elements, and the supernatant was stored at 80 C for subsequent assay. The concentrations of interleukin 1, interleukin 6, interleukin 8, and tumor necrosis factor (TNF)- were March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume

34 Interleukin 8 for Postherpetic Neuralgia/Kotani et al measured by an enzyme-linked immunosorbent assay (Toray-Fuji Biochemicals, Tokyo, Japan). Data Analysis First, we evaluated the association between the prevalence of chronic pain and several possible predictive factors (2 6), including age, sex, comorbid conditions, prodromal pain, localization and severity of herpetic rash, number of skin lesions, degree of acute pain, and cerebrospinal fluid interleukin 8 concentrations during and at the time of healing of the herpetic rash. The values of the other measured cerebrospinal fluid cytokines (interleukin 1, interleukin 6, and TNF- ) were not included in the analysis because they were all below our detection levels. Factors that were associated significantly with the prevalence of herpes zoster associated pain and postherpetic neuralgia (P 0.05) were included as potential independent predictors in multivariate logistic regression models (StatView 5; SAS Institute Inc., Cary, North Carolina). Adjusted odds ratios and 95% confidence intervals were calculated. Odds ratios for cerebrospinal fluid interleukin 8 are reported for a 20- g/l increase and odds ratios for age are reported for a 10-year increase. We compared the predictive ability of significant predictive factors for postherpetic neuralgia using receiver operating characteristic curve (ROC) analysis (14). The area under the ROC curve was assessed (14); values range from 0.5 (no apparent accuracy) to 1.0 (perfect accuracy) as the curve moves toward the top left boundary of the graph. Variables in the cerebrospinal fluid were expressed as medians with 25th and 75th percentiles. For statistical analysis, we log transformed these variables. We then used paired t tests to analyze changes in each variable. We used unpaired t tests to compare these variables with those of control patients. We used regression analysis to determine the correlation of cerebrospinal fluid interleukin 8 concentration with cerebrospinal fluid cells and protein. These log-transformed parameters were expressed as mean SD. Differences were considered significant when P 0.05 (two-sided). RESULTS We enrolled 170 patients with herpes zoster; no patients dropped out of the study (Table 1). None of the participants experienced serious side effects during herpes zoster, and none had any treatment-related side effects. Seven patients had mild headaches after lumbar puncture. No other complications were noted. Fifty-one patients (30%) had herpes zoster associated pain at 2 months after healing of the herpetic rash; 31 (18%) had postherpetic neuralgia at 6 months posthealing; and 27 (16%) at 1 year. Table 1. Characteristic of the 170 Patients with Herpes Zoster Characteristic Number (%) or Mean SD Age (years) 65 9 Male sex 88 (52) Comorbid disorders* 42 (25) Prodromal pain 110 (65) Location Cervical 24 (14) Thoracic 116 (68) Lumbar 18 (11) Sacral 12 (7) Severity of acute pain (cm) Severity of skin rash 1 38 (22) 2 54 (32) 3 78 (46) Rating of skin lesions 1 67 (39) 2 42 (25) 3 61 (36) Cerebrospinal fluid interleukin 8 concentration at healing of herpetic rash ( g/l) * Diabetes mellitus, malignancy, and autoimmune disease. The degree of acute pain was evaluated on a 10-cm visual analog scale. The severity of the herpetic rash in the dermatomes was graded as mild (1 point), moderate (2 points), or severe (3 points). Skin lesions were counted and rated as mild ( 25 lesions; 1 point), moderate (25 to 50 lesions; 2 points), or severe ( 50 lesions; 3 points). In univariate models, several factors including age and female sex were associated with an increased risk of postherpetic neuralgia (Table 2). Patients with postherpetic neuralgia were also more likely to have experienced prodromal pain, severe acute herpetic pain, severe herpetic rash, and a large number of skin lesions. In addition, a high concentration of cerebrospinal fluid interleukin 8 concentration at healing of the herpetic rash increased the risk. For example, there was a 26% risk (12/47) of postherpetic neuralgia in patients more than 70 years old, but only a 12% risk (15/123) in younger patients. Patients reporting pain greater than the group mean (6.9 cm on the visual analog scale) had a 25% risk (23/93) of postherpetic neuralgia, compared with only 5% (4/77) in those who reported milder pain. Forty-five percent (25/56) of patients with an interleukin 8 concentration exceeding 70 g/l had postherpetic neuralgia, whereas only 2% (2/ 114) of the patients with lower interleukin 8 concentration did. In multivariate models, only three variables advanced age, degree of acute pain, and cerebrospinal fluid interleukin 8 concentration at healing of the herpetic rash were significant predictors of postherpetic neuralgia at 6 months and 1 year (Table 3). 320 March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume 116

35 Interleukin 8 for Postherpetic Neuralgia/Kotani et al Table 2. Independent Predictors of Postherpetic Neuralgia Variable 2-Month Herpes Zoster- Associated Pain (n 52) Odds Ratio (95% Confidence Interval) P Value 6-Month Postherpetic Neuralgia (n 31) Odds Ratio (95% Confidence Interval) P Value 1-Year Postherpetic Neuralgia (n 27) Odds Ratio (95% Confidence Interval) P Value Age (per 10-year increase) 2.2 ( ) ( ) ( ) 0.02 Prodromal pain 4.1 (1.1 16) (1.0 32) (0.7 28) 0.11 Degree of acute pain (per unit increase) 1.5 ( ) ( ) ( ) Severity of skin rash (per unit increase) 4.6 (1.2 19) ( ) ( ) 0.32 Cerebrospinal fluid interleukin 8 concentration at healing of herpetic rash (per 20- g/l increase) 1.8 ( ) ( ) ( ) The area under the ROC curve was significantly greater for cerebrospinal fluid interleukin 8 than for age or acute pain (Figure 1; P 0.001). In the control patients, the median (25th and 75th percentile) protein concentration in the cerebrospinal fluid was 550 mg/l (440 to 660 mg/l), cell counts were 2 per L (1to7per L), and the interleukin 8 concentration was 18 g/l (7 to 24 g/l). In comparison, the protein concentration in patients during the herpetic rash was 610 mg/l (462 to 770 mg/l), the cell count was 13 per L (5to75per L), and the interleukin 8 concentration was 81 g/l (57 to 380 g/l); all of these variables were significantly higher than in control patients (Figure 2). At the time of healing of herpes zoster, the cerebrospinal fluid protein concentration became similar to that of control patients (580 mg/l [463 to 770 mg/l]). Although the number of cells (5 per L [2to12per L]) and the concentration of interleukin 8 (44 g/l [28 to 101 g/l]) decreased at healing of the herpetic rash, these values were still higher than in control patients (Figure 2). Logtransformed cerebrospinal fluid interleukin 8 values correlated significantly with cerebrospinal fluid protein and cells during herpes zoster (Figure 3), but not at the time of healing of the herpetic rash (Figure 4). DISCUSSION Advanced age is the only established predictor of postherpetic neuralgia. Although we confirmed this association in our study, we also found that the cerebrospinal fluid interleukin 8 concentration at healing of the herpetic rash was more accurate than age for predicting postherpetic neuralgia. As in patients with serious meningitis and myeloencephalitis (15,16), we found that the cerebrospinal fluid interleukin 8 concentration during acute herpes zoster correlated significantly with the number of cells and the protein concentration in cerebrospinal fluid. The concentration of cerebrospinal fluid interleukin 8 during acute herpes zoster, however, failed to predict postherpetic neuralgia. Similarly, Haanpaa and coworkers (9) reported that the increase in cells and protein during herpes zoster was not associated with the severity of pain or the development of postherpetic neuralgia. Although cerebrospinal fluid interleukin 8 concentrations decreased over time, the concentration at healing of the herpetic rash was still significantly higher than in normal controls and was similar to previous reports (10,11). Studies with magnetic resonance imaging (9) demonstrate structural changes in the central nervous system in patients with postherpetic neuralgia, but not in patients Table 3. Area Under the Receiver Operating Characteristic Curves Outcome Cerebrospinal Fluid Interleukin 8 Age Acute Pain Area (95% Confidence Interval) 2-Month herpes zoster-associated pain 0.83 ( )* 0.73 ( ) 0.74 ( ) 6-Month postherpetic neuralgia 0.84 ( )* 0.72 ( ) 0.66 ( ) 1-Year postherpetic neuralgia 0.85 ( )* 0.69 ( ) 0.67 ( ) * P from other variables. March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume

36 Interleukin 8 for Postherpetic Neuralgia/Kotani et al Figure 1. Receiver operating characteristic curves for age (squares), acute pain (triangles), and interleukin 8 in the cerebrospinal fluid at the time of healing of the herpetic rash (circles). Only those patients having herpes zoster associated pain 2 months after full crusting of the herpetic rash, postherpetic neuralgia 6 months after full crusting of the rash, and postherpetic neuralgia 1 year after full crusting of the rash were included. For this analysis, age cutoffs were 55, 60, 65, 70, 75, 80, and 85 years; acute pain cutoffs were 5, 6, 7, 8, 9, and 10 cm on a visual analog scale; and interleukin 8 concentration cutoffs were 5, 20, 30, 40, 60, 80, 100, 120, 150, and 300 g/l. who recovered without neuralgia. Recent findings from animal and human studies suggest that many of the neurodestructive disorders are mediated by interleukin 8. Increases in cerebrospinal fluid concentrations of interleukin 8 and other cytokines have been observed in human traumatic nerve injury (17,18) and subarachnoid hemorrhage (19). Whalen and coworkers (18) reported that elevated cerebrospinal fluid interleukin 8 concentrations correlated with mortality in traumatic brain injury. Furthermore, inhibition of interleukin 8 activity in an ischemic stroke model in rats reduced infarct size and secondary brain edema (20). Recombinant neutrophil inhibitory factor, which has a pharmacologic function similar to that of an interleukin 8 antagonist, improved the neurologic outcome after ischemic brain injury in rats (21). Figure 2. Protein, inflammatory cells, and interleukin 8 concentrations from normal patients (n 50, squares) and patients with herpes zoster (n 170, circles). Log-transformed data are expressed as mean SD. Acute phase and healing are during the acute phase of herpes zoster and at the time of full crusting of the herpetic rash. Asterisks (*) indicate statistically significant differences (P 0.05) between acute phase and healing. Pound signs (#) indicate statistically significant differences (P 0.01) from the normal values. Interleukin 8 in the cerebrospinal fluid is derived from various types of cells in the spinal cord, including nerve, glial, and vascular epithelial cells. Therefore, an elevated interleukin 8 concentration at the time of healing could have been the result of mechanisms other than an inflammatory response (22). One possibility is that interleukin 8 acts as an appropriate defensive process against viral nerve injury. Anti-inflammatory cytokines such as interleukin 10 are elevated after acute brain injury (23) and can improve neurologic recovery in animal models (23). Interleukin 8 also has the potential to promote neuronal growth and repair (24). For example, in cultured astrocytes, interleukin 8 stimulates the production of nerve growth factor (25), and after brain injury the concentra- 322 March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume 116

37 Interleukin 8 for Postherpetic Neuralgia/Kotani et al Figure 3. Log-transformed interleukin 8 concentration correlated with the number of inflammatory cells and the concentration of protein in the cerebrospinal fluid during the acute phase of herpes zoster. Figure 4. Log-transformed interleukin 8 concentration did not correlate with the number of inflammatory cells or the concentration of protein in the cerebrospinal fluid at the time of healing of the herpetic rash. tion of interleukin 8 correlates closely with nerve growth factor concentration in cerebrospinal fluid (25). Interleukin 8 also promotes the survival of cultured hippocampal neurons (26). We found several other independent risk factors for intractable postherpetic neuralgia, including age and degree of acute pain. Some (27,28), but not all (3), authors have reported that acute pain is a risk factor. Prodromal pain and the severity of the skin rash were risk factors for postherpetic neuralgia 2 months after full crusting of the herpetic rash, but did not correlate with postherpetic neuralgia 6 and 12 months later. This finding emphasizes that definition of postherpetic neuralgia is critical in evaluating predictive factors. The major limitation of our study is that we excluded patients with trigeminal herpes zoster who have a greater risk of postherpetic neuralgia (29,30) because elevation of cerebrospinal fluid interleukin 8 is inconsistent during and after trigeminal zoster (10,11). Additional research would be necessary to determine whether cerebrospinal fluid interleukin 8 concentrations are associated with the subsequent development of trigeminal postherpetic neuralgia. In summary, we found an increase in the concentration of interleukin 8 in the cerebrospinal fluid during and at the time of healing of herpes zoster. An elevated concentration at the time of healing of the herpetic rash was highly reliable for predicting postherpetic neuralgia 6 months and 1 year later. We conclude that measurement of cerebrospinal fluid interleukin 8 at healing of the herpetic rash may be an effective tool for predicting the development of postherpetic neuralgia and possibly for planning early treatment for this debilitating condition. ACKNOWLEDGMENT We are grateful to Elisabeth F. Lanzl, AM, University of Chicago, for editing the manuscript, and to Gilbert Haugh, MS, University of Louisville, for statistical assistance. REFERENCES 1. Kost RG, Straus SE. Postherpetic neuralgia. Predicting and preventing risk. Arch Intern Med. 1997;157: Kost RG, Straus SE. Postherpetic neuralgia pathogenesis, treatment, and prevention. N Engl J Med. 1996;335: Choo PW, Galil K, Donahue JG, Walker AM, Spiegelman D, Platt R. Risk factors for postherpetic neuralgia. Arch Intern Med. 1997; 157: Meister W, Neiss A, Gross G, et al. A prognostic score for postherpetic neuralgia in ambulatory patients. Infection. 1998;26: Whitley RJ, Shukla S, Crooks RJ. The identification of risk factors associated with persistent pain following herpes zoster. J Infect Dis. 1998;178(suppl 1):S71 S75. March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume

38 Interleukin 8 for Postherpetic Neuralgia/Kotani et al 6. Dworkin RH, Boon RJ, Griffin DR, Phung D. Postherpetic neuralgia: impact of famciclovir, age, rash severity, and acute pain in herpes zoster patients. J Infect Dis. 1998;178(suppl 1):S76 S Schmidbauer M, Budka H, Pilz P, Kurata T, Hondo R. Presence, distribution and spread of productive varicella zoster virus infection in nervous tissues. Brain. 1992;115: Watson CPN, Deck JH, Morshead C, Van der Kooy D, Evans RJ. Post-herpetic neuralgia: further post-mortem studies of cases with and without pain. Pain. 1991;44: Haanpaa M, Dastidar P, Weinberg A, et al. CSF and MRI findings in patients with acute herpes zoster. Neurology. 1998;51: Kikuchi A, Kotani N, Sato T, Takamura K, Sakai I, Matsuki A. Comparative therapeutic evaluation of intrathecal versus epidural methylprednisolone for long-term analgesia in patients with intractable postherpetic neuralgia. Reg Anesth Pain Med. 1999;24: Kotani N, Kushikata T, Hashimoto H, et al. Intrathecal methylprednisolone for intractable postherpetic neuralgia. N Engl J Med. 2000;343: Higa K, Mori M, Hirata K, Hori K, Manabe H, Dan K. Severity of skin lesions of herpes zoster at the worst phase rather than age and involved region most influences the duration of acute herpetic pain. Pain. 1997;69: Whitley RJ, Weiss HL, Soong SJ, Gnann JW. Herpes zoster: risk categories for persistent pain. J Infect Dis. 1999;179: Hanley JA, McNeil BJ. A method of comparing the areas under receiver operating characteristic curves derived from the same cases. Radiology. 1983;148: Lopez Cortes LF, Marquez Arbizu R, Jimenez Jimenez LM, et al. Cerebrospinal fluid tumor necrosis factor-alpha, interleukin-1beta, interleukin-6, and interleukin-8 as diagnostic markers of cerebrospinal fluid infection in neurosurgical patients. Crit Care Med. 2000;28: Ostergaard C, Benfield T, Gesser B, et al. Pretreatment with granulocyte colony-stimulating factor attenuates the inflammatory response but not the bacterial load in cerebrospinal fluid during experimental pneumococcal meningitis in rabbits. Infect Immun. 1999;67: Xia M, Qin S, McNamara M, Mackay C, Hyman BT. Interleukin-8 receptor B immunoreactivity in brain and neuritic plaques of Alzheimer s disease. Am J Pathol. 1997;150: Whalen MJ, Carlos TM, Kochanek PM, et al. Interleukin-8 is increased in cerebrospinal fluid of children with severe head injury. Crit Care Med. 2000;28: Kikuchi T, Okuda Y, Kaito N, Abe T. Cytokine production in cerebrospinal fluid after subarachnoid haemorrhage. Neurol Res. 1995; 17: Matsumoto T, Ikeda K, Mukaida N, et al. Prevention of cerebral edema and infarct in cerebral reperfusion injury by an antibody to interleukin-8. Lab Invest. 1997;77: Jiang N, Chopp M, Chahwala S. Neutrophil inhibitory factor treatment of focal cerebral ischemia in the rat. Brain Res. 1998;788: Sherwood ER, Prough DS. Interleukin-8, neuroinflammation, and secondary brain injury. Crit Care Med. 2000;28: Bell MJ, Kochanek PM, Doughty LA, et al. Interleukin-6 and interleukin-10 in cerebrospinal fluid after severe traumatic brain injury in children. J Neurotrauma. 1997;14: Knoblach SM, Faden AI. Interleukin-10 improves outcome and alters proinflammatory cytokine expression after experimental traumatic brain injury. Exp Neurol. 1998;153: Kossmann T, Stahel PF, Lenzlinger PM, et al. Interleukin-8 released into the cerebrospinal fluid after brain injury is associated with blood-brain barrier dysfunction and nerve growth factor production. J Cereb Blood Flow Metab. 1997;17: Araujo DM, Cotman CW. Trophic effects of interleukin-4, -7 and -8 on hippocampal neuronal cultures: potential involvement of glial-derived factors. Brain Res. 1993;600: Dworkin RH, Hartstein G, Rosner HL, Walther RR, Sweeney EW, Brand L. A high-risk method for studying psychosocial antecedents of chronic pain: the prospective investigation of herpes zoster. J Abnorm Psychol. 1992;101: Whitley RJ, Weiss H, Gnann JW Jr, et al. Acyclovir with and without prednisone for the treatment of herpes zoster. A randomized, placebo-controlled trial. The National Institute of Allergy and Infectious Diseases Collaborative Antiviral Study Group. Ann Intern Med. 1996;125: Ragozzino MW, Melton LJ III, Kurland LT, Chu CP, Perry HO. Population-based study of herpes zoster and its sequelae. Medicine (Baltimore). 1982;61: Galil K, Choo PW, Donahue JG, Platt R. The sequelae of herpes zoster. Arch Intern Med. 1997;157: March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume 116

39 SPECIAL ARTICLE Cost-effectiveness of Inhaled Corticosteroids for Chronic Obstructive Pulmonary Disease According to Disease Severity Donald D. Sin, MD, MPH, Kamran Golmohammadi, MD, Philip Jacobs, PhD PURPOSE: Inhaled corticosteroids reduce exacerbations in patients with chronic obstructive pulmonary disease (COPD), but their cost-effectiveness is not known. METHODS: We used a Markov model to determine, from a societal perspective, the cost-effectiveness of four treatment strategies involving inhaled corticosteroids: no use regardless of COPD severity; use in all disease stages; use in patients with stage 2 or 3 disease (forced expiratory volume in 1 second [FEV 1 ] 50% of predicted); and use in patients with stage 3 disease (FEV 1 35% of predicted). Data from the literature were used to estimate mortality, exacerbation, and disease progression rates, as well as the costs associated with care and quality-adjusted life-years (QALYs), according to disease stage and use or nonuse of inhaled corticosteroids. A time horizon of 3 years was used. RESULTS: Use of inhaled corticosteroids in patients with stage 2 or 3 disease was associated with a cost of $17,000 per QALY gained. In stage 3 patients, use resulted in a cost of $11,100 per QALY gained. Providing inhaled corticosteroids to all COPD patients was associated with a less favorable cost-effectiveness ratio. Results were robust to various assumptions in a Monte Carlo simulation. CONCLUSION: In patients with COPD, use of inhaled corticosteroids in those with stage 2 or 3 disease for 3 years results in improved quality-adjusted life expectancy at a cost that is similar to that of other therapies commonly used in clinical practice. Am J Med. 2004;116: by Excerpta Medica Inc. Chronic obstructive pulmonary disease (COPD) affects more than 16 million people in the United States, at a societal cost of over $23 billion in direct and $1 billion in indirect costs annually (1). Indeed, morbidity and mortality from COPD are expected to increase rapidly over the next 20 years (2 4). Recent studies suggest that inhaled corticosteroids may attenuate airway responsiveness (5), reduce exacerbation rates (6), and slow the decline in health status (7) in patients with COPD. Although these data are promising, additional information regarding costs, resource use, and cost-effectiveness would be helpful in determining from a societal From the Institute of Health Economics (DDS, KG, PJ), and Departments of Medicine (Pulmonary Division) (DDS) and Public Health Sciences (PJ), University of Alberta, Edmonton, Alberta, Canada. This project was funded in part by an unrestricted research grant from GlaxoSmithKline Canada Inc., Mississauga, Ontario, Canada, and the Institute of Health Economics, Edmonton, Alberta. Dr. Sin is supported by a New Investigator Award from the Canadian Institutes of Health Research and a Population Health Investigator Award from the Alberta Heritage Foundation for Medical Research. Drs. Sin and Jacobs have received honoraria and research funding from AstraZeneca and GlaxoSmithKline. Requests for reprints should be addressed to Donald D. Sin, MD, MPH, 2E4.29 Walter C. Mackenzie Centre, University of Alberta, Edmonton T6G 2B7, Canada, or don.sin@ualberta.ca. Manuscript submitted March 13, 2003, and accepted in revised form September 18, perspective in which groups of patients with COPD, if any, these medications should be recommended. METHODS Design of the Model We designed a simulation model to estimate, from a societal perspective, the clinical effects and costs of using (or not using) inhaled corticosteroids in patients with varying severity of COPD. The baseline model used an analytic horizon of 3 years divided in 3-month increments. The model was limited to 3 years because there were insufficient data on the efficacy of inhaled corticosteroids beyond this time horizon. We used a 3-month window for calculating transitional probabilities to allow maximal flexibility for movement of patients between disease severity categories. Although the model was originally developed for findings pertaining to a Canadian population, we used data from the United States, the United Kingdom, and other Western European countries to ensure that the findings would be generalizable to other populations. Disease History To model the natural history of COPD, patients were divided into three mutually exclusive disease severity strata based on the recommendations from the American 2004 by Excerpta Medica Inc /04/$ see front matter 325 All rights reserved. doi: /j.amjmed

40 Cost-effectiveness of Inhaled Steroids in COPD/Sin et al Thoracic Society (8): stage 1 disease was defined as forced expiratory volume in 1 second (FEV 1 ) 50% of predicted; stage 2 disease as FEV 1 of 35% to 50% of predicted; and stage 3 disease as FEV 1 35% of predicted. From the Third National Health and Nutrition Examination Survey, we estimated that 93% of the total pool of patients had stage 1 disease, 4% had stage 2 disease, and 3% had stage 3 disease (9). Once the baseline severity was established, the model assumed that FEV 1 declined with time. Based on data from the Lung Health Study (10), we assumed that the mean rate of FEV 1 decline in each severity group was 47 ml per patient per year (or ml over 3 months). Using this assumption, the probability for a person with stage 1 disease to progress to stage 2 during a 3-month period was 0.74%, and 2.48% for progression from stage 2 to 3 disease. As FEV 1 declined, the number and severity of exacerbations as well as mortality risk increased. Because there is little evidence that inhaled corticosteroids have a substantial effect on the rate of decline in FEV 1 (except during the first 6 to 12 months of therapy), the same transitional probability values (of moving from a lower to higher disease stage) were applied to patients who did or did not use inhaled corticosteroids after the first year of therapy (7,11). Modeled Strategies In a Markov model, the effects of inhaled corticosteroids were evaluated using four strategies. In the first (base case) strategy, none of the patients were treated during the follow-up period. In the second strategy, all patients regardless of disease severity were treated. In the third strategy, inhaled corticosteroids were given only to patients with stage 2 or 3 disease. In the fourth strategy, inhaled corticosteroids were given only to those with stage 3 disease. All patients were followed for 3 years in 3-month blocks. For each 3-month period, the probabilities of death and exacerbation were assessed for each patient within each disease category. In moving from one 3-month cycle to the next, a small proportion of patients progressed to a higher disease severity category, based on the expected declines in FEV 1 for each severity group. Patients who died during the 3-month cycle were censored from further analysis. Survivors of each cycle were passed through another 3-month cycle wherein a similar set of probabilities (survival, exacerbation, and disease progression) were applied. In total, there were 12 cycles, translating into 36 months of follow-up. The aggregated probabilities of deaths, exacerbations, and disease progression were calculated for the entire follow-up period. In addition, we calculated the costs associated with clinical outcome, including the costs of inhaled corticosteroids. We assumed that the cost of routine follow-up COPD care applied equally between those who did and did not receive inhaled corticosteroids. Costs were discounted at an annual rate of 5%. Health Outcomes The two main health outcomes were patient health-related quality of life (HRQL) and all-cause mortality. We assumed that a patient s health status declined with decrements in FEV 1 (12). We also assumed that there were further reductions in health status with each exacerbation (13). For analytic purposes, we subclassified exacerbations into three mutually exclusive categories (14): mild, defined as worsening of symptoms requiring outpatient physician services and institution of systemic corticosteroids or antimicrobial agents (i.e., exacerbation therapy); moderate, defined as clinical episodes requiring emergency department utilization or urgent physician office visits (and institution of exacerbation therapy); and severe, defined as symptoms requiring in-patient care (and institution of exacerbation therapy). The average HRQL for a stable patient with stage 3 disease was estimated to be 0.84 quality-adjusted life-years (QALYs) (15). The average HRQL for a patient with stage 2 disease was estimated to be 0.92 QALYs. For stage 1 disease, 1.0 QALYs were estimated. Because exacerbations related to COPD are characterized by an increase in cough and dyspnea, we used the Oregon Survey to estimate the reductions in QALY during exacerbations (16), assuming the reduction to be 0.32 QALYs for each episode. We assumed the average duration of mild exacerbations to be 1 week, moderate exacerbations to be 2 weeks, and severe exacerbations to be 4 weeks. All of these data were used to determine the cost-effectiveness (in QALYs) for each strategy. In the base case analysis, the overall rate of exacerbations, regardless of severity, was assumed to be 4.97% in patients with stage 1 disease (7,17) and 14.55% in those with stage 2 or 3 disease (7,14,18), over a 30-day period (Table 1). In stage 1, most of the exacerbations were assumed to be self-manageable (93.7% of total exacerbations); whereas in stage 2, 74% of the clinically apparent exacerbations were assumed to be moderate or severe. For patients with stage 3 disease, we assumed that 30% of the exacerbations would require hospitalization (18). Based on data from a recent review (6), inhaled corticosteroids were assumed to reduce the rate of all types of exacerbations by 30% (relative risk [RR] 0.70; 95% confidence interval: 0.58 to 0.84). Probabilities for exacerbations were reduced accordingly for strategies in which inhaled corticosteroid therapy was instituted for COPD patients (Table 2). All-cause mortality was estimated from data reported by Anthonisen et al (25). We chose all-cause mortality because patients with COPD may die from complications of COPD as well as from other causes, and the distinction is not always clear. We assumed that mortality varied ac- 326 March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume 116

41 Cost-effectiveness of Inhaled Steroids in COPD/Sin et al Table 1. Monthly Rate of Exacerbations and the Average Cost Associated with Each Class of Exacerbation* Mild Exacerbations Moderate Exacerbations Severe Exacerbations References Exacerbation rate in stage % 0.19% 0.13% 7,17 Exacerbation rate in stage % 9.05% 1.76% 14 Exacerbation rate in stage % 4.42% 18 Direct cost of exacerbation $53.08 $ $ Indirect cost of exacerbation $0 $52 $369 * Costs expressed in Canadian dollars (Canadian $1 is equivalent to U.S. $0.72). No reliable data available. Mild exacerbations assumed a visit with a physician and use of medications. Physician visits were estimated to be $24.78 (Fee code item 0303A, Alberta Health Care Insurance Plan, Schedule of Medical Benefits, 2000) (19). Medication costs included 7 days of antibiotics, at a cost of $4.40 per day (20), which was a weighted average of daily use of the following drugs: ampicillin, trimethoprim-sulfamethoxazole, cephalosporin, penicillin, fluoroquinolones, macrolides, and tetracycline. The weights were calculated based on data from reference 21. Services included those used for mild exacerbations, plus emergency services. Emergency visits were $164 (22). Emergency physician billing was $ Services included those used for mild and moderate exacerbations as well as those related to hospitalizations. Inpatient hospitalization was based on weighted average resource intensity for Case Mix Group 140 (23), estimated to be $2842 per weighted case, for a value of $3581 per case. Physician fees for in-hospital care (of a 7-day stay) were estimated to be $ Based on data from the Canada National Population Health Survey (24). Approximately 55% of all persons with COPD are under 65 years old, and only 65% of those younger than 65 years are employed. Therefore, 36% ( ) of all persons in the COPD age range are employed. The average daily wage for all persons in Canada is $148 ($742 weekly/5-day work week). We assumed that persons with moderate exacerbations lost 1 day of work. Persons with severe exacerbations lost 7 days of work. Table 2. Probabilities of Various Events, by Use of Inhaled Corticosteroid Therapy Variable Standard Therapy Standard Therapy Inhaled Corticosteroids No. of Events per 100 Person-Years (95% Confidence Interval) Mortality Stage 1 (FEV 1 50% of predicted) 4.64 ( ) 3.92 ( )* Stage 2 (FEV 1, 35% 49% of predicted) 7.32 ( ) 6.16 ( )* Stage 3 (FEV 1 35% of predicted) ( ) 9.24 ( )* Mild exacerbation rate Stage ( ) 38.9 ( ) Stage ( ) 31.3 ( ) Moderate exacerbation rate Stage ( ) 1.6 ( ) Stage ( ) 74.6 ( ) Stage ( ) 85.1 ( ) Severe exacerbation rate Stage ( ) 1.1 ( ) Stage ( ) 14.5 ( ) Stage ( ) 37.1 ( ) Progression of disease Stage 1 to ( ) 2.96 ( ) Stage 2 to ( ) 9.92 ( ) * Expressed as mode (range), as the distribution of responses to inhaled corticosteroids was non-normally distributed (6). FEV 1 forced expiratory volume in 1 second. March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume

42 Cost-effectiveness of Inhaled Steroids in COPD/Sin et al Table 3. Marginal Cost-effectiveness Values Associated with Various Strategies for the 3-Year and Lifetime Models* Strategy Cost Per QALY Gained Assuming Mortality Benefits of Inhaled Steroids Cost Per QALY Gained Assuming No Mortality Benefits of Inhaled Steroids Cost Per Exacerbation Reduced 3-Year model All disease stages $46,200 $123,700 $10,000 Stage 2 or 3 disease (FEV 1 50% of predicted) $17,000 $34,100 $1000 Stage 3 disease (FEV 1 35% of predicted) $11,100 $30,200 $600 Lifetime model All disease stages $4600 $26,200 $5200 Stage 2 or 3 disease $2900 $21,200 $900 Stage 3 disease $2000 $15,000 $600 * QALYs expressed in U.S. dollars. In all analyses, the referent is the strategy of not treating anyone with inhaled corticosteroids. Only moderate and severe exacerbations were considered. FEV 1 forced expiratory volume in 1 second; QALY quality-adjusted life-year. cording to disease severity (25) and treatment with inhaled corticosteroids (6). In the base case analysis (i.e., no inhaled corticosteroids used), the imputed 3-month mortality was 1.2% in patients with stage 1 disease, 1.8% in those with stage 2 disease, and 2.8% in those with stage 3 disease (Table 2). All-cause mortality was assumed to be 16% lower in patients treated with inhaled corticosteroids than in the control group (RR 0.84) (6). Calculation of Costs Direct medical expenditures were determined for all disease states in Canadian dollars (Canadian $1 is equivalent to U.S. $0.72), stratified according to whether or not patients were treated with inhaled corticosteroids. All direct costs pertaining to office, emergency, and hospital visits were calculated based on the 1999 to 2000 fee schedules from Alberta Health and Wellness (19); only marginal costs (i.e., costs above and beyond routine care) were considered for this analysis. The estimated marginal costs for exacerbations are summarized in Table 1. Because most randomized controlled trials used a dose of inhaled corticosteroids equivalent to 1 mg per day of fluticasone propionate, for analytic purposes we assumed the marginal daily medication costs to be $3.20 (20). We assumed that patients would use the drug for the entire follow-up period once assigned to the inhaled corticosteroid arm. In a secondary analysis, we also calculated the marginal indirect costs associated with work loss during exacerbations for those younger than 65 years. Cost-effectiveness ratios were then calculated using standard methods. Sensitivity Analyses To determine the robustness of our data to a different (but clinically plausible) range of assumptions, we performed multivariate sensitivity analyses where we made simultaneous adjustments for the effects of the relevant covariates for each of the strategies. Using Monte Carlo simulation, we randomly sampled all variables and produced 100,000 sample sets. We performed an economic analysis using a lifetime horizon, based on similar assumptions as those for the 3-year model. Finally, we evaluated the cost-effectiveness of inhaled corticosteroids with the assumption that they had no effect on all-cause mortality for both 3-year and lifetime horizons. All data analyses were conducted using Data Pro (TREEAGE Software Inc; Williamstown, Massachusetts). All QALYs are expressed to the nearest $100 using U.S. dollars. RESULTS The characteristics of the representative sample of patients with COPD used in the models were as follows: mean age, 61 years; 21% female; 87% white; 97% former or current smokers; 40% current smokers; and 54 packyears of smoking (25). Quality-Adjusted Life-years The base case strategy (patients not receiving inhaled corticosteroids) was associated with 2.71 QALYs, compared with 2.72 QALYs for the strategy in which inhaled corticosteroids were administered to patients with stage 3 disease, 2.72 QALYs when corticosteroids were given to those with stage 2 or 3 disease, and 2.75 QALYs when corticosteroids were given to all patients regardless of disease severity. Costs If inhaled corticosteroid therapy was given only to patients with stage 2 or 3 disease, the total marginal costs over 3 years would be $922 per patient. Cost would be $3612 per patient if corticosteroids were given to all COPD patients. The total marginal costs would be $774 if only stage 3 patients were targeted for therapy. Cost-effectiveness Analysis In comparisons of the marginal cost-effectiveness ratios for the various strategies over 3-year and lifetime hori- 328 March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume 116

43 Cost-effectiveness of Inhaled Steroids in COPD/Sin et al zons (Table 3), the base case strategy served as the comparator for each comparison. In the 3-year model, if corticosteroids were given only to patients with stage 3 disease, the cost per QALY gained would be $11,100. By giving corticosteroids to patients with stage 2 or 3 disease, the cost increased to $17,000 for every QALY gained; and by giving corticosteroids to patients regardless of disease severity, the cost increased to $46,200 per QALY gained. If we assumed that inhaled corticosteroids had no effect on mortality, the cost-effectiveness values became less favorable. Providing these drugs to patients with stage 3 disease, for instance, resulted in a figure of $30,200 per QALY gained. The cost increased to $34,100 when corticosteroids were given to patients with stage 2 or 3 disease. The lifetime model produced more favorable cost-effectiveness values. Even when corticosteroids were given to all patients, the cost per QALY gained was only $4600 (assuming a mortality benefit of inhaled corticosteroids) or $26,200 (assuming no mortality benefit). Sensitivity Analysis If we used U.S. $50,000 per QALY gained as a benchmark for determining the cost-effectiveness of inhaled corticosteroids, then there was a 57% probability that the strategy of providing inhaled corticosteroids to all patients regardless of severity would be cost-effective. The probability increased to 95% if inhaled corticosteroids were given to only stage 2 or 3 patients, whereas it would be 84% if corticosteroids were provided to only patients with stage 3 disease. We also evaluated the approach of empirically treating all patients with stage 2 or 3 disease for 6 months and then discontinuing treatment in those who were deemed nonresponders by their attending physicians. If 25% were nonresponders, the overall cost per QALY gained would be $18,100 (or $36,400 assuming no mortality reduction) in the 3-year model and $2900 (or $21,300 assuming no mortality reduction) in the lifetime model. If 50% were nonresponders, the cost per QALY gained would be $19,200 (or $38,600 assuming no mortality reduction) in the 3-year model and $3000 (or $21,500 assuming no mortality reduction) in the lifetime model. DISCUSSION We developed an economic model to determine the costeffectiveness of using inhaled corticosteroids in patients with varying severity of COPD, and found that treatment was most cost-effective when given to patients with stage 2 or 3 disease (FEV 1 50% of predicted). Compared with the base case strategy (no treatment for any COPD patients), therapy restricted to patients with stage 3 disease (FEV 1 35% of predicted) was associated with a cost of $11,100 per QALY gained; therapy restricted to those with stage 2 or 3 disease was associated with a cost of $17,000 per QALY gained; whereas therapy provided to all patients, regardless of disease severity, was associated with a cost of $46,200 per QALY gained. We tested the robustness of these findings to various different (but plausible) assumptions and conditions using a Monte Carlo simulation model, and found these estimates to be stable. Data from clinical trials suggest a modest effect of inhaled corticosteroids in reducing clinically important exacerbations, including hospitalizations (6). Although the effect on all-cause mortality is not known, a recent metaanalysis indicated a point estimate of 0.84 but with a 95% confidence interval that overlapped 1 (6). We incorporated the uncertainties of these point estimates in our model (26). Even so, there were few changes in our results of sensitivity analyses, suggesting that our findings for stage 2 and 3 disease were robust. Findings were less stable for stage 1 disease, probably reflecting the increased heterogeneity of clinical disease within this group of patients. When we assumed no direct mortality benefits of inhaled corticosteroids, the use of inhaled corticosteroids in patients with stage 2 or 3 disease was still cost-effective. If these benefits extended beyond 3 years to a lifetime, use of these medications in all patients with COPD would be cost-effective. However, there is little empiric evidence to support this assumption; therefore, caution should be exercised when interpreting data from the lifetime model. Nevertheless, even using the most conservative set of assumptions (i.e., no mortality benefit and a 3-year time horizon), use of inhaled corticosteroids in patients with stage 2 or 3 disease would be as cost-effective as instituting tissue plasminogen activator therapy in those presenting with acute myocardial infarction, or using angiotensin-converting enzyme inhibitors following a myocardial infarction (27). Our study had several strengths. First, the natural history of COPD has been well mapped in many large observational studies. Thus, our assumptions are largely supported by very robust data. This enabled us to create a relatively stable economic model to evaluate the role of inhaled corticosteroids in COPD. Second, our economic model met the standards and recommendations of expert panels and committees (26,28), which makes our findings not only valid but also likely to be generalizable to other groups of patients. The generalizability was further enhanced by use of data from the United States, Canada, and Western European countries. Our study also had limitations. Although robust data were available for most key variables, there were a few variables for which we did not have high-quality estimates. For instance, we could not find a standard agreement on how best to classify exacerbations. As a result, our ability to obtain measures of cost per exacerbation March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume

44 Cost-effectiveness of Inhaled Steroids in COPD/Sin et al and exacerbation rates (by severity of exacerbations) was limited. Similarly, there was limited information on rates of exacerbation within each stage of disease, especially for stage 1 disease. Thus, we could not determine if inhaled corticosteroids might be a cost-effective strategy for certain subgroups within stage 1 disease. Also, our model was limited to a 3-year time horizon because of a paucity of clinical efficacy data on inhaled corticosteroids beyond this time period. As with other accepted therapies, if the clinical benefits persisted beyond 3 years, inhaled corticosteroid therapy may be even more cost-effective than shown in our study. On the other hand, there is growing concern that long-term use may be associated with the development of cataracts and osteoporosis (29,30), which may diminish the appropriateness of these medications. We did not include the costs of these potential adverse effects, as we considered that these events were unlikely to occur within the 3-year time frame. Moreover, there is still little consensus on whether inhaled corticosteroids are in fact associated with these adverse effects or whether the associations found in previous observational studies merely reflected confounding by indication (31 33). Despite the ongoing controversy over the efficacy of inhaled corticosteroids in clinical practice (34,35), they are being used increasingly (23). With the expected rise in the prevalence and severity of COPD over the next 20 years (2 4), the cost-effectiveness of these medications is an important consideration. Our study demonstrated that use of inhaled corticosteroids in patients with stage 2 or 3 disease is associated with a cost-effectiveness that is similar to that for other accepted medical therapies. Indeed, certain subgroups of patients, such as those with increased disease severity, may benefit from such therapy. REFERENCES 1. Hurd S. The impact of COPD on lung health worldwide: epidemiology and incidence. Chest. 2000;117(suppl 2): Murray CJ, Lopez AD. Alternative projections of mortality and disability by cause : Global Burden of Disease Study. Lancet. 1997;349: Michaud CM, Murray CJ, Bloom BR. Burden of disease implications for future research. JAMA. 2001;285: Pauwels RA, Buist AS, Calverley PM, et al. Global strategy for the diagnosis, management, and prevention of chronic obstructive pulmonary disease. NHLBI/WHO Global Initiative for Chronic Obstructive Lung Disease (GOLD) Workshop summary. Am J Respir Crit Care Med. 2001;163: The Lung Health Study Research Group. Effect of inhaled triamcinolone on the decline in pulmonary function in chronic obstructive pulmonary disease. N Engl J Med. 2000;343: Alsaeedi A, Sin DD, McAlister FA. The effects of inhaled corticosteroids in chronic obstructive pulmonary disease: a systematic review of randomized placebo-controlled trials. Am J Med. 2002;113: Burge PS, Calverley PM, Jones PW, et al. Randomised, double blind, placebo controlled study of fluticasone propionate in patients with moderate to severe chronic obstructive pulmonary disease: the ISOLDE trial. BMJ. 2000;320: American Thoracic Society. Standards for the diagnosis and care of patients with chronic obstructive pulmonary disease. Am J Respir Crit Care Med. 1995;152(suppl):S77 S Sin DD, Stafinski T, Ng YC, et al. The impact of chronic obstructive pulmonary disease on work loss in the United States. Am J Respir Crit Care Med. 2002;165: Anthonisen NR, Connett JE, Kiley JP, et al. Effects of smoking intervention and the use of an inhaled anticholinergic bronchodilator on the rate of decline of FEV1. The Lung Health Study. JAMA. 1994;272: Calverley P, Pauwels R, Vestbo J, et al. Combined salmeterol and fluticasone in the treatment of chronic obstructive pulmonary disease: a randomised controlled trial. Lancet. 2003;361: Spencer S, Calverley PM, Burge PS, et al. Inhaled steroids in obstructive lung disease. Health status deterioration in patients with chronic obstructive pulmonary disease. Am J Respir Crit Care Med. 2001;163: Aaron SD, Vandemheen KL, Clinch JJ, et al. Measurement of shortterm changes in dyspnea and disease-specific quality of life following an acute COPD exacerbation. Chest. 2002;121: Paggiaro PL, Dahle R, Bakran I, et al. Multicentre randomised placebo-controlled trial of inhaled fluticasone propionate in patients with chronic obstructive pulmonary disease. International COPD Study Group. Lancet. 1998;351: McBride A, Milne R. Hospital-Based Pulmonary Rehabilitation Programmes For Patients with Severe Chronic Obstructive Pulmonary Disease. Southampton, United Kingdom: Wessex Institute for Health Research and Development; Development and Evaluation Committee Report No United States Congress Office of Technology Assessment. Evaluation of the Oregon Medicaid Proposal. Washington, DC: U.S. Government Printing Office; Report No. OTA-H Vestbo J, Sorensen T, Lange P, et al. Long-term effect of inhaled budesonide in mild and moderate chronic obstructive pulmonary disease: a randomised controlled trial. Lancet. 1999;353: Niewoehner DE, Erbland ML, Deupree RH, et al. Effect of systemic glucocorticoids on exacerbations of chronic obstructive pulmonary disease. Department of Veterans Affairs Cooperative Study Group. N Engl J Med. 1999;340: Alberta Health and Wellness. Medical Benefits Price List Available at: Price_List.pdf. Accessed October 8, Alberta Health and Wellness Drug Benefit List Available at: Accessed October 8, Sin DD, Tu JV. Outpatient antibiotic therapy and short term mortality in elderly patients with chronic obstructive pulmonary disease. Can Respir J. 2000;7: Alberta Health and Wellness. Health Costing in Alberta: 2001 Annual Report. Available at: document/health_costing_2001.pdf. Accessed October 8, Jackevicius CA, Chapman KR. Prevalence of inhaled corticosteroid use among patients with chronic obstructive pulmonary disease: a survey. Ann Pharmacother. 1997;31: Statistics Canada. National Population Health Survey. Available at: Accessed October 8, Anthonisen NR, Wright EC, Hodgkin JE. Prognosis in chronic obstructive pulmonary disease. Am Rev Respir Dis. 1986;133: March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume 116

45 Cost-effectiveness of Inhaled Steroids in COPD/Sin et al 26. Guidelines for Economic Evaluation of Pharmaceuticals: Canada. 2nd ed. Ottawa, Canada: Canadian Coordinating Office for Health Technology Assessment (CCOHTA); Chapman RH, Stone PW, Sandberg EA, Bell C, Neumann PJ. A comprehensive league table of cost-utility ratios and a subtable of panel-worthy studies. Med Decis Making. 2000;20: Gold MR, Siegel JE, Russell LB, Weinstein MC, eds. Cost-Effectiveness in Health and Medicine. New York, New York: Oxford University Press; Cumming RG, Mitchell P, Leeder SR. Use of inhaled corticosteroids and the risk of cataracts. N Engl J Med. 1997;337: Baltzan MA, Suissa S, Bauer DC, Cummings SR. Hip fractures attributable to corticosteroid use. Study of Osteoporotic Fractures Group. Lancet. 1999;353: Lau E, Mamdani M, Tu K. Inhaled or systemic corticosteroids and the risk of hospitalization for hip fracture among elderly women. Am J Med. 2003;114: Jones A, Fay JK, Burr M, et al. Inhaled corticosteroid effects on bone metabolism in asthma and mild chronic obstructive pulmonary disease. Cochrane Database Syst Rev. 2002;1:CD Delcourt C, Cristol JP, Tessier F, et al. Risk factors for cortical, nuclear, and posterior subcapsular cataracts: the POLA study. Pathologies Oculaires Liees a l Age. Am J Epidemiol. 2000;151: Calverley PM. Inhaled corticosteroids are beneficial in chronic obstructive pulmonary disease. Am J Respir Crit Care Med. 2000;161: Barnes PJ. Inhaled corticosteroids are not beneficial in chronic obstructive pulmonary disease. Am J Respir Crit Care Med. 2000;161: March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume

46 REVIEWS An Approach to the Diagnosis and Treatment of Cryofibrinogenemia Tshering D. Amdo, MD, James A. Welker, DO Cryofibrinogenemia is a rarely symptomatic disorder that is underrecognized due to the infrequency with which it causes symptoms. Although completely reversible, this disorder can be life threatening when untreated. In this review, the classification, pathophysiology, and clinical presentation of cryofibrinogenemia are described, based on case reports and prospective observational data. Diagnostic criteria are outlined, and therapies are assessed critically. This information should help clinicians in establishing a diagnosis of cryofibrinogenemia and initiating treatment. Am J Med. 2004;116: by Excerpta Medica Inc. Patients with cutaneous ulcers and gangrene have a limited differential diagnosis that includes essential cryofibrinogenemia, a rare disorder characterized by cryoprecipitation of the patients native fibrinogen, which causes thrombotic occlusion of the small to medium arteries. When recognized early, cryofibrinogenemia is a treatable and completely reversible disease. The goal of this paper is to increase awareness of essential and secondary cryofibrinogenemia, describe a typical presentation of this syndrome, and provide an approach to diagnosis and treatment. METHODS Articles were obtained via a MEDLINE search, interview of two experts, and review of references obtained from articles. Searches were limited to human studies written in English. Search terms included cryofibrinogenemia, cryoprotein, plasmapheresis, and cryofiltration. The search results yielded case reports, case series, and observational data. Due to the infrequency of symptomatic disease, there are no randomized trials. A total of 46 articles were obtained, 21 of which were eliminated due to lack of new data. All obtained articles were assessed by two reviewers who were not blinded to author, institution, or journal. Interpretation was established by consensus. From the Hospitalist Section, Department of Internal Medicine, Harbor Hospital Center, Baltimore, Maryland. Requests for reprints should be addressed to James A. Welker, DO, Hospitalist Section, Department of Internal Medicine, Harbor Hospital Center, 3001 S. Hanover Street, Baltimore, Maryland 21225, or jim.welker@medstar.net. Manuscript submitted January 14, 2003, and accepted in revised form September 11, All reported treatments have been critically assessed. Treatments were considered indicated if the symptoms were relieved with the treatment, recurred with cessation of treatment, and relieved again with subsequent initiation of treatment. CLASSIFICATION Cryofibrinogenemia has been classified into an essential (primary) and a secondary form. Clinically relevant essential cryofibrinogenemia is rare and its prevalence is not known, although one study reported a prevalence of 3% in 135 healthy residents in Oklahoma City, Oklahoma (1). Two epidemiological studies estimated the prevalence of secondary cryofibrinogenemia in patients without symptoms of cryofibrinogenemia who were hospitalized for another illness (1,2). One study, which involved 36,000 hospitalized patients between the ages of 15 and 100 years (2), reported a prevalence of 3.4%. However, the blood samples were maintained at 4 C for only 24 hours prior to determining the presence of cryofibrinogens. In a study of 665 hospitalized patients in which samples were observed for 48 hours, the prevalence was 13% (1). Since most authors agree that the sensitivity of this assay is improved by monitoring the specimen 72 hours for cryofibrinogens, 13% is more likely to be an accurate representation of the prevalence of secondary cryofibrinogenemia. Unfortunately, the epidemiologic data for both essential and secondary cryofibrinogenemia are quite old, and it is not known whether they are representative of patients today. Changes in the composition of hospitalized patients, such as a decreasing incidence of infectious diseases among immunocompetent patients and an increasing number of immunocompromised patients, is by Excerpta Medica Inc /04/$ see front matter All rights reserved. doi: /j.amjmed

47 An Approach to the Diagnosis and Treatment of Cryofibrinogenemia/Amdo and Welker Figure 1. Photograph of the feet showing edema and bullae. expected to modify the incidence of cryofibrinogenemia. This supports the need for further epidemiologic studies. There are limited reports suggesting a familial predilection for the disease (3 5). Presently, there seems to be insufficient information to confirm or delineate a pattern of inheritance. PATHOPHYSIOLOGY Cryofibrinogen is the term that was used by Korst and Kratochvil in 1955 to describe an abnormal, cold, precipitable protein (6). The substance is a cold, insoluble complex of fibrin, fibrinogen, and fibrin split products with albumin, plasma proteins, and immunoglobulins. Cryofibrinogen clots with thrombin and reversibly precipitates in the plasma on cooling to 4 C, then redissolves on warming to 37 C. Serum is the fluid remaining after plasma is allowed to clot. The proteins consumed in the clotting process are the necessary substrates for cryofibrinogens. Therefore, unlike cryoglobulins, which precipitate in both cooled plasma and serum, cryofibrinogens do not precipitate in cooled serum. The pathogenesis of cryofibrinogenemia is not known. The most plausible hypothesis is based on the high plasma levels of the protease inhibitors 1- antitrypsin and 2 -macroglobulin that have been found in patients with cryofibrinogenemia (2). These substances inhibit the fibrinolytic agent plasmin, thereby inhibiting fibrinolysis. The result is the accumulation of cryofibrinogen, which clots with thrombin and leads to the thrombotic occlusion of small and medium arteries. Additional vascular occlusion may be caused by the development of reflex vasospasm, vascular stasis, and hyperviscosity. When this occurs in end arteries, such as distal extremities, tissue ischemia and gangrene develop. CLINICAL PRESENTATION Essential cryofibrinogenemia develops spontaneously in previously healthy persons. Too few cases have been reported, however, to determine the clinical presentation by patient characteristics. Secondary cryofibrinogenemia occurs with a female to male ratio of 1.4 to 1, but with no age or racial predilection (2). These patients suffer from an underlying inflammatory disease, such as a malignancy, diabetes mellitus, collagen vascular disease, or active infection. They live in colder climates and report a temporal association between exposure to cold and the onset of symptoms. The most common symptoms are due to cutaneous ischemia, and include purpura, livido reticularis, ecchymosis, ulcerations, ischemic necrosis, and, less often, gangrene. Although any area of the body can be affected, areas that maintain a lower temperature and that are more sensitive to further cooling upon exposure to cold are more often affected. These include the hands, feet, ears, nose, and buttocks (Figure 1). Constitutional symptoms such as malaise and fever are common, and are due to the cytokine release associated with tissue ischemia and death. The presence of secondary cryofibrinogens is associated with increased mortality; however, death is not directly attributable to cryofibrinogenemia (2,7). The most directly associated cause of death is sepsis resulting from the secondary infection of gangrenous cutaneous tissue. March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume

48 An Approach to the Diagnosis and Treatment of Cryofibrinogenemia/Amdo and Welker Thrombotic events occur in 25% of patients with secondary cryofibrinogenemia, irrespective of the amount of cryofibrinogen in serum (1). Although thrombotic cutaneous events are directly attributable to cryofibrinogens, the cause and effect relation of cryofibrinogens with other thrombotic events is not known. These other infrequently reported thrombotic events include cerebrovascular thrombus; myocardial infarct; thrombophlebitis; obstruction of the aorta, iliac, or femoral artery; pulmonary emboli; mesenteric artery thrombosis; and retinal artery thrombosis. Similar to disseminated intravascular coagulation, paradoxical spontaneous bleeding may occur due to the depletion of clotting factors. Bleeding events have been found to be directly proportional to the amount of cryofibrinogen in serum, and have been documented in 45% of patients with heavy levels as compared with 22% of those with trace levels. The combination of thrombosis and bleeding has been only seen in 5% of patients with heavy concentrations (1). (The terms heavy and trace in reference to cryofibrinogen levels or concentrations are defined below in the section on diagnosis.) Procedures that include cooling have been complicated by cryofibrinogenemia. Complications have been seen with superficial cryosurgery, cooling blankets, and the generalized body cooling that occurs during prolonged surgical procedures (5,8,9). There has been one report of dramatic complications during cosmetic superficial cryosurgery, which required the patient to undergo numerous treatments to improve the acquired facial deformity (9). The authors recommend assessing all patients for cryofibrinogenemia and testing cryosurgery on inconspicuous cutaneous locations prior to performing the definitive treatment (9). DIAGNOSIS Table 1. Differential Diagnosis for Essential Cryofibrinogenemia Secondary cryofibrinogenemia Cryoglobulinemia Peripheral vascular disease Frostbite Thrombotic thrombocytopenia purpura Disseminated intravascular coagulation Coumarin necrosis Hereditary hypercoagulable states Embolic diseases, such as endocarditis or cholesterol emboli Vasculitis Calciphylaxis in end-stage renal disease Antiphospholipid antibody syndrome Purpura fulminans The diagnosis of cryofibrinogenemia should be considered in all previously healthy persons presenting with unexplained areas of tissue ischemia and gangrene. The differential diagnosis includes cryoglobulinemia, peripheral vascular disease, frostbite, thrombotic thrombocytopenia purpura, disseminated intravascular coagulation, coumarin necrosis, hereditary hypercoagulable states, antiphospholipid antibody syndrome, embolic diseases such as endocarditis or cholesterol emboli, vasculitis, calciphylaxis in end-stage renal disease, and purpura fulminans (Table 1). Diagnostic criteria for essential cryofibrinogenemia have been developed based on the published literature and clinical experience (Table 2). The evidence has been divided into that which is essential and that which is supportive. The essential evidence, which is required to be present prior to making the diagnosis, includes the appropriate clinical presentation as described above, the presence of cryofibrinogens in the plasma, the absence of cryoglobulins in serum, and the exclusion of secondary causes of cryofibrinogenemia as well as other vaso-occlusive diseases. The supportive evidence is nonspecific and not compulsory, but when present along with the essential evidence, improves the accuracy of the diagnosis. These include an angiogram displaying the abrupt occlusion of small to medium arteries (Figure 2), the appropriate biopsy findings of the involved tissue as described below, and elevated serum levels of 1 -antitrypsin and 2 -macroglobulin. The diagnostic criteria for secondary cryofibrinogenemia are the same, except that the exclusion of secondary causes of cryofibrinogens is removed from the essential evidence. The detection of cryofibrinogens requires special attention, is often done poorly by reference laboratories, and should be performed by the clinicians who suspect the diagnosis. If the sample is not centrifuged immediately, it must be kept at 37 C to prevent autoabsorption of cryofibrinogens by the red blood cells. Centrifugation that is delayed or subsequent to sample cooling leads to Table 2. Diagnostic Criteria for Essential Cryofibrinogenemia Essential Evidence Appropriate clinical presentation: sudden onset of skin changes and constitutional symptoms, with or without thrombosis, bleeding, or exposure to cold Presence of cryofibrinogen in the plasma Absence of cryoglobulins No secondary causes of cryofibrinogens and no evidence of other vaso-occlusive disease Supportive Evidence Angiogram with abrupt occlusion of small to medium arteries Typical skin biopsy findings: cryofibrinogen plugging vessels, leukocytoclastic vasculitis, or dermal necrosis Elevation of serum levels of 1 -antitrypsin and 2 - macroglobulin 334 March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume 116

49 An Approach to the Diagnosis and Treatment of Cryofibrinogenemia/Amdo and Welker Figure 2. Arteriogram of both feet demonstrating complete obstruction of the small and medium arteries beyond the midmetatarsal level. discarding of cryofibrinogens with the red blood cells, and reporting of false-negative results. Blood should be collected in tubes containing oxalate, citrate, or ethylenediaminetetraacetic acid as the anticoagulant. Tubes containing heparin should not be used because of the likelihood of false-positive results due to the formation of a cryoprecipitate containing fibronectin, fibrin, and fibrinogen in combination with heparin, known as heparin precipitable fraction (1). After collection, the blood should be stored at 37 C until centrifuged. After centrifugation, the plasma should be stored at 4 C for 72 hours. Cryofibrinogens will develop between 24 and 72 hours after cooling (Figure 3). Figure 3. Initial plasma with cryofibrinogen (left); initial serum without cryoglobulins (middle); plasma after streptokinase showing decreased cryofibrinogens (right). Cryofibrinogens can be quantitated in two ways. In the first method, none represents no precipitate; heavy indicates a precipitate greater than 100 mg percent; and trace indicates a precipitate present in a quantity less than 100 mg percent (1). In the second method, which refers to the approximate percent volume of plasma supernate containing precipitate, 15% to 25% equals 1, 50% equals 2, 75% equals 3, and 100% equals 4 (2). Simultaneously, a serum sample should be prepared by collecting blood in a tube free of anticoagulant, allowing the blood to clot, and then centrifuging. This serum sample should be cooled in the same manner as for cryofibrinogens to demonstrate the absence of precipitable cryoglobulins. Cryoglobulins form in both the serum and plasma, thereby hindering the ability to identify cryofibrinogens accurately and necessitating their exclusion prior to making the diagnosis of cryofibrinogenemia. In addition, Western blot analyses to purify the cryoprecipitate components suggest that cryoglobulins most likely promote the formation of cryofibrinogens. This technique demonstrates that 62% of cryofibrinogens occur in isolation, whereas 70% of cryoglobulins occur together with cryofibrinogens (10). This promotion of cryofibrinogen production indicates that treatment should be directed toward the cryoglobulins when both entities are present. Therefore, visual quantitation of precipitate accurately identifies the presence of cryofibrinogens and cryoglobulins in a manner that allows treatment decisions to be made quickly and inexpensively. Numerous patients with cryofibrinogenemia have had to undergo skin biopsies. The most specific finding is the plugging of the superficial and deep blood vessels with thrombi containing cryofibrinogen, which stains eosinophilic with hematoxylin-eosin stain and purple-red positive with periodic acid Schiff stain (11,12). This finding is only seen in patients with thrombotic diseases, including thrombotic thrombocytopenic purpura, disseminated intravascular coagulation, coumarin necrosis, protein C and protein S deficiency, or monoclonal cryoglobulinemia, or the presence of lupus anticoagulant. The nonspecific findings of leukocytoclastic vasculitis and necrosis of the dermis and epidermis are more common. All of these findings are absent in asymptomatic cases of cryofibrinogenemia; when present, they are consistent with a broad differential diagnosis. This lack of sensitivity and specificity relegates histology to the status of a secondary criterion. Still, a timely biopsy can provide support for the diagnosis, and should be performed in patients in whom the diagnosis is uncertain. It is essential that the biopsy be performed prior to the development of necrosis. The presence of elevated serum levels of 1 -antitrypsin and 2 -macroglobulin is less well established. These proteins were found to be elevated in a subgroup of 11 randomly selected patients with elevated plasma cryofi- March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume

50 An Approach to the Diagnosis and Treatment of Cryofibrinogenemia/Amdo and Welker Figure 4. Serum protein electrophoresis demonstrating elevated 2 protein levels. A/G albumin:globulin ratio. brinogen levels but no symptoms of the disease (2). Of these proteins, 2 -macroglobulin seems to be elevated more frequently (13). Screening for these proteins can be performed with serum protein electropheresis (Figure 4). TREATMENT The level of the evidence on which treatment decisions for cryofibrinogenemia are based is limited to case reports. Avoidance of cold exposure and placing the symptomatic patient in an environment of about 37 C is a reasonable, inexpensive measure that is without risk, and that has been found to be partially efficacious. Antiseptic wound care is imperative. Cutaneous lesions should be managed in the same way as for any gangrenous soft tissue injury or burn. All patients require lifelong monitoring because symptoms have a tendency to recur, especially with the cessation of therapy. The most commonly used pharmaceutical treatments include stanozolol, streptokinase, and steroids. A dose of 50,000 to 80,000 units of the oral fibrinolytic agent streptokinase was the first reported successful treatment (12). In patients with essential cryofibrinogenemia in whom the diagnosis was well established, the disease was observed to remit when streptokinase was administered, recur when treatment was discontinued, and then remit again when the medication was initiated. In the United States, the oral form is unavailable; hence, streptokinase must be administered intravenously. An intravenous dose as low as 25,000 units and given every 24 hours should be sufficient, since the orally administered form does not enter the circulation in sufficient amounts to produce a measurable thrombolytic effect. Based on this evidence, streptokinase at a daily dose of either 50,000 to 80,000 units orally or 25,000 to 200,000 units intravenously is indicated in the treatment of cryofibrinogenemia. Stanozolol is also indicated in the treatment of cryofibrinogenemia. Successful treatment with stanozolol (2 to 4 mg administered orally, twice daily) has been described in several case reports (14 17). Stanozolol has been noted to have an onset of action of several days, which limits its usefulness in patients suffering from an acute onset of severe symptoms (15). As a testosterone derivative, it is an androgenic steroid that has an established fibrinolysis-enhancing effect, which is believed to result from its alteration of the synthesis of liver proteins involved in the fibrinolytic process. However, it is associated with adverse effects, such as sodium retention, hirsutism, acne, liver function abnormalities, and hyperlipidemia. Based on the available evidence, steroids alone do not appear to be a valid treatment, but may be appropriate in combination with immunosuppressants. The majority of case reports do not demonstrate a beneficial effect with high-dose steroids (10,17,18). However, reports in which the diagnosis of essential cryofibrinogenemia was well established demonstrated a benefit with prednisone (60 mg/d) in combination with azathioprine (150 mg/d) (10,11), and with prednisone (10 mg/d) in combination with chlorambucil (4 mg/d) (19). Azathioprine treatment has been shown to be effective. In contrast, cyclophosphamide alone has not been shown to be beneficial (10). Steroids should be used as needed for the treatment of an underlying inflammatory condition (e.g., collagen vascular diseases), which can help in the control of cryofibrinogenemia. There is limited evidence to support the use of plasmapheresis. This technique was shown to be effective in one case report in which the patient s diagnosis was well established (20). However, other case reports have been complicated by the absence of cryofibrinogens in plasma and poor outcomes (10,21,22). In one instance, a patient had a documented decrease in cryofibrinogens, but the cutaneous ulcers never healed and became infected, and the patient died of sepsis (22). Several articles have suggested that cryofiltration apheresis might be effective (23 25). However, there have been no published reports of the use of this technique in patients with cryofibrinogenemia, and this recommendation is based on the successful treatment of patients with cryoglobulinemia. Furthermore, the success of other treatments vary for each of these two diseases, and an unexpected lack of benefit of cryofiltration apheresis has been observed in cold agglutinin disease (23). At this time, plasmapheresis is a recommended treatment, whereas cryofiltration apheresis should be considered an unproven albeit potentially beneficial therapy. The cost and limited availability of these modalities make them a second-line therapy at most hospitals. 336 March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume 116

51 An Approach to the Diagnosis and Treatment of Cryofibrinogenemia/Amdo and Welker The use of heparin is not supported by the literature. Indeed, symptoms have been found to progress with use of heparin (10,19). Warfarin is potentially beneficial. Precursors to warfarin (bishydroxycourmin) have been used successfully. However, subsequent treatment with warfarin in patients with a well-established diagnosis of cryofibrinogenemia has had both successful and unsuccessful outcomes (10,18). One trial found Dextran 40 to be beneficial (18). However, it is the only report of this treatment, and neither dosage nor duration was noted. Aspirin and colchicine have been studied repeatedly and found to have no apparent benefit, and thus should not be used (10). We treated a patient with amlodipine (5 mg administered once daily), based on the principle that vasospasm may be involved in the pathogenesis of the disease, and observed that the patient s symptoms progressed. Amlodipine is thus not recommended. CONCLUSION Cryofibrinogenemia is an underrecognized and unnecessarily life-threatening disease. The information provided in this paper should help clinicians to become more efficient at establishing a diagnosis and initiating treatment. ACKNOWLEDGMENT We would like to thank Robert Marcus, MD, for his expertise in rheumatology and special interest in cryofibrinogenemia; Sameer Bade, MD, whose images and posters brought clarity to this project s presentations; and Satish Chandra, MD, for his support with radiological studies. REFERENCES 1. McKee PA, Kalbfleisch JM, Bird RM. Incidence and significance of cryofibrinogenemia. J Lab Clin Med. 1963;61: Smith SB, Arkin C. Cryofibrinogenemia: incidence, clinical correlations, and a review of the literature. Am J Clin Pathol. 1972;58: Wulffraat N, Meyer KJ, Zegers BJ, Kuis W. Familial presence of primary cryofibrinogenaemia, a report of three cases. Br J Rheumatol. 1996;35: van Geest AJ, van Dooren-Greebe RJ, Andriessen MPM, et al. Familial primary cryofibrinogenemia. J Eur Acad Dermatol Venereol. 1999;12: Lolin Y, Razis PA, Gorman PO, et al. Transient nephrotic syndrome after anesthesia resulting from a familial cryofibrinogen precipitating at 35 C. J Med Genet. 1989;26: Korst DR, Kratochvil CH. Cryofibrinogen formation in case of lung neoplasm associated with thrombophlebitis migrans. Blood. 1955; 10: Goodall HB, Todd AS, Maclean D, et al. Proceedings: cryofibrinogenaemia and activation of the coagulation/lysis systems in accidental hypothermia of the elderly. J Clin Pathol. 1975;28: Waxman S, Dove JT. Cryofibrinogenemia aggravated during hypothermia. N Engl J Med. 1969;281: Stewart RH, Graham GF. Cryo corner: a complication of cryosurgery in a patient with cryofibrinogenemia. J Dermatol Surg Oncol. 1978;4: Blain H, Cacoub P, Musset L, et al. Cryofibrinogenaemia: a study of 49 patients. Clin Exp Immunol. 2000;120: Beightler E, Diven DG, Sanchez RL, Solomon AR. Thrombotic vasculopathy associated with cryofibrinogenemia. J Am Acad Dermatol. 1991;24: Rachmilewitz EA, Sacks MI, Zlotnick A. Essential cryofibrinogenemia: clinical, pathological and immunological studies. Isr J Med Sci. 1970;6: Martin S. Cryofibrinogenemia, monoclonal gammopathy, and purpura. Report of a case and review of the literature. Arch Dermatol. 1979;115: Kirsner RS, Eaglstein WH, Katz MH, et al. Stanozolol causes rapid pain relief and healing of cutaneous ulcers caused by cryofibrinogenemia. J Am Acad Dermatol. 1993;28: Revenga F, Aguilar C, Gonzalez R, et al. Cryofibrinogenaemia with a good response to stanozolol. Clin Exp Dermatol. 2000;25: Rubegni P, Flori ML, Fimiani M, Andreassi L. A case of cryofibrinogenaemia responsive to stanozolol. Br J Haematol. 1996;93: Williamson AE, Cone LA, Huard S. Spontaneous necrosis of the skin associated with cryofibrinogenemia, cryoglobulinemia, and homocystinuria. Ann Vasc Surg. 1996;10: Ball GV, Goldman LN. Chronic ulcerative colitis, skin necrosis, and cryofibrinogenemia. Ann Intern Med. 1976;85: Zouboulis CC, Gollnick H, Weber S, et al. Intravascular coagulation necrosis of the skin associated with cryofibrinogenemia, diabetes mellitus, and cardiolipin autoantibodies. J Am Acad Dermatol. 1991;25: Euler HH, Zeuner RA, Beress R, et al. Monoclonal cryo-antifibrinogenemia. Arthritis Rheum. 1996;39: Copeman PW. Cryofibrinogenaemia and skin ulcers: treatment with plasmapheresis. Br J Dermatol. 1979;101: Sankarasubbaiyan S, Scott G, Holley JL. Cryofibrinogenemia: an addition to the differential diagnosis of calciphylaxis in end-stage renal disease. Am J Kidney Dis. 1998;32: Siami GA, Siami FS. Cryofiltration apheresis in the United States. Ther Apher. 1998;2: Siami FS, Siami GA. Cryofiltration apheresis in the treatment of cryoprecipitate induced diseases. Ther Apher. 1997;1: Siami FS, Siami GA. Plasmapheresis and paraproteinemia: cryoprotein-induced diseases, monoclonal gammopathy, Waldenstrom s macroglobulinemia, hyperviscosity syndrome, multiple myeloma, light chain disease, and amyloidosis. Ther Apher. 1999;3: March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume

52 Leukotriene Receptor Antagonists for Allergic Rhinitis: A Systematic Review and Meta-analysis Andrew M. Wilson, MD, Paul M. O Byrne, MB, Krishnan Parameswaran, MD PURPOSE: To compare the clinical efficacy of leukotriene receptor antagonists with that of placebo, antihistamines, and nasal corticosteroids in patients with allergic rhinitis or nasal polyposis. METHODS: We performed a systematic review and metaanalysis of randomized controlled trials of the effectiveness of leukotriene receptor antagonists in patients with rhinitis. Composite daily rhinitis symptom scores (as a percentage of the maximum score) and rhinitis-specific quality of life (unit scores ranging from 0 to 6) were pooled after assessing heterogeneity among studies. The pooled estimates were expressed as weighted mean differences between treatments in a randomeffects model. We considered a difference of 10% in nasal score and 0.6 units in quality-of-life score to be clinically relevant. RESULTS: Of the 196 citations, 11 studies on seasonal allergic rhinitis were used in the analysis: eight evaluating leukotriene receptor antagonists alone or in combination with other treatments versus placebo or other treatments (n 3924) and three evaluating leukotriene receptor antagonists plus an antihistamine (n 80). Leukotriene receptor antagonists reduced mean daily rhinitis symptom scores (in absolute terms) 5% (95% confidence interval [CI]: 3% to 7%) more than did placebo. However, antihistamines improved the nasal symptoms score 2% (95% CI: 0% to 4%) more than did leukotriene receptor antagonists, and nasal corticosteroids improved the score 12% (95% CI: 5% to 18%) more than did leukotriene antagonists. Leukotriene receptor antagonists significantly improved rhinoconjunctivitis quality of life by 0.3 units (95% CI: 0.24 to 0.36 units) when compared with placebo. There were no randomized controlled trials evaluating the effect of leukotriene receptor antagonists on perennial allergic rhinitis or polyposis. CONCLUSION: Leukotriene receptor antagonists are modestly better than placebo, as effective as antihistamines, but less effective than nasal corticosteroids in improving symptoms and quality of life in patients with seasonal allergic rhinitis. Am J Med. 2004;116: by Excerpta Medica Inc. Allergic rhinitis is a common inflammatory condition of the upper airways (1). Although rarely responsible for hospitalization, it reduces quality of life (2), especially when associated with polyposis (3), and can be a considerable cost to society (4 6). Allergic rhinitis is also associated with other medical conditions such as sinusitis and asthma (7). Cysteinyl leukotrienes are involved in the pathophysiology of asthma, as evidenced by elevated levels in bronchoalveolar lavage (8) and induced sputum (9), when compared with levels in healthy volunteers at baseline and following exercise (10) or allergen challenge (11). It is not surprising, therefore, that leukotriene receptor antagonists are used in the management of asthma. However, they are not as effective as inhaled corticosteroids (12). Patients with allergic rhinitis have been shown to have elevated levels of leukotriene, both during the pollen season or following allergen (13) or aspirin challenge (14). There is also evidence to suggest that leukotriene receptor antagonists are beneficial in rhinitis; indeed, they have From the Firestone Institute for Respiratory Health, St. Joseph s Healthcare, and Department of Medicine, McMaster University, Hamilton, Ontario, Canada. Dr. Parameswaran is supported by a postdoctoral fellowship from the Canadian Institutes of Health Research. Requests for reprints should be addressed to Krishnan Parameswaran, MD, Firestone Institute for Respiratory Health, St. Joseph s Healthcare, 50 Charlton Avenue East, Hamilton, Ontario, L8N 4A6, Canada, or parames@mcmaster.ca. Manuscript submitted April 3, 2003, and accepted in revised form September 19, recently been licensed in the United States for this indication. We conducted a systematic review and meta-analysis to assess the effectiveness of leukotriene receptor antagonists, in comparison with that of antihistamines and nasal corticosteroids, in patients with rhinitis. As histamine is another important inflammatory mediator in allergic rhinitis, we also evaluated the benefits of therapy that combined leukotriene receptor antagonists and histamine receptor antagonists. METHODS We searched MEDLINE (1966 to February 2003), EM- BASE (1980 to February 2003), and Cumulated Index to Nursing and Allied Health (CINAHL) (1956 to February 2003) for original articles published in any language, using the following terms: rhinitis, allergic rhinitis, seasonal rhinitis, perennial rhinitis, nasal polyp, leukotriene receptor antagonist, montelukast, zafirlukast, pranlukast, cinalukast, pomilukast, Singulair, Accolate, and Onon. The reference lists of relevant primary studies and review articles were searched for articles of potential interest. Potentially relevant articles were independently identified using the title, abstract, keywords, and journal type by two reviewers (KP, AMW), and the full text of all potentially relevant articles were independently reviewed for inclusion. Since poor trial designs such as inadequate allocation concealment can overestimate the treatment effect, the reviewers assessed methodological quality using by Excerpta Medica Inc /04/$ see front matter All rights reserved. doi: /j.amjmed

53 Leukotriene Antagonists for Allergic Rhinitis/Wilson et al a scale described by Jadad et al (15). The score ranges from 0 to 5, based on randomization, blinding, allocation concealment, withdrawals, and dropouts. The two reviewers independently extracted data from the text, tables, or figures of included trials. Data Analysis Agreement between reviewers in selecting studies to be pooled was calculated using the test. The primary outcome measure was the composite nasal symptom score. A symptom score index was calculated as the composite daytime nasal symptom score expressed as a percentage of the maximum possible symptom score (higher score indicating worse symptoms). In studies in which the daytime nasal symptom score was not reported separately (16 20), the 24-hour symptom score was used to calculate the composite score. It was possible to obtain data for daytime nasal symptoms from the raw data from a further study in which these data were collected but not reported (21). All such studies used the same metric to report daytime and nighttime nasal symptoms; they collected data on the same components of nasal symptoms (e.g., nasal blockage, itching, sneezing, and rhinorrhea). The pooled standard deviation was calculated, as the square root of the average of the variance of the two samples, where appropriate. The secondary outcome was standardized rhinoconjunctivitis quality of life, expressed as a unit score ranging from 0 (no symptoms) to 6 (worst symptoms) (22). The mean scores at the end of treatment were calculated from the data provided in tables and graphs. Weighted means and 95% confidence intervals for both outcomes were calculated using standard metaanalysis software (RevMan 4.1; Update Software Ltd., Oxford, United Kingdom), which uses the inverse of the variance to assign a weight to the mean of the withinstudy treatment effect. This was estimated in terms of percentage difference for symptom score and absolute units for quality of life. Summary estimates of the treatment effect were calculated using a random-effects model and reported as weighted mean differences (23). For example, if one treatment reduces nasal symptoms by 10% and the second treatment reduces nasal symptoms by 8%, the weighted mean difference is 2% in favor of the first treatment. A chi-squared test for heterogeneity (Breslow- Day) was performed to test whether the distribution of the results was compatible with the assumption that intertrial differences were attributable to chance alone. Since trials with a Jadad score 3 have been reported to show consistently greater effects of treatment than those with a score 3, we performed sensitivity analyses that used methodological quality (i.e., studies with a score 3 vs. those with a score 3). Publication bias was examined by visual inspection of a funnel plot. Figure 1. Flow chart illustrating the results of the research strategy. RESULTS The search strategy yielded 11 studies for analysis (Figure 1). Eight studies evaluated the efficacy of a leukotriene receptor antagonist (alone or in combination with other therapy) compared with placebo or other therapy for rhinitis (Table 1) and three studies evaluated the combination of a leukotriene receptor antagonist and an antihistamine versus other therapy (Table 2). The two reviewers were in good agreement about the inclusion of articles ( 0.89); all disagreements regarding inclusion were resolved by discussion and third-party adjudication was therefore not required. Leukotriene Receptor Antagonists versus Placebo Leukotriene receptor antagonists improved the composite nasal symptom score 5% (95% confidence interval [CI]: 3% to 7%) more than did placebo (Figure 2). The effects varied significantly among the studies (P 0.01), most likely because the lower-quality studies (Jadad score 3) had variable results. When a subgroup analysis was performed using only studies with a Jadad score 3, the results were similar (weighted mean difference 4%; 95% CI: 3% to 6%) and there was no heterogeneity (P 0.44). In four studies that reported treatment response in terms of standardized rhinoconjunctivitis quality of life (24 26,28), leukotriene antagonists improved quality of life by 0.3 units (95% CI: 0.24 to 0.36 units) compared with placebo. There were insufficient data pertaining to the individual components of nasal symptom score (e.g., March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume

54 Leukotriene Antagonists for Allergic Rhinitis/Wilson et al Table 1. Details of Studies Comparing Leukotriene Receptor Antagonists with Placebo, Antihistamines, or Intranasal Corticosteroids First Author (Reference) Drug Duration Comparison Parallel Pullerits (16) Zafirlukast 20 mg twice daily (n 11) Wilson (20) Montelukast 10 mg once daily ( cetirizine 10 mg) (n 11) Meltzer (24)* Montelukast 10 mg (n 95) or 20 mg (n 90) Nayak (25)* Philip (26)* Pullerits (27) van Adelsberg (28)* Crossover Wilson (19) * Pharmaceutical industry sponsored. Montelukast 10 mg once daily (n 155) Montelukast 10 mg once daily (n 348) Montelukast 10 mg once daily (n 16) Montelukast 10 mg once daily (n 522) Montelukast 10 mg once daily (n 12) 50 days Placebo (n 11); beclomethasone dipropionate 400 g (n 11) 4 weeks Placebo (n 13); mometasone furoate 200 g(n 14) 2 weeks Placebo (n 91); loratadine 10 mg (n 92); montelukast 10 mg loratadine 10 mg (n 90) 2 weeks Placebo (n 149); loratadine 10 mg (n 301); montelukast 10 mg loratadine 10 mg (n 302) 2 weeks Placebo (n 352); loratadine 10 mg (n 602) 50 days Placebo (n 18); fluticasone propionate 200 g(n 13); montelukast 10 mg loratadine 10 mg (n 15) 2 weeks Placebo (n 521); loratadine 10 mg (n 171) 2 weeks Placebo (n 12); inhaled (400 g) and intranasal (200 g) budesonide (n 12) Endpoints In Addition to Nasal Symptoms Jadad Score Nasal biopsy 4 Nasal peak inspiratory flow Quality of life 5 Quality of life, blood eosinophil count Quality of life, blood 4 eosinophil count Nasal biopsy 3 Quality of life, blood eosinophil count, physician score Daily activity, nasal nitric oxide, nasal peak inspiratory flow, asthma symptoms blockage, sneezing, nasal itching, or discharge) to perform a subgroup analysis. Leukotriene Receptor Antagonists versus Intranasal Corticosteroids and Antihistamines Intranasal corticosteroids improved composite nasal symptoms scores 12% (95% CI: 5% to 18%) more than did leukotriene receptor antagonists (Figure 3). The four studies pooled were heterogeneous in terms of their methodological rigor and evaluated four different steroids with different potency. None of the studies evaluated quality of life. Antihistamines improved the composite nasal symptoms score 2% more than did leukotriene antagonists (Figure 4), although the difference was not statistically significant (95% CI: 0% to 4%). Antihistamines improved quality of life 0.11 units more than did leukotriene antagonists (95% CI: 0.04 to 0.18 units). Combination of Leukotriene Receptor Antagonists Plus Antihistamines In terms of 24-hour composite nasal symptoms, the combination of a leukotriene receptor antagonist and antihistamine was significantly more effective than use of either a leukotriene receptor antagonist or antihistamine alone (Table 3). However, the differences were not significant in terms of standardized rhinoconjunctivitis quality of life (Table 3). There was no difference between use of an intranasal corticosteroid and combination therapy (leukotriene receptor antagonist plus antihistamine) in terms of nasal symptoms. DISCUSSION This systematic review shows that leukotriene receptor antagonists produce a small but statistically significant 340 March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume 116

55 Leukotriene Antagonists for Allergic Rhinitis/Wilson et al Table 2. Details of Studies Comparing a Leukotriene Receptor Antagonist Plus an Antihistamine versus an Antihistamine, Leukotriene Receptor Antagonist, or Intranasal Corticosteroid Alone First Author (Reference) Drug Duration Comparison Parallel Meltzer (24)* Nayak (25)* Pullerits (27) Crossover Wilson (21) Wilson (17) Wilson (18) Montelukast 10 mg loratadine 10 mg (n 90) Montelukast 10 mg loratadine 10 mg (n 302) Montelukast 10 mg loratadine 10 mg once daily (n 15) Montelukast 10 mg loratadine 10 mg once daily (n 37) Montelukast 10 mg cetirizine 10 mg once daily (n 21) Montelukast 10 mg cetirizine 10 mg once daily (n 22) 2 weeks Loratadine 10 mg (n 92); montelukast 10 mg (n 95); montelukast 20 mg (n 90) 2 weeks Loratadine 10 mg (n 301); montelukast 10 mg once daily (n 155) 50 days Fluticasone propionate 200 g(n 13); montelukast 10 mg (n 16) 2 weeks Fexofenadine 120 mg (n 37) 2 weeks Inhaled (400 g) and intranasal (200 g) budesonide (n 21) 2 weeks Mometasone furoate 200 g (n 22) Endpoints In Addition to Nasal Symptoms Jadad Score Quality of life 5 Quality of life, blood eosinophils Nasal biopsy 3 Nasal peak inspiratory flow, daily activity Nasal peak inspiratory flow, daily activity, acoustic rhinometry, rhinomanometry Nasal peak inspiratory flow, daily activity, rhinomanometry, acoustic rhinometry, nasal nitric oxide, blood eosinophil count improvement in nasal symptom and standardized rhinoconjunctivitis quality-of-life scores when compared with placebo. However, there was no significant difference between leukotriene receptor antagonists and antihistamines, and leukotriene receptor antagonists were less effective than intranasal corticosteroids. The combination Figure 2. Comparison of leukotriene receptor antagonists and placebo. The plot of weighted mean difference with 95% confidence intervals for composite rhinitis symptom scores is expressed as a percentage of the maximum score. There was significant variation in results among studies (P 0.01 for heterogeneity). Figure 3. Comparison of leukotriene receptor antagonists and nasal corticosteroids. The plot of weighted mean difference with 95% confidence intervals for composite rhinitis symptom scores is expressed as a percentage of the maximum score. There was significant variation in results among studies (P for heterogeneity). March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume

56 Leukotriene Antagonists for Allergic Rhinitis/Wilson et al Figure 4. Comparison of leukotriene receptor antagonists and antihistamines. The plot of weighted mean difference with 95% confidence intervals for composite rhinitis symptom scores is expressed as a percentage of the maximum score. The results did not vary among the studies (P 0.13 for heterogeneity). There was no significant difference between leukotriene receptor antagonists and antihistamines. of an antihistamine and leukotriene receptor antagonist was superior to either therapy when given alone, but there was no difference when compared with intranasal corticosteroids. This suggests that leukotriene receptor antagonists may have a role in the management of rhinitis, although intranasal corticosteroids should be the first-line therapy. There was a lack of evaluable data on nasal polyposis (29 34) or perennial allergic rhinitis (35,36), with no adequate randomized controlled trials. Furthermore, there are only a few studies providing objective measurements of nasal patency and inflammation. Also, there were insufficient data to assess the effects of sex, duration of therapy, or comorbidity with asthma or aspirin sensitivity on response to leukotriene receptor antagonist therapy. Still, despite the small number of studies providing clinical data on leukotriene receptor antagonists in rhinitis, there are a considerable number of review articles, many of which discuss the role of leukotriene in rhinitis or asthma and the theoretical role of leukotriene receptor antagonists. However, none involve a systematic review and meta-analysis. It is clear that further research is required, with more emphasis on objective assessment, the effects on individual nasal symptoms, and the effects of leukotriene receptor antagonists on nasal polyposis and perennial rhinitis. We performed a thorough literature search that included manuscripts that were not written in English, to avoid bias (37). However, our study had several limitations. Scientific meeting abstracts were not reviewed as we could not validate the quality of those results. We were unable to obtain data from unpublished studies, which may have biased our results (38). Furthermore, we did not contact the authors to obtain data that had not been reported, which might have been useful in identifying the effects of the different therapies on the individual nasal symptoms. A visual inspection of a funnel plot of the effect sizes of the pooled studies against the inverse of the variances (a measure of the precision of the estimates) suggested a low probability for a publication bias. Four of the eight studies comparing leukotriene receptor antagonists with placebo (24 26,28) were industry sponsored. Since these were the largest studies involving more than 90% of the study sample, we did not perform a subgroup analysis by excluding these data. Also, since all the studies involved patients with seasonal allergic rhinitis, the conclusions may not be applicable to patients with perennial allergic rhinitis. The difference in standardized rhinoconjunctivitis quality of life between leukotriene receptor antagonists and placebo was 0.30 units, and 0.11 units between leukotriene receptor antagonists and antihistamines, both less than the minimal important difference for change, which has been calculated to be 0.57 units (39). The changes in rhinitis symptoms, when standardized for maximum score, were also small. Although the changes were statistically significant for leukotriene receptor antagonists versus placebo, they may not be clinically relevant. The study by Meltzer et al (24) was powered to achieve a difference of 8.3%, compared with 5% in the Table 3. Details of Analysis of Studies Comparing a Leukotriene Receptor Antagonist Plus an Antihistamine versus Use of a Leukotriene Receptor Antagonist, Antihistamine, or Intranasal Corticosteroid Alone Comparison with: Number of Studies Measure Weighted Mean Difference* (95% Confidence Interval) Leukotriene receptor antagonists 3 Symptoms 3% ( 6% to 1%) Leukotriene receptor antagonists 2 Quality of life 0.17 units ( 0.56 to 0.22 units) Antihistamines 3 Symptoms 4% ( 6% to 3%) Antihistamines 2 Quality of life 0.06 units ( 0.27 to 0.15 units) Intranasal corticosteroids 3 Symptoms 3% ( 6% to 11%) * A negative weighted mean difference favors a leukotriene receptor antagonist plus an antihistamine. P March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume 116

57 Leukotriene Antagonists for Allergic Rhinitis/Wilson et al study by van Adelsberg et al (28). If we assume that a difference between the two treatments of 10% is clinically relevant, then only the difference between intranasal corticosteroids and leukotriene receptor antagonists is clinically important. Many of the studies used different scoring systems for rhinitis symptoms. We created an index of the symptom score, which we believe is valid because all the studies reported a linear score from no symptoms (0) to a maximum score (which was variable). Furthermore, all the symptom scores used the same individual symptoms, starting from the same point, and were unidirectional with higher values representing more severe symptoms. We used a linear equation to convert all scores into a standard score with a gradient of 1 (or 100% when expressed as a percentage). Since we converted all measurements to the same scale, we combined the mean differences weighted on the precision of the estimate (weighted mean difference), rather than using a standardized mean difference, which is usually used to combine outcomes expressed in different scales. The random-effects model was a conservative way of representing the data. It assumes that the studies pooled are a representation of a random sample of a hypothetical group of studies (40). When heterogeneous studies were pooled, we examined the sources of heterogeneity, which included methodological rigor and the different medications used. For example, the source of heterogeneity in the comparison of leukotriene antagonists was the Jadad score, whereas the differences in potency of various nasal steroids were an additional source of heterogeneity in the comparison of nasal corticosteroids. In the evaluation of combination therapy, two studies showed a beneficial effect of montelukast and cetirizine (17,18), whereas a third study showed the opposite effect with montelukast and loratadine (27). This difference in effect may be due to the relative potency of the antihistamine, as cetirizine is more potent than loratadine (41). However, we were unable to compare the different leukotriene receptor antagonists, antihistamines, or intranasal corticosteroids due to the lack of direct comparative studies and the relatively small number of studies using leukotriene receptor antagonists other than montelukast. Our findings are in keeping with the conclusion of previous review articles. Meltzer (42) highlighted the efficacy and safety of leukotriene receptor antagonists in placebocontrolled studies. Mygind et al (43) commented on the lack of data on use of leukotriene receptor antagonists in rhinitis and nasal polyposis. They believed that antihistamine therapy would be more beneficial than leukotriene receptor antagonists, but that the combination of the two would have a role in the management of patients with asthma or rhinitis. More recently, Nathan (44) reviewed the effects of leukotriene receptor antagonists alone or in combination with an antihistamine, and reported results similar to ours. Since a meta-analysis was not performed, the magnitude of the treatment effects was not estimated. In conclusion, we found that leukotriene receptor antagonists have significantly greater efficacy than placebo, similar efficacy to antihistamines, but less efficacy than intranasal corticosteroids in reducing nasal symptoms and improving rhinoconjunctivitis quality of life. The magnitude of benefits is less than what is considered clinically important, but there is an additive effect of leukotriene receptor antagonists and antihistamines, with combination therapy being more efficacious than either therapy alone. However, intranasal corticosteroids are still more effective than combination therapy, and are likely to remain the first-line therapy for allergic rhinitis. ACKNOWLEDGMENT We thank L. Ripendelli, F. Braccioni, M. Katayama, and T. Kawayama for their help with the translation of non-english manuscripts. REFERENCES 1. Settipane RA. Demographics and epidemiology of allergic and nonallergic rhinitis. Allergy Asthma Proc. 2001;22: Meltzer EO. Quality of life in adults and children with allergic rhinitis. J Allergy Clin Immunol. 2001;108(suppl):S45 S Radenne F, Lamblin C, Vandezande LM, et al. Quality of life in nasal polyposis. J Allergy Clin Immunol. 1999;104: Wasserman B. Allergic rhinitis and wellness opportunities. Manag Care Interface. 2001;14: Law AW, Reed SD, Sundy JS, Schulman KA. Direct costs of allergic rhinitis in the United States: estimates from the 1996 medical expenditure panel survey. J Allergy Clin Immunol. 2003;111: Fineman SM. The burden of allergic rhinitis: beyond dollars and cents. Ann Allergy Asthma Immunol. 2002;88: Grupp-Phelan J, Lozano P, Fishman P. Health care utilization and cost in children with asthma and selected comorbidities. J Asthma. 2001;38: Wenzel SE, Larsen GL, Johnston K, et al. Elevated levels of leukotriene C4 in bronchoalveolar lavage fluid from atopic asthmatics after endobronchial allergen challenge. Am Rev Respir Dis. 1990; 142: Pavord ID, Ward R, Woltmann G, et al. Induced sputum eicosanoid concentrations in asthma. Am J Respir Crit Care Med. 1999; 160: Broide DH, Eisman S, Ramsdell JW, et al. Airway levels of mast cell-derived mediators in exercise-induced asthma. Am Rev Respir Dis. 1990;141: Macfarlane AJ, Dworski R, Sheller JR, et al. Sputum cysteinyl leukotrienes increase 24 hours after allergen inhalation in atopic asthmatics. Am J Respir Crit Care Med. 2000;161: Busse W, Raphael GD, Galant S, et al. Low-dose fluticasone propionate compared with montelukast for first-line treatment of persistent asthma: a randomized clinical trial. J Allergy Clin Immunol. 2001;107: Shaw RJ, Fitzharris P, Cromwell O, et al. Allergen-induced release of sulphidopeptide leukotrienes (SRS-A) and LTB4 in allergic rhinitis. Allergy. 1985;40: Kowalski ML, Sliwinska-Kowalska M, Igarashi Y, et al. Nasal secretions in response to acetylsalicylic acid. J Allergy Clin Immunol. 1993;91: March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume

58 Leukotriene Antagonists for Allergic Rhinitis/Wilson et al 15. Jadad AR, Moore RA, Carroll D, et al. Assessing the quality of reports of randomized clinical trials: is blinding necessary? Control Clin Trials. 1996;17: Pullerits T, Praks L, Skoogh BE, et al. Randomized placebo-controlled study comparing a leukotriene receptor antagonist and a nasal glucocorticoid in seasonal allergic rhinitis. Am J Respir Crit Care Med. 1999;159: Wilson AM, Sims EJ, Orr LC, et al. Effects of topical corticosteroid and combined mediator blockade on domiciliary and laboratory measurements of nasal function in seasonal allergic rhinitis. Ann Allergy Asthma Immunol. 2001;87: Wilson AM, Orr LC, Sims EJ, et al. Effects of monotherapy with intra-nasal corticosteroid or combined oral histamine and leukotriene receptor antagonists in seasonal allergic rhinitis. Clin Exp Allergy. 2001;31: Wilson AM, Dempsey OJ, Sims EJ, Lipworth BJ. A comparison of topical budesonide and oral montelukast in seasonal allergic rhinitis and asthma. Clin Exp Allergy. 2001;31: Wilson A, Dempsey OJ, Sims EJ, et al. Evaluation of treatment response in patients with seasonal allergic rhinitis using domiciliary nasal peak inspiratory flow. Clin Exp Allergy. 2000;30: Wilson AM, Orr LC, Coutie WJ, et al. A comparison of once daily fexofenadine versus the combination of montelukast plus loratadine on domiciliary nasal peak flow and symptoms in seasonal allergic rhinitis. Clin Exp Allergy. 2002;32: Juniper EF, Guyatt GH. Development and testing of a new measure of health status for clinical trials in rhinoconjunctivitis. Clin Exp Allergy. 1991;21: DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials. 1986;7: Meltzer EO, Malmstrom K, Lu S, et al. Concomitant montelukast and loratadine as treatment for seasonal allergic rhinitis: a randomized, placebo-controlled clinical trial. J Allergy Clin Immunol. 2000; 105: Nayak AS, Philip G, Lu S, et al. Efficacy and tolerability of montelukast alone or in combination with loratadine in seasonal allergic rhinitis: a multicenter, randomized, double-blind, placebo-controlled trial performed in the fall. Ann Allergy Asthma Immunol. 2002;88: Philip G, Malmstrom K, Hampel FC, et al. Montelukast for treating seasonal allergic rhinitis: a randomized, double-blind, placebocontrolled trial performed in the spring. Clin Exp Allergy. 2002;32: Pullerits T, Praks L, Ristioja V, Lotvall J. Comparison of a nasal glucocorticoid, antileukotriene, and a combination of antileukotriene and antihistamine in the treatment of seasonal allergic rhinitis. J Allergy Clin Immunol. 2002;109: van Adelsberg J, Philip G, LaForce CF, et al. Randomized controlled trial evaluating the clinical benefit of montelukast for treating spring seasonal allergic rhinitis. Ann Allergy Asthma Immunol. 2003;90: Di Rienzo L, Artuso A, Cerqua N. Antileukotrienes in the prevention of postoperative recurrence of nasal polyposis in ASA syndrome [in Italian]. Acta Otorhinolaryngol Ital. 2000;20: Grundman T, Topfner M. Treatment of ASS-associated Polyposis (ASSAP) with a cysteinyl leukotriene receptor antagonist a prospective drug study on its anti-inflammatory effects. Laryngorhinootologie. 2001;80: Parnes SM, Chuma AV. Acute effects of antileukotrienes on sinonasal polyposis and sinusitis. Ear Nose Throat J. 2000;79: Ragab S, Parikh A, Darby YC, Scadding GK. An open audit of montelukast, a leukotriene receptor antagonist, in nasal polyposis associated with asthma. Clin Exp Allergy. 2001;31: Wilson AM, White PS, Gardiner Q, et al. Effects of leukotriene receptor antagonist therapy in patients with chronic rhinosinusitis in a real life rhinology clinic setting. Rhinology. 2001;39: Malerba M, Radaeli A, Ceriani L, et al. Comparison of oral montelukast and inhaled fluticasone in the treatment of asthma associated with chronic rhinopolyposis: a single blind randomised pilot study. Curr Ther Res Clin Exp. 2002;63: Narita S, Kurose M, Obayashi K. Pranlukast produces beneficial effects in nasal symptoms on patients with perennial allergic rhinitis when used in combination with an antihistamine. Practica Otologica [in Japanese]. 2001;94: Numata T, Konno A, Yamakoshi T, et al. Comparative role of peptide leukotrienes and histamine in the development of nasal mucosal swelling in nasal allergy. Ann Otol Rhinol Laryngol. 1999;108: Egger M, Zellweger-Zahner T, Schneider M, et al. Language bias in randomised controlled trials published in English and German. Lancet. 1997;350: Cook DJ, Guyatt GH, Ryan G, et al. Should unpublished data be included in meta-analyses? Current convictions and controversies. JAMA. 1993;269: Juniper EF, Guyatt GH, Griffith LE, Ferrie PJ. Interpretation of rhinoconjunctivitis quality of life questionnaire data. J Allergy Clin Immunol. 1996;98: Petitti DB. Meta-Analysis, Decision Analysis and Cost-Effectiveness Analysis: Methods for Quantitative Synthesis in Medicine. New York, New York: Oxford University Press; 1994: Statistical Methods in Meta-Analysis. 41. Meltzer EO, Weiler JM, Widlitz MD. Comparative outdoor study of the efficacy, onset and duration of action, and safety of cetirizine, loratadine, and placebo for seasonal allergic rhinitis. J Allergy Clin Immunol. 1996;97: Meltzer EO. Clinical evidence for antileukotriene therapy in the management of allergic rhinitis. Ann Allergy Asthma Immunol. 2002;88: Mygind N, Dahl R, Bisgaard H. Leukotrienes, leukotriene receptor antagonists, and rhinitis. Allergy. 2000;55: Nathan RA. Pharmacotherapy for allergic rhinitis: a critical review of leukotriene receptor antagonists compared with other treatments. Ann Allergy Asthma Immunol. 2003;90: March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume 116

59 Radioiodine Ablation of the Thyroid to Allow the Reintroduction of Amiodarone Treatment in Patients with a Prior History of Amiodarone- Induced Thyrotoxicosis Jean-Sylvain Hermida, MD, Géneviève Jarry, MD, Emmanuel Tcheng, MD, Véronique Moullart, MD, Sylvie Arlot, MD, Jean-Luc Rey, MD, Jean Delonca, MD, Claire Schvartz, MD Long-term use of amiodarone is often limited by side effects on thyroid, pulmonary, and hepatic function. The incidence of amiodarone-induced thyrotoxicosis, which occurs in 2% to 20% of patients, depends on dietary iodine content (1,2). In type 1 thyrotoxicosis, an iodine load triggers autonomous hormone production; in type 2 thyrotoxicosis, hormone release may be linked to destructive thyroiditis (3 5). In patients with mildly impaired thyroid function and a stable cardiac rhythm, amiodarone can be withdrawn safely, resulting in spontaneous resolution of thyrotoxicosis, and the antiarrhythmic treatment strategy can be modified. In other patients, amiodarone withdrawal is inadequate due to life-threatening thyrotoxicosis or arrhythmias. Near total thyroidectomy may be an option (6), although a recent study suggested that continuing amiodarone may be possible in patients who are treated for hyperthyroidism (7). At a later stage, after full recovery of thyroid function, reintroduction of amiodarone may be desirable for controlling worsening arrhythmias. However, previous amiodarone-induced thyrotoxicosis is generally considered to be a contraindication to doing so, despite the lack of published studies in this area. An alternative could be to treat patients with radioiodine 131 (I 131 ), as in patients with recurrent episodes of lymphocytic thyroiditis (8). The aim of this study was to describe the results of this approach. METHODS We reviewed the medical records of 15 patients (12 men; mean [ SD] age, years) who received I 131 to prevent the recurrence of amiodarone-induced thyrotoxicosis. Patients were treated between 1995 and Thyrotoxicosis was defined as a thyroid-stimulating BRIEF OBSERVATION hormone (TSH) level 0.1 IU/mL and a free triiodothyronine level 6.5 pmol/l (422 pg/dl). Patients were treated with I 131 if the following criteria were fulfilled: previous episode of amiodarone-induced thyrotoxicosis with normal thyroid function after the cessation of amiodarone; I 131 uptake rate 10%; recurrence of symptomatic atrial or life-threatening ventricular tachyarrhythmia; failure of conventional antiarrhythmic strategy; and previous efficacy of amiodarone. Depending on the size of the thyroid and the iodine uptake, a standard dose of I 131 was administered orally (370 to 740 MBq [10 to 20 mci]). RESULTS Amiodarone was initially prescribed for atrial fibrillation or ventricular tachycardia in patients with a variety of cardiac disorders (Table 1). Two patients had isolated atrial fibrillation. The mean left ventricular ejection fraction was 47% 10%. Amiodarone-induced thyrotoxicosis occurred after months of treatment. I 131 uptake was absent in 14 of the 15 patients; 1 patient (patient 4 in Table 1) had heterogeneous thyroid nodules. Antithyroid peroxidase and TSH receptor antibodies were not detected in the 6 patients in whom they were measured. All but 1 patient were thought to have type 2 amiodaroneinduced thyrotoxicosis. Amiodarone withdrawal was possible in all patients. Thyrotoxicosis resolved spontaneously in 4 patients; 5 patients received prednisone, 3 patients received carbimazole, and 1 patient received both. The management of the initial episode of thyrotoxicosis was unknown for 2 patients. I 131 Therapy and Amiodarone Reintroduction The indications for reintroduction of amiodarone were multiple episodes of symptomatic paroxysmal atrial fibrillation in 10 patients and recurrent ventricular tachycardia in 5 patients. Of the 10 patients with recurrent atrial fibrillation, 8 occurred before the technique of pulmonary vein ablation was available and amiodarone rechallenge was preferred to ablation of the His bundle. Among the 5 patients with ventricular tachycardia, I 131 and amiodarone rechallenge was proposed in 2 patients because of mapping difficulties during radiofrequency ablation. In 1 patient, radiofrequency ablation failed. One patient had recurrent discharges of an automatic internal defibrillator and another patient was not eligible for implantation of an automatic internal defibrillator. All patients were euthyroid before I 131 treatment, which was administered for a mean of months (range, 5 to 147 months) after the episode of thyrotoxicosis (Table 2). Early, mild, transient hyperthyroidism developed in 2 patients after I 131 treatment. Fourteen 2004 by Excerpta Medica Inc /04/$ see front matter 345 All rights reserved. doi: /j.amjmed

60 Radioiodine Ablative Therapy and Reintroduction of Amiodarone/Hermida et al Table 1. Characteristics of Patients Patient No. Year of Treatment Sex Age (years) Type of Heart Disease Arrhythmia Left Ventricular Ejection Fraction (%) M 49 Arrhythmogenic Ventricular tachycardia 55 right ventricular dysplasia F 66 Hypertensive Atrial fibrillation M 61 Ischemic Ventricular tachycardia M 41 Hypertensive Atrial fibrillation M 60 Ischemic Atrial fibrillation M 70 Cardiomyopathy Atrial fibrillation M 43 None Atrial fibrillation M 40 Arrhythmogenic Ventricular tachycardia 50 right ventricular dysplasia M 80 Cardiomyopathy Atrial fibrillation F 74 Valvular Atrial fibrillation M 76 Ischemic Atrial fibrillation M 82 Cardiomyopathy Atrial fibrillation M 77 Ischemic Ventricular tachycardia M 58 None Atrial fibrillation F 70 Cardiomyopathy Ventricular tachycardia 40 F female; M male. (93%) of the 15 patients developed hypothyroidism and were treated with L-thyroxine. Only 1 patient remained euthyroid. Amiodarone was reintroduced in all but 1 patient, who developed permanent atrial fibrillation and became asymptomatic. Amiodarone (200 to 400 mg/d) was Table 2. Outcome after Radioiodine Ablation of the Thyroid Patient No. I 131 Dose (MBq) Serum TSH Level before I 131 ( IU/mL) Interval after Amiodarone Withdrawal (months)* TSH Level 1 Month after I 131 ( IU/mL) Amiodarone Reintroduction Interval (days) Duration of Hypothyroidism (months) Length of Follow-up (months) Arrhythmia Control after Amiodarone Reintroduction Yes Yes Yes Yes Yes Yes Yes Yes Yes No Euthyroid 20 Yes Yes No No Yes Mean SD * From amiodarone withdrawal to radioiodine ablation of the thyroid. From radioablation to reintroduction. Died of nonarrhythmic causes. I 131 iodine 131; TSH thyroid-stimulating hormone. 346 March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume 116

61 Radioiodine Ablative Therapy and Reintroduction of Amiodarone/Hermida et al started within a month after I 131 treatment in most patients (Table 2). No recurrence of amiodarone-induced thyrotoxicosis was observed. Cardiac Outcome Arrhythmia control was achieved in 12 (86%) of 14 patients after the reintroduction of amiodarone. The arrhythmias were not controlled in 1 patient who had ventricular tachycardia and in 1 patient who developed permanent atrial fibrillation. Amiodarone was not withdrawn from any patient. Two patients died of low cardiac output, unrelated to the recurrence of cardiac arrhythmias. DISCUSSION In our series of patients with amiodarone-induced thyrotoxicosis, I 131 treatment allowed the reintroduction of amiodarone and control of the tachyarrythmia in 12 of 14 patients. Transient hyperthyroidism and permanent hypothyroidism were the major disadvantages of this approach. Despite the paucity of data about the risk of recurrent thyrotoxicosis, few clinicians are willing to reintroduce amiodarone in patients whose arrhythmias recur after its withdrawal. However, there are only a few options for preventing amiodarone-induced thyrotoxicosis. These include thyroidectomy (6), which should be reserved for patients requiring urgent amiodarone treatment. Another possibility is the reintroduction of amiodarone under close monitoring. About 10% of patients develop recurrent thyrotoxicosis within the first 2 years of treatment; that incidence increases with time (9). Finally, patients may be treated with I 131 radioablation of the thyroid; this requires a long washout period (several months) because the iodine load and absent iodine uptake, which precludes radioiodine therapy, may persist for 6 to 12 months or longer after amiodarone is stopped. Amiodarone should be reintroduced, when possible, within a month after I 131 treatment. I 131 is the treatment of choice for patients with Graves disease and for most patients with toxic multinodular or uninodular goiters (10,11). Despite attempts to design dosage schedules aimed at euthyroidism, permanent hypothyroidism is the main complication of I 131 (12). A recent study found that the incidence of hypothyroidism was 61% a year after a single dose of 370 MBq (13). In our series, we observed a higher risk (93%) of hypothroidism. This could be a specific characteristic of patients with a history of amiodarone-induced thyrotoxicosis. However, hypothyroidism should be viewed as a goal, rather than a complication, of treatment in these patients. A less frequent complication of I 131 treatment is early transient hyperthyroidism, which might be a direct consequence of the radiation or an immunological phenomenon (13). Patients should be followed closely during that time. Three large studies suggest that the overall risk of cancer is not related to the dose of I 131 or to the time after exposure (14 16). The mean follow-up now extends to 21 years in one study (16). Although a small increase in the individual risk of certain types of cancer has been reported (14 16), these observations are controversial (10,12). In view of the potential benefit-risk ratio, we believe that the preventive use of I 131 should be reserved for patients with all of the following characteristics: prior documented episode of amiodarone-induced thyrotoxicosis; symptomatic atrial or life-threatening ventricular tachyarrhythmias that responded to previous treatment with amiodarone; and resistance to all other antiarrhythmic strategies. For patients with atrial fibrillation, this method could be used in cases of pulmonary vein ablation failure, in the absence of clear ablation indication (age), or because of patient refusal. It could be preferred to His bundle ablation to avoid pacemaker implantation. For patients with ventricular tachycardia, this approach could be useful in those with frequent automatic internal defibrillator discharges or when radiofrequency ablation is not possible. REFERENCES 1. Borowski GD, Garofano CD, Rose LI, et al. Effect of long-term amiodarone therapy on thyroid hormone levels and thyroid function. Am J Med. 1985;78: Martino E, Safran M, Aghini-Lombardi F, et al. Environmental iodine intake and thyroid dysfunction during chronic amiodarone therapy. Ann Intern Med. 1984;101: Martino E, Bartalena L, Bogazzi F, et al. The effects of amiodarone on the thyroid. Endocr Rev. 2001;22: Bogazzi F, Bartalena L, Gasperi M, et al. The various effects of amiodarone on thyroid function. Thyroid. 2001;11: Chiovato L, Martino E, Tonacchera M, et al. Studies on the in vitro cytotoxic effect of amiodarone. Endocrinology. 1994;134: Hamoir E, Meurisse M, Defechereux T, et al. Surgical management of amiodarone-associated thyrotoxicosis: too risky or too effective? World J Surg. 1998;22: Osman F, Franklyn JA, Sheppard MC, et al. Successful treatment of amiodarone-induced thyrotoxicosis. Circulation. 2002;105: Nikolai TF, Coombs GJ, McKenzie AK, et al. Treatment of lymphocytic thyroiditis with spontaneously resolving hyperthyroidism (silent thyroiditis). Arch Intern Med. 1982;142: Trip MD, Wiersinga W, Plomp TA. Incidence, predictability, and pathogenesis of amiodarone-induced thyrotoxicosis and hypothyroidism. Am J Med. 1991;91: Cooper DS. Radioiodine for hyperthyroidism: where do we stand after 50 years? JAMA. 1998;280: March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume

62 Radioiodine Ablative Therapy and Reintroduction of Amiodarone/Hermida et al 11. Singer PA, Cooper DS, Levy EG, et al. Treatment guidelines for patients with hyperthyroidism and hypothyroidism. Standards of Care Committee, American Thyroid Association. JAMA. 1995;273: Pauwels EK, Smit JW, Slats A, et al. Health effects of therapeutic use of 131I in hyperthyroidism. Q J Nucl Med. 2000;44: Allahabadia A, Daykin J, Sheppard MC, et al. Radioiodine treatment of hyperthyroidism prognostic factors for outcome. J Clin Endocrinol Metab. 2001;86: Holm LE, Hall P, Wiklund K, et al. Cancer risk after iodine-131 therapy for hyperthyroidism. J Natl Cancer Inst. 1991;83: Ron E, Doody MM, Becker DV, et al. Cancer mortality following treatment for adult hyperthyroidism. Cooperative Thyrotoxicosis Therapy Follow-up Study Group. JAMA. 1998;280: Franklyn JA, Maisonneuve P, Sheppard M, et al. Cancer incidence and mortality after radioiodine treatment for hyperthyroidism: a population-based cohort study. Lancet. 1999;353: From Hôpital Sud (JSH), Centre Hospitalier Universitaire, Amiens, France; the Cardiology Department (GJ, ET, JLR, JD), Nuclear Medicine Center (VM), Endocrinology Department (SA), University Hospital, Amiens, France; and the Biophysics and Nuclear Medicine Unit (CS), Jean Godinot Institute, Reims, France. Requests for reprints should be addressed to Jean-Sylvain Hermida, MD, Hôpital Sud, Centre Hospitalier Universitaire, Amiens Cedex, France, or hermida.jean-sylvain@chu-amiens.fr. Manuscript submitted January 28, 2003, and accepted in revised form August 22, March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume 116

63 IMAGES OF OSLER Cases from the Osler Medical Service at Johns Hopkins University Prepared by Emily Schopick, MD, and Kerri Cavanaugh, MD PRESENTING FEATURES: A 42-year-old woman presented to the emergency department at the Johns Hopkins Hospital complaining of a fever and a rash. The patient reported that 9 days before admission she noticed a painful rash on her left buttock. She was diagnosed with zoster and given a 10-day course of valacyclovir. However, on day 6 of therapy, the patient developed a new rash that started on her palms and soles, spreading up her arms and legs to her trunk, and moved to the mucous membranes in her mouth. The patient reported that the new rash began as erythematous circles with central vesicles, that it was itchy and painful, and that it seemed to be associated with fevers, chills, and rigors. Concurrently, she developed worsening of her migraine headaches and right eye redness and itching. The patient denied any nausea, vomiting, or abdominal symptoms, but had decreased oral intake due to painful oral lesions. The patient denied any important past medical history. She had chicken pox twice, once as a child and once in her 20s. She had all of her childhood immunizations. She had no history of herpes infection, although her husband recently had been diagnosed with oral herpes. She had no recent travel outside of the state of Maryland. On day 4 of the valacyclovir therapy, the patient had hiked in a state park, but did not notice any tick bites. She had two dogs at home that were not new to the family. The patient denied any allergies to medications, and at the time of presentation was taking acetaminophen, natural estrogen supplementation (estroven) for the past 6 weeks, hydroxyzine as needed for itching, and sumatriptan as needed. Valacyclovir was stopped on the day before admission. On physical examination, the patient was febrile (101.0 F), with a heart rate of 129 beats per minute, blood pressure of 127/80 mmhg, respiratory rate of 18 breaths per minute, and an oxygen saturation of 98% on room air. She appeared well and in no acute distress. On HEENT examination, her right eye was diffusely erythematous and watering, but had no exudate. Her nasopharynx was clear, and examination of her oropharynx was notable for multiple ulcers on the buccal mucosa. There was Figure 1. Target lesions: round erythematous macules with raised centers and three zones of color changes best demonstrated on patient s right third finger (arrow) by Excerpta Medica Inc /04/$ see front matter 349 All rights reserved. doi: /j.amjmed

64 Images of Osler/Schopick and Cavanaugh tender left cervical and bilateral axillary lymphadenopathy. Lungs, heart, abdomen, neurologic, and extremity examinations were normal. Skin examination was notable for generalized blanchable, erythematous, round macules, with some confluence to plaque formation, greater on the extremities, palms, and soles than on the trunk (Figures 1 and 2). Most lesions were erythematous, targetoid lesions measuring 0.5 to 2.0 cm in diameter. There were occasional papules and pustules. On the left buttock, there was a 5- to 6-cm plaque of eroded skin without vesicles, as well as some vesicular lesions in the same area. Nikolsky sign was negative for both rashes. On laboratory analysis, the patient had normal electrolytes, an alanine aminotransferase level of 103, and an aspartate aminotransferase level of 55, with no previous values for comparison. Her white blood cell count was 7370/mm 3, with 12% lymphocytes, 9% monocytes, 78% neutrophils, 1% eosinophils, and no bands. Platelet count was 234 x 10 3 / L. Rapid plasma reagent was nonreactive, and urinalysis was unremarkable. Tzanck smear of the generalized rash was negative for giant cells. Human immunodeficiency virus antibodies and polymerase chain reaction were negative. Lyme disease and Rocky Mountain spotted fever antibodies were negative. Epstein Barr virus immunoglobulin (Ig) G was positive and IgM was negative, cytomegalovirus IgG was positive and IgM was negative, and herpes simplex virus IgG was positive. Serum varicella-zoster virus polymerase chain reaction was negative. Hepatitis A, B, and C serologies were all negative. What is the diagnosis? DIAGNOSIS: Erythema multiforme that is likely caused by herpes simplex viral infection, although valacyclovir or estrogen-induced erythema multiforme cannot be excluded. DISCUSSION: After consultation with the dermatology department, it was decided that the patient s rash was classic for erythema multiforme. The rash fit into the category of erythema multiforme major, given the oral mucosal involvement. Valacyclovir and estrogen were discontinued at the time of admission, and the patient was started on prednisone 60 mg per day. Viscous lidocaine and sucralfate were used for oral pain relief. She was seen by the ophthalmology service, which believed that there was no ocular involvement. During the first 2 hospital days there was very little resolution of the rash, but subsequently the vesicles began to close and the oral ulcers began to heal. The patient s history and physical examination were very consistent with the diagnosis of erythema multiforme, likely herpes simplex virus induced. As happened in our patient, malaise, fevers, itching, and burning sensations on the skin may precede erythema multiforme. It is characterized by lesions that begin as maculopapules and develop over 24 to 48 hours into target lesions. Lesions often begin on the hands, feet, forearms, and legs. Lesions also develop on the palms and soles, Figure 2. Erythema multiforme: extensive lesions on legs in different stages of development. All are circular erythematous macules with a raised center. 350 March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume 116

65 Images of Osler/Schopick and Cavanaugh but these lesions appear more urticarial than target-like (1,2). Herpes simplex virus is believed to cause up to 90% of cases of erythema multiforme through an immune-mediated process. Viral DNA fragments are carried to distant skin sites by mononuclear cells in the blood. CD4 T-lymphocytes specific to herpes simplex virus antigens react with these pieces of DNA, producing interferon gamma and initiating an inflammatory cascade. It is this process of recruiting inflammatory mediators to the skin surface that causes the classic erythema multiforme rash (3). Erythema multiforme has an incidence of approximately 0.01% to 1% per year. It is most often seen in young males between the ages of 20 to 40 years, but can be seen in all groups (3). It is sometimes considered to be the mildest form in a continuum of skin rashes including Stevens-Johnson syndrome and toxic epidermal necrolysis. Erythema multiforme is often divided into erythema multiforme minor, which consists of only skin involvement, and erythema multiforme major, which includes mucosal involvement and which begins to overlap with Stevens-Johnson syndrome (1). Erythema multiforme is considered to have a benign course, whereas Stevens- Johnson syndrome can evolve into toxic epidermal necrolysis, which is life threatening. The histology of these two processes is also quite different. In erythema multiforme, the histology consists of lymphocytic infiltration into the dermo-epidermal junction. In Stevens- Johnson syndrome, the histology shows epidermal necrosis (4). Therefore, although our patient had some mucosal involvement, her diagnosis was much more consistent with erythema multiforme. The differential diagnosis for causes of erythema multiforme includes both infectious and drug-related etiologies. One of the major differences between them is that erythema multiforme is most often of infectious etiology and Stevens-Johnson syndrome/toxic epidermal necrolysis is almost always caused by drug reactions. However, erythema multiforme can also be caused by drug reactions, and less commonly by malignancies, radiation therapy, and idiopathic causes. As mentioned previously, the most common cause of erythema multiforme is herpes simplex virus. Other infectious causes include mycoplasma, Epstein-Barr virus, adenovirus, coxsackievirus, varicella virus, variola virus, influenza A, and many other viruses and bacteria (1,2). The most common drug-induced causes of erythema multiforme are sulfa medications, barbiturates, and tetracycline. Other drugs, however, can also cause these symptoms, most notably allopurinol, dapsone, nonsteroidal anti-inflammatory drugs, oral contraceptives, metronidazole, and pseudoephedrine. (5) In this patient, we considered both infectious and drug-induced causes. Although she had not taken any of the drugs that most commonly cause this presentation, valacyclovir and estrogen have been implicated in erythema multiforme. Valacyclovir is most often associated with erythema multiforme as a modality of treatment for herpes simplex virus induced erythema multiforme, but in some cases it has actually been the primary cause (6). Estrogen, in some case reports, has been found to cause erythema multiforme and lesions that mimic erythema multiforme clinically and histologically. Although the patient had been on estrogen for 6 weeks, it can sometimes take 30 to 40 days for this type of reaction to a new drug to develop (7). The most likely infectious causes in our patient were varicella-zoster virus infection and herpes simplex virus infection. The patient had been given a diagnosis of zoster 9 days before admission. However, the rash on the patient s buttocks was more likely herpes simplex virus than zoster. The clustering of lesions was not consistent with any dermatome and therefore was less likely to be a varicella-zoster virus infection, although it is not impossible (8). The patient s rash and systemic symptoms resolved on prednisone therapy and withdrawal of both valacyclovir and the estrogen supplement. REFERENCES 1. Weber DJ, Cohen MS, Fine JD. The acutely ill patient with fever and rash. In: Mandell GL, Bennet JE, Dolin R, eds. Principles and Practice of Infectious Diseases. 5th ed. Philadelphia, Pennsylvania: Churchill Livingstone Inc.; 2000; Habif TP. Hypersensitivity syndromes and vasculitis. Clinical Dermatology. 3rd ed. St. Louis, Missouri: Mosby-Year Book Inc; 1996: Aurelian L, Ono F, Burnett J. Herpes simplex virus (HSV)-associated erythema multiforme (HAEM): A viral disease with an autoimmune component. Dermatol Online J [serial online]. 2003;9:1. Available at: 4. Revuz J. New advances in severe adverse drug reactions. Dermatol Clin. 2001;19, Svensson CK, Cowen EW, Gaspari AA. Cutaneous drug reactions. Pharmacol Rev. 2000;53: Brown TJ, McCrary M, Tyrin SK. Antiviral agents: Non-antiviral drugs. J Am Acad Dermatol. 2002;47: Mutasim D, Baumbach JL. Bullous autoimmune estrogen dermatitis. J Am Acad Dermatol. 2003;49: Gnann JW Jr, Whitley RJ. Herpes zoster. N Engl J Med. 2002;347: March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume

66 EDITORIAL Lessening the Diagnostic Uncertainty in Patients with Suspected Pulmonary Embolism Russell D. Hull, MBBS, MSc, William A. Ghali, MD, MPH, Graham F. Pineo, MD The diagnosis of pulmonary embolism remains one of the greatest challenges confronting the clinician in internal medicine. The search for a single test to diagnose or exclude pulmonary embolism presently remains unfulfilled. Indeed, for the clinician, the major frustration is that diagnostic testing frequently fails to provide certainty (1 4). Recent clinical studies, however, have evaluated several diagnostic protocols that can, in many circumstances, provide reasonable clinical certainty for the diagnosis of pulmonary embolism based on estimates of clinical probability, D-dimer assays, duplex ultrasonography, helical computed tomography (CT), ventilation perfusion lung scanning, and pulmonary angiography (5). The optimal sequencing and combination of these tests has not yet been fully determined, but the approach to the patient with suspected pulmonary embolism can be aided by management studies (5 7) that enhance the diagnostic approach. In this issue, Perrier et al (8) make an important contribution by prospectively assessing a diagnostic protocol that combines quantitative rapid enzyme-linked immunoadsorbent assay (ELISA) D-dimer testing, duplex ultrasonography, helical CT scanning, and a clinical probability estimate to diagnose or exclude pulmonary embolism. The multicenter nature of this study is a strength because it demonstrates the algorithm s performance in multiple test sites, thereby enhancing the validity of the algorithm. Many such algorithms have been assessed and described in the literature (5), but the work and algorithm of Perrier and colleagues are notable on at least two fronts. First, the large patient sample permits a clinically useful upper 95% confidence limit and narrow confidence intervals around the follow-up venous thromboembolism event rate in patients with a negative quantitative rapid ELISA D-dimer result. Second, the study provides clinically important insight into the value of helical CT. Perrier et al establish the safety of using a quantitative rapid ELISA D-dimer assay as a single test in large patient samples (8,9). A particularly informative event rate (and upper 95% confidence limit) that warrants specific mention here is the 0% (upper 95% confidence limit 1.4%; Am J Med. 2004;116: range, 0% to 1.4%) follow-up event rate in patients with a negative quantitative rapid ELISA D-dimer. This provides reasonable confidence that further testing is not necessary and presents a uniquely attractive diagnostic option, particularly in centers that do not have ready access to imaging facilities. The second front that is intensively evaluated by Perrier et al is the use of single stage helical CT. Considerable data on the isolated test performance of helical CT suggest that its sensitivity may be as low as 70%. (10). Although newer generation technology may improve the sensitivity somewhat, clinicians need information on how to incorporate helical CT into diagnostic algorithms combining D-dimer, ultrasound, and imaging results. Such algorithms are well-described for lung scanning (5), but the current report suggests that helical CT may safely avoid lung scanning in an effective diagnostic algorithm. Although the follow-up event rate of 1.7% for venous thromboembolism is low in patients with negative duplex ultrasound, negative helical CT, and low to intermediate clinical probability, the corresponding upper confidence limit of 3.5% is just above the accepted safety standard of 3% (5). The practical advantages of helical CT relative to perfusion lung scanning are notable: helical CT is frequently available 24 hours, 7 days a week; is faster to perform; gives insight to alternative diagnoses; and is less influenced by the presence of lung disease. The findings of Perrier and colleagues (8) support two recent studies (6,7) that show, in aggregate, that a negative helical CT together with a negative lower extremity duplex ultrasound and low or moderate clinical probability can be used safely to exclude the diagnosis of pulmonary embolism without requiring lung scanning. Not all D-dimer assays are the same; indeed, the sensitivity, specificity, and likelihood ratios vary among the assays. The quantitative rapid ELISA assay (8) is associated with a high sensitivity and a low negative likelihood ratio (11), which in most circumstances provides a high certainty for excluding pulmonary embolism. Uniquely, Perrier et al have validated the quantitative rapid ELISA as a stand-alone test if negative. This finding should not be generalized to the non-elisa assays as, in general, the sensitivities and negative likelihood ratios of non-elisa tests are insufficient for stand-alone use. Non-ELISA assays have, however, functioned well in protocol when by Excerpta Medica Inc /04/$ see front matter All rights reserved. doi: /j.amjmed

67 Diagnostic Uncertainty in Pulmonary Embolism/Hull et al combined with a low clinical probability to exclude pulmonary embolism safely. Although absolute diagnostic certainty remains elusive, the frontier of pulmonary embolism is becoming simpler and more practical. The validated use of helical CT a fast, readily accessible, and widely available test and the applied early use of D-dimer testing represent major steps forward in the clinical paradigm. Clinicians and their patients will benefit on an ongoing basis from these more practical approaches to diagnosis. REFERENCES 1. Hull RD, Hirsh J, Carter CJ, et al. Pulmonary angiography, ventilation lung scanning, and venography for clinically suspected pulmonary embolism with abnormal perfusion lung scan. Ann Intern Med. 1983;98: Hull RD, Raskob GE, Coates G, Panju AA. Clinical validity of a normal perfusion lung scan in patients with suspected pulmonary embolism. Chest. 1990;97: PIOPED Investigators. Value of the ventilation/perfusion scan in pulmonary embolism. Results of the prospective investigation of pulmonary embolism diagnosis (PIOPED). JAMA. 1990;263: Stein PD, Hull RD, Pineo GF. Strategy that includes serial noninvasive leg tests for diagnosis of thromboembolic disease in patients with suspected acute pulmonary embolism based on data from PIOPED. Arch Intern Med. 1995;155: Kruip MJHA, Leclercq MGL, Heul CVD, Prins MH, Buller HR. Diagnostic strategies for excluding pulmonary embolism in clinical outcome studies. Ann Intern Med. 2003;138: Musset D, Parent F, Meyer G, et al. Diagnostic strategy for patients with suspected pulmonary embolism: a prospective multicentre outcome study. Lancet. 2002;360: Van Strijen MJL, de Monye W, Schiereck J, et al. Single-detector helical computed tomography as the primary diagnostic test in suspected pulmonary embolism: a multicenter clinical management study of 510 patients. Ann Intern Med. 2003;138: Perrier A, Roy PM, Aujesky D, et al. Diagnosing pulmonary embolism in outpatients with clinical assessment, D-dimer, venous ultrasound and helical computed tomography: a multicenter management study. Am J Med 2004;116: Perrier A, Desmarais S, Miron MJ, et al. Non-invasive diagnosis of venous thromboembolism in outpatients. Lancet. 1999;353: Rathbun SW, Raskob GE, Whitsett TL. Sensitivity of helical computed tomography in the diagnosis of pulmonary embolism: a systematic review. Ann Intern Med. 2000;132: Brown MD, Rowe BH, Reeves MJ, Bermingham JM, Goldhaber SZ. The accuracy of the enzyme-linked immunosorbent assay D-dimer test in the diagnosis of pulmonary embolism: a meta-analysis. Ann Emerg Med. 2003;41: March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume

68 CORRESPONDENCE LONG-TERM OUTCOME OF PATIENTS TREATED WITH HEḾATOPOIETIC GROWTH FACTORS FOR IDIOSYNCRATIC DRUG-INDUCED AGRANULOCYTOSIS To the Editor: In the long term, hematopoietic growth factors granulocyte colonystimulating factor (G-CSF) and granulocyte-macrophage colony-stimulating factor (GM-CSF) may induce several hematological disorders, such as acute myeloid leukemia (1,2). This is seen especially in patients who have myelodysplasia in association with induction chemotherapy regimens for lymphoma or acute leukemia and in cases of bone marrow transplantation (1,2). We report the long-term outcome of patients treated with G-CSF or GM-CSF for idiosyncratic drug-induced agranulocytosis. The patients were from a cohort study of drug-induced agranulocytosis, at the Hôpitaux Universitaires de Strasbourg, France, a tertiary referral center. Data from this cohort of more than 100 patients have been published elsewhere (3,4). In our study, 48 patients (53%) were divided into two groups for G-CSF or GM-CSF treatment (Table). The endpoints were long-term overall mortality and morbidity (evidence of any abnormality in blood count, particularly myelodysplasia, appearance of hematological proliferative disorder, or cancer). These data were obtained by contacting the general practitioner of each patient. The studied data are presented in the Table. The mean SD follow-up for all patients was 52 3 months (range, 13 to 102 months). The median age of the patients was 66.5 years (range, 19 to 92 years); about two thirds were women; and one third had chronic underlying diseases, particularly heart and vascular diseases. Hematological proliferative disorders were found in 2 patients: chronic lymphocytic leukemia and follicular non Hodgkin s lymphoma; and cancer in 4 elderly patients: breast cancer, prostate cancer, colon cancer, and skin cancer (Table). After a follow-up of more than 52 months, this study did not show any evidence of malignant blast cell proliferation with G-CSF and GM-CSF therapy in this homogenous population of 48 patients who have idiosyncratic drug-induced agranulocytosis. This confirms the published data on the tolerability of hematopoietic growth factors in chronic neutropenia and, in particular, the absence of Table. Long-term Outcome of 48 Patients with Drug-Induced Agranulocytosis* Outcome Patients with G-CSF or GM-CSF Treatment (n 22) Patients without G-CSF or GM-CSF Treatment (n 26) Number (%) or Mean SD Follow-up (months) Overall mortality 1 (5) 2 (8) Evidence of any abnormality 4 (18) 5 (19) in blood count Myelodysplasia 0 1 (4) Hematological proliferative 1 (5) 1 (4) disorder Cancer 2 (9) 2 (8) * Difference not significant for all variables. G-CSF granulocyte colony-stimulating factor; GM-CSF granulocyte-macrophage colonystimulating factor. long-term hematological adverse events, such as myelodysplasia or hematological proliferative disorders (5). Given the apparent good tolerability of G-CSF and GM-CSF therapy and the possible benefits of their use in several hematological disorders (1), including drug-induced agranulocytosis (3), further studies with larger sample sizes and longer follow-up are warranted. Emmanuel Andrès, MD Esther Noel, MD Department of Internal Medicine University Hospital of Strasbourg France Frédéric Maloisel, MD Department of Hematology University Hospital of Strasbourg France 1. Costa JJ. The therapeutic use of hematopoietic growth factors. J Allergy Clin Immunol. 1998;101: Gutierrez-Delgado F, Bensinger W. Safety of granulocyte-colony stimulating factor in normal donors. Curr Opin Hematol. 2001;8: Andrès E, Kurtz JE, Martin-Hunyadi C, et al. Non-chemotherapy drug-induced agranulocytosis in elderly patients: the effects of Granulocyte Colony-Stimulating Factor. Am J Med. 2002;112: Andrès E, Maloisel F, Kurtz JE, et al. Modern management of non-chemotherapy drug- induced agranulocytosis: a monocentric cohort study of 90 cases and review of the literature. Eur J Intern Med. 2002;13: Palmblad J, Papadaki HA, Eliopoulos G. Acute and chronic neutropenias. What is new? J Intern Med. 2001;250: HYPOTHYROIDISM AND PULMONARY ARTERIAL HYPERTENSION To the Editor: In 1999, our group reported a high prevalence of hypothyroidism in patients with primary pulmonary arterial hypertension (22.5%) (1). More recently, evidence for autoimmune antibodies, including antithyroglobulin antibodies, was found in 67% of 24 patients with primary pul by Excerpta Medica Inc /04/$ see front matter All rights reserved.

69 Letters to the Editor monary arterial hypertension (2). Since our prior report, we have instituted thyroid function screening for all outpatients with pulmonary hypertension. Between May 1999 and April 2001, 197 patients with primary pulmonary hypertension, scleroderma-associated pulmonary hypertension, pulmonary hypertension due to Eisenmenger phenomenon, or chronic thromboembolic pulmonary hypertension were evaluated. Blood was collected to obtain a thyroid-stimulating hormone (TSH) level on all patients (third-generation assay with a reference range of 0.40 to 5.50 IU/ ml). Hypothyroidism was defined as a TSH level greater than 5.50 IU/mL or the use of oral thyroid replacement therapy for a previously confirmed diagnosis of primary hypothyroidism. The study group comprised patients with primary pulmonary hypertension or scleroderma-associated pulmonary hypertension. Patients with chronic thromboembolic pulmonary hypertension or pulmonary hypertension due to Eisenmenger phenomenon constituted the control group. A chi-squared test was used to compare the hypothyroidism prevalence in the two groups. Odds ratio (OR) with 95% confidence intervals (CI) were calculated. The study group included 144 patients and the control group included 53 patients. Of the 144 patients in the study population, 29 (20.1%) had hypothyroidism versus 4 (7.6%) of the 53 patients in the control group (OR 3.09; 95% CI: 1.03 to 9.26; P 0.03). Of the 29 patients with hypothyroidism in the study group, 17 were taking oral thyroid replacement and 12 had newly diagnosed hypothyroidism based on an elevated TSH level (mean [ SD] TSH level, IU/mL). The primary pulmonary hypertension only subgroup consisted of 121 patients and they had a hypothyroidism prevalence of 19.0% (n 23), which was still higher than the 7.6% in the control group (OR 2.88; 95% CI: 0.94 to 8.77; P 0.05). The 20.1% prevalence rate in this study is comparable to the 22.5% found in our 1999 cross-sectional study (1). Whether autoimmunity is a shared insult in the pathogenesis of both diseases and whether thyroid hormone deficiency contributes to the vasomotor instability in pulmonary hypertension remains to be seen. Case reports of Raynaud s phenomenon that improved after hormonal replacement in patients with hypothyroidism suggest a vasomotor role for the thyroid hormone (3). Onthe other hand, pulmonary hypertension occurs in patients with connective tissue diseases (4); and autoimmune antibodies, including antithyroglobulin antibodies, have been found with increased frequency in patients with primary pulmonary hypertension (2,5). This may explain the link with hypothyroidism because an autoimmune injury is the cause of most cases of hypothyroidism in the young (6). Although we were unable to verify the actual hypothyroidism diagnosis in patients who had already started hormonal replacement therapy, our screening did not include an assessment of the TSH response to thyrotropin-releasing hormone administration, which could have uncovered additional cases of subclinical hypothyroidism. Another limitation is that patients with scleroderma have been reported to have an increased prevalence of hypothyroidism, although with a lower prevalence than in our scleroderma-associated pulmonary hypertension group (7). The primary pulmonary hypertension group alone still had an increased prevalence of hypothyroidism when compared with the control group, although with borderline statistical significance. Because patients with primary pulmonary hypertension and scleroderma-associated pulmonary hypertension have an increased prevalence of hypothyroidism, TSH screening is warranted. Ziad W. Ghamra, MD Raed A. Dweik, MD Alejandro C. Arroliga, MD Department of Pulmonary and Critical Care Medicine The Cleveland Clinic Foundation Cleveland, Ohio 1. Curnock AL, Dweik RA, Higgins BH, Saadi HF, Arroliga AC. High prevalence of hypothyroidism in patients with primary pulmonary hypertension. Am J Med Sci. 1999;318: Chu JW, Kao PN, Faul JL, Doyle RL. High prevalence of autoimmune thyroid disease in pulmonary arterial hypertension. Chest. 2002;122: Lateiwish AM, Feher J, Baraczka K, Racz K, Kiss R, Glaz E. Remission of Raynaud s phenomenon after L-thyroxine therapy in a patient with hypothyroidism. J Endocr Invest. 1992;15: Gurubhagavatula I, Palevsky HI. Pulmonary hypertension in systemic autoimmune disease. Rheum Dis Clin North Am. 1997;23: Yanai-Landau H, Amital H, Bar-Dayan Y, et al. Autoimmune aspects of primary pulmonary hypertension. Pathobiology. 1995;63: Singer PA, Cooper DS, Levy EG, et al. Treatment guidelines for patients with hyperthyroidism and hypothyroidism. Standards of Care Committee, American Thyroid Association. JAMA. 1995;273: Kahl LE, Medsger TA, Klein I. Prospective evaluation of thyroid function in patients with systemic sclerosis. J Rheum. 1986;13: LONGITUDINAL MYELITIS IN A PREGNANT PATIENT WITH SLE To the Editor: Myelopathy is a rare central nervous system complication associated with systemic lupus erythematosus (SLE), with transverse myelopathy being the most common form (1). Although still a subject of debate, administering a corticosteroid together with an immunosuppressive agent appears to be the best treatment approach (2). Longitudinal myelitis is an uncommon myelopathy with dis- March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume

70 Letters to the Editor Figure. Spinal cord magnetic resonance image showing multiple nodules with increased signal intensity on T2-weighted images and cord swelling. tinct magnetic resonance imaging (MRI) findings and a generally poor prognosis (3). We report a more uncommon case of longitudinal myelitis during pregnancy and its favorable outcome after corticosteroid and plasma exchange treatment. The patient, a 35-year-old woman, was diagnosed as having SLE at age 19 years on the basis of acute chorea, oral ulceration, arthralgias, and strongly positive antinuclear antibody level. Her first pregnancy, in 2000, was complicated by preeclampsia, necessitating delivery of a live male infant by cesarean section at 34 weeks. During her second pregnancy, in 2002, she was admitted at 13 weeks for paraplegia. The patient displayed complete motor deficit, spastic reflexes, bilateral extensor plantar responses, and sphincter disorders. Lower extremity weakness was first noted 4 days before admission. She had taken no drug other than aspirin, as hydroxychloroquine had been discontinued at the beginning of the second pregnancy, and corticosteroids some years before the first pregnancy. The test results showed the following values: white blood cell count, / L; C-reactive protein level, 5 mg/l; normal urea electrolyte and liver function; antinuclear antibody level, 1/1280, with native DNA level, 43 IU/L; normal complement and negative lupus anticoagulant, anticardiolipin, antiphosphatidylethanolamine, antiprothrombin, and 2 -glycoprotein antibody. Viral titers and syphilis screening were negative. Spinal cord MRI showed more than four spinal segments to be involved, with increased signal intensity on T2- weighted images, suggestive of longitudinal myelitis (Figure). Lumbar puncture showed clear cerebrospinal fluid, white blood cell count of 10/ L, elevated protein level (100 mg/dl), and negative microbiological findings. Obstetrical ultrasonography was normal. The diagnosis was SLE-related longitudinal myelitis. Due to the severity of the neurological symptoms, aggressive treatment was initiated immediately. The patient received intravenous methylprednisolone (1 g/d for 3 days) followed by orally administered prednisone (1 mg/kg/d), hydroxychloroquine (200 mg twice daily), and heparin anticoagulation therapy. Cyclophosphamide was not used due to its teratogenicity, but plasma exchange (3 L/d) was carried out for 10 days. Her health gradually improved, and she was able to walk after 10 days of treatment. To prevent pregnancyassociated hypertension, gravid diabetes, and SLE flare-up, azathioprine treatment (50 mg twice daily) was started at 16 weeks and corticosteroid dosage was progressively decreased to 20 mg/d. At 35 weeks, a live female infant weighing 2.3 kg was delivered by cesarean section. Six months after delivery, clinical examination and MRI of the mother s spinal cord were normal. Antiphospholipid antibody was still negative, so anticoagulation therapy was terminated. Corticosteroid dosage was decreased to 10 mg/d, but azathioprine treatment has been continued. One year after delivery, the mother s and child s health remain satisfactory. Early aggressive treatment for longitudinal myelitis with corticosteroid and plasma exchange, which is well tolerated during pregnancy (4), in this case ensured a normal pregnancy, with complete recovery of the mother. The use of plasma exchange instead of cyclophosphamide appears to have had a beneficial effect. In fact, corticosteroid and cyclophosphamide treatment for SLE-related longitudinal myelitis has only been completely successful in one reported case (5), with slow improvement noted by Tellez-Zenteno in 1 of 6 patients (3). In our patient, the underlying mechanism involved nonvascular injury (6). Moreover, the discontinuation of hydroxychloroquine treatment before the onset of longitudinal myelitis could be implicated in this flare-up (7). In addition, no SLE flare-up recurrence was observed after the decrease of corticosteroid dosage before and after delivery, probably because of the use of azathioprine (8). Olivier Moranne, MD Eric Hachulla, MD, PhD Anne Sophie Valat, MD Gustavo Sotoares, MD Dominique Pagniez, MD Eric Boulanger, MD Centre Hospitalier Régional Universitaire Lille, France 1. Kovacs B, Lafferty TL, Brent LH, et al. Transverse myelopathy in systemic lupus erythematosus: an analysis of 14 cases and review of the literature. Ann Rheum Dis. 2000;59: Mok CC, Lau CS, Chan EY, et al. Acute transverse myelopathy in systemic lupus erythematosus: clinical presentation, treatment and outcome. J Rheumatol. 1998;25: Tellez-Zenteno JF, Remes-Troche JM, Negrete-Pulido RO, et al. Longitudinal myelitis associated with systemic lupus erythematosus: clinical features and magnetic resonance imaging of six cases. Lupus. 2001;10: March 1, 2004 THE AMERICAN JOURNAL OF MEDICINE Volume 116