NEW ENGLAND JOURNAL of MEDICINE

Volume 357, July 26, 2007, Number 4, pp.321- 428

Article Summaries

Understanding Hospice — An Underutilized Option for Life's Final Chapter G. Gazelle

Letting Go of the Rope — Aggressive Treatment, Hospice Care, and Open Access A. A. Wright and I. T. Katz

A Day in the Life of Oscar the Cat D. M. Dosa

A Multicenter, Randomized, Controlled Trial of Dexamethasone for Bronchiolitis H. M. Corneli and Others

High-Dose Chemotherapy and Stem-Cell Rescue for Metastatic Germ-Cell Tumors L. H. Einhorn and Others

Partial Thrombosis of the False Lumen in Patients with Acute Type B Aortic Dissection T. T. Tsai and Others

Rofecoxib and Cardiovascular Adverse Events in Adjuvant Treatment of Colorectal Cancer D. J. Kerr and Others

The Spread of Obesity in a Large Social Network over 32 Years N. A. Christakis and J. H. Fowler

Skin and Soft-Tissue Infections Caused by Methicillin-Resistant Staphylococcus aureus R. S. Daum

Positive-Pressure Ventilation with a Face Mask and a Bag-Valve Device R. Ortega, A. K. Mehio, A. Woo, and D. H. Hafez

Metastatic Germ-Cell Cancer J. Menke and E. Grabbe

Ascending Aortic Aneurysm in a Young Adult P. D. Harris and C. Cokis

Case 23-2007 — A 9-Year-Old Boy with Bone Pain, Rash, and Gingival Hypertrophy C. P. Duggan, S. J. Westra, and A. E. Rosenberg

Therapy for Bronchiolitis: When Some Become None C. B. Hall

Network Medicine — From Obesity to the "Diseasome" A.-L. Barabási

Cancer and the Constitution — Choice at Life's End G. J. Annas

PCI for Stable Coronary Disease

Diabetic Gastroparesis

Case 15-2007: A Woman with Asthma and Cardiorespiratory Arrest Folate Deficiency and Plasma Homocysteine during Increased Oxidative Stress

Radiofrequency Ablation of a Tumor Causing Oncogenic Osteomalacia

Culturing Life: How Cells Became Technologies

Bioethics in Law

Survival of the Sickest: A Medical Maverick Discovers Why We Need Disease

Diabetic Gastroparesis

Posaconazole or Fluconazole for Prophylaxis in Severe Graft-versus-Host Disease

The NEW ENGLAND JOURNAL of MEDICINE

Perspective july 26, 2007

Understanding Hospice — An Underutilized Option for Life’s Final Chapter Gail Gazelle, M.D. t was Mr. G.’s third exacerbation of congestive many aspects of hospice care are heart failure in the past 6 months. Eighty-three still misunderstood by both phy- I sicians and patients. years old, he had New York Heart Association class For instance, many would not IV heart failure, end-stage coronary artery disease, consider Mr. G. to be a candidate for hospice care. He did not and insulin-dependent diabetes. of life. Relieved, Mr. G. acknowl- have cancer, and his death was Although he had never wanted edged that he would prefer to probably months, not days, away. to be put on a ventilator, this avoid rehospitalization. The fact is, however, that slight- time his shortness of breath was Introduced in the United States ly less than half of hospice pa- so terrifying that he felt he had as a grassroots movement more tients have terminal cancer; near no choice. After having a good re- than 30 years ago and added as ly 40% of hospice admissions are sponse to diuresis, he was success a Medicare entitlement in 1983, for end-stage cardiac disease, fully extubated and transferred hospice care is now considered end-stage dementia, debility, pul- out of the coronary care unit. part of mainstream medicine, as monary disease, and stroke.1 Two days later, a hospitalist evidenced by growing patient en- Patients and clinicians may suggested to Mr. G. and his wife rollment and Medicare expendi- also not realize that hospice care that given his advanced disease, tures (see table). In 2005, more at home is free. Medicare is the he should consider going home than 1.2 million Americans re- primary payer for hospice care and receiving hospice care there. ceived hospice care, and between in approximately 80% of cases, Sensing the couple’s fear, she re- 2000 and 2004, the percentage with care most often provided assured them that death was not of Medicare decedents that had in the patient’s home. Commer- imminent and that members of been enrolled in hospice pro- cial insurers also provide hospice the hospice staff would work to grams increased by almost 50%. benefits, but the specifics of cov- ensure the best possible quality But despite its increased use, erage vary. Under Medicare, most

n engl j med 357;4 www.nejm.org july 26, 2007 321 PERSPECTIVE understanding hospice — an underutilized option for life’s final chapter

possible in the face of advanced Use of Hospice Care among Medicare Beneficiaries from 2000 to 2004.* incurable disease. Variable 2000 2004 % Increase, 2000–2004 To determine eligibility, the Beneficiaries in hospice care 534,261 797,117 49 attending physician and hospice (no.) medical director must certify Payment (billions of $) 2.9 6.7 130 that to the best of their judg- Time in hospice care 26 52 101 ment, the patient is more likely (millions of days) than not to die within 6 months. Decedents who had been in 22 31 — Responsibility for determining hospice care (%) ongoing eligibility rests with the director. To assist physicians in * Data are from the Center for Medicare and Medicaid Services and the Medicare Payment Advisory Commission (MedPAC) and include Puerto Rico. prognosticating, Medicare provides broad guidelines for many medical conditions (see box), expenses related to the terminal length of hospice service is only but these guidelines do not rep- diagnosis are paid in full, includ- 26 days, with one third of pa- resent hard-and-fast require- ing all medication and equip- tients referred to hospice care ments. Coexisting conditions or ment and all visits by hospice during the last week of life.1 a particularly rapid functional de- nurses and home health aides. (Ex Factors contributing to late re- cline can outweigh strict adher- penses related to other diagnoses ferral include application of a ence to written requirements. remain covered by the patient’s curative model to end-stage in- After enrollment, a plan of primary insurance provider.) Oth- curable illnesses; Medicare’s per care is developed in accordance er hallmark hospice services in- diem hospice reimbursement, with the needs and wishes of clude intensive emotional and which precludes costly, aggres- the patient and family, often spiritual counseling, 24-hour cri- sive therapies; and the mistaken tempered by the presence or ab- sis management, and bereave- view that patients must have a sence of caregivers to participate ment support for at least 1 year do-not-resuscitate order. in day-to-day care. The primary after the patient’s death. However, the most important goal is to ensure that pain and Hospice care can successfully factors in delayed referrals ap- such symptoms as insomnia, address the critical end-of-life pear to relate to physician atti- dyspnea, depression, constipa- concerns that have been identi- tudes. In its first position paper tion, agitation, nausea, and fied in numerous studies: dying on the topic of cancer and dy- emotional and spiritual distress with dignity, dying at home and ing, the American Society of are aggressively addressed. Most without unnecessary pain, and Clinical Oncology acknowledged clinical care is provided by a reducing the burden placed on that many oncologists and other hospice nurse, and the vast ma- family caregivers.2‑4 Evaluation physicians regard the death of a jority of patients are not seen by studies reveal consistently high patient as a professional failure.5 a physician. Mr. G.’s plan of care family satisfaction, with 98% of Many also fear that they will de- included continuing high-dose family members willing to rec- stroy their patients’ hope, which furosemide, adding low-dose ommend hospice care to others physicians may believe lies only lorazepam for the anxiety that in need.1 And the extensive ex- in efforts to increase the quan- typically accompanies shortness pertise of physicians specializ- tity rather than quality of life. of breath, and initiating lowing in hospice and palliative Furthermore, physicians receive dose liquid opioids, a mainstay medicine was recognized in little training in the compas- in the management of dyspnea. 2006, when the field was ac- sionate discussion of bad news. To address Mr. G.’s nonmedi- credited as a fully independent But perhaps the most critical cal needs, a home health aide medical subspecialty. factor is that physicians view provided assistance with per- Despite these benefits and hospice care as something re- sonal hygiene and dressing for the general understanding by served for the imminently dying an hour each day, 5 days a week. clinicians that at least 6 months instead of as a service designed The hospice social worker of- of care are provided, the median to help people live as well as fered to have a volunteer shop

322 n engl j med 357;4 www.nejm.org july 26, 2007 PERSPECTIVE understanding hospice — an underutilized option for life’s final chapter

Medicare Hospice Eligibility Guidelines for Selected Diagnoses.*

Alzheimer’s disease Eligibility is based on Reisberg Functional Assessment Staging (FAST), stage 7 or greater. Stage 7 is defined by the following criteria: need for assistance with at least three activities of daily living increased frequency of incontinence of bowel and bladder ability to speak only six or fewer intelligible words in the course of an average day In addition to the FAST criteria, the patient must have one of the following: a history of upper urinary tract infection, sepsis, or pneumonia within the past 12 months multiple stage III or IV decubitus ulcers within the past 12 months If the patient meets neither of the two preceding criteria, the patient must show nutritional decline, as evidenced by one of the following: unintentional progressive weight loss of 10% of body weight over the past 6 months a serum albumin level of <2.5 g/dl Pulmonary disease Eligibility is based on severe chronic lung disease, as defined by the following criteria: disabling dyspnea at rest and poor response to bronchodilators, resulting in decreased functional capacity disease progression reflected in increased emergency department or physician visits or by increased hospitalizations hypoxemia at rest — oxygen saturation ≤88% with patient breathing ambient air Heart disease Eligibility is based on New York Heart Association classification of class IV, as defined by the following criterion: inability to carry out any physical activity without discomfort (documentation of an ejection fraction of <20% is helpful but not required) In addition, the patient must be optimally treated with diuretics and vasodilators as tolerated in relation to blood pressure and renal function If the criteria for class IV do not apply, eligibility can be based on one of the following: the patient has angina, which must be present at rest or resistant to standard nitrate therapy the patient is not a candidate for or declines invasive procedures Debility (no one specific terminal diagnosis identified) Eligibility is based on the progression of disease, as documented by one of the following: recurrent or intractable infections, such as pneumonia, upper urinary tract infections, or sepsis progressive weight loss of >10% of body weight over the past 6 months dysphagia leading to recurrent aspiration or inadequate nutritional intake progressive deep decubitus ulcers

* The guidelines do not represent hard-and-fast requirements. The presence of clinically significant coexisting conditions or rapid functional decline can substitute for some criteria. The guidelines may also vary with the Medicare fiscal intermediary. Information is from National Government Services (www.ugsmedicare.com/providers/LMRP/documents/Hospice%20REVISED%2010-01-06.pdf) and is based on personal communication with James Cope, medical director of National Government Services.

for groceries and provide com- for past acts, she encouraged coordinator, and the medical di- panionship. The social worker visits by the hospice chaplain. rector — meet to discuss the also talked with the family and Hospice emphasizes an inter- needs of the patient and family. identified the need to address disciplinary approach to care. In In the interim, nurses call attend- Mr. G.’s anxiety and his wife’s most cases, at least once every ing physicians with their recom- fears about the future. Noting other week, the hospice team mendations. that Mr. G. had concerns about — nurses, social workers, a pas- One serious challenge in hos- whether God was punishing him toral counselor, the bereavement pice care is that attending physi-

n engl j med 357;4 www.nejm.org july 26, 2007 323 PERSPECTIVE understanding hospice — an underutilized option for life’s final chapter

cians typically receive little to no frequently; during one such epi- hospice care offers and work to training in the use of medica- sode, he received intravenous fu- overcome barriers in talking frank tions for pain and symptom rosemide in his home, since he ly with patients about what lies management and thus rely on a wanted to avoid further hospital- ahead. presumed level of expertise on izations. During 4 months of the part of the hospice nurse. hospice care, Mr. G.’s condition Dr. Gazelle is a member of the Division of Given the current nursing short- gradually deteriorated, with in- General Medicine and Primary Care at Brigham and Women’s Hospital and presi- age, however, such an assump- creasing weakness, dyspnea, and dent of MD Can Help — both in Boston. tion of competency may or may cardiac cachexia. Near the end, not be well founded. Attending his family and friends gathered, 1. NHCPO’s facts and figures — 2005 find- physicians should routinely eval- and he died peacefully with his ings. Alexandria, VA: National Hospice and Palliative Care Organization, 2006. (Ac- uate recommendations and wife and nurse at his side. De- cessed July 6, 2007, at http://nhpco.org/ should have a low threshold for spite his family’s grief, they ex- files/public/2005-facts-and-figures.pdf.) reviewing cases with the hospice pressed their appreciation that 2. Singer PA, Martin DK, Kelner M. Quality end-of-life care: patients’ perspectives. JAMA medical director. Mr. G. had maintained a reason- 1999;281:163-8. As a patient’s disease progress ably high quality of life and had 3. Emanuel LL, Alpert HR, Baldwin DC, es, the hospice plan shifts to ac- died in his home as he had Emanuel EJ. What terminally ill patients care about: toward a validated construct of pa- commodate decreasing indepen- wished. tients’ perspectives. J Palliat Med 2000;3: dence, alterations in symptoms, With the growing number of 419-31. and changing psychosocial needs. baby boomers seeking more con- 4. Steinhauser KE, Clipp EC, McNeilly M, Christakis NA, McIntyre LM, Tulsky JA. In In Mr. G.’s case, the realization trol over all aspects of their search of a good death: observations of pa- that his symptoms could be health care, the use of hospice tients, families, and providers. Ann Intern managed at home lessened his care will probably continue to Med 2000;132:825-32. 5. American Society of Clinical Oncology. anxiety, which in turn decreased increase. It is especially impor- Cancer care during the last phase of life. episodes of chest pain. Flash tant, therefore, that physicians J Clin Oncol 1998;16:1986-96. pulmonary edema occurred less become more familiar with what Copyright © 2007 Massachusetts Medical Society.

Letting Go of the Rope — Aggressive Treatment, Hospice Care, and Open Access Alexi A. Wright, M.D., and Ingrid T. Katz, M.D., M.H.S.

ore Americans are choosing and continuing to receive total not negotiate pricing or spread Mhospice for end-of-life care, parenteral nutritional support. the cost of expensive medications but ironically, hospice patients Unfortunately, treatment op- across many patients. A few large increasingly are forced to give tions are often limited by the hospices offer what is called up effective palliative treatments economic constraints of hospice open-access care, which allows along with aggressive medical in- care. The hospice that was the patients to add hospice care to tervention. For Joanne Doolin, a closest to Doolin’s Boston-area their current medical treatment, 64-year-old mother of three who home would accept only patients but this option is not available spent her last 2 years of life willing to forgo life-sustaining in Massachusetts. fighting colon cancer that even- treatments, including chemo- The Medicare hospice benefit tually made it impossible to eat, therapy and parenteral nutrition. reimburses hospices on a per enrollment in hospice care in- It cares for only about 20 pa- diem basis, paying fixed inpatient volved a difficult trade-off: with tients at a time with three nurs- and outpatient fees regardless of only a few weeks left to live and es, a manager, a part-time chap- services provided. Despite adjust- her daughter’s wedding approach lain, and a medical director who ments for inflation, the fees ing, Doolin was forced to choose works there one morning a have not kept up with the cost between entering hospice care week. As a small program, it can of cutting-edge palliative treat-

324 n engl j med 357;4 www.nejm.org july 26, 2007 PERSPECTIVE letting go of the rope — aggressive treatment, hospice care, and open access

ments. Many patients who meet represents 2.7% of Medicare’s had cancer, although patients the criterion for hospice care — total hospice payments) (see Table with dementia, heart disease, having less than 6 months to 2). The average outpatient fee and fatal lung conditions are in- live — still opt for palliation was $126 for a typical day of creasingly entering hospice care. from oral chemotherapies, radi- care, an amount that must cover Diane Meier, director of the Cen- ation, antiemetics, or blood trans- nursing care; contributions from ter to Advance Palliative Care at fusions. But these treatments can social workers, chaplains, and the Mount Sinai School of Medi- cost more than $10,000 per month volunteers; and all drugs and cine in New York, argues that — too much for most hospice durable medical equipment, as “palliative care and hospice are programs (see Table 1). well as 13 months of bereave- the only medical disciplines Although some observers wor- ment support. where nurses and physicians fo- ry that nationwide open access Despite differences among cus on the whole person.” could bankrupt Medicare, most hospice programs, patient and Most patients, however, wait agree that per diem reimburse- family satisfaction is high; in until the last few weeks of life ment rates remain unacceptably 2005, one third of the 2.4 mil- to enroll. In 2005, the median low: in 2006, hospices were paid lion Americans who died were hospice stay was 26 days. One an average of $563 per patient receiving hospice care.1 The contributing factor is late refer- per day for inpatient care (which largest proportion of patients rals by oncologists, who routinely

Table 1. Approximate Costs of Drugs Commonly Used by Hospices and Oncologists for Palliative Treatment.*

Class and Drug Dose/24 Hr Price $ Emergency hospice pack 1-night supply of morphine oral concentrate, lorazepam, haloperidol, 50.00 prochlorperazine, and Senokot Pain relief and laxatives Morphine oral concentrate 200 mg 186.00 MS Contin generic 200 mg 294.00 Oxycodone ER 160 mg 562.20 Fentanyl patch 100 µg 533.60 Senna 2 tablets 6.60 Antiemetics, anxiolytics, anticholinergics Lorazepam 6 mg 115.20 Prochlorperazine 20 mg 53.70 Haloperidol oral concentrate 2 mg 22.56 Scopolamine patches 1 patch 9.18† Ondansetron 8 mg 1,113.90 Oral chemotherapy and supportive care Temozolomide 200 mg 1,867.40 Capecitabine 2500 mg 1,883.70 Erlotinib 150 mg 3,906.60 Zoledronic acid 4 mg 824.56 Erythropoietin 40,000 units/wk 2,504.00

* Prices represent estimates for a 1-month supply of medicine at average wholesale price (Medi-Span, http://www.medi-span.com, accessed July 27, 2007). Many hospice programs, hospitals, and physicians are able to negotiate lower prices than those listed here; some hospices negotiate a fixed daily rate per patient to cover all necessary medications. † Each patch lasts 72 hours. The patch is given very close to the end of life.

n engl j med 357;4 www.nejm.org july 26, 2007 325 PERSPECTIVE letting go of the rope — aggressive treatment, hospice care, and open access

versity–Massey Cancer Center, Table 2. Medicare Hospice Payment Categories and Rates, Fiscal Year 2006.* in Richmond, “that it requires Type of Care Base Payment Rate % of Days divorcing yourself from your pa- $ tient’s care because you can’t be Home care on a typical day 126/day 93.0 their cancer doctor anymore. As Home care during periods of patient crisis 31/hr 4.1 soon as you enroll in hospice, Inpatient care for a short period to provide 131/day 0.2 there goes your Aranesp, your respite for primary caregiver Zometa, and your Zofran. . . . I Inpatient care to treat symptoms that can- 563/day 2.7 can’t do anything but adjust pain not be managed at home meds and hold hands. These are wonderful things to do, but they * “Home care during periods of patient crisis” is defined as crisis care delivered in won’t keep my office running.” the home for 8 or more hours within a 24-hour period beginning at midnight; care must be delivered by a nurse for more than half of the hours to qualify for this Many hospice directors counter level of payment. The minimum daily payment rate for such care is $246 per day that oncologists abandon their and the maximum is $738 per day. Data are from the Medicare Payment Advisory patients when they can no lon- Commission (MedPAC). ger visit the office. A few large hospices and in- overestimate patients’ lifespans.2 home nursing care and access to surance companies are trying to Many patients are referred only an infusion company for nutri- prevent these situations with when no other option remains. tion until she was ready for hos- open-access programs. Last year, In addition, many patients fear pice care. She spent more than a Capital Hospice, based in Wash- that they will not receive enough month at home, visited her fa- ington, D.C., paid for palliative medical services in hospice care. vorite casino, and attended her chemotherapy, radiation, dialysis, “It felt like I was trading in the daughter’s wedding. But then blood transfusions, parenteral nu- Lamborghini of medical care for Doolin’s health suddenly deteri- trition, antibiotics, and other ex- an old pick-up truck driving down orated, and she needed urgent pensive intravenous medications. a rutted road,” said one patient medical care and pain manage- With an average daily census of with cancer. Optimal end-of-life ment. Her family contacted the 606 patients, the program can support often necessitates careful bridge program’s hospice but spread out the expense. Presi- titration of opioid, antipsychotic, could not enroll her in time. Dur- dent and chief executive officer and anxiolytic drugs, which can ing Doolin’s last few hours, care Malene Davis likens open access sometimes require a doctor’s was provided by a haphazard mix to “two ropes hanging from the presence. But few patients ever of people, including her family, ceiling. We’ve asked people to meet a physician after enrolling a covering oncologist, a pharma- hold on to the aggressive-treat- for hospice care; there are no cist, and compassionate local fire ment rope with both hands,” she rules mandating the degree of fighters. One year later, Doolin’s says, “but when they go on hos- physician involvement. Medicare family is still angry over the pice we tell them to let go com- does not even collect information forced choice between parenteral pletely. Open access gives people on the number, frequency, or du- nutrition and hospice care. They the choice to let go of active ration of visits or on which per- believe she would have suffered treatment with one hand and sonnel provide which aspects of less in an open-access hospice grab on to the hospice rope un- care. Each hospice program de- program. til they feel comfortable letting cides what services to offer, and The disconnect between pre- the other hand go.” family members often must fill hospice and hospice care seems The large insurance company in the gaps. absolute to physicians as well. UnitedHealth offers a basic open- Like most patients with termi- The Medicare hospice benefit “is access hospice benefit to nearly nal illness, Joanne Doolin chose so restrictive,” says Thomas Smith, 26 million members and a small- ongoing medical treatment over chair of the division of hematol- er hospice program in 11 cities hospice. She entered a bridge-to- ogy–oncology and palliative care that includes physician home vis- hospice program that provided at Virginia Commonwealth Uni- its and reviews of care. A com-

326 n engl j med 357;4 www.nejm.org july 26, 2007 PERSPECTIVE letting go of the rope — aggressive treatment, hospice care, and open access

access (see pie chart). Elsewhere, by Medicare, particularly for pa- >400 Patients 2.5% patients and hospice directors tients with dementia or other ill- 125–400 must make tough choices. nesses, who often live with the Patients The only randomized trial to disease for more than 6 months.

15.0% <25 Patients date examining standard cancer “Baby boomers are going to want care both with and without hos- everything — from death coaches 35.0% 75–124 14.5% pice support showed no signifi- to powerful drugs,” says Davis, Patients cant difference in survival rates, “but we’ve got to begin grap- but it did show significant im- pling with tough choices if we’re 33.0% provements in quality of life going to stay in business for when cancer care and hospice $150 a day.” 25–74 Patients care were combined. Preliminary Some choices will undoubt- analysis revealed a 27% cost re- edly involve better definitions of Size of Hospice. duction in the combined-care palliative treatment. Currently, Data are from the National Hospice and Palliative Care Organization. group, which received less chemo oncologists focus on how well a therapy and diagnostic testing and tumor responds to chemotherapy, 3 ICM AUTHOR: Wright RETAKE 1st required fewer hospitalizations. but they will soon have to exam- 2nd REG F FIGUREpany spokesperson1 of 1 says that the Nevertheless, many experts ine improvement of symptoms 3rd CASE cost is negligible as comparedRevised worry that open access may be and quality of life to justify treat- Line 4-C EMail with the cost of its otherSIZE pro- prohibitively expensive. A 1990 ment costs. Meanwhile, patients ARTIST: ts H/T H/T Enon grams. In 2004,Combo Aetna started study showed that most patients will simply have to hope for ac- its AUTHOR, Compassionate PLEASE NOTE: Care Pro- with cancer would choose to un- cess to a hospice that is large Figure hasgram, been redrawn which and usestype has International been reset. dergo toxic chemotherapy de- enough to help them. Please check carefully. Classification of Diseases, Ninth Revi- spite marginal potential bene- 4 Dr. Wright is a fellow in hematology–oncol- JOB: 35704 sion, codes and pharmaceuticalISSUE: 07-26-07 fits ; a study in 2004 reported ogy at the Dana–Farber Cancer Institute, information to identify members increasingly aggressive care at the and Dr. Katz is a fellow in infectious disease with terminal illnesses; the end of life.5 Patients with con- at the Beth Israel Deaconess Medical Cen- members are then contacted by gestive heart failure also face ter — both in Boston. nurse case managers, who offer difficult choices, since life-sus- 1. 2005 National summary of hospice care: emotional support, care coordi- taining medications can cost statistics and trends from the 2005 national nation, and information about $1,300 per day. “Whoever wrote data set and the 2005 NHPCO membership end-of-life planning and symp- [Medicare’s hospice] policy has survey. Alexandria, VA: National Hospice and Palliative Care Organization, Novem- tom relief. Early results suggest never taken care of sick pa- ber 2006:4. that members appreciate the ad- tients,” argues Diane Meier. 2. Lamont EB, Christakis NA. Prognostic ditional support that tailored case “Our patients are fighting for disclosure to patients with cancer near the end of life. Ann Intern Med 2001;134:1096- management provides; more mem their lives and will do anything 105. bers are enrolling in hospice, and to extend the length of time they 3. Finn JW, Pienta KJ, Parzuchowski J, Wor- the program is reducing rates of live, as long as they have some den F. Palliative care project: bridging active treatment and hospice for terminal cancer. unnecessary hospitalizations. quality of life.” Proc Am Soc Clin Oncol 2002;21. abstract. But these programs remain the CMS foresees an annual in- 4. Slevin ML, Stubbs L, Plant HJ, et al. Atti- exception. According to the Cen- crease of 9% in hospice spend- tudes to chemotherapy: comparing views of patients with cancer with those of doctors, ter for Medicare and Medicaid ing over the next decade, which nurses, and general public. BMJ 1990;300: Services (CMS), only 2.5% of the will outpace increases for hospi- 1458-60. country’s 4100 hospices have an tals, physicians, skilled nursing 5. Earle CC, Neville BA, Landrum MB, Aya- nian JZ, Block SD, Weeks JC. Trends in the average daily census above 400 facilities, and home health ser- aggressiveness of cancer care near the end — commonly considered the vices. Many expect closer scruti- of life. J Clin Oncol 2004;22:315-21. minimum requirement for open ny of hospice reimbursements Copyright © 2007 Massachusetts Medical Society.

n engl j med 357;4 www.nejm.org july 26, 2007 327 PERSPECTIVE a day in the life of oscar the cat

A Day in the Life of Oscar the Cat David M. Dosa, M.D., M.P.H.

scar the Cat awakens from decides to head down the west Oscar takes no notice of the Ohis nap, opening a single wing first, along the way side- woman and leaps up onto the eye to survey his kingdom. From stepping Mr. S., who is slumped bed. He surveys Mrs. T. She is atop the desk in the doctor’s over on a couch in the hallway. clearly in the terminal phase of charting area, the cat peers With lips slightly pursed, he illness, and her breathing is la- down the two wings of the nurs- snores peacefully — perhaps bored. Oscar’s examination is in- ing home’s advanced dementia blissfully unaware of where he terrupted by a nurse, who walks unit. All quiet on the western is now living. Oscar continues in to ask the daughter whether and eastern fronts. Slowly, he Mrs. T. is uncomfortable and rises and extravagantly stretches needs more morphine. The his 2-year-old frame, first back- daughter shakes her head, and ward and then forward. He sits the nurse retreats. Oscar returns up and considers his next move. to his work. He sniffs the air, In the distance, a resident ap- gives Mrs. T. one final look, then proaches. It is Mrs. P., who has jumps off the bed and quickly been living on the dementia leaves the room. Not today. unit’s third floor for 3 years Making his way back up the now. She has long forgotten her hallway, Oscar arrives at Room family, even though they visit 313. The door is open, and he her almost daily. Moderately di- proceeds inside. Mrs. K. is rest- sheveled after eating her lunch, ing peacefully in her bed, her COLOR FIGURE Draft 1 06/22/07 half of which she now wears on down the hallway untilAuthor he Dosareach- breathing steady but shallow. Fig # 1 Title Oscar her shirt, Mrs. P. is taking one es its end and Room ME 310. The She is surrounded by photo- DE Malina Artist KMK of her many aimless strolls to door is closed, so Oscar AUTHORsits PLEASE and NOTE: graphs of her grandchildren and Figure has been redrawn and type has been reset Please check carefully nowhere. She glides toward Os- waits. He has importantIssue date 7/19/0 busi7 - one from her wedding day. De- car, pushing her walker and ness here. spite these keepsakes, she is muttering to herself with com- Twenty-five minutes later, the alone. Oscar jumps onto her bed plete disregard for her surround- door finally opens, and out and again sniffs the air. He ings. Perturbed, Oscar watches walks a nurse’s aide carrying pauses to consider the situation, her carefully and, as she walks dirty linens. “Hello, Oscar,” she and then turns around twice be- by, lets out a gentle hiss, a rat- says. “Are you going inside?” fore curling up beside Mrs. K. tlesnake-like warning that says Oscar lets her pass, then makes One hour passes. Oscar waits. “leave me alone.” She passes his way into the room, where A nurse walks into the room to him without a glance and con- there are two people. Lying in a check on her patient. She pauses tinues down the hallway. Oscar corner bed and facing the wall, to note Oscar’s presence. Con- is relieved. It is not yet Mrs. P.’s Mrs. T. is asleep in a fetal posi- cerned, she hurriedly leaves the time, and he wants nothing to tion. Her body is thin and wast- room and returns to her desk. do with her. ed from the breast cancer that She grabs Mrs. K.’s chart off the Oscar jumps down off the has been eating away at her or- medical-records rack and begins desk, relieved to be once more gans. She is mildly jaundiced to make phone calls. alone and in control of his do- and has not spoken in several Within a half hour the family main. He takes a few moments days. Sitting next to her is her starts to arrive. Chairs are to drink from his water bowl daughter, who glances up from brought into the room, where and grab a quick bite. Satisfied, her novel to warmly greet the the relatives begin their vigil. he enjoys another stretch and visitor. “Hello, Oscar. How are The priest is called to deliver sets out on his rounds. Oscar you today?” last rites. And still, Oscar has

328 n engl j med 357;4 www.nejm.org july 26, 2007 PERSPECTIVE a day in the life of oscar the cat

not budged, instead purring and plaque is awarded to Oscar the habilitation Center in Providence, gently nuzzling Mrs. K. A young Cat.” Oscar takes a quick drink Rhode Island. His mere presence at the grandson asks his mother, of water and returns to his desk bedside is viewed by physicians and “What is the cat doing here?” to curl up for a long rest. His nursing home staff as an almost abso- The mother, fighting back tears, day’s work is done. There will lute indicator of impending death, al- tells him, “He is here to help be no more deaths today, not in lowing staff members to adequately Grandma get to heaven.” Thirty Room 310 or in any other room notify families. Oscar has also pro- minutes later, Mrs. K. takes her for that matter. After all, no one vided companionship to those who last earthly breath. With this, dies on the third floor unless would otherwise have died alone. For Oscar sits up, looks around, Oscar pays a visit and stays his work, he is highly regarded by the then departs the room so quietly awhile. physicians and staff at Steere House that the grieving family barely and by the families of the residents notices. Note: Since he was adopted by staff whom he serves. On his way back to the chart- members as a kitten, Oscar the Cat has Dr. Dosa is a geriatrician at Rhode Island ing area, Oscar passes a plaque had an uncanny ability to predict Hospital and an assistant professor of med- mounted on the wall. On it is when residents are about to die. Thus icine at the Warren Alpert Medical School engraved a commendation from far, he has presided over the deaths of of Brown University — both in Providence. a local hospice agency: “For his more than 25 residents on the third Copyright © 2007 Massachusetts Medical Society. compassionate hospice care, this floor of Steere House Nursing and Re-

n engl j med 357;4 www.nejm.org july 26, 2007 329 Study. Social-network phenomena seem relevant to obesity, which appears to spread through social ties. See P. 370; Editorial, P. 404 Clinical Practice original article MRSA Skin and Soft-Tissue Infections Dexamethasone for Bronchiolitis A 37-year-old man presents with localized swelling and Infants with a first episode of wheezing diagnosed as tenderness of the left leg just below the knee; he sus- bronchiolitis are often treated with oral dexametha- pects a spider bite. Examination shows a 5-by-7-cm area sone. In this trial, children with bronchiolitis and no of erythema and warmth. A small area of necrotic skin history of asthma received oral dexamethasone or covers a central, fluctuant 2-by-2-cm area. The tem- placebo. There were no clinically significant differenc- perature is 38.3°C, the pulse rate 115 beats per minute, es in outcomes between the two groups. and the blood pressure 116/78 mm Hg. See P. 331; Editorial, P. 402; CME, P. 431 See P. 380; CME, P. 429 CASE RECORDS OF THE MASSACHUSETTS GENERAL HOSPITAL Original Article Salvage Therapy for Metastatic Germ-Cell Tumors A Boy with Bone Pain, Rash, and Gingival Hypertrophy This article summarizes the experience of a single institution in treating patients with metastatic testicular tu- A 9-year-old boy with autism was admitted to the hos- mors that did not respond to cisplatin-based chemo- pital because of pain in the hip, rash, and swelling of therapy. High-dose chemotherapy with hematopoietic the gingiva. Three months earlier, pain in his right hip stem-cell rescue was potentially curative in such cases. developed, and he eventually refused to walk. Physical examination disclosed no joint swelling, and radio- See P. 340 graphs of the pelvis and spine were normal. Indometh- Original Article acin was prescribed; the pain improved, but a petechial Partial Thrombosis of the False Lumen rash on the legs and gingival swelling developed. in Type B Aortic Dissection See P. 392; CME, P. 430 A cohort of 201 patients with type B acute aortic dis- Health Law, Ethics, and Human Rights section was classified according to whether the false Cancer and the Constitution lumen of the aorta was patent, partially thrombosed, The Abigail Alliance for Better Access to Developmen- or completely thrombosed. Patients with partial throm- tal Drugs sued the FDA, objecting to its policy prohib- bosis had a significantly higher mortality rate at 3 years. iting the sale of unapproved drugs and arguing that See P. 349 terminally ill patients with cancer should have access Original Article to experimental treatments after phase 1 studies. The Toxicity of Rofecoxib in Colorectal Cancer author discusses this case and explains why he thinks In this clinical trial, rofecoxib (25 mg per day) was stud- a ruling in favor of the Abigail Alliance is unlikely. ied in the prevention of recurrent colorectal cancer. see P. 408 Although the median duration of study treatment was Videos in Clinical Medicine only 7.4 months, rofecoxib therapy was associated with Face Mask and Bag-Valve Ventilation an increased risk of cardiovascular adverse events (rel- Providing positive-pressure ven- ative risk as compared with placebo, 2.66; P = 0.04). tilation with a face mask and a The results indicate that even short-term treatment bag-valve device can be a life- with rofecoxib may result in cardiovascular toxicity. saving maneuver. Although See P. 360 seemingly simple, the technique Special Article requires an understanding of The Spread of Obesity in a Large Social Network the airway anatomy, the equip- This article, which describes the person-to-person spread ment, and the indications. This video demonstrates the of obesity as a potential contributing factor in the U.S. equipment and technique used to provide positive- obesity epidemic, analyzes a densely interconnected pressure ventilation with a face mask and a bag-valve social network, using repeated assessments performed device. from 1971 to 2003 as part of the Framingham Heart See P. e4

330 n engl j med 357;4 www.nejm.org july 26, 2007 The new england journal of medicine

established in 1812 july 26, 2007 vol. 357 no. 4

A Multicenter, Randomized, Controlled Trial of Dexamethasone for Bronchiolitis

Howard M. Corneli, M.D., Joseph J. Zorc, M.D., Prashant Mahajan, M.D., M.P.H., Kathy N. Shaw, M.D., M.S.C.E., Richard Holubkov, Ph.D., Scott D. Reeves, M.D., Richard M. Ruddy, M.D., Baqir Malik, M.D., Kyle A. Nelson, M.D., M.P.H., Joan S. Bregstein, M.D., Kathleen M. Brown, M.D., Matthew N. Denenberg, M.D., Kathleen A. Lillis, M.D., Lynn Babcock Cimpello, M.D., James W. Tsung, M.D., Dominic A. Borgialli, D.O., M.P.H., Marc N. Baskin, M.D., Getachew Teshome, M.D., M.P.H., Mitchell A. Goldstein, M.D., David Monroe, M.D., J. Michael Dean, M.D., and Nathan Kuppermann, M.D., M.P.H., for the Bronchiolitis Study Group of the Pediatric Emergency Care Applied Research Network (PECARN)*

Abstr act

Background Bronchiolitis, the most common infection of the lower respiratory tract in infants, From the University of Utah (H.M.C.) and is a leading cause of hospitalization in childhood. Corticosteroids are commonly Central Data Management and Coordinat ing Center (R.H., J.M.D.), Salt Lake City; used to treat bronchiolitis, but evidence of their effectiveness is limited. the Children’s Hospital of Philadelphia, Methods Philadelphia (J.J.Z., K.N.S.); Children’s Hospital of Michigan, Detroit (P.M., B.M.); We conducted a double-blind, randomized trial comparing a single dose of oral Cincinnati Children’s Hospital Medical dexamethasone (1 mg per kilogram of body weight) with placebo in 600 children Center, Cincinnati (S.D.R., R.M.R.); Wash ington University, St. Louis (K.A.N.); Co (age range, 2 to 12 months) with a first episode of wheezing diagnosed in the emer- lumbia University, New York (J.S.B.); gency department as moderate-to-severe bronchiolitis (defined by a Respiratory Children’s National Medical Center, Wash Distress Assessment Instrument score ≥6). We enrolled patients at 20 emergency ington, DC (K.M.B.); Devos Children’s Hospital, Grand Rapids, MI (M.N.D.); departments during the months of November through April over a 3-year period. Women and Children’s Hospital of Buf The primary outcome was hospital admission after 4 hours of emergency department falo, Buffalo, NY (K.A.L.); University of observation. The secondary outcome was the Respiratory Assessment Change Score Rochester Medical Center, Rochester, NY (L.B.C.); Bellevue Hospital Center, New (RACS). We also evaluated later outcomes: length of hospital stay, later medical York (J.W.T.); Hurley Medical Center, Flint, visits or admissions, and adverse events. MI (D.A.B.); Children’s Hospital, Boston (M.N.B.); University of Maryland, Balti Results more (G.T.); Johns Hopkins Children’s Cen Baseline characteristics were similar in the two groups. The admission rate was ter, Baltimore (M.A.G.); Howard County General Hospital, Columbia, MD (D.M.); 39.7% for children assigned to dexamethasone, as compared with 41.0% for those and the University of California, Davis, assigned to placebo (absolute difference, −1.3%; 95% confidence interval [CI], −9.2 Medical Center, Sacramento (N.K.). Ad to 6.5). Both groups had respiratory improvement during observation; the mean dress reprint requests to Dr. Corneli at P.O. 4-hour RACS was −5.3 for dexamethasone, as compared with −4.8 for placebo (abso- Box 581289, Salt Lake City, UT 84158-1289. lute difference, −0.5; 95% CI, −1.3 to 0.3). Multivariate adjustment did not signifi- *Other investigators in the PECARN Bron cantly alter the results, nor were differences detected in later outcomes. chiolitis Study Group are listed in the Appendix. Conclusions N Engl J Med 2007;357:331-9. In infants with acute moderate-to-severe bronchiolitis who were treated in the emer- Copyright © 2007 Massachusetts Medical Society. gency department, a single dose of 1 mg of oral dexamethasone per kilogram did not significantly alter the rate of hospital admission, the respiratory status after 4 hours of observation, or later outcomes. (ClinicalTrials.gov number, NCT00119002.)

n engl j med 357;4 www.nejm.org july 26, 2007 331 T h e new england journal o f medicine

ronchiolitis is the leading cause of from the parent or guardian of each infant in- hospitalization for infants in the United cluded in the study. BStates,1 accounting for 100,000 admissions We included infants 2 to 12 months of age annually, with hospital charges alone estimated who were brought to the emergency department at $700 million.2 Hospitalization rates for infants with a first episode of bronchiolitis, defined as with bronchiolitis more than doubled between wheezing (with no prior bronchiolitis, wheezing, 1980 and 1996, and the proportion of infant hos- or asthma and no bronchodilator use before the pitalizations that were due to bronchiolitis more current illness), within 7 days after the onset of than tripled, from 5% to 16%.2 symptoms. In addition, the episode had to be mod Treatment for bronchiolitis is controversial. erate or severe as defined by a score on the Respi- Bronchodilators are commonly used,3,4 but they ratory Distress Assessment Instrument (RDAI)23 have not been shown to have consistent bene- of 6 or more (on a scale of 0 to 17, with higher fits.5-8 Although studies suggest that approximate scores indicating more severe respiratory symp- ly a quarter of infants hospitalized with bronchi- toms) (Table 1). We excluded infants with a prior olitis receive corticosteroids,3,4,9 the efficacy of adverse reaction to dexamethasone, known heart these agents has also not been consistently demon or lung disease, premature birth (defined as birth strated.5,8,10 Most positive11-15 and negative16‑19 before 36 weeks of gestation), immunosuppres- studies of corticosteroids have been small and sion or immunodeficiency, treatment with corti- heterogeneous in design, but a controlled trial costeroids in the previous 14 days, active varicella involving 70 infants with moderate-to-severe bron- or recent exposure to varicella, or inability of the chiolitis was reported by Schuh and colleagues in parent or guardian to speak English or Spanish. 2002.20 They found significant reductions in re- Critically ill infants were also excluded. spiratory scores after 4 hours of observation in Infants were screened for eligibility during infants who received 1 mg of oral dexamethasone times when a research assistant and study clini- per kilogram of body weight, as compared with cian (emergency department faculty, fellow, or those who received placebo. Moreover, the admis- nurse practitioner) were available. Each center sion rate was 19% in the dexamethasone group, kept a record of all screened infants, including as compared with 44% in the placebo group. those who arrived when study staff were available A number of experts and reviews5,8,10,21 have and who underwent screening but were not en- called for further study of corticosteroids for rolled. Research assistants and all study clinicians bronchiolitis. The 2003 report published by the received yearly training from site lead investiga- Agency for Healthcare Research and Quality tors in study procedures and respiratory scoring. (AHRQ)5 stated that there is “no evidence that any Site monitors visited each site during and after single agent can be recommended for treatment data collection to audit all study records. of bronchiolitis,” and it called for a “rigorously Before enrollment, study clinicians confirmed designed and adequately sized trial” of agents clinical bronchiolitis and determined the duration to include dexamethasone. The goal of our study of symptoms and the RDAI score. Research assis was to determine the effectiveness of a single tants or clinicians obtained the medical history dose of oral dexamethasone in infants with mod- from parents or guardians on a standardized erate-to-severe bronchiolitis. data-collection form, which included questions about a history of eczema in the patient, a family Methods history of asthma in the immediate family, and the presence of smokers or pets at home. Infants Patients with eczema or a family history of asthma were We conducted the study in 20 emergency depart- considered to have possible atopy and, as recom- ments of the Pediatric Emergency Care Applied mended in the AHRQ report,5 were treated as a Research Network (PECARN)22 during bronchio prespecified subgroup in the analysis. At enroll- litis season (November through April) from Janu- ment and 1 hour and 4 hours after administra- ary 2, 2004, through April 30, 2006. Planned start tion of the study medication, a nurse recorded and end dates were the same for all centers. The clinical variables (respiratory and heart rates, institutional review boards at all sites approved temperature, and oxygen saturation while the the study. Written informed consent was obtained infant was breathing ambient air). A study clini-

332 n engl j med 357;4 www.nejm.org july 26, 2007 Trial of Dexamethasone for Bronchiolitis

Table 1. Wheezing and Retraction Scales for the Respiratory Distress Assessment Instrument (RDAI).*

Symptom Points 0 1 2 3 4 Maximum Wheezing During expiration None End First half First three Throughout 4 quarters During inspiration None Part Throughout — — 2 No. of involved lung fields 0 1 or 2 3 or 4 — — 2 Retractions Supraclavicular None Mild Moderate Marked — 3 Intercostal None Mild Moderate Marked — 3 Subcostal None Mild Moderate Marked — 3 Total 17

* Both wheezing and retractions were scored. The total score on the RDAI is the sum of the scores for each row, with a range of 0 to 17; higher scores indicate more severe disease. cian repeated respiratory scoring 1 hour and according to the clinician’s preference and local 4 hours after the administration of the study standards. Any diagnostic testing, including viral medication and assessed each child for discharge testing, was also left to the clinician’s discretion, or admission after 4 hours. and tests were performed with assays available at the participating center. Because testing for Randomization other viruses varied, only the results for respira- We performed computerized randomization by tory syncytial virus were recorded. telephone, using the keypad for data entry. Infants After 7 to 10 days, a research assistant, who were assigned in equal numbers to the dexameth- was unaware of the group assignments, reviewed asone and placebo groups with the use of random the chart and conducted a brief standardized permuted blocks stratified by center. All emer- telephone interview with the parent or guardian. gency department staff, study personnel, and par- The interview included questions about whether ents and guardians were unaware of the group hospitalization, unscheduled medical visits, or ad assignments. Randomization codes were secured verse reactions to the study drug (as judged by a until all data entry was complete. physician or the parent or guardian) had occurred within 7 days after the initial emergency depart- Study Intervention ment visit. Using the same formulation as in prior studies,20,24 research pharmacies prepared oral dexa- Outcome Measures methasone solutions (1 mg per milliliter of liquid) The primary outcome was the decision to hospi- from generic dexamethasone phosphate injection talize or discharge the infant 4 hours after the solution and identical oral placebo solutions. The administration of the study medication. Infants preparations were packaged in identical clear requiring admission to an intensive care unit be- plastic vials labeled only with the randomization fore 4 hours of observation had been completed numbers. A nurse orally administered 1 ml of were included in the analysis of admissions. The solution per kilogram, providing 1 mg of dexa- secondary outcome was the Respiratory Assess- methasone per kilogram in the dexamethasone ment Change Score (RACS) at 4 hours.23 The RACS group (maximum, 12 mg). Any episode of vom- is calculated as the sum of the change in the RDAI iting within 20 minutes after administration of score and a standardized score for the change in the study medication was recorded, but the dose the respiratory rate, with a reduction of 1 unit for was not repeated. a decrease of 5 to 15%, 2 units for a decrease of For ethical, practical, and scientific reasons, 16 to 25%, and so on.20,23 Thus, negative RACS all other bronchiolitis treatments were provided values signify improvement. Other investigators

n engl j med 357;4 www.nejm.org july 26, 2007 333 T h e new england journal o f medicine

absolute reduction in hospital admission rates of 12% or more in the dexamethasone group. A bio- 8686 Infants were assessed for eligibility statistician participated in the study design and performed all analyses. The primary analysis was based on the intention-to-treat principle, with all 8086 Were not enrolled 7352 Did not meet inclusion patients included in their assigned group. A sec- criteria ondary, per-protocol analysis examined the results 707 Were not enrolled because consent among infants who actually received the assigned was not provided study medication. 27 Were not enrolled for other reasons Hospital admission rates were compared with the use of Pearson’s chi-square test. The RACS was compared in the two groups by means of a 600 Were enrolled and underwent two-sample t-test. Adjusted measures and sub- randomization group effects for admission and RACS were analyzed with the use of logistic regression and linear regression, respectively. Generalized estimating equations and linear mixed models were used to test for an interaction between treatment group 305 Were assigned to dexamethasone 295 Were assigned to placebo 299 Received dexamethasone 293 Received placebo and site in the admission and RACS outcomes, respectively. Changes in clinical variables after 4 hours of observation were regressed against baseline values and treatment group as predictors. Hospitalization data available for 305 Hospitalization data available for 295 Length-of-stay measures were compared by means RACS available for 283 RACS available for 271 Follow-up data available for 284 Follow-up data available for 265 of the two-sample Wilcoxon test. The alpha level was set at 0.05 for all analyses, 95% confidence Figure 1. Eligibility, Randomization, and Follow-up. intervals were calculated, and all comparisons Among the 7352 infants who did not meet the criteria for inclusion in the were two-tailed. study, two thirds had either prior wheezing or mild disease; the remainder met other exclusion criteria. For the primary outcome, hospital admission, Results data were available for all 600 infants in the intention-to-treat analysis. The secondary outcome, the Respiratory Assessment Change Score (RACS), in RETAKE 1st ICM AUTHOR: Corneli A total of 8686 infants were screened for study cluded two variables, respiratory rate and Respiratory Distress2nd Assessment REG F FIGURE: 1 of 2 eligibility (Fig. 1). Among the 7352 infants who Instrument (RDAI) score, that were compared at baseline and 3rdafter 4 hours of observation.CASE Follow-up involved a single telephone Revisedinterview with each did not meet the inclusion criteria, two thirds EMail Line 4-C SIZE infant’s parent or guardian.ARTIST: ts had either prior wheezing (41%) or an RDAI score H/T H/T 22p3 Enon Combo of less than 6 (25%). Of the 600 infants who AUTHOR, PLEASE NOTE: underwent randomization, 305 were assigned to Figure has been redrawn and type has been reset. have interpretedPlease check a carefully.change of 2 units or more as the dexamethasone group and 295 to the placebo clinically important.20 group; all were included in the intention-to-treat JOB: 35704 ISSUE: 07-26-07 analysis. Two randomly assigned infants were Adverse Events hospitalized before administration of the study Study clinicians and research assistants monitored drug, leaving 598 treated infants. Five of these the infants for adverse events during observation infants received the wrong medication because in the emergency department. Subsequent adverse of errors in vial selection, and 1 received an in- events were determined at follow-up. A patient- sufficient dose of dexamethasone, leaving 592 pa- safety committee, made up of people not involved tients in the per-protocol analysis. The results of with patient enrollment, tracked all adverse events. this analysis did not differ qualitatively from those of the intention-to-treat analysis. A detailed Statistical Analysis subanalysis according to the results of tests for Assuming a 40% admission rate in the placebo respiratory syncytial virus revealed no significant group, we calculated the sample size that would differences in any of the studied outcomes be- be required to provide more than 80% power tween infants with positive results and those with (with a two-sided alpha level of 0.05) to detect an negative results (Fig. 2).

334 n engl j med 357;4 www.nejm.org july 26, 2007 Trial of Dexamethasone for Bronchiolitis

Baseline demographic and clinical characteristics were similar in the dexamethasone and Risk of Hospitalization placebo groups (Table 2). Similar proportions of Overall infants received inhaled bronchodilator treat- Atopy ment, either with albuterol (77.0% and 80.3%, History of eczema or family history of asthma respectively) or epinephrine (15.5% and 16.7%, No history of eczema or family respectively). The number of such treatments re- history of asthma ceived was also similar, with a mean of 2.0 treat- RSV ments with albuterol and 1.2 treatments with Positive epinephrine in each study group. Negative Not tested Age Hospital Admission ≤6 mo We found no significant difference between the >6 mo study groups with respect to hospitalization. Of 0.5 1.0 1.5 2.0 the 305 infants in the dexamethasone group, 121 Dexamethasone Placebo (39.7%) were admitted, as compared with 121 of Better Better the 295 infants (41.0%) in the placebo group (absolute difference, −1.3%; 95% confidence interval Figure 2. Risk Ratios for Hospital Admission. [CI], −9.2 to 6.5; P = 0.74). Neither was there a Estimated risk ratios (the risk of hospitalization in the significant difference when admission was ana- dexamethasone group as compared with that in the placebo group) are shown for the overall study groups, lyzed in the prespecified subgroup with eczema or as well as for specific subgroups evaluated in bivariate a family history of asthma (absolute difference, analysis. The horizontal lines represent the 95% confi −1.3%; 95% CI, −11.1 to 8.5). Figure 2 shows the dence intervals. Risk ratios of less than 1.0 favor the use RETAKE 1st of dexamethasone;ICM AUTHOR: theCorneli value 1.0 represents equivalence relative risk of admission overall and in this and 2nd betweenREG F theFIGURE: dexamethasone2 of 2 group and the placebo other subgroups. 3rd group.CASE Testing for the respiratory syncytial virusRevised (RSV) wasEMail performed in 269 infants;Line 325 did4-C not undergoSIZE ARTIST: ts Respiratory Assessment Change Score testing, and in the case of 6 H/Tinfants, itH/T is not known16p6 Enon Combo The respiratory status of both study groups im- whether testing was performed. AUTHOR, PLEASE NOTE: proved during treatment and observation in the Figure has been redrawn and type has been reset. emergency department, but mean RACS values Please check carefully. did not differ significantly between the groups The results for admission and RACS outcomes (Table 3). Neither was there a significant differ- in JOB:multivariate35704 regression models ISSUE:were virtually07-26-07 ence when RACS values were analyzed in the sub- identical to those in the unadjusted models (data group with eczema or a family history of asthma not shown). The site had no significant effect on (absolute difference, −0.4; 95% CI, −1.3 to 0.6). treatment outcomes.

Other Outcomes Adverse Events Table 3 also shows differences in the clinical There were few adverse events. Vomiting within variables between the baseline and 4-hour obser- 20 minutes after administration of the study med- vations. Changes in the respiratory rate did not ication occurred in 5.5% of the dexamethasone differ significantly between the two study groups. group and 4.7% of the placebo group. No infant Although changes in the RDAI score, oxygen satu- had gastrointestinal bleeding, hypertension, or ration, temperature, and heart rate did differ sig- complicated varicella. Pneumonia was diagnosed nificantly between the groups, these differences in three infants; two were in the placebo group, were small. and an empyema developed in one of these two The mean length of stay for hospitalized pa- infants. tients was 2.55 days in the dexamethasone group and 2.27 days in the placebo group (P = 0.10). Discussion Subsequent hospital admissions in the 7 days after the intervention were reported for 12 of 284 Our multicenter, randomized, double-blind study children in the dexamethasone group (4.2%) and of 600 infants with acute, moderate-to-severe 10 of 265 children in the placebo group (3.8%). bronchiolitis in the emergency department showed

n engl j med 357;4 www.nejm.org july 26, 2007 335 T h e new england journal o f medicine

our observed limit of uncertainty (a 9.2% reduc- Table 2. Baseline Characteristics of the Infants According to the Assigned Study Group.* tion in admissions), 11 patients would need to be treated with dexamethasone to prevent one ad- Dexamethasone Placebo mission. Characteristic (N = 304) (N = 294) Because the response to corticosteroids might Male sex — no./total no. (%) 190/304 (62.5) 178/294 (60.5) differ in children with possible atopy, our pre- Age — mo 5.1±2.6 5.1±2.8 specified subgroup analysis examined outcomes RDAI score 9.0±2.1 9.2±2.4 according to whether the infants had eczema or a Respiratory rate — breaths/min 53±13 53±13 family history of asthma. Because bronchiolitis is Heart rate — beats/min 157±20 158±21 a clinical syndrome caused by several viruses, we also examined outcomes according to whether Temperature — °C 37.6±0.8 37.7±0.8 the infants were positive or negative for respira- Oxygen saturation — % 96±4 96±4 tory syncytial virus. No significant differences No. of days of illness 3.7±2.5 3.6±2.5 between the study groups were found in these RSV-positive — no. positive/ 85/127 (66.9) 81/142 (57.0) analyses, which suggests that none of these fac- no. tested (%) tors identify a subgroup of infants who have a History — no./total no. (%) response to dexamethasone. Family history of asthma 165/295 (55.9) 170/284 (59.9) Small differences between groups (as shown History of eczema 76/292 (26.0) 77/281 (27.4) in Table 3) may be statistically significant when Either family history of asthma or 187/295 (63.4) 196/284 (69.0) large samples are compared. An antipyretic effect history of eczema of dexamethasone may account for the small Smoker in home — no./total no. (%) 117/300 (39.0) 103/287 (35.9) differences observed in temperature and heart 21,25 Pet in home — no./total no. (%) 97/299 (32.4) 90/282 (31.9) rate. Although these two variables are not ty pically central to the assessment of bronchiolitis, * The data do not include two patients who were hospitalized before the adminis oxygen saturation is considered central to this tration of study medication. Plus–minus values are means ±SD. RDAI denotes assessment. For this variable, however, the small Respiratory Distress Assessment Instrument, and RSV respiratory syncytial virus. difference actually favored placebo. The small difference in RDAI scores between the study groups no significant reduction in hospital admissions or became nonsignificant when incorporated into improvement in respiratory status after 4 hours the overall RACS. of observation when infants given a single oral Our findings are consistent with studies that dose of 1 mg of dexamethasone per kilogram were failed to demonstrate the efficacy of corticoste- compared with infants given placebo. This was roids in bronchiolitis.8,16-19 A collective review by true whether or not the infants had markers of the Cochrane Collaboration of 13 studies of the possible atopy. Furthermore, among the infants use of corticosteroids for bronchiolitis10 showed who received dexamethasone, as compared with no significant differences between corticosteroid those who received placebo, there was no reduc- and placebo groups in respiratory rates, oxygen tion in the duration of hospitalization for those saturation, initial admission rates, length of stay, who were initially admitted, and there were no subsequent visits, or readmission rates.10 The reductions in later, unscheduled admissions or AHRQ report5 analyzed five placebo-controlled visits to an emergency department or physician studies of oral corticosteroids, including dexa- for those who were initially discharged. methasone, and two placebo-controlled studies Several studies have already noted the use of of parenteral dexamethasone. Only one study20 corticosteroids in bronchiolitis despite the absence showed a significant difference between groups. of definitive evidence of any benefit.3,4,9 The au- In 2006, a subcommittee of the American Acad- thors of the AHRQ report5 thought this use of emy of Pediatrics reviewed the evidence from pre corticosteroids would persist “unless a large sim vious studies and recommended that corticoste- ple trial of the most common interventions is roids not be used routinely for bronchiolitis.8 All mounted.” Following their recommendations, we of these reviews, however, note the inconclusive designed a study with sufficient power to evaluate nature of the available evidence. hospitalization rates, an “important outcome to Some studies11-14,26 have suggested a benefit of parents, clinicians, and health systems.”5 Even at corticosteroid therapy, and the size, methods, and

336 n engl j med 357;4 www.nejm.org july 26, 2007 Trial of Dexamethasone for Bronchiolitis

Table 3. Hospital Admission and Changes in Clinical Variables from Baseline to 4 Hours after Intervention.*

Dexamethasone Placebo Difference between Variable Group Group Groups (95% CI) P Value Hospital admission (%) 39.7 41.0 −1.3 (−9.2 to 6.5) 0.74 RACS −5.3±4.7 −4.8±4.6 −0.5 (−1.3 to 0.3) 0.21 RDAI score −4.4±3.1 −3.9±3.2 −0.5 (−1.0 to −0.1) 0.03 Respiratory rate (breaths/min) −8±15 −7±14 −1.0 (−3.0 to 1.0) 0.39 Oxygen saturation (%) 0.3±3.3 0.9±3.2 −0.6 (−1.0 to −0.1) 0.02 Heart rate (beats/min) −13±24 −5±25 −8.0 (−12.0 to −5.0) <0.001 Temperature (°C) −0.6±0.9 −0.2±1.0 −0.4 (−0.6 to −0.3) <0.001

* Data for all variables except hospital admission are expressed as the change from baseline to 4 hours. RACS denotes Respiratory Assessment Change Score, and RDAI Respiratory Distress Assessment Instrument. findings of these studies have been carefully re- Our study had some limitations. For both ethi- viewed.5,10 A meta-analysis27 of six trials of sys- cal and scientific reasons, we sought to exclude temic corticosteroids in infants with bronchiolitis children with possible early asthma, who might showed a small benefit in the corticosteroid have benefited from dexamethasone. We there- groups, but this effect did not persist when out- fore studied only young infants with first-time comes were analyzed separately or when only wheezing. Older children or children with recur- studies of first-time wheezing were examined. rent wheezing might have a different response to The results of our study differ from those of dexamethasone. the study by Schuh et al.20 The two studies used We studied a single oral dose of dexametha- the same dose of dexamethasone (1 mg per kilo- sone (1 mg per kilogram). The size of the dose gram), with 4 hours of observation in the emer- makes it unlikely that more medication would be gency department and outcomes of hospitaliza- effective. Oral administration and a 4-hour observa tion and RACS, but they differ in certain respects. tion period were chosen to replicate the methods In the study by Schuh et al., oral dexamethasone used by Schuh et al.20 Although the biologic basis (0.6 mg per kilogram) or placebo was continued of the effect is not clear, their study and ours were for 5 days in patients discharged home. This predicated on extensive evidence that oral cortico- would not, however, affect the study’s main out- steroids are effective within 4 hours in patients comes — hospitalization and RACS after 4 hours. with asthma28‑33 and those with croup.34‑36 We studied infants in the first year of life, where- It is unlikely that we missed a later benefit of as Schuh et al. included children up to 24 months dexamethasone. We collected data on later out- of age. By chance, the dexamethasone group in comes, including the length of the hospital stay their study had a significantly higher proportion among infants who were initially admitted, sub- of infants with family histories of atopy than did sequent admissions or unscheduled medical vis- the placebo group. Our study groups were bal- its, and adverse events in the two study groups. anced in this regard. Their study was substantial If dexamethasone had been effective after the pas- ly smaller and conducted at a single institution, sage of 4 hours, this result should have been ap- where all infants were treated with a standard- parent in one or more of these later outcomes. ized bronchodilator regimen. Given the current In summary, in our multicenter study of 600 wide variation in the use of bronchodilators3,4 infants from 2 to 12 months of age who had and uncertainty regarding their effectiveness,5,10 moderate-to-severe bronchiolitis, we found that we did not try to control bronchodilator use. Our treatment with 1 mg of oral dexamethasone per study was not powered to examine possible inter- kilogram did not significantly alter the rate of actions between bronchodilators and dexameth- hospital admission or the respiratory status after asone. We did, however, confirm that the types 4 hours of observation. Neither did such treatment and numbers of bronchodilator treatments were affect the length of the hospital stay among in- similar in the two groups. fants who were initially admitted, subsequent ad

n engl j med 357;4 www.nejm.org july 26, 2007 337 T h e new england journal o f medicine

missions or unscheduled medical visits, or adverse Dr. Holubkov, consulting fees from St. Jude Medical, Tyco Health Care, and Ucyclyd Pharma and grant support from St. Jude events. We recommend evaluation of other treat- Medical and GlaxoSmithKline. No other potential conflict of ments and preventive strategies for bronchiolitis. interest relevant to this article was reported. Supported by a grant (R40MC042980100) from the Maternal We thank Kammy K. Jacobsen and Laurie F. Larsen of the and Child Health Research program and by cooperative agree- University of Utah; Sally Jo Zuspan, Kym Call, Li Dong, Rene ments (U03MC00001, U03MC00003, U03MC00006, U03MC00007, Enriquez, Stacey Knight, and Rita Gerard of the Central Data and U03MC00008) with the Emergency Medical Services for Management and Coordinating Center, Salt Lake City; Helena Children program of the Maternal and Child Health Bureau, Rincón, Emily Kim, and the PECARN research assistants and Health Resources and Services Administration. nodal administrators; and all the study clinicians, without whose Dr. Zorc reports receiving grant support from Sepracor; and help this project could not have been completed.

APPENDIX In addition to the authors, the following investigators participated in the PECARN bronchiolitis study: Washington University and St. Louis Children’s Hospital, St. Louis — D. Jaffe; Children’s National Medical Center and George Washington University, Washington, DC — S. Teach; Devos Children’s Hospital and Michigan State University, Grand Rapids — J. Hoyle, Jr.; Bellevue Hospital Center and New York University, New York — M. Tunik; University of Maryland, Baltimore — R. Lichenstein; SUNY Upstate Medical University, Syracuse, NY — J. Callahan; University of Michigan, Ann Arbor – D. Treloar; St. Barnabas Health Care System, Livingston, NJ — N. Schamban. Patient Safety Committee: N. Kuppermann, Kathleen Mahackian, M. Gorelick, W. Schalick III, R. Woods, J.P. Joad. For PECARN: Steering Committee: N. Kuppermann (chair), E. Alpern, J. Chamberlain, J.M. Dean, M. Gerardi, J. Goepp, M. Gorelick, J. Hoyle, D. Jaffe, C. Johns, N. Levick, P. Mahajan, R. Maio, K. Melville, S. Miller (deceased), D. Monroe, R. Ruddy, R. Stanley, D. Treloar, M. Tunik, A. Walker. Maternal and Child Health Bureau liaisons: D. Kavanaugh, H. Park. Central Data Management and Coordinating Center: M. Dean, R. Holubkov, S. Knight, A. Donaldson. Data Analysis and Management Subcommittee: J. Chamberlain (chair). M. Brown, H. Corneli, J. Goepp, R. Holubkov, P. Mahajan, K. Melville, E. Stremski, M. Tunik. Grants and Publications Subcommittee: M. Gorelick (chair), E. Alpern, J.M. Dean, G. Foltin, J. Joseph, S. Miller (deceased), F. Moler, R. Stanley, S. Teach. Protocol Concept Review and Development Subcommittee: D. Jaffe (chair), K. Brown, A. Cooper, J.M. Dean, C. Johns, R. Maio, N.C. Mann, D. Monroe, K. Shaw, D. Teitelbaum, D. Treloar. Quality Assurance Subcommittee: R. Stanley (chair), D. Alexander, J. Brown, M. Gerardi, M. Gregor, R. Holubkov, K. Lillis, B. Nordberg, R. Ruddy, M. Shults, A. Walker. Safety and Regulatory Affairs Subcommittee: N. Levick (chair), J. Brennan, J. Brown, J.M. Dean, J. Hoyle, R. Maio, R. Ruddy, W. Schalick, T. Singh, J. Wright.

References 1. Leader S, Kohlhase K. Recent trends ment of bronchiolitis. Pediatrics 2006;118: Gorichovsky Y, Bibi H. Dexamethasone in severe respiratory syncytial virus (RSV) 1774-93. inhalations in RSV bronchiolitis: a double- among US infants, 1997 to 2000. J Pediatr 9. Behrendt CE, Decker MD, Burch DJ, blind, placebo-controlled study. Acta Pae- 2003;143:Suppl 5:S127-S132. Watson PH. International variation in the diatr 2005;94:866-71. 2. Shay DK, Holman RC, Newman RD, management of infants hospitalized with 16. Berger I, Argaman Z, Schwartz SB, et Liu LL, Stout JW, Anderson LJ. Bronchi- respiratory syncytial virus. Eur J Pediatr al. Efficacy of corticosteroids in acute bron- olitis-associated hospitalizations among 1998;157:215-20. chiolitis: short-term and long-term follow- US children, 1980-1996. JAMA 1999;282: 10. Patel H, Platt R, Lozano JM, Wang EE. up. Pediatr Pulmonol 1998;26:162-6. 1440-6. Glucocorticoids for acute viral bronchiolitis 17. De Boeck K, Van der Aa N, Van Lierde 3. Christakis DA, Cowan CA, Garrison in infants and young children. Cochrane S, Corbeel L, Eeckels R. Respiratory syn- MM, Molteni R, Marcuse E, Zerr DM. Database Syst Rev 2004;3:CD004878. cytial virus bronchiolitis: a double-blind Variation in inpatient diagnostic testing 11. van Woensel JB, Wolfs TF, van Aal- dexamethasone efficacy study. J Pediatr and management of bronchiolitis. Pediat- deren WM, Brand PL, Kimpen JL. Ran- 1997;131:919-21. rics 2005;115:878-84. domised double blind placebo controlled 18. Klassen TP, Sutcliffe T, Watters LK, 4. Willson DF, Horn SD, Hendley JO, trial of prednisolone in children admitted Wells GA, Allen UD, Li MM. Dexametha- Smout R, Gassaway J. Effect of practice to hospital with respiratory syncytial virus sone in salbutamol-treated inpatients with variation on resource utilization in infants bronchiolitis. Thorax 1997;52:634-7. acute bronchiolitis: a randomized, con- hospitalized for viral lower respiratory ill- 12. Goebel J, Estrada B, Quinonez J, Nagji trolled trial. J Pediatr 1997;130:191-6. ness. Pediatrics 2001;108:851-5. N, Sanford D, Boerth RC. Prednisolone 19. Roosevelt G, Sheehan K, Grupp-Phelan 5. Management of bronchiolitis in infants plus albuterol versus albuterol alone in J, Tanz RR, Listernick R. Dexamethasone and children. Rockville, MD: Agency for mild to moderate bronchiolitis. Clin Pedi- in bronchiolitis: a randomised controlled Healthcare Research and Quality, January atr (Phila) 2000;39:213-20. trial. Lancet 1996;348:292-5. 2003. (AHRQ publication no. 03-E014.) 13. Csonka P, Kaila M, Laippala P, Iso- 20. Schuh S, Coates AL, Binnie R, et al. 6. Kellner JD, Ohlsson A, Gadomski AM, Mustajarvi M, Vesikari T, Ashorn P. Oral Efficacy of oral dexamethasone in outpa- Wang EE. Bronchodilators for bronchioli prednisolone in the acute management of tients with acute bronchiolitis. J Pediatr tis. Cochrane Database Syst Rev 2000;2: children age 6 to 35 months with viral re- 2002;140:27-32. CD001266. spiratory infection-induced lower airway 21. McBride JT. Dexamethasone and bron- 7. Wainwright C, Altamirano L, Cheney disease: a randomized, placebo-controlled chiolitis: a new look at an old therapy? M, et al. A multicenter, randomized, dou- trial. J Pediatr 2003;143:725-30. J Pediatr 2002;140:8-9. ble-blind, controlled trial of nebulized 14. Kuyucu S, Unal S, Kuyucu N, Yilgor E. 22. The Pediatric Emergency Care Applied epinephrine in infants with acute bronchi- Additive effects of dexamethasone in neb- Research Network (PECARN): rationale, olitis. N Engl J Med 2003;349:27-35. ulized salbutamol or L-epinephrine treat- development, and first steps. Pediatr Emerg 8. American Academy of Pediatrics Sub- ed infants with acute bronchiolitis. Pedi- Care 2003;19:185-93. committee on Diagnosis and Management atr Int 2004;46:539-44. 23. Lowell DI, Lister G, Von Koss H, Mc- of Bronchiolitis. Diagnosis and manage- 15. Bentur L, Shoseyov D, Feigenbaum D, Carthy P. Wheezing in infants: the re-

338 n engl j med 357;4 www.nejm.org july 26, 2007 Trial of Dexamethasone for Bronchiolitis

sponse to epinephrine. Pediatrics 1987;79: 29. Scarfone RJ, Fuchs SM, Nager AL, 33. Rowe BH, Spooner C, Ducharme FM, 939-45. Shane SA. Controlled trial of oral predni- Bretzlaff JA, Bota GW. Early emergency 24. Chou JW, Decarie D, Dumont RJ, En- sone in the emergency department treat- department treatment of acute asthma som MHH. Stability of dexamethasone in ment of children with acute asthma. Pedi- with systemic corticosteroids. Cochrane extemporaneously prepared oral suspen- atrics 1993;92:513-8. Database Syst Rev 2001;1:CD002178. sions. Can J Hosp Pharm 2001;54:96-101. 30. Scarfone RJ, Loiselle JM, Wiley JF II, 34. Johnson DW, Jacobson S, Edney PC, 25. Hanna CM, Greenes DS. How much Decker JM, Henretig FM, Joffe MD. Nebu- Hadfield P, Mundy ME, Schuh S. A com- tachycardia in infants can be attributed to lized dexamethasone versus oral predni- parison of nebulized budesonide, intra- fever? Ann Emerg Med 2004;43:699-705. sone in the emergency treatment of asth- muscular dexamethasone, and placebo for 26. Bentur L, Canny GJ, Shields MD, et al. matic children. Ann Emerg Med 1995;26: moderately severe croup. N Engl J Med Controlled trial of nebulized albuterol in 480-6. 1998;339:498-503. children younger than 2 years of age with 31. Rodrigo G, Rodrigo C. Inhaled fluni- 35. Klassen TP, Craig WR, Moher D, et al. acute asthma. Pediatrics 1992;89:133-7. solide for acute severe asthma. Am J Respir Nebulized budesonide and oral dexameth- 27. Garrison MM, Christakis DA, Harvey Crit Care Med 1998;157:698-703. asone for treatment of croup: a random- E, Cummings P, Davis RL. Systemic corti- 32. Lin RY, Pesola GR, Bakalchuk L, et al. ized controlled trial. JAMA 1998;279:1629- costeroids in infant bronchiolitis: a meta- Rapid improvement of peak flow in asth- 32. analysis. Pediatrics 2000;105:E44. matic patients treated with parenteral 36. Russell K, Wiebe N, Saenz A, et al. 28. Ellul-Micallef R. The acute effects of methylprednisolone in the emergency de- Glucocorticoids for croup. Cochrane Da- corticosteroids in bronchial asthma. Eur J partment: a randomized controlled study. tabase Syst Rev 2004;1:CD001955. Respir Dis Suppl 1982;122:118-25. Ann Emerg Med 1999;33:487-94. Copyright © 2007 Massachusetts Medical Society.

clinical trial registration The Journal requires investigators to register their clinical trials in a public trials registry. The members of the International Committee of Medical Journal Editors (ICMJE) will consider most clinical trials for publication only if they have been registered (see N Engl J Med 2004;351:1250-1). Current information on requirements and appropriate registries is available at www.icmje.org/faq.pdf.

n engl j med 357;4 www.nejm.org july 26, 2007 339 T h e new england journal o f medicine

original article

High-Dose Chemotherapy and Stem-Cell Rescue for Metastatic Germ-Cell Tumors

Lawrence H. Einhorn, M.D., Stephen D. Williams, M.D., Amy Chamness, B.A., Mary J. Brames, R.N., Susan M. Perkins, Ph.D., and Rafat Abonour, M.D.

Abstr act

Background From the Division of Hematology–Oncol Metastatic testicular tumors that have not been successfully treated by means of ogy, Indiana University School of Medi initial chemotherapy are potentially curable with salvage chemotherapy. cine (L.H.E., S.D.W., A.C., M.J.B., R.A.); the Walther Cancer Institute (L.H.E., S.D.W., A.C., M.J.B., R.A.); and the Division of Bio Methods statistics, Indiana University (S.M.P.) — all We conducted a retrospective review of 184 consecutive patients with metastatic in Indianapolis. Address reprint requests to Dr. Einhorn at the Indiana University Can testicular cancer that had progressed after they received cisplatin-containing com- cer Center, 535 Barnhill Dr., Rm. 473, Indi bination chemotherapy. We gave 173 patients two consecutive courses of high-dose anapolis, IN 46202-5289, or at leinhorn@ chemotherapy consisting of 700 mg of carboplatin per square meter of body-sur- iupui.edu. face area and 750 mg of etoposide per square meter, each for 3 consecutive days, N Engl J Med 2007;357:340-8. and each followed by an infusion of autologous peripheral-blood hematopoietic Copyright © 2007 Massachusetts Medical Society. stem cells; the other 11 patients received a single course of this treatment. In 110 patients, cytoreduction with one or two courses of vinblastine plus ifosfamide plus cisplatin preceded the high-dose chemotherapy.

Results Of the 184 patients, 116 had complete remission of disease without relapse during a median follow-up of 48 months (range, 14 to 118). Of the 135 patients who received the treatment as second-line therapy, 94 were disease-free during follow-up; 22 of 49 patients who received treatment as third-line or later therapy were disease- free. Of 40 patients with cancer that was refractory to standard-dose platinum, 18 were disease-free. A total of 98 of 144 patients who had platinum-sensitive disease were disease-free, and 26 of 35 patients with seminoma and 90 of 149 patients with nonseminomatous germ-cell tumors were disease-free. Among the 184 patients, there were three drug-related deaths during therapy. Acute leukemia developed in three additional patients after therapy.

Conclusions Testicular tumors are potentially curable by means of high-dose chemotherapy plus hematopoietic stem-cell rescue, even when this regimen is used as third-line or later therapy or in patients with platinum-refractory disease.

340 n engl j med 357;4 www.nejm.org july 26, 2007 Salvage Therapy for Metastatic Germ-Cell Tumors

erm-cell tumors are curable even described previously.8 We harvested the cells after in the presence of metastatic disease.1-3 stimulation of the patient’s marrow with granulo GAn international collaboration has estab- cyte colony-stimulating factor and enriched these lished that metastatic germ-cell tumors can be cells for CD34+ hematopoietic cells. Patients who classified into good-, intermediate-, and poor-risk received high-dose chemotherapy as initial sal- disease, with corresponding cure rates of 90 to vage treatment and whose tumor had not pro- 95%, 75%, and 40 to 50%, respectively.4 Herein- gressed within 4 weeks after the last round of after, we refer to these categories as low-, interme- this treatment received standard doses of vin- diate-, and high-risk disease, respectively. Patients blastine plus ifosfamide plus cisplatin6 to reduce with tumors that relapse or with tumors that prog- the bulk of the tumor and prevent progression ress despite initial chemotherapy are candidates before we administered high-dose chemotherapy. for salvage therapy. The few patients with anatom- Patients who had already received ifosfamide- ically confined disease are amenable to surgical based salvage chemotherapy were offered high- extirpation.5 For most patients, however, the op- dose chemotherapy without any further treatment. tions include salvage chemotherapy with cisplatin Patients with primary mediastinal nonseminoma- plus ifosfamide plus vinblastine6 or paclitaxel7 for tous germ-cell tumors or tumors with late relapse four courses, or high-dose chemotherapy with (>2 years after previous therapy) were not offered autologous hematopoietic stem-cell transplanta- high-dose chemotherapy during the specified tion to rescue the bone marrow from the myeloab- period. lative effects of chemotherapy.8-10 We began treat- All patients received therapy in the outpatient ing patients with carboplatin-based high-dose clinic except for patients with complications re- chemotherapy in 1986.11 Earlier studies of such quiring inpatient care. Antibiotics, including pro- treatment based on high-dose cyclophosphamide phylactic acyclovir, fluconazole, ciprofloxacin, were unsuccessful.12 and either penicillin or vancomycin, were used Initially, we used autologous bone marrow cells routinely. to rescue the hematopoietic system after high- High-dose chemotherapy consisted of two cy- dose chemotherapy. In February 1996, we shifted cles of 700 mg of carboplatin per square meter of to peripheral-blood stem cells, which rapidly en- body-surface area plus 750 mg of etoposide per grafted, thereby permitting a second course of square meter, both given intravenously 5, 4, and high-dose chemotherapy with fewer delays. In this 3 days before the infusion of peripheral-blood article, we present a retrospective study of 184 stem cells. A minimum of 1 million CD34+ cells consecutive patients treated at Indiana University per kilogram of body weight was required for with high-dose carboplatin plus etoposide (high- each cycle of high-dose chemotherapy. There were dose chemotherapy) and rescue of the hemato- no planned reductions or escalations in the doses poietic system by infusion of peripheral-blood of chemotherapy. The second cycle of high-dose stem cells. Our goal was to examine the efficacy chemotherapy was given after recovery of granu- of this treatment for cisplatin-resistant, progres locyte and platelet counts, unless there was a sively growing testicular cancer. grade 4 nonhematologic toxic effect or no response to the first course. Most patients who had Methods a complete or partial remission after the two cycles of high-dose chemotherapy and who had Patients normal serum levels of human chorionic gonado- We conducted a review of 184 consecutive patients tropin (hCG) and alpha-fetoprotein received a who received high-dose chemotherapy and periph- maintenance oral dose of 50 mg of etoposide per eral-blood stem-cell rescue between February 1996 square meter daily for 21 consecutive days every and December 2004. Patient interviews and med- 4 weeks for three cycles.13 ical charts were used to obtain the data analyzed Complete remission was defined as no clinical in this study. The institutional review board at ly or radiographically detectable disease and nor- Indiana University approved the study. The re- mal serum levels of hCG and alpha-fetoprotein. quirement for informed consent was waived. Patients with a resectable residual mass after high- Before high-dose chemotherapy was initiated, dose chemotherapy were offered surgery. Disease- peripheral-blood stem cells were collected and free survival, defined as no evidence of disease at purified according to a technique that has been all follow-up visits after high-dose chemotherapy,

n engl j med 357;4 www.nejm.org july 26, 2007 341 T h e new england journal o f medicine

was the primary end point for this analysis. Sur- The Kaplan–Meier method14 was used to calculate vival time was measured from the first day of overall and disease-free survival according to the high-dose chemotherapy. Disease-free survival risk group (low, intermediate, or high), which was measured from the initiation of high-dose was defined on the basis of the prognostic scor- chemotherapy to tumor progression or death. ing algorithm described below. The association of disease-free survival with each prognostic variable Statistical Analysis was assessed with the use of Fisher’s exact test. Analyses were carried out with the use of SAS A multivariate proportional-hazards regression (version 9.1) and S-PLUS (version 7.0) software. analysis was used to construct a prognostic scoring algorithm for disease-free survival. Because Table 1. Characteristics of 184 Patients at the Beginning of High-Dose the individual prognostic variables correlated with Chemotherapy.* each other significantly, and because our goal was Characteristic No. of Patients (%) to devise a prognostic scoring algorithm that was No. of previous chemotherapy regimens easy to implement clinically, we used the branch- 15 1 135 (73.4) and-bound algorithm of Furnival and Wilson to find the best model among all possible models 2 45 (24.4) that contained three prognostic variables. This ≥3 4 (2.2) algorithm fits all possible three-variable models Histologic type and identifies the model with the highest likeli- Seminoma 35 (19.0) hood score (chi-square statistic). The best four- Nonseminomatous germ-cell tumor 149 (81.0) variable model was also obtained to determine Response to initial chemotherapy whether the value added by including a fourth Complete remission (with or without resection of tumor) 75 (40.8) variable was meaningful. This model did not per- Partial remission (normal hCG and alpha-fetoprotein 9 (4.9) form better than the best three-variable model. levels) We used a bootstrap method in which 1000 sam Other (less than complete remission or partial remission 100 (54.3) ples of 184 patients were randomly selected from with normal hCG and alpha-fetoprotein levels) the original patient cohort.16 For each sample, the Initial IGCCCG stage branch-and-bound algorithm was applied. Low risk 71 (38.6) Next, the model with the highest frequency of Intermediate risk 38 (20.7) selection among the 1000 samples was identified. High risk 75 (40.8) The best three- and four-variable models identi- Platinum sensitivity fied by the bootstrap algorithm were both identical to the best three- and four-variable models for Sensitive 144 (78.3) the original sample, providing support for the Refractory 40 (21.7) validity of the final model. With the use of the Serum hCG level method of Rassi et al.,17 the β regression coef- ≥1000 IU/liter 22 (12.0) ficients from the final model applied to the origi- <1000 IU/liter 162 (88.0) nal sample were used to develop prognostic scores Serum alpha-fetoprotein level (with low scores reflecting a greater probability ≥1000 μg/liter 8 (4.3) of disease-free survival) for each variable in the <1000 μg/liter 176 (95.7) model. These prognostic scores were then summed Beyer score† and split three ways (by comparing the overall patterns of survival according to prognostic score 0 136 (73.9) with a Kaplan–Meier plot and forming three 1 24 (13.0) groups with similar patterns) to create an overall 2 9 (4.9) stratification of patients into low-, intermediate-, 3–4 15 (8.2) and high-risk groups.

* IGCCCG denotes International Germ Cell Cancer Collaborative Group, and hCG human chorionic gonadotropin. Percentages may not sum to 100 because of Results rounding. † A score of 1 point was given for platinum-refractory or absolute platinum- We analyzed the disease-free and overall survival refractory disease, and 2 points were given for each primary mediastinal tumor and for a serum hCG level ≥1000 IU/liter. of patients who received high-dose chemotherapy for at least 1 day. The one patient who was lost to

342 n engl j med 357;4 www.nejm.org july 26, 2007 Salvage Therapy for Metastatic Germ-Cell Tumors follow-up was counted in the group of patients in whom treatment failed. 100 Table 1 lists characteristics of the patients at the start of high-dose chemotherapy. The me- 80 dian age was 31 years (range, 15 to 58). Patients with a low Eastern Cooperative Oncology Group (ECOG) performance status were not excluded, 60 but most patients had an ECOG performance status of 0 or 1. Thirty-five patients had semi- 40

noma only, with no elements of nonseminoma- Overall Survival (%) tous germ-cell tumor, and the remaining 149 pa 20 tients had nonseminomatous germ-cell tumors. The Beyer score was used to stratify patients in 0 low-, intermediate-, or high-risk categories.18 010 2030 4050 6070 80 90 100 110 120 With this scoring system, 1 point was assigned Months since First Day of High-Dose Chemotherapy for platinum-refractory or absolute platinum- No. at Risk 184 161 128 103 80 61 49 27 23 17 7 4 refractory disease and 2 points were assigned Figure 1. Kaplan–Meier Estimates of Overall Survival. for each primary mediastinal tumor and for a The top and bottom lines show the 95% confidence interval. serum hCG level greater than or equal to 1000 IU per liter. Platinum-refractory disease was defined as tumor progression within 4 weeks after the most recent cisplatin-based chemotherapy. Table 2 lists prognostic variables for the 184 Absolute platinum-refractory disease was defined patients in our study. Variables that were signifi- as no response to the initial cisplatin-based che- cantly associated with progression-free survival motherapy. were the use of high-dose chemotherapy as sec- Eleven patients did not receive the scheduled ond-line as compared with third-line chemother- second course of high-dose chemotherapy; five apy, response of the tumor to cisplatin (platinum RETAKE 1st ICM AUTHOR: Einhorn had progressive disease, and four of these five sensitivity), response to initial chemotherapy, 2nd REG F FIGURE: 1 of 2 patients died within 10 months after the initia- favorable prognosis,20 and favorable International 3rd CASE Revised tion of high-dose chemotherapy. The fifth pa- Germ Cell CancerEMail Collaborative GroupLine (IGCCCG)4-C SIZE 4 ARTIST: ts tient has no evidence of disease more than 33 score. Of the 61 patients whoH/T had favorableH/T 22p3 Enon Combo months after two salvage surgeries. Five of the prognostic features, 49 had disease that was in AUTHOR, PLEASE NOTE: 11 patients did not receive a second course of continuous remissionFigure has for been a median redrawn and of type 46 hasmonths been reset. high-dose chemotherapy because of grade 4 toxic (range, 25 to 112). A totalPlease of check18 of carefully. 40 patients effects (nephrotoxicity in 2, hepatotoxicity in 2, with progressive metastatic disease and tumors JOB: 35623 ISSUE: 06-14-07 and pulmonary toxicity in 1). Of these five pa- that were refractory to platinum remained dis- tients, two remained continuously disease-free ease-free for a median of 49 months (range, 22 65 and 43 months after recovering from the toxic to 110). effects of the single course of high-dose chemo- To develop the prognostic scoring algorithm therapy. One patient who received a single course with the use of multivariate proportional-hazards had apparently false elevations of serial serum regression, we included all individual variables alpha-fetoprotein levels in the absence of liver (timing of high-dose chemotherapy, hCG level, disease; he has been disease-free for more than platinum sensitivity, response to initial chemo- 100 months. therapy, histologic types, and IGCCCG score) in During a median follow-up of 48 months the branch-and-bound algorithm except for an (range, 14 to 118), 116 of 184 patients (63%) were alpha-fetoprotein level greater than or equal to continuously disease-free. Of these 116 patients, 1000 μg per liter, since only eight patients had 104 (90%) were disease-free for more than 2 years. an alpha-fetoprotein level of that elevation. The Six additional patients had complete remission best three-variable model included timing of of disease, four after receiving paclitaxel plus high-dose chemotherapy, platinum sensitivity or gemcitabine19 and two after undergoing subse- refractoriness, and IGCCCG stage (Table 3). The quent resection of a germ-cell tumor. Figure 1 prognostic scoring algorithm, based on the shows the Kaplan–Meier estimate of survival. three-variable model, assigned a score of 3 points

n engl j med 357;4 www.nejm.org july 26, 2007 343 T h e new england journal o f medicine

Table 2. Prognostic Variables in 184 Patients.*

Disease-free Death or Survival Survival with Disease P Hazard Ratio Variable (N = 116) (N = 68) Value† (95% CI)‡ no./total no. (%) High-dose chemotherapy Second-line 94/135 (69.6) 41/135 (30.4) 1.00 Third-line or subsequent 22/49 (44.9) 27/49 (55.1) 0.003 2.22 (1.37– 3.62) Markers hCG <1000 IU/liter 102/162 (63.0) 60/162 (37.0) 1.00 hCG ≥1000 IU/liter 14/22 (63.6) 8/22 (36.4) 1.00 1.05 (0.50– 2.20) Alpha-fetoprotein <1000 μg/liter 113/176 (64.2) 63/176 (35.8) 1.00 Alpha-fetoprotein ≥1000 μg/liter 3/8 (37.5) 5/8 (62.5) 0.15 2.40 (0.96– 5.99) Platinum sensitivity Sensitive 98/144 (68.1) 46/144 (31.9) 1.00 Refractory 18/40 (45.0) 22/40 (55.0) 0.01 2.16 (1.30– 3.60) Response to initial chemotherapy Complete remission or partial remission 61/84 (72.6) 23/84 (27.4) 1.00 with normal serum markers Less than complete remission or less than 55/100 (55.0) 45/100 (45.0) 0.01 1.85 (1.12– 3.05) partial remission with normal serum markers Histologic type Seminoma 26/35 (74.3) 9/35 (25.7) 1.00 Nonseminomatous germ-cell tumor 90/149 (60.4) 59/149 (39.6) 0.17 1.56 (0.77– 3.14) IGCCCG stage (at start of initial chemotherapy) Low risk 54/71 (76.1) 17/71 (23.9) 1.00 Intermediate risk 24/38 (63.2) 14/38 (36.8) 1.62 (0.80– 3.29) High risk 38/75 (50.7) 37/75 (49.3) 0.006 2.39 (1.35– 4.25) Prognosis Favorable§ 49/61 (80.3) 12/61 (19.7) 1.00 Unfavorable 67/123 (54.5) 56/123 (45.5) <0.001 2.76 (1.48– 5.16) Beyer score¶ 0 90/136 (66.2) 46/136 (33.8) 1.00 1–2 19/33 (57.6) 14/33 (42.4) 1.37 (0.75– 2.49) 3–4 7/15 (46.7) 8/15 (53.3) 0.25 2.02 (0.95– 4.29)

* IGCCCG denotes International Germ Cell Cancer Collaborative Group, and hCG human chorionic gonadotropin. Percent ages may not sum to 100 because of rounding. † P values were calculated with the use of Fisher’s exact test and are for the comparison of the patients with disease-free survival with patients who died or survived with disease. ‡ The hazard ratio for disease progression was calculated with the use of the univariate proportional-hazards regression model. § Patients with a favorable prognosis had complete remission or partial remission with normal serum markers and received high-dose chemotherapy as initial salvage therapy. ¶ A score of 1 point was given for platinum-refractory or absolute platinum-refractory disease, and 2 points were given for each primary mediastinal tumor and for a serum hCG level ≥1000 IU/liter.

344 n engl j med 357;4 www.nejm.org july 26, 2007 Salvage Therapy for Metastatic Germ-Cell Tumors

Table 3. Results of Multivariate Cox Proportional-Hazards Analysis and Prognostic Score.*

Prognostic Variable Hazard Ratio (95% CI) P Value β Regression Coefficient Prognostic Score† Third-line or subsequent chemotherapy 2.19 (1.35–3.56) 0.002 0.78 3 Platinum-refractory disease 1.74 (1.01–3.00) 0.05 0.55 2 IGCCCG high-risk stage 1.67 (1.00–2.78) 0.05 0.51 2

* The hazard ratio is for disease progression. IGCCCG denotes International Germ Cell Cancer Collaborative Group. † The score was calculated by dividing the regression coefficient by 0.51, multiplying by 2.0, and rounding to the nearest whole number. for third-line chemotherapy, 2 for platinum refractoriness, and 2 for advanced IGCCCG stage. 100 High scores indicated a low probability of dis- Low risk (0 points) 80 ease-free survival. Figure 2 illustrates overall survival according to this scoring algorithm, which 60 stratifies patients according to the level of risk Intermediate risk (2–3 points) (low, 0 points; intermediate, 2 to 3 points; or 40 high, 4 to 7 points). The results of log-rank tests High risk (4–7 points) comparing the disease-free survival curves among Overall Survival (%) 20 the three risk groups were all statistically signifi- 0 cant (P<0.05). 0 10 20 30 40 50 60 70 80 90 100 110 120 The toxic effects of high-dose chemotherapy Months since First Day of High-Dose Chemotherapy were primarily myelosuppression, mucositis, nau- No. at Risk sea, vomiting, dehydration, peripheral neuropa- Low risk 73 63 59 44 34 26 21 10 9 7 4 2 8 Intermediate 64 45 37 35 28 22 19 12 10 7 3 2 thy, and otologic abnormalities. There were three risk sudden drug-related deaths; two were due to he- High risk 47 23 17 13 10 7 5 2 1 1 patic failure, and one was due to pulmonary toxic effects. Table 4 lists the toxic effects that were Figure 2. Disease-free Survival. grade 3 or higher. Acute leukemia developed in The prognostic scoring algorithm, based on the three-variable model, as signed a score of 3 points for third-line chemotherapy, 2 points for plati three patients 11, 21, and 60 months after high- num refractoriness, and 2 points for advanced International Germ Cell dose chemotherapy. None of these patients had Cancer Collaborative Group stage. High scores indicated a low probability of disease-free survival. received maintenance therapy with oral etoposide. RETAKE 1st ICM AUTHOR: Einhorn All three had been treated with high-dose chemo- 2nd REG F FIGURE: 2 of 2 3rd therapy as third-line or later chemotherapy. Two CASE Revised of these patients died during follow-up; the third static germ-cellEMail tumors. PrognosticLine factors,4-C based SIZE ARTIST: ts H/T H/T 22p3 remained alive with cancer that was in complete on data fromEnon more than 5000 suchCombo patients, are 4 remission after antileukemia therapy plus allo- well established. The AUTHOR, factors PLEASE associated NOTE: with geneic bone marrow transplantation. Two of these long-term survivalFigure after has salvagebeen redrawn therapy, and type however, has been reset. three patients have been described previously.21 are less well established. Please check carefully. 18 Glioblastoma multiforme developed in one pa- In 1996,JOB: Beyer35623 et al. reported a multivariateISSUE: 06-14-07 tient 6 years after high-dose chemotherapy. He analysis of factors that can predict cure after had received radiation therapy for brain metasta- high-dose chemotherapy. Adverse prognostic var ses of the germ-cell tumor at the time of the iables in the 283 patients in that study includ initial diagnosis. ed refractoriness to platinum (progression within 4 weeks after treatment with cisplatin), absolute Discussion refractoriness to platinum (no response to initial platinum-based chemotherapy), mediastinal non- Cisplatin-containing combination chemotherapy seminomatous germ-cell tumor, and a serum hCG cures 70% of patients with newly diagnosed meta- level greater than or equal to 1000 IU per liter.

n engl j med 357;4 www.nejm.org july 26, 2007 345 T h e new england journal o f medicine

The improved results in our series, as com- Table 4. Toxic Effects of High-Dose Chemotherapy (Grade 3 or Higher) among 184 Patients.* pared with the 283 patients in the analysis by Beyer et al., reflect the benefit of high-dose chemo No. of No. of therapy given as second-line rather than third- Toxic Effect Patients Deaths line chemotherapy and the administration of two Hematologic (leukemia) 3 2 consecutive rounds of high-dose chemotherapy Renal (serum creatinine, 3–6× ULN) 4 0 with hematopoietic stem-cell rescue. We did not Gastrointestinal 30 0 use a third agent such as cyclophosphamide, ifos- 22-27 Hepatic 6 2 famide, or thiotepa, as others have, because Neurologic 9 0 adding a third agent requires dose reductions of the two most active drugs, carboplatin and etopo- Pulmonary 3 1 side. Only a randomized study could show wheth- * Gastrointestinal toxic effects included colitis, diarrhea, mucositis, nausea, and er the addition of a third agent is beneficial. vomiting. Hepatic toxic effects included a serum albumin level that was less Some variables were associated with very high than 2 g per deciliter; a value for alkaline phosphatase, aspartate aminotrans ferase, or alanine aminotransferase that was >5 to 20 times the upper limit rates of continuous disease-free survival. Of 35 of the normal range (ULN); and a bilirubin level that was >3 to 10 times the patients with a pure seminoma (defined as no ULN. Two patients died from acute hepatic failure that developed after the other cell types and normal serum alpha-feto- first course of high-dose chemotherapy. Neurologic toxic effects included sensory alterations, parasthesias, and weakness interfering with daily activity. protein levels) that relapsed after first- or second- Pulmonary toxic effects included dyspnea interfering with daily activity and line chemotherapy, 26 remained disease-free for symptomatic hiccups interfering with sleep. The acute respiratory distress a median of 43 months (range, 19 to 118). Sixty- syndrome developed in one patient. one patients had favorable prognostic features; these features included cancer that was in com- A prognostic score based on these variables was plete or partial remission and normal serum developed. Primary mediastinal tumors and an hCG and alpha-fetoprotein levels for more than hCG level greater than or equal to 1000 IU per 6 months after the first chemotherapy treatment liter were each assigned 2 points, whereas all and then high-dose chemotherapy as the initial other variables were each assigned 1 point. For salvage therapy.7 Of these 61 patients, 49 had patients with a score of 3 or higher, the rate of disease that was in continuous remission for a disease-free survival at 2 years was only 5%, as median of 46 months (range, 25 to 112). In a compared with 51% for patients with a Beyer recently reported series, 29 of 46 similar patients score of 0. This scoring system, however, was had disease that was in a durable complete re- based on data from patients who were treated mission (median, 69 months) after treatment with between 1984 and 1993; most of these patients paclitaxel plus ifosfamide plus cisplatin.20 The had received only a single course of high-dose minimum follow-up was 2 years. chemotherapy. A total of 91% of these patients Patients with primary mediastinal nonsemi- received two or more induction regimens before nomatous germ-cell tumors were not eligible for receiving high-dose chemotherapy. high-dose carboplatin plus etoposide during this There are some similarities between the prog- study. From 1988 to 1996, 13 patients with this nostic variables in our study (Table 2) and the tumor received high-dose chemotherapy with Beyer scores, but our list of variables reflects bone marrow transplantation. None of them sur- some additions and deletions. Beyer and col- vived disease-free for more than 2 years.28 A larger leagues assigned 2 points for “absolute refrac- international study had poor results with any tory disease,” defined as progression with initial type of salvage chemotherapy, with only 9 of 79 cisplatin-based chemotherapy. Forty-three of the patients (11%) alive and disease-free for more 283 patients (15%) in that study were assigned than 2 years.29 to this category. In our series, only 2 of 184 pa- In a recent study, 219 patients in the high-risk tients had absolute refractory disease. In the study group were randomly assigned to receive initial reported by Beyer et al., serum hCG levels greater treatment with four courses of bleomycin plus than or equal to 1000 IU per liter were assigned etoposide plus cisplatin or two courses of this 2 points. In our series, there was no difference initial treatment followed by two courses of high- in disease-free survival on the basis of hCG levels dose chemotherapy with stem-cell rescue.30 There (P = 1.00). was no benefit from the high-dose chemotherapy.

346 n engl j med 357;4 www.nejm.org july 26, 2007 Salvage Therapy for Metastatic Germ-Cell Tumors

Pico et al. randomly assigned 280 patients to re- germ-cell tumor that is refractory to platinum- ceive salvage chemotherapy with either four cours based chemotherapy or that is not cured by a es of vinblastine (or etoposide) plus ifosfamide cisplatin–ifosfamide regimen as salvage chemo- plus cisplatin or three similar cycles followed by therapy. In our study, 18 of 40 patients with pro- a single course of high-dose carboplatin plus eto- gressive metastatic disease and tumors that were poside plus cyclophosphamide.31 The single cycle refractory to platinum remained disease-free for of high-dose chemotherapy had no effect on the a median of 49 months (range, 22 to 110), and 22 outcome. To our knowledge, these two studies of 49 patients who received high-dose chemo- are the only randomized trials of high-dose che- therapy as third-line or later therapy remained motherapy in patients with germ-cell tumors, and disease-free for a median of 46 months (range, it is disappointing that they did not show an ad- 25 to 112). vantage of such treatment. Presented in part at the 42nd Annual Meeting of the Ameri- can Society of Clinical Oncology, Atlanta, June 2–6, 2006. There should be little or no debate on the use No potential conflict of interest relevant to this article was of high-dose chemotherapy for a patient with a reported.

References 1. Einhorn LH. Treatment of testicular fractory germ cell cancer — a phase I/II tory germ cell tumor patients undergoing cancer: a new and improved model. J Clin trial of high-dose carboplatin and etopo- autologous stem-cell transplantation us- Oncol 1990;8:1777-81. side with autologous bone marrow trans- ing high dose etoposide. J Clin Oncol 2004; 2. Bosl GJ, Motzer RJ. Testicular germ- plantation. J Clin Oncol 1989;7:932-9. 22:2155-8. cell cancer. N Engl J Med 1997;337:242-53. 12. Buckner CD, Clift RA, Fefer A, et al. 22. Siegert W, Beyer J, Strohscheer I, et al. [Erratum, N Engl J Med 1997;337:1403.] High-dose cyclophosphamide (NSC-26271) High dose treatment with carboplatin, 3. Williams SD, Birch R, Einhorn LH, for the treatment of metastatic testicular etoposide, and ifosfamide by autologous Irwin L, Greco FA, Loehrer PJ. Treatment neoplasms. Cancer Chemother Rep 1974; stem-cell transplantation in relapsed or of dissemination germ-cell tumors with 58:709-14. refractory germ cell cancer: a phase I/II cisplatin, bleomycin, and either vinblas- 13. Cooper MA, Einhorn LH. Maintenance study. J Clin Oncol 1994;12:1223-31. tine or etoposide. N Engl J Med 1987;316: chemotherapy with daily oral etoposide 23. Beyer J, Kingreen D, Krause M, et al. 1435-40. following salvage therapy in patients with Long-term survival of patients with recur- 4. International Germ Cell Collaborative germ cell tumors. J Clin Oncol 1995;13: rent or refractory germ cell tumors after Group. International Germ Cell Consensus 1167-9. high dose chemotherapy. Cancer 1997;79: Classification: a prognostic factor-based 14. Kaplan EL, Meier P. Nonparametric 161-8. staging system for metastatic germ cell estimation from incomplete observations. 24. Rodenhuis S, de Wit R, de Mulder cancers. J Clin Oncol 1997;15:594-603. J Am Stat Assoc 1958;53:457-81. PHM, et al. A multi-center prospective 5. Murphy BR, Breeden ES, Donohue JP, 15. Furnival GM, Wilson RW. Regression phase II study of high-dose chemotherapy et al. Surgical salvage of chemorefractory by leaps and bounds. Technometrics 1974; in germ-cell cancer patients relapsing germ cell tumors. J Clin Oncol 1993;11: 16:499-511. from complete remission. Ann Oncol 1999; 324-9. 16. Efron B, Tibshirani RJ. An introduc- 10:1467-74. 6. Loehrer PJ Jr, Gonin R, Nichols CR, tion to the bootstrap. Boca Raton, FL: 25. Rick O, Bokemeyer C, Beyer J, et al. Weathers T, Einhorn LH. Vinblastine plus Chapman & Hall/CRC, 1998. Salvage treatment with paclitaxel, ifosfa ifosfamide plus cisplatin as initial salvage 17. Rassi A Jr, Rassi A, Little WC, et al. mide, and cisplatin plus high-dose carbo- therapy in recurrent germ cell tumor. Development and validation of a risk score platin, etoposide, and thiotepa followed J Clin Oncol 1998;16:2500-4. for predicting death in Chagas’ heart dis- by autologous stem-cell rescue in patients 7. Motzer RJ, Sheinfeld J, Mazumdar M, ease. N Engl J Med 2006;355:799-808. with relapsed or refractory germ cell can- et al. Paclitaxel, ifosfamide, and cisplatin 18. Beyer J, Kramar A, Mandanas R, et al. cer. J Clin Oncol 2001;19:81-8. second-line therapy for patients with re- High-dose chemotherapy as salvage treat- 26. Rosti G, DeGiorgi U, Salvioni R, et al. lapsed testicular germ cell cancer. J Clin ment in germ cell tumors: a multivariate Salvage high-dose chemotherapy in pa- Oncol 2000;18:2413-8. analysis of prognostic variables. J Clin tients with germ cell tumors: an Italian 8. Bhatia S, Abonour R, Porcu P, et al. Oncol 1996;14:2638-45. experience with 84 patients. Cancer 2002; High-dose chemotherapy as initial salvage 19. Einhorn LH, Brames MJ, Juliar B, Wil- 95:309-15. chemotherapy in patients with relapsed liams SD. Phase II study of paclitaxel 27. Lotz JP, Bui B, Gomez F, et al. Sequen- testicular cancer. J Clin Oncol 2000;18: plus gemcitabine salvage chemotherapy tial high-dose chemotherapy protocol for 3346-51. for germ cell tumors after progression relapsed poor prognosis germ cell tumor 9. Motzer RJ, Mazumdar M, Sheinfeld J, following high-dose chemotherapy with combining two mobilization and cytore- et al. Sequential dose-intensive paclitaxel, tandem transplant. J Clin Oncol 2007;25: ductive treatments followed by three high- ifosfamide, carboplatin, and etoposide sal- 513-6. dose chemotherapy regimens supported vage therapy for germ cell tumor patients. 20. Kondagunta GV, Bacik J, Donadio A, by autologous stem cell transplantation: J Clin Oncol 2000;18:1173-80. et al. Combination of paclitaxel, ifosfa results of the phase II multi-centric TAXIF 10. Rick O, Beyer J, Kingreen D, et al. mide, and cisplatin is an effective second- trial. Ann Oncol 2005;16:411-8. High-dose chemotherapy in germ cell tu- line therapy for patients with relapsed 28. Vaena DA, Abonour R, Einhorn LH. mours: a large single centre experience. testicular germ cell tumors. J Clin Oncol Long-term survival after high-dose sal- Eur J Cancer 1998;34:1883-8. 2005;23:6549-55. vage chemotherapy for germ cell malig- 11. Nichols CR, Tricot G, Williams SD, et 21. Houck W, Abonour R, Vance G, Ein- nancies with adverse prognostic variables. al. Dose-intensive chemotherapy in re- horn LH. Secondary leukemias in refrac- J Clin Oncol 2003;21:4100-4.

n engl j med 357;4 www.nejm.org july 26, 2007 347 Salvage Therapy for Metastatic Germ-Cell Tumors

29. Hartmann JT, Einhorn L, Nichols CR, et al. Phase III randomized trial of con- 31. Pico JL, Rosti G, Kramar A, et al. A ran- et al. Second-line chemotherapy in patients ventional-dose chemotherapy with or with- domized trial of high-dose chemotherapy with relapsed extragonadal nonsemino- out high-dose chemotherapy and autolo- in the salvage treatment of patients fail- matous germ cell tumors: results of an gous hematopoietic stem-cell rescue as ing first-line platinum chemotherapy for international multicenter analysis. J Clin first-line treatment for patients with poor- advanced germ cell tumors. Ann Oncol Oncol 2001;19:1641-8. prognosis metastatic germ cell tumors. 2005;16:1152-9. 30. Motzer RJ, Nichols CJ, Margolin KA, J Clin Oncol 2007;25:247-56. Copyright © 2007 Massachusetts Medical Society.

348 n engl j med 357;4 www.nejm.org july 26, 2007 T h e new england journal o f medicine

original article

Partial Thrombosis of the False Lumen in Patients with Acute Type B Aortic Dissection

Thomas T. Tsai, M.D., M.Sc., Arturo Evangelista, M.D., Christoph A. Nienaber, M.D., Truls Myrmel, M.D., Gabriel Meinhardt, M.D., Jeanna V. Cooper, M.S., Dean E. Smith, Ph.D., Toru Suzuki, M.D., Rossella Fattori, M.D., Alfredo Llovet, M.D., James Froehlich, M.D., Stuart Hutchison, M.D., Alessandro Distante, M.D., Thoralf Sundt, M.D., Joshua Beckman, M.D., James L. Januzzi, Jr., M.D., Eric M. Isselbacher, M.D., and Kim A. Eagle, M.D., for the International Registry of Acute Aortic Dissection*

ABSTRACT

Background Patency or thrombosis of the false lumen in type B acute aortic dissection has been From the Department of Internal Medicine, found to predict outcomes. The prognostic implications of partial thrombosis of the Division of Cardiovascular Medicine, Uni- versity of Michigan Medical Center, Ann false lumen have not yet been elucidated. Arbor (T.T.T., J.V.C., D.E.S., J.F., K.A.E.); the Hospital General Universitari Vall Methods d’Hebron, Barcelona (A.E.); the Univer- sity of Rostock, Rostock, Germany (C.A.N.); We examined 201 patients with type B acute aortic dissection who were enrolled in Tromsø University Hospital, Tromsø, Nor- the International Registry of Acute Aortic Dissection between 1996 and 2003 and way (T.M.); Robert-Bosch Krankenhaus, who survived to hospital discharge. Kaplan–Meier mortality curves were stratified Stuttgart, Germany (G.M.); the University of Tokyo, Tokyo (T. Suzuki); University according to the status of the false lumen (patent, partial thrombosis, or complete Hospital S. Orsola, Bologna, Italy (R.F.); thrombosis) as determined during the index hospitalization. Cox proportional- Hospital Universitario 12 de Octubre, hazards analysis was performed to identify independent predictors of death. Madrid (A.L.); St. Michael’s Hospital, To- ronto (S.H.); the National Research Coun- cil, Lecce, Italy (A.D.); the Mayo Clinic, Results Rochester, MN (T. Sundt); Brigham and During the index hospitalization, 114 patients (56.7%) had a patent false lumen, Women’s Hospital, Boston (J.B.); and Massachusetts General Hospital, Boston 68 patients (33.8%) had partial thrombosis of the false lumen, and 19 (9.5%) had (J.L.J., E.M.I.). Address reprint requests complete thrombosis of the false lumen. The mean (±SD) 3-year mortality rate for to Dr. Tsai at the University of Michigan patients with a patent false lumen was 13.7±7.1%, for those with partial thrombosis Cardiovascular Center, 1500 E. Medical Center Dr., Ann Arbor, MI 48109-5853, or was 31.6±12.4%, and for those with complete thrombosis was 22.6±22.6% (median at [email protected]. follow-up, 2.8 years; P = 0.003 by the log-rank test). Independent predictors of postdischarge mortality were partial thrombosis of the false lumen (relative risk, 2.69; *The investigators for the International Registry of Acute Aortic Dissection are 95% confidence interval [CI], 1.45 to 4.98; P = 0.002), a history of aortic aneurysm listed in the Appendix. (relative risk, 2.05; 95% CI, 1.07 to 3.93; P = 0.03), and a history of atherosclerosis (relative risk, 1.87; 95% CI, 1.01 to 3.47; P = 0.05). N Engl J Med 2007;357:349-59. Copyright © 2007 Massachusetts Medical Society. Conclusions Mortality is high after discharge from the hospital among patients with type B acute aortic dissection. Partial thrombosis of the false lumen, as compared with complete patency, is a significant independent predictor of postdischarge mortality in these patients.

n engl j med 357;4 www.nejm.org july 26, 2007 349 T h e new england journal o f medicine

cute aortic dissection is a danger- Study Population ous condition with high in-hospital and We examined data from all patients with type B A follow-up mortality rates. Dissections con- acute aortic dissection enrolled in IRAD between fined to the descending aorta (type B) have better January 1, 1996, and December 31, 2003. Type B in-hospital survival than those involving the as- acute aortic dissection was defined as any non- cending aorta (type A). Up to 89% of patients traumatic dissection not involving the ascending with uncomplicated type B dissections survive to aorta and presenting within 14 days of symptom hospital discharge after receiving effective anti- onset.13,14 Patients were identified prospectively hypertensive therapy.1 However, despite a low in- at presentation or retrospectively from discharge hospital mortality, the short- and long-term prog- diagnoses and from imaging and surgical data- nosis of patients with type B acute aortic dissection bases. Diagnosis was based on confirmatory im- after discharge from the hospital is heterogene aging, intraoperative visualization, or autopsy. ous, with reported survival rates ranging from 56 Of the 532 patients enrolled in IRAD with to 92% at 1 year and from 48 to 82% at 5 years.2‑9 type B acute aortic dissection, 466 were discharged Given the variable prognosis of type B acute from the hospital alive. Postdischarge mortality aortic dissection with current management strat- data were available for 342 patients. Of these, egies, predictors of poor outcomes have been 141 were excluded from our study, including 64 sought. In addition to aortic diameter, a reported with a diagnosis of intramural hematoma, 46 for predictor of outcomes in type B acute aortic dis- whom imaging data on the false lumen were lack- section has been the patency of the false lu- ing, and 31 for whom consensus on the status of men.3,10,11 Studies have suggested that patients the false lumen on imaging was lacking. The with complete thrombosis of the false lumen have distinction between an intramural hematoma and improved outcomes, whereas those with a patent a true dissection with complete thrombosis of the false lumen have an increased risk of aortic ex- false lumen was made by experts at local IRAD pansion and death.3,10,11 To our knowledge, par- centers. Patients were considered to have an intra- tial thrombosis of the false lumen, defined as the mural hematoma if the hematoma extended out- concurrent presence of both flow and thrombus, ward from the lumen in a crescent shape and has not been studied. The purpose of this analy- maintained a constant circumferential relationship sis was to evaluate the incidence of partial throm- with the aortic wall without a demonstrable inti- bosis of the false lumen on cross-sectional imag- mal flap and with no radiologically apparent inti- ing and to assess its effect on mortality in patients mal tear. Our final study population included 201 presenting with type B acute aortic dissection. patients (38% of those enrolled in the registry).

Methods Data Collection A standardized form was used to record clinical The International Registry of Acute Aortic Dis- variables, including information on patient demo- section (IRAD) is a multinational registry of pa- graphics and history, clinical presentation, phys- tients with acute aortic dissection evaluated at 22 ical findings, imaging results, medical and surgi- aortic centers in 11 countries. The registry is cal treatment, and outcomes, including mortality. supported by grants and receives no commercial The data forms were forwarded to the IRAD co- funding. Treatment during the index hospitaliza- ordinating center at the University of Michigan, tion is not standardized but is conducted at the reviewed for internal consistency and face valid- discretion of each patient’s treating physician. ity, and then scanned electronically into a Micro- Full details of the IRAD structure and the meth- soft Access database. ods used have been previously published.1,12 The imaging results were interpreted at each The registry was approved by the institutional patient’s respective tertiary care center by experi- review board or ethics committee at each partici- enced radiologists and echocardiographers and pating center, and a waiver of informed consent were entered on the data form. Each patient un- for retrospective chart review was granted for the derwent spiral computed tomography, transesoph- registry. Individual written informed consent was ageal echocardiography, magnetic resonance im- obtained for the follow-up study. aging, or a combination of these procedures.

350 n engl j med 357;4 www.nejm.org july 26, 2007 Partial Thrombosis of the False Lumen in Type B Aortic Dissection

The status of the false lumen on imaging was aneurysm (21.0%), and prior aortic dissection classified as patent if flow was present in the (11.7%), were not uncommon. In comparison with absence of thrombus, as partially thrombosed if the study population, patients in the IRAD data- both flow and thrombus were present, or as com- base who were not included in the analysis were pletely thrombosed if no flow was present. older (65.9±12.6 years), more likely to have a his- Yearly follow-up data were obtained with the tory of atherosclerosis (39.7%), and less likely to use of standardized forms after the patient was have had a previous aortic dissection (5.4%). discharged. We obtained clinical and imaging In more than 90% of patients, the diagnosis data as well as information about mortality, with of type B acute aortic dissection was confirmed the date of death when known. When applicable, by cross-sectional imaging within 1 day of pre- missing data on mortality were obtained from the sentation. The remaining cases were diagnosed Social Security Death Index.15 within 7 days. On cross-sectional imaging, the false lumen was found to be patent in 114 pa- Statistical Analysis tients (56.7%), partially thrombosed in 68 (33.8%), Three comparison groups were created on the ba- and completely thrombosed in 19 (9.5%) (Table 1). sis of the status of the false lumen: patent, partial The mean number of imaging studies performed thrombosis, or complete thrombosis. The clini- per patient was 1.5; the most frequent procedure cal characteristics of each of the three groups was computed tomography, which was performed were presented as frequencies and percentages for in three quarters of the patients. There were no categorical variables and as means ±SD for con- significant differences in the frequency of per- tinuous variables. Univariate differences among formance of different imaging procedures be- the three groups were compared by the chi-square tween patients with patent, partially thrombosed, test for categorical variables and by analysis of and completely thrombosed false lumens. variance for continuous variables. Univariate associations between all clinical Clinical Features Associated with Status variables, including false-lumen status, and post- of the False Lumen discharge mortality were calculated by Cox re- Patients with complete thrombosis of the false gression analysis. No imputation of missing lumen were significantly older than those in the variables was performed. Stepwise Cox propor- other two false-lumen groups, with a mean age of tional-hazards analysis was performed to iden- 70.9±12.1 years, versus 57.6±13.5 years in patients tify independent predictors of postdischarge with a patent false lumen and 63.3±13.5 years in mortality. The initial modeling used variables patients with partial thrombosis of the false lu- marginally suggestive of an unadjusted associa- men (P<0.001). There were no significant differ- tion with mortality (P<0.20). Variables were re- ences among the three groups in symptoms or viewed for clinical significance before testing. physical findings at presentation (Table 1). With Forward-ascending stepwise selection of variables regard to diagnostic testing, patients with com- after adjustment for age, sex, and in-hospital plete thrombosis of the false lumen were more treatment (medical, surgical, or endovascular) likely to have abnormal electrocardiograms and was performed sequentially, with a default value pleural effusions on chest radiography than pa- for inclusion set at P<0.05. SAS software, version tients with a patent or partially thrombosed false 8.2, was used for all analyses. lumen.

Results In-hospital Treatment and Outcomes All patients initially received medical therapy to Baseline and Imaging Characteristics regulate systolic blood pressure and the velocity The mean age (±SD) of the 201 patients exam- of left ventricular ejection (change in pressure ÷ ined in this analysis was 60.8±13.9 years (Table 1). change in time [dP/dt]). One hundred forty-six The majority of patients (69.1%) were male, 28.4% patients (72.6%) received medical therapy only, 36 were 70 years of age or older, and 77.0% had a (17.9%) underwent surgery, and 19 (9.5%) had history of hypertension. Other coexisting condi- endovascular treatment, defined as stent place- tions, such as atherosclerosis (29.8%), prior aortic ment, fenestration, or both (Table 1). There were

n engl j med 357;4 www.nejm.org july 26, 2007 351 T h e new england journal o f medicine 0.28 0.12 0.42 0.39 0.06 0.19 0.81 0.20 0.41 0.62 0.71 0.24 0.69 0.54 0.42 0.27 0.55 0.37 0.96 0.86 0.31 0.31 0.02 <0.001 <0.001 P Value 19)

7 (36.8) 7 (15.8) 3 12 (63.2) 12 (N 87.8±19.3 70.9±12.1 1/18 (5.6) 1/18 (5.6) 1/18 8/18 (44.4) 8/18 (21.1) 4/19 (16.7) 3/18 (23.5) 4/17 (13.3) 2/15 (17.6) 3/17 (12.5) 2/16 (56.2) 9/16 (35.7) 5/14 0/18 162.4±30.9 14/17 (82.4) 14/17 (89.5) 17/19 (58.8) 10/17 (66.7) 12/18 (82.4) 14/17 (77.8) 14/18 (66.7) 10/15 Complete Thrombosis Complete 68)

5 (7.4) 5 22 (32.4) 22 (36.8) 25 (N 96.8±20.7 63.3±13.5 1/65 (1.5) 1/65 3/68 (4.4) 3/68 (7.4) 5/68 (5.9) 4/68 (7.8) 5/64 174.5±36.4 16/68 (23.5) 16/68 (14.7) 10/68 (18.8) 12/64 (20.6) 13/63 (18.2) 12/66 50/62 (80.6) 50/62 (76.5) 52/68 (36.8) 25/68 (70.6) 48/68 (73.1) 49/67 (81.5) 53/65 (78.8) 52/66 (51.6) 33/64 (54.0) 34/63 Partial Thrombosis Partial Status of the False Lumen 114)

11 (9.7) 11 23 (20.2) 23 (26.3) 30 Patent (N 57.6±13.5 94.7±20.9 6/111 (5.4) 6/111 (1.8) 2/111 8/113 (7.1) 8/113 169.5±37.1 17/111 (15.3) 17/111 75/111 (67.6) 75/111 85/113 (75.2) 85/113 (23.2) 26/112 (19.4) 21/108 (69.9) 79/113 (69.6) 78/112 (20.0) 21/105 (17.9) 19/106 (14.6) 15/103 95/109 (87.2) 95/109 (22.6) 24/106 (22.0) 24/109 (41.3) 43/104 (47.1) 49/104 100/111 (90.1) 100/111 19 (9.5) 19 57 (28.4) 57 (30.8) 62 (N=201) 94.8±20.7 60.8±13.9 170.5±36.3 All Patients 4/194 (2.1) 4/194 11/199 (5.5) 11/199 41/195 (21.0) 41/195 (6.6) 13/197 (19.9) 38/191 (13.8) 25/181 59/198 (29.8) 59/198 (11.7) 23/197 (20.2) 38/188 (20.0) 37/185 (17.6) 33/188 (46.2) 85/184 (51.1) 93/182 159/188 (84.6) 159/188 (69.2) 137/198 (69.4) 136/196 (86.5) 167/193 (73.5) 144/196 154/200 (77.0) 154/200 Characteristics of Patients Stratified According to the Status of the False Lumen.* False the of Status the to According Stratified Patients of Characteristics Normal mediastinum Widened contour aortic Abnormal effusion Pleural Table 1. Table Characteristic Age — yr — Age (%) no. — yr ≥70 Age (%) no. — sex Female (%)† no. no./total — race White (%) no. no./total — syndrome Marfan’s (%) no. no./total — Hypertension (%)‡ no. no./total — Atherosclerosis (%) no. no./total — dissection aortic Previous (%)§ no. no./total — aneurysm aortic Previous (%) no. no./total — Diabetes (%) no. no./total — surgery cardiovascular Previous presentation Clinical (%) no. no./total — pain Chest (%) no. no./total — pain Back (%) no. no./total — pain of onset Abrupt (%) no. no./total — pain Migrating (%) no. — deficit neurologic Any Hg mm — pressure blood Systolic Hg mm — pressure blood Diastolic (%) no. no./total — shock or Hypotension (%) no. no./total — Hypertension (%) no. no./total — deficit pulse Any imaging Diagnostic (%) no. no./total — radiograph Chest

352 n engl j med 357;4 www.nejm.org july 26, 2007 Partial Thrombosis of the False Lumen in Type B Aortic Dissection 0.003 0.50 0.66 0.02 0.84 0.29 0.97 0.41 0.15 0.34 0.37¶ 0.13 0.43 0.15 0.48 0.55 0.23 7 (36.8) 7 (15.8) 3 (15.8) 3 0 17 (89.5) 17 (84.2) 16 1.4±0.7 4.1±0.8 1/7 (14.3) 1/7 1/11 (9.1) 1/11 1/16 (6.2) 1/16 (6.2) 1/16 2/15 (13.3) 2/15 (11.8) 2/17 (13.3) 2/15 0/16 0/15 11/17 (64.7) 11/17 17/17 (100.0) 17/17 9 (13.2) 9 (11.8) 8 16 (23.5) 16 (75.0) 51 49 (72.1) 49 (39.7) 27 4.7±1.1 1.5±0.8 2/38(5.3) 6/61 (9.8) 6/61 6/63 (9.5) 6/63 (4.8) 3/63 (6.8) 4/59 (6.8) 4/59 10/61 (16.4) 10/61 10/65 (15.4) 10/65 (42.4) 14/33 50/67 (74.6) 50/67 (45.3) 29/64 11 (9.6) 11 18 (15.8) 18 85 (74.6) 85 (38.6) 44 (21.1) 24 (69.3) 79 1.5±0.6 4.5±1.2 1/66 (1.5) 1/66 6/101 (5.9) 6/101 5/107 (4.7) 5/107 (8.7) 9/103 30/58 (51.7) 30/58 68/111 (61.3) 68/111 (10.7) 11/103 (14.9) 15/101 18/107 (16.8) 18/107 (12.0) 13/108 33/105 (31.4) 33/105 19 (9.5) 19 78 (38.8) 78 (18.4) 37 (17.9) 36 1.5±0.7 4.5±1.2 151 (75.1) 151 146 (72.6) 146 4/115 (3.5) 4/115 45/98 (45.9) 45/98 14/179 (7.8) 14/179 (6.5) 12/184 (7.1) 13/183 (9.3) 17/182 23/185 (12.4) 23/185 (39.2) 73/186 (11.2) 20/179 (15.8) 29/184 135/195 (69.2) 135/195 Abnormal wave Q Old ischemia or elevations, ST wave, Q New changes wave T or ST Nonspecific Endovascular treatment includes fenestration and stent placement. Plus–minus values are means ±SD. Race was determined by the investigator. Atherosclerosis includes coronary, peripheral, and cerebrovascular disease. Aortic aneurysm includes thoracic and abdominal aneurysm. The P value is for the comparison among surgical, endovascular, and medical management. Electrocardiogram — no./total no. (%) no. no./total — Electrocardiogram patient per performed studies of No. (%) no. — tomography Computed (%) no. — echocardiography Transesophageal (%) no. — imaging resonance Magnetic cm — aorta descending of diameter Widest (%) no. no./total — cm ≥6 aorta ascending of diameter Widest Treatment (%) no. — Surgical (%)‖ no. — Endovascular (%) no. — only Medical complications In-hospital (%) no. no./total — deficit Neurologic (%) no. no./total — Hypotension (%) no. no./total — Malperfusion (%) no. no./total — ischemia Mesenteric (%) no. no./total — failure renal Acute (%) no. no./total — ischemia Limb

‖ * † ‡ § ¶

n engl j med 357;4 www.nejm.org july 26, 2007 353 T h e new england journal o f medicine

P Values 0.50 Overall, 0.003 Partial thrombosis vs. patent, <0.001 Complete thrombosis vs. patent, 0.17 Partial thrombosis vs. complete thrombosis, 0.41

Partial thrombosis

0.25 Complete thrombosis

Mortality Rate Patent

0.00 0 400 800 1200 Days since Follow-up No. at Risk Patent 111 96 75 56 Partial thrombosis 67 48 32 24 Complete thrombosis 19 17 12 4

Figure 1. Kaplan–Meier Mortality Curve Stratified According to the Status of the False Lumen. AUTHOR: Tsai RETAKE 1st P values were calculated by the log-rankICM test. Overall denotes comparison of all three curves. 2nd REG F FIGURE: 1 of 2 3rd CASE Revised EMail Line 4-C SIZE ARTIST: ts no significant differences among the three false- H/Trespectively.H/T However,33p9 the difference between the Enon Combo lumen groups with regard to choice of therapy or mortality curves according to the log-rank test AUTHOR, PLEASE NOTE: rates of specific in-hospital complications.Figure has been redrawn didand typenot has change been reset. significantly. The median follow-up for the 201 patientsPlease ex check- carefully. Predictors of Postdischarge Mortality amined in this analysis wasJOB: 2.835704 years. Figure 1 ISSUE: 07-26-07 shows the Kaplan–Meier mortality curves strati- Candidate univariate predictors of postdischarge fied according to false-lumen status. The mortal- mortality are shown in Table 2. Patients with ity rate was highest in patients with partial throm type B acute aortic dissection who died during bosis of the false lumen, with 1- and 3-year the follow-up period were significantly older and mortality rates of 15.4±8.8% and 31.6±12.4%, significantly more likely to have a history of aor- respectively, versus 5.4±4.2% and 13.7±7.1% in tic aneurysm or atherosclerosis. In addition, they patients with a patent false lumen and 0% and were more likely to have pleural effusions on chest 22.6±22.6% in patients with complete thrombo- radiography. Independent predictors of mortality sis of the false lumen. Separate log-rank testing are shown in Table 3. After adjustment for age, revealed a significant increase in mortality in pa sex, and in-hospital treatment, the key predictors tients with partial thrombosis of the false lumen of mortality were partial thrombosis of the false as compared with patients with a completely pat- lumen versus a patent false lumen, a history of aor ent false lumen (P<0.001). Log-rank testing did tic aneurysm, and a history of atherosclerosis. not reveal significant differences between patients We also assessed the potential effect of surgi- with complete thrombosis of the false lumen cal or endovascular therapy (both of which tend and those with a completely patent false lumen ed to be performed independently of false-lumen (P = 0.17) or with partial thrombosis of the false status) on the relationship between false-lumen lumen (P = 0.41). status and survival. Thirty-six patients underwent In a further analysis, we included patients with surgery and 19 received endovascular therapy; intramural hematoma in the group classified as among the remaining 146 patients, the relative having complete thrombosis of the false lumen. risk of partial thrombosis of the false lumen The number of patients in this group was in- remained a significant independent predictor of creased to 41, and the 1- and 3-year mortality postdischarge mortality (relative risk, 4.01; 95% rates were increased to 5.6±7.5% and 32.4±19.2%, confidence interval, 1.87 to 8.64; P<0.001).

354 n engl j med 357;4 www.nejm.org july 26, 2007 Partial Thrombosis of the False Lumen in Type B Aortic Dissection

Table 2. Candidate Univariate Predictors of Mortality.*

Variable Categorical Variables No. of Deaths/ No. of Patients (%) Risk Ratio (95% CI) P Value Age ≥70 yr 20/57 (35.1) 1.77 (0.99–3.13) 0.05 Female sex 17/62 (27.4) 1.11 (0.62–2.00) 0.73 Atherosclerosis† 23/59 (39.0) 2.26 (1.29–3.96) 0.005 Previous aortic aneurysm‡ 17/41 (41.5) 2.41 (1.34–4.33) 0.003 Chest pain 29/137 (21.2) 0.65 (0.37–1.14) 0.13 Pleural effusion on chest radiograph 11/25 (44.0) 2.00 (1.02–3.95) 0.04 Status of false lumen Patent§ 20/114 (17.5) Partial thrombosis 25/68 (36.8) 2.67 (1.48–4.82) 0.001 Complete thrombosis 5/19 (26.3) 1.86 (0.70–4.98) 0.22 Treatment Medical only§ 35/146 (24.0) Surgical 10/36 (27.8) 1.17 (0.58–2.36) 0.67 Endovascular¶ 5/19 (26.3) 1.07 (0.42–2.75) 0.89 Hypotension after admission 5/12 (41.7) 1.85 (0.73–4.70) 0.19 Continuous Variables Patients Who Died during Follow-up (N = 50) Risk Ratio (95% CI)‖ P Value Age — yr 64.8±14.2 1.03 (1.00–1.05) 0.02 Diameter of ascending aorta — cm 4.6±1.0 1.71 (1.22–2.39) 0.004 Diameter of descending aorta — cm 4.8±1.5 1.29 (0.97–1.73) 0.08

* CI denotes confidence interval. † Atherosclerosis includes coronary, peripheral, and cerebrovascular disease. ‡ Aortic aneurysm includes thoracic and abdominal aneurysm. § Patent false lumen and medical-only treatment are the reference groups. ¶ Endovascular treatment includes fenestration and stent placement. ‖ The risk ratio is per additional year or centimeter.

3,10,11,16,17 Discussion false lumen. These studies have linked patency of the false lumen to adverse events re- In this large cohort of patients with type B acute sulting from aneurysmal dilatation and rupture aortic dissection, mortality was high after dis- during the chronic phase.5,18-20 In our study, par- charge from the hospital, with nearly one in four tial thrombosis of the false lumen was defined patients (24.9%) dying within 3 years. Partial as the concurrent presence of both flow and thrombosis of the false lumen was common thrombus in the false lumen; this condition was (present in a third of patients) and was the stron- not considered a distinct physiological state in gest independent predictor of postdischarge mor- most previous studies and has not been previous tality that we identified: the risk of death in these ly associated with increased mortality. patients was increased by a factor of 2.7 in com- In our study, complete thrombosis of the false parison with patients with a patent false lumen. lumen occurred in a small number of patients Previous small observational studies have sug- who were, on average, 13 years older than pa- gested a lower risk of adverse events and better tients with a completely patent false lumen, a outcomes in patients with complete thrombosis finding similar to that of previous studies.3,19,21 of the false lumen than in those with a patent Since only 19 patients (9.5%) were classified as

n engl j med 357;4 www.nejm.org july 26, 2007 355 T h e new england journal o f medicine

adequate outflow, despite similar systolic pres- Table 3. Independent Predictors of Death after Adjustment with the Use 27-29 of Multivariate Models.* sure. An increase in pressure within the false lumen Hazard Ratio will increase wall tension, which may elevate the Variable (95% CI) P Value risk of aneurysm expansion, redissection, and Age ≥70 yr 1.42 (0.74–2.74) 0.29 rupture and would thus explain the increased Female sex 1.16 (0.62–2.17) 0.64 mortality seen in these patients. Complete throm- Surgical treatment 1.33 (0.49–3.58) 0.57 bosis of the false lumen (Fig. 2C) excludes the Endovascular treatment† 1.38 (0.64–3.01) 0.41 false lumen from the circulation and is thought Previous aortic aneurysm‡ 2.05 (1.07–3.93) 0.03 to be a prerequisite for complete healing. This is the principle on which endovascular stent ther- Atherosclerosis§ 1.87 (1.01–3.47) 0.05 apy is based.24,30 Patent false lumen¶ 1.00 Partial thrombosis may also have a role in the Partial thrombosis of the false lumen 2.69 (1.45–4.98) 0.002 rupture of the false lumen similar to its role in Complete thrombosis of the false lumen 1.02 (0.32–3.22) 0.98 the rupture of abdominal aortic aneurysms. Pre- vious studies have suggested a direct relationship * CI denotes confidence interval. † Endovascular treatment includes fenestration and stent placement. between intraluminal thrombosis and the risk ‡ Aortic aneurysm includes thoracic and abdominal aneurysm. of rupture of an abdominal aortic aneurysm as § Atherosclerosis includes coronary, peripheral, and cerebrovascular disease. a result of hypoxia of the arterial wall adjacent ¶ Patent false lumen is the reference group. to the intraluminal thrombus, which leads to increased local inflammation, neovascularization, having complete thrombosis of the false lumen, and localized wall weakening.31-34 This mecha- comparisons with this group lack statistical nism may be even more pertinent to the false power. We did not find a significant difference lumen of a dissected aorta, since in this setting in mortality between patients with complete the residual outer layers of the aortic wall already thrombosis of the false lumen and patients with have diminished strength. a completely patent false lumen. The small num- In addition to partial thrombosis of the false ber of cases has also impaired the ability to draw lumen, independent predictors of mortality in our conclusions about this group in other studies of study included a history of atherosclerosis and a type B acute aortic dissection.3,7,22-26 Only one pre history of aortic aneurysm. Atherosclerosis has vious study found that complete thrombosis of been previously linked to mortality in patients the false lumen was a predictor of less aortic en- recovering from type B acute aortic dissection. largement; another study found that dissection- Pathological studies have suggested that the wall related mortality was lower among patients with of an aneursymal aortic segment has decreased complete thrombosis of the false lumen.10,19 collagen synthesis, reduced elastin content, and a Determination of the mechanism by which thinner wall as part of a systemic problem through partial thrombosis of the false lumen portends out the peripheral vasculature.33,35,36 These bio- a poor outcome is beyond the scope of this ob- physical properties of the aorta predispose the servational study. However, two possible contrib- entire aorta and its branches to dissection, further uting factors deserve mention. One potential ex- aneurysm formation, or rupture in the future planation relates the pressure within the false and may contribute to the increased mortality in lumen to the presence of partial thrombosis. this group. Whereas a patent false lumen may be perfused As with all observational studies, this investi- by a proximal entry tear and decompressed gation has limitations that must be kept in mind through distal reentry tears (Fig. 2A), formation when the data are interpreted. First, our cohort of a partial thrombus may occlude these distal consists of patients who were treated at aortic tears, impeding outflow and, in the most extreme specialty centers and for whom imaging data on situation (Fig. 2B), resulting in a blind sac. Stud- false-lumen status as well as follow-up vital sta- ies have shown that pulsatile inflow into a lumen tistics were available. As a result, these findings with impaired outflow may lead to a significant do not represent the entire cohort of the IRAD or increase in the mean arterial and diastolic pres- patients followed at community hospitals. Second, sure as compared with that in a lumen with the mortality data available to us did not include

356 n engl j med 357;4 www.nejm.org july 26, 2007 Partial Thrombosis of the False Lumen in Type B Aortic Dissection

A Patent False Lumen without Thrombus B False Lumen with Partial Thrombosis C False Lumen with Complete Thrombosis

BP, 140/70 mm Hg BP, 140/70 mm Hg BP, 140/70 mm Hg

200 mm Hg 200 mm Hg 200 mm Hg

140 120

100 80 BP, 10/10 mm Hg MAP, 10 mm Hg BP, 140/80 mm Hg MAP, 100 mm Hg BP, 120/100 mm Hg MAP, 107 mm Hg 0 mm Hg 0 mm Hg 0 mm Hg

Figure 2. Conceptual Model of Risk According to the Status of the False Lumen. The figure shows a proposed model of the physiological consequences of false-lumen patency or thrombosis, basedCOLOR on FIGUR hemodynamicE studies in ex vivo models and in patients with aortic dissection.27-29 Panel A shows type B aortic dissectionRev6 with patent proximal07/09/07 and

patent distal reentry tears in the absence of thrombus. The blood-pressure tracing shows systolic, diastolic,Author and Drmean. Tsai arterial pressures in the false lumen similar to the pressures in the true lumen. Panel B shows type B aortic dissection with a patent entry tear Fig # 2 and partial thrombosis that occupies the inner circumference of the false lumen and obstructs the reentry tears, forming a blind sac. Title The blood-pressure tracing shows diastolic and mean arterial pressures in the false lumen that exceed the pressures seen in Panel A, with identical pressures in the true lumen. Panel C shows type B aortic dissection with a false lumenME filled with thrombus and no longer communicating with the true lumen. The pressure within the false lumen is likely to be low andDE nonpulsatile.Dr . JarchoBP denotes blood pressure, and MAP mean arterial pressure. Artist Daniel Muller AUTHOR PLEASE NOTE: Figure has been redrawn and type has been reset Please check carefully information on the cause of death. We were vious studies have shown that theIssue majority date 07/26/2007 of therefore unable to evaluate cause-specific mor- deaths in such patients are related to catastro- tality or other end points, such as freedom from phes of the aorta.2-4,9 reoperation, rupture, or redissection, which would Third, imaging techniques were not standard- be necessary to give more plausibility to our ized among centers, and imaging data were col- mechanistic hemodynamic theories. However, pre lected before our study was designed and were

n engl j med 357;4 www.nejm.org july 26, 2007 357 T h e new england journal o f medicine

not subsequently reevaluated. Thus, misclassifica- thrombosis of the false lumen. In this analysis, tion of false-lumen status is possible. Follow-up the differences between mortality curves did not mortality was recorded independently of the des- change significantly. ignation of false-lumen status on the in-hospital In summary, we analyzed data from the IRAD data forms, thereby minimizing any systematic to evaluate the prognosis in patients with type B misclassification bias. acute aortic dissection who survive their initial Fourth, false-lumen status was determined hospitalization. Mortality is high after discharge once during the hospitalization and does not re- from the hospital, with nearly one in four patients flect false-lumen status during the follow-up pe- dying within 3 years. In patients with partial riod. In small studies, false-lumen status changed thrombosis of the false lumen, the risk of death in a minority of patients (18 to 25%) over 10 to is increased by a factor of 2.7 in comparison with 15 months and was not studied in the acute patients with a completely patent false lumen. period.3,21,22 Therefore, we believe the likelihood Supported by the University of Michigan Faculty Group Prac- of a change in false-lumen status is low in the tice, the Varbedian Fund for Aortic Research, and the Endowed Fund for Clinical Research. Dr. Tsai is supported by a training early window of interest. However, surgery or grant from the National Institutes of Health (5T32HL007853-08). endovascular therapy used in a complication- Dr. Nienaber reports receiving lecture fees from Medtronic; specific approach may alter the status of the Dr. Fattori, consulting and lecture fees from Medtronic; Dr. Froehlich, consulting and lecture fees from Sanofi-Aventis and false lumen before discharge. We therefore per- Pfizer, as well as lecture fees from Merck; and Dr. Eagle, con- formed a separate analysis confined to patients sulting fees from GenTac and lecture fees from CV Therapeu- receiving only medical treatment to demonstrate tics, as well as grant support from Sanofi-Aventis and Biosite. No other potential conflict of interest relevant to this article was that the effect of false-lumen status on survival reported. remains significant in this subgroup. We thank Daniel G. Montgomery, B.S., for his work on an Fifth, although intramural hematoma was earlier version of Figure 2. strictly defined in the IRAD, the true ability of cross-sectional imaging to distinguish between intramural hematoma and a completely throm- An interactive ani bosed false lumen in an acute dissection is largely mation showing unknown because of the absence of a gold stan- partial and complete dard. Moreover, only pathoanatomic studies would thrombosis of the be capable of determining whether an intimal false lumen can be viewed at www. tear was present. We therefore reanalyzed the nejm.org. data, including patients with intramural hematoma in the group classified as having complete

Appendix The investigators for the International Registry of Acute Aortic Dissection are as follows: Coprincipal investigators: K.A. Eagle, University of Michigan, Ann Arbor; E.M. Isselbacher, Massachusetts General Hospital, Boston; C.A. Nienaber, University of Rostock, Rostock, Germany. Coinvestigators: E. Bossone, National Research Council, Lecce, Italy; A. Evangelista, Hospital General Universitari Vall d’Hebron, Barcelona; R. Fattori, University Hospital S. Orsola, Bologna, Italy; J. Froehlich, University of Michigan, Ann Arbor; D. Gilon, Hadassah University Hospital, Jerusalem, Israel; S. Hutchison, St. Michael’s Hospital, Toronto; J.L. Januzzi, Jr., Massachusetts General Hospital, Boston; A. Llovet, Hospital Universitario 12 de Octubre, Madrid; D. Mukherjee, University of Kentucky, Lexington; T. Myrmel, Tromsø University Hospital, Tromsø, Norway; P. O’Gara and J. Beckman, Brigham and Women’s Hospital, Boston; J.K. Oh, Mayo Clinic, Roch- ester, MN; L.A. Pape, University of Massachusetts Hospital, Worcester; U. Sechtem and G. Meinhardt, Robert-Bosch Krankenhaus, Stuttgart, Germany; T. Suzuki, University of Tokyo, Tokyo; S. Trimarchi, Policlinico San Donato, San Donato, Italy. Data management and biostatistical support: J.V. Cooper and D.E. Smith, University of Michigan, Ann Arbor.

References 1. Hagan PG, Nienaber CA, Isselbacher 3. Bernard Y, Zimmermann H, Chocron 5. Haverich A, Miller DC, Scott WC, et al. EM, et al. The International Registry of S, et al. False lumen patency as a predictor Acute and chronic aortic dissections — Acute Aortic Dissection (IRAD): new in- of late outcome in aortic dissection. Am J determinants of long-term outcome for sights into an old disease. JAMA 2000;283: Cardiol 2001;87:1378-82. operative survivors. Circulation 1985;72: 897-903. 4. Gysi J, Schaffner T, Mohacsi P, Aesch- II-22–II-34. 2. Elefteriades JA, Lovoulos CJ, Coady bacher B, Althaus U, Carrel T. Early and 6. Glower DD, Speier RH, White WD, MA, Tellides G, Kopf GS, Rizzo JA. Man- late outcome of operated and non-operat- Smith LR, Rankin JS, Wolfe WG. Manage- agement of descending aortic dissection. ed acute dissection of the descending aor- ment and long-term outcome of aortic Ann Thorac Surg 1999;67:2002-5. ta. Eur J Cardiothorac Surg 1997;11:1163-9. dissection. Ann Surg 1991;214:31-41.

358 n engl j med 357;4 www.nejm.org july 26, 2007 Partial Thrombosis of the False Lumen in Type B Aortic Dissection

7. Lansman SL, McCullough JN, Nguyen 18. Masuda Y, Yamada Z, Morooka N, 28. Chung JW, Elkins C, Sakai T, et al. KH, et al. Subtypes of acute aortic dissec- Watanabe S, Inagaki Y. Prognosis of pa- True-lumen collapse in aortic dissection: tion. Ann Thorac Surg 1999;67:1975-8. tients with medically treated aortic dis- part I. Evaluation of causative factors in 8. Schor JS, Yerlioglu ME, Galla JD, Lans- sections. Circulation 1991;84:Suppl:III-7– phantoms with pulsatile flow. Radiology man SL, Ergin MA, Griepp RB. Selective III-13. 2000;214:87-98. management of acute type B aortic dis- 19. Marui A, Mochizuki T, Mitsui N, Koya- 29. Williams DM, Lee DY, Hamilton BH, section: long-term follow-up. Ann Thorac ma T, Kimura F, Horibe M. Toward the et al. The dissected aorta: part III. Anato- Surg 1996;61:1339-41. best treatment for uncomplicated patients my and radiologic diagnosis of branch- 9. Umana JP, Lai DT, Mitchell RS, et al. with type B acute aortic dissection: a con- vessel compromise. Radiology 1997;203: Is medical therapy still the optimal treat- sideration for sound surgical indication. 37-44. ment strategy for patients with acute type B Circulation 1999;100:Suppl:II-275–II-280. 30. Nienaber CA, Fattori R, Lund G, et al. aortic dissections? J Thorac Cardiovasc 20. Sueyoshi E, Sakamoto I, Hayashi K, Nonsurgical reconstruction of thoracic Surg 2002;124:896-910. Yamaguchi T, Imada T. Growth rate of aortic dissection by stent-graft placement. 10. Erbel R, Oelert H, Meyer J, et al. Effect aortic diameter in patients with type B N Engl J Med 1999;340:1539-45. of medical and surgical therapy on aortic aortic dissection during the chronic 31. Satta J, Laara E, Juvonen T. Intralumi- dissection evaluated by transesophageal phase. Circulation 2004;110:Suppl 1:II-256– nal thrombus predicts rupture of an ab- echocardiography: implications for prog- II-261. dominal aortic aneurysm. J Vasc Surg nosis and therapy. Circulation 1993;87: 21. Erbel R, Engberding R, Daniel W, 1996;23:737-9. 1604-15. Roelandt J, Visser C, Rennollet H. Echo- 32. Wolf YG, Thomas WS, Brennan FJ, 11. Pretre R, Von Segesser LK. Aortic dis- cardiography in diagnosis of aortic dissec- Goff WG, Sise MJ, Bernstein EF. Com- section. Lancet 1997;349:1461-4. tion. Lancet 1989;1:457-61. puted tomography scanning findings asso 12. Mehta RH, O’Gara PT, Bossone E, et 22. Mohr-Kahaly S, Erbel R, Rennollet H, ciated with rapid expansion of abdominal al. Acute type A aortic dissection in the et al. Ambulatory follow-up of aortic dis- aortic aneurysms. J Vasc Surg 1994;20: elderly: clinical characteristics, manage- section by transesophageal two-dimen- 529-35. [Erratum, J Vasc Surg 1995;21: ment, and outcomes in the current era. sional and color-coded Doppler echocar- 295.] J Am Coll Cardiol 2002;40:685-92. diography. Circulation 1989;80:24-33. 33. Kazi M, Thyberg J, Religa P, et al. In- 13. Daily PO, Trueblood HW, Stinson EB, 23. Akutsu K, Nejima J, Kiuchi K, et al. fluence of intraluminal thrombus on Wuerflein RD, Shumway NE. Management Effects of the patent false lumen on the structural and cellular composition of ab- of acute aortic dissections. Ann Thorac long-term outcome of type B acute aortic dominal aortic aneurysm wall. J Vasc Surg Surg 1970;10:237-47. dissection. Eur J Cardiothorac Surg 2004; 2003;38:1283-92. 14. Ehrlich M, Fang WC, Grabenwoger M, 26:359-66. 34. Vorp DA, Lee PC, Wang DH, et al. As- Cartes-Zumelzu F, Wolner E, Havel M. 24. Dake MD, Kato N, Mitchell RS, et al. sociation of intraluminal thrombus in ab- Perioperative risk factors for mortality in Endovascular stent-graft placement for the dominal aortic aneurysm with local hy- patients with acute type A aortic dissec- treatment of acute aortic dissection. N Engl poxia and wall weakening. J Vasc Surg tion. Circulation 1998;98:Suppl:II-294– J Med 1999;340:1546-52. 2001;34:291-9. II-298. 25. Eggebrecht H, Naber CK, Bruch C, et 35. Satta J, Laurila A, Paakko P, et al. 15. Sesso HD, Paffenbarger RS, Lee IM. al. Value of plasma fibrin D-dimers for Chronic inflammation and elastin deg- Comparison of National Death Index and detection of acute aortic dissection. J Am radation in abdominal aortic aneurysm World Wide Web death searches. Am J Coll Cardiol 2004;44:804-9. disease: an immunohistochemical and Epidemiol 2000;152:107-11. 26. Juvonen T, Ergin MA, Galla JD, et al. electron microscopic study. Eur J Vasc En- 16. Dinsmore RE, Willerson JT, Buckley Risk factors for rupture of chronic type B dovasc Surg 1998;15:313-9. MJ. Dissecting aneurysm of the aorta: dissections. J Thorac Cardiovasc Surg 1999; 36. Carmo M, Colombo L, Bruno A, et al. aortographic features affecting prognosis. 117:776-86. Alteration of elastin, collagen and their Radiology 1972;105:567-72. 27. Parodi JC, Berguer R, Ferreira LM, La cross-links in abdominal aortic aneu- 17. Ergin MA, Phillips RA, Galla JD, et al. Mura R, Schermerhorn ML. Intra-aneurys- rysms. Eur J Vasc Endovasc Surg 2002;23: Significance of distal false lumen after mal pressure after incomplete endovascu- 543-9. type A dissection repair. Ann Thorac Surg lar exclusion. J Vasc Surg 2001;34:909- Copyright © 2007 Massachusetts Medical Society. 1994;57:820-4. 14.

full text of all journal articles on the world wide web Access to the complete text of the Journal on the Internet is free to all subscribers. To use this Web site, subscribers should go to the Journal’s home page (www.nejm.org) and register by entering their names and subscriber numbers as they appear on their mailing labels. After this one-time registration, subscribers can use their passwords to log on for electronic access to the entire Journal from any computer that is connected to the Internet. Features include a library of all issues since January 1993 and abstracts since January 1975, a full-text search capacity, and a personal archive for saving articles and search results of interest. All articles can be printed in a format that is virtually identical to that of the typeset pages. Beginning 6 months after publication, the full text of all Original Articles and Special Articles is available free to nonsubscribers who have completed a brief registration.

n engl j med 357;4 www.nejm.org july 26, 2007 359 T h e new england journal o f medicine

original article

Rofecoxib and Cardiovascular Adverse Events in Adjuvant Treatment of Colorectal Cancer

David J. Kerr, M.D., Janet A. Dunn, Ph.D., Michael J. Langman, M.D., Justine L. Smith, B.Sc., Rachel S.J. Midgley, M.D., Andrew Stanley, M.Phil., Joanne C. Stokes, B.Sc., Patrick Julier, M.Sc., Claire Iveson, B.Sc., Ravi Duvvuri, B.Sc., and Christopher C. McConkey, M.Sc., for the VICTOR Trial Group*

ABSTRACT

Background From the Oncology Clinical Trials Office, Selective cyclooxygenase inhibitors may retard the progression of cancer, but they University of Oxford, Oxford (D.J.K., J.L.S., have enhanced thrombotic potential. We report on cardiovascular adverse events in R.S.J.M., J.C.S., P.J., C.I., R.D.); the Statis- tical Centre at Warwick Medical School patients receiving rofecoxib to reduce rates of recurrence of colorectal cancer. Clinical Trials Unit, University of Warwick, Coventry (J.A.D., M.J.L., C.C.M.); and Methods City Hospital, Birmingham (A.S.) — all in All serious adverse events that were cardiovascular thrombotic events were reviewed the United Kingdom. Address reprint requests to Dr. Kerr at the Department of in 2434 patients with stage II or III colorectal cancer participating in a randomized, Clinical Pharmacology, University of Ox- placebo-controlled trial of rofecoxib, 25 mg daily, started after potentially curative ford, Radcliffe Infirmary, Woodstock Rd., tumor resection and chemotherapy or radiotherapy as indicated. The trial was termi- Oxford OX2 6HA, United Kingdom, or at [email protected]. nated prematurely owing to worldwide withdrawal of rofecoxib. To examine possible persistent risks, we examined cardiovascular thrombotic events reported up to 24 *Members of the Vioxx in Colorectal Can- months after the trial was closed. cer Therapy: Definition of Optimal Re- gime (VICTOR) Trial Group are listed in Results the Appendix. The median duration of active treatment was 7.4 months. The 1167 patients receiving N Engl J Med 2007;357:360-9. rofecoxib and the 1160 patients receiving placebo were well matched, with a median Copyright © 2007 Massachusetts Medical Society. follow-up period of 33.0 months (interquartile range, 27.6 to 40.1) and 33.4 months (27.7 to 40.4), respectively. Of the 23 confirmed cardiovascular thrombotic events, 16 occurred in the rofecoxib group during or within 14 days after the treatment period, with an estimated relative risk of 2.66 (from the Cox proportional-hazards model; 95% confidence interval [CI], 1.03 to 6.86; P = 0.04). Analysis of the Anti- platelet Trialists’ Collaboration end point (the combined incidence of death from cardiovascular, hemorrhagic, and unknown causes; of nonfatal myocardial infarction; and of nonfatal ischemic and hemorrhagic stroke) gave an unadjusted relative risk of 1.60 (95% CI, 0.57 to 4.51; P = 0.37). Fourteen more cardiovascular thrombotic events, six in the rofecoxib group, were reported within the 2 years after trial closure, with an overall unadjusted relative risk of 1.50 (95% CI, 0.76 to 2.94; P = 0.24). Four patients in the rofecoxib group and two in the placebo group died from thrombotic causes during or within 14 days after the treatment period, and during the follow-up period, one patient in the rofecoxib group and five patients in the placebo group died from cardiovascular causes. Conclusions Rofecoxib therapy was associated with an increased frequency of adverse cardiovascular events among patients with a median study treatment of 7.4 months’ duration. (Current Controlled Trials number, ISRCTN98278138.)

360 n engl j med 357;4 www.nejm.org july 26, 2007 Cardiovascular Toxicity of Rofecoxib in Colorectal Cancer

pproximately half of all patients Methods undergoing potentially curative surgery A for colorectal cancer ultimately have a re- Patients lapse and die of metastatic disease. This has led Patients were randomly assigned to receive rofe- to the introduction of cytotoxic adjuvant therapy,1 coxib or placebo at 151 hospitals in the United the benefits of which are relatively small (5 to Kingdom. Inclusion criteria were as follows: histo- 10% improvement in the 5-year survival rate).2-4 logically proven colorectal carcinoma of stage III A range of laboratory investigations suggest (any tumor stage, N1 or 2, and M0) or stage II that cyclooxygenase-2 (COX-2) plays an important (T3 or 4, N0, and M0) in patients who had under- role in colorectal carcinogenesis during the tran- gone complete resection of the primary tumor sition from adenoma to carcinoma and subse- without gross or microscopical evidence of resid- quently during invasion and metastasis.5-7 Epide- ual disease; World Health Organization perfor- miologic studies have indicated that the incidence mance status 0 or 1; and hematologic and bio- of colorectal cancer is reduced by 30 to 70%8-10 chemical function within the normal range. In in subjects taking nonsteroidal antiinflammatory addition, all patients had to have completed their drugs (NSAIDs). It has been hypothesized that the potentially curative therapy (surgery alone or sur- antineoplastic effects of NSAIDs are mediated gery plus radiotherapy, chemotherapy, or both) by the inhibition of COX-2, and the gastrointes- 12 or fewer weeks previously and had to have given tinal side effects of NSAIDs by the inhibition of written informed consent. Patients with active pep COX-1, suggesting that any anticancer interven- tic ulceration or gastrointestinal bleeding in the tion involving selective COX-2 inhibitors, as com- past year, a history of adverse reactions to NSAIDs, pared with traditional NSAIDs, would reduce the or a known sensitivity to rofecoxib were excluded, risks of complications from peptic ulcer.11 as were those receiving long-term NSAID therapy Rofecoxib (Vioxx, Merck), a potent inhibitor (except for low-dose aspirin, ≤100 mg per day), of COX-2, was hypothesized to reduce rates of those younger than 18 years, and women who tumor recurrence in our randomized, placebo- were pregnant, lactating, or premenopausal but controlled trial — the Vioxx in Colorectal Cancer not using contraception. Patients with a history Therapy: Definition of Optimal Regime (VICTOR) of cancer (other than adequately treated in situ trial — of patients who had undergone poten- carcinoma of the cervix or basal or squamous-cell tially curative surgery for colorectal cancer. Re- carcinoma), inflammatory bowel disease, or se- cruitment for the VICTOR trial stopped in Sep vere congestive heart failure were also excluded. tember 2004, when Merck withdrew the drug worldwide after a significant increase in con- Trial Design firmed cardiovascular thrombotic events was We planned to randomly assign the use of rofe- noted in the Adenomatous Polyp Prevention on coxib (one 25-mg tablet daily) or placebo to 7000 Vioxx (APPROVe) trial.12 An excess of vascular patients, with half of each group receiving the events was then found in the Adenoma Preven- study drug for 2 years and the other for 5 years, tion with Celecoxib (APC) polyp-prevention trial effectively a four-group design. Local investiga- of celecoxib.13 Evidence from well-designed, ran- tors randomly assigned patients through the domized trials, and their meta-analysis,14 pro- VICTOR Trial Office, and each patient was as- vides support for a moderate increase in vascular signed to a study drug in a double-blind fashion. event rates associated with the use of COX-2 in- The VICTOR Trial Office supplied rofecoxib and hibitors, but there is less clarity about the dura- placebo to participating hospitals every 6 months. tion of drug exposure that is responsible for the risk and whether it is equivalent in patients with Protocol Modifications and in those without established cancer, for Data-collection forms were amended in January whom the potential for benefit from reduced can 2004 (22 months into the trial) to solicit baseline cer progression may be large. In our study, we data on cardiovascular risk factors, both for newly compared rates of cardiovascular thrombotic recruited patients and for patients already entered events and death during the period of study (Fig. 1 in the Supplementary Appendix, available treatment and for 2 years after trial closure. with the full text of this article at www.nejm.

n engl j med 357;4 www.nejm.org july 26, 2007 361 T h e new england journal o f medicine

org). Information sheets were amended twice to search Campaign, the West Midlands Multicenter reflect evolving data on the possible adverse car- Research Ethics Committee, and local research diovascular effects of rofecoxib. After the world- ethics committees at participating centers. The wide withdrawal of rofecoxib, all investigators and trial was supported by an unrestricted grant from patients were informed, all study treatment was Merck, which also provided the rofecoxib and stopped, and follow-up was initiated. matching placebo and stood to provide indemnity but otherwise had no input into data ac- Cardiovascular Events crual or data analysis or control over manuscript Blinded systematic review of all reported serious preparation. The randomization, data collection, adverse events that were potentially cardiovascu- monitoring, and follow-up were coordinated by lar thrombotic events and that were reported dur- the VICTOR Trial Office (initially located at the ing treatment or within 14 days after the treat- University of Birmingham and then relocated to ment period — the primary cardiovascular event the University of Oxford). Study data were analyzed end point — was conducted by an independent, by three of the authors at the Statistical Centre, expert physician panel selected from academic University of Warwick, with an agreement to pro- centers in the United States and Europe by Merck vide Merck with reports of serious adverse events (Table 1 in the Supplementary Appendix). The use that occurred after randomization and within 14 of published reporting systems12,15 and adjudi- days after the end of the treatment period, as well cation by this previously assembled review panel as with a copy of the trial database at prespecified allowed for consistency of adverse-event report- analysis points. An independent data and safety ing across the placebo-controlled rofecoxib trials monitoring committee was appointed. Pharmaco- (VICTOR, APPROVe, and Vioxx in Prostate Cancer vigilance reports were reviewed by the Medicines [VIP]), permitting a patient-level meta-analysis. and Healthcare Products Regulatory Agency (the Thrombotic events were defined as fatal and non- U.K. regulatory authority) and by the West Mid- fatal myocardial infarction, unstable angina, sud- lands Multicenter Research Ethics Committee. All den death from cardiac causes, fatal and nonfatal authors contributed to the writing of the manu- ischemic stroke, transient ischemic attack, periph- script, and the VICTOR Trial Office vouches for eral arterial thrombosis, peripheral venous throm- the accuracy and completeness of the data and bosis, and pulmonary embolism. Also analyzed analysis. were serious adverse events meeting the Anti- platelet Trialists’ Collaboration (APTC) criteria15: Statistical Analysis the combined incidence of death from cardiovas- Although our trial was not designed to analyze cular, hemorrhagic, and unknown causes; of non- cardiovascular data, it had a statistical power of fatal myocardial infarction; and of nonfatal is 90% to detect an increase by a factor of 2 in the chemic and hemorrhagic stroke. To avoid the overall risk of cardiovascular events, from 0.5 to possibility that censoring cardiovascular-event 1%, in patients taking rofecoxib for up to 2 years. data 14 days after drug discontinuation, which The statistical power was insufficient for com- was the approach used in the APPROVe trial,12 parisons of risk according to duration of study might distort the outcome data by ruling out later treatment. toxicity, further analysis was performed on all Kaplan–Meier curves were used to assess the cardiovascular events that occurred during the time from the start of study treatment to the time treatment period or within 24 months after trial of reporting of confirmed cardiovascular events closure. These events were adjudicated by two of (serious adverse events that were cardiovascular the authors. All patients who consented to inclu- thrombotic events or Antiplatelet Trialists’ Col- sion in the study were registered centrally, and laboration end points). Relative risks for con- death certificates were automatically forwarded to firmed cardiovascular events in the rofecoxib the VICTOR Trial Office. group, as compared with the placebo group, were calculated with the use of Cox proportional- Ethics and Indemnity hazards regression analysis, with study treatment Our study was designed by the investigators, and as a factor. Relative risks were also adjusted ac- the protocol was peer reviewed and endorsed by cording to age, use or nonuse of adjuvant chemo- the Clinical Trials Committee of the Cancer Re- therapy, and presence or absence of cardiovascu-

362 n engl j med 357;4 www.nejm.org july 26, 2007 Cardiovascular Toxicity of Rofecoxib in Colorectal Cancer lar risk factors at baseline. All reported P values in the placebo group and one in the rofecoxib are two-sided, and P values less than 0.05 were group received the incorrect medication for the considered to indicate statistical significance. first 6 months, and one patient in the rofecoxib group switched to placebo for a period of 3 weeks Results before switching back. Fifty patients in the rofecoxib group and 57 in the placebo group had not Between April 2002 and September 2004, a total yet started the study treatment when rofecoxib of 2434 patients were recruited at 151 hospitals was withdrawn. The intention-to-treat population in the United Kingdom. One patient in the rofe- included in analyses therefore comprised 1167 coxib group was found to be ineligible because of patients in the rofecoxib group and 1160 patients an incomplete resection, and one patient in the in the placebo group (Fig. 2 in the Supplementary placebo group was ineligible because random as- Appendix). The duration of study treatment was signment took place more than 12 weeks after sur- known only approximately for 65 patients receiving gery. These patients were included in the analyses. rofecoxib and for 62 patients receiving placebo. The intention-to-treat population comprised Table 1 shows that the assignment of study 1217 patients in the rofecoxib group and 1217 treatment was balanced on the basis of age, sex, patients in the placebo group. Owing to an ad- disease site, cancer stage, and receipt or non ministrative error at a site pharmacy, one patient receipt of previous adjuvant chemotherapy. The

Table 1. Baseline Characteristics of the Patients, According to Study Group.

Rofecoxib Group Placebo Group Characteristic (N = 1167) (N = 1160) P Value Stage of cancer — no. (%) 0.79 II 551 (47.2) 554 (47.8) III 616 (52.8) 606 (52.2) Site of cancer — no. (%) 0.46 Colon 756 (64.8) 767 (66.1) Rectum and sigmoid colon 93 (8.0) 77 (6.6) Rectum 318 (27.2) 316 (27.2) Adjuvant chemotherapy — no. (%) 755 (64.7) 748 (64.5) 0.91 Age 0.73 Median — yr 65 65 Interquartile range — yr 58–71 57–71 <65 yr — no. (%) 574 (49.2) 579 (49.9) Male sex — no. (%) 745 (63.8) 742 (64.0) 0.95 White race — no. (%)* 1147 (98.3) 1139 (98.2) 0.86 No history of hypertension — no. (%)† 825 (70.7) 852 (73.4) 0.14 History of diabetes — no. (%)† 102 (8.7) 65 (5.6) 0.003 No history of hyperlipidemia — no. (%)‡ 601 (51.5) 588 (50.7) 0.70 Current smoker — no. (%)§ 141 (12.1) 129 (11.1) 0.47 History of symptomatic atherosclerotic disease — no. (%)¶ 249 (21.3) 222 (19.1) 0.19 Low-dose aspirin use at baseline — no. (%) 101 (8.7) 80 (6.9) 0.11

* Race was self-reported. † The presence or absence of a history was unknown for three patients. ‡ The presence or absence of a history of hyperlipidemia was unknown for 999 patients. § The current smoking status was unknown for 20 patients. ¶ A history of symptomatic atherosclerotic disease was a composite measure defined as the presence of at least one risk factor in the “Documented History of Vascular Disease” section and at least two risk factors in the “Cardiac Risk Factors” section of the VICTOR Cardiovascular Risk Assessment Form (Fig. 2 in the Supplementary Appendix).

n engl j med 357;4 www.nejm.org july 26, 2007 363 T h e new england journal o f medicine

risk-assessment forms were all received after ran- group). Since only 4% of patients received the domization (half before the trial was unblinded), study drug for more than 2 years, the randomiza- and it took approximately 18 months after trial tion variable originally planned to reflect the du- closure to construct a validated database for ration of study treatment (2 years or 5 years) is analysis. Slightly more patients in the rofecoxib unimportant. group than in the placebo group had predefined Thirty-five potential cardiovascular thrombotic cardiovascular risk factors (Table 1). The median events occurring during or within 14 days after duration of trial treatment was 7.4 months (inter- the treatment period were adjudicated in a blind- quartile range, 3.1 to 14.0) in the rofecoxib group ed fashion by the independent panel convened by and 8.2 months (interquartile range, 3.7 to 15.0) Merck (Table 1 in the Supplementary Appendix), in the placebo group, with 33% of all patients including 11 events that were reported after the having received the study drug for at least 12 unblinding of study treatment in nine patients. months (Table 2). The median duration of follow- Sixteen events in 15 patients receiving rofecoxib up was 33.0 months (interquartile range, 27.6 to were confirmed to be cardiovascular thrombotic 40.1) in the rofecoxib group and 33.4 months events, as compared with seven events in 6 pa- (interquartile range, 27.7 to 40.4) in the placebo tients receiving placebo. Three of these events group. occurred while the patients were taking other Before the worldwide withdrawal of rofecoxib, NSAIDs (sudden death from cardiac causes in a 342 patients receiving rofecoxib and 268 patients patient in the placebo group who was taking di- receiving placebo discontinued the study drug be clofenac, transient ischemic attack in a patient in fore its intended completion, but all of these pa- the rofecoxib group who was taking diclofenac, tients were included in the intention-to-treat pop- and peripheral venous thrombosis in a patient in ulation of 1167 patients in the rofecoxib group the rofecoxib group who was taking meloxicam). and 1160 patients in the placebo group. The most There were 10 events qualifying as an APTC end common medical reasons for early discontinua- point in nine patients receiving rofecoxib, as com- tion of study drug were gastrointestinal pain or pared with 6 events in six patients receiving pla- heartburn (15 patients in the rofecoxib group cebo. Rates of confirmed cardiovascular throm- and 5 in the placebo group), analgesia required botic events per 100 patient-years are presented for arthritis (4 and 15, respectively), hypertension in Table 3 and in Kaplan–Meier plots in Figure 1. (7 and 1), renal impairment (7 and 1), diarrhea The median duration of study treatment before (4 and 4), and heart failure (2 in the rofecoxib a cardiovascular event was reported was 157 days in the rofecoxib group and 195 days in the placebo group (Table 2 in the Supplementary Appendix). Table 2. Reported Duration of Study Treatment.* The relative risk of a cardiovascular thrombotic Rofecoxib Placebo event during or within 14 days after the treat- Group Group ment period was 2.66 (95% confidence interval Duration (N = 1167) (N = 1160) [CI], 1.03 to 6.86) among patients receiving ro- <30 days (no.) 112 72 fecoxib, as compared with those receiving placebo 30 days to <6 mo (no.) 381 368 (P = 0.04). The relative risk was slightly reduced 6 to <12 mo (no.) 307 323 after adjustment for cardiovascular risk factors 12 to <24 mo (no.) 321 348 (2.41; 95% CI, 0.93 to 6.26; P = 0.07). Analysis of ≥24 mo (no.) 45 48 the APTC end points showed an unadjusted relative risk of 1.60 (95% CI, 0.57 to 4.51; P = 0.37) Unknown (no.) 1 1 and an adjusted relative risk of 1.42 (95% CI, 0.50 Median duration (days) to 4.03; P = 0.52). No. 226 249 Further analysis, in which all cardiovascular Interquartile range 94–425 112–456 events that occurred during the treatment period Total patient-yr 889 946 and all those reported within 24 months after trial closure were combined, was performed. An * The data include those for 65 patients in the rofecoxib additional 14 cardiovascular events were noted group and 62 in the placebo group whose durations of study treatment were known only approximately. (Table 3, and Table 2 in the Supplementary Appen dix) after adjudication of all adverse events re-

364 n engl j med 357;4 www.nejm.org july 26, 2007 Cardiovascular Toxicity of Rofecoxib in Colorectal Cancer

No. of Events/ of No. 0.61 (0.22–1.32) 0.61 (0.22–1.32) 0.61 (0.11–1.04) 0.41 (0.00–0.57) 0.10 (0.00–0.57) 0.10 (0.15–0.56) 0.31 0.20 (0.02–0.73) 0.20 (0.23–0.72) 0.43 1160)

100 Patient-Yr (95% CI) (95% Patient-Yr 100 =

1 986 986 3257 3265 Patient-Yr Total No. of No. Total Placebo Group (N

1 (0.1) 1 (0.1) 1 6 (0.5) 6 (0.5) 6 (0.3) 4 (0.2) 2 1 1 1 1 1 1 1 2 No. of No. 14 (1.2) 14 (0.9) 10 Patients (%) Patients

No. of Events/ of No. 1.62 (0.91–2.67) 1.62 0.97 (0.44–1.84) 0.97 (0.37–1.70) 0.86 (0.07–0.95) 0.32 (0.18–1.26) 0.54 (0.00–0.40) 0.00 (0.40–0.99) 0.65 (0.21–0.68) 0.40 1167)

100 Patient-Yr (95% CI) (95% Patient-Yr 100

927 928 3252 3269 Patient-Yr Total No. of No. Total Rofecoxib Group (N

9 (0.8) 9 (0.7) 8 (0.3) 3 (0.4) 5 1 1 1 3 3 2 3 2 0 No. of No. 15 (1.3) 15 (1.8) 21 (1.1) 13 Patients (%) Patients 15 Incidence of Confirmed Cardiovascular Adverse Events during and after the Treatment Period.* Treatment the after and during Events Adverse Cardiovascular Confirmed of Incidence closure Myocardial infarction Myocardial infarction myocardial Fatal causes cardiac from death Sudden pectoris angina Unstable thrombus Cardiac thrombosis venous Peripheral embolism Pulmonary stroke cerebrovascular Ischemic attack ischemic Transient stroke cerebrovascular hemorrhagic Fatal aneurysm cerebral Ruptured Table 3. Table Event During or within 14 days after treatment period treatment after days 14 within or During event thrombotic Cardiovascular point end APTC events cardiac All events vascular peripheral All events cerebrovascular All events hemorrhagic All trial after months 24 within or period treatment During event thrombotic Cardiovascular point end APTC The APTC end point was defined as the combined incidence of death from cardiovascular, hemorrhagic, and unknown causes; of nonfatal myocardial infarction; and of nonfatal ische mic and hemorrhagic stroke.

n engl j med 357;4 www.nejm.org july 26, 2007 365 T h e new england journal o f medicine

A Cardiovascular Thrombotic Events, Censored 14 Days B Cardiovascular Thrombotic Events, Censored 2 Years after the Treatment Period after Trial Closure 5.5 5.5 5.0 5.0 4.5 4.5 4.0 4.0 Rofecoxib 3.5 group 3.5 3.0 3.0 2.5 2.5 Rofecoxib group 2.0 P=0.04 2.0 P=0.24 1.5 1.5 Cumulative Incidence (%) Cumulative Incidence (%) 1.0 Placebo 1.0 Placebo 0.5 group 0.5 group 0.0 0.0 0 6 12 18 24 0 12 24 36 48 Months Months No. at Risk No. at Risk Rofecoxib 1166 710 383 181 55 Rofecoxib 1167 1134 1064 445 64 Placebo 1159 748 411 211 63 Placebo 1160 1131 1060 458 67

C APTC End Points, Censored 14 Days after the Treatment Period D APTC End Points, Censored 2 Years after Trial Closure 5.5 5.5 5.0 5.0 4.5 4.5 4.0 4.0 3.5 3.5 3.0 3.0 Rofecoxib 2.5 group 2.5 2.0 2.0 P=0.37 P=0.54 Rofecoxib 1.5 1.5 group Cumulative Incidence (%) 1.0 Placebo Cumulative Incidence (%) 1.0 group 0.5 0.5 Placebo group 0.0 0.0 0 6 12 18 24 0 12 24 36 48 Months Months No. at Risk No. at Risk Rofecoxib 1166 711 383 181 55 Rofecoxib 1167 1139 1071 448 64 Placebo 1159 748 411 212 63 Placebo 1160 1134 1063 459 67

Figure 1. Cumulative Incidence of Cardiovascular Adverse Events. Cardiovascular thrombotic events (Panels A and B) and Antiplatelet Trialists’ Collaboration (APTC) end points (Panels C and D) are shown according to treatment group. In Panels A and C, one patient in each group was not included in the analysis because treatment duration was unknown. Neither patient had an adverse event. The APTC end point was defined as the combined incidence of death from cardio- AUTHOR: Kerr RETAKE 1st vascular, hemorrhagic, and unknown causes; ofICM nonfatal myocardial infarction; and of nonfatal ischemic and hemorrhagic stroke.15 I bars 2nd REG F FIGURE: 1 of 1 indicate 95% confidence intervals. 3rd CASE Revised EMail Line 4-C SIZE ARTIST: sw ported within 24 months after trial closure.H/T The H/Tthree patients36p6 in the rofecoxib group presented Enon Combo resulting relative risk of a cardiovascular throm- with congestive cardiac failure during the treat- AUTHOR, PLEASE NOTE: botic event, unadjusted Figurefor cardiovascular has been redrawn and risk type hasment been reset.period. Overall, four patients in the rofe- factors, was 1.50 (95% CI, 0.76 to 2.94;Please checkP = 0.24). carefully. coxib group and two in the placebo group died as Analysis of the APTC end points showed an un- a result of thrombotic events occurring during or JOB: 35704 ISSUE: 07-26-07 adjusted relative risk of 1.29 (95% CI, 0.57 to 2.95). within 14 days after the treatment period. An ad- Although information on blood pressure and re- ditional six deaths from cardiovascular throm- nal function was not collected systematically, botic events (one patient in the rofecoxib group

366 n engl j med 357;4 www.nejm.org july 26, 2007 Cardiovascular Toxicity of Rofecoxib in Colorectal Cancer and five in the placebo group), identified by death more common in each of the celecoxib groups certification, were reported within 24 months than in the placebo group (200-mg group: hazard after trial closure. The total numbers of patients ratio, 2.3; 95% CI, 0.9 to 5.5; 400-mg group: haz- who died from cardiovascular causes — five in ard ratio, 3.4; 95% CI, 1.4 to 7.8). the rofecoxib group and seven in the placebo A recent meta-analysis of 138 randomized group — did not differ significantly. trials involving 145,373 participants assessed the risk of vascular events from the use of selective Discussion COX-2 inhibitors and traditional NSAIDs.14 In all, 121 placebo-controlled trials of selective COX-2 The chief limitations of our study are the relative inhibitors (predominantly rofecoxib and celecox- ly small number of events available for analysis ib) were analyzed, and the authors found a pro- and the relatively short duration of exposure to portional increase by nearly a factor of 2 in the the study drug (median, 7.4 months). However, risk of myocardial infarction among patients re- our findings suggest an increased risk of a cardio- ceiving COX-2 inhibitors as compared with pla- vascular thrombotic event in patients randomly cebo (hazard ratio, 1.86; 95% CI, 1.33 to 2.59; assigned to receive rofecoxib (as compared with P = 0.003) but no significant difference in the in- those randomly assigned to receive placebo) as cidence of stroke. Too few vascular events were adjuvant treatment for the prevention of a recur- available to study the influence of dose, but the rence of colorectal cancer (relative risk of an event investigators noted that two thirds of the vascu- during or within 14 days after the treatment peri- lar events had occurred in the nine long-term od, 2.66; 95% CI, 1.03 to 6.86; P = 0.04). These trials with treatment periods of 1 year or more. findings are consistent with those obtained in It would appear from our study that patients tak- other placebo-controlled studies of treatment with ing rofecoxib for fewer than 18 months may be COX-2 inhibitors in patients with colorectal ade- at increased risk for a cardiovascular thrombotic noma. Extension of the period of observation of event, since 50% of all such reported episodes cardiovascular events to 24 months after trial clo- occurred in patients treated for fewer than 12 sure did not show a statistically significant ad- months. There has been a thorough statistical verse effect of rofecoxib (relative risk, 1.50; 95% critique of the original interpretation of the time- CI, 0.76 to 2.94; P = 0.24). This information was to-event data of the APPROVe investigators, which generated by the serious-adverse-event reporting has challenged their assessment of the data.16 system that was maintained throughout the fol- Observational information, typically from the low-up period and by central notification of examination of large databases, suggests that death certification. It is possible that worldwide treatment with COX-2 inhibitors may enhance withdrawal of rofecoxib led to the underreporting cardiovascular risk and that such risk may be of adverse events during the follow-up period. greater than, or the same as, that associated with Bresalier et al.12 reported results of the the use of nonselective NSAIDs.17-19 A recent sys- APPROVe trial, in which 2586 patients with a his- tematic review of observational data reported a tory of colorectal adenoma were randomly as- dose-related relative risk of cardiovascular events signed to receive placebo or 25 mg of rofecoxib of 1.33 (95% CI, 0.91 to 1.23) with 25 mg of daily (as in our study). The authors found an in- rofecoxib per day or less, as compared with place creased relative risk of cardiovascular thrombotic bo, and a relative risk of 2.19 (95% CI, 1.64 to events in the rofecoxib group (1.92; 95% CI, 1.19 2.91) with more than 25 mg of rofecoxib per to 3.11) and claimed that this risk was apparent day.20 The ability of population-based studies to after 18 months of treatment. Solomon et al.13 give definitive answers is inevitably limited be- reviewed all potentially serious cardiovascular cause of the difficulty in controlling for con- events in 2035 patients with a history of colorec- founders; however, the summary relative risk in tal adenomatous polyps who had been enrolled in the systematic review20 is similar to that found the APC study, which compared two doses of the in the meta-analysis of randomized trials.14 A re selective COX-2 inhibitor celecoxib (200 mg or cent report from the European Medicines Agency21 400 mg twice daily) with placebo. The composite concluded that epidemiologic evidence and up- end point of death from cardiovascular causes, dated trial data continue to point to an increased myocardial infarction, stroke, or heart failure was thrombotic risk with COX-2 inhibitors, possibly

n engl j med 357;4 www.nejm.org july 26, 2007 367 T h e new england journal o f medicine

accounting for about three events per 1000 patient- or death from cancer, but prolonged follow-up years. will provide more informative data. The mechanism linking the use of COX-2 inhibitors to an increased incidence of thrombotic Sponsored by the University of Oxford and an educational vascular events has not been precisely elucidated. study grant from Merck. Drs. Kerr and Dunn report receiving grant support from Can- Cyclooxygenase and its prostanoid products have cer Research UK; Dr. Dunn, receiving consulting fees from important roles in regulating factors affecting the Merck and Johnson & Johnson; and Dr. Langman, serving as an risk of thrombosis; for instance, thromboxane advisor to Merck and to Novartis and as a consultant to lawyers representing Merck. No other potential conflict of interest rele- synthase is regulated by COX-2. It is generally vant to this article was reported. accepted that patients with cancer have a higher We thank the VICTOR trial staff (F. Duchesne, K. Reed, S. risk of thrombosis than does the general popu- Pendlebury, I. Kennedy, G. Davis, J. Birtwistle, L. Blair, A. Bange, K. Bell, L. Dunham, R. Deacon, A. Mellor, and B. Mikolajczyk); lation, and a positive correlation has been found the clinical coordinators (S. Grumett and D. Rea); the VICTOR between increased expression of thromboxane Trial Advisory Group; the U.K. National Cancer Research Insti- synthase and shortened survival from bladder tute and Cancer Research Campaign Clinical Trials Awards and 22 Advisory Committee for permitting the support from the U.K. cancer. Current data are insufficiently mature to National Cancer Research Network; and all the nurses, clini- permit commentary on the risks of recurrence of cians, and patients who participated in this trial.

Appendix Members of the VICTOR trial were as follows. Independent Data and Safety Monitoring Committee: Finsen Institute, Copenhagen — H.H. Hansen; Centre for Statistics in Medicine, Oxford, United Kingdom — D. Altman; Churchill Hospital, Oxford, United Kingdom — D. Talbot; Principal investigators: Mayday University Hospital — M. Abulafi; North Staffordshire Royal Infirmary — F. Adab; Queen Elizabeth Hospital King’s Lynn — A. Ahmad; St. James University Hospital — S. Ambrose; Wexham Park Hospital — R. Ashford; Singleton Hospital — C. Askill; North Tyneside General Hospital — P. Atherton; Royal Shrewsbury Hospital — S. Awwad; South Tyneside District Hospital — A. Azzabi; Torbay Hospital — N. Bailey; Withybush General Hospital — A. Barnes; St. Mary’s Hospital Newport — C. Baughan; Beesley Conquest Hospital — S. Harriet; Gloucester Royal Hospital — K. Benstead; Nottingham City Hospital — E. Bessell; Great Western Hospital — C. Blesing; Bradford Royal Infirmary — C. Bradley; Royal Gwent Hospital — A. Brewster; North Middlesex University Hospital — J. Bridgewater; Macclesfield District General Hospital — W. Brough; Norfolk & Norwich University Hospital — A. Bulman; Glasgow Royal Infirmary and Western Infirmary — J. Cassidy; Derbyshire Royal Infirmary — P.R. Chakraborti; Kidderminster Hospital — M. Churn; St. Mary’s Hospital London — S. Cleator; Pinderfields General Hospital — R. Cooper; Addenbrooke’s Hospital — P. Corrie; Newcastle General Hospital — F. Coxon; Airedale General Hospital — M. Crawford; Princess Royal Hospital — A. Crawshaw; Royal Glamorgan Hospital — T. Crosby; Royal Marsden Hospital Sutton — D. Cunningham; Derriford Hospital — F. Daniel; Wansbeck General Hospital — W. Dobrowsky; Belfast City Hospital — M. Eatock; Royal Cornwall Hospital — R. Ellis; Lincoln County Hospital — J. Eremin; Yeovil District Hospital — S. Falk; Worcester Royal Hospital — D. Farrugia; Russells Hall Hospital — D. Ferry; Leeds General Infirmary — P. Finan; Countess of Chester Hospital — G. Foster; Royal Berkshire Hospital — A. Freebairn; Birmingham Heartlands Hospital — I. Geh; Oldchurch Hospital — A. Gershuny; Royal United Hospital Bath — E. Gilby; Good Hope Hospital — J. Glaholm; Mount Vernon Hospital — R. Glynne-Jones; Glan Clwyd Hospital — S. Gollins; Princess Elizabeth Hospital Guernsey — P. Gomes; Walsgrave Hospital — R. Grieve; Darlington Memorial Hospital — K. Gunning; Royal Hampshire County Hospital — V. Hall; Manor Hospital — A. Hartley; Poole Hospital — T. Hickish; Maidstone Hospital — M. Hill; Manchester Royal Infirmary — J. Hill; Royal Sussex County Hospital — N. Hodson; Weston Park Hospital — J. Hornbuckle; Salisbury District Hospital — T. Iveson; Bronglais General Hospital — D. Jackson; Huddersfield Royal Infirmary — J. Joffe; Warwick Hospital — D. Jones; Leicester Royal Infirmary — S. Khanna; Whittington Hospital — J. Ledermann; St. Thomas’ Hospital — M. Leslie; Royal Bolton Hospital — E. Levine; St. George’s Hospital — F. Lofts; Leighton Hospital — J. Logue; Diana Princess of Wales Hospital — P. Mack; Luton & Dunstable Hospital — A. Makris; Whiston Hospital — E. Marshall; Velindre Hospital — T. Maughan; Trafford General Hospital — F. Mazarelo; Peterborough District Hospital — K. McAdam; Eastbourne District General Hospital — F. McKinna; Royal Free Hospital — T. Meyer; Worthing Hospital — T. Miles; Castle Hill Hospital — J. Monson; West Suffolk Hospital — M. Moody; Ninewells Hospital — A. Munro; North Devon District Hospital — M. Napier; Cumberland Infirmary — J. Nicoll; Northwick Park Hospital — J. Northover; St. Mary’s Hospital Portsmouth — A. O’Callaghan; Milton Keynes General Hospital — R. O’Hara; Dorset County Hospital — R. Osbourne; Sunderland Royal Hospital — I. Pedley; Charing Cross Hospital — R.H. Phillips; Western General Hospital — H. Phillips; Epsom General Hospital — M. Raja; Pennine Acute Hospitals National Health Service Trust — A. Rate; City Hospital — D. Rea; Southern General Hospital — G. Robertson; Southend Hospital — A. Robinson; Hope Hospital — M. Saunders; York District Hospital — D. Sebag-Montefiore; Cookridge Hospital — M. Seymour; Ipswich Hospital National Health Service Trust — K. Sherwin; Tameside General Hospital — K. Siddiqui; University Hospital Aintree — D. Smith; Pontefract General Infirmary — M. Snee; King George Hospital — S. Snooks; Alexandra Hospital — S. Sothi; Scunthorpe General Hospital — T. Sreenivasan; Noble’s Hospital — S. Stock; Ysbyty Gwynedd — N. Stuart; Southport & Formby District General Hospital — A. Sun-Myint; University Hospital Hartlepool — M. Tabaq chali; University Hospital of North Tees — M. Tabaqchali; Broomfield Hospital — S. Tahir; Hinchingbrooke Hospital — L.T. Tan; Bedford General Hospital — R. Thomas; Halton Hospital — M. Tighe; Taunton & Somerset Hospital — M. Tighe; Weston General Hospital — M. Tomlinson; Royal Surrey County Hospital — C. Topham; Friarage Hospital — J.C.M. Van der Voet; James Cook University Hospital — N. Wadd; Stoke Mandeville Hospital — N. Warner; Hammersmith Hospital — H. Wasan; Royal Liverpool University Hospital — A. Watson; Churchill Hospital — A. Weaver; St. Bartholomew’s Hospital — P. Wells; Stepping Hill Hospital — P. Wilkinson; Withington Hospital — M. Wilson; Royal Albert Edward Infirmary — G. Wilson; George Eliot Hospital — J. Worlding; Hairmyres Hospital — H. Yosef; Queen Elizabeth Hospital London — C.-Y. Yui.

368 n engl j med 357;4 www.nejm.org july 26, 2007 Cardiovascular Toxicity of Rofecoxib in Colorectal Cancer

References

1. Midgley RS, Kerr DJ. ABC of colorec- 9. Thun MJ, Namboodiri MM, Calle EE, 16. Lagakos SW. Time-to-event analyses tal cancer: adjuvant therapy. BMJ 2000; Flanders DW, Heath CW Jr. Aspirin use and for long-term treatments — the APPROVe 321:1208-11. risk of fatal cancer. Cancer Res 1993;53: trial. N Engl J Med 2006;355:113-7. 2. QUASAR Collaborative Group. Com- 1322-7. 17. Johnsen SP, Larsson H, Tarone RE, et parison of fluorouracil with additional 10. Langman MJS, Cheng KK, Gilman al. Risk of hospitalization for myocardial levamisole, higher-dose folinic acid, or EA, Lancashire RJ. Effect of anti-inflam- infarction among users of rofecoxib, cele- both, as adjuvant chemotherapy for colo matory drugs on overall risk of common coxib, and other NSAIDs: a population- rectal cancer: a randomised trial. Lancet cancer: case-control study in general prac- based case-control study. Arch Intern Med 2000;355:1588-96. tice research database. BMJ 2000;320: 2005;165:978-84. 3. QUASAR Collaborative. QUASAR: 1642-6. 18. Mamdani M, Rochon P, Juurlink DN, a randomized study of adjuvant chemo- 11. Langman MJ, Jensen DM, Watson DJ, et al. Effect of selective cyclooxygenase 2 therapy (CT) vs observation including 3238 et al. Adverse upper gastrointestinal ef- inhibitors and naproxen on short-term risk colorectal cancer patients. J Clin Oncol fects of rofecoxib compared with NSAIDs. of acute myocardial infarction in the el- 2004;22:Suppl 14:3501a. abstract. JAMA 1999;282:1929-33. derly. Arch Intern Med 2003;163:481-6. 4. Sargent DJ, Wieand HS, Haller DG, et 12. Bresalier RS, Sandler RS, Quan H, et 19. Andersohn F, Suissa S, Garbe E. Use al. Disease-free survival versus overall sur- al. Cardiovascular events associated with of first- and second-generation cyclooxy- vival as a primary end point of adjuvant rofecoxib in a colorectal adenoma chemo- genase-2-selective drugs and risk of acute colon cancer studies: individual patient prevention trial. N Engl J Med 2005;352: myocardial infarction. Circulation 2006; data from 20,898 patients on 18 random 1092-102. [Erratum, N Engl J Med 2006; 113:1950-7. ised trials. J Clin Oncol 2005;23:8664-70. 355:221.] 20. McGettigan P, Henry D. Cardiovascu- 5. Oshima M, Murai N, Kargman S, et al. 13. Solomon SD, McMurray JJV, Pfeffer lar risk and inhibition of cyclooxygenase: Chemoprevention of intestinal polyposis MA, et al. Cardiovascular risk associated a systematic review of observational stud- in the Apcdelta716 mouse by rofecoxib, with celecoxib in a clinical trial for colo ies of selective and nonselective inhibi- a specific cycooxygenase-2 inhibitor. Can- rectal adenoma prevention. N Engl J Med tors of cyclooxygenase 2. JAMA 2006;296: cer Res 2001;61:1733-40. 2005;352:1071-80. 1633-44. 6. Sheehan KM, Sheahan K, O’Donoghue 14. Kearney PM, Baigent C, Goodwin J, 21. Public CHMP assessment report for DP, et al. The relationship between cyclo- Halla H, Emberson JR, Patrono C. Do se- medicinal products containing non-selec- oxygenase-2 expression and colorectal lective cyclo-oxygenase-2 inhibitors and tive non-steroidal anti-inflammatory drugs cancer. JAMA 1999;282:1254-7. [Erratum, traditional non-steroidal anti-inflamma- (NSAIDs). London: European Medicines JAMA 2000;283:1427.] tory drugs increase the risk of athero- Agency, November 7, 2006. (Accessed July 7. Fenwick SW, Toogood GJ, Lodge PA, thrombosis? Meta-analysis of randomised 2, 2007, at http://www.emea.europa.eu/ Hull MA. The effect of the selective cyclo- trials. BMJ 2006;332:1302-5. pdfs/human/opiniongen/44213006en. oxygenase-2 inhibitor rofecoxib on human 15. Antiplatelet Trialists’ Collaboration. pdf.) colorectal cancer liver metastases. Gastro- Collaborative overview of randomised 22. Moussa O, Yordy JS, Abol-Enein H, et enterology 2003;125:716-29. trials of antiplatelet therapy. I. Preven- al. Prognostic and functional significance 8. Giovannucci E, Rimm EG, Stampfer tion of death, myocardial infarction, and of thromboxane synthase gene overexpres- MJ, Colditz GA, Ascherio A, Willett WC. stroke by prolonged antiplatelet therapy sion in invasive bladder cancer. Cancer Aspirin use and the risk of colorectal can- in various categories of patients. BMJ Res 2005;65:11581-7. cer and adenoma in male health profes- 1994;308:81-106. [Erratum, BMJ 1994;308: Copyright © 2007 Massachusetts Medical Society. sionals. Ann Intern Med 1994;121:241-6. 1540.]

electronic access to the journal’s cumulative index At the Journal’s site on the World Wide Web (www.nejm.org), you can search an index of all articles published since January 1975 (abstracts 1975–1992, full text 1993–present). You can search by author, key word, title, type of article, and date. The results will include the citations for the articles plus links to the full text of articles published since 1993. For nonsubscribers, time-limited access to single articles and 24-hour site access can also be ordered for a fee through the Internet (www.nejm.org).

n engl j med 357;4 www.nejm.org july 26, 2007 369 T h e new england journal o f medicine

special article

The Spread of Obesity in a Large Social Network over 32 Years

Nicholas A. Christakis, M.D., Ph.D., M.P.H., and James H. Fowler, Ph.D.

ABSTRACT

Background From the Department of Health Care The prevalence of obesity has increased substantially over the past 30 years. We Policy, Harvard Medical School, Boston performed a quantitative analysis of the nature and extent of the person-to-person (N.A.C.); the Department of Medicine, Mt. Auburn Hospital, Cambridge, MA (N.A.C.); spread of obesity as a possible factor contributing to the obesity epidemic. the Department of Sociology, Harvard Uni- versity, Cambridge, MA (N.A.C.); and the Methods Department of Political Science, Univer- sity of California, San Diego, San Diego We evaluated a densely interconnected social network of 12,067 people assessed (J.H.F.). Address reprint requests to Dr. repeatedly from 1971 to 2003 as part of the Framingham Heart Study. The body- Christakis at the Department of Health mass index was available for all subjects. We used longitudinal statistical models to Care Policy, Harvard Medical School, 180 Longwood Ave., Boston, MA 02115, or at examine whether weight gain in one person was associated with weight gain in his [email protected]. or her friends, siblings, spouse, and neighbors.

N Engl J Med 2007;357:370-9. Results Copyright © 2007 Massachusetts Medical Society. Discernible clusters of obese persons (body-mass index [the weight in kilograms divided by the square of the height in meters], ≥30) were present in the network at all time points, and the clusters extended to three degrees of separation. These clusters did not appear to be solely attributable to the selective formation of social ties among obese persons. A person’s chances of becoming obese increased by 57% (95% confidence interval [CI], 6 to 123) if he or she had a friend who became obese in a given interval. Among pairs of adult siblings, if one sibling became obese, the chance that the other would become obese increased by 40% (95% CI, 21 to 60). If one spouse became obese, the likelihood that the other spouse would become obese increased by 37% (95% CI, 7 to 73). These effects were not seen among neighbors in the immediate geographic location. Persons of the same sex had relatively greater influence on each other than those of the opposite sex. The spread of smoking cessation did not account for the spread of obesity in the network.

Conclusions Network phenomena appear to be relevant to the biologic and behavioral trait of obesity, and obesity appears to spread through social ties. These findings have implications for clinical and public health interventions.

370 n engl j med 357;4 www.nejm.org july 26, 2007 The Spread of Obesity in a Large Social Network Over 32 Years

he prevalence of obesity has in- inal cohort.16 The Framingham Offspring Study creased from 23% to 31% over the recent began in 1971, when most of the children of past in the United States, and 66% of adults members of the original cohort and their spouses T 1,2 17 are overweight. Proposed explanations for the were enrolled in the offspring cohort. There has obesity epidemic include societal changes that been almost no loss to follow-up other than death promote both inactivity and food consumption.3 in this cohort of 5124 people; only 10 people left The fact that the increase in obesity during this the study. In 2002, the third-generation cohort, period cannot be explained by genetics4,5 and consisting of 4095 children of the offspring co- has occurred among all socioeconomic groups1 hort, was initiated. All participants undergo phys- provides support for a broad set of social and ical examinations (including measurements of environmental explanations. Since diverse phe- height and weight) and complete written ques- nomena can spread within social networks,6-10 tionnaires at regular intervals. we conducted a study to determine whether obesity might also spread from person to person, Network Ascertainment possibly contributing to the epidemic, and if so, For our study, we used the offspring cohort as how the spread might occur. the source of 5124 key subjects, or “egos,” as they Whereas obesity has been stigmatized in the are called in social-network analysis. Any persons past, attitudes may be changing.11,12 To the extent to whom the egos are linked — in any of the that obesity is a product of voluntary choices or Framingham Heart Study cohorts — can, how- behaviors, the fact that people are embedded in ever, serve as “alters.” Overall, 12,067 living egos social networks and are influenced by the evident and alters were connected at some point during appearance and behaviors of those around them the study period (1971 to 2003). suggests that weight gain in one person might To create the network data set, we entered in- influence weight gain in others. Having obese formation about the offspring cohort into a com- social contacts might change a person’s tolerance puter. This information was derived from archived, for being obese or might influence his or her handwritten administrative tracking sheets that adoption of specific behaviors (e.g., smoking, eat- had been used since 1971 to identify people close ing, and exercising). In addition to such strictly social mechanisms, it is plausible that physiolog- Glossary ical imitation might occur; areas of the brain that correspond to actions such as eating food Ego: The person whose behavior is being analyzed. may be stimulated if these actions are observed Alter: A person connected to the ego who may influence the behavior of the ego. 13 in others. Even infectious causes of obesity are Node: An object that may or may not be connected to other objects in a net- conceivable.14,15 work. In this study, nodes represent people in the Framingham Heart We evaluated a network of 12,067 people who Study cohorts. underwent repeated measurements over a period Tie: A connection between two nodes that can be either one-way (directed) of 32 years. We examined several aspects of the or two-way (bilateral). In this study, all family ties (e.g., between siblings and parents) as well as marital ties are bilateral, but friendship ties are di- spread of obesity, including the existence of clus- rectional since a subject may identify someone as a friend who does not ters of obese persons within the network, the necessarily identify that person as a friend in return. association between one person’s weight gain and Degree of separation: The social distance between two people as measured weight gain among his or her social contacts, the by the smallest number of intermediaries between an ego and other dependence of this association on the nature of members of the network. For a given ego, alters are degree 1, since they are directly connected to the ego. Nodes that are connected to the alters the social ties (e.g., ties between friends of differ- but not to the ego are degree 2 (alters’ alters). Nodes that are connected ent kinds, siblings, spouses, and neighbors), and to the alters’ alters but not to the ego are degree 3, and so on. the influence of sex, smoking behavior, and geo- Homophily: The tendency for people to choose relationships with people graphic distance between the domiciles of per- who have similar attributes. sons in the social network. Induction: The spread of a behavior or trait from one person to another. Cluster: A group of nodes, each of which is connected to at least one other Methods node in the group. Connected component: Part of a social network in which all persons have Source Data a social tie to at least one other person and no person is connected to a The Framingham Heart Study was initiated in member of any other component of the network. 1948, when 5209 people were enrolled in the orig-

n engl j med 357;4 www.nejm.org july 26, 2007 371 T h e new england journal o f medicine

to the study participants to facilitate follow-up. lowed by alter-perceived friendships. Our reason- These sheets contain valuable, previously unused ing was that the person making the identification social-network information because they system- esteems the other person and may wish to emu- atically and comprehensively identify relatives and late him or her. friends named by the ego. The tracking sheets We included only persons older than 21 years provide complete information about all first-order of age at any observation point and subsequently. relatives (parents, spouses, siblings, and children), At the inception of the study, 53% of the egos whether they are alive or dead, and at least one were women, the mean age of the egos was 38 “close friend” at each of seven examinations be- years (range, 21 to 70), and their mean educa- tween 1971 and 2003. The examinations took tional level was 13.6 years (range, no education place during 3-year periods centered in 1973, 1981, to ≥17 years of education). 1985, 1989, 1992, 1997, and 1999. Detailed home The study data are available from the Framing- addresses were also recorded at each time point; ham Heart Study. The study was approved by the we used this information to calculate the geo- institutional review board at Harvard Medical graphic distance between people. School; all subjects provided written informed Many of the named alters on these sheets also consent. were members of Framingham Heart Study cohorts. This newly computerized database thus Statistical Analysis identifies the network links among participants We graphed the network with the use of the at each examination and longitudinally from one Kamada–Kawai18 algorithm in Pajek software.19 examination to the next. As a person’s family We generated videos of the network by means of changed because of birth, death, marriage, or the Social Network Image Animator (known as divorce, and as contacts changed because of resi- SoNIA).20 We examined whether our data con- dential moves or new friendships, this informa- formed to theoretical network models such as the tion was recorded. Furthermore, dates of birth small-world,10 scale-free,21 and hierarchical types22 and death were available from separate Framing- (see the Supplementary Appendix, available with ham Heart Study files. the full text of this article at www.nejm.org). Overall, there were 38,611 observed social and We defined obesity as a body-mass index (the family ties to the 5124 egos, yielding an average weight in kilograms divided by the square of of 7.5 ties per ego (not including neighbors). For the height in meters) of 30 or more. Analyses in example, 83% of the spouses of egos were direct which the body-mass index was a continuous var ly and repeatedly observed at the time of exami- iable did not yield different results. nation, and 87% percent of egos with siblings We considered three explanations for the clus- had at least one sibling in the network. For 10% tering of obese people. First, egos might choose of the egos, an immediate neighbor also partici- to associate with like alters (“homophily”).21,23,24 pated in the study; more expansive definitions of Second, egos and alters might share attributes neighbors yielded similar results. or jointly experience unobserved contemporane- A total of 45% of the 5124 egos were con- ous events that cause their weight to vary at the nected through friendship to another person in same time (confounding). Third, alters might the network. There were 3604 unique, observed exert social influence or peer effects on egos friendships, for an average of 0.7 friendship tie per (“induction”). Distinguishing the interpersonal ego. Because friendship identifications are direc- induction of obesity from homophily requires tional, we studied three different kinds of friend- dynamic, longitudinal network information about ships: an “ego-perceived friendship,” in which an the emergence of ties between people (“nodes”) ego identifies an alter as a friend; an “alter-per- in a network and also about the attributes of ceived friendship,” in which an alter identifies nodes (i.e., repeated measures of the body-mass an ego as a friend; and a “mutual friendship,” in index).25 which the identification is reciprocal. We hypoth- The basic statistical analysis involved the spec esized that a friend’s social influence on an ego ification of longitudinal logistic-regression mod- would be affected by the type of friendship, with els in which the ego’s obesity status at any given the strongest effects occurring in mutual friend- examination or time point (t + 1) was a function of ships, followed by ego-perceived friendships, fol- various attributes, such as the ego’s age, sex, and

372 n engl j med 357;4 www.nejm.org july 26, 2007 The Spread of Obesity in a Large Social Network Over 32 Years educational level; the ego’s obesity status at the ing both their weights. We estimated these mod- previous time point (t); and most pertinent, the els in varied ego–alter pair types. alter’s obesity status at times t and t + 1.25 We To evaluate the possibility that omitted vari- used generalized estimating equations to account ables or unobserved events might explain the as- for multiple observations of the same ego across sociations, we examined how the type or direc- examinations and across ego–alter pairs.26 We tion of the social relationship between the ego assumed an independent working correlation and the alter affected the association between the structure for the clusters.26,27 ego’s obesity and the alter’s obesity. For example, The use of a time-lagged dependent variable if unobserved factors drove the association be- (lagged to the previous examination) eliminated tween the ego’s obesity and the alter’s obesity, serial correlation in the errors (evaluated with a then the directionality of friendship should not Lagrange multiplier test28) and also substantial have been relevant. ly controlled for the ego’s genetic endowment and We evaluated the role of a possible spread in any intrinsic, stable predisposition to obesity. The smoking-cessation behavior as a contributor to use of a lagged independent variable for an alter’s the spread of obesity by adding variables for the weight status controlled for homophily.25 The smoking status of egos and alters at times t and key variable of interest was an alter’s obesity at t + 1 to the foregoing models. We also analyzed time t + 1. A significant coefficient for this vari- the role of geographic distance between egos able would suggest either that an alter’s weight and alters by adding such a variable. affected an ego’s weight or that an ego and an We calculated 95% confidence intervals by sim alter experienced contemporaneous events affect- ulating the first difference in the alter’s contem-

Figure 1. Largest Connected Subcomponent of the Social Network in the Framingham Heart Study in the Year 2000. Each circle (node) represents one person in the data set. There are 2200 persons in this subcomponent of the social network. Circles with red borders denote women, and circles with blue borders denote men. The size of each circle is proportional to the person’s body-mass index. The interior color of the circles indicates the person’s obesity status: yellow denotes an obese person (body-mass index, ≥30) and green denotes a nonobese person. The colors of the ties between the nodes indicate the relationship between them: purple denotes a friendship or marital tie and orange denotes a familial tie.

n engl j med 357;4 www.nejm.org july 26, 2007 373 T h e new england journal o f medicine

A 1975 B 1980

C 1985 D 1990

E 1995 F 2000

Figure 2. Part of the Social Network from the Framingham Heart Study with Information about Body-Mass Index According to Year. Each circle (node) represents one person in the data set. Circles with red borders denote women, and circles with blue borders denote men. The size of each circle is proportional to the person’s body-mass index. The interior color of the circles indicates the person’s obesity status: RETAKE 1st ICM AUTHOR: Christaki yellow denotes an obese person (body-mass index, ≥30) and green denotes a nonobese person.2nd The colors of the ties between the circles REG F FIGURE: 2 of 4 indicate the relationship between them: purple denotes a friendship or a marital tie and orange3rd denotes a familial tie. The disappearance of a circle from one year to another indicatesCASE the person’s death, and the disappearanceRevised of a tie between the circles indicates that the re- EMail Line 4-C SIZE lationship between the two persons no longer exists.ARTIST: The largestts connected subcomponent of the whole network and the change in obe- H/T H/T 33p9 sity over the 32-year study period are shown Enonin an animation that is availableCombo with the full text of this article at www.nejm.org. AUTHOR, PLEASE NOTE: Figure has been redrawn and type has been reset. 374 n englPlease j med check 357;4 carefully. www.nejm.org july 26, 2007

JOB: 356xx ISSUE: 07-26-07 The Spread of Obesity in a Large Social Network Over 32 Years

Figure 3. Effect of Social and Geographic Distance from A Obese Alters on the Probability of an Ego’s Obesity in 100 the Social Network of the Framingham Heart Study. Examination 1 Examination 2 Panel A shows the mean effect of an ego’s social prox- 80 Examination 3 imity to an obese alter; this effect is derived by compar- Examination 4 ing the conditional probability of obesity in the observed Examination 5 network with the probability of obesity in identical net- 60 Examination 6 works (with topology preserved) in which the same Examination 7 number of obese persons is randomly distributed. The social distance between the alter and the ego is repre- 40 sented by degrees of separation (1 denotes one degree of separation from the ego, 2 denotes two degrees of separation from the ego, and so forth). The examina- 20 tion took place at seven time points. Panel B shows the mean effect of an ego’s geographic proximity to an obese in an Ego if Alter becomes Obese (%) 0

alter. We ranked all geographic distances (derived from Relative Increase in Probability of Obesity geocoding) between the homes of directly connected egos and alters (i.e., those pairs at one degree of sepa- 1 2 3 4 5 6 ration) and created six groups of equal size. This figure Social Distance between the Ego and the Alter (degree of separation) shows the effects observed for the six mileage groups (based on their average distance): 1 denotes 0 miles B (i.e., closest to the alter’s home), 2 denotes 0.26 mile, 100 3 denotes 1.5 miles, 4 denotes 3.4 miles, 5 denotes 9.3 miles, and 6 denotes 471 miles (i.e., farthest from the alter’s home). There is no trend in geographic dis- 80 tance. I bars for both panels show 95% confidence intervals based on 1000 simulations. To convert miles to kilometers, multiply by 1.6. 60 poraneous obesity (changing from 0 to 1), using 40 1000 randomly drawn sets of estimates from the coefficient covariance matrix and assuming mean 20 values for all other variables.29 All tests were two-tailed. The sensitivity of the results was as-

in an Ego if Alter becomes Obese (%) 0

sessed with multiple additional analyses (see the Relative Increase in Probability of Obesity Supplementary Appendix). 1 2 3 4 5 6 Results Geographic Distance between the Ego and the Alter (mileage group)

Figure 1 depicts the largest connected subcom- RETAKE 1st ICM AUTHOR: Christakis ponent of the social network in the year 2000. lence of obesity as the observed network, but with 2nd REG F FIGURE: 3 of 4 This network is sufficiently dense to obscure the incidence of obesity randomly distributed 3rd CASE Revised much of the underlying structure, although re- among the nodesEMail (in what we callLine “random4-C body- SIZE ARTIST: ts H/T H/T gions of the network with clusters of obese or mass–index networks”).Enon If clustering is occur- 22p3 Combo nonobese persons can be seen. Figure 2 illus- ring, then the probability that an alter will be AUTHOR, PLEASE NOTE: trates the spread of obesity between adjoining obese, given thatFigure an hasego been is redrawnknown and to type be has obese, been reset. nodes in a part of the network over time. A video should be higher in the observedPlease check network carefully. than

(available with the full text of this article at www. in the randomJOB: body-mass–index35704 networks. ISSUE:What 07-26-07 nejm.org) depicts the evolution of the largest we call the “reach” of the clusters is the point, in component of the network and shows the prog- terms of an alter’s degree of separation from any ress of the obesity epidemic over the 32-year study given ego, at which the probability of an alter’s period. obesity is no longer related to whether the ego Figure 3A characterizes clusters within the is obese. In all of the examinations (from 1971 entire network more formally. To quantify these through 2003), the risk of obesity among alters clusters, we compared the whole observed net- who were connected to an obese ego (at one de- work with simulated networks with the same gree of separation) was about 45% higher in the network topology and the same overall preva- observed network than in a random network. The

n engl j med 357;4 www.nejm.org july 26, 2007 375 T h e new england journal o f medicine

unobserved contemporaneous events by separate Alter Type ly analyzing models of subgroups of the data in- Ego-perceived friend Mutual friend volving various ego–alter pairings. Figure 4 sum- Alter-perceived friend marizes the associations. Same-sex friend If an ego stated that an alter was his or her Opposite-sex friend friend, the ego’s chances of becoming obese ap- Spouse peared to increase by 57% (95% confidence in- Sibling Same-sex sibling terval [CI], 6 to 123) if the alter became obese. Opposite-sex sibling However, the type of friendship appeared to be Immediate neighbor important. Between mutual friends, the ego’s risk 0 100 200 300 of obesity increased by 171% (95% CI, 59 to 326) Increase in Risk of Obesity in Ego (%) if an alter became obese. In contrast, there was no statistically meaningful relationship when the Figure 4. Probability That an Ego Will Become Obese According to the Type friendship was perceived by the alter but not the of Relationship with an Alter Who May Become Obese in Several Subgroups ego (P = 0.70). Thus, influence in friendship ties of the Social Network of the Framingham Heart Study.RETAKE 1st ICM AUTHOR: Christakis 2nd appeared to be directional. The closenessREG of friendshipF FIGURE: is4 relevantof 4 to the spread of obesity. Persons in 3rd The sex of the ego and alter also appeared to closer, mutual CASEfriendships have more of an effect on each other than persons Revised be important. When the sample was restricted to in other types EMailof friendships. The dependentLine variable4-C in eachSIZE model is the obesity of the ego. IndependentARTIST: ts variablesH/T includeH/T a time-lagged measure- Enon 22p3 same-sex friendships (87% of the total), the prob- ment of the ego’s obesity; the obesity Comboof the alter; a time-lagged measure- ability of obesity in an ego increased by 71% ment of the alter’s obesity; AUTHOR,the ego’s PLEASEage, sex, NOTE: and level of education; and (95% CI, 13 to 145) if the alter became obese. indicator variablesFigure (fixed has effects) been redrawn for each and examination. type has been reset.Full models and Please check carefully. equations are available in the Supplementary Appendix. Mean effect sizes For friends of the opposite sex, however, there and 95% confidence intervals were calculated by simulating the first differ- was no significant association (P = 0.64). Among JOB: 35704 ISSUE: 07-26-07 ence in the contemporaneous obesity of the alter (changing from 0 to 1) friends of the same sex, a man had a 100% (95% with the use of 1000 randomly drawn sets of estimates from the coefficient CI, 26 to 197) increase in the chance of becom- covariance matrix and with all other variables held at their mean values. ing obese if his male friend became obese, where as the female-to-female spread of obesity was risk of obesity was also about 20% higher for al- not significant (38% increased chance; 95% CI, ters’ alters (at two degrees of separation) and −39 to 161). about 10% higher for alters’ alters’ alters (at three Among pairs of adult siblings, one sibling’s degrees of separation). By the fourth degree of chance of becoming obese increased by 40% separation, there was no excess relationship be- (95% CI, 21 to 60) if the other sibling became tween an ego’s obesity and the alter’s obesity. obese. This phenomenon appeared to be more Hence, the reach of the obesity clusters was three marked among siblings of the same sex (55%; degrees. 95% CI, 26 to 88) than among siblings of the Figure 3B indicates that the effect of geo- opposite sex (27%; 95% CI, 3 to 54), although the graphic distance is different from the effect of difference was not significant (P = 0.16). Among social distance. Whereas increasing social dis- brothers, an ego’s chance of becoming obese in- tance appeared to decrease the effect of an alter creased by 44% (95% CI, 6 to 91) if his alter be- on an ego, increasing geographic distance did not. came obese, and among sisters, an ego’s chance The obesity of the most geographically distant of becoming obese increased by 67% (95% CI, alters correlated as strongly with an ego’s obesity 27 to 114) if her alter became obese. Obesity in as did the obesity of the geographically closest a sibling of the opposite sex did not affect the alters. These results suggest that social distance chance that the other sibling would become plays a stronger role than geographic distance in obese. the spread of behaviors or norms associated with Among married couples, when an alter became obesity. obese, the spouse was 37% more likely (95% CI, We evaluated the extent of interpersonal asso 7 to 73) to become obese. Husbands and wives ciation in obesity with the use of regression appeared to affect each other similarly (44% and analysis. Our models account for homophily by 37%, respectively). Finally, we observed no effect including a time-lagged measurement of the on the risk that an ego would become obese if alter’s obesity. We evaluated the possible role of an immediate neighbor became obese.

376 n engl j med 357;4 www.nejm.org july 26, 2007 The Spread of Obesity in a Large Social Network Over 32 Years

We also investigated two factors that might any such exposures should have an equally strong mediate or modify the effect of an alter’s weight influence regardless of the directionality of friend gain: his or her smoking behavior and geograph- ship. This observation also points to the specifi- ic distance from the ego (see the Supplementary cally social nature of these associations, since the Appendix). We added measures of smoking be- asymmetry in the process may arise from the fact havior for the ego and the alter at both the cur- that the person who identifies another person as rent and previous examinations. The coefficient a friend esteems the other person. for the effect of the alter’s obesity was virtually Finally, pairs of friends and siblings of the unchanged; smoking behavior does not appear to same sex appeared to have more influence on be instrumental to the spread of obesity. Models the weight gain of each other than did pairs of that included the geographic distance between friends and siblings of the opposite sex. This the ego and alter corroborated the result shown finding also provides support for the social na- in Figure 3B: geographic distance did not modify ture of any induction of obesity, since it seems the intensity of the effect of the alter’s obesity on likely that people are influenced more by those the ego. they resemble than by those they do not. Con- versely, spouses, who share much of their phys Discussion ical environment, may not affect each other’s weight gain as much as mutual friends do; in the Our study suggests that obesity may spread in case of spouses, the opposite-sex effects and social networks in a quantifiable and discernable friendship effects may counteract each another. pattern that depends on the nature of social ties. Obesity in alters might influence obesity in Moreover, social distance appears to be more im- egos by diverse psychosocial means, such as chang portant than geographic distance within these ing the ego’s norms about the acceptability of networks. Although connected persons might being overweight, more directly influencing the share an exposure to common environmental fac- ego’s behaviors (e.g., affecting food consump- tors, the experience of simultaneous events, or tion), or both. Other mechanisms are also pos- other common features (e.g., genes) that cause sible. Unfortunately, our data do not permit a them to gain or lose weight simultaneously, our detailed examination. However, some insight into observations suggest an important role for a pro- possible mechanisms can be gained from a con- cess involving the induction and person-to-person sideration of the roles of smoking and geograph- spread of obesity. ic distance in obesity. The tendency of persons Our findings that the weight gain of immedi- to gain weight when they stop smoking is well ate neighbors did not affect the chance of weight known,31 and the coincidence of a decrease in gain in egos and that geographic distance did not smoking and an increase in obesity in the over- modify the effect for other types of alters (e.g., all population has been noted.32 However, the friends or siblings) helps rule out common expo- present study indicates that regardless of wheth- sure to local environmental factors as an explana- er smoking cessation causes weight gain in indi- tion for our observations. Our models also con- vidual persons, and regardless of whether smok- trolled for an ego’s previous weight status; this ing-initiation or smoking-cessation behavior itself helps to account for sources of confounding that spreads from person to person,33 any spread in are stable over time (e.g., childhood experiences smoking behavior is not a significant factor in or genetic endowment).30 In addition, the control the spread of obesity. This finding indicates that in our models for an alter’s previous weight sta- smoking behavior does not mediate the interper- tus accounts for a possible tendency of obese sonal effect in the spread of obesity. However, in people to form ties among themselves. Finally, addition, it suggests that the psychosocial mech- the findings regarding the directional nature of anisms of the spread of obesity may rely less on the effects of friendships are especially important behavioral imitation than on a change in an ego’s with regard to the interpersonal induction of general perception of the social norms regarding obesity because they suggest that friends do not the acceptability of obesity. This point is further simultaneously become obese as a result of con- reinforced by the relevance of the directionality temporaneous exposures to unobserved factors. of friendship. If the friends did become obese at the same time, Hence, an ego may observe that an alter gains

n engl j med 357;4 www.nejm.org july 26, 2007 377 T h e new england journal o f medicine

weight and then may accept weight gain in him- that provide peer support — that is, that modify self or herself. This weight gain in an ego might, the person’s social network — are more success- in turn, be determined by various behaviors that ful than those that do not.34,35,38,39 People are an ego chooses to evince, and these behaviors connected, and so their health is connected.40,41 need not be the same behaviors that an alter Consequently, medical and public health interven evinces. The observation that geographic distance tions might be more cost-effective than initially does not modify the effect of an alter’s obesity supposed, since health improvements in one per- also provides support for the concept that norms son might spread to others.42 The observation may be particularly relevant here. Behavioral ef- that people are embedded in social networks sug- fects might rely more on the frequency of contact gests that both bad and good behaviors might (which one might reasonably expect to be attenu- spread over a range of social ties. This highlights ated with distance), whereas norms might not. the necessity of approaching obesity not only as The spread of obesity in social networks ap- a clinical problem but also as a public health pears to be a factor in the obesity epidemic. Yet problem. the relevance of social influence also suggests Supported by a grant from the National Institutes of Health (NIH R-01 AG24448-01). that it may be possible to harness this same force No potential conflict of interest relevant to this article was to slow the spread of obesity. Network phenom- reported. ena might be exploited to spread positive health We thank Laurie Meneades, Rebecca Joyce, Molly Collins, Marian Bellwood, and Karen Mutalik for the expert assistance 34-36 behaviors, in part because people’s percep- required to build the data set, and Emelia Benjamin, Virginia tions of their own risk of illness may depend on Chang, Scott Desposato, Felix Elwert, Peter Marsden, Joanne 37 Murabito, James O’Malley, Barbara McNeil, Mark Pachucki, Mark the people around them. Smoking- and alcohol- Pletcher, Mason A. Porter, Darren Schreiber, Richard Suzman, cessation programs and weight-loss interventions and Alan Zaslavsky for helpful comments.

References 1. Chang VW, Lauderdale DS. Income 11. Sobal J. The size acceptance move- 19. Batagelj V, Mrvar A. Pajek — analysis disparities in body mass index and obe- ment and the social construction of body and visualization of large networks. In: sity in the United States, 1971-2002. Arch weight. In: Sobal J, Maurer D, eds. Weighty Junger M, Mutzel P, eds. Graph drawing Intern Med 2005;165:2122-8. issues: fatness and thinness as social software. Berlin: Springer, 2003:77-103. 2. Hedley AA, Ogden CL, Johnson CL, problems. New York: Aldine de Gruyter, 20. Moody J, McFarland DA, Bender- Carroll MD, Curtin LR, Flegal KM. Preva- 1999:231-49. deMoll S. Visualizing network dynamics. lence of overweight and obesity among US 12. Chang VW, Christakis NA. Medical Am J Sociol 2005;110:1206-41. children, adolescents, and adults, 1999- modeling of obesity: a transition from 21. Barabasi AL, Albert R. Emergence of 2002. JAMA 2004;291:2847-50. action to experience in a 20th century scaling in random networks. Science 1999; 3. Hill JO, Peters JC. Environmental con- American medical textbook. Sociol Health 286:509-12. tributions to the obesity epidemic. Science Illness 2002;24:151-77. 22. Ravasz E, Barabasi AL. Hierarchical 1998;280:1371-4. 13. Fogassi L, Ferrari PF, Gesierich B, Roz organization in complex networks. Phys 4. Stunkard AJ, Sorensen TI, Hanis C, zi S, Chersi F, Rizzolatti G. Parietal lobe: Rev E Stat Nonlin Soft matter Phys 2003; et al. An adoption study of human obesity. from action organization to intention 67:026112. N Engl J Med 1986;314:193-8. understanding. Science 2005;308:662-7. 23. McPherson M, Smith-Lovin L, Cook 5. Stunkard AJ, Harris JR, Pedersen NL, 14. Whigham LD, Israel BA, Atkinson RL. JM. Birds of a feather: homophily in social McClearn GE. The body-mass index of Adipogenic potential of multiple human networks. Ann Rev Sociol 2001;27:415-44. twins who have been reared apart. N Engl adenoviruses in vivo and in vitro in ani- 24. Sackett DL, Anderson GD, Milner R, J Med 1990;322:1483-7. mals. Am J Physiol Regul Integr Comp Feinleib M, Kannel WB. Concordance for 6. Newman MEJ. The structure and func- Physiol 2006;290:R190-R194. coronary risk factors among spouses. Cir- tion of complex networks. SIAM Review 15. Turnbaugh PJ, Ley RE, Mahowald culation 1975;52:589-95. 2003;45:167-256. MA, Magrini V, Mardis ER, Gordon JI. An 25. Carrington PJ, Scott J, Wasserman S. 7. Nowak MA, Sigmund K. Evolution obesity-associated gut microbiome with Models and methods in social network of indirect reciprocity. Nature 2005;437: increased capacity for energy harvest. Na- analysis. New York: Cambridge University 1291-8. ture 2006;444:1027-31. Press, 2005. 8. Bearman PS, Moody J, Stovel K. Chains 16. Dawber TR. The Framingham Study: 26. Liang K-Y, Zeger SL. Longitudinal data of affection: the structure of adolescent the epidemiology of atherosclerotic dis- analysis using generalized linear models. romantic and sexual networks. Am J Sociol ease. Cambridge, MA: Harvard University Biometrika 1986;73:13-22. 2004;110:44-91. Press, 1980. 27. Schildcrout JS, Heagerty PJ. Regression 9. Liljeros F, Edling CR, Nunes Amaral 17. Feinleib M, Kannel WB, Garrison RJ, analysis of longitudinal binary data with LA. Sexual networks: implications for the McNamara PM, Castelli WP. The Framing- time-dependent environmental covariates: transmission of sexually transmitted in- ham Offspring Study: design and prelimi- bias and efficiency. Biostatistics 2005;6: fections. Microbes Infect 2003;5:189-96. nary data. Prev Med 1975;4:518-25. 633-52. 10. Watts DJ, Strogatz SH. Collective dy- 18. Kamada T, Kawai S. An algorithm for 28. Beck N. Time-series–cross-section namics of ‘small-world’ networks. Nature drawing general undirected graphs. Infor- data: what have we learned in the past few 1998;393:440-2. mation Processing Letters 1989;31:7-15. years? Annu Rev Polit Sci 2001;4:271-93.

378 n engl j med 357;4 www.nejm.org july 26, 2007 The Spread of Obesity in a Large Social Network Over 32 Years

29. King G, Tomz M, Wittenberg J. Mak- The influence of peers on young adult sub- disease: their impact on perceived risk. ing the most of statistical analyses: improv stance use. Health Psychol 2002;21:349-57. Prev Med 2003;37:242-9. ing interpretation and presentation. Am J 34. Wing RR, Jeffery RW. Benefits of re- 38. McKnight AJ, McPherson K. Evalua- Polit Sci 2000;44:341-55. cruiting participants with friends and in- tion of peer intervention training for high 30. Atwood LD, Heard-Costa NL, Fox CS, creasing social support for weight loss school alcohol safety education. Accid Anal Jaquish CE, Cupples LA. Sex and age and maintenance. J Consult Clin Psychol Prev 1986;18:339-47. specific effects of chromosomal regions 1999;67:132-8. 39. Wechsler H, Moeykens B, Davenport A, linked to body mass index in the Fram- 35. Malchodi CS, Oncken C, Dornelas EA, Castillo S, Hansen J. The adverse impact ingham Study. BMC Genet 2006;7:7. Caramanica L, Gregonis E, Curry SL. The of heavy episodic drinkers on other college 31. Eisenberg D, Quinn BC. Estimating effects of peer counseling on smoking students. J Stud Alcohol 1995;56:628-34. the effect of smoking cessation on weight cessation and reduction. Obstet Gynecol 40. Christakis NA, Allison PD. Mortality gain: an instrumental variable approach. 2003;101:504-10. after the hospitalization of a spouse. Health Serv Res 2006;41:2255-66. 36. Bruckner H, Bearman PS. After the N Engl J Med 2006;354:719-30. 32. Philipson TJ, Posner RA. The long-run promise: the STD consequences of ado- 41. Granovetter MS. The strength of weak growth in obesity as a function of tech- lescent virginity pledges. J Adolesc Health ties. Am J Sociol 1973;78:1360-80. nological change. Perspect Biol Med 2003; 2005;36:271-8. 42. Christakis NA. Social networks and col 46:Suppl 3:S87-S107. 37. Montgomery GH, Erblich J, DiLorenzo lateral health effects. BMJ 2004;329:184-5. 33. Andrews JA, Tildesley E, Hops H, Li F. R, Bovbjerg DH. Family and friends with Copyright © 2007 Massachusetts Medical Society.

n engl j med 357;4 www.nejm.org july 26, 2007 379 T h e new england journal o f medicine

clinical practice

Skin and Soft-Tissue Infections Caused by Methicillin-Resistant Staphylococcus aureus

Robert S. Daum, M.D., C.M.

This Journal feature begins with a case vignette highlighting a common clinical problem. Evidence supporting various strategies is then presented, followed by a review of formal guidelines, when they exist. The article ends with the author’s clinical recommendations.

A 37-year-old man presents for the evaluation of localized swelling and tenderness of the left leg just below the knee. He suspects this lesion developed after a spider bite, although he did not see a spider. Examination of the leg reveals an area of erythema and warmth measuring approximately 5 by 7 cm. At the center of the lesion is a fluctuant area measuring approximately 2 by 2 cm, overlaid by a small area of necrotic skin. The man’s temperature is 38.3°C. The pulse rate is 115 beats per minute. The blood pressure is 116/78 mm Hg. How should this patient be evaluated and treated?

The Clinical Problem

From the Section of Infectious Diseases, Methicillin-resistant Staphylococcus aureus (MRSA) refers to isolates that are resistant Department of Pediatrics, University of to all currently available β-lactam antibiotics, including penicillins and cephalo- Chicago, Chicago. sporins.1 MRSA isolates were first recognized shortly after the introduction of N Engl J Med 2007;357:380-90. methicillin into clinical practice in the early 1960s. Their prevalence slowly in- Copyright © 2007 Massachusetts Medical Society. creased during the next three decades,2 although they remained confined almost exclusively to patients who frequented health care facilities; other persons at risk for MRSA colonization or infection included those in contact with a person who had an MRSA infection or with a history of illicit drug use. In the mid-1990s, MRSA infections began to be detected in the community in persons who did not have contact with the health care system.3 Molecular typing of isolates from these community-associated cases of MRSA infection has shown that they are largely caused by new MRSA strains.4 As compared with health care–associated MRSA isolates, community-associated MRSA isolates are usually susceptible to clindamycin, and they are less often multiply resistant to other non–β-lactam antibiotics.5 Other distinguishing features of community-associated MRSA isolates include a high prevalence of genes encoding the two-component Panton–Valentine leukocidin6; this exotoxin is associated with necrosis of the skin, severe necrotizing pneumonia,7 and abscess formation, although its role in the pathogenesis of community-associated MRSA infections remains controversial.8,9 In addition, small DNA cassettes mediating methicillin resis- tance4,10,11 have been detected in community-associated MRSA isolates of multiple genetic backgrounds, suggesting easy transfer. These cassettes differ from those in hospital-associated MRSA strains, which are larger and presumably less mobile. The classification of circulating community-associated MRSA strains according to pulsed-field electrophoretic patterns12 has revealed global, geographic variations. In most areas of the United States, a community-associated MRSA genotype called USA300 has emerged as the major circulating strain and has even emerged as a nosocomial strain in many areas.13

380 n engl j med 357;4 www.nejm.org july 26, 2007 clinical practice

Numerous reports have suggested the easy transmission of these new community-associated MRSA isolates in settings where people are in close contact. These settings include households,14 day-care centers,15,16 and military installations.17 These isolates also may be spread among prison and jail detainees18 and athletes.19 Before the 1990s, such evidence of contagion among otherwise healthy members of the community was documented infrequently. Other groups reported to be at increased risk for community-associated MRSA infection include Native Americans20 and Pacific Islanders21 and men who have sex with men.22 Figure 1. Anterior Abdominal-Wall Abscess in a 15-Year- There has been a dramatic increase in the Old Boy. occurrence of S. aureus infections in general and There was a 3-day history of drainage from this abscess, which had increased in size to 1 cm in length and be- community-associated MRSA infections in par- come more painful. It was fluctuant and tender on ticular. At Driscoll Children’s Hospital in Corpus examination. Incision and drainage were performed; Christi, Texas, the number of community-asso- about 2 ml of purulent material was obtained. A culture ciated MRSA infections increased from 9 in 1999 yielded MRSA that was susceptible to clindamycin. The to 459 in 200323; in 2003, these infections consti- results of the D-zone test were negative. tuted 98% of S. aureus infections overall in that institution. In most, but not all, U.S. cities, community-associated MRSA is now the most common pathogen cultured from patients with skin and soft-tissue infections in emergency departments.24 Epidemic community-associated MRSA disease has also been reported from some rural areas, although epidemic disease has not yet spread to all regions of the United States. Consistent with the occurrence of epidemic, symptomatic, community-associated MRSA disease in the United States are observations of the increasing prevalence of asymptomatic colonization of MRSA among children25 and adults26 in the community. Recent data indicate that 9.2% of healthy children in Nashville have asymptomatic Figure 2. Swelling and a Small Amount of Drainage Involving the Left Naris in a 10-Year-Old Girl. colonization23 (74% of these infections are com- There was a 5-day history of drainage from the lesion. munity-associated MRSA [Creech CB: personal The child appeared well and did not have a fever. Inci- communication]), as compared with 0.8% in 2001, sion and drainage yielded 0.5 ml of purulent material. and 7.3% of adolescents and adults in Atlanta have A culture yielded MRSA that was susceptible to clinda- asymptomatic colonization, including both hos- mycin. The results of the D-zone test were negative. pital- and community-acquired MRSA isolates.24 Skin and soft-tissue infections represent the majority of the community-associated MRSA dis- (even in areas where these spiders do not live) or ease burden27 and are the focus of this article. insect bites. In addition, necrotizing pneumo- Examples of such infections are shown in Figures nia,28 pleural empyema, necrotizing fasciitis,29 1 and 2. (Other examples are in the Supplemen- septic thrombophlebitis with pulmonary emboli- tary Appendix, available with the full text of this zation,30 myositis,31 and severe sepsis with pur- article at www.nejm.org.) Necrotic skin lesions pura fulminans and the Waterhouse–Friderichsen are a common presentation and are often incor- syndrome32 have been described in association rectly attributed to bites by brown recluse spiders with community-associated MRSA.28,29,33

n engl j med 357;4 www.nejm.org july 26, 2007 381 T h e new england journal o f medicine

Strategies and Evidence drainage alone may suffice, particularly for abscesses that are small. Lee et al.41 have defined Evaluation small abscesses as those that are less than 5 cm Suspicion that community-associated MRSA may in length, but this definition may not be appropri- be the cause of a skin and soft-tissue infection ate for skin and soft-tissue infections in infants should be heightened by a history of previous and in certain areas of the body (e.g., the head MRSA infection in the patient or a household con- and neck). In patients with larger abscesses, sys- tact. Table 1 lists other groups likely to be at risk temic signs of infection, or both, antimicrobial for community-associated MRSA transmission. therapy is generally recommended in addition to However, many patients with community-associ- incision and drainage (for purulent lesions). The ated MRSA infection have none of these risk fac- type and route of therapy should be guided by the tors. Furthermore, no clinical features distinguish severity of the clinical syndrome. with certainty skin and soft-tissue infections caused by MRSA from those caused by methicillin-suscep- Outpatient Therapies tible S. aureus.38 Topical antimicrobial therapy is sometimes used Information on local antibiotic-resistance pat- to treat superficial MRSA skin infections such as terns (e.g., from local hospitals) can help clinicians impetigo, although comparative outcome data are to assess the likelihood of community-associated lacking. Bacitracin, alone or in combination with MRSA infection and guide decisions regarding polymyxin and neomycin, mupirocin (Bactroban), empirical treatment. Some have suggested that and retapamulin (Altabax) are commercially avail- management strategies should be tailored to the able for this purpose. For bacitracin, in vitro sus- possibility of community-associated MRSA infec- ceptibility factors that predict the clinical outcome tion on the basis of an arbitrary threshold of 10% have not been defined.42 For mupirocin, isolates or more methicillin resistance among S. aureus with low-level resistance and those with high-level isolates. resistance have been identified; the latter do pre- Obtaining a specimen for culture and suscep- dict clinical failure and may be increasing in prev- tibility testing, which was considered to be un- alence among MRSA isolates.43,44 Retapamulin is necessary when the prevalence of MRSA was low, newly licensed for children 9 months of age or is useful in guiding therapy. Specimens are most older. It has good in vitro activity against MRSA commonly obtained at the time of incision and infection, but mutants with decreased susceptibil- drainage of purulent skin and soft-tissue lesions. ity can be selected in vitro.45 In nonpurulent cellulitis that is not amenable For oral systemic treatment, β-lactam anti- to incision and drainage, a possible approach is biotics can no longer be considered to be reliable a biopsy with culture of the material obtained. as empirical therapy for community-acquired skin In practice, this procedure is infrequently per- and soft-tissue infections. The optimal antibiotic formed.39 Moreover, although many patients with MRSA bacteremia also have nasal colonization 40 Table 1. Persons at Risk for Skin and Soft-Tissue Infections with the organism, it is not known whether Caused by Community-Associated MRSA. screening for such colonization in a patient with a skin and soft-tissue infection has useful predic- Household contacts of a patient with proven community- associated MRSA infection14 tive value. Such screening is not currently recom- 34 mended. Children Day-care center contacts of hospitalized patients with MRSA infections15,16 Treatment Men who have sex with men22 The recommended treatment of community- Soldiers17,18 associated MRSA infection depends on an assess- Incarcerated persons18 ment of the severity of the clinical presentation Athletes, particularly those involved in contact sports19 and the type of skin and soft-tissue infection. Native Americans20 Purulent skin and soft-tissue infections without Pacific Islanders21 associated systemic signs, such as fever, tachycar- Persons with a previous community-associated MRSA dia, or hemodynamic instability, are generally infection35,36 managed with incision and drainage, with or Intravenous drug users37 without oral antimicrobial therapy; incision and

382 n engl j med 357;4 www.nejm.org july 26, 2007 clinical practice therapy when community-associated MRSA infec- resistance of group A streptococci to these agents. tion is suspected is not clear. Results of suscep- Such resistance is well documented for tetracy- tibility testing and clinical experience provide clines, although it is less clear for trimethoprim– support for a primary role of older antibiotics sulfamethoxazole.39 However, these agents are such as clindamycin, trimethoprim–sulfamethox reasonable choices in cases in which community- azole, and tetracyclines, although their effective- associated MRSA infection is confirmed or strong- ness for skin and soft-tissue infections due to ly suggested by the presence of purulent material. community-associated MRSA has not been rigor- Some clinicians suggest the addition of a β-lac- ously evaluated or compared in clinical trials. tam antibiotic, that is active against streptococci Table 2 lists oral agents that are useful in the if trimethoprim–sulfamethoxazole or a tetracy- outpatient management of community-associated cline is used for a nonpurulent cellulitis of uncer- MRSA infections. An observational study showed tain cause. that clindamycin, a lincosamide antibiotic, was Testing of nearly all community-associated uniformly effective in 39 patients with clindamy- MRSA isolates shows susceptibility to trimetho cin-susceptible community-associated MRSA in- prim–sulfamethoxazole, but data on the out- fection who were mildly to moderately ill.46 The comes of treatment are limited. In a study at an disadvantages of this medication include its asso outpatient clinic in Boston where almost half of ciation with diarrhea caused by Clostridium difficile community-associated MRSA isolates were clinda- and increasing rates of clindamycin resistance in mycin-resistant and where trimethoprim–sulfa- some regions of the world.11,47,48 Clindamycin methoxazole became the most frequently used resistance among community-associated MRSA antimicrobial agent for skin and soft-tissue in- isolates should be monitored locally, and some fections caused by community-associated MRSA,47 experts recommend avoiding empirical therapy the percentage of patients with clinical resolution with clindamycin when local rates of clindamy- of the MRSA infection increased in parallel with cin resistance exceed 10 to 15% among MRSA trimethoprim–sulfamethoxazole use during the isolates causing skin and soft-tissue infections. study period (1998 to 2005). In another study, Moreover, the results of testing for clindamy- however, treatment failure occurred in 6 of 12 cin susceptibility may be misleading; occasional adults who received double-strength trimetho treatment failures have been documented when prim–sulfamethoxazole.51 Few data are available the results of tests showed that an MRSA isolate to provide support for the efficacy of doxycycline was susceptible to clindamycin but resistant to or minocycline. In one retrospective review of erythromycin.46,49 In such cases, use of the D‑zone skin and soft-tissue infections caused by commu test (Fig. 3) is warranted to detect inducible clin nity-associated MRSA, the cure rate was 83%.52 damycin resistance; positive results in 10 to 20% Linezolid, a newer antimicrobial agent in the of tested isolates (with one notable outlier 49) have oxazolidinone family, is active against almost all been reported, but these rates may be increasing. community-associated MRSA isolates and group The Clinical and Laboratory Standards Institute A streptococci. The disadvantages of this agent suggests that isolates that are positive on the include its high cost, the lack of routine avail- D‑zone test should be reported as being resistant ability, hematologic side effects, and the potential to clindamycin despite a positive result of single- for resistance among S. aureus strains, possibly by agent susceptibility testing.50 The institute sug- multiple mechanisms. Prolonged linezolid admin- gests permissive language to accompany the result istration increases the likelihood of resistance, of the susceptibility testing: “The isolate is pre- probably through the accumulation of mutations sumed to be resistant based on detection of induc- in multiple copies of the 23S ribosomal RNA ible clindamycin resistance. Clindamycin might S. aureus gene.53 still be effective in some patients.” In practice, Rifampin is highly active against susceptible when the results of the D-zone test become community-associated MRSA isolates, but a high known, the use of clindamycin should be recon- frequency of mutations to rifampin resistance sidered on the basis of the clinical response. is a contraindication for the use of rifampin Neither trimethoprim–sulfamethoxazole nor alone.54 Thus, a combination of trimethoprim– tetracyclines are generally recommended as sole sulfamethoxazole or doxycycline with rifampin empirical therapy for a nonpurulent cellulitis of is sometimes used for the treatment of skin and unknown cause because of concerns regarding the soft-tissue infections caused by community-asso-

n engl j med 357;4 www.nejm.org july 26, 2007 383 T h e new england journal o f medicine - - Comments taste of the suspension the of taste not may suspension oral available immediately be pharmacies many at capsule available; cially sprin be may powder as such food on kled applesauce Many patients dislike the dislike patients Many high; relatively is cost The commer is suspension No ------

ally thrombo ally because resis because fluids, abnor fluids, u

Clostridium diffi Clostridium Main Side Effects Effects Side Main hematologic suppression suppression hematologic

and Contraindications and

tivity, cile thrombocytopenia), (especially syndrome Stevens–Johnson the bones and teeth in of because age of years 9 than and teeth in deposition potential bones vestibular bones, and teeth in toxicity of because age of years 9 than and teeth in deposition potential bones ane cause can but cytopenia), with mostly neutropenia, or mia use prolonged drug– function, liver in malities interactions drug un an at selected are mutants tant rate high acceptably Myelosuppression (us Myelosuppression Discoloration of body of Discoloration Cannot be used alone used be Cannot Diarrhea caused by caused Diarrhea photosensi rash, vomiting, Nausea, deposition photosensitivity, Nausea, younger children in Contraindicated deposition photosensitivity, Nausea, younger children in Contraindicated - - - sion sion suspension suspension sion Formulations Tablet, suspen Tablet, suspen Tablet, tablet, Capsule, tablet, Capsule, suspen Tablet, Capsule - - - Children day, in three or four di four or three in day, doses vided sulfamethoxa and day, mg/kg/day, 40–60 zole, doses divided two in doses divided four doses vided doses; divided four 600 dose, maximum mg/day divided doses divided 30 mg/kg of body weight/ body of mg/kg 30 mg/kg/ 8–12 Trimethoprim, or two in mg/kg/day, 2–4 two in mg/kg/day, 4 di three in mg/kg/day, 30 or two in mg/kg/day, 20 - - Usual Dose* divided doses divided

Adults twice daily (each tablet con tablet (each daily twice mg, 160 trimethoprim, taining 800 sulfamethoxazole, and mg) doses divided dose, maximum doses; vided mg/day 600 200 mg/day, in two in mg/day, 200 300 mg thrice daily thrice mg 300 tablets double-strength 2 to 1 four or two in mg/day, 100–200 daily twice mg 600 di four or two in mg/kg/day, 20 - Oral Agents for the Outpatient Treatment of Putative Community-Associated MRSA Infections. MRSA Community-Associated Putative of Treatment Outpatient the for Agents Oral Doxy-100, Monodox, Doxy-100, Vibra-Tabs) Vibramycin, Minocin) zole (Bactrim, Septra) (Bactrim, zole Adoxa, (Doryx, Doxycycline (Dynacin, Minocycline Table 2. Table Medication (Cleocin) Clindamycin Trimethoprim–sulfamethoxa Tetracyclines (Zyvox) Linezolid Rimactane) (Rifadin, Rifampin Optimal doses have not been established for all drugs listed.

384 n engl j med 357;4 www.nejm.org july 26, 2007 clinical practice ciated MRSA,51 although data are lacking to provide support for this approach. Fluoroquinolones should not be used to treat skin and soft-tissue infections caused by community-associated MRSA. Resistance to them develops readily in S. aureus and is already widely prevalent.24

Inpatient Therapies Some patients with community-associated MRSA infection will require more aggressive treatment than incision and drainage with or without oral antimicrobial therapy on an outpatient basis. A decision to hospitalize a patient for parenteral therapy (Table 3) depends on several factors, including clinical judgment regarding the severity of the illness. The presence of a large abscess, fever, other Figure 3. The D-Zone Test for Erythromycin-Resistant, signs of systemic infection, or high-risk charac- Clindamycin-Susceptible Isolates. teristics such as an age younger than 6 months, The Clinical and Laboratory Standards Institute advises diabetes, or immunodeficiency should prompt clinical microbiology laboratories to perform a D-zone consideration of hospitalization. The detailed man test on erythromycin-resistant, clindamycin-susceptible isolates. This test detects inducible clindamycin resis- agement of invasive disease due to community- tance; blunting of the clindamycin zone of inhibition associated MRSA is beyond the scope of this (arrow) suggests the presence of an erm gene in the test review. isolate that is inducible by erythromycin. The erm gene Vancomycin is still considered the first-line can confer the macrolide–lincosamide–streptogramin B treatment for hospitalized patients with invasive phenotype to an isolate with cross-resistance to macrolide antibiotics such as erythromycin, lincosamide S. aureus infection. However, this drug should be antibiotics such as clindamycin, and streptogramin B switched if susceptibility testing indicates that a antibiotics. E denotes erythromycin, and CC clindamy- more rapidly bactericidal β-lactam agent such cin concentration. as oxacillin would be appropriate. Microbiologic treatment failure may occur with vancomycin even if there is no increase in the minimal inhib Intravenous trimethoprim–sulfamethoxazole itory concentration (MIC) on susceptibility test- has undergone minimal evaluation for invasive ing.55,56 S. aureus isolates with low-level (so-called S. aureus infection. A study of intravenous drug intermediate) resistance to vancomycin (MIC, abusers with serious S. aureus infections antedated >2 μg per milliliter) as well as those with high- the epidemic of community-associated MRSA in- level resistance (MIC, >16 μg per milliliter) have fection, and it indicated that intravenous trimeth been described, and they may not be identified by oprim–sulfamethoxazole was significantly less ef- means of routine techniques for susceptibility test fective than vancomycin.61 ing.57 Although resistant isolates are believed to Parenteral linezolid lacks bactericidal activity, be infrequent, global decreased susceptibility which some experts believe is important in treat- (so-called MIC creep) among S. aureus isolates ing intravascular infection, a common feature of has been documented in several locations in the invasive disease. Moreover, reports of a case of United States,58-60 and this decreased susceptibility endocarditis caused by a susceptible organism may limit the continued effectiveness of vanco- during linezolid therapy and of clinical failure in mycin. Some experts have proposed that the use patients treated with linezolid for endocarditis of a higher dose and maintenance of high serum have raised concerns about its use alone for se- levels of vancomycin may be beneficial, but the vere, invasive S. aureus infections62,63 (an exception efficacy of these strategies has not been proved. is health care–associated MRSA pneumonia, for Parenteral clindamycin may be useful in re- which linezolid has proved efficacious64). gions where the likelihood of a resistant organ- Tigecycline, a parenteral glycylcycline–mino- ism is low. It should not be used as sole therapy cycline derivative, was also recently approved by when the patient is moderately to severely ill. the Food and Drug Administration (FDA) for the

n engl j med 357;4 www.nejm.org july 26, 2007 385 T h e new england journal o f medicine

Table 3. Parenteral Agents for the Treatment of Putative Community-Associated MRSA Infections.

Main Side Effects Medication Usual Dose* and Contraindications Comments Adults Children Vancomycin 2–4 g/day, in two 40 mg/kg/day, in The red-man syndrome (a hista- Slowing the rate of administration is usually (Vancocin) to four divided three to four mine-release syndrome usu- sufficient management for the red-man doses divided doses ally manifested as flushing) syndrome, but accompanying hypotension may require discontinuation of the drug or additional intervention in rare cases Excretion is slowed in patients with renal failure, and serum levels should be monitored in such patients to avoid drug accumulation; whether such monitoring is routinely necessary in patients with normal renal function is not clear, but it should be performed when multiple nephrotoxic drugs are administered simultaneously Clindamycin 300 mg thrice 30 mg/kg/day, in Diarrhea caused by C. difficile (Cleocin) daily three divided doses Daptomycin 4–6 mg/kg/day, Unknown Potential muscle toxicity Resistance was documented in 6 of 120 patients (Cubicin) in four divided receiving this therapy† doses Excretion is slowed in patients with renal failure, and dosage adjustment is recommended Tigecycline 100 mg loading Unknown Nausea, vomiting, photosensitivi- (Tygacil) dose, then 50 ty, deposition in teeth and mg every 12 hr bones Contraindicated in children younger than 9 years of age because of potential deposition in teeth and bones Linezolid 600 mg/day, in 30 mg/kg/day, in Myelosuppression (usually The cost is relatively high (Zyvox) two divided two to three thrombocytopenia, but also doses divided doses anemia or neutropenia), mostly with pro-longed use Quinupristin 7.5 mg/kg, every 7.5 mg/kg, every Hyperbilirubinemia, arthralgias Dosage adjustment may be necessary in pa- and dalfopris- 8–12 hr 8–12 hr and myalgias, phlebitis, drug– tients with hepatic impairment tin (Synercid) drug interactions (especially with cytochrome P450 3A4 substrates)

* Optimal doses have not been established for all drugs listed. † Data are from Fowler et al.55

treatment of skin and soft-tissue infections caused approved by the FDA for use in patients with by MRSA.65 This approval was granted on the basis skin and soft-tissue infections. The success rate of data showing microbiologic eradication in 25 of with the use of daptomycin for these infections 32 adults (78%) with complicated skin and soft- is 75% — similar to that of vancomycin. It is also tissue infections. approved for MRSA bacteremia,55 including that A fixed combination of the streptogramins associated with right-sided endocarditis, but it quinupristin and dalfopristin (Synercid) was li- should not be used for pneumonia, for which its censed by the FDA for the treatment of skin and efficacy has been limited by its propensity for soft-tissue infections caused by methicillin-sus- binding surfactant.66 ceptible S. aureus. Its use has been limited by the potential for drug–drug interactions and by side Areas of Uncertainty effects (including arthralgias, myalgias, and gastrointestinal toxic effects). The optimal oral antimicrobial regimen for the Daptomycin, a cyclic lipodepsipeptide, has been treatment of skin and soft-tissue infections is not

386 n engl j med 357;4 www.nejm.org july 26, 2007 clinical practice known. A trial addressing this question, sponsored fective if the solution is permitted to remain on by the National Institutes of Health, is expected the skin before rinsing. to be initiated this year. Contagion among the close household con- The optimal management of recurrent com- tacts of patients, as well as correctional facility, munity-associated MRSA disease is also uncer- school, and sports-team contacts, is well recog- tain. Although not well studied, the recurrence nized. Although the risk of transmission has not rate is believed to be 10% or higher. It is not been well quantified, anecdotal evidence suggests clear whether recurrences represent autoinocula- that more than 60% of households of children tion or a new MRSA infection. At present, recur- hospitalized with community-associated MRSA rent episodes are generally treated in the same infections have one or more members with a his- way as the initial episode. In addition, “decoloni- tory of a putative MRSA infection in the previous zation” strategies are frequently recommended 6 months. If this estimate proves to be correct, in such cases, although neither the indications for it will lend support to the empirical treatment of their use nor their effectiveness in reducing the an entire household (perhaps even including pets) risk of recurrences is clear. One such strategy is if an effort to eradicate community-associated the use of intranasal mupirocin to reduce nasal MRSA colonization in a patient is undertaken. carriage of MRSA; however, eradication of nasal The efficacy of such an approach has not been colonization appears to be transient, and the use studied. of this agent remains controversial. Moreover, The role of fomites needs to be clarified. the recent identification of a mupirocin-resistance Hospital-acquired MRSA isolates can survive on gene in USA300 isolates (which accounted for a variety of inanimate surfaces, sometimes for 97% of isolates in a recent study24) and of mupi- weeks. It is unclear whether this is also true for rocin resistance among 11 community-associated community-associated MRSA isolates; if it is, MRSA isolates in Boston raises serious concern their presence on such items as clothing, towels, about exposing populations of staphylococci to and athletic equipment might contribute to out- this agent.67 Some experts have also proposed breaks. Pets (including dogs and cats), livestock, adjunctive attempts at skin decolonization. Topi- and birds have been identified as MRSA carriers; cal chlorhexidine gluconate or 1 tsp (3.4 g) of their role in MRSA transmission to humans re- bleach diluted in 1 gallon (3.8 liters) of bath quires further evaluation.68 Local hygiene mea- water is commonly suggested, although these ap- sures recommended by an expert panel from the proaches have not been rigorously evaluated. The Centers for Disease Control and Prevention (CDC) optimal strength of the chlorhexidine solution is are shown in Table 4. not known, nor is it clear whether it is more ef- No vaccine is currently available for S. aureus. Many experts believe that it is unlikely that a single-antigen approach will prove to be effective. Table 4. Recommended Measures to Limit the Spread of Community-Associated MRSA Isolates.* Guidelines Cover draining wounds with clean bandages. Wash hands, especially after contact with a contaminated The CDC has issued guidelines for the prevention wound. and management of community-associated MRSA Launder clothing after contact with a contaminated area infections.69 The recommendations in this article on the skin. are largely concordant with this review. Bathe regularly with use of soap. Avoid sharing items (e.g., towels, bedding, clothing, ra- zors, or athletic equipment) that may become contami- conclusions and nated by contact with wounds or skin flora. Recommendations Clean sports equipment with agents that are effective against staphylococci (e.g., a detergent or disinfec- With the increasing prevalence of community- tant registered by the Environmental Protection associated MRSA infection, the management of Agency, such as quaternary ammonium compounds skin and soft-tissue infections requires knowl- or a solution of dilute bleach). edge of local rates of MRSA infection. Many experts * Information is modified from Gorwitz et al.69 suggest an arbitrary threshold of more than 10% methicillin resistance among S. aureus isolates

n engl j med 357;4 www.nejm.org july 26, 2007 387 T h e new england journal o f medicine

causing skin and soft-tissue infections acquired Although data directly comparing antimicro- in the community and recommend inclusion of bial agents for the treatment of community-asso antimicrobial therapy against community-associated MRSA infection are lacking, clindamycin, ciated MRSA when managing a putative S. aureus trimethoprim–sulfamethoxazole, or a long-acting infection. tetracycline such as doxycycline is a reasonable In a patient such as the man described in the initial choice; linezolid is another possibility. vignette, presenting with an abscess or a purulent Follow-up is essential, since relapse or recurrence and necrotic skin lesion, incision and drainage may occur. are the cornerstones of therapy; purulent material Supported by grants from the National Institute of Allergy should be cultured. In many patients, particularly and Infectious Diseases (RO1AI40481 and 5RO1AI067584), the Centers for Disease Control and Prevention (RO1CCR 523379, those with small lesions (<5 cm in length), inci- RO1CI000373, and U01-CI000384), and the Grant HealthCare sion and drainage alone will be adequate therapy. Foundation. If the skin lesions are large or accompanied by Dr. Daum reports serving on paid advisory boards for Clorox, Sanofi Pasteur, GlaxoSmithKline, Pfizer, and the MRSA National systemic signs of infection or if there is evidence Faculty Meeting (sponsored by Astellas and Theravance); re- of an increased risk of complicated community- ceiving lecture fees from Nabi Biopharmaceuticals and Pfizer; associated MRSA disease, antimicrobial thera- and receiving grant support from Clorox, Pfizer, Sage Products, and Sanofi Pasteur. No other potential conflict of interest rele- py that is active against community-associated vant to this article was reported. MRSA is also recommended. Therapy ultimately I thank Michael David, M.D., Daniel Glikman, M.D., Stephen should be guided by the results of susceptibility Weber, M.D., Loren Miller, M.D., and Sharmeen Younus, Pharm.D., for helpful critical comments and Mark A. Hostetler, M.D., of testing of cultures obtained before the initiation the Department of Pediatrics, University of Chicago, for the pho- of therapy. tographs of patients.

References 1. Crawford SE, Boyle-Vavra S, Daum RS. coccus aureus disease? J Infect Dis 2006;194: son JE, Murphy TV. Methicillin-resistant Community associated methicillin-resis- 1761-70. Staphylococcus aureus in two child care cen- tant Staphylococcus aureus. In: Hooper D, 9. Labandeira-Rey M, Couzon F, Boisset ters. J Infect Dis 1998;178:577-80. Scheld M, eds. Emerging infections. Vol. 7. S, et al. Staphylococcus aureus Panton Valen- 16. Shahin R, Johnson IL, Jamieson F, Washington, DC: ASM Press, 2007:153-79. tine leukocidin causes necrotizing pneu- McGreer A, Tolkin J, Ford-Jones EL. Methi- 2. Chambers HF. The changing epidemi- monia. Science 2007;315:1130-3. cillin-resistant Staphylococcus aureus carriage ology of Staphylococcus aureus? Emerg Infect 10. Ito T, Ma XX, Takeuchi F, Okuma K, in a child care center following a case of Dis 2001;7:178-82. Yuzawa H, Hiramatsu K. Novel type V disease. Arch Pediatr Adolesc Med 1999; 3. Herold BC, Immergluck LC, Maranan staphylococcal cassette chromosome mec 153:864-8. MC, et al. Community-acquired methicil- driven by a novel cassette chromosome 17. Ellis MW, Hospenthal DR, Dooley DP, lin-resistant Staphylococcus aureus in chil- recombinase, ccrC. Antimicrob Agents Che- Gray PJ, Murray CK. Natural history of dren with no predisposing risk. JAMA mother 2004;48:2637-51. community-acquired methicillin-resistant 1998;279:593-8. 11. Boyle-Vavra S, Ereshefsky B, Wang CC, Staphylococcus aureus colonization and infec- 4. Daum RS, Ito T, Hiramatsu K, et al. Daum RS. Successful multiresistant com- tion in soldiers. Clin Infect Dis 2004;39: A novel methicillin-resistance cassette in munity-associated methicillin-resistant 971-9. community-acquired methicillin-resistant Staphylococcus aureus lineage from Taipei, 18. Aiello AE, Lowy FD, Wright LN, Lar- Staphylococcus aureus isolates of diverse ge- Taiwan, that carries either the novel staphy son EL. Meticillin-resistant Staphylococcus netic backgrounds. J Infect Dis 2002;186: lococcal chromosome cassette mec (SCCmec) aureus among US prisoners and military

1344-7. type VT or SCCmec type IV. J Clin Microbiol personnel: review and recommendations 5. Naimi TS, LeDell KH, Como-Sabetti 2005;43:4719-30. [Erratum, J Clin Micro- for future studies. Lancet Infect Dis 2006; K, et al. Comparison of community- and biol 2005;43:6223.] 6:335-41. health care-associated methicillin-resis- 12. McDougal LK, Steward CD, Killgore 19. Kazakova SV, Hageman JC, Matava M, tant Staphylococcus aureus infection. JAMA GE, Chaitram JM, McAllister SK, Tenover et al. A clone of methicillin-resistant Staphy 2003;290:2976-84. FC. Pulsed-field gel electrophoresis typing lococcus aureus among professional football 6. Vandenesch F, Naimi T, Enright MC, of oxacillin-resistant Staphylococcus aureus players. N Engl J Med 2005;352:468-75. et al. Community-acquired methicillin- isolates from the United States: establish- 20. Groom AV, Wolsey DH, Naimi TS, et al. resistant Staphylococcus aureus carrying Pan- ing a national database. J Clin Microbiol Community-acquired methicillin-resistant ton-Valentine leukocidin genes: worldwide 2003;41:5113-20. Staphylococcus aureus in a rural American emergence. Emerg Infect Dis 2003;9:978- 13. Maree CL, Daum RS, Boyle-Vavra S, Indian community. JAMA 2001;286:1201- 84. Matayoshi K, Miller LG. Community-asso- 5. 7. Gillet Y, Issartel B, Vanhems P, et al. ciated methicillin-resistant Staphylococcus 21. Castrodale LJ, Beller M, Gessner BD. Association between Staphylococcus aureus aureus isolates causing healthcare-associ- Over-representation of Samoan/Pacific Is- strains carrying gene for Panton-Valentine ated infections. Emerg Infect Dis 2007;13: landers with methicillin-resistant Staphy leukocidin and highly lethal necrotising 236-42. lococcus aureus (MRSA) infections at a large pneumonia in young immunocompetent 14. Dietrich DW, Auld DB, Mermel LA. family practice clinic in Anchorage, Alas- patients. Lancet 2002;359:753-9. Community-acquired methicillin-resistant ka, 1996-2000. Alaska Med 2004;46: 8. Voyich JM, Otto M, Mathema B, et al. Staphylococcus aureus in southern New En 88-91. Is Panton-Valentine leukocidin the ma- gland children. Pediatrics 2004;113(4): 22. Lee NE, Taylor MM, Bancroft E, et al. jor virulence determinant in community- e347-e352. Risk factors for community-associated associated methicillin-resistant Staphylo 15. Adcock PM, Pastor P, Medley F, Patter- methicillin-resistant Staphylococcus aureus

388 n engl j med 357;4 www.nejm.org july 26, 2007 clinical practice skin infections among HIV-positive men man-Roberts H, Mussa HR, Elliott AC. ty-acquired Staphylococcus aureus infections who have sex with men. Clin Infect Dis Prospective comparison of methicillin- in children. Clin Infect Dis 2005;40: 2005;40:1529-34. [Erratum, Clin Infect Dis susceptible and methicillin-resistant com- 1785-91. 2005;41:135.] munity-associated Staphylococcus aureus in- 49. Frank AL, Marcinak JF, Mangat PD, et 23. Purcell K, Fergie J. Epidemic of com- fections in hospitalized patients. J Infect al. Clindamycin treatment of methicillin- munity-acquired methicillin-resistant Staph 2007;54:427-34. resistant Staphylococcus aureus infections ylococcus aureus infections: a 14-year study 37. Young DM, Harris HW, Charlebois ED, in children. Pediatr Infect Dis J 2002;21: at Driscoll Children’s Hospital. Arch Pe- et al. An epidemic of methicillin-resistant 530-4. diatr Adolesc Med 2005;159:980-5. Staphylococcus aureus soft tissue infections 50. Performance standards for antimicro- 24. Moran GJ, Krishnadasan A, Gorwitz among medically underserved patients. bial susceptibility testing: seventeenth in- RJ, et al. Methicillin-resistant S. aureus in- Arch Surg 2004;139:947-53. formational supplement. M100-S17. Wayne, fections among patients in the emergency 38. Miller LG, Perdreau-Remington F, PA: Clinical and Laboratory Standards In- department. N Engl J Med 2006;355:666- Bayer AS, et al. Clinical and epidemiologic stitute, 2007:52. 74. characteristics cannot distinguish commu 51. Iyer S, Jones DH. Community-acquired 25. Creech CB II, Kernodle DS, Alsentzer nity-associated methicillin-resistant Staph methicillin-resistant Staphylococcus aureus A, Wilson C, Edwards KM. Increasing rates ylococcus aureus infection from methicillin- skin infection: a retrospective analysis of of nasal carriage of methicillin-resistant susceptible S. aureus infection: a prospective clinical presentation and treatment of a Staphylococcus aureus in healthy children. investigation. Clin Infect Dis 2007;44:471- local outbreak. J Am Acad Dermatol 2004; Pediatr Infect Dis J 2005;24:617-21. 82. 50:854-8. 26. Hidron AI, Kourbatova EV, Halvosa JS, 39. Swartz MN. Cellulitis. N Engl J Med 52. Ruhe JJ, Monson T, Bradsher RW, Me- et al. Risk factors for colonization with 2004;350:904-12. non A. Use of long-acting tetracyclines for methicillin-resistant Staphylococcus aureus 40. von Eiff C, Becker K, Machka K, Stam- methicillin-resistant Staphylococcus aureus (MRSA) in patients admitted to an urban mer H, Peters G. Nasal carriage as a source infections: case series and review of the hospital: emergence of community-associ- of Staphylococcus aureus bacteremia. N Engl literature. Clin Infect Dis 2005;40:1429-34. ated MRSA nasal carriage. Clin Infect Dis J Med 2001;344:11-6. 53. Peeters MJ, Sarria JC. Clinical char- 2005;41:159-66. 41. Lee MC, Rios AM, Aten MF, et al. acteristics of linezolid-resistant Staphylo 27. Kaplan SL, Hulten KG, Gonzalez BE, Management and outcome of children coccus aureus infections. Am J Med Sci 2005; et al. Three-year surveillance of commu- with skin and soft tissue abscesses caused 330:102-4. nity-acquired Staphylococcus aureus infections by community-acquired methicillin-resis- 54. Strausbaugh LJ, Jacobson C, Sewell DL, in children. Clin Infect Dis 2005;40:1785- tant Staphylococcus aureus. Pediatr Infect Potter S, Ward TT. Antimicrobial therapy 91. Dis J 2004;23:123-7. for methicillin-resistant Staphylococcus au 28. Mongkolrattanothai K, Boyle S, Ka- 42. Jones RN, Li Q, Kohut B, Biedenbach reus colonization in residents and staff of a hana MD, Daum RS. Severe Staphylococcus DJ, Bell J, Turnidge JD. Contemporary anti- Veterans Affairs nursing home care unit. aureus infections caused by clonally related microbial activity of triple antibiotic oint- Infect Control Hosp Epidemiol 1992;13: community-acquired methicillin-suscep- ment: a multiphased study of recent 151-9. tible and methicillin-resistant isolates. clinical isolates in the United States and 55. Fowler VG Jr, Boucher HW, Corey GR, Clin Infect Dis 2003;37:1050-8. Australia. Diagn Microbiol Infect Dis 2006; et al. Daptomycin versus standard therapy 29. Miller LG, Perdreau-Remington F, Rieg 54:63-71. for bacteremia and endocarditis caused by G, et al. Necrotizing fasciitis caused by 43. Han LL, McDougal LK, Gorwitz RJ, Staphylococcus aureus. N Engl J Med 2006; community-associated methicillin-resis- et al. High frequencies of clindamycin and 355:653-65. tant Staphylococcus aureus in Los Angeles. tetracycline resistance in methicillin-resis- 56. Moore MR, Perdreau-Remington F, N Engl J Med 2005;352:1445-53. tant Staphylococcus aureus pulsed-field type Chambers HF. Vancomycin treatment fail- 30. Gonzalez BE, Teruya J, Mahoney DH Jr, USA300 isolates collected at a Boston am- ure associated with heterogeneous vanco- et al. Venous thrombosis associated with bulatory health center. J Clin Microbiol mycin-intermediate Staphylococcus aureus in staphylococcal osteomyelitis in children. 2007;45:1350-2. a patient with endocarditis and in the Pediatrics 2006;117:1673-9. 44. Diep BA, Gill SR, Chang RF, et al. rabbit model of endocarditis. Antimicrob 31. Pannaraj PS, Hulten KG, Gonzalez BE, Complete genome sequence of USA300, Agents Chemother 2003;47:1262-6. Mason EO Jr, Kaplan SL. Infective pyomyo- an epidemic clone of community-acquired 57. Centers for Disease Control and Pre- sitis and myositis in children in the era of meticillin-resistant Staphylococcus aureus. vention. VISA/VRSA: vancomycin-inter- community-acquired, methicillin-resistant Lancet 2006;367:731-9. mediate/resistant Staphylococcus aureus. (Ac- Staphylococcus aureus infection. Clin Infect 45. Gentry DR, Rittenhouse SF, McCloskey cessed July 2, 2007, at http://www.cdc. Dis 2006;43:953-60. L, Holmes DJ. Stepwise exposure of Staphy gov/ncidod/dhqp/ar_visavrsa.html.) 32. Adem PV, Montgomery CP, Husain lococcus aureus to pleuromutilins is associat 58. Wang G, Hindler JF, Ward KW, Bruck- AN, et al. Staphylococcus aureus sepsis and ed with stepwise acquisition of mutations ner DA. Increased vancomycin MICs for the Waterhouse–Friderichsen syndrome in in rplC and mimimally affects susceptibil- Staphylococcus aureus clinical isolates from a children. N Engl J Med 2005;353:1245- ity to retapamulin. Antimicrob Agents university hospital during a 5-year period. J 51. Chemother 2007;51:2048-52. Clin Microbiol 2006;44:3883-6. 33. Kravitz GR, Dries DJ, Peterson ML, 46. Martinez-Aguilar G, Hammerman WA, 59. Kapadia M, Coyle E, Prince R, et al. Schlievert PM. Purpura fulminans due to Mason EO Jr, Kaplan SL. Clindamycin Declining in vitro activity of vancomycin Staphylococcus aureus. Clin Infect Dis 2005; treatment of invasive infections caused by against Staphylococcus aureus isolates from 40:941-7. community-acquired methicillin-resistant cancer patients. In: Programs and abstracts 34. Fridkin SK, Hageman JC, Morrison M, and methicillin-susceptible Staphylococcus of the 45th Interscience Conference on et al. Methicillin-resistant Staphylococcus aureus in children. Pediatr Infect Dis J 2003; Antimicrobial Agents and Chemotherapy, aureus disease in three communities. N Engl 22:593-8. Washington, DC, December 16–19, 2005. J Med 2005;352:1436-44. 47. Szumowski JD, Cohen DE, Kanaya F, Washington, DC: American Society for 35. Szumowski JD, Cohen DE, Kanaya F, Mayer KH. Treatment and outcomes of Microbiology, 2005. abstract. Mayer KH. Treatment and outcomes of infections by methicillin-resistant Staphy 60. Golan Y, Baiez-Giangreco C, O’Sullivan infections by methicillin-resistant Staphy lococcus aureus at an ambulatory clinic. Anti- C, Snydman DR. Trends in vancomycin lococcus aureus at an ambulatory clinic. Anti- microb Agents Chemother 2007;51:423-8. susceptibility among consecutive MRSA microb Agents Chemother 2007;51:423-8. 48. Kaplan SL, Hulten KG, Gonzalez BE, bacteremic isolates. In: Proceedings of the 36. Skiest DJ, Brown K, Cooper TW, Hoff- et al. Three-year surveillance of communi- 44th Annual Meeting of the Infectious Dis-

n engl j med 357;4 www.nejm.org july 26, 2007 389 clinical practice

eases Society of America, Toronto, October ditis with linezolid. Scand J Infect Dis Complete genome sequence of USA300, 12–15, 2006. Arlington, VA: Infectious 2005;37:946-9. an epidemic clone of community-acquired Diseases Society of America, 2006:LB11. 64. Wunderink RG, Rello J, Cammarata meticillin-resistant Staphylococcus aureus. abstract. SK, Croos-Dabrera RV, Kollef MH. Linezo Lancet 2006;367:731-9. 61. Markowitz N, Quinn EL, Saravolatz lid vs vancomycin: analysis of two double- 68. Leonard FC, Meticillin-resistant BKM. LD. Trimethoprim-sulfamethoxazole com- blind studies of patients with methicillin- Staphylococcus aureus in animals: a review. pared with vancomycin for the treatment resistant Staphylococcus aureus nosocomial Vet J (in press). of Staphylococcus aureus infection. Ann In- pneumonia. Chest 2003;124:1789-97. 69. Gorwitz RJ, Jernigan DB, Powers JH, tern Med 1992;117:390-8. 65. Stein GE, Craig WA. Tigecycline: a criti- Jernigan JA. Strategies for clinical manage- 62. Ben Mansour EH, Jacob E, Monchi M, cal analysis. Clin Infect Dis 2006;43:518-24. ment of MRSA in the community: sum- et al. Occurrence of MRSA endocarditis 66. Silverman JA, Mortin LI, Vanpraagh mary of an experts’ meeting convened by during linezolid treatment. Eur J Clin Mi- AD, Li T, Alder J. Inhibition of daptomy- the Centers for Disease Control and Pre- crobiol Infect Dis 2003;22:372-3. cin by pulmonary surfactant: in vitro mod- vention. March 2006. (Accessed July 2, 63. Corne P, Marchandin H, Macia JC, eling and clinical impact. J Infect Dis 2007, at http://www.cdc.gov/ncidod/dhqp/ Jonquet O. Treatment failure of methicil- 2005;191:2149-52. pdf/ar/CAMRSA_ExpMtgStrategies.pdf.) lin-resistant Staphylococcus aureus endocar- 67. Diep BA, Gill SR, Chang RF, et al. Copyright © 2007 Massachusetts Medical Society.

collections of articles on the journal’s web site The Journal’s Web site (www.nejm.org) sorts published articles into more than 50 distinct clinical collections, which can be used as convenient entry points to clinical content. In each collection, articles are cited in reverse chronologic order, with the most recent first.

390 n engl j med 357;4 www.nejm.org july 26, 2007 T h e new england journal o f medicine

images in clinical medicine

Metastatic Germ-Cell Cancer

A B

60-year-old man presented with progressive swelling of the Jan Menke, M.D.

right side of theICM scrotum.AUTHOR He Menkereported no historyRETAKE of trauma1st to this area and Eckhardt Grabbe, M.D. no sexually transmittedREG F FIGURE infections. 1a-c The alpha-fetoprotein2nd level was elevated at University Hospital 3rd 37075 Göttingen, Germany A CASE TITLE 3100 μg per liter (normal value, <8), and the serum level ofRevised the beta subunit of human EMail Line 4-C chorionic gonadotropin was less than 1 IU per liter. MultidetectorSIZE computed tomog- Enon ARTIST: mst H/T H/T raphy with multiplanarFILL reformation (PanelC omboA) showed a large28p testicular tumor with lymphangitic spread along the testicularAUTHOR, PLEASE vessels NOTE: to the associated draining lymph nodes below the renal hilus. Figure Right-sidedhas been redrawn hydronephrosis and type has been reset. is present because of com- Please check carefully. pression of the ureter by the lymphatic metastases. Imaging also showed lung metastases (Panel B, arrows) JOB:and ileosacral35624 osseous metastasesISSUE: 6-14-07 (Panel C, long arrows) with infiltration of the iliac muscle (Panel C, short arrow) and the sacral spinal channel. Right-sided radical inguinal orchiectomy was performed, and histologic examination confirmed the diagnosis of advanced nonseminomatous germ-cell cancer. The patient subsequently received chemotherapy, which was complicated by fulminant septic shock in the setting of aplasia, which led to the patient’s death from multiorgan failure. These images show classic lymphatic and hematogenous metastatic pathways of advanced testicular cancer. Copyright © 2007 Massachusetts Medical Society.

n engl j med 357;4 www.nejm.org july 26, 2007 391 T h e new england journal o f medicine

case records of the massachusetts general hospital

Founded by Richard C. Cabot Nancy Lee Harris, m.d., Editor Eric S. Rosenberg, m.d., Associate Editor Jo-Anne O. Shepard, m.d., Associate Editor Alice M. Cort, m.d., Associate Editor Sally H. Ebeling, Assistant Editor Christine C. Peters, Assistant Editor

Case 23-2007: A 9-Year-Old Boy with Bone Pain, Rash, and Gingival Hypertrophy

Christopher P. Duggan, M.D., M.P.H., Sjirk J. Westra, M.D., and Andrew E. Rosenberg, M.D.

Presentation of Case

From the Department of Medicine, Divi- Dr. Laura Chapman (Pediatrics): A 9-year-old boy with autism was admitted to this hos- sion of Gastroenterology and Nutrition, pital because of pain in the hip, refusal to walk, and the recent onset of a rash and Children’s Hospital (C.P.D.); the Depart- ments of Radiology (S.J.W.) and Pathol gingival swelling. ogy (A.E.R.), Massachusetts General Hos- The patient was in his usual state of health until approximately 3 months before pital; and the Departments of Pediatrics admission, when he had an upper respiratory illness with fever; shortly thereafter, (C.P.D.), Radiology (S.J.W.), and Pathol ogy (A.E.R.), Harvard Medical School — he began to have right hip pain, to limp, and to have decreased energy. During the all in Boston. next several weeks, he began to have difficulty climbing stairs, became increasingly irritable, and began moaning in his sleep at night; his appetite decreased, and he lost N Engl J Med 2007;357:392-400. Copyright © 2007 Massachusetts Medical Society. 3.6 kg. Magnetic resonance imaging (MRI) of the hips at another hospital 7 weeks before admission to this hospital reportedly showed multifocal areas of hyperinten- sity in the bones and bone marrow edema. A test for antinuclear antibody factor was negative; results of other laboratory tests are shown in Table 1. One month before admission, he saw an orthopedist at this hospital. On examination, the range of motion of both hips was normal, although he seemed to have pain at the end point of abduction of the right hip. He walked with a wide- based gait. Reflexes were normal. There were no petechiae. Radiographs of the spine and pelvis showed normal alignment of the bones without fracture or dislocation. Consultation with a neurologist and a hematologist were scheduled; however, 4 days later, the symptoms had markedly worsened; he ceased playing and refused to walk or sit up, and his mother noted a rash on his legs. He was admitted to this hospital. He did not have fever, night sweats, dysphagia, nausea, emesis, dyspnea, cough, or change in urinary or bowel habits. At the age of 5 months, computed tomography (CT) of the head had shown communicating hydrocephalus, which had remained stable and had not required surgical intervention. Autism had been diagnosed at the age of 2 years. At the time of admission, his verbal communication was limited to squealing, single words, and echolalia. He had marked anxiety to strangers and engaged in stereotyped behaviors, including head banging. At baseline, his motor function was normal, and he was incontinent of stool and urine. There was a cat in the home, which had not been sick. His parents and older sister were well. There was no exposure to insects and no recent travel. There was

392 n engl j med 357;4 www.nejm.org july 26, 2007 case records of the massachusetts general hospital

Table 1. Results of Laboratory Tests.*

Reference Range, Seven Weeks First Third Test Age-Adjusted† before Admission Admission Admission White-cell count (per mm3) 4,500–13,500 5,590 8,400 10,600 Differential count (%) Neutrophils 33–59 43 48 59 Lymphocytes 33–50 47 42 36 Monocytes 4–11 4 5 4 Eosinophils 0–8 5 5 1 Basophils 0–3 1 0 0 Hematocrit (%) 35.0–45.0 28.9 (reference 32.3 25.8 range, 33.4–40.1) Hemoglobin (g/dl) 11.5–15.5 9.8 11.4 8.4 Erythrocyte count (per mm3) 4.0×106–5.20×106 4.57×106 3.68×106 Mean corpuscular hemoglobin (pg/cell) 25.0–33.0 24.9 22.9 Mean corpuscular volume (μm3) 80–100 71 70 Platelet count (per mm3) 150,000–450,000 311,000 327,000 408,000 Erythrocyte sedimentation rate (mm/hr) 0–17 59 59 95 Rheumatoid factor <12.0 3.9 C-reactive protein (mg/liter) <8.0 24.5 Calcium (mg/dl) 8.5–10.5 9.8 Ferritin (ng/ml) 30–300 46 Iron (µg/dl) 45–160 22 Iron-binding capacity (µg/dl) 228–428 320 Magnesium (mmol/liter) 0.7–1.0 1.35 Phosphorus (mg/dl) 4.5–5.5 4.6

* To convert values for calcium to millimoles per liter, multiply by 0.250. To convert values for iron and iron-binding capacity to micromoles per liter, multiply by 0.1791. To convert values for magnesium to millimoles per liter, multiply by 0.5. To convert values for phosphorus to millimoles per liter, multiply by 0.3229. † Reference values are affected by many variables, including the patient population and the laboratory methods used. Pediatric ranges are estimates derived from a combination of published normal ranges and internal data from the Massachusetts General Hospital for these age groups. a maternal family history of leukemia and breast range of passive motion in all joints of the arms and bone cancer, a paternal family history of and legs, but he grimaced at the end point of ab- leukemia and uterine and bone cancer, and no duction of the left hip. There were no deformities, family history of neurologic or autoimmune dis- redness, or swelling of the joints. Motor strength ease. His only medication was clonidine, and he was 4/5 throughout, and there was no muscular had no known drug allergies. atrophy. He was able to bear weight, but he was On physical examination, the patient was alert, unwilling to bend his knees or hips to sit or to and his vital signs were normal. The pupils were bend over. His gait was wide-based with out- reactive to light, and a dilated funduscopic exami- stretched arms; he walked slowly and reached for nation was normal. The oropharyngeal and nasal support often, but there was no truncal ataxia. mucosa were normal. The lungs and abdomen were Neurologic examination was normal. normal. There was a punctate, erythematous rash Levels of serum electrolytes, lactate dehydro- on the dorsa of the feet that extended to the tops genase, creatine kinase, albumin, and total protein of the thighs and was scattered on the arms and and results of renal- and liver-function tests were legs; it was not raised or excoriated. There was no normal. Levels of C3 and C4 and anticardiolipin tenderness to vertebral palpation. There was full IgM and IgG were normal; tests for antinuclear

n engl j med 357;4 www.nejm.org july 26, 2007 393 T h e new england journal o f medicine

and anti–double-stranded DNA antibodies were On the second hospital day, a dermatology con- negative. Other laboratory values are shown in sultant noted that the lesions on the right leg had Table 1. MRI of the brain on the fourth hospital the appearance of perifollicular petechiae, which day showed ventriculomegaly, unchanged from a were slightly palpable. Additional diagnostic tests study performed 6 years earlier. MRI of the cervi- were performed. cal, thoracic, and lumbar spine showed no abnor- mality. On MRI of the pelvis, T2-weighted images Differential Diagnosis showed multifocal, hyperintense, enhancing lesions throughout the bony pelvis with associated Dr. Christopher P. Duggan: This 9-year-old child with abnormal periosteal enhancement, most striking autism presented with limb pain and a progres- around the acetabula. A bone scan showed mini- sive decrease in ambulation, followed by a rash and mal increased uptake in the right sacroiliac joint. gingival hypertrophy; there were radiographic Acetaminophen with codeine was prescribed, changes of periosteal inflammation and elevated and the patient was discharged on the fifth day to acute-phase reactants. follow up with a hematologist. During the next May we review the radiographic studies? several days, he ceased walking altogether, and Dr. Sjirk J. Westra: Plain radiographs of the pelvis his mother believed that his knees were swollen. and the thoracolumbar spine were normal. On He was readmitted 4 days later. On the second T1-weighted MRI, there was normal architecture hospital day, biopsies of iliac crest bone were per- of the bone and no gross destructive lesions within formed under anesthesia; examination of the the pelvic skeleton. On fat-suppressed, T2-weighted specimens disclosed apophyseal cartilage under- imaging with short-inversion-time inversion recov- going endochondral ossification, cancellous bone, ery (STIR), there were multifocal areas of high and hematopoietic marrow with focal edema; flow T2-weighted signal in the bone marrow of both cytometry showed no abnormal T cells or B cells. sacral wings and in the iliac bones around both Synovial fluid aspirated from the left knee con- sacroiliac joints (Fig. 1A) and around the triradiate tained 917 white cells, 13% polymorphonuclear cartilages (Fig. 1B). These signal changes are leukocytes, 29% lymphocytes, no blasts, and no indicative of bone marrow edema, located in the red cells, and there was no bacterial growth on metaphyseal equivalents of the pelvic bones. In ad- culture. Indomethacin (at a dose of 3 mg per kilo- dition, there was abnormal T2-weighted signal in gram of body weight per day) was begun; his gait the periosteum of the left iliac bone (Fig. 1B) and improved, and he was discharged home with a in the left pubic bone adjacent to the hip joint. walker on the fourth day. There were no soft-tissue abnormalities. Fat-sup- During the next 10 days, his mother noted gin- pressed T1-weighted images after the administra- gival swelling with bleeding, and the rash be- tion of gadolinium showed abnormal enhancement came confluent. He was readmitted to this hospi- of all lesions and abnormal periosteal enhance- tal. The vital signs were normal. The left maxillary ment in the posterior aspect of the left acetabu- gingiva was hypertrophic, covered the molars, lum. A radionuclide bone scan showed only slight- and protruded into the mouth. The right mandibu- ly increased uptake over the right sacroiliac joint lar gingiva was hypertrophic and had a deep pur- but was otherwise normal for the patient’s age. ple discoloration. There were a few palatal pete- The differential diagnosis of multifocal lesions chiae. There was a diffuse palpable petechial rash in the metaphyseal equivalents of the pelvic bones over the legs and feet and two 10-mm purpuric includes multifocal osteomyelitis or a hemato- lesions on the left foot. There was full passive logic cancer. We considered Langerhans-cell his- range of motion of all joints except for the right tiocytosis, although the absence of lytic lesions hip, which could not be flexed or abducted with- on the plain radiographs would argue against this out causing pain. He was unable to bear weight on diagnosis. The lack of soft-tissue edema and the the right leg. The remainder of the general physical relatively normal bone scan argue against multi- and neurologic examination was unchanged from focal osteomyelitis. Because of the predilection of the previous admission. Levels of electrolytes, the abnormalities to the metaphyseal equivalents, renal- and liver-function tests, coagulation studies, where bone growth normally takes place, we con- and immunoglobulins were normal. Additional sidered metabolic disorders. laboratory-test results are shown in Table 1. Dr. Zachary M. Grinspan (Pediatrics): The patient

394 n engl j med 357;4 www.nejm.org july 26, 2007 case records of the massachusetts general hospital

A A

Figure 2. Clinical Images of the Patient. Petechiae are present on both legs up to the thigh (Panel A). The petechiae are localized around hair fol liclesICM (inset),AUTHOR and hairless Duggan areas are spared.RET AAKE photo1st- graphREG obtained F FIGURE during 2a&b examination of 4 of the oral cavity,2nd CASE 3rd with the useTITLE of a cheek and lip retractor whileRevised the patient wasEMail under general anesthesia,Line shows 4-Cthe gingival lesions SIZE Figure 1. MRI of the Pelvis. (PanelEnon B). ARTISTThe right: mst lower, rightH/T upper,H/T and left upper 16p6 Coronal images obtained with the STIR technique gingivaFILL (arrows) are swollen,C andombo the gingiva bleed easily. (short-inversion-time inversion recovery) demonstrate AUTHOR, PLEASE NOTE: fociICM of increasedAUTHOR T -weightedDuggan signal in theRET boneAKE mar1st- Figure has been redrawn and type has been reset. 2 Please check carefully. rowREG in theF FIGUREmetaphyseal 1a&b ofanalogues 4 of the pelvis, nota2nd- 3rd bly CASEadjacentTITLE to both sacroiliac joints (Panel A) and the Revised both JOB:localized 35704 and systemic disordersISSUE: 7-26-07 (Table 2). triradiateEMail cartilages of both Lineacetabula4-C (Panel B). There SIZE Several features of this case allow us to limit pos- is high-intensityEnon ARTIST :signalmst in theH/T periosteumH/T of the left FILL 16p6 sible diagnoses. Child abuse or other trauma can iliac bone (Panel B, arrow). Combo AUTHOR, PLEASE NOTE: be observed in patients with developmental delay, Figure has been redrawn and type has been reset. Please check carefully. but the bilateral radiographic abnormalities argue had a petechial rash on both legs up to his thighs, against isolated trauma as a cause of his pain. In- whichJOB: localized 35704 around the hairISSUE: follicles,7-26-07 spar- fectious causes including septic arthritis and os- ing hairless areas of his ankles (Fig. 2A). He also teomyelitis are less likely, since the bone lesions had two purpuric lesions (7 to 8 mm in diame- are multifocal and there was no recurrent fever. ter) on his left lateral malleolus. General anesthe- A normal neuromuscular examination reduces the sia was administered, and the trachea was intu- likelihood of myopathies and neuropathies. Leu- bated for a fluoroscopically guided corticosteroid kemia or lymphoma can present with limb pain injection of the sacroiliac joint and biopsy of the or limp, and the white-cell count may be normal, iliac bone. A cheek and lip retractor provided vi- but the bone marrow biopsy did not reveal a ma- sualization of the gingival lesions. The right lower, lignant tumor. Endocrinopathies, including diabe- right upper, and left upper gingiva were swollen, tes mellitus and hyperparathyroidism, are not and the gingiva bled easily (Fig. 2B). suggested by the clinical presentation or labora- Dr. Duggan: The differential diagnosis of ac- tory studies. The imaging studies do not suggest a quired limb pain, limp, or both in a child includes bone tumor.

n engl j med 357;4 www.nejm.org july 26, 2007 395 T h e new england journal o f medicine

Several features of this child’s presentation Table 2. Differential Diagnosis of Acquired Limp or Limb Pain. are consistent with spondyloarthropathy, includ- Localized Diseases Systemic Diseases ing his male sex, young age (although presenta- Infection Connective-tissue disorders tion is typically in the second decade of life), Osteomyelitis Dermatomyositis gradual onset of symptoms, and elevated eryth- Septic arthritis Henoch–Schönlein purpura Myositis Inflammatory bowel disease rocyte sedimentation rate. The history does not Lyme disease Kawasaki’s disease note morning stiffness or improvement over the Chronic recurrent multifocal Polyarteritis nodosa course of the day, which are common in spondy- osteomyelitis Systemic lupus erythematosus Trauma Juvenile rheumatoid arthritis loarthropathies. In addition, knee or other joint Chondroma or sarcoma Spondyloarthropathy swelling was not noted on physical examination. Bone tumors Endocrinopathy The absence of joint swelling, warmth, and red- Osteoblastoma Diabetes Ewing’s or other sarcoma Hyperparathyroidism ness argues against a diagnosis of rheumatic Other disorders Hematologic diseases disease. Legg–Calvé–Perthes disease Sickle cell disease Toxic synovitis Hemoglobinopathy Conditions with altered intestinal Leukemia Lymphoma permeability Neuropathies Arthritis is a common complication of diseases Guillain–Barré syndrome Heavy-metal or vitamin A intoxication with altered intestinal permeability, including Vitamin deficiency celiac disease, infectious enteropathies, and idio- Rickets pathic inflammatory bowel disease. Could this Scurvy child have involvement of the sacroiliac joint as a component of undiagnosed colitis or enteritis? Chronic recurrent multifocal osteomyelitis A possible link between gastrointestinal diseases Chronic recurrent multifocal osteomyelitis is an and autism has been pursued for many years but inflammatory condition that often affects the me- never proved. Studies have suggested associa- taphyses of long bones and has a remitting and tions with celiac disease, diarrhea, abdominal relapsing course.1 It is characterized by the insidi- pain, gastroesophageal reflux, and increased out- ous onset of local joint swelling and pain, and input of pancreaticobiliary fluid after secretin stim- 5,6 creased signal intensity on T2-weighted images ulation. One report suggested that secretin are seen with MRI.2 Girls are affected more often infusion7 led to improvement in symptoms of than boys, and the disorder has an overall favor- autism, but controlled clinical studies subsequent- able prognosis.3 Chronic recurrent multifocal os- ly disproved its efficacy.8,9 Another study10 impli- teomyelitis could explain several of the findings in cated exposure to measles–mumps–rubella (MMR) this case, but it remains a diagnosis of exclusion vaccine, which led to a fall in rates of MMR vac- since there are no pathognomonic findings. Once cination in several countries. This study was later the rash and gingival lesions appeared, this diag- retracted,11 and a causal relationship between nosis became less likely. MMR vaccination and autism was subsequently refuted.12 Although gastrointestinal symptoms in Rheumatologic diseases children with autism can be subtle, this child had Juvenile rheumatoid arthritis can present with limb no history of abdominal pain, diarrhea, abdominal pain and difficulty walking. The negative rheuma- distention, or other symptoms that would suggest toid factor does not rule out rheumatologic condi- gastrointestinal disease. tions since only a minority of children with juvenile Lost in the controversy about the purported rheumatoid arthritis are seropositive at presenta- relationship between autism and MMR vaccination. Spondyloarthropathies are a family of related tion was the finding that urinary levels of meth- disorders that include ankylosing spondylitis, Reit- ylmalonic acid were higher in the children with er’s syndrome, reactive arthritis, psoriatic arthritis, autism than in age-matched controls, a finding 10 spondyloarthropathy associated with inflamma- consistent with vitamin B12 deficiency. Indeed, tory bowel disease, undifferentiated spondyloar- a number of studies have shown that children thropathy, Whipple’s disease, and Behçet’s disease. with autism, who characteristically consume a In children, spondyloarthropathies are the second diet highly restricted in color and consistency, most commonly diagnosed rheumatologic condi- are at risk for micronutrient deficiencies, includ- tion, after juvenile rheumatoid arthritis.4 ing deficiencies of vitamin A,13 vitamin D,14 and

396 n engl j med 357;4 www.nejm.org july 26, 2007 case records of the massachusetts general hospital vitamin C.15 Could this patient’s multiorgan pre- tamin is generally recommended. This child’s sentation (joints, skin, and oral mucosa) be relat symptoms began after an upper respiratory infec- ed to a vitamin deficiency? When a much smaller tion; possibly the increased metabolic needs asso- differential-diagnosis list that might explain his ciated with this infection unmasked a subclinical gingival lesions is considered (Table 3), vitamin C vitamin C deficiency. deficiency is the one diagnosis that could explain Dr. Nancy Lee Harris (Pathology): Dr. Whelan and all of his symptoms. Dr. Seashore, would you comment on your im- pressions when you saw this child? Scurvy Dr. J. Patrick Whelan (Pediatric Rheumatology): Scurvy is one of the earliest recorded diseases in This patient presented with hip pain that was humans, first described in the Ebers papyrus in striking, despite a relatively normal examination. 1550 b.c. Its successful treatment with oranges Compression of the pelvis, which puts pressure on and lemons was established by Scottish surgeon the sacroiliac joints, elicited pain, suggesting the James Lind in one of the first recorded con- sacroiliac joints as a source of the pain. The ele- trolled clinical trials, published in 1753. Depri- vated inflammatory indexes, including C-reactive vation of vitamin C, named ascorbic acid because protein and erythrocyte sedimentation rate, sug- of its antiscorbutic properties, was subsequently gested an inflammatory cause. Since pain in the shown to produce symptoms of scurvy after 30 sacroiliac joint may not respond to naproxen, I be to 40 days.16,17 gan indomethacin, another good treatment for Ascorbic acid is a reversible reducing agent this type of pain. In retrospect, the indomethacin that is an essential cofactor for the hydroxylation may have precipitated the gingival hemorrhage of proline to hydroxyproline in collagen synthesis and the petechial rash, both of which developed and for the hydroxylation of the neurotransmitter quite dramatically during the 10 days after start- dopamine to noradrenaline.18 Many signs and ing the drug. symptoms of vitamin C deficiency relate to its es- Dr. Carl Seashore (Pediatrics): The medical team sential role in collagen synthesis, including derma- on the child’s third hospitalization had the bene tologic manifestations of petechia, ecchymoses, fit of having watched the results of his physical corkscrew hairs, and hyperkeratosis. Perifollicular examinations change over time. The development hemorrhages, which were seen in this case, are of skin lesions and gingival swelling raised the particularly characteristic. Systemic symptoms in- suspicion of vitamin C deficiency. The medical clude lassitude and fatigue (as was noted in this team reinterviewed the patient’s mother, with the case), and neurologic symptoms can include de- patient present, specifically about the patient’s pression and vasomotor instability. Gingival swell- dietary history. We learned that the patient had ing and hemorrhage and bone disease due to increasingly limited his diet since the onset of subperiosteal bleeding are common and were his bony pain several months earlier, and for the manifested in this patient.19 past month or two had consumed only toaster Why did scurvy develop in this child at this pastries and cola drinks, neither of which con- time? Many children with autism have limited di- tain vitamin C; he refused to eat fruits or vege- etary intake of energy, macronutrients and micro- tables, drank no juice, and did not take a multivi- nutrients, which is why a daily complete multivi- tamin. We obtained plain radiographs of the hands and feet, which we thought might show Table 3. Differential Diagnosis of Gingival Swelling. more classic diagnostic features of vitamin C deficiency. The skin lesions were also biopsied. Phenytoin exposure Dr. Harris: Dr. Westra, will you show us the Pyogenic granuloma bone images? Aphthous ulcers Dr. Westra: It is ironic that in this hospital, where Infectious gingivitis we may begin with the most advanced techniques, Crohn’s disease MRI and CT, the diagnosis in this case was made on plain films. On the hand film (Fig. 3A), this Behçet’s disease patient’s bone age was 1 year behind his chrono- Dental abscess logic age, and there was widening of and irregu- Scurvy larity around the growth plate. Most of the growth of the skeleton takes place around the

n engl j med 357;4 www.nejm.org july 26, 2007 397 T h e new england journal o f medicine

Figure 3. Plain Radiographs of the Hands and Knees. A radiograph of the left wrist (Panel A) shows irregu- A larity with widening of the distal ulnar physis (arrow). However, there is normal mineralization of the zone of provisional calcification on the metaphyseal side of the growth plates and surrounding the epiphyses. (The curved band is a tube outside the patient’s hand.) A radiograph of the right knee (Panel B) shows additional findings typical of scurvy: metaphyseal irregularities with spurring (Pelkan’s sign, black arrows); white lines surrounding the epiphyses (Wimberger’s sign), indicative of osteoporosis; a white line of Frankl in the zone of provisional calcification (white arrowhead) with a lucent line immediately below this (Trummer- feld zone or scurvy line, black arrowheads); and peri osteal reactions along the metaphyses (white arrows). The estimated bone age is 2 years behind the patient’s chronologic age.

knee; this patient was delayed 2 years in skeletal maturation in the knee. There is osteoporosis of the epiphysis, which is surrounded by a scle- rotic ring (ring sign). Magnified views (Fig. 3B) B reveal periosteal elevations due to subperiosteal hemorrhages and irregularities, fragmentation, and spurs at the margins of the metaphysis (Pel- kan’s sign), which is typical of scurvy, especially in the healing phase. These irregularities can mimic the metaphyseal corner fractures seen in cases of child abuse. There is a dense zone of provisional calcification at the margins of the growth plate (white line, or Fränkel’s sign), and immediately below that a lucent line (Trummer- feld zone or scurvy line), which is pathognomonic of scurvy.19 These radiologic findings only become manifest after 3 to 6 months of nutritional deficiency. Dr. Andrew E. Rosenberg: Can you rule out a component of rickets? Dr. Westra: There was good mineralization at the ends of the bones; in rickets there would be a lack of mineralization in that area.

Clinical Diagnosis

Scurvy.

DR. CHRISTOPHER P. DUGGAN’S DIAGNOSIS

Scurvy.

ICM AUTHOR Duggan RETAKE 1st REG F FIGURE 3a&b of 4 2nd CASE 3rd TITLE Revised EMail Line 4-C SIZE 398 n engl j med 357;4 www.nejm.orgEnon julyARTIST 26,: 2007mst H/T H/T FILL Combo 16p6 AUTHOR, PLEASE NOTE: Figure has been redrawn and type has been reset. Please check carefully.

JOB: 00000 ISSUE: 00-00-00 case records of the massachusetts general hospital

Pathological Discussion A Dr. Rosenberg: The specimen of the bone biopsy from the iliac crest performed on an earlier admission (Fig. 4A) contained marrow that was abnormal, with numerous extravasated red cells, edema, and early fibrosis; these correlate with the abnormal findings seen on MRI and CT scanning. The skin biopsy performed on the current admission appears relatively normal at first glance; however, extravasated red cells are present in the dermis around hair follicles (Fig. 4B). This finding indicates recent bleeding and underlies the perifollicular petechiae noted in this patient. The B pathologic alterations of the skin and bone are non specific, but in conjunction with the other clinical findings, they support the diagnosis of scurvy. The manifestations of scurvy in this patient were caused by deficient production of collagen in connective tissue,15,17,19 particularly that of the supportive sheaths around small blood vessels, and of rapidly growing bone. The skeletal changes in scurvy are most severe in young children such as this patient, because their skeletal tissues are still growing, and the periosteum is not as tightly bound to the surface of the cortex as it is Figure 4. Biopsy Specimens of Bone Marrow and Skin. in adults. Deficient collagen in subperiosteal The bone-biopsy specimen from the iliac crest performed at the time of the second hospital admission blood vessels leads to rupture and hemorrhage, includesICM theAUTHOR hyaline Duggan cartilage of the apophysis,RETAKE the 1st which mechanically lifts the periosteum from the underlyingREG F FIGURE primary 4a&b and ofsecondary 4 spongiosa, and the2nd CASE 3rd underlying cortex. Reparative osteoblasts deposit interveningTITLE marrow (Panel A, hematoxylin andRevised eosin). reactive bone, which can be seen on plain radiog- TheEMail marrow is abnormal, withLine edema,4-C numerous ex- Enon SIZE raphy, as in this case. Similar findings occur in travasatedARTIST red cells: mst (indicatingH/T acuteH/T marrow hemorrhage),FILL and focal fibrosis. TheCombo skin-biopsy specimen16p6 the metaphysis at the base of the growth plate; (Panel B, hematoxylinAUTHOR, and PLEASE eosin) NOTE:includes the stratified reduced collagen production results in decreased keratinized Figure squamoushas been redrawn epithelium, and type papillary has been dermis,reset. Please check carefully. bone deposition, in structural weakness, and in reticular dermis, and the superficial portion of the subcutis that contains scattered adnexal structures. both hemorrhage and fractures with minimal JOB: 35704 ISSUE: 7-26-07 stress. With treatment, these pathologic changes of scurvy will completely resolve. Dr. Seashore: The serum vitamin C level was less fortable, began sitting up in bed, and was even than 0.12 mg per deciliter (normal range, 0.20 to able to bear weight on his legs. The next day he 1.90 [7 μmol per liter; normal range, 11 to 108]). was discharged home with physical therapy, and The levels of 25-hydroxyvitamin D and parathy- at a follow-up visit 1 month later, he showed con- roid hormone were also low at 13 ng per mil- tinuing improvement. liliter (normal range, 20 to 100) and 9 pg per milliliter (normal range, 10 to 60), but the 1,25- PATHOLOGICAL DIAGNOSIS dihydroxyvitamin D level was normal, at 10 pg per milliliter (normal range, 6 to 62). Vitamin C, Scurvy. at a dose of 160 mg daily, and a pediatric multi- Dr. Duggan reports receiving consulting fees from and serving vitamin were begun on the third hospital day, and on the paid advisory board of Groupe Danone. No other poten- by the fourth day the patient appeared more com- tial conflict of interest relevant to this article was reported.

n engl j med 357;4 www.nejm.org july 26, 2007 399 case records of the massachusetts general hospital

References 1. Robertson LP, Hickling P. Chronic re- spectrum disorders. J Assoc Acad Minor an autistic child. Arch Ophthalmol 1998; current multifocal osteomyelitis is a dif- Phys 1998;9:9-15. 116:392-3. ferential diagnosis of juvenile idiopathic 8. Sandler AD, Sutton KA, DeWeese J, 14. Clark JH, Rhoden DK, Turner DS. arthritis. Ann Rheum Dis 2001;60:828-31. Girardi MA, Sheppard V, Bodfish JW. Symptomatic vitamin A and D deficien- 2. Jurik AG, Egund N. MRI in chronic Lack of benefit of a single dose of syn- cies in an eight-year-old with autism. JPEN recurrent multifocal osteomyelitis. Skele- thetic human secretin in the treatment of J Parenter Enteral Nutr 1993;17:284-6. tal Radiol 1997;26:230-8. autism and pervasive developmental dis- 15. Monks G, Juracek L, Weigand D, 3. Huber AM, Lam PY, Duffy CM, et al. order. N Engl J Med 1999;341:1801-6. Magro C, Cornelison R, Crowson AN. Chronic recurrent multifocal osteomyeli- 9. Lightdale JR, Hayer C, Duer A, et al. A case of scurvy in an autistic boy. J Drugs tis: clinical outcomes after more than five Effects of intravenous secretin on lan- Dermatol 2002;1:67-9. years of follow-up. J Pediatr 2002;141:198- guage and behavior of children with 16. Hodges RE, Baker EM, Hood J, Sauber 203. autism and gastrointestinal symptoms: lich HE, March SC. Experimental scurvy 4. Denardo BA, Tucker LB, Miller LC, a single-blinded, open-label pilot study. in man. Am J Clin Nutr 1969;22:535-48. Szer IS, Schaller JG. Demography of a re- Pediatrics 2001;108(5):e90. 17. Baker EM, Hodges RE, Hood J, Sauber gional pediatric rheumatology patient pop 10. Wakefield AJ, Murch SH, Anthony A, lich HE, March SC, Canham JE. Metabo- ulation. J Rheumatol 1994;21:1553-61. et al. Ileal-lymphoid-nodular hyperplasia, lism of 14C- and 3H-labeled L-ascorbic 5. Horvath K, Papadimitriou JC, Rabsz- non-specific colitis, and pervasive devel- acid in human scurvy. Am J Clin Nutr tyn A, Drachenberg C, Tildon JT. Gastro- opmental disorder in children. Lancet 1971;24:444-54. intestinal abnormalities in children with 1998;351:637-41. 18. Jacob R. Vitamin C. In: Shils ME, autistic disorder. J Pediatr 1999;135:559- 11. Murch SH, Anthony A, Casson DH, et Shike M, Ross AC, eds. Modern nutrition 63. al. Retraction of an interpretation. Lancet in health and disease. 9th ed. Baltimore: 6. Asperger H. Psychopathology of chil- 2004;363:750. Lippincott Williams & Wilkins, 1999:467- dren with coeliac disease. Ann Paediatr 12. Madsen KM, Hviid A, Vestergaard M, 83. 1961;197:346-51. (In German.) et al. A population-based study of mea- 19. Weinstein M, Babyn P, Zlotkin S. An 7. Horvath K, Stefanatos G, Sokolski KN, sles, mumps, and rubella vaccination and orange a day keeps the doctor away: scur- Wachtel R, Nabors L, Tildon JT. Improved autism. N Engl J Med 2002;347:1477-82. vy in the year 2000. Pediatrics 2001;108(3): social and language skills after secretin 13. Steinemann TL, Christiansen SP. Vi- E55. administration in patients with autistic tamin A deficiency and xerophthalmia in Copyright © 2007 Massachusetts Medical Society.

Lantern Slides Updated: Complete PowerPoint Slide Sets from the Clinicopathological Conferences Any reader of the Journal who uses the Case Records of the Massachusetts General Hospital as a teaching exercise or reference material is now eligible to receive a complete set of PowerPoint slides, including digital images, with identifying legends, shown at the live Clinicopathological Conference (CPC) that is the basis of the Case Record. This slide set contains all of the images from the CPC, not only those published in the Journal. Radiographic, neurologic, and cardiac studies, gross specimens, and photomicrographs, as well as unpublished text slides, tables, and diagrams, are included. Every year 40 sets are produced, averaging 50-60 slides per set. Each set is supplied on a compact disc and is mailed to coincide with the publication of the Case Record. The cost of an annual subscription is $600, or individual sets may be purchased for $50 each. Application forms for the current subscription year, which began in January, may be obtained from the Lantern Slides Service, Department of Pathology, Massachusetts General Hospital, Boston, MA 02114 (telephone 617-726-2974) or e-mail [email protected].

400 n engl j med 357;4 www.nejm.org july 26, 2007 T h e new england journal o f medicine

e di t or i a l s

Therapy for Bronchiolitis: When Some Become None Caroline Breese Hall, M.D.

In past centuries, infant mortality in the first year knowledged benefit of diminishing the respiratory of life was described as the greatest plague.1 In distress from other obstructive airway illnesses, fants in the 21st century continue to be plagued such as asthma.11 Furthermore, corticosteroids by high rates of hospitalization, and bronchiolitis may affect the pathogenesis of bronchiolitis at is the number one cause.2,3 Despite therapeutic tributed to respiratory syncytial virus. Both clin advancements, hospitalizations for bronchiolitis ical and experimental studies indicate that an have continued to increase. The annual rate of altered and augmented immune response of the emergency department visits for bronchiolitis dur young infant contributes to the phenotypic ex ing the season of respiratory syncytial virus, the pression and severity of respiratory syncytial vi major cause of this condition, has been estimated rus disease and may predispose the infant to as 22.8 per 1000 for children 12 months of age or pulmonary sequelae.7,8 The variable compliance younger.2 Yearly costs for bronchiolitis hospitali with the recommendations against the routine zations exceed $700 million. Yet even these strik use of corticosteroids may also relate to the ing figures underestimate the health care burden conflicting results of the studies.5,6 They have imposed by bronchiolitis. They do not reflect the been heterogeneous in design and generally in costs associated with the many more frequent of volved small numbers of patients, thus reducing fice visits of less ill children or with the second the strength of their evidence-based recommen ary infections among high-risk family contacts.4 dations. A better indicator of the need for control of In this issue of the Journal, Corneli and col bronchiolitis is the increasing accumulation of leagues12 report on a randomized trial examining therapeutic studies and the continued conun whether dexamethasone administered to previous drums and controversies they engender.5,6 The ly healthy infants presenting with a first episode clinical manifestations, course, and pathology of of wheezing diagnosed as moderate-to-severe bronchiolitis have been well described, but the bronchiolitis reduces the need for hospitalization. pathogenesis has not. Whether the pulmonary The results, although negative, are important and disease results from the acute lytic effects of the will probably be accompanied by close scrutiny virus, the infant’s immune response, or both re of the study’s design and the extent of its clini mains unclear.7,8 This may partially explain why cal relevance. effective interventions remain elusive. The study, which involved 20 emergency de Current therapeutic options for bronchiolitis partments, was designed to avoid many of the are few — primarily antiviral drugs, bronchodila deficits of previous corticosteroid trials. It was tors, and corticosteroids — and all are controver conducted over three respiratory seasons, and sial. Not one is routinely recommended.5,6 Yet personnel at each site received yearly training in one or more of these agents, as well as antibi study procedures. The infants in the trial received otics, are administered to as many as 50 to 80% a single oral dose of dexamethasone (1 mg per of infants with bronchiolitis admitted to hospi kilogram of body weight) or placebo, and during tals in the United States and other countries.9,10 the next 4 hours their respiratory status was as Part of the rationale for their use is their ac sessed with use of a standardized scoring instru

402 n engl j med 357;4 www.nejm.org july 26, 2007 editorials ment. The study is singular in the number of in emergency departments within these age groups fants enrolled (600) and in the broad spectrum of would be of interest. The majority of infants hos contributing emergency departments. pitalized with bronchiolitis are less than 6 months Nonetheless, variations in practice among the of age, and most are 2 to 3 months old.3,14 These sites probably existed. The primary outcome — youngest infants are the most likely to have acute the decision to hospitalize or discharge the infant complications, such as apnea, and altered respons — was determined by individual physicians, but es to corticosteroid therapy. Children older than it was corroborated as a valid measure of the 1 year of age, although less likely to be hospital effect of dexamethasone by the secondary out ized because of their larger airways and more come — the change in the respiratory status competent immune systems, continue to have after 4 hours as measured by the Respiratory As considerable rates of hospitalization.3,15 The char sessment Change Score. No significant difference acteristics of bronchiolitis among these younger was found between the infants treated with the and older children may differ from those of the corticosteroid and those receiving placebo. The illness as it occurs in children between 2 and 12 trial was designed to closely duplicate the meth months of age. For this very reason, their exclu ods of a previous controlled study of 70 infants sion from the study is warranted, but applying that demonstrated, in contrast with this trial, sig the findings of the study to these children is nificant benefit from the oral administration of not justified. dexamethasone as compared with placebo.13 How An important question is whether corticoste ever, the degree of change in score that is pre roid administration diminishes the risk of the re dictive of a relevant change in clinical status has current wheezing that develops in 20 to 50% of not been determined. The short period of obser infants diagnosed with bronchiolitis.7,8 Although vation also potentially limits the study’s value for these episodes of recurrent wheezing are most predicting the subsequent prognosis. frequent in the first several years after an infant The strict criteria for inclusion in the study contracts bronchiolitis, they have been linked to by Corneli et al. enhanced the uniformity of the later occurrence of asthma and chronic lung the enrolled population, but these criteria also abnormalities.7,8 The effect of corticosteroid limited the study’s representation of the overall therapy on the development of later pulmonary population of infants presenting with bronchi dysfunction is unresolved. Among the limited olitis. What should be considered, therefore, is number of studies examining this issue, a few whether the results of this carefully defined seg have suggested benefit, but more have not.16 ment of bronchiolitis cases can and should be Follow-up of the children enrolled in the Cor applied to the treatment of other infants with a neli study could extend its value and answer a first episode of bronchiolitis. If not, will the con major clinical question. clusions be clinically and economically relevant? These questions and the potential limitations Of 8686 patients initially assessed, 93% were of the study should not diminish recognition of considered ineligible, 91% of whom did not meet the soundness of the study’s results and their the inclusion criteria. A notably large proportion, clinical pertinence. The findings are bolstered by 41%, was excluded because the infants had had the study’s size and the controlled, randomized a previous episode of wheezing; one fourth were design. The conclusions are corroborated by not considered ill enough to meet the criteria for the additional demonstration that the findings moderate-to-severe bronchiolitis. Still, these in were independent of site and extended beyond fants were judged by their parents or primary the 4 hours of initial observation, according to physician to be ill enough to require evaluation an interview with the parent or guardian and in an emergency department, and they would chart review. Corticosteroid therapy did not sig make up an even larger proportion of the bron nificantly affect the duration of hospitalization chiolitis cases seen in office practices than the or the need for subsequent medical visits or re infants who were enrolled in the study. admission. Infants younger than 2 months and children Despite the value of this study, the long his older than 12 months were also excluded from tory of therapies and recommendations attending the study. Knowing the proportion of children bronchiolitis suggest that the study’s results will with bronchiolitis presenting to the participating not appreciably change the nature of the care pro

n engl j med 357;4 www.nejm.org july 26, 2007 403 T h e new england journal o f medicine

vided by the primary physician faced with a young, cal and research approaches. Pediatr Infect Dis J 2003;22:Suppl: S58-S65. distressed infant and anxious parents. Withhold 8. Martinez FD. Respiratory syncytial virus bronchiolitis and ing therapy is much more difficult than giving it. the pathogenesis of childhood asthma. Pediatr Infect Dis J 2003; 22:Suppl:S76-S82. No potential conflict of interest relevant to this article was re 9. Behrendt CE, Decker MD, Burch DJ, Watson PH. Interna ported. tional variation in the management of infants hospitalized with respiratory syncytial virus. Eur J Pediatr 1998;157:215-20. From the Department of Infectious Diseases, University of Roch- 10. Christakis DA, Cowan CA, Garrison MM, Molteni R, Mar ester School of Medicine and Dentistry, Rochester, NY. cuse E, Zerr DM. Variation in inpatient diagnostic testing and management of bronchiolitis. Pediatrics 2005;115:878-84. 1. Knapp VJ. Major medical explanations for high infant mor 11. Rowe BH, Spooner C, Ducharme FM, Bretzlaff JA, Bota GW. tality in nineteenth-century Europe. Can Bull Med Hist 1998;15: Early emergency department treatment of acute asthma with 317-36. systemic corticosteroids. Cochrane Database Syst Rev 2001;1: 2. Leader S, Kohlhase K. Recent trends in severe respiratory CD002178. syncytial virus (RSV) among US infants, 1997-2000. J Pediatr 12. Corneli HM, Zorc JJ, Mahajan P, et al. A multicenter, random 2003;143:Suppl:S127-S132. ized, controlled trial of dexamethasone for bronchiolitis. N Engl 3. Shay DK, Holman RC, Newman RD, Liu LL, Stout JW, Ander J Med 2007;357:331-9. son LJ. Bronchiolitis-associated hospitalizations among US chil 13. Schuh S, Coates AL, Binnie R, et al. Efficacy of oral dexa dren, 1980-1996. JAMA 1999;282:1440-6. methasone in outpatients with acute bronchiolitis. J Pediatr 4. Falsey AR, Hennessey PA, Formica MA, Cox C, Walsh EE. 2002;140:27-32. Respiratory syncytial virus infection in elderly and high-risk 14. Parrott RH, Kim HW, Arrobio JO, et al. Epidemiology of adults. N Engl J Med 2005;352:1749-59. respiratory syncytial virus infection in Washington, D.C. II. Infec 5. American Academy of Pediatrics Subcommittee on Diagno tion and disease with respect to age, immunologic status, race sis and Management of Bronchiolitis. Diagnosis and management and sex. Am J Epidemiol 1973;98:289-300. of bronchiolitis. Pediatrics 2006;118:1774-93. 15. Nicholson KG, McNally T, Silverman M, Simons P, Stockton 6. Management of bronchiolitis in infants and children. Evi JD, Zambon MC. Rates of hospitalisation for influenza, respira dence report/technology assessment. No. 69. Rockville, MD: tory syncytial virus and human metapneumovirus among infants Agency for Healthcare Research and Quality, January 2003:1-5. and young children. Vaccine 2006;24:102-8. (AHRQ publication no. 03-E009.) 16. Gern JE. Mechanisms of virus-induced asthma. J Pediatr 2003; 7. Openshaw PJ, Dean GS, Culley FJ. Links between respira 142:Suppl:S9-S14. tory syncytial virus bronchiolitis and childhood asthma: clini Copyright © 2007 Massachusetts Medical Society.

Network Medicine — From Obesity to the “Diseasome” Albert-László Barabási, Ph.D.

A recent study reported that among people who Study, making use of the fact that the participants carried a single copy of the high-risk allele for had been asked to name their friends to facilitate the FTO gene, which is associated with fat mass follow-up in the study. The authors observed that and obesity, the risk of obesity increased by 30%. when two persons perceived each other as friends, The risk of obesity increased by 67% among peo if one friend became obese during a given time ple who carried two alleles, and on average they interval, the other friend’s chances of following gained 3.0 kg (6.6 lb) or more.1 Given that ap suit increased by 171%. Among pairs of adult sib proximately one sixth of the population of Euro lings, if one sibling became obese, the chance pean descent is homozygous for this allele, this that the other would become obese increased by link between the FTO gene and obesity appears 40%. The results of this study also indicate that to be one of the strongest genotype–phenotype obesity is clustered in communities. For example, associations detected by modern genome-screen the risk that the friend of a friend of an obese ing techniques. person would be obese was about 20% higher in That obesity has a genetic component is not the observed network than in a random network; surprising: researchers have long known that it this effect vanished only by the fourth degree of often runs in families. In this issue of the Journal, separation. Christakis and Fowler suggest that friends have In the past 7 years, our understanding of net an even more important effect on a person’s risk works has undergone a revolution because of the of obesity than genes do.2 The authors recon emergence of a new array of theoretical tools structed a social network showing the ties between and techniques for mapping out real networks. friends, neighbors, spouses, and family members These advances have included some surprises in among participants of the Framingham Heart dicating that most real networks in technologi

404 n engl j med 357;4 www.nejm.org july 26, 2007 editorials cal, social, and biologic systems have common gins. Human diseases, therefore, themselves form designs that are governed by simple and quanti a network in which two diseases are connected fiable organizing principles.3 The growing inter if they share at least one gene.7 In this disease est in interconnectedness has brought into focus network, obesity has links to seven diseases, in an often ignored issue: networks pervade all as cluding asthma, lipodystrophy, and glioblastoma pects of human health. One example of this trend (Fig. 1). Thus, the network concept reveals a num involves social networks and their impact on the ber of surprising connections between diseases, spread of obesity or pathogens — from influenza forcing us to rethink the way in which we classify to the severe acute respiratory syndrome or the hu and separate them. man immunodeficiency virus. The role of neural In the long run, networks may affect all as networks in various psychiatric and neurodegen pects of medical research and practice.8 Indeed, erative diseases is another example. In fact, net the fundamental question of where function lies work analysis is poised to play the biggest role within a cell is slowly shifting from a single- at the cellular level,4 since most cellular compo minded focus on genes to the understanding that nents are connected to each other through intri behind each cellular function there is a discerni cate regulatory, metabolic, and protein–protein ble network module consisting of genes, transcrip interactions. Because of these many functional tion factors, RNAs, enzymes, and metabolites. links, the defects of various genes spread through This understanding forces us to view diseases as out the intracellular network, affecting the activ the breakdown of selected functional modules ity of genes that otherwise carry no defects. rather than as single or small groups of genes. To understand various disease mechanisms, Given the many components of such functional it is not sufficient to know the precise list of modules, there are different paths to disease- “disease genes”; instead, we should try to map inducing systems failure; this explains why often out the detailed wiring diagram of the various many genes are linked to the same disease pheno cellular components that are influenced by type. Similarly, the effects of drugs are not lim these genes and gene products. Such network- ited to the molecules to which they directly bind; based thinking has already provided insights into instead, these effects can spread throughout the the pathogenesis of several diseases. For exam cellular network in which they act, causing un ple, a recent study suggested that 18 of the 23 wanted side effects. Therefore, drug side effects genes known to be associated with ataxia are are inherently network phenomena. part of a highly interlinked subnetwork5; in an Naturally, network-based thinking may account other example, a reverse-engineered subnetwork for the environmental and social influences on indicated that the androgen-receptor gene might disease as well. In this context, we must under be used to detect the aggressiveness of primary stand the human interactions encompassing so prostate cancer.6 cial and family links, proximity-based contacts, The existence of intricate molecular links be and transportation networks.9 For example, recent tween subcellular components and disease genes advances in the study of sexual networks have raises another possibility: that is, diseases may led to new protocols for drug dispersion. These not be as independent of each other as medical protocols are expected to be more efficient in practitioners currently consider them to be. For combating the acquired immunodeficiency syn example, could a genetic origin account for the drome in underdeveloped countries than current fact that obesity is a risk factor for diabetes? A protocols that are based on social need.10 quick look at the list of genes associated with The Human Genome Project has revolution these two diseases indicates that several genes, ized gene hunting, leading to an explosion in including ectoenzyme nucleotide pyrophosphate the number of detected associations between phosphodiesterase (ENPP1), peroxisome-prolifera genes and disease phenotypes. The beauty of ge tor–activated receptor γ (PPARγ), and — more re nomewide association studies lies in their ability cently — FTO, may be implicated in both diseas to quantify their own limitations. For instance, es. In addition to the well-known link between many of the newfound disease-associated ge diabetes and obesity, the large number of genes netic mutations account for only a tiny fraction shared by often quite distinct disorders indicates of disease occurrences. There is a tendency to that these diseases may have common genetic ori believe that the rest are hidden in more genes.

n engl j med 357;4 www.nejm.org july 26, 2007 405 T h e new england journal o f medicine

Figure 1. Complex Networks of Direct Relevance to Network Medicine. Although they are often treated separately, most human diseases are not independent of each other. Many diseases are associated with the breakdown of functional modules that are best described as subnetworks of a complex network connecting many cellular components. Therefore, an understanding of the functionally relevant genetic, regulatory, metabolic, and protein–protein interactions in a cellular network will play an important role in understanding the pathophysiology of human diseases (bottom layer). One way to visualize the ensuing potential interrelationships among human diseases is to construct a disease network (middle layer) in which two diseases are connected if they have a common genetic or functional origin. For example, on the basis of our current knowledge of disease genes, obesity is connected to at least seven other diseases such as diabetes, asthma, and insulin resistance, since genes associated with these diseases are known to affect obesity as well. The third network of key importance to human disease is the social network, which encompasses all human-to-human interactions (e.g., familial, friendship, sexual, and proximity-based contacts) that play a role in the spread of pathogens (top layer). These networks also have an important role in the spread of obesity. Efforts to understand the interactions between the cellular, disease, and social networks are part of network medicine, which aims to quantify the complex interlinked factors that may contribute to individual diseases.

406 n engl j med 357;4 www.nejm.org july 26, 2007 editorials

As the article by Christakis and Fowler shows,2 ant in the FTO gene is associated with body mass index and pre disposes to childhood and adult obesity. Science 2007;316:889-94. the answer is not always as simple as that. Net 2. Christakis NA, Fowler JH. The spread of obesity in a large works, in this case those that pertain to social social network over 32 years. N Engl J Med 2007;357:370-9. influence, may have just as strong an impact on 3. Barabási A-L. Linked. New York: Plume, 2003. 4. Barabási A-L, Oltvai ZN. Network biology: understanding the development of obesity as the otherwise the cell’s functional organization. Nat Rev Genet 2004;5:101-15. strong genetic effects. The role of links and 5. Lim J, Hao T, Shaw C, et al. A protein-protein interaction connections does not stop here. In the past few network for human inherited ataxias and disorders of Purkinje cell degeneration. Cell 2006;125:801-14. years, we learned that network effects increas 6. Ergün A, Lawrence CA, Kohanski MA, Brennan TA, Collins ingly affect all aspects of biologic and medical JJ. A network biology approach to prostate cancer. Mol Syst Biol research, from disease mechanisms to drug dis 2007;3:82. 3 7. Goh K-I, Cusick ME, Valle D, Childs B, Vidal M, Barabási A-L. covery. It is only a matter of time until these The human disease network. Proc Natl Acad Sci U S A 2007;104: advances will start to affect medical practice as 8685-90. well, marking the emergence of a new field that 8. Loscalzo J, Kohan I, Barabási A-L. Human diseases classifi cation in the postgenomic era: a complex systems approach to may be aptly called network medicine. human pathobiology. Mol Syst Biol (in press). No potential conflict of interest relevant to this article was re 9. Colizza V, Barthelemy M, Barrat A, Valleron A-J, Vespignani A. ported. Modeling the worldwide spread of pandemic influenza: baseline case and containment interventions. PLoS Med 2007;4(1):e13. From the Center for Complex Network Research, Departments 10. Cohen R, Havlin S, Ben-Avraham D. Efficient immunization of Physics and Computer Science, University of Notre Dame, strategies for computer networks and populations. Phys Rev Lett Notre Dame, IN. 2003;91:247901. 1. Frayling TM, Timpson NJ, Weedon MN, et al. A common vari Copyright © 2007 Massachusetts Medical Society.

powerpoint slides of journal figures and tables At the Journal’s Web site, subscribers can automatically create PowerPoint slides. In a figure or table in the full-text version of any article at www.nejm.org, click on Get PowerPoint Slide. A PowerPoint slide containing the image, with its title and reference citation, can then be downloaded and saved.

n engl j med 357;4 www.nejm.org july 26, 2007 407 T h e new england journal o f medicine

Health Law, Ethics, and Human Rights

Cancer and the Constitution — Choice at Life’s End George J. Annas, J.D., M.P.H.

J.M. Coetzee’s violent, anti-apartheid Age of Iron, families have focused their frustrations on the a novel the Wall Street Journal termed “a fierce pag Food and Drug Administration (FDA), which they eant of modern South Africa,” is written as a let see as a government agency denying them access ter by a retired classics professor, Mrs. Curren, to treatments they need. to her daughter, who lives in the United States. In May 2006 these families won an apparent Mrs. Curren is dying of cancer, and her daughter major victory when the Court of Appeals for the advises her to come to the United States for treat District of Columbia, in the case of Abigail Alli- ment. She replies, “I can’t afford to die in Amer ance v. Von Eschenbach (hereafter referred to as Abigail ica. . . . No one can, except Americans.” 1 Dying Alliance),3 agreed with their argument that patients of cancer has been considered a “hard death” for with cancer have a constitutional right of access at least a century, unproven and even quack rem to investigational cancer drugs. In reaction, the edies have been common, and price has been a FDA began the process of rewriting its own regu secondary consideration. Efforts sponsored by the lations to make it easier for terminally ill patients federal government to find cures for cancer date not enrolled in clinical trials to have access to from the establishment of the National Cancer investigational drugs.4 In November 2006, the Institute (NCI) in 1937. Cancer research was in full bench of the Court of Appeals vacated the tensified after President Richard Nixon’s decla May 2006 opinion, and the case was reheard in ration of a “war on cancer” and passage of the March 2007.5 The decision of the full bench, ex National Cancer Act of 1971.2 Most recently, pected by the fall, will hinge on the answer to a calls for more cancer research have followed the central question: Do terminally ill adult patients announcement by Elizabeth Edwards, wife of pres with cancer for whom there are no effective treat idential candidate John Edwards, that her cancer ments have a constitutional right of access to in is no longer considered curable. vestigational drugs their physicians think might Frustration with the methods and slow prog be beneficial? ress of mainstream medical research has helped fuel a resistance movement that distrusts both the constitutional controversy conventional medicine and government and that has called for the recognition of a right for ter The Abigail Alliance for Better Access to Devel minally ill patients with cancer to have access to opmental Drugs (hereafter called the Abigail Alli any drugs they want to take. Prominent examples ance) sued the FDA to prevent it from enforcing include the popularity of Krebiozen in the 1950s its policy of prohibiting the sale of drugs that had and of laetrile in the 1970s. As an NCI spokes not been proved safe and effective to competent person put it more than 20 years ago, when thou adult patients who are terminally ill and have no sands of people were calling the NCI hotline alternative treatment options. The Abigail Alli pleading for access to interleukin-2, “What the ance is named after Abigail Burroughs, whose callers are saying is, ‘Our mother, our brother, squamous-cell carcinoma of the head and neck our sister is dying at this very moment. We have was diagnosed when she was only 19 years old. nothing to lose.’”2 Today, families search the In Two years later, in 2001, she died. Before her ternet for clinical trials, and even untested chem death she had tried unsuccessfully to obtain in icals such as dichloroacetate, that seem to offer vestigational drugs on a compassionate use basis them some hope. In addition, basing advocacy on from ImClone and AstraZeneca and was accepted their personal experiences with cancer, many for a clinical trial only shortly before her death.

408 n engl j med 357;4 www.nejm.org july 26, 2007 health law, ethics, and human rights

Her father founded the Abigail Alliance in her volve treatment by the government or a memory.6 government subsidy. Rather, much as the The district court dismissed the Abigail Alli guardians of the comatose [sic] patient in ance lawsuit. The appeals court, in a two-to-one Cruzan did, the Alliance seeks to have the opinion written by Judge Judith Rogers, who was government step aside by changing its pol joined by Judge Douglas Ginsburg, reversed the icy so the individual right of self-determi decision. It concluded that competent, terminally nation is not violated.3 ill adult patients have a constitutional “right to access to potentially life-saving post-Phase I in The appeals court concluded that the Supreme vestigational new drugs, upon a doctor’s advice, Court’s 1979 unanimous decision on laetrile,8 in even where that medicine carries risks for the pa which the Court concluded that Congress had tient,” and remanded the case to the district court made no exceptions in the FDA law for termi to determine whether the FDA’s current policy nally ill cancer patients, was not relevant because violated that right.3 laetrile had never been studied in a phase 1 trial and because the Court did not address the ques the right to life tion of whether terminally ill cancer patients have a constitutional right to take whatever drugs their The appeals court found that the relevant consti physicians prescribe. tutional right was determined by the due-process clause of the Fifth Amendment: “no person shall the dissent be . . . deprived of life, liberty, or property with out due process of law.” In the court’s words, the Judge Thomas Griffith, the dissenting judge, ar narrow question presented by Abigail Alliance is gued that the suggested constitutional right sim whether the due-process clause “protects the right ply does not exist. He noted, for example, that the of terminally ill patients to make an informed de self-defense cases relied on are examples of “ab cision that may prolong life, specifically by use stract concepts of personal autonomy,” and can of potentially life-saving new drugs that the FDA not be used to craft new rights. As to the nation’s has yet to approve for commercial marketing but history and traditions, he concluded that the that the FDA has determined, after Phase I clini FDA’s drug-regulatory efforts have been reasona cal human trials, are safe enough for further test ble responses “to new risks as they are present ing on a substantial number of human beings.”3 ed.”3 Accepting his argument leaves the majority The court answered yes, finding that this right resting squarely on Cruzan and the laetrile case. has deep legal roots in the right to self-defense, As to Cruzan, the dissent argued that “A tradition and that “Barring a terminally ill patient from of protecting individual freedom from life-saving, the use of a potentially life-saving treatment im but forced, medical treatment does not evidence pinges on this right of self-preservation.”3 In a a constitutional tradition of providing affirma footnote, the court restated this proposition: “The tive access to a potentially harmful, even fatal, fundamental right to take action, even risky ac commercial good.”3 As to the laetrile case, the tion, free from government interference, in order judge noted simply that the Court had agreed to save one’s own life undergirds the court’s de with the FDA that, “For the terminally ill, as for cision.”3 The court relied primarily on the Cruzan anyone else, a drug is unsafe if its potential for case,7 in which the Supreme Court recognized inflicting death or physical injury is not offset by the right of a competent adult to refuse life-sus the possibility of therapeutic benefit.”3,8 taining treatment, including a feeding tube: Finally, the dissenting judge argued that if the new constitutional right were accepted, it was too The logical corollary is that an individual vague to be applied only to terminally ill patients must also be free to decide for herself seeking drugs that had been tested in phase 1 tri whether to assume any known or unknown als. Specifically, the judge asked, must the right risks of taking a medication that might pro also apply to patients with “serious medical con long her life. Like the right claimed in Cruzan, ditions,” to patients who “cannot afford potential the right claimed by the [Abigail] Alliance ly life-saving treatment,” or to patients whose to be free of FDA imposition does not in physicians believe “marijuana for medicinal pur

n engl j med 357;4 www.nejm.org july 26, 2007 409 T h e new england journal o f medicine

poses . . . is potentially life saving?”3 In other ule I controlled substances, like marijuana or ly words, there is no principled reason to restrict sergic acid diethylamide (LSD)? If it is a constitu the constitutional right the majority created to tional right, these should be available too, at least either terminally ill patients or to post–phase 1 unless the state can demonstrate a “compelling drugs. interest” in regulating them. My prediction is that after rehearing this case discussion en banc, the full Circuit Court will reject the po sition of the Abigail Alliance for the same rea The facts as illustrated by stories of patients dying sons that the Supreme Court rejected the “right” of cancer while trying unsuccessfully to enroll in of terminally ill patients to have access to physi clinical trials are compelling, and our current cian-prescribed drugs they could use to end their system of ad hoc exceptions is deeply flawed. lives.9-11 To decide otherwise would entirely un The central constitutional issue, however, rests dermine the legitimacy of the FDA. Patients in primarily on determining whether this case is or the United States have always had a right to re is not like the right-to-refuse-treatment case of fuse any medical treatment, but we have never Nancy Cruzan, a woman in a permanent vegeta had a right to demand mistreatment, inappropri tive state whose family wanted tube feeding dis ate treatment, or even investigational or experi continued because they believed that discontinu mental interventions. This will not, however, be ation was what she would have wanted. I do not the end of the matter. After the physician-assist think Abigail Alliance is like Cruzan. Rather, it is sub ed–suicide cases, the fight appropriately shifted stantially identical to cases involving physician- to the states, although so far only one, Oregon, assisted suicide, in which a terminally ill patient has provided its physicians with immunity for claims a constitutional right of access to physi prescribing life-ending drugs to their competent, cian-prescribed drugs to commit suicide. terminally ill patients.12 In the Abigail Alliance The Supreme Court has decided, unanimously, case, the debate will continue in the forum in that no right to physician-prescribed drugs for which it began — the FDA — and in Congress. suicide exists.9,10 There is no historical tradition of support for this right. And although the right congress seems to be narrowly defined, it is unclear to whom it should apply — why only to terminally Congressional action also had its birth with the ill patients? Don’t patients in chronic pain have story of one patient with cancer and was also even a stronger interest in suicide? Why is the heavily influenced by another individual patient physician necessary, and why are physician-pre involved in a controversy over removal of a feed scribed drugs the only acceptable method of sui ing tube. “Terri’s Law” was enacted in Florida in cide? None of these questions can be answered 2003 to try to prevent the removal of a feeding by examining the Constitution.11 tube from Terri Schiavo; the case was substan Similarly, in Abigail Alliance, the new constitu tially similar to Cruzan. Terri’s case gained nation tional right proposed has no tradition in the Unit al attention 2 years later.13 In the midst of it, in ed States, and it cannot be narrowly applied. For March 2005, the Wall Street Journal asserted, in an example, why should a constitutional right apply editorial titled “How About a ‘Kianna’s Law’?,” only to people who have a particular medical “If Terri Schiavo deserves emergency federal inter status? And why should a physician be involved vention to save her life, people like Kianna Karnes at all? If patients have a right to autonomy, why deserve it even more.”14 At the time, Kianna Karnes isn’t the requirement of a government-licensed was a 44-year-old mother of four who was dying physician’s recommendation at least as burden of kidney cancer. Her only hope of survival, ac some as the requirement of the FDA’s approval cording to the editorial, was to gain access to one of the investigational drug? And why would the of two experimental drugs in clinical trials, but Constitution apply only to investigational drugs neither of the two companies running the trials for which phase 1 trials have been completed? (Bayer and Pfizer) would make the drugs avail Why not include access to investigational medical able to her on a compassionate-use basis. This devices, like the artificial heart, or even to Sched was because, according to the Wall Street Journal,

410 n engl j med 357;4 www.nejm.org july 26, 2007 health law, ethics, and human rights the FDA “makes it all but impossible” for the ers of rights — are jettisoned, with the require manufacturers “to provide [drugs] to terminal pa ment of such waivers becoming the price of ob tients on a ‘compassionate use’ basis.”14 taining the investigational agent from an otherwise Almost immediately after the editorial was reluctant drug company. published, both drug manufacturers contacted Kianna’s physicians to discuss releasing the drugs fda proposal to her. But within 2 days after publication, she was dead. The Wall Street Journal editorialized, “Isn’t In direct response to Abigail Alliance, the FDA pro it a national scandal that cancer sufferers should posed amending its rules to encourage more drug have to be written about in the Wall Street Journal companies to offer their investigational drugs to be offered legal access to emerging therapies through compassionate-use programs.4 These pro once they’ve run out of other options?”15 It not grams first came into prominence during the ed that Mrs. Karnes’ father, John Rowe — him early days of infection with the human immuno self a survivor of leukemia — was working with deficiency virus (HIV) and AIDS, when there were the Abigail Alliance on a “Kianna’s Law.” That no effective treatments and AIDS activists insist law, formally titled the “Access, Compassion, Care, ed that they have early access to investigational and Ethics for Seriously Ill Patients Act” or the drugs because, in the words of their inaccurate “ACCESS Act,” was introduced in November 2005 slogan, “A Research Trial Is Treatment Too.”18 and is an attempt to make it much easier for seri Because the FDA could not stand the political ously ill patients to gain access to experimental pressure generated by the activists, the compas drugs.16,17 sionate-use program was developed as a kind of The act begins with a series of congressional political safety valve to provide enough exceptions findings, including that “Seriously ill patients to save their basic research rules. In early Decem have a right to access available investigational ber 2006, the FDA continued this political-safety- drugs, biological products, and devices.” The act valve approach by issuing new proposed regula permits the sponsor to apply for approval to make tions with a title that could have been taken an investigational drug, biologic product, or de directly from the AIDS Coalition to Unleash Pow vice available on the basis of data from a complet er (ACT-UP): “Expanded Access to Investigational ed phase 1 trial, “preliminary evidence that the Drugs for Treatment Use.”19 product may be effective against a serious or life- The FDA’s expanded-access proposal applies threatening condition or disease,” and an assur to “seriously ill patients when there is no com ance that the clinical trial will continue.17 The parable or satisfactory alternative therapy to diag patient, who must have exhausted all approved nose, monitor, or treat the patient’s disease or treatments, must provide written informed con condition.” 4 Manufacturers are required to file sent and must also sign “a written waiver of the an “expanded access submission,” and the prod right to sue the manufacturer or sponsor of the uct must be administered or dispensed by a li drug, biological product, or device, or the physi censed physician who will be considered an “in cians who prescribed the product or the institu vestigator,” with all the reporting requirements tion where it was administered, for an adverse that role entails.3 event caused by the product, which shall be bind Whether or not the proposal is adopted, it will ing in every State and Federal court.”17 do little to increase access, since the major bottle Although Congress is the proper forum to ad neck in the compassionate-use program has nev dress this issue, this initial attempt has some of er been the FDA. The manufacturers have no in the same problems as the Abigail Alliance decision: centives to make their investigational products the patients to whom it applies are ambiguously available outside clinical trials. This is because classified, and clinical research seems to be equat direct access to investigational drugs by individu ed with clinical care. Also troubling is that the als may make it more difficult to recruit research patients (and would-be subjects) are asked to as subjects, and thus to conduct the clinical trials sume all of the risks of the uncontrolled experi necessary for drug approval, and could also sub ments, and current rules of research — which ject the drug manufacturer to liability for serious protect subjects by prohibiting mandatory waiv adverse reactions. Even without a lawsuit, a seri

n engl j med 357;4 www.nejm.org july 26, 2007 411 T h e new england journal o f medicine

ous reaction to a drug outside a trial could ad simply being responsive to patient needs, the in versely affect the trial itself.4,16,20 The drug com terventions “may turn out to be a projection of panies are right to worry that the approaches of their own needs onto patients.”21 the judiciary, Congress, and the FDA will prob ably make clinical trials more difficult to con government and the market duct, because few seriously ill patients who have exhausted conventional treatments would rather Another recurrent theme is the belief that gov be randomly assigned to an investigational drug ernment regulation is evil, a central tenet of the than have a guarantee that they will receive the laetrile litigation of the 1970s. The court hearing investigational drug their physician recommends Abigail Alliance was correct to note that laetrile for them. This could result in significant delays in never underwent a phase 1 trial, but every indica the approval and overall availability of drugs that tion was that the drug, also known as vitamin demonstrate effectiveness — a result no one fa B17, was harmless, albeit also ineffective against vors. Even if patients with cancer are willing buy cancer. Laetrile became a legal cause celebre in ers, drug manufacturers are not willing sellers. 1972, when California physician John A. Richard son was prosecuted for promoting laetrile. Rich physicians and patients ardson was a member of the John Birch Society, which quickly formed the Committee for Freedom The cover story for all the proposed changes is of Choice in Cancer Therapy, with more than 100 patients’ choice. But without scientific evidence committees nationwide.24 It took another 7 years of the risks and benefits of a drug, choice can before the FDA prevailed in its case against lae not be informed, and for seriously ill patients, trile before the Supreme Court.8 The basic argu fear of death will predictably overcome fear of ments against FDA regulation remain the same unknown risks. This is understandable. As psy today: the FDA follows a “paternalistic public chiatrist Jay Katz, the leading scholar on informed policy that prevents individuals from exercising consent, has noted, when medical science seems their own judgment about risks and benefits. If impotent to fight nature, “all kinds of senseless the FDA must err, it should be on the side of interventions are tried in an unconscious effort to patients’ freedom to choose.”25 cure the incurable magically through a ‘wonder drug,’ a novel surgical procedure, or a penetrating public policy psychological interpretation.”21 Another Wall Street Journal article, entitled “Saying No to Penelope,”22 The FDA will prevail again today, not only be illustrates the impossibility of limiting access to cause there is no constitutional right of access to unproven cancer drugs to competent adults. The unapproved drugs but also because even if there article tells the story of 4-year-old Penelope, who were, the state has the same compelling interest is dying from neuroblastoma that has proved re in approving drugs as it has in licensing physi sistant to all conventional treatments. Her par cians. From a public policy view, the Abigail Alli- ents seek “anything [that] has a prayer of saving ance court, the Congress, and the FDA all seem her.” In her father’s words, “The chance of any to be suffering from the “therapeutic illusion” in thing bringing her back from the abyss now is which research, designed to test a hypothesis for very low. But the only thing I know for sure is society, is confused with treatment, administered if we don’t treat her, she will die.” With Penelope in the best interests of individual patients.21,26,27 hospitalized and in pain, her parents continue Of course there is a continuum, and it is perfectly “searching Penelope’s big brown eyes for clues as understandable that many patients with cancer, to how long she wants to continue to battle for told that there is nothing conventional medicine life.” can do for them, will want access to whatever is It is suggested that the requirement of a phy available in or outside the context of clinical tri sician’s recommendation can safeguard against als. But this is a problem for patients, physicians, “magical thinking” and help make informed con the FDA, and drug manufacturers. First, because sent real.23 But as Katz has noted, although phy terminally ill patients can be harmed and exploit sicians (and, he could have added, drug compa ed, there are better and worse ways to die.21,26 nies) often justify such last-ditch interventions as Second, it is only through research, not “treat

412 n engl j med 357;4 www.nejm.org july 26, 2007 health law, ethics, and human rights ment,” that cancer may become a chronic illness 7. Cruzan v. Director, Missouri Dept. of Health, 497 U.S. 261 (1990). that is treated with a complex array of drugs, 8. United States v. Rutherford, 442 U.S. 544 (1979). 28,29 given either together or in a progression. The 9. Washington v. Glucksberg, 521 U.S. 702 (1997). right to choose in medicine is a central right of 10. Vacco v. Quill, 521 U.S. 793 (1997). 11. Annas GJ. The bell tolls for a constitutional right to assisted patients, but the choices can and should be lim suicide. N Engl J Med 1997;337:1098-103. ited to reasonable medical alternatives, which 12. Gonzales v. Oregon, 546 U.S. 243 (2006). themselves are based on evidence. 13. Annas GJ. “I want to live”: medicine betrayed by ideology in the political debate over Terri Schiavo. Stetson Law Rev 2005;35: This is, I believe, good public policy. But it is 49-80. 30 also much easier said than done. Death is 14. How about a “Kianna’s Law”? Wall Street Journal. March 24, feared and even dreaded in our culture, and few 2005:A14. 15. Kianna’s legacy. Wall Street Journal. March 29, 2005:Al4. Americans are able to die at home, at peace, 16. Groopman J. The right to a trial: should dying patients have with our loved ones in attendance, without seek access to experimental drugs? The New Yorker. December 18, ing the “latest new treatment.” There always 2006:40-7. 17. ACCESS Act (Access, Compassion, Care, and Ethics for Seri seems to be something new to try, and there is ously Ill Patients), S. 1956, 109th Cong (2005). almost always anecdotal evidence that it could 18. Annas GJ. Faith (healing), hope and charity at the FDA: the help. This is one reason that even extremely high politics of AIDS drug trials. Villanova Law Rev 1989;34:771-97. 19. FDA proposes rules overhaul to expand availability of ex prices do not affect demand for cancer drugs, perimental drugs: the agency also clarifies permissible charges 31,32 even ones that add little or no survival time. to patients. Rockville, MD: Food and Drug Administration, De When does caring for the patient demand pri cember 11, 2006. (Accessed July 6, 2007, at http://www.fda.gov/ bbs/topics/NEWS/2006/NEW01520.html.) mary attention to palliation rather than to long- 20. Prud’homme A. The cell game: Sam Waksal’s fast money and shot, high-risk, investigational interventions? Coet false promises — and the fate of ImClone’s cancer drug. New zee’s Mrs. Curren, who rejected new medical York: Harper Business, 2004. 21. Katz J. The silent world of doctor and patient. New Haven, treatment for her cancer and insisted on dying at CT: Yale University Press, 1984:151. home, told her physician, whom she saw as “with 22. Anand G. Saying no to Penelope: father seeks experimental drawing” from her after giving her a terminal cancer drug, but a biotech firm says risk is too high. Wall Street Journal. May 1, 2007:A1. prognosis — “His allegiance to the living, not 23. Robertson J. Controversial medical treatment and the right the dying” — “I have no illusions about my con to health care. Hastings Cent Rep 2006;36:15-20. dition, doctor. It is not [experimental] care I need, 24. Culbert ML. Vitamin B17: Forbidden weapon against cancer. 1 New Rochelle, NY: Arlington House, 1974. just help with the pain.” 25. Miller HI. Paternalism costs lives. Wall Street Journal. March No potential conflict of interest relevant to this article was re 2, 2006:A15. ported. 26. Annas GJ. The changing landscape of human experimenta From the Department of Health Law, Bioethics, and Human tion: Nuremberg, Helsinki, and beyond. Health Matrix J Law Rights, Boston University School of Public Health, Boston. Med 1992;2:119-40. 27. Appelbaum PS, Lidz CW. Re-evaluating the therapeutic mis 1. Coetzee JM. Age of iron. London: Seeker & Warburg, 1990. conception: response to Miller and Joffe. Kennedy Inst Ethics J 2. Patterson JT. The dread disease: cancer and modern Ameri 2006;16:367-73. can culture. Cambridge, MA: Harvard University Press, 1987. 28. Nathan D. The cancer treatment revolution: how smart drugs 3. Abigail Alliance v. Von Eschenbach, 445 F.3d 470 (DC Cir and other therapies are renewing our hope and changing the 2006). Vacated 469 F.3d 129 (DC Cir 2006). face of medicine. New York: John Wiley, 2007. 4. Proposed rules for charging for investigational drugs and 29. Brugarolas J. Renal-cell carcinoma — molecular pathways expanded access to investigational drugs for treatment use. and therapies. N Engl J Med 2007;356:185-6. Rockville, MD: Food and Drug Administration, 2006. (Accessed 30. Callahan D. False hopes: why America’s quest for perfect July 6, 2007, at http://www.fda.gov/cder/regulatory/applications/ health is a recipe for failure. New York: Simon and Schuster, IND_PR.htm.) 1998. 5. Abigail Alliance v. Von Eschenbach, 429 F.3d 129 (DC Cir 31. Berenson A. Hope, at $4,200 a dose: why a cancer drug’s cost 2006). doesn’t hurt demand. New York Times. October 1, 2006:BU1. 6. Jacobson PD, Parmet WE. A new era of unapproved drugs: 32. Anand G. From Wall Street, a warning about cancer drug the case of Abigail Alliance v Von Eschenbach. JAMA 2007;297: prices. Wall Street Journal. March 15, 2007:A1. 205-8. Copyright © 2007 Massachusetts Medical Society.

n engl j med 357;4 www.nejm.org july 26, 2007 413 T h e new england journal o f medicine

previous myocardial infarction, bypass grafting, ing death, myocardial infarction, or stroke in and other factors. long-term follow-up. As suggested by Kiat, we are performing post William E. Boden, M.D. hoc analyses to better delineate a high-risk sub- University at Buffalo School of Medicine group on the basis of the ischemic burden as as- and Biomedical Sciences sessed on MPI at baseline and during a follow-up Buffalo, NY 14214 period of 6 to 14 months. Koon K. Teo, M.B., B.Ch., Ph.D. Nagajothi et al. acknowledge that our inten- McMaster University Medical Center tion-to-treat analysis was appropriate but encour- Hamilton, ON L8N 3Z5, Canada age further analysis according to actual treatment William S. Weintraub, M.D. received. A detailed analysis of the “crossover” Christiana Care Health System Newark, DE 19718 population is under way. Finally, Mak emphasizes the incremental ben- for the COURAGE Trial Investigators efit of the use of dual antiplatelet therapy on the 1. Gibbons RJ, Abrams J, Chatterjee K, et al. ACC/AHA 2002 composite end point of death, myocardial infarc- guideline update for the management of patients with chronic 2 stable angina — summary article: a report of the American Col- tion, or stroke at 1 year in the CREDO trial. En- lege of Cardiology/American Heart Association Task Force on rollment in our trial antedated these results. practice guidelines (Committee on the Management of Patients However, the Clopidogrel for High Atherothrom- with Chronic Stable Angina). J Am Coll Cardiol 2003;41:159-68. 2. Steinhubl SR, Berger PB, Mann JT, et al. Early and sustained botic Risk and Ischemic Stabilization, Manage- dual oral antiplatelet therapy following percutaneous coronary ment, and Avoidance (CHARISMA) trial,3 involv- intervention: a randomized controlled trial. JAMA 2002;288: ing patients with stable coronary artery disease, 2411-20. [Erratum, JAMA 2003;289:987.] 3. Bhatt DL, Fox KAA, Hacke W, et al. Clopidogrel and aspirin did not show a compelling benefit of combined versus aspirin alone for the prevention of atherothrombotic treatment with aspirin and clopidogrel in reduc- events. N Engl J Med 2006;354:1706-17.

Diabetic Gastroparesis

To the Editor: In his review of diabetic gastro- Robert J. Heine, M.D., Ph.D. paresis, Camilleri (Feb. 22 issue)1 states that in VU University Medical Center our clinical trial of exenatide,2 nausea and vomit- 1007 MB Amsterdam, the Netherlands ing led to the “cessation of treatment in about one Robert Brodows, M.D. third of patients.” This information is incorrect. Lilly Research Laboratories Of 282 patients randomly assigned to receive ex- Indianapolis, IN 46285 enatide, 54 withdrew from the study (27 withdrew 1. Camilleri M. Diabetic gastroparesis. N Engl J Med 2007;356: because of an adverse event, 10 because of proto- 820-9. 2. Heine RJ, Van Gaal LF, Johns D, et al. Exenatide versus insu- col violations, 7 because of the patient’s decision, lin glargine in patients with suboptimally controlled type 2 dia- 4 because of loss of glucose control, and 1 because betes: a randomized trial. Ann Intern Med 2005;143:559-69. of the physician’s decision, and 5 were lost to 3. DeFronzo RA, Ratner RE, Han J, Kim DD, Fineman MS, Baron AD. Effects of exenatide (exendin-4) on glycemic control follow-up). As stated in our article, 18 of these and weight over 30 weeks in metformin-treated patients with patients withdrew from the study because of nau- type 2 diabetes. Diabetes Care 2005;28:1092-100. sea or other gastrointestinal symptoms. This rate 4. Kendall DM, Riddle MC, Rosenstock J, et al. Effects of exenatide (exendin-4) on glycemic control over 30 weeks in patients is in line with the dropout rates in other recent with type 2 diabetes treated with metformin and a sulfonylurea. clinical trials, which range from 1.8 to 5.1%.3‑5 Diabetes Care 2005;28:1083-91. Mild-to-moderate nausea is the most common 5. Nauck MA, Duran S, Kim D, et al. A comparison of twice daily exenatide and biphasic insulin aspart in patients with type adverse event in patients receiving exenatide. How- 2 diabetes who were suboptimally controlled with sulfonylurea ever, the development of tolerance to these ad- and metformin: a non-inferiority study. Diabetologia 2007;50: verse gastrointestinal effects of exenatide has been 259-67. suggested in our trial as well as in other long- term phase 3 trials of the drug that have been To the Editor: In Table 2 of his article, Camilleri reported.3‑5 lists the motilin-receptor agonists erythromycin,

418 n engl j med 357;4 www.nejm.org july 26, 2007 correspondence clarithromycin, and azithromycin as recommend- dyskinesia are increased after long-term treated agents for the treatment of diabetic gastropa- ment.2,3 Diabetes is associated with an increased resis. I agree with the recommendation regarding frequency and severity of tardive dyskinesia.2,4 erythromycin. However, clarithromycin, azithro- The risk of tardive dyskinesia is underrecognized mycin, and other more acid-stable macrolides and by physicians who use metoclopramide for long- azalides are likely to be poor alternatives. Eryth- term treatment3; this is especially important in romycin A in an acidic medium such as gastric view of the increased use of metoclopramide juice is degraded into its anhydrous hemiketal and since the withdrawal of cisapride in 2000.5 spiroketal forms.1 Both forms are inactive micro- Daniel Tarsy, M.D. biologically, but they have motilin-like activity that Beth Israel Deaconess Medical Center is several times greater than that of erythromy- Boston, MA 02215 cin A.1,2 It is mostly because of this characteristic [email protected] that erythromycin has a low oral bioavailability 1. Tarsy D, Baldessarini RJ. Epidemiology of tardive dyskine- of the microbiologically active drug and consid- sia: is risk declining with modern antipsychotics? Mov Disord 2006;21:589-98. erable gastrointestinal adverse effects. Clarithro- 2. Ganzini L, Heintz RT, Hoffman WF, Casey DE. The prevalence mycin and azithromycin have modified chemical of tardive dyskinesia in neuroleptic-treated diabetics: a controlled structures that are more acid-stable and do not study. Arch Gen Psychiatry 1991;48:259-63. 3. Pasricha PJ, Pehlivanov N, Sugumar A, Jankovic J. Drug in- 2,3 form the highly active motilin mimics ; conse- sight: from disturbed motility to disordered movement — a re- quently, neither drug is likely to be as useful as view of the clinical benefits and medicolegal risks of metoclo- erythromycin for the treatment of motilin-respon- pramide. Nat Clin Pract Gastroenterol Hepatol 2006;3:138-48. 4. Woerner MG, Saltz BL, Kane JM, Lieberman JA, Alvir JM. sive disorders. Both clarithromycin and azithromy Diabetes and development of tardive dyskinesia. Am J Psychiatry cin are also more expensive than erythromycin. 1993;150:966-8. 5. Shaffer D, Butterfield M, Pamer C, Mackey AC. Tardive dys- Eufronio G. Maderazo, M.D. kinesia risks and metoclopramide use before and after U.S. mar- William W. Backus Hospital ket withdrawal of cisapride. J Am Pharm Assoc (2003) 2004;44: Norwich, CT 06360 661-5. 1. Weymouth-Wilson AC. The role of carbohydrates in biologically active natural products. Nat Prod Rep 1997;14:99-110. The Author Replies: Heine and Brodows correct 2. Tsuzuki K, Sunazuka T, Marui S, et al. Motilides, macrolides with gastrointestinal motor stimulating activity. I. O-substituted ly point out that the withdrawal rate attributed to and tertiary N-substituted derivatives of 8,9-anhydroerythromy- nausea in their clinical trial was overstated in my cin A 6,9-hemiacetal. Chem Pharm Bull (Tokyo) 1989;37:2687-700. article. However, in trials involving longer-term 3. Gill CJ, Abruzzo GK, Flattery AM, et al. In vivo evaluation of three acid-stable azalide compounds, L-701,677, L-708,299 and treatment that may better reflect clinical experi- 1-4 L-708,365 compared to erythromycin, azithromycin and clarith- ence, as well as in the study by Nauck et al. romycin. J Antibiot (Tokyo) 1995;48:1141-7. cited by Heine and Brodows, there is a high prevalence of nausea (average prevalence, approximate To the Editor: Camilleri compares gastrointes- ly 33%). The all-cause withdrawal rate in these tinal prokinetic agents with regard to their effi- randomized, controlled, or open-label treatment cacy and side effects. For metoclopramide, neuro- trials involving exenatide ranged from 21 to 45%. logic side effects are reported from one 4-week Approximately 4% of these withdrawals were at- trial. Tardive dyskinesia and other extrapyramidal tributed to a loss of glucose control and 7% were effects are not mentioned in the text and are cit- attributed to adverse events. It is unclear whether ed only briefly in Table 2 of the article. Since the withdrawal of consent by approximately 11% clinical trials of metoclopramide have been short- of patients and withdrawals because of protocol term trials, this review understates the risk of violations by approximately 10% of patients were tardive dyskinesia among patients treated with due to nausea, which was by far the most fre- metoclopramide on a long-term basis. quent adverse event (57% in the article by Heine Metoclopramide is a dopamine-receptor antag et al. cited by Heine and Brodows). Nausea and onist that causes parkinsonism, acute dystonia, vomiting in patients with diabetes may be wrong- akathisia, and tardive dyskinesia. The risk of tar- ly attributed to gastroparesis rather than to iatro- dive dyskinesia after long-term treatment with genic disease, which may be reversible. antipsychotic drugs that are dopamine-receptor I agree with Maderazo’s statement regarding antagonists is 5% per year of drug exposure.1 Al the different potencies of macrolides in the stim though prospective data for metoclopramide are ulation of gastric emptying. I included clarithro- unavailable, the prevalence and severity of tardive mycin and azithromycin in Table 2 because this

n engl j med 357;4 www.nejm.org july 26, 2007 419 T h e new england journal o f medicine

table reflects national society guidelines, as noted Since publication of his article, Dr. Camilleri reports having received research support from Novartis. in one of the table footnotes. 1. Riddle MC, Henry RR, Poon TH, et al. Exenatide elicits sus- The potential for tardive dyskinesia during tained glycaemic control and progressive reduction of body treatment with metoclopramide is important. weight in patients with type 2 diabetes inadequately controlled However, as indicated in the article by Shaffer by sulphonylureas with or without metformin. Diabetes Metab Res Rev 2006;22:483-91. et al. (to which Tarsy refers in his letter), well- 2. Blonde L, Klein EJ, Han J, et al. Interim analysis of the ef- described risk factors are common in reports fects of exenatide treatment on A1C, weight and cardiovascular of metoclopramide-associated tardive dyskinesia. risk factors over 82 weeks in 314 overweight patients with type 5 2 diabetes. Diabetes Obes Metab 2006;8:436-47. Moreover, Ganzini et al. calculated that the rela- 3. Ratner RE, Maggs D, Nielsen LL, et al. Long-term effects of tive risk of tardive dyskinesia was not signifi- exenatide therapy over 82 weeks on glycaemic control and weight cantly elevated with use of metoclopramide (rela- in over-weight metformin-treated patients with type 2 diabetes mellitus. Diabetes Obes Metab 2006;8:419-28. tive risk, 1.67; 95% confidence interval, 0.93 to 4. Buse JB, Klonoff DC, Nielsen LL, et al. Metabolic effects of 2.97), although the risk appeared to be higher two years of exenatide treatment on diabetes, obesity, and hepatic among patients with diabetes. biomarkers in patients with type 2 diabetes: an interim analysis of data from the open-label, uncontrolled extension of three double- Michael Camilleri, M.D. blind, placebo-controlled trials. Clin Ther 2007;29:139-53. 5. Ganzini L, Casey DE, Hoffman WF, McCall AL. The preva- Mayo Clinic College of Medicine lence of metoclopramide-induced tardive dyskinesia and acute Rochester, MN 55905 extrapyramidal movement disorders. Arch Intern Med 1993;153: [email protected] 1469-75.

Case 15-2007: A Woman with Asthma and Cardiorespiratory Arrest

To the Editor: In the Case Record presented by To the Editor: Dr. Shepard is quoted as saying, Wechsler et al. (May 17 issue),1 many possible caus “The pneumothorax results from the Macklin es of death in a patient with asthma are consid- effect.” I suspect that what she really said was, ered. However, the discussants do not sufficiently “The pneumomediastinum results from the Mack- emphasize that lung mechanics and hemodynamic lin effect.” effects associated with airflow obstruction are the Wallace T. Miller, M.D. most important physiological disturbances in pa- Hospital of the University of Pennsylvania tients with severe acute asthma. Hemodynamic Philadelphia, PA 19104 collapse related to dynamic hyperinflation is a common cause of obstructive shock, pulseless electric activity, and cardiac arrest in intubated The Discussants Reply: There are many causes patients with asthma. of death in young patients with asthma. In addi- Thiago Lisboa, M.D. tion to hypoxemia from prolonged bronchospasm, University Rovira Virgili we agree with Lisboa et al. that a progressive in- 43007 Tarragona, Spain crease in intrathoracic pressure associated with dynamic hyperinflation can lead to both dimin- Diego de Mendoza, M.D. ished venous return and mechanical compression Centro de Investigación Biomédica en Red (CIBER) 43007 Tarragona, Spain of the heart and associated vasculature, with re- sultant hemodynamic collapse. It is important to Jordi Rello, M.D., Ph.D. recognize these potential causes of complications Joan XXIII Hospital 43007 Tarragona, Spain in patients with asthma who have respiratory fail- [email protected] ure, since several ventilatory strategies (including 1. Case Records of the Massachusetts General Hospital (Case the use of low tidal volumes and low respiratory 15-2007). N Engl J Med 2007;356:2083-91. rates) may minimize these risks.

420 n engl j med 357;4 www.nejm.org july 26, 2007 correspondence

Since there is extrapulmonary air in the tho- Michael E. Wechsler, M.D., M.M.Sc. rax, resulting from rupture of a peripheral alveo- Brigham and Women’s Hospital lus, we agree with Miller that “pneumomediasti- Boston, MA 02115 num” is a more precise term than “pneumothorax” Jo-Anne O. Shepard, M.D. to describe the location of air due to the Macklin Massachusetts General Hospital effect in this case. Boston, MA 02114

Folate Deficiency and Plasma Homocysteine during Increased Oxidative Stress

To the Editor: Several reports have indicated levels were also determined in normal mice and that elevated plasma levels of the amino acid mice expressing the human APOE ε3 allele. homocysteine are associated with, or are a primary A comparison of values indicated that plasma risk factor for, coronary and vascular disease and homocysteine levels were identical in all mice Alzheimer’s disease. However, a number of addi- under normal dietary conditions (Fig. 1). As ex- tional studies do not show such a correlation.1 This pected, after 1 month of dietary folate depriva- discrepancy may arise because there are multiple tion, which has previously been shown to result metabolic fates for homocysteine, depending on in a folate-deficient state, normal and APOE ε3 the nature and extent of cellular oxidative stress.

Most homocysteine is converted to methionine 14 Complete diet Deficient diet through folate-dependent and vitamin B12–depen 12 dent reactions. Compromise of this conversion as a consequence of folate deficiency can lead to in- 10 creased export of homocysteine from cells, which 8 fosters an ultimate increase in total plasma homo 6 cysteine, the fraction measured in the conventional laboratory assay for homocysteine. How- 4 ever, a growing body of literature underscores 2

that under conditions of oxidative stress, compen- Plasma Homocysteine (µmol/liter) 0 satory amplification of the redox-sensitive trans- Normal APOE ε3 APOE ε4 APOE−/− sulfuration pathway will consume homocysteine to generate the endogenous antioxidant gluta- Figure 1. Plasma Homocysteine Levels in Normal AUTHOR: Rogers RETAKE 1st 2 and ICMApolipoprotein E (APOE)–Variant Mice, According thione. Such conditions may obviate any increase 2nd to WhetherREG F FIGURE: the Diet1 Wasof 1 Complete or Associated in plasma homocysteine despite folate deficiency. 3rd withCASE Oxidative Stress. Revised We set out to determine whether oxidative NormalEMail C57B/6 mice and miceLine of the 4-Csame geneticSIZE back ARTIST: ts stress could prevent the anticipated increase in ground but lacking murine APOEH/T (APOEH/T−/−) or 16p6express Enon Combo plasma homocysteine levels under folate-deficient ing human APOE ε3 or APOE ε4 were fed a complete diet for 9 to 12 months.AUTHOR, Each PLEASE mouse NOTE: type was then di- conditions. Plasma homocysteine levels were de- Figure has been redrawn and type has been reset. vided into two groupsPlease and check either carefully. maintained on this termined in mice that were homozygously lacking complete diet (AIN-76, Purina/Mother Hubbard) or apolipoprotein E (APOE−/−) and transgenic mice switchedJOB: 35704 to an otherwise identical but oxidativelyISSUE: 07-26-07 stress- expressing the human APOE ε4 allele. Notably, ful diet, deficient in folate and vitamin E and containing the presence of one or more APOE ε4 alleles is dietary iron (50 mg per gram of diet), for 1 month. At correlated with coronary artery disease, stroke, the end of this period, blood was obtained, and plasma 3 levels of homocysteine were determined by high-perfor- atherosclerosis, and Alzheimer’s disease. Both mance liquid chromatography. Values shown are the models show increased oxidative stress and are mean (±SE) plasma homocysteine levels in 8 to 20 mice used in studies of age-related coronary and neu- per genotype from two independent experiments. rodegenerative disorders.4,5 Plasma homocysteine

n engl j med 357;4 www.nejm.org july 26, 2007 421 T h e new england journal o f medicine

mice had a significant increase in plasma homo- Eugene J. Rogers, Ph.D. cysteine levels (P<0.05, by analysis of variance ShanShan Chen, B.Sc. with Tukey’s post hoc multiple-comparison test). Amy Chan, Ph.D. In contrast, plasma homocysteine levels in the University of Massachusetts Lowell oxidatively stressed APOE−/− and APOE ε4 mice Lowell, MA 01854 [email protected] remained unchanged from values obtained in the 1. Ravaglia G, Forti P, Maioli F, et al. Homocysteine and folate presence of dietary folate. as risk factors for dementia and Alzheimer disease. Am J Clin These results provide evidence that plasma Nutr 2005;82:636-43. homocysteine may be an accurate predictor of 2. Tchantchou F. Homocysteine increase folate oxidative brain homocysteine metabolism and various consequences of folate folate deficiency under otherwise normal condi- deficiency. J Alzheimers Dis 2006;9:421-7. tions but may not reflect folate deficiency in the 3. Marin DB, Breuer B, Marin ML, et al. The relationship be- presence of underlying risk factors known to pro- tween apolipoprotein E, dementia, and vascular illness. Athero- sclerosis 1998;140:173-80. mote increased oxidative stress. In the future, an 4. Ramassamy C, Krzywkowski P, Averill D, et al. Impact of accompanying test or series of tests to assess the apolipoprotein E deficiency on oxidative insults and antioxidant degree of cellular oxidative stress may be warrant levels in the brain. Brain Res Mol Brain Res 2001;86:76-83. 5. Zhu X, Raina AK, Lee HG, Casadesus G, Smith MA, Perry G. ed in order to interpret plasma homocysteine Oxidative stress signaling in Alzheimer’s disease. Brain Res 2004; values correctly. 1000:32-9.

Radiofrequency Ablation of a Tumor Causing Oncogenic Osteomalacia

To the Editor: Oncogenic osteomalacia is a rare femoral head that was detected on coregistration syndrome that is usually driven by small, mesen- of positron-emission tomography and computed chymal tumors that express phosphatonins, tomography (known as PET–CT).3 Laboratory proteins that decrease the abundance of sodium– analyses revealed hypophosphatemia (serum phos phosphate cotransporters in the proximal renal phate level, 0.45 mmol per liter; normal range, tubule. This decrease causes renal phosphate wast 0.83 to 1.67) and a reduced renal tubular maxi- ing and leads to the clinical features of onco- mum for the reabsorption of phosphate normal- genic osteomalacia, which include hyperphospha- ized to the glomerular filtration rate (0.2 mmol turia, hypophosphatemia, reduced or abnormal per liter; normal range, 0.8 to 1.4). In our patient, serum 1,25-dihydroxyvitamin D levels, and osteo- complete tumor removal by means of an open malacia.1,2 surgical procedure would have required total hip The standard treatment of oncogenic osteo- arthroplasty. Because of her age, we attempted to malacia is surgical excision of the mesenchymal preserve the healthy hip joint, which was not af- tumor, which rapidly and permanently abrogates fected by the tumor. Multiple tissue samples were all symptoms. However, tumor removal can be obtained with the use of needle biopsy guided by complicated, because the lesion is usually small CT. Histologic analysis revealed a benign mesen- and difficult to distinguish from the surrounding chymal tumor, confirming the diagnosis. We per- tissue. Complete excision is necessary, because formed CT-guided radiofrequency ablation to de- the syndrome typically persists if any tumor tissue stroy the lesion, thereby avoiding arthroplasty. remains. To ensure complete excision, surgery may Complete tumor ablation was achieved 6 days later, require wide resection margins, causing iatrogen after a second round of radiofrequency ablation, ic tissue damage that is disproportional to the tu and was confirmed on the following day by mag- mor, which is small and benign in most cases. netic resonance imaging (MRI) (Fig. 1). A few We present the case of a 40-year-old woman weeks later, levels of biochemical markers returned with an acquired hypophosphatemic, vitamin D– to normal, and all symptoms resolved. Clinical resistant osteomalacia due to a tumor in the right follow-up at 1 year was unremarkable.

422 n engl j med 357;4 www.nejm.org july 26, 2007 correspondence

A B C

D E F

G H I

Figure 1. MRI and CT Scans of the Tumor before, during, and after Radiofrequency Ablation.

The tumor was 1.7 by 1.5 cm, as measured on a T1-weighted MRI scan (Panel A, arrow), a fat-saturated, T2-weighted MRI scan (Panel B, arrow), and a fat-saturated, T1-weighted MRI scan with gadolinium (Panel C, arrow). Probe placement is shown on a CT scan obtainedICM duringAUTHOR the firstHesse radiofrequency ablationRETAKE (Panel1st D). Follow-up MRI studies REG F FIGURE 1a-i 2nd revealed remaining tumor tissue, as shown on a T1-weighted scan (Panel E, arrow) and a fat-saturated, T1-weighted CASE 3rd scan with gadolinium (Panel F, arrow). TheTITLE intralesional positioning of the electrodeRevised was confirmed on a CT scan EMail obtained during the second round of radiofrequency ablationLine (Panel G).4-C Complete ablation of the tumor was achieved Enon SIZE with the second round of radiofrequency ablation,ARTIST: asmst shownH/T on a T -weightedH/T MRI scan (Panel H, arrow) and a fat- FILL Combo 1 33p9 saturated, T1-weighted MRI scan with gadolinium (Panel I, arrow). AUTHOR, PLEASE NOTE: Figure has been redrawn and type has been reset. Please check carefully. Radiofrequency ablation is a well-established This case demonstrates that CT-guided radio- procedure for selectively removingJOB: small 35704 volumes frequencyISSUE: ablation7-26-07 may be used to treat a tumor of tissue and is an effective palliative treatment causing oncogenic osteomalacia. We believe that for skeletal metastases. Radiofrequency ablation radiofrequency ablation may broaden the scope of benign primary bone tumors is generally re- of treatment options for oncogenic osteomalacia stricted to osteoid osteoma and is the standard and may offer an effective, less invasive alterna- of care in many centers.4,5 tive to classic surgery.

n engl j med 357;4 www.nejm.org july 26, 2007 423 correspondence

Eric Hesse, M.D. phosphate homeostasis: frizzled related protein-4, matrix extra- Yale University School of Medicine cellular phosphoglycoprotein, and fibroblast growth factor 23. New Haven, CT 06520-8044 Endocr Rev 2006;27:221-41. 2. [email protected] Carpenter TO. Oncogenic osteomalacia — a complex dance of factors. N Engl J Med 2003;348:1705-8. Herbert Rosenthal, M.D. 3. Hesse E, Moessinger E, Rosenthal H, et al. Oncogenic oste- Hannover Medical School omalacia: exact tumor localization by co-registration of positron 30625 Hannover, Germany emission and computed tomography. J Bone Miner Res 2007;22: 158-62. Leonard Bastian, M.D. 4. Rosenthal DI. Radiofrequency treatment. Orthop Clin North Klinikum Leverkusen Am 2006;37:475-84. 51375 Leverkusen, Germany 5. Rosenthal DI, Hornicek FJ, Torriani M, Gebhardt MC, Mankin HJ. Osteoid osteoma: percutaneous treatment with ra- 1. White KE, Larsson TE, Econs MJ. The roles of specific genes diofrequency energy. Radiology 2003;229:171-5. implicated as circulating factors involved in normal and disordered Correspondence Copyright © 2007 Massachusetts Medical Society.

instructions for letters to the editor Letters to the Editor are considered for publication, subject to editing and abridgment, provided they do not contain material that has been submitted or published elsewhere. Please note the following: •Letters in reference to a Journal article must not exceed 175 words (excluding references) and must be received within 3 weeks after publication of the article. Letters not related to a Journal article must not exceed 400 words. All letters must be submitted over the Internet at http://authors.nejm.org. •A letter can have no more than five references and one figure or table. •A letter can be signed by no more than three authors. •Financial associations or other possible conflicts of interest must be disclosed. (Such disclosures will be published with the letters. For authors of Journal articles who are responding to letters, this information appears in the published articles.) •Include your full mailing address, telephone number, fax number, and e-mail address with your letter.

Our Web site: http://authors.nejm.org

We cannot acknowledge receipt of your letter, but we will notify you when we have made a decision about publication. Letters that do not adhere to these instructions will not be considered. Rejected letters and figures will not be returned. We are unable to provide prepublication proofs. Submission of a letter constitutes permission for the Massachusetts Medical Society, its licensees, and its assignees to use it in the Journal’s various print and electronic publications and in collections, revisions, and any other form or medium.

424 n engl j med 357;4 www.nejm.org july 26, 2007 T h e new england journal o f medicine

book reviews

Culturing Life: How Cells Became Technologies

By Hannah Landecker. 276 pp., illustrated. Cambridge, MA, Harvard University Press, 2007. $35. ISBN 978-0-674-02328-4.

flask or dish containing nutrient A medium and living cells is such a usual sight in a biologic or medical laboratory that we hardly notice it. Hannah Landecker’s fascinating, beauti- fully written account of the history of cell culture restores the sense of wonder felt by the first scientists who grew living cells apart from organisms and by the people who read about their achievements in scientific journals, popular mag- Color-Enhanced Scanning Electron Micrograph azines, and newspapers. But this book does much of a Human Embryo at Day 3. more than that; it sheds a unique light on the his- The egg was fertilized in vitro and developed in cul- tory of biology in the 20th century, the rise of bio- ture. An acid solution was used to make a hole in the zona pellucida, which allowed for the removal of indi- technology, and our understanding of what life is. vidual cells. Landecker is interested in the discovery of the Yorgos Nikas/Wellcome Images. plasticity of life and the far-reaching consequenc es of this discovery. Culturing Life follows the trans formation of living cells into techno-scientific multiple meanings of “culture” and “milieu,” but objects between 1907 and 1970. She starts by she does not use these terms as mere symbols or describing the work of Ross Harrison, the Amer- metaphors. She is interested instead in straight- ican embryologist who was the first to success- forward descriptions of technical devices, from fully cultivate isolated cells outside the body. She the development of the mixtures of nutrients that then investigates the next steps in the transforma made it possible for cells to survive in a test tube tion of cells in the laboratory — the establish- to the use of glycerol in the freezing and thaw- ment of the first cell lines, the development of ing of living cells. Landecker focuses on practical, methods of freezing and thawing living cells, the material issues — what scientists do, what mate- massive diffusion of cell lines in biology labora- rials they use, and what the consequences of their tories, and studies on the somatic hybridization activities are. The scientific activity at the center of cells. She discusses in parallel the technologi- of the book is the transformation of cells into cal applications of cells — such as their use in the entities that are detachable from the body, can be production of other living entities (viruses, for kept in a frozen state, are able to multiply indefi- example) and of devices such as viral vaccines. nitely in an identical or a quasi-identical form, Finally, Landecker’s chapter on the HeLa cell line and are circulated, exchanged, and appropriated follows the rich cultural and symbolic meanings by biologists. of a cell line made to stand for technological This is a book about biology, not medicine. progress, contamination and insufficient quality Landecker concludes with an invitation to reflect control, ownership of body parts, patients’ rights, on the ways in which the cultivation of cells race, and gender. apart from organisms has changed our under- Landecker focuses on the various material standing of living matter. Debates about stem techniques that made cell culture possible. An cells or cloning tend to focus on a philosophical anthropologist of science, she is attuned to the question: How do new biotechnologies change

n engl j med 357;4 www.nejm.org july 26, 2007 425 The new england journal of medicine

the ways we understand the meaning of being committees and decisions made by institutional human? It may be useful, Landecker proposes, to review boards. start by thinking about how these technologies All of this raises vexing questions. Who has have changed what it means to be biologic. The the authority to call themselves a bioethicist? discovery that the biologic is a fundamentally What expertise do bioethicists actually offer? plastic entity indeed explains much of the real When is a bioethicist’s opinion sufficiently reli- power of biotechnology in present-day culture. In able and helpful that a court should consider it? her closing chapter, however, Landecker tells a And what weight should a court of law give to different story — about the use of cells cultured bioethics? There is a danger of confusing bioeth- from amniotic fluid in prenatal diagnosis. In a ics and law — both speak of rights and duties, cytology laboratory, the capacity to grow cells but they are governed by different core docu- apart from organisms may have far-reaching prac ments, norms, and processes. tical consequences. By reflecting on the catastroph Bioethics in Law helpfully collects judicial opin- ic, artificial, and radically new variety of life that ions considering bioethics material, testimony, arises in the laboratory, Landecker opens new and briefs to explore the ways in which bioethics ways to articulate the ideas of the French philoso- is entering U.S. courtrooms. Spielman explores pher of science Georges Canguilhem about the how bioethics can help and hurt the judicial pro- distinction between the normal and the patho- cess. She argues that when bioethics informa- logic. Cultured cells bring into question our un- tion and opinion support established legal rights derstanding of the biologic, but their infinite and norms, judges use bioethics, but when it runs plasticity, as Landecker eloquently shows, often counter to law, judges tend to reject it. Spielman unfolds in the normative space of medicine. generally approves of this, writing, “Legal and bio Ilana Löwy, Ph.D. ethical reasoning will, one hopes, remain dis- Centre de Recherche Médecine, Science, Santé et Société (CERMES) tinct, continuing to operate alongside each other 94801 Villejuif, France and to influence each other.” She catalogues a [email protected] variety of ways in which judges consider bioethics, sometimes to help establish the facts of the case, and sometimes to help clarify the relevant policy and rules. She quotes telling exchanges Bioethics in Law among bioethics witnesses, lawyers, and judges By Bethany J. Spielman. 181 pp. Totowa, NJ, Humana Press, — some bioethics testimony is clear and relevant, 2007. $69.50. ISBN 978-1-58829-434-0. whereas other testimony is painful to read. Spielman aims this book first at lawyers and hroughout its nearly 40-year history, bioethicists but writes accessibly for a broader au Tmodern bioethics has been intimately connect dience. The book will advance the conversation ed with law. The establishment of U.S. regulations on the relationship between law and bioethics. governing human subjects research and the judi- There are nonetheless some gaps in the book’s cial recognition of patients’ rights to consent to coverage. It never addresses the question of who or refuse treatment are foundational events in the can claim to be a bioethicist (a big controversy in rise of bioethics. Bioethics grows out of philoso- a field that lacks formal credentialing); nor does phy, theology, and medicine as well, but its debt the book address the role of the bioethicist as a to law is substantial. consultant at the lawyer’s table (where influen- As bioethics reaches middle age, a number of tial bioethicists have sat) or as a defendant when scholars are trying to dissect this complex rela- bioethicists have been sued (a development that tionship. Bethany Spielman has written a range of is dismaying to bioethicists but inevitable). The articles on the topic and now this book, which book also would have benefited from more ex- focuses on an important piece of the puzzle — pansive discussion, especially to synthesize the the role of bioethics in the courtroom. Increas- trends Spielman sees and to elaborate normative ingly, bioethicists serve as expert witnesses, sub- recommendations. She thanks bioethicists who mit amicus briefs, and see their work cited in have shared with her their expert testimony and judicial opinions. Judges also consider whether to legal briefs. Given the important material she has allow into evidence advice from health care ethics collected, it would have been helpful to include a

426 n engl j med 357;4 www.nejm.org july 26, 2007 correction table documenting how often each type of bioeth The authors liken their approach to a “magical ics evidence or testimony has been offered, wheth medical mystery tour,” which seems apt given er the court allowed it, and how it was used. their tendency to devote similar degrees of atten- Bioethics in Law offers a treasure trove of infor- tion to the incidental and the significant. A sub- mation that cannot readily be found elsewhere, theme that runs through several chapters is that shedding light on the complex interaction be- the scientific “maverick” — which is how Moalem tween bioethics and law. Spielman’s pathbreak- describes himself — will eventually win out ing research and thoughtful organization of the against the skeptical establishment “expert.” Sug- issues provide an important resource for those gestive findings, such as E. Fuller Torrey’s re- concerned about the confusion over the roles of search on the relationship between infection with bioethics and law. After all, as Spielman suggests, Toxoplasma gondii and schizophrenia, are presented we should not want judges to defer to bioethi- not quite as established science but with rather cists to determine our legal rights, but neither more implication than the original researcher should we want judges determining our rights as might claim for them. patients, research subjects, or health professionals The uncritical assemblage of fascinating facts by ignoring the insights that bioethics can offer. and intriguing preliminary research in Survival of Susan M. Wolf, J.D. the Sickest reflects a tendency I have observed in University of Minnesota other popular texts on evolutionary biology, which Minneapolis, MN 55455 is to veer away from science — which depends on [email protected] tight arguments that withstand the scrutiny of empirical repetition — and toward the beguiling hypotheticals of scientism. In certain contexts, genetic diseases may confer some evolutionary Survival of the Sickest: advantages, and the authors provide some en- A Medical Maverick Discovers gaging case studies, but this is hardly news. To Why We Need Disease their credit, the authors explain complicated phe- By Sharon Moalem, with Jonathan Prince. 267 pp. New York, nomena in clear and jargon-free prose. Nonethe- William Morrow, 2007. $25.95. ISBN 978-0-06-088965-4. less, they and their publisher should have sought a higher standard, as both their subject and the his book, intended for general read- intelligent lay audience deserve better. Ters, reviews recent research on a major ques- Robert Martensen, M.D., Ph.D. tion at the intersection of evolutionary biology and Brody School of Medicine at East Carolina University medicine: Why does evolution, which, as Moalem Greenville, NC 27858 and Prince write in their introduction, “is sup- [email protected] posed to weed out harmful traits and promote Book Reviews Copyright © 2007 Massachusetts Medical Society. helpful ones,” allow serious genetic diseases to persist? The answers, as the authors make clear, Corrections are complicated and incomplete, requiring responses from many research communities. Diabetic Gastroparesis (February 22, 2007;356:820-9). The last Moalem, a researcher in neurogenetics and sentence of the second paragraph under Impaired Gastric Emp- evolutionary medicine, and Prince, a former advi tying in Patients with Diabetes (page 821) should have read “In ser and speechwriter in the Clinton White House, a clinical trial of exenatide, nausea occurred in 57% of patients, have not written a comprehensive review but in- and vomiting occurred in 17% of patients; nausea or other gastrointestinal symptoms were identified as the reason for with- stead focus on specific examples of genetic traits drawal from the study in 6% of patients,” rather than “vomiting that can prove helpful or harmful depending on occurred in 19% of patients, leading to the cessation of treat- the environmental context or general evolutionary ment in about one third of patients.” Also, the second sentence process, such as the coevolution of parasites and under Areas of Uncertainty (page 825) should have read “Agents such as the 5-HT4–receptor agonist tegaserod (which is approved their hosts. The result is a grab bag that con- for the treatment of patients with the irritable bowel syndrome tains discussions of Barbara McClintock’s “jump- in whom constipation is predominant and patients with chronic ing genes,” bubonic plague, pinworms, telo- constipation) and acetylcholinesterase inhibitors (e.g., pyridostig- meres, work on heritability in the former Soviet mine) have been used off-label in patients with gastroparesis, but data from clinical trials providing support for their use are Union, frozen wood frogs, and Jean-Baptiste La- lacking,” rather than “acetylcholine inhibitors.” The text has marck, to name just a few. been corrected on the Journal’s Web site at www.nejm.org.

n engl j med 357;4 www.nejm.org july 26, 2007 427 notices

Posaconazole or Fluconazole for Prophylaxis in Severe Graft- CALL FOR APPLICATIONS versus-Host Disease (January 25, 2007;356:335-47). The second The National Rosacea Society is accepting applications for its sentence of the third paragraph under Efficacy (page 341) should research grants program. Deadline for submission is Sept. 15. have read “In other analyses, posaconazole was superior to flu- Contact the National Rosacea Society, 800 S. Northwest Hwy., conazole in reducing the incidence of proven or probable asper- Suite 200, Barrington, IL 60010; or call (888) 662-5874; or e-mail gillosis (odds ratio, 0.31; 95% CI, 0.13 to 0.75; P = 0.006) during [email protected]; or see http://www.rosacea.org/grants/. the treatment period and was superior to fluconazole in reduc- 9TH SEMINAR OF THE EUROPEAN SOCIETY ing the incidence of breakthrough proven or probable invasive OF CONTRACEPTION fungal infections (odds ratio, 0.30; 95% CI, 0.12 to 0.71; P = 0.004) The seminar, entitled “From Abortion to Contraception,” and invasive aspergillosis during the exposure period (odds ra- will be held in Bucharest, Romania, Sept. 21 and 22. tio, 0.17; 95% CI, 0.05 to 0.57; P = 0.001)” rather than “invasive Contact Mrs. Nancy Hablis, European Society of Contracep- aspergillosis during the treatment period.” The text has been tion, Opalfeneweg 3, 1740 Ternat, Belgium; or call (32) 2 582 08 corrected on the Journal’s Web site at www.nejm.org. 52; or fax (32) 2 582 55 15; or e-mail congress@contraception-esc. com; or see http://[email protected]. VIENNA MEDICAL ACADEMY notices The following meetings will be held: “XX International Workshop on Helicobacter and Related Bacteria in Chronic Di- gestive Inflammation (EHSG)” (Istanbul, Turkey, Sept. 20–22) and “2nd Joint Meeting of the European Federation of Auto- Notices submitted for publication should contain a mailing nomic Societies and the American Autonomic Society” (Vienna, address and telephone number of a contact person or depart- Oct. 10–13). ment. We regret that we are unable to publish all notices Contact Vienna Medical Academy, Alser Strasse 4, 1090 Vi- enna, Austria; or call (43) 1 405 13 83 0; or fax (43) 1 407 82 74; received. Notices also appear on the Journal’s Web site or e-mail [email protected] or [email protected], (www.nejm.org/meetings). The listings can be viewed in respectively; or see http://www.helicobacter.org or http://www. their entirety or searched by location, month, or key word. efasweb.com/2007, respectively. UNIVERSITY OF TORONTO NORTH CAROLINA OCCUPATIONAL SAFETY The following meetings will be held in Toronto, unless other AND HEALTH EDUCATION & RESEARCH CENTER wise indicated: “4th Annual G-I-N Conference 2007: Collabora- The following courses will be offered in Chapel Hill, NC, tion in Clinical Practice Guidelines” (Aug. 22–25; sponsored by unless otherwise indicated: “Supervising Lead Abatement Pro- the Guidelines International Network); “International Ocular grams” (Aug. 14); “Comprehensive Industrial Hygiene (CIH) Blood Flow Symposium” (Oct. 13); “Update in Surgical Oncol- Review Course” (Sept. 17–21); “Asbestos Operations and Main- ogy 2007” (Oct. 26); “International Interprofessional Wound tenance” (Oct. 2 and 3); “Building Inspection and Management Care Course” (Oct. 19–22, April 11–14); “CME Update in Pa- Planning for Asbestos” (Oct. 8–12; refresher course, Sept. 11, thology 2007” (Nov. 2 and 3); and “Cardiovascular Pathology: Dec. 4); “Occupational Health Nursing: Introduction to Prin- Current Concepts” (July 25–27, 2008). ciples and Practice” (Oct. 30–Nov. 1); “Supervising Asbestos Contact the Office of Continuing Education & Professional Abatement Projects” (Nov. 5–9; refresher course, Sept. 10, Dec. Development, Faculty of Medicine, University of Toronto, 500 3); “Certified Safety and Professional (CSP) Review Course” University Ave., Suite 650, Toronto, ON M5G 1V7, Canada; or (Nov. 12–16); “Designing Asbestos Abatement Projects” (Dec. call (416) 978-2719 or (888) 512-8173; or fax (416) 946-7028; or 5); and “29th Annual Occupational Safety and Health Update: e-mail [email protected]; or see http://www.cme.toronto.ca. Compliance and Beyond” (Dec. 6 and 7). Contact Occupational Safety and Health Education & Re- MAYO CLINIC SCOTTSDALE search Center, University of North Carolina, 3300 Hwy. 54 W., The following courses will be offered: “Mayo Clinic Practi- Chapel Hill, NC 27516-6248; or call (888) 235-3320 (national) cal 21st Century Clinical Neurology Review” (Kohala Coast, HI, or (919) 962-2101 (NC); or fax (919) 966-7579; or see http:// Aug. 1–4); “20th Annual Advanced Techniques in Endoscopic www.sph.unc.edu/osherc/; or e-mail [email protected]. & Robotic Gynecologic Surgery” (Maui, HI, Sept. 26–29); “2007 Parkinson’s Disease & Other Movement Disorders for the Prac- MAYO CLINIC titioner” (Kauiai, HI, Nov. 8–10); “Update in Hospital Medicine The following courses will be offered in Rochester, MN, 2007” (Tucson, AZ, Nov. 14–17); and “Mayo Clinic Interactive unless otherwise indicated: “Mayo Clinic Gastroenterology & Surgery Symposium” (Wailea-Maui, HI, Feb. 10–15). Hepatology Board Review” (Chicago, Sept. 6–9); “Mayo Clinic Contact CME Department, Mayo Clinic Scottsdale, 13400 ERCP A to Z” (Sept. 13 and 14); “Mayo Clinic Hepatology State E. Shea Blvd., Scottsdale, AZ 85259; or call (480) 301-4580; or of the Art” (Sept. 13 and 14); “Internal Medicine Review for e-mail [email protected]; or see http://www.mayo.edu/cme/. Nurse Practitioners and Physician Assistants” (Sept. 27 and 28); “Mayo Clinic Professionalism at Academic Medical Cen- ters: Challenges and Opportunities” (Oct. 4 and 5); “Mayo the journal’s web and e-mail addresses: Clinic Human Factors in Health Care: Practical Applications to Improve Patient Safety” (St. Paul, MN, Oct. 17–19); “Neurora- For letters to the Editor: authors.nejm.org diology: Practice to Innovation” (Scottsdale, AZ, Nov. 5–9); For information about the status of a submitted manuscript: and “Mayo Clinic OB GYN Clinical Reviews” (Nov. 8 and 9). authors.nejm.org Contact Mayo School of CME, 200 First St. SW, Rochester, [email protected] MN 55905; or call (800) 323-2688 (national) or (507) 280-2509 To submit a meeting notice: (MN); or fax (507) 284-0532; or see http://www.mayo.edu/cme; The Journal’s Web pages: www.nejm.org or e-mail [email protected].

428 n engl j med 357;4 www.nejm.org july 26, 2007 T h e new england journal o f medicine

videos in clinical medicine

Positive-Pressure Ventilation with a Face Mask and a Bag-Valve Device

Rafael Ortega, M.D., Abdel K. Mehio, M.D., Albert Woo, M.D., and Dina H. Hafez, B.A.

Indications From the Department of Anesthesiology, Providing positive-pressure ventilation with a face mask and a bag-valve device can Boston University Medical Center, Boston. be a lifesaving maneuver. Although seemingly simple, the technique requires an Address reprint requests to Dr. Ortega at the Department of Anesthesiology, Boston understanding of the airway anatomy, the equipment, and the indications. University Medical Center, 88 E. Newton Face-mask ventilation is used in patients who have respiratory failure but are St., Boston, MA 02118, or at rortega@bu. still breathing spontaneously and in patients with complete apnea.1 Face-mask edu. ventilation can be indicated in any situation in which spontaneous breathing is N Engl J Med 2007;357:e4. failing or has ceased, including cardiopulmonary arrest. Copyright © 2007 Massachusetts Medical Society. Contraindications Face-mask ventilation is rarely contraindicated. However, caution is advised in patients with severe facial trauma and eye injuries.2 In addition, foreign material (e.g., gastric contents) in the airway may lead to aspiration pneumonitis. In these circumstances, alternative approaches, including endotracheal intubation, may be necessary.

Equipment There are many types of face masks, varying in design, size, and construction ma- terials. Transparent masks are preferred because they allow for inspection of lip color, condensation, secretions, and vomitus.3 To maintain a good seal, the mask’s size and shape must conform to the facial anatomy. Thus, several mask shapes and sizes should be readily available. Various bag-valve designs are available. All have a self-inflating bag and a non- rebreathing, unidirectional valve. The valve is designed to function during both spontaneous and manually controlled ventilation. Because bag-valve devices can operate without an oxygen source, it is important to ascertain that supplemental oxygen is flowing through the bag-valve device when supplemental oxygen is indicated and available. Test the bag-valve device’s capability for delivering positive-pressure ventilation before use. This can be achieved by sealing the bag-valve device connector with your thumb and squeezing the bag with reasonable force. If it is difficult to compress the bag or if air is forced between the connector and your thumb, positive pressure can be delivered. Whenever possible during face-mask ventilation, suction should be readily available. You may need to use airway-management adjuncts, such as disposable oral or nasal airways. Before beginning face-mask ventilation, examine the patient’s oral cavity. If possible, remove any dental prostheses or other foreign bodies that might be swallowed or aspirated.1

e n engl j med 357;4 www.nejm.org july 26, 2007 Positive-Pressure Ventilation with a Face Mask and a Bag-Valve Device

One-hand Technique The most common method used to hold the mask requires placing your thumb and index finger on the body of the mask while your other fingers pull the jaw forward and extend the head. Place your middle and ring fingers on the ridge of the mandible and the fifth finger behind the angle of the mandible.1,3 4 The tongue is the most common cause of airway obstruction. It is important to 4.5 cm minimize the pressure applied to the submandibular soft tissues because pressure may further obstruct the airway by pushing the tongue against the palate.1 Main- taining an adequate seal while extending the patient’s head, thrusting the jaw forward, and squeezing the bag with the other hand may require considerable manual strength and coordination. Extreme caution is advised in patients with cer- vical spine injuries, in which flexion or extension of the neck is contraindicated. In this situation, the jaw-thrust maneuver alone, without head extension, is recommended.4 Figure 1. Mask sizes

TWO-hand Technique You might find it difficult or impossible to maintain an adequate seal with only one hand. This is particularly true in the case of obese or edentulous patients or those with abundant facial hair. In these situations, hold the mask with two hands, with each hand positioned as described in the one-hand technique. A second person should compress the bag-valve device.1 Regardless of the technique you use to ventilate the patient with a face mask, you can assess adequate ventilation by inspecting and auscultating the chest and abdomen. The rising and falling of the chest and breath sounds synchronous with the delivered tidal volume suggest adequate ventilation. Epigastric sounds and abdominal distension indicate gastric insufflation and poor ventilation. Figure 2. Positioning the mask

Using Oropharyngeal and Nasopharyngeal Airways Occasionally, it may be difficult or impossible to provide ventilation unless a disposable oral or nasal airway is inserted. These devices are most helpful when the cough and gag reflexes are absent. Insertion in patients with intact reflexes may precipi- tate coughing, vomiting, and laryngospasm. When the use of a disposable oral or nasal airway is necessary, you must select the appropriate-sized device to avoid wors- ening the airway obstruction.1 Estimate the correct size of an oral airway by hold- ing it next to the patient’s mouth. The tip should reach the angle of the mandible. Insert the oropharyngeal airway by depressing the tongue with a tongue blade and advancing the airway toward the base of the tongue. Alternatively, you can insert the airway upside down and then rotate it 180 degrees as it is being advanced Figure 3. Measuring the oral airway posteriorly. Nasopharyngeal airways are better tolerated than oral airways when airway reflexes are present. They are useful when the patient’s mouth cannot ºbe opened. The simplest method of estimating their appropriate length is by correlating it with the external anatomy of the face and neck. Nasopharyngeal airways should be lubricated and advanced perpendicular to the face. Use them only with extreme caution in patients with facial injuries, basilar skull fractures, and coagulopathy, weigh- ing the risk of further injury and bleeding against the need for oxygenation.1 When the patient is breathing spontaneously, you must synchronize the delivered tidal volume with the patient’s inspiration. Regardless of the presence of spontaneous respiratory effort, when excessive pressure is delivered to the airway, gastric insufflation may occur. This may lead to a vicious cycle of increased intra abdominal pressure, which requires higher peak inspiratory pressures, predisposing patients to vomiting or regurgitation.

n engl j med 357;4 www.nejm.org july 26, 2007 Positive-Pressure Ventilation with a Face Mask and a Bag-Valve Device

Complications References Complications, including corneal abrasions and blindness in the presence of eye 1. Barash PG, Cullen BF, Soelting RK. Clinical anesthesia. 5th ed. Philadelphia: injury, can occur. Soft-tissue injuries, including injuries to the nose and lips, may Lippincott Willams & Wilkins, 2006. result when excessive pressure is applied. 2. Morgan GE Jr, Mikhail MS, Murray MJ. Clinical anesthesiology. 3rd ed. New York: McGraw-Hill, 2002. Summary 3. Miller RD, ed. Miller’s anesthesia. Discontinuation of face-mask ventilation depends on clinical circumstances. Pa- 6th ed. New York: Churchill Livingstone, tients may require a more permanent or effective method of airway management, 2005. 4. Wilson WC, Grande CM, Hoyt DB. such as endotracheal intubation or tracheostomy. On other occasions, all that is Trauma: emergency resuscitation, periop- needed for patients to recover completely is effective face-mask ventilation with erative anesthesia, surgical management. New York: Informa Healthcare, 2007. oxygen. Copyright © 2007 Massachusetts Medical Society. No potential conflict of interest relevant to this article was reported.

n engl j med 357;4 www.nejm.org july 26, 2007 T h e new england journal o f medicine

images in clinical medicine

Ascending Aortic Aneurysm in a Young Adult

A B

C D

37-year-old man presented to the emergency department with Paul D. Harris, M.B., Ch.B. chest pain of 2 ICMdays’ AUTHORduration. HarrisThe pain was heavyRETAKE in character,1st intermittent, Chris Cokis, F.A.N.Z.C.A. REG F FIGURE a-c 2nd and made worse by deep inspiration and lying flat. The man3rd had no shortness CASE TITLE Royal Perth Hospital A Revised of breath and had beenEMail actively working asLine a gardener4-C until the onset of the pain. Perth, WA, Australia SIZE He was an active smoker,Enon andARTIST his :fathermst hadH/T had aH/T myocardial infarction at the age of 52 years. He did notFILL have hypertension.Combo Auscultation28p of the precordium was AUTHOR, PLEASE NOTE: notable for an aortic regurgitant Figure has been murmur. redrawn and Electrocardiography type has been reset. showed left ventricular hypertrophy. Chest radiographyPlease checkrevealed carefully a .widened mediastinum (Panel A). 3-D reconstruction computed tomographic angiography of the chest showed a 9.5- cm ascending aortic aneurysmJOB: 35704 (Panel B). The patientISSUE: underwent7-26-07 an ascending aortic root replacement, during which the aneurysm was visible through the median sternotomy (Panel C), and a Dacron graft was successfully placed (Panel D). The patient had a full and uneventful recovery. Copyright © 2007 Massachusetts Medical Society.

n engl j med 357;4 www.nejm.org july 26, 2007 e