New England Journal of Medicine(March 2005)

New England Journal Of Medicine

Volume 352 March 24, 2005 Number 12

Article Summaries Renal Papillary Necrosis Unmasking Sickle Cell Disease M. Voulgarelis and P. D. Ziakas

The Trouble with Uncompensated Hospital Severe Fecal Impaction Care Y. Loubières and O. Chereau J. S. Weissman

Safety in Numbers — Monitoring Risk in Approved Drugs Case 9-2005 — A 67-Year-Old Man with S. Okie Acute Respiratory Failure D. M. Systrom and C. Wittram Herbal Medicine in Europe — Relaxing Regulatory Standards P. A.G.M. De Smet Concurrent Antiplatelet and Fibrinolytic Therapy R. A. Lange and L. D. Hillis

Addition of Clopidogrel to Aspirin and Fibrinolytic Therapy for Myocardial Defining the Stages of Aggressive Non- Infarction with ST-Segment Elevation Hodgkin's Lymphoma — A Work in M. S. Sabatine and Others Progress J. O. Armitage

Hospitalization for Mental Illness among Financing Health Care — Finding the Parents after the Death of a Child Money Is Hard and Spending It Well Is J. Li and Others Even Harder R. Kronick ACVBP versus CHOP plus Radiotherapy for Localized Aggressive Lymphoma F. Reyes and Others Health Care Vouchers — A Proposal for Universal Coverage Day–Night Pattern of Sudden Death in E. J. Emanuel and V. R. Fuchs Obstructive Sleep Apnea A. S. Gami, D. E. Howard, E. J. Olson, and Do We Really Want Broad Access to Health V. K. Somers Care? J. J. Mongan and T. H. Lee

Palifermin and Chemotherapy-Induced Oral Pertussis — Not Just for Kids Mucositis E. L. Hewlett and K. M. Edwards Care of the Potential Organ Donor

Medical Progress: Polycystic Ovary Contagious Acute Gastrointestinal Infections Syndrome D. A. Ehrmann Cancer of the Ovary

Case 37-2004: Postmenopausal Bleeding and a Cystic Ovarian Mass

Why "Why" Matters

Nosocomial Transmission of Cryptococcosis

Szasz Under Fire: The Psychiatric Abolitionist Faces His Critics

Message in a Bottle: The Making of Fetal Alcohol Syndrome

Hideous Absinthe: A History of the Devil in a Bottle

Masters of the Mind: Exploring the Story of Mental Illness from Ancient Times to the New Millennium

The Effect of Air Pollution on Lung Development from 10 to 18 Years of Age

Combination of Isosorbide Dinitrate and Hydralazine in Blacks with Heart Failure

Anti–Interleukin-12 Antibody for Active Crohn's Disease

The new england journal of medicine

correspondence 1264 Palifermin and Chemotherapy- Induced Oral Mucositis 1266 Care of the Potential Organ This Donor 1267 Contagious Acute Gastro- intestinal Infections 1268 Cancer of the Ovary We ek 1269 Case 37-2004: Postmenopausal Bleeding and a Cystic Ovarian Mass 1270 Why “Why” Matters 1271 Nosocomial Transmission in the of Cryptococcosis book reviews 1273 Szasz under Fire: The Psychiat- Journal ric Abolitionist Faces His Critics march 24, 2005 1274 Message in a Bottle: The Making of Fetal Alcohol Syndrome 1275 Hideous Absinthe: A History of the Devil in a Bottle 1276 Masters of the Mind: Exploring original article the Story of Mental Illness from Ancient Times to the New Clopidogrel with Fibrinolysis in Myocardial Infarction Millennium To establish reperfusion within the infarct-related continuing medical education coronary artery, acute 1277 Polycystic Ovary Syndrome myocardial infarction is 1278 Pertussis — Not Just for Kids often treated with a 1279 Day–Night Pattern of Sudden combination of fibrino- Death in Obstructive Sleep Apnea lytic agents, heparin, and aspirin. Despite this therapy, reperfusion is unsuccessful in some patients and reocclusion occurs in others. The ad- Next Week dition of the antiplatelet in the Journal agent clopidogrel to the march 31, 2005 regimen substantially improved the rate of reperfusion without a sig- Integrity of the NIH nificant increase in bleeding complications. The results of this study should prompt rethinking of the protocol for reperfusion therapy in pa- Robert Steinbrook tients with acute myocardial infarction. see p. 1179; editorial, p. 1248

n engl j med 352;12 www.nejm.org march 24, 2005 1169

The new england journal of medicine

original article original article original article Mental Illness Chemotherapy Alone Sudden Death from Cardiac Causes after Parental Bereavement for Localized Lymphoma in Obstructive Sleep Apnea In a large study that made use of This large trial of the treatment of lo- The risk of sudden death from cardiac national registers in Denmark, the calized aggressive lymphoma in pa- causes is highest between the hours authors show that the relative risk of tients under the age of 60 years com- of 6 a.m. and noon and is lowest from a first hospitalization for psychiatric pared intensive chemotherapy with midnight to 6 a.m. This study found illness was significantly higher among chemotherapy plus involved-field ra- that in patients with obstructive sleep parents who lost a child than among diotherapy. Superior overall and dis- apnea, this pattern is altered — the those who did not and was higher ease-free survival rates were obtained risk of sudden death from cardiac among bereaved mothers than among with intensive chemotherapy alone. causes was much higher from mid- bereaved fathers. Treatment of the B-cell lymphomas is night to 6 a.m. than during the other see p. 1190 in flux. Is a trial of chemotherapy of hours of the day. The increase in risk B-cell lymphomas outdated because may be due to a higher incidence of it does not include rituximab, a mono- episodes of apnea and hypopnea dur- clonal antibody against B cells? We ing these hours. cannot be sure. Given a five-year see p. 1206; cme, p. 1279 overall survival rate of 90 percent with intensive chemotherapy alone, there is little room for substantial improvement. see p. 1197; editorial, p. 1250

clinical practice medical progress sounding board Pertussis — Not Just for Kids Polycystic Ovary Syndrome Health Care Vouchers — A Proposal for Universal Coverage Six weeks ago, a 45-year-old woman The polycystic ovary syndrome is one noticed a scratchy feeling in her throat of the most common hormonal dis- Emanuel and Fuchs outline their pro- that has now progressed to more than orders affecting women. It has mul- posal for a universal health care 20 episodes of severe, spasmodic tiple components — reproductive, voucher system. Americans would re- coughing per day. Her coughing spells metabolic, and cardiovascular — with ceive vouchers to purchase coverage are worse at night and are sometimes health implications for the patient’s from their choice of private health associated with gagging and vomiting. entire life span. This review addresses plans. The system would replace Med- Her adolescent son and several of his current concepts regarding the diag- icaid, Medicare, and employer-based friends had similar illnesses several nosis, cause, and treatment of the insurance. The authors propose fund- weeks before the onset of her symp- condition. ing vouchers with a national value- toms, and they continue to cough. see p. 1223; cme, p. 1277 added tax and argue that this would How should the patient be assessed control costs, because increases in health care spending would require for possible pertussis? Should she be case records of the treated and, if so, how? Could this ill- massachusetts general hospital public approval of tax increases. see p. 1255; editorial, p. 1252 ness have been prevented? A Man with Acute Respiratory Failure see p. 1215; cme, p. 1278 A 67-year-old man was admitted to the hospital with acute respiratory failure. Do We Really Want Broad Access He had never smoked, had not trav- to Health Care? eled, but had recently been exposed to In the United States, about 45 million bird droppings. Cultures and serologic people do not have health insurance. test results were negative. With antibi- Mongan and Lee argue that even otic therapy and respiratory support, though most Americans say they be- his symptoms rapidly improved over lieve in access to health care for all, the 48 hours. Three days after discharge, public is not willing to pay for it. They his symptoms returned shortly after urge health care providers to assume using his hot tub, and again abated a leadership role and advocate for the rapidly after admission to his local tax increases necessary to provide B. pertussis hospital. health insurance to everyone. see p. 1238 see p. 1260; editorial, p. 1252

1170 n engl j med 352;12 www.nejm.org march 24, 2005

PERSPECTIVE The Trouble with Uncompensated Hospital Care

bacco litigation several years ago. In March 2004, 18 8.0 Scruggs was approached by Dr. Charles Bagnado, a 17 7.5 Uninsured Mississippi surgeon, and Charles Rehberg, the fi- 7.0 16 nancial administrator of Bagnado’s practice. After 6.5 15 being blocked from opening an outpatient surgery 6.0 center across the street from the hospital where 14 5.5 Bagnado worked, the two began investigating the 13 Uncompensated 5.0 hospital’s finances and found evidence of abusive care Uncompensated Care (%) Nonelderly Uninsured (%) 12 4.5 billing practices. The issue has since become something of a crusade for all three men. In addition to 0 0 citing aggressive billing practices similar to those 1991 1993 1995 1997 1999 2001 2003 at Yale–New Haven Hospital, Scruggs claims that Year some low-income, uninsured patients are being Trends in the Percentage of Nonelderly Americans Who overcharged for services. Rather than being billed Are Uninsured and the Percentage of Total Hospital at the same discounted rate as most insured pa- Expenses That Are Uncompensated. tients, some are billed at the “sticker price,” which Data on uncompensated care are from the American can be much higher than the actual cost of care. Hospital Association, and data on the uninsured are Why would hospitals continue trying to collect from the Employee Benefit Research Institute. from patients who obviously cannot afford to pay, rather than writing the expenses off as free care? According to Scruggs, a number of current and Journal chronicled what appeared to be the overly former hospital financial officers have told him aggressive billing practices of nonprofit hospitals. that the most common reason is to discourage in- Soon afterward, the advocacy group Community digent patients from returning to the hospital. Catalyst reported that many hospitals rarely publi- Since last fall, Scruggs has negotiated a settlement cized free or discounted care and were not forth- with one hospital, although a similar case brought coming in offering it to indigent patients. Then, in by other lawyers in Pennsylvania was dismissed. November 2003, students from a legal clinic at Yale Whatever the eventual outcome of the lawsuits, the Law School spearheaded a lawsuit against Yale– action has inspired many hospitals to review and New Haven Hospital, arguing that the hospital’s formalize their financial-assistance policies. debt-collection tactics were draconian and includ- Ultimately, uncompensated hospital care is a ed such practices as garnisheeing wages and plac- weak substitute for insurance and a poor method of ing liens or foreclosing on patients’ homes. distributing costs among providers. Even hospitals The AHA responded that Medicare rules did not whose mission includes treating indigent patients permit discounts to uninsured patients — or that, are reluctant to make the process too easy or too at least, the regulations were confusing. The AHA public for fear of becoming magnets for uninsured requested clarification from Medicare, distributed patients. Relying on uncompensated care is also un- guidelines for billing and collection procedures, economical, since it encourages uninsured, poor and asked its members to sign a pledge to review patients to seek care in emergency rooms. Because their policies. In February 2004, Health and Human such care detracts from the bottom line, it places Services Secretary Tommy Thompson responded hospitals that serve many uninsured patients at a that Medicare rules do not prohibit discounts, and disadvantage in the marketplace. The challenge, in June 2004, after a year of investigations, the then, is to encourage hospitals to act charitably House Energy and Commerce Committee began a without harming their ability to function competi- series of hearings on hospitals’ charging practices tively. and tax-exempt status. One option is to reduce the demand for uncom- Certainly the most contentious development be- pensated care by expanding coverage — an ap- gan last summer, when 13 related class-action law- proach that has met with some success in Minne- suits were brought against nonprofit hospitals in sota and Maine. But universal health coverage on a seven states on behalf of uninsured patients. The national scale is unlikely to be embraced in the near lawsuits are being coordinated by the law firm of future. Incremental reforms such as the creation Richard Scruggs, the group that won the major to- of health savings accounts (HSAs) probably won’t

1172 n engl j med 352;12 www.nejm.org march 24, 2005

PERSPECTIVE The Trouble with Uncompensated Hospital Care

have a large effect on uncompensated care, either, tals that provide large amounts of free care, but because poor people tend not to take advantage of there is some danger that they will be used merely such options and because well-insured people to shore up failing or inefficient hospitals. In addi- might switch to policies with high deductibles and tion, programs designed to expand access to spe- then find that they cannot pay their bills. cialists who work at hospitals might be cost-effec- Short of universal coverage, however, a number tive if they prevented unnecessary hospitalizations. of policy goals may well be worth pursuing. First, At the federal level, perhaps Medicare should con- transparency in pricing, financial records, and hos- sider changes to the DSH funding formula to ensure pital policies could lead to more consistent practices that funds reach hospitals with large uninsured for reporting and awarding free care and bad debt populations. and to greater accountability. It would be helpful if Most hospitals and doctors are surely trying to hospitals distinguished more clearly between the do the right thing. But serving as a safety net while two and if the AHA made hospitals’ data available still functioning as a business is a challenge. Until for monitoring purposes. The financial-assistance the country decides to provide health coverage for and collection policies of hospitals could be for- all residents, the problem of uncompensated care malized and made public and could be better coor- will not go away. dinated with public programs such as Medicaid. Several groups — including Community Catalyst 1. Hadley J, Holahan J. Who pays and how much? The cost of caring for the uninsured. Washington, D.C.: Kaiser Commission and the California Hospital Association — have on Medicaid and the Uninsured, 2003. proposed models that appear to balance the rights 2. Gabel J, Claxton G, Holve E, et al. Health benefits in 2003: and needs of low-income patients with the realities premiums reach thirteen-year high as employers adopt new forms of cost sharing. Health Aff (Millwood) 2003;22(5):117-26. of hospital survival. 3. Tu HT. Rising health costs, medical debt and chronic condi- Second, low-income, uninsured patients ought tions. Issue brief no. 88. Washington, D.C.: Center for Studying not to be asked to pay inflated prices. Third, uncom- Health System Change, 2004. 4. Overview of uncompensated care: an audio-visual confer- pensated care has to be financed somehow, and ence event sponsored by the Agency for Healthcare Research charitable contributions are generally not sufficient, and Quality. Washington, D.C.: AHRQ/MedPAC, 2002. in part because they are often earmarked for other 5. Weissman JS, Gaskin DJ, Reuter J. Hospitals’ care of uninsured patients during the 1990s: the relation of teaching status purposes. Uncompensated-care pools do a reason- and managed care to changes in market share and market con- ably good job of leveling the playing field for hospi- centration. Inquiry 2003;40:84-93.

Safety in Numbers — Monitoring Risk in Approved Drugs Susan Okie, M.D.

In most cases, when a new drug is approved, al- than 1 in 1000 patients, that effect will generally be most everything known about its safety in humans recognized only after the medication is being wide- is based on the responses of a few thousand people ly used. Moreover, if the drug increases the inci- who took it during clinical trials. But once the drug dence of a common condition, such as myocardial is on the market, the real safety testing gets under infarction, that risk, too, is unlikely to be identified way. Within a year or two, the number of people until millions of people have taken the drug. About who are exposed to the medication may climb into half the drugs that enter the market have serious ad- the millions, especially if the manufacturer pro- verse effects that are detected only after approval.1 motes it aggressively with television or print adver- And these days, more often than not, Americans tisements that target consumers. If the drug has a are the test population. Fifteen years ago, most dangerous but rare side effect — for example, liver new drugs were first approved in other countries. If failure or aplastic anemia — that occurs in fewer life-threatening side effects showed up after approval, the products never made it to the U.S. mar- Dr. Okie is a contributing editor of the Journal. ket. Today, because of speedier review of product

n engl j med 352;12 www.nejm.org march 24, 2005 1173

PERSPECTIVE Safety in Numbers — Monitoring Risk in Approved Drugs

applications by the Food and Drug Administration Brian L. Strom, a professor of public health and (FDA), more than 60 percent of new drugs are ap- preventive medicine at the University of Pennsylva- proved first in the United States. nia, favors a similar requirement and believes that That shift is a major reason why drug-policy ex- manufacturers should be prohibited from advertis- perts, lawmakers, consumer advocates, and feder- ing drugs directly to consumers until the compa- al officials are all calling for better ways of monitor- nies have gathered observational data on at least ing drug safety. The best ways to expand and 20,000 users. Consumer-targeted advertising of improve the current system will be the focus of a new drugs tends to boost the number of prescrip- new investigation by the Institute of Medicine. tions written for patients other than those for The urgency of this effort is clear: more Ameri- whom they are indicated. For example, the explo- cans are taking prescription medications than ever sive growth in sales of rofecoxib (Vioxx) was fueled before. In 2004, pharmacists filled 3.1 billion pre- chiefly by its use for pain due to arthritis in patients scriptions, 60 percent more than a decade earlier. who were at low risk for gastrointestinal bleeding Reports to the FDA of drug-related adverse events and thus could have taken a nonspecific nonsteroi- have increased correspondingly and now total dal antiinflammatory drug instead.2 “Misuse and about 375,000 per year — more than twice as many overuse of new drugs is the central source of much as a decade ago — even though the agency’s cur- of the problem,” said Strom. “The risk–benefit bal- rent surveillance system is passive, relying on the ance for a new drug is much more acceptable if it is diligence of drug companies, health care provid- used only in the people who need it.” ers, and consumers. If companies were required to collect safety data “Given how many people are exposed to drugs, after a drug has been marketed, they would be like- how quickly they’re taken up in the population, ly to pick up the rare but serious side effects that how many people take multiple drugs... we’re generally are not identified before approval. Such under no illusions that we have a good postmarket surveillance would also provide information on the system right now,” said deputy FDA commissioner drug’s behavior in groups of users, such as the el- Janet Woodcock. derly, who tend to be inadequately represented in Woodcock and other policy experts suggest that clinical trials. But if a biologic or epidemiologic the new system should include ways to gather ob- signal suggested that a drug might increase the servational data on large numbers of people who risk of a common disease — as rofecoxib increased are exposed to medications once they are on the the risk of myocardial infarction — then federal market. Such information might be collected from regulators would also need the legal authority to databases that are increasingly becoming available require that the manufacturer conduct a random- as managed-care networks and other providers ized, controlled trial to define that potential risk move to the use of electronic medical records. further. Currently, the FDA has no legal power to “The preapproval system is really designed and mandate additional safety studies once a drug has powered to detect efficacy” rather than safety, ac- been approved (see diagram). In Europe, by con- cording to Alastair J.J. Wood, a professor of medi- trast, drug approvals are reviewed again every five cine and pharmacology at Vanderbilt University years, and pharmaceutical companies pay post- School of Medicine, who chaired the recent FDA marketing fees that contribute to the cost of safety advisory committee hearing on the safety of cyclo- surveillance.3 oxygenase-2 (COX-2) inhibitors. “That’s probably Bruce M. Psaty, a professor of medicine, epide- not an inappropriate balance,” Wood said in an in- miology, and health services at the University of terview, “but we’d be more comfortable if we had a Washington in Seattle, believes that for drugs that better postapproval monitoring system.” Wood patients are likely to take for years, companies suggests allowing new medications to be marketed should be required to initiate long-term trials be- with limited FDA approval and then requiring the fore approval and to continue them after the drugs manufacturer, as a condition of retaining its patent are marketed. “For drugs that are going to be used exclusivity, to collect extensive additional data on by millions of people for many years, six-week stud- safety for several years. He favors such an approach ies are not adequate to assess the trade-off between over one that severely delays access to new drugs by risks and benefits,” he said. In the case of statins, forcing companies to conduct studies involving Psaty pointed out, long-term trials that were com- tens of thousands of users before approval. pleted after approval identified additional, unex-

1174 n engl j med 352;12 www.nejm.org march 24, 2005

PERSPECTIVE Safety in Numbers — Monitoring Risk in Approved Drugs

when it comes to assessing the risks of medica- Preclinical Research FDA Post-Marketing tions that it has approved. Graham maintains that (in vitro and in vivo) Clinical Studies Review Surveillance safety monitoring should therefore be moved out Short-term testing of the FDA. Drug-policy experts outside the federal in animals government are divided on the question of whether Long-term testing in animals a new, separate board or agency is needed. “To spin off the ODS, I think, would actually be a disaster,” said Strom. “Part of the problem now Phase 1 is a lack of communication and coordination” be- Small trials studying safety tween the review teams responsible for the approval and toxic and the labeling of drugs and the ODS epidemiolo- effects in healthy gists who search for drug-related adverse events. volunteers The creation of a separate agency could exacerbate Phase 2 that problem, he predicted. Instead, the safety office Trials of safety “needs more people, more resources, and more le- and efficacy in gal clout.” patients with the condition to But Vanderbilt’s Wood believes that creating a be treated drug-safety board separate from the FDA, similar Phase 3 to the National Transportation Safety Board, would At least 2 larger, randomized help to restore public trust and would provide a “pivotal” trials of efficacy and safety mechanism for the impartial assessment of risks Phase 4 and the discovery of effective ways to reduce them. FDA Office If a serious problem developed with an approved of Drug medication, the board would conduct an investiga- Safety monitors tion and issue a report. Pharmaceutical companies reported would face severe penalties if they withheld infor- adverse reactions mation about their products from the safety board. Usually 2000 to 5000 “When a plane crashes, we don’t turn over the inves- patients are exposed FDA cannot tigation to [the airline] and the air-traffic control- to a drug during mandate clinical studies before further safety lers,” Wood said. “We get someone else to do it.” FDA approval studies Last month, Health and Human Services Secre- tary Mike Leavitt and the newly nominated FDA Company obtains FDA Application FDA commissioner, Lester Crawford, announced a plan permission for trials for new approval in humans drug to pursue a middle ground, establishing a new approval Drug Safety Oversight Board within the FDA that would draw some of its members from inside the Outline of the Process of Obtaining FDA Approval for a New Drug. agency and some from outside. FDA employees who are involved in reviewing drugs for approval would not serve on the board, which would be free pected benefits of the drugs. “They expanded the to seek advice from members of the agency’s advi- market in ways that helped public health,” he said. sory committees and from public-interest groups. As post-marketing surveillance of drugs ex- Crawford also promised greater openness, saying pands, who should be in charge of minding the that the FDA will begin sharing much more of its safety data? The current system of identifying im- drug-safety data with the public, even in cases in portant risks depends heavily on reporting by phar- which such information is considered preliminary. maceutical companies, which have a conflict of in- “Our culture, which has received some criticism in terest when sifting through adverse-event reports past months, is not to alarm the public when we related to their own products. Whistle-blower Da- get a signal,” he told agency employees. “That era vid J. Graham, an epidemiologist in the FDA’s Of- is sort of past. What the public, we think, is de- fice of Drug Safety (ODS) who made headlines last manding is to know as soon as we know what’s go- fall when he criticized his agency’s safety standards, ing on.”4 believes that the FDA also has a conflict of interest It remains to be seen whether the new board will

n engl j med 352;12 www.nejm.org march 24, 2005 1175

PERSPECTIVE Safety in Numbers — Monitoring Risk in Approved Drugs

be truly independent. Senator Charles E. Grassley Kessler. “They’re different methodologies. Which (R-Iowa), one of the agency’s sharpest critics on one adequately reflects the reality? How much data Capitol Hill, has announced that he will introduce do you need, and how solid do the data have to be legislation to give the board the authority it needs on cause and effect?” and has called on Congress to require the registra- If the Drug Safety Oversight Board functions as tion of all clinical trials. advertised, physicians and patients may be able to The changes under way at the FDA are likely to review the evidence, listen to the debate, and judge focus public attention on a long-simmering debate for themselves. “The expectations would be that all within the agency over the level of scientific evi- viewpoints would be represented there,” said Wood. dence needed to justify restricting access to a drug “The FDA would be in the happy position of letting or removing it from the market. That tension was it all hang out.” evident during the recent advisory committee hearing on COX-2 inhibitors, as panel members and 1. FDA drug review: postapproval risks, 1976-85. Washington, FDA officials wrangled over how to weigh the find- D.C.: General Accounting Office, April 1990. (GAO/PEMD-90- 15.) ings of clinical trials against those of epidemiolog- 2. Dai C, Stafford RS, Alexander GC. National trends in cyclo- ic studies in assessing the drugs’ cardiovascular oxygenase-2 inhibitor use since market release: nonselective dif- risks and deciding whether to allow the drugs to re- fusion of a selectively cost-effective innovation. Arch Intern Med 2005;165:158-60. main on the market. 3. Psaty BM, Furberg CD, Ray WA, Weiss NS. Potential for con- “Within the agency, the really fierce debates that flict of interest in the evaluation of suspected adverse drug reac- I remember were when the pharmacoepidemiolo- tions: use of cerivastatin and risk of rhabdomyolysis. JAMA 2004; 292:2622-31. gists and the clinical-trials folks were in the same 4. Harris G. F.D.A. moves toward more openness with the pub- room,” said former FDA commissioner David A. lic. New York Times. February 20, 2005:A28.

Herbal Medicine in Europe — Relaxing Regulatory Standards Peter A.G.M. De Smet, Pharm.D., Ph.D.

Herbal medicine is big — and relatively mainstream in the United Kingdom, on the other hand, pre- — business in Europe: in 2003, European coun- scribe herbal medicines, which are generally not tries spent almost $5 billion (at manufacturers’ covered by the National Health Service, although prices to wholesalers) on over-the-counter herbal approximately 1300 herbal practitioners may law- medicines. But not all European countries have em- fully sell unlicensed herbal remedies, provided that braced herbal treatments with equal warmth. Ger- they do so after consultation with a patient. many and France are indisputably in the lead in Companies that make herbal preparations have over-the-counter sales (see graph), and they have usually found it difficult to meet the conventional re- also had noteworthy markets for prescription herb- quirements for proof of medical efficacy, and Euro- al preparations. In 2003, German health insurance pean countries have also varied in their approaches paid $283 million in reimbursements for prescribed to this issue. On their own, some countries, such as ginkgo, St. John’s wort, mistletoe, saw palmetto, ivy, Germany and France, created simplified registra- hawthorn, stinging nettle root, myrtol, phytoster- tion procedures for herbal products, whereby con- ols, and cucurbita, and in 2002, French health insur- clusive evidence of efficacy was no longer required. ance paid $91 million in partial reimbursements for Other countries, such as the United Kingdom, clung ginkgo, saw palmetto, and pygeum prescriptions to the principle that industrial herbal preparations with a total value of $196 million. Few physicians should meet the same requirements as conventional medicines, even if this meant that most herbal products could not be licensed and would there- Dr. De Smet is from the Scientific Institute of Dutch Phar- macists, The Hague; and the Department of Clinical Phar- fore continue to be sold without firm regulatory macy, University Medical Center St. Radboud, Nijmegen control. — both in the Netherlands. The European Community has taken two legis-

1176 n engl j med 352;12 www.nejm.org march 24, 2005

PERSPECTIVE Safety in Numbers — Monitoring Risk in Approved Drugs

Herbal Medicine in Europe — Relaxing Regulatory Standards Peter A.G.M. De Smet, Pharm.D., Ph.D.

1176 n engl j med 352;12 www.nejm.org march 24, 2005

PERSPECTIVE Herbal Medicine in Europe — Relaxing Regulatory Standards

well documented and their action is ancillary to

Germany 2.06 billion that of the herbs. The proposed indications for use must not require the diagnostic or therapeutic su- France 1.13 billion pervision of a medical practitioner. Products must have a specified strength and dosage and be admin- Italy 543 million istered orally, externally, or through inhalation. The Poland 252 million data on traditional use must be sufficient to show that they are not harmful in the specified conditions United Kingdom 211 million of use and that pharmacologic effects are plausi- ble. And they must have been used medicinally for Spain 170 million at least 30 years, including at least 15 years within Belgium 127 million the European Community. The new directive also called for the creation of Switzerland 93 million a special committee on herbal products within the European Agency for the Evaluation of Medicinal Austria 88 million Products, which will, among other tasks, generate The Netherlands 81 million or approve monographs on herbs and draft a list of herbal substances and preparations that may be li- Czech Republic 76 million censed as traditional herbal medicines. This list

Sales of Over-the-Counter will provide the appropriate indications, specified Herbal Medicines ($) strength and dosage, route of administration, and any other information necessary for safe use. Distribution of the $4.96 Billion European Market for Over-the-Counter The new legislation will offer public health bene- Herbal Medicines in 2003. fits, especially for member states that did not yet The remaining $132 million in sales was divided among Portugal, Hungary, have a simplified registration procedure for herbal Ireland, Slovakia, Finland, and Norway. Data are from IMS Health. The spend- medicines. It will help to clarify which preparations ing per capita (calculated on the basis of the manufacturers’ prices to wholesalers and recent United Nations projections of 2003 populations) was $25.00 are on the market and who is legally responsible for in Germany, $18.80 in France, $9.50 in Italy, $6.50 in Poland, $3.60 in the United putting them there. This clarification will facilitate Kingdom, $4.10 in Spain, $12.30 in Belgium, $13.00 in Switzerland, $10.90 recalls and may open new venues for cooperation in Austria, $5.00 in the Netherlands, and $7.40 in the Czech Republic. between health authorities and the industry for safety improvement. The directive will guarantee a pre- marketing check of product quality and safety by lative steps to harmonize these differences. In 1999, the health authorities, reducing the risk that prep- it adopted a directive permitting greater flexibility arations with an unsafe ingredient, adulterant, or in the use of bibliographic data in proving the safe- contaminant will enter the market. The planned ty and efficacy of so-called well-established medi- positive list of acceptable herbs is preferable to any cines. Applicants for a marketing authorization are negative listing, which could only swab up leaks no longer required to submit data from new preclin- while leaving the tap running. The directive also ical tests and clinical trials if they can demonstrate permits health authorities to require labeling that through detailed references to published scientif- includes all necessary warnings against incorrect ic literature that their herbal products have a well- or unsafe uses. Furthermore, companies that hold established medicinal use with recognized effica- licenses for herbal medicines will be obliged to cy and an acceptable level of safety. conduct adequate post-marketing surveillance and In March 2004, another directive was passed, to report all suspected adverse events to the author- creating a simplified registration procedure for all ities. traditional herbal medicines that, despite long By introducing a list of acceptable herbs and use, do not fulfill the requirements for classifica- mandating adverse-event reporting, the European tion as well-established medicines. To be eligible legislation goes beyond the new rule for Good Man- for this registration, herbal medicines must meet ufacturing Practice that the U.S. Food and Drug Ad- several criteria. They must contain exclusively one ministration proposed in 2003 to improve the qual- or more herbal ingredients; the addition of vita- ity of dietary supplements. It should therefore be an mins or minerals is allowed only if their safety is inspiration for anyone striving for better regulation

n engl j med 352;12 www.nejm.org march 24, 2005 1177

PERSPECTIVE Herbal Medicine in Europe — Relaxing Regulatory Standards

Second, the new directive explicitly allows non- medicinal herbal products to be regulated under legislation covering foods. This raises the question of to what extent it will remain possible to market the same herb as both a medicine and a food sup- plement. Furthermore, positive listings and general monographs on herbs may not necessarily guarantee that all products licensed on their basis have the same chemical and pharmacologic properties. In Germany, which has had a simplified registration procedure based on general monographs for many years, the same herb may appear in both phy- tomedicines and food supplements. And recent studies of St. John’s wort preparations on the Ger- man market have shown substantial variation in the level and in vitro dissolution of a major constit- uent, hyperforin. Third, traditional experience is not always a reliable tool for the detection of rare or late reactions, and it has limited value in predicting risks associated with nontraditional preparations or with use under nontraditional circumstances (e.g., in combination with conventional drugs). Finally, and most controversially, the new Euro- Ginkgo Leaf, St. John’s Wort, and Stinging Nettle. pean directive results in a double standard for dif- These are among the herbs that were most highly re- ferent types of medicines, since the scientific proof imbursed by German health insurance in 2003 (for of efficacy required for conventional medicines is $87 million, $40 million, and $14 million, respectively). no longer needed for herbal medicines, as long as Ginkgo and St. John’s Wort are among the few excep- they are used for minor health problems and la- tions to a new German rule that herbal “pharmacy-only” beled as “traditionally used.” It could be argued that medicines no longer qualify for reimbursement. conventional efficacy requirements are not an end in themselves but a means of safeguarding public health and that the relaxation of these requirements of herbal products in the United States. Besides ex- for self-care with herbs reflects a permissible prin- pected public health benefits, however, the new Eu- ciple of proportionality. But the new legislation ropean legislation also carries important caveats. relaxes an iron grip on claims of efficacy for med- First, the focus on industrial preparations ne- icines and may provide a further disincentive for cessitates additional measures to help ensure the manufacturers to submit their herbal products to quality and safety of crude herbal ingredients used randomized, controlled trials. Perhaps this disin- in locally compounded medicines. Additional mea- centive may be countered best by allowing herbal sures are also needed to ensure that nonmedical medicines to qualify for public reimbursement only practitioners who prescribe herbal treatments re- after their efficacy has been satisfactorily proved. ceive adequate training and continuous education In this context, it is noteworthy that as of April 2004, to reach and maintain a high standard of practice. most herbal medicines no longer qualify for reim- The United Kingdom Department of Health is al- bursement in Germany, because they are “pharma- ready making progress in this domain, having pro- cy-only” medicines, unless they appear on a list of posed statutory self-regulation of herbal practi- exceptions. Currently, only four herbs are included tioners. on this list.

1178 n engl j med 352;12 www.nejm.org march 24, 2005

The new england journal of medicine

established in 1812 march 24, 2005 vol. 352 no. 12

Addition of Clopidogrel to Aspirin and Fibrinolytic Therapy for Myocardial Infarction with ST-Segment Elevation

Marc S. Sabatine, M.D., M.P.H., Christopher P. Cannon, M.D., C. Michael Gibson, M.D., Jose L. López-Sendón, M.D., Gilles Montalescot, M.D., Pierre Theroux, M.D., Marc J. Claeys, M.D., Ph.D., Frank Cools, M.D., Karen A. Hill, B.A., Allan M. Skene, Ph.D., Carolyn H. McCabe, B.S., and Eugene Braunwald, M.D., for the CLARITY–TIMI 28 Investigators*

abstract background A substantial proportion of patients receiving fibrinolytic therapy for myocardial in- From the TIMI Study Group, Cardiovas- farction with ST-segment elevation have inadequate reperfusion or reocclusion of the cular Division, Department of Medicine, Brigham and Women’s Hospital and Har- infarct-related artery, leading to an increased risk of complications and death. vard Medical School, Boston (M.S.S., C.P.C., C.M.G., C.H.M., E.B.); Hospital Universi- methods tario Gregorio Marañon, Madrid (J.L.L.-S.); We enrolled 3491 patients, 18 to 75 years of age, who presented within 12 hours after Institut de Cardiologie, Hôpital Pitié-Salpê- the onset of an ST-elevation myocardial infarction and randomly assigned them to re- trière, Paris (G.M.); the Montreal Heart Institute, Montreal (P.T.); the Department ceive clopidogrel (300-mg loading dose, followed by 75 mg once daily) or placebo. Pa- of Cardiology, University Hospital Antwerp, tients received a fibrinolytic agent, aspirin, and when appropriate, heparin (dispensed Edegem, Belgium (M.J.C.); Academisch according to body weight) and were scheduled to undergo angiography 48 to 192 hours Ziekenhuis Klina, Brasschaat, Belgium (F.C.); and Nottingham Clinical Research Group, after the start of study medication. The primary efficacy end point was a composite of Nottingham, United Kingdom (K.A.H., an occluded infarct-related artery (defined by a Thrombolysis in Myocardial Infarction A.M.S.). Address reprint requests to Dr. flow grade of 0 or 1) on angiography or death or recurrent myocardial infarction before Sabatine at the TIMI Study Group, Cardio- vascular Division, Brigham and Women’s angiography. Hospital, 75 Francis St., Boston, MA 02115, or at [email protected]. results The rates of the primary efficacy end point were 21.7 percent in the placebo group and *The participants in the Clopidogrel as Ad- junctive Reperfusion Therapy (CLARITY)– 15.0 percent in the clopidogrel group, representing an absolute reduction of 6.7 per- Thrombolysis in Myocardial Infarction centage points in the rate and a 36 percent reduction in the odds of the end point with (TIMI) 28 study are listed in the Appendix. clopidogrel therapy (95 percent confidence interval, 24 to 47 percent; P<0.001). By 30 This article was published at www.nejm. days, clopidogrel therapy reduced the odds of the composite end point of death from org on March 9, 2005. cardiovascular causes, recurrent myocardial infarction, or recurrent ischemia leading to the need for urgent revascularization by 20 percent (from 14.1 to 11.6 percent, N Engl J Med 2005;352:1179-89. P=0.03). The rates of major bleeding and intracranial hemorrhage were similar in the Copyright © 2005 Massachusetts Medical Society. two groups. conclusions In patients 75 years of age or younger who have myocardial infarction with ST-segment elevation and who receive aspirin and a standard fibrinolytic regimen, the addition of clopidogrel improves the patency rate of the infarct-related artery and reduces ischemic complications.

n engl j med 352;12 www.nejm.org march 24, 2005 1179

The new england journal of medicine

he benefit of fibrinolytic thera- nolytic agent, an anticoagulant (if a fibrin-specif- t py for myocardial infarction with ST-seg- ic lytic agent was prescribed), and aspirin. ment elevation is limited by inadequate Exclusion criteria were as follows: treatment reperfusion or reocclusion of the infarct-related ar- with clopidogrel within seven days before enroll- tery in a sizable proportion of patients. Initial re- ment or planned treatment with clopidogrel or a perfusion fails to occur in approximately 20 percent glycoprotein IIb/IIIa inhibitor before angiography; of patients1-3 and is associated with a doubling of contraindications to fibrinolytic therapy (includ- mortality rates.4 The artery becomes reoccluded in ing documented stroke, intracranial hemorrhage, an additional 5 to 8 percent of patients during their and intracranial neoplasm); a plan to perform an- index hospitalization, and this event is associated giography within 48 hours in the absence of a new with an increase in mortality rates by a factor of near- clinical indication; cardiogenic shock; prior coro- ly three.5 nary-artery bypass grafting; and a weight of 67 kg Platelet activation and aggregation play a key or less and receipt of more than a 4000-U bolus of role in initiating and propagating coronary-artery unfractionated heparin, a weight of more than 67 kg thrombosis. In the Second International Study of and receipt of more than a 5000-U bolus of unfrac- Infarct Survival, conducted in patients with acute tionated heparin, or receipt of more than a standard myocardial infarction, aspirin reduced the odds of dose of low-molecular-weight heparin. death from vascular causes by 23 percent and the The protocol was approved by the institutional odds of reinfarction by 46 percent.6 Aspirin has also review board at each participating center. Written been shown to reduce the rate of angiographic re- informed consent was obtained from all patients. occlusion by 22 percent, as compared with placebo.7 Clopidogrel is an adenosine diphosphate– study protocol receptor antagonist, a class of oral antiplatelet agents Patients were randomly assigned in a 1:1 ratio to that block the P2Y12 component of the adenosine receive either clopidogrel (Plavix, Sanofi-Aventis diphosphate receptor and thus inhibit the activa- and Bristol-Myers Squibb; a 300-mg loading dose tion and aggregation of platelets.8 Clopidogrel has followed by 75 mg once daily) or placebo in a dou- been shown to prevent death and ischemic com- ble-blind fashion by means of a central, computer- plications in patients with symptomatic atheroscle- ized system of randomization. Patients were to re- rotic disease, patients who have undergone percu- ceive study medication daily up to and including the taneous coronary intervention, and patients with day of coronary angiography. For patients who did unstable angina or myocardial infarction without not undergo angiography, study drug was to be ad- ST-segment elevation.9-11 A major remaining ques- ministered up to and including day 8 or hospital tion is whether the addition of clopidogrel is bene- discharge, whichever came first. ficial in patients who have myocardial infarction All patients were to be treated with a fibrinolyt- with ST-segment elevation and who are receiving ic agent (selected by the treating physician), aspi- a standard fibrinolytic regimen, including aspirin. rin (recommended dose, 150 to 325 mg on the first day and 75 to 162 mg daily thereafter), and for those methods receiving a fibrin-specific lytic agent, heparin for 48 hours. The recommended dose of unfractionat- patient population ed heparin was a bolus of 60 U per kilogram of body Between February 10, 2003, and October 31, 2004, weight given intravenously (maximum, 4000 U), 3491 patients were enrolled at 319 sites in 23 coun- followed by infusion at a rate of 12 U per kilogram tries (listed in the Appendix). As described previous- per hour (maximum, 1000 U per hour).13 The use of ly,12 men and women 18 to 75 years of age were el- low-molecular-weight heparin instead of unfrac- igible if they had begun to have ischemic discomfort tionated heparin and the use of heparin in patients at rest within 12 hours before randomization and it receiving streptokinase were at the discretion of the had lasted more than 20 minutes; if they had ST-seg- treating physician. Unless clinically indicated, the ment elevation of at least 0.1 mV in at least two con- use of glycoprotein IIb/IIIa inhibitors was permit- tiguous limb leads, ST-segment elevation of at least ted only after coronary angiography. 0.2 mV in at least two contiguous precordial leads, Coronary angiography was to be performed or left bundle-branch block that was not known to according to the protocol during the index hospi- be old; and if they were scheduled to receive a fibri- talization, 48 to 192 hours after the start of study

1180 n engl j med 352;12 www.nejm.org march 24, 2005

clopidogrel and fibrinolysis in myocardial infarction medication, to assess for late patency of the in- Table 1. Baseline Characteristics of the Patients.* farct-related artery. Angiography was permitted before 48 hours had elapsed only if clinically indi- Clopidogrel Placebo cated.12,13 For patients who underwent coronary Characteristic (N=1752) (N=1739) stenting, it was recommended that open-label clo- Age — yr 57.7±10.3 57.2±10.3 pidogrel be administered after angiography, with Male sex — no. (%) 1400 (79.9) 1403 (80.7) the use of a loading dose of at least 300 mg, followed White race — no. (%)† 1569 (89.6) 1556 (89.5) by a daily dose of 75 mg. Patients were to undergo Weight — kg 80.1±14.7 80.1±14.6 electrocardiography at baseline and 90 and 180 min- Hypertension — no. (%) 750 (42.8) 764 (43.9) utes after the administration of the loading dose of Hyperlipidemia — no. (%) 564 (32.2) 574 (33.0) study drug. Current smoker — no. (%) 887 (50.7) 865 (49.9) Patients were followed for clinical end points Diabetes mellitus — no. (%) 289 (16.5) 286 (16.4) and adverse events during their index hospitaliza- Prior myocardial infarction — no. (%) 159 (9.1) 159 (9.1) Prior percutaneous coronary intervention — 84 (4.8) 85 (4.9) tion. Telephone follow-up was performed at 30 days no. (%) to identify clinical end points or adverse events, Anterior myocardial infarction — no. (%) 722 (41.2) 697 (40.1) which were verified by means of medical records. Fibrinolytic agent — no. (%)‡ Vital status was ascertained in 3487 of the 3491 pa- Tenecteplase 838 (47.8) 822 (47.3) tients (99.9 percent). Reteplase 209 (11.9) 214 (12.3) Alteplase 159 (9.1) 155 (8.9) end points Streptokinase 542 (30.9) 543 (31.2) The primary efficacy end point was the composite None 4 (0.2) 6 (0.3) of an occluded infarct-related artery (defined by a Initial aspirin — no. (%) 1726 (98.5) 1715 (98.6) Thrombolysis in Myocardial Infarction [TIMI] flow Initial heparin — no. (%)§ grade of 0 or 1) on angiography, death from any Unfractionated heparin 808 (46.1) 792 (45.5) cause before angiography could be performed, or Low-molecular-weight heparin 528 (30.1) 506 (29.1) recurrent myocardial infarction before angiogra- Both 85 (4.9) 90 (5.2) phy — the last two of which served as surrogates Neither 331 (18.9) 351 (20.2) for failed reperfusion or reocclusion of the infarct- Time from onset of symptoms to start of related artery. For patients who did not undergo an- fibrinolytic therapy — hr giography, the primary end point was death or re- Median 2.7 2.6 current myocardial infarction by day 8 or hospital Interquartile range 1.8–4.2 1.7–4.0 discharge, whichever came first. The TIMI flow Angiography — no. (%) 1645 (93.9) 1638 (94.2) grade1 in the infarct-related artery was determined Time to angiography — hr in a blinded fashion by the TIMI Angiographic Core Median 84 84 Laboratory. The definitions of recurrent myocardial Interquartile range 55–123 50–124 infarction and other efficacy end points have been Cardiac medications during index hospitalization — no. (%) described previously.12 Beta-blockers 1554 (88.7) 1559 (89.6) The primary safety end point was the rate of ma- 14 Statins 1408 (80.4) 1410 (81.1) jor bleeding (according to TIMI criteria ) by the ACE inhibitors or angiotensin-receptor 1273 (72.7) 1254 (72.1) end of the calendar day after angiography or, if an- blockers¶ giography was not performed, by day 8 or hospital Open-label clopidogrel after completion 954 (54.5) 967 (55.6) discharge, whichever came first. Other safety end of study drug points included the rates of intracranial hemor- Ticlopidine after completion of study drug 62 (3.5) 50 (2.9) rhage and minor bleeding (according to TIMI cri- * Plus–minus values are means ±SD. None of the differences between groups teria). All ischemic and any clinically significant were statistically significant. Data on weight were missing for 61 patients (31 bleeding events were adjudicated in a blinded fash- in the clopidogrel group and 30 in the placebo group), and data on smoking ion by members of an independent clinical-events status were missing in 7 patients (2 and 5, respectively). † Race was self-reported. committee. ‡ One patient in the placebo group was treated with both reteplase and streptokinase. statistical analysis § Initial heparin includes any heparin that was given immediately before or during the first two hours after randomization. We estimated that the enrollment of 3500 patients ¶ACE denotes angiotensin-converting enzyme. would provide the study with a statistical power of

n engl j med 352;12 www.nejm.org march 24, 2005 1181

The new england journal of medicine

Table 2. Efficacy Outcomes.*

Clopidogrel Placebo Odds Ratio Outcome (N=1752) (N=1739) (95% CI) P Value

Primary efficacy end point — no. 262 (15.0) 377 (21.7) 0.64 (0.53 to 0.76) <0.001 of patients (%)† TIMI flow grade 0 or 1 192 (11.7) 301 (18.4) 0.59 (0.48 to 0.72) <0.001 Death 45 (2.6) 38 (2.2) 1.17 (0.75 to 1.82) 0.49 Recurrent myocardial infarction 44 (2.5) 62 (3.6) 0.70 (0.47 to 1.04) 0.08 Other angiographic measurement — no. of patients (%) TIMI flow grade 3 1112 (67.8) 993 (60.8) 1.36 (1.18 to 1.57) <0.001 TIMI myocardial-perfusion grade 3 885 (55.8) 817 (51.2) 1.21 (1.05 to 1.40) 0.008 Intracoronary thrombus 697 (43.0) 822 (50.8) 0.73 (0.64 to 0.84) <0.001 Mean stenosis — % 68.4 70.8 ¡2.3 (¡3.8 to ¡0.9)‡ 0.001 Mean minimal luminal diameter — mm 0.82 0.75 0.07 (0.03 to 0.11)‡ 0.001

* Data on the Thrombolysis in Myocardial Infarction (TIMI) flow grade were available for 1640 patients in the clopidogrel group and 1634 patients in the placebo group; data on TIMI myocardial-perfusion grade were available for 1585 and 1596 patients, respectively; data on thrombus were available for 1622 and 1619 patients, respectively; and data on stenosis and the minimal luminal diameter were available for 1560 and 1559 patients, respectively. CI denotes confidence interval. † The primary efficacy end point was ascertained through the start of coronary angiography (at a median of 3.5 days) or, among patients who did not undergo angiography, at hospital discharge or day 8, whichever came first. ‡ This value is the mean difference between groups, rather than the odds ratio.

95 percent to detect a relative reduction in the rate trial by the TIMI Study Group, which designed the of the primary end point of 24 percent (from 19.0 trial and had free and complete access to the data. to 14.4 percent) with the use of a two-sided test at Data were coordinated by the Nottingham Clinical the 5 percent level. All efficacy analyses were based Research Group. Members of the TIMI Study Group on the intention-to-treat principle. The prospec- and of the Nottingham group carried out the pre- tively defined analyses of the primary and second- specified analyses, and the sponsors independent- ary end points involved a logistic-regression model ly validated them. that included terms for the treatment group, the type of fibrinolytic agent used, the type of heparin results used, and the location of the infarct. For continuous variables, differences between the treatment groups A total of 3491 patients underwent randomization, were assessed by analysis of variance. Safety analy- and the two groups were well matched with regard ses were performed according to the treatment ac- to baseline characteristics (Table 1). Their average tually received by each patient. The rates of the safety age was 57 years, 80.3 percent were men, 50.3 per- end points and stroke were compared with the use cent were current smokers, and 9.1 percent had a of Fisher’s exact test. history of myocardial infarction. A total of 99.7 per- An independent data and safety monitoring cent of the patients received a fibrinolytic agent, of board monitored the incidence of the safety end whom 68.8 percent received a fibrin-specific agent. points after the enrollment of every 500 patients, The median time from the onset of symptoms to with one formal interim analysis after 50 percent of the administration of a fibrinolytic agent was 2.7 the patients had been enrolled. No stopping rules hours. A total of 98.6 percent of the patients received were specified; therefore, the overall significance aspirin. For initial anticoagulation, 45.8 percent levels were not adjusted as a result of the formal in- received unfractionated heparin, 29.6 percent low- terim analysis. molecular-weight heparin, 5.0 percent both, and The study was an investigator-initiated clinical 19.5 percent neither.

1182 n engl j med 352;12 www.nejm.org march 24, 2005

clopidogrel and fibrinolysis in myocardial infarction

Reduction Patients with End Point No. of in Odds Clopidogrel Placebo Characteristic Patients Odds Ratio % % Overall 3491 36 15.0 21.7 Age <65 yr 2466 42 13.2 21.0 ≥65 yr 1015 22 19.0 23.1 Sex Male 2796 35 14.5 20.8 Female 685 38 16.9 24.7 Infarct location Anterior 1416 33 15.0 20.7 Nonanterior 2065 38 15.0 22.2 Types of fibrinolytic agent Fibrin-specific 2397 31 14.7 20.1 Non–fibrin-specific 1084 44 15.7 24.9 Predominant type of heparin Low-molecular-weight 1429 31 11.4 15.7 Unfractionated 1431 42 17.8 27.1 None 621 26 17.1 21.9

0.4 0.6 0.8 1.0 1.2 1.6

Clopidogrel Better Placebo Better

Figure 1. Rates of and Odds Ratios for the Primary Efficacy End Point Overall and in Various Subgroups. The primary efficacy end point was a composite of a Thrombolysis in Myocardial Infarction (TIMI) flow grade of 0 or 1 on angiography or death or recurrent myocardial infarction before angiography. For the logistic-regression models to con- verge correctly, the 10 patients who did not receive a fibrinolytic agent were excluded from the subgroup analyses. The analysis of subgroups according to the type of heparin used was based on the predominant heparin used from randomization to the time of angiography. All P values for interactions were not significant. The overall treatment effect is represented by the diamond, the left and right borders of which indicate the 95 percent confidence interval. The dotted line represents the point estimate of the overall treatment effect. For subgroups, the size of each box is proportional to the number of patients in the individual analyses. The horizontal lines represent the 95 percent confidence intervals.

In all, 98.9 percent of the patients received study ascertainment of the primary end point, open-label medication. The median time from the adminis- clopidogrel or ticlopidine was given to 56.7 percent tration of a fibrinolytic agent to the administration of the patients in the clopidogrel group and 57.4 of study medication was 10 minutes (interquartile percent of those in the placebo group. range, 5 to 25). Patients received a median of four doses of study medication. The rate of use of oth- efficacy end points er cardiac medications was high and similar in the The rates of the prespecified primary efficacy end two groups (Table 1). Angiography was performed point were 21.7 percent in the placebo group and in 93.9 percent of the patients in the clopidogrel 15.0 percent in the clopidogrel group, representing group and 94.2 percent of those in the placebo an absolute reduction of 6.7 percentage points in group, at a median of 84 hours after randomiza- the rate and a 36 percent reduction in the odds of tion in each group. Percutaneous coronary inter- the end point in favor of treatment with clopidogrel vention and coronary-artery bypass grafting were (95 percent confidence interval, 24 to 47 percent; performed in 57.2 percent and 5.9 percent, respec- P<0.001). Among the individual components of tively, of the patients in the clopidogrel group and the primary end point (Table 2), clopidogrel had in 56.6 percent and 6.0 percent, respectively, of the greatest effect on the rate of an occluded in- those in the placebo group. After angiography and farct-related artery (reducing it from 18.4 percent

n engl j med 352;12 www.nejm.org march 24, 2005 1183

The new england journal of medicine

the degree of resolution was 59 percent (median, 15 73 percent) with clopidogrel, as compared with 61 Placebo percent (median, 72 percent) with placebo (P=0.22). As compared with placebo, clopidogrel therapy was Clopidogrel 10 associated with a 21 percent reduction in the odds of the need for early angiography (i.e., within 48 hours after randomization) for clinical indications (15.4 percent vs. 18.6 percent, P=0.01) and a 21 End Point (%) 5 percent reduction in the odds of the need for revascularization on an urgent basis during the index P=0.03 hospitalization, as assessed by local investigators 0 (19.5 percent vs. 23.3 percent, P=0.005). Among 0 5 10 15 20 25 30 the patients who underwent percutaneous coro- Days nary intervention, the rates of use of glycoprotein No. at Risk IIb/IIIa were 29.3 percent in the clopidogrel group Placebo 1739 1529 1504 1494 Clopidogrel 1752 1569 1555 1550 and 33.0 percent in the placebo group (P=0.07). By the time of the ascertainment of the primary end Figure 2. Cumulative Incidence of the End Point of Death from Cardiovascular point (median, 3.5 days), the rate of the composite Causes, Recurrent Myocardial Infarction, or Recurrent Ischemia Leading end point of death, recurrent myocardial infarction, to the Need for Urgent Revascularization. or recurrent myocardial ischemia was 8.3 percent The odds ratio for this end point was significantly lower in the clopidogrel in the clopidogrel group and 9.3 percent in the group than in the placebo group at 30 days (11.6 percent vs. 14.1 percent; placebo group (reduction in the odds, 12 percent; odds ratio, 0.80 [95 percent confidence interval, 0.65 to 0.97]; P=0.03). P=0.27). By 30 days, clopidogrel therapy had reduced the odds of the composite end point of death from car- to 11.7 percent; 41 percent reduction in the odds; diovascular causes, recurrent myocardial infarction, P<0.001) and the rate of recurrent myocardial in- or recurrent ischemia leading to the need for urgent farction (reducing it from 3.6 to 2.5 percent; 30 per- revascularization by 20 percent (from 14.1 to 11.6 cent reduction in the odds; P= 0.08), but it had no percent, P=0.03) (Fig. 2). In terms of the individual significant effect on the rate of death from any cause end points (Fig. 3), there were the following: no dif- (2.2 percent in the placebo group vs. 2.6 percent in ference in the rate of death from cardiovascular the clopidogrel group, P=0.49). The beneficial ef- causes; a statistically significant, 31 percent reduc- fect of clopidogrel on the incidence of the primary tion in the odds of recurrent myocardial infarction end point was consistent across the prespecified in the clopidogrel group as compared with the pla- subgroups, as defined on the basis of age, sex, the cebo group (P=0.02); a 24 percent reduction in type of fibrinolytic agent used, the type of heparin the odds of recurrent myocardial ischemia leading used, and the location of the infarct (Fig. 1). to the need for urgent revascularization (P=0.11); Clopidogrel improved all angiographic measure- and a 46 percent reduction in the odds of stroke ments (Table 2). Specifically, as compared with pla- (P=0.052). cebo, treatment with clopidogrel increased the odds of achieving optimal epicardial flow (defined by a safety end points TIMI flow grade of 3) by 36 percent (P<0.001) and The rates of the primary safety end point, TIMI- the odds of achieving optimal myocardial reperfu- defined major bleeding through the day after angi- sion (defined by a TIMI myocardial-perfusion grade ography, were 1.3 percent in the clopidogrel group of 3) by 21 percent (P=0.008) and reduced the odds and 1.1 percent in the placebo group (P=0.64) (Ta- of intracoronary thrombus by 27 percent (P<0.001). ble 3). There were no significant increases in the As compared with placebo, treatment with clopid- risk of major bleeding with clopidogrel in any of ogrel also resulted in less severe stenosis (P=0.001) the subgroups prespecified according to the type and a larger minimal luminal diameter of the in- of fibrinolytic agent used, the type of heparin used, farct-related artery (P=0.001). Clopidogrel therapy age, sex, or weight (data not shown). The rates of had no significant effect on the mean degree of res- TIMI-defined major bleeding or the need for the olution of ST-segment elevation by 180 minutes: transfusion of at least 2 units of blood were 1.8

1184 n engl j med 352;12 www.nejm.org march 24, 2005

clopidogrel and fibrinolysis in myocardial infarction

Reduction Patients with End Point in Odds Clopidogrel Placebo End Point Odds Ratio % % Death from CV causes 3 4.4 4.5 Recurrent MI 31 4.1 5.9 Recurrent ischemia leading 24 3.5 4.5 to urgent revascularization Stroke 46 0.9 1.7 Death from CV causes or recurrent MI 17 8.4 9.9 Death from CV causes, recurrent MI, 18 9.1 10.9 or stroke Death from CV causes, recurrent MI, or recurrent ischemia leading 20 11.6 14.1 to urgent revascularization Death from CV causes, recurrent MI, stroke, or recurrent ischemia leading 21 12.3 15.0 to urgent revascularization 0.4 0.6 0.8 1.0 1.2 1.6

Clopidogrel Better Placebo Better

Figure 3. Odds Ratios for Individual and Composite Clinical End Points through 30 Days in the Clopidogrel Group as Compared with the Placebo Group. The horizontal lines represent the 95 percent confidence intervals. CV denotes cardiovascular, and MI myocardial infarction. percent in the clopidogrel group and 1.3 percent ment with a loading dose of 300 mg of clopidogrel in the placebo group (P=0.28), and the rates of followed by a daily dose of 75 mg resulted in a 36 TIMI-defined minor bleeding through the day af- percent reduction in the odds of an occluded infarct- ter angiography were 1.0 percent and 0.5 percent, related artery or death or recurrent myocardial in- respectively (P=0.17) (Table 3). The rates of intra- farction by the time of angiography. The benefit was cranial hemorrhage were 0.5 percent in the clopido- consistent across a broad range of subgroups, in- grel group and 0.7 percent in the placebo group cluding those categorized according to the type of (P=0.38). At 30 days, there were no significant dif- fibrinolytic agent used and the type of heparin used. ferences in the rates of major or minor bleeding By 30 days, clopidogrel therapy led to a significant, between the two groups (Table 3). Among the 136 20 percent reduction (from 14.1 to 11.6 percent) in patients who underwent coronary-artery bypass the odds of the composite end point of death from grafting during the index hospitalization, treat- cardiovascular causes, recurrent myocardial infarc- ment with clopidogrel was not associated with a tion, or recurrent ischemia leading to the need for significant increase in the rate of major bleeding urgent revascularization. Treatment with clopido- through 30 days of follow-up (7.5 percent in the grel was not associated with an increased rate of clopidogrel group, as compared with 7.2 percent major bleeding or intracranial hemorrhage. in the placebo group; P=1.00), even among those Arterial thrombi that are rich in platelets are who underwent coronary-artery bypass grafting relatively resistant to fibrinolysis and prone to in- within 5 days after the discontinuation of study duce reocclusion after initial reperfusion.15 Despite medication (9.1 percent and 7.9 percent, respec- the inhibition of cyclooxygenase by aspirin, plate- tively; P=1.00). let activation can still occur through thromboxane A2–independent pathways, leading to the aggrega- 16 discussion tion of platelets and the formation of thrombin. Clopidogrel is a potent antiplatelet agent that has Our study demonstrates the benefit of adding clo- a synergistic antithrombotic effect when combined pidogrel to aspirin and fibrinolytic therapy for my- with aspirin.17 Clopidogrel has been shown to ben- ocardial infarction with ST-segment elevation. Treat- efit patients with documented atherosclerosis (re-

n engl j med 352;12 www.nejm.org march 24, 2005 1185

The new england journal of medicine

cacy and safety of enoxaparin are currently being Table 3. Safety Outcomes.* tested in a large clinical trial.25 The benefit we ob- Clopidogrel Placebo served with clopidogrel was equally apparent in pa- Outcome (N=1733) (N=1719) P Value tients treated with unfractionated heparin and patients who received low-molecular-weight heparin. no. of patients (%) The trial was not powered to detect a survival ben- Through the day after angiography efit, and none was seen. However, we did observe Major bleeding 23 (1.3) 19 (1.1) 0.64 consistent effects of clopidogrel in improving mul- Minor bleeding 17 (1.0) 9 (0.5) 0.17 tiple angiographic outcomes and reducing ische- Major or minor bleeding 40 (2.3) 28 (1.6) 0.18 mic events, all of which have been shown to be associated with improved long-term survival after Intracranial hemorrhage 8 (0.5) 12 (0.7) 0.38 myocardial infarction.2,4,5,26-28 The use of proto- At 30 days col-driven angiography and its attendant high rate Major bleeding 33 (1.9) 30 (1.7) 0.80 of revascularization in our trial may have attenuat- Minor bleeding 27 (1.6) 16 (0.9) 0.12 ed the translation of the angiographic benefit into Major or minor bleeding 59 (3.4) 46 (2.7) 0.24 an immediate reduction in mortality. Whether a mortality benefit would emerge in the setting of fi- * Safety end points were assessed in the treated population. The incidence of brinolysis without mandatory angiography is the bleeding was ascertained through the calendar day after angiography and subject of a separate study specifically powered to at 30 days. For patients who did not undergo angiography, the incidence of assess mortality.29 bleeding was ascertained through day 8 or hospital discharge, whichever came first. The prespecified primary bleeding end point was major bleeding, We excluded patients who presented more than according to Thrombolysis in Myocardial Infarction (TIMI) criteria,14 through 12 hours after the onset of symptoms, those old- the calendar day after angiography. TIMI-defined major bleeding includes er than 75 years of age, and those with a history of intracranial hemorrhage. coronary-artery bypass grafting. The efficacy and safety of adding treatment with clopidogrel to aspirin and fibrinolytic therapy in these groups remain cent myocardial infarction, recent stroke, or estab- to be established. There was a low rate of bleeding lished peripheral arterial disease), patients who have complications in our trial, most likely because of undergone percutaneous coronary intervention, our emphasis on adherence to weight-based guide- and patients with unstable angina or myocardial lines for heparin dosing.13,30 Still, the administra- infarction that is not associated with ST-segment tion of a fibrinolytic agent in conjunction with hep- elevation.9-11 We now extend those findings to pa- arin and two antiplatelet agents must be performed tients with the most severe manifestation of ath- with caution. As with any clinical trial, application erosclerotic coronary artery disease: myocardial in- of the results to a different population outside the farction that is associated with ST-segment setting of the trial should be done carefully. elevation. In conclusion, we found that, in patients 75 years Since the use of aspirin, heparin, and fibrin-spe- of age or younger who have myocardial infarction cific lytic therapy became established for myocar- with ST-segment elevation and who receive fibri- dial infarction with ST-segment elevation in the nolytic therapy, aspirin, and (when appropriate) late 1980s and early 1990s,6,18,19 there have been weight-based heparin, clopidogrel offers an effec- many attempts to improve on this regimen, with lit- tive, simple, inexpensive, and safe means by which tle success. Newer fibrinolytic agents are equivalent to improve the rate of patency of the infarct-related but not superior to older fibrin-specific agents.20,21 artery and to reduce the rate of ischemic compli- Aggressive antiplatelet therapy with glycoprotein cations. IIb/IIIa inhibitors improves the rate of patency and Supported in part by the pharmaceutical partnership of Sanofi- reduces the risk of reinfarction, but at the cost of Aventis and Bristol-Myers Squibb. Dr. Sabatine is the recipient of a grant (R01 HL072879) from the National Heart, Lung, and Blood doubling the rates of major bleeding and, in pa- Institute. tients older than 75 years of age, intracranial hem- 22,23 Dr. Sabatine reports having received research grant support from orrhage. Low-molecular-weight heparin has Bristol-Myers Squibb; having received lectures fees from Bristol- emerged as an attractive alternative to unfraction- Myers Squibb and Sanofi-Aventis; and having served on paid advisory boards for Bristol-Myers Squibb, Sanofi-Aventis, and AstraZeneca. ated heparin in patients who have myocardial in- 24 Dr. Cannon reports having received research grant support from As- farction with ST-segment elevation, and the effi- traZeneca, Bristol-Myers Squibb, Merck, and Sanofi-Aventis and

1186 n engl j med 352;12 www.nejm.org march 24, 2005

clopidogrel and fibrinolysis in myocardial infarction having received lecture fees from and having served on paid advisory served on paid advisory boards for Sanofi-Aventis and Bristol-Myers boards for AstraZeneca, Bristol-Myers Squibb, GlaxoSmithKline, Squibb. Dr. Theroux reports having received lectures fees from, own- Guilford Pharmaceuticals, Merck, Millennium, Pfizer, Sanofi-Aven- ing equity or stock options in, having served on paid advisory boards tis, Schering-Plough, and Vertex. Dr. Gibson reports having received for, and having received lecture fees from Sanofi-Aventis, as well as research grant support from Bristol-Myers Squibb and Millennium; having received lecture fees from Bristol-Myers Squibb. Dr. Cools having received lecture fees from Bristol-Myers Squibb, Genentech, reports having received lectures fees from Bristol-Myers Squibb and and Millennium; and having served on paid advisory boards for Sanofi-Aventis. Ms. McCabe reports having received research grant Genentech and Millennium. Dr. López-Sendón reports having re- support from Bristol-Myers Squibb, Sanofi-Aventis, AstraZeneca, and ceived research grant support from Sanofi-Aventis; having received Millennium. Dr. Braunwald reports having received research grant lecture fees from Guidant and Pfizer; and having served on paid ad- support from Bristol-Myers Squibb, Sanofi-Aventis, Eli Lilly, and visory boards for Sanofi-Aventis, GlaxoSmithKline, and Pfizer. Dr. AstraZeneca and having received lectures fees from Bristol-Myers Montalescot reports having received lecture fees from and having Squibb.

appendix The participants in the CLARITY–TIMI 28 study were as follows: Operations Committee — E. Braunwald (chair), C. Cannon (principal investigator), M. Sabatine (co–principal investigator), C. McCabe, A. McCagg, B. Job, C. Gaudin, I. Thizon-de Gaulle, M. Blumenthal, R. Saini, I. Delaet, L. Townes, D. Anhalt, K. van Holder, A. Skene, K. Hill; Steering Committee — E. Braunwald, C. Cannon, M. Sabatine, C. McCabe, B. Job, C. Gaudin, I. Thizon-de Gaulle, M. Blumenthal, R. Saini, I. Delaet, L. Townes, A. Skene, D. Ardissino, P. Aylward, M. Bertrand, C. Bode, A. Budaj, M. Claeys, M. Dellborg, R. Ferreira, A. Gershlick, K. Huber, M. Keltai, N. Kleiman, B. Lewis, J. Lopez-Sendon, J. Marx, G. Mon- talescot, J. Nicolau, Z. Ongen, E. Paolasso, J. Leiva Pons, M. Ruda, P. Theroux, F. Van de Werf, F. Verheugt, R. Wilcox, U. Zeymer; TIMI Study Group — E. Braunwald, C. Cannon, M. Sabatine, M. Gibson, C. McCabe, A. McCagg; Sponsors: Sanofi-Aventis (Paris) — B. Job, C. Gaudin, I. Thizon-de Gaulle; Bristol-Myers Squibb (Princeton, N.J.) — M. Blumenthal, R. Saini, J. Froehlich, I. Delaet, L. Townes, D. Anhalt, K. van Holder, A. Pieters; Data Coordinating Center (Nottingham Clinical Research Group, Nottingham, United Kingdom) — A. Skene, K. Hill, A. Charlesworth, M. Goulder, S. Stead; Clinical Events Committee — S. Wiviott (chair); Physician Reviewers — J. Aroesty, C. Berger, L. Garcia, D. Greer, D. Leeman, H. Lyle, G. Philippides, L. Schwamm; TIMI Angiographic Core Laboratory (Brigham and Women’s Hospital, Boston) — M. Gibson (director), S. Marble, J. Aroesty, J. Buros, L. Ciaglo, A. Kirtane, A. Shui, A. Wilson; TIMI Static Electrocardiographic (ECG) Core Laboratory (Brigham and Women’s Hospital, Boston) — B. Scirica (director), D. Morrow, R. Giugliano, S. Wiviott; Biomarker Core Laboratory (Brigham and Women’s Hospital, Boston) — D. Morrow, P. Ridker, N. Rifai; Genetics Core Laboratory (Harvard–Partners Center for Genetics and Geno- mics, Cambridge, Mass.) — M. Sabatine, R. Kucherlapati, D. Kwiatkowski; Continuous ECG Core Laboratory (Brigham and Women’s Hospital, Boston) — P. Stone (director), G. Maccallum, M. Lucier, A. Garriga; J. Anderson (chair), K. Fox, S. Kelsey; Clinical Centers (grouped by countries, which are listed in order of number of patients enrolled): Spain: A. Batalla; E. Lopez de Sa y Areses; M. Fiol, A. Bethencourt; I. Lozano; J. Figueras; J. Guiterrez Cortés, A. Garcia Jimenez; A. Rivera Fernandez; I. Ferreira Montero, J. Casasnovas; F. Ortigosa Aso; R. Carbonell de Blas; J. Audi- cana; I. Echanove Errazti; J. Sala Montero, I. Rohlfs; C. Piñero; M. Martinez Rodriguez; L. Velasco Alvarez; A. Castro Belras, S. Barros; J. Diaz Fernández, A. Tobaruela; France: A. Bonneau, L. Soulat; Y. Lambert, J. Caussanel, B. Livareck, C. Ramaut, J. Ricome; F. Lapostolle, A. Beruben; F. Thieleux; J. Martelli, Etori; M. Martelet, J. Mouallem; F. Ahmed; D. Galley, S. Alhabaj, Strateman; D. Doucos; E. Bearez, S. Spe- tebroodt; C. Gully; J. Dujardin, B. Averland; B. Emmonot; D. Pollet, P. Vallet; R. Mossaz, A. Marquand; A. Kermarrec, C. Le Lay, R. Douillet; L. Olliver, O. Matas; Canada: S. Kassam, F. Halperin, P. Parsons, B. Bovak, B. Hart; B. Sussex, S. Newman; G. Gosselin, M. David; K. Sidhar, D. Wiseman; C. Lefkowitz, M. Thornley; R. Bhargava, A. McCullum; D. Grandmond, D. Carignan; S. Kouz, M. Roy; P. Polasek, V. Stedham; K. Lai, B. Paquette; D. Raco, M. Sayles; K. Sridhar, D. Wiseman; B. Tremblay, C. Darveau; B. Sussex, S. Newman; M. Senaratne, M. Holland; I. Bata, M. Hulan; C. Constance, S. Pouliot; D. Gossard, L. Day; A. Glanz, C. Vilag; W. Hui, L. Kvill; D. Dion, A. Morisette; F. Sandrin, S. Mit- ges; Belgium: F. Cools, S. Vanhagendoren; A. De Meester, O. Marcovitch; E. Allaf, F. Gits; J. Verrostte; J. Thoeng; M. Quinonez, H. Appel- tants; G. D’Hooghe; M. Eycken, I. Swennen; G. Hollanders; E. Michaël; P. Decroly; E. Rombaut, D. Carlaire; D. Vermander; St. F. Marenne, C. Gavray; P. Michel, A. Poth; J. Paquay, S. Lutasu; D. Dirk; J. Vermeulen, E. Govaerts; Russia: S. Boldueva; A. Sherenkov; M. Ruda; V. Kos- tenko; S. Tereschenko; G. Zalevsky; M. Boyarkin; B. Tankhilevitch; M. Glezer; N. Gratsiansky; E. Shlyakhto; B. Sidorenko; V. Zadiontchen- ko; N. Perepech; A. Gruzdev; S. Chernov; Germany: B. Pollock; D. Andresen, S. Hoffmann; D. Andresen; H. Arntz; H. Klein; H. Neuss; E. Wilhelms; R. Henzgen, H. Kuckuck; H. Darius, C. Hausdorf; H. Ochs; H. Olbrich; R. Willenbrock; A. Hepp, P. Wucherpfennig; H. Wiech- mann, H. Schmidtendorf; B. Kohler; W. Sehnert; A. Ueberreiter; R. Uebis, M. Stockmann; U. Rommele; A. Meissner; L. Schneider; United Kingdom: J. Adgey, T. McAllister; D. McEneaney, A. Mackin; N. Sulke, R. Dixon; R. Wilcox, S. Congreave; A. Flapan, L. Flint; J. Purvis, G. McCorkell; P. Stubbs, C. Steer, P. McKee; R. Senior, L. Chester; T. Levy, S. Kennard; A. Bishop, S. Hlaing; M. de Belder, N. Cunningham; T. Gilbert, J. Young; R. Henderson, D. Falcoln-Lang; A. Gershlick, K. Fairbrother; Israel: T. Rosenfeld, S. Atar; B. Lewis, R. Yuval; S. Meisel, I. Alony; Y. Rozenman, S. Logrineuko; E. Goldhammer, S. Harel; D. David, N. Erez; J. Jafari, O. Tubul; A. Marmor, L. Pritulo; A. Caspi, N. Roit- berg; E. Kaluski, R. Amar; D. Tzivoni, A. Rojanski, S. Yedid-Am; A. Keren, L. Datiashvily; Brazil: J. Kerr Saraiva, C. Travaini Garcia; M. Mark- man, A. Chaves; R. Cecin Vaz, O. Dutra; O. Rizzi Coelho, R. Sciampaglia; C. Pereira da Cunha, M. Ribas de Brito; P. Lotufo, E. Ribeiro; A. Cicogna; Y. Lage Michalaros, C. Eduardo Ornellas; S. Carlos de Moreas Santos, P. Rosatelli; W. Pimentel Filho, M. Grudzinski; F. Guimaraes Filho, C. Gustavo; M. Moreira, J. Gosmao; the Netherlands: D. Hertzberger, A. Schut; C. van der Zwaan, H. Havenaar; J. van Wijngaarden, H. Verheij; E. Wajon, M. Bosschaart; A. Oomen; M. Daniëls, A. Coppes; F. Vergeught, H. Dieker; D. Jochemsen, I. Fijlstra; C. Leenders, T. Gelder; United States: D. Janicke, J. Bass; G. Hamroff, K. Keeler; M. Chandra, B. Van Hoose; S. Minor, P. Mock; M. Chandra, M. Olliges; A. Unwala, C. Fisher; W. French, O. Barillas; H. Chandna, D. Holly; H. Chandna, D. Holly; G. Grewal, L. Gurnsey; P. Mehta, A. Cruse; R. Perl- man, D. Palazzo; J. Puma, C. Payne; J. Torres, L. Patten; L. Lefkovic, N. Viswanathan; A. Rees, A. Yoches; L. Rodriguez-Ospina, J. Santos; A. Chohan, A. Kemp; R. Perlman, D. Palazzo; R. Weiss, B. Brennan; Y. Aude, P. Harrison; G. Brogan, J. Ayan; D. Cragg, A. Murawka; M. Garg, J. McKelvy; A. Hulyalkar, C. Kuchenrither; S. Jackson, J. Whitaker, J. Richardson; R. Vicari, M. Howard; Y. Aude, P. Harrison; H. Chadow, R. Hauptman; J. Gelormini, T. Giambra; N. Lakkis, S. Jiang; Mexico: I. Hernandez, A. Vazquez; E. Lopez Rosas; J. Cortes Lawrenz, F. Busta- mante; C. Arean Martinez; M. Antonio Alcocer, E. Garcia Hernandez; C. Wabi Dogre, M. de Los Reyes Barrera Bustillos; R. Alvarado, M. Casares Ramirez; P. Hinojosa, O. Omar Rivera; J. Leiva Pons, J. Carrillo; J.A. Velasco, F. Quintana; F. Petersen, R. Palomera; M. Ibarra, C. Jerjes Sanchez; L. Delgado, E. Bayram; G. Mendez Machado; Argentina: M.A. Berli, J.C. Vinuela; O. Allall, V. Fuentealba, M. Reguero; G. Covelli, E. San Martin, E. Francia; F. Gadaleta, M. Haehnel; R. Iglesias, S. Blumberg; A. Alves de Lima, R. Henquin; J. Navarro Estrada, D. de Arenaza; R. Fernandez, M. Abud, C. Becker; R. Milesi, R. Schamuck, J. Manuel Doyharzabal; D.G. Caime, L.C. Teresa; E. Kuschnir, H. Sgammini, J. Sgammini, J. David; C.H. Sosa, E. Redcozub, V. Avalos; D. Nul, S. Ramos, P. T. Castro; A. Rodriguez, C. Sosa; J. Gant Lopez,

n engl j med 352;12 www.nejm.org march 24, 2005 1187

The new england journal of medicine

F. Novo; F. Diez; R. Piraino, G. Godoy, C. Weller; J. Pomposiello; A. Sanchez; Sweden: J. Karlsson, K. Collberg; K. Malmqvist, E. Cedergren; M. Dellborg, G. Coops; A. Kallryd, P. Laudon; S. Bandh, C. Andersson; S. Soome, H. Classon; J. Lugnegard, H. Ahlmark; M. Risenfors, L. Ortgren; B. Fjelstad, S. Osberg; J. Frisell, M. Sandstrom; Italy: O. Silvestri, A. Sasso; A. Rolli, G. Paoli; F. Chiarella, L. Olivotte; B. Doronzo, A. Coppolino; G. Pivoccari, A. Pesaresi; G. Ambrosio, M. Del Pinto; I. De Luca, M. Laico; G. De Ferrari, E. Baldini; U. Guiducci, G. Tortorel- la; M. Romano, R. Villani; F. Casazza, A. Capozi; L. Moretti, M. Del Pinto; S. Perelli; G. Sinagra, P. Maras; M. Porcu, G. Scorcu; F. Proietti; S. Severi; Poland: J. Gessek, L. Pawlowicz; R. Szelemej, M. Gorski; M. Janion, J. Sielski; A. Kleinrok, J. Rodzik; J. Hiczkiewicz; M. Mikulicz Pasler, M. Soltsiak; J. Gorny; Austria: K. Huber, N. Jordanova; M. Heigert, K. Kopp; H. Frank; A. Laggner, W. Schreiber; W. Weihs; M. Heinz Drexel; South Africa: A. Doubell, L. Burgess; J. Bennett, F. Snyders; I. Ebrahim, C. Bobak; J. Marx, M. deBeer; S. Schmidt, H. Cyster; P. van der Berg, W. Groenewald; L. Herbst, A. Britz; M. Snyman, M. Schonken; J. Badenhorst, P. Blomerus; M. Mpe, C. Bobak; J. Mynhardt, C. Groenewald; C. van Dyk, J. Meiring; L. Steingo, L. Glaeser; Australia: A. Whelan, G. Tulloch; B. Gunalingham, B. Conway; A. Soward, S. Hales; J. Lefkovits, M. Sallaberger; D. Owesby, J. Kesby; J. Amerena, K. Fogarty; D. Chew, M. Austin, S. Kovaricek; N. Banham, G. Tul- loch; Hungary: L. Jobbagy; J. Dinnyes; J. Lippai; B. Oze; I. Szakal; G. Bartal, A. Foldi; Ireland: D. Foley, F. McGrath; B. Meany, M. Geraghty; K. Daly, S. Henessey; N. Mulvihill, C. Schilling; P. Kearney, P. Kelly; P. Kearney, P. Kelly; Turkey: Z. Ongen, A. Vura; O. Kozan, E. Ozaydin; H. Kultursay, M. Kayikcioglu; M. Degertekin, B. Mutlu; Portugal: F. Monteiro; D. Ferreira; M. Carrageta.

references 1. The TIMI Study Group. The Thrombol- 11. The Clopidogrel in Unstable Angina to Effectiveness of intravenous thrombolytic ysis in Myocardial Infarction (TIMI) trial: Prevent Recurrent Ischemic Events Trial Inves- treatment in acute myocardial infarction. Phase I findings. N Engl J Med 1985;312: tigators. Effects of clopidogrel in addition Lancet 1986;1:397-402. 932-6. to aspirin in patients with acute coronary 19. The GUSTO Investigators. An interna- 2. The GUSTO Angiographic Investigators. syndromes without ST-segment elevation. tional randomized trial comparing four The effects of tissue plasminogen activator, N Engl J Med 2001;345:494-502. [Erratum, thrombolytic strategies for acute myocardial streptokinase, or both on coronary-artery N Engl J Med 2001;345:1506, 1716.] infarction. N Engl J Med 1993;329:673-82. patency, ventricular function, and survival 12. Sabatine MS, McCabe CH, Gibson CM, 20. The Global Use of Strategies to Open after acute myocardial infarction. N Engl J Cannon CP. Design and rationale of Clopi- Occluded Coronary Arteries. (GUSTO III) Med 1993;329:1615-22. [Erratum, N Engl J dogrel as Adjunctive Reperfusion Therapy Investigators. A comparison of reteplase with Med 1994;330:516.] (CLARITY)–Thrombolysis in Myocardial Inf- alteplase for acute myocardial infarction. 3. Braunwald E. The open-artery theory is arction (TIMI) 28 trial. Am Heart J 2005; N Engl J Med 1997;337:1118-23. alive and well — again. N Engl J Med 1993; 149:227-33. 21. Assessment of the Safety and Efficacy 329:1650-2. 13. Antman EM, Anbe DT, Armstrong PW, of a New Thrombolytic Investigators. Sin- 4. Dalen JE, Gore JM, Braunwald E, et al. et al. ACC/AHA guidelines for the manage- gle-bolus tenecteplase compared with Six- and twelve-month follow-up of the ment of patients with ST-elevation myo- front-loaded alteplase in acute myocardial phase I Thrombolysis in Myocardial Infarc- cardial infarction: a report of the American infarction: the ASSENT-2 double-blind tion (TIMI) trial. Am J Cardiol 1988;62:179- College of Cardiology/American Heart Asso- randomised trial. Lancet 1999;354:716-22. 85. [Erratum, Am J Cardiol 1988;62:1151.] ciation Task Force on Practice Guidelines 22. Antman EM, Giugliano RP, Gibson CM, 5. Ohman EM, Califf RM, Topol EJ, et al. for the Management of Patients with Acute et al. Abciximab facilitates the rate and extent Consequences of reocclusion after success- Myocardial Infarction. Circulation 2004;110: of thrombolysis: results of the Thromboly- ful reperfusion therapy in acute myocardial e82-e292. sis in Myocardial Infarction (TIMI) 14 trial. infarction. Circulation 1990;82:781-91. 14. Bovill EG, Terrin ML, Stump DC, et al. Circulation 1999;99:2720-32. 6. ISIS-2 (Second International Study of Hemorrhagic events during therapy with 23. Topol EJ, GUSTO V Investigators. Reper- Infarct Survival) Collaborative Group. Ran- recombinant tissue-type plasminogen acti- fusion therapy for acute myocardial infarc- domised trial of intravenous streptokinase, vator, heparin, and aspirin for acute myocar- tion with fibrinolytic therapy or combina- oral aspirin, both or neither among 17,187 dial infarction: results of the Thrombolysis tion reduced fibrinolytic therapy and platelet cases of suspected acute myocardial infarc- in Myocardial Infarction (TIMI), Phase II glycoprotein IIb/IIIa inhibition: the GUSTO V tion: ISIS-2. Lancet 1988;2:349-360. Trial. Ann Intern Med 1991;115:256-65. randomised trial. Lancet 2001;357:1905-14. 7. Meijer A, Verheugt FW, Werter CJ, Lie 15. Yasuda T, Gold HK, Leinbach RC, et al. 24. Assessment of the Safety and Efficacy of KI, van der Pol JM, van Eenige MJ. Aspirin Lysis of plasminogen activator-resistant a New Thrombolytic Regimen (ASSENT)-3 versus coumadin in the prevention of re- platelet-rich coronary artery thrombus with Investigators. Efficacy and safety of tenec- occlusion and recurrent ischemia after suc- combined bolus injection of recombinant teplase in combination with enoxaparin, cessful thrombolysis: a prospective placebo- tissue-type plasminogen activator and anti- abciximab, or unfractionated heparin: the controlled angiographic study: results of the platelet GPIIb/IIIa antibody. J Am Coll Car- ASSENT-3 randomised trial in acute myo- APRICOT Study. Circulation 1993;87:1524- diol 1990;16:1728-35. cardial infarction. Lancet 2001; 358:605-13. 30. 16. Patrono C, Coller B, FitzGerald GA, Hirsh 25. Antman EM, Morrow DA, McCabe CH, 8. Quinn MJ, Fitzgerald DJ. Ticlopidine J, Roth G. Platelet-active drugs: the relation- et al. Enoxaparin versus unfractionated hep- and clopidogrel. Circulation 1999;100:1667- ships among dose, effectiveness, and side arin as antithrombin therapy in patients 72. effects: the Seventh ACCP Conference on receiving fibrinolysis for ST elevation MI: 9. CAPRIE Steering Committee. A random- Antithrombotic and Thrombolytic Therapy. design and rationale for the Enoxaparin and ised, blinded, trial of clopidogrel versus aspi- Chest 2004; 126:234S-264S. Thrombolysis Reperfusion for Acute Myocar- rin in patients at risk of ischaemic events 17. Herbert JM, Dol F, Bernat A, Falotico R, dial Infarction Treatment Thrombolysis in (CAPRIE). Lancet 1996;348:1329-39. Lale A, Savi P. The antiaggregating and anti- Myocardial Infarction–Study 25 (ExTRACT– 10. Steinhubl SR, Berger PB, Mann JT III, et thrombotic activity of clopidogrel is potenti- TIMI 25). Am Heart J 2005;149:217-26. al. Early and sustained dual oral antiplatelet ated by aspirin in several experimental mod- 26. Gibson CM, Cannon CP, Murphy SA, et therapy following percutaneous coronary els in the rabbit. Thromb Haemost 1998;80: al. Relationship of TIMI myocardial perfu- intervention: a randomized controlled trial. 512-8. sion grade to mortality after administration JAMA 2002;288:2411-20. [Erratum, JAMA 18. Gruppo Italiano per lo Studio della Strep- of thrombolytic drugs. Circulation 2000; 2003;289:987.] tochinasi nell’Infarto Miocardico (GISSI). 101:125-30.

1188 n engl j med 352;12 www.nejm.org march 24, 2005 clopidogrel and fibrinolysis in myocardial infarction

27. Mueller HS, Forman SA, Menegus MA, point in thrombolytic trials of acute myocar- among patients with suspected acute myo- Cohen LS, Knatterud GL, Braunwald E. Prog- dial infarction (TIMI 4 and 5): Thrombosis cardial infarction. J Cardiovasc Risk 2000;7: nostic significance of nonfatal reinfarc- in Myocardial Infarction. Am J Cardiol 1997; 435-41. tion during 3-year follow-up: results of 80:696-9. 30. Giugliano RP, McCabe CH, Antman the Thrombolysis in Myocardial Infarction 29. Second Chinese Cardiac Study (CCS-2) EM, et al. Lower-dose heparin with fibrino- (TIMI) phase II clinical trial. J Am Coll Car- Collaborative Group. Rationale, design and lysis is associated with lower rates of intra- diol 1995;26:900-7. organization of the Second Chinese Cardiac cranial hemorrhage. Am Heart J 2001;141: 28. Cannon CP, Sharis PJ, Schweiger MJ, et Study (CCS-2): a randomized trial of clo- 742-50. al. Prospective validation of a composite end pidogrel plus aspirin, and of metoprolol, Copyright © 2005 Massachusetts Medical Society.

n engl j med 352;12 www.nejm.org march 24, 2005 1189

The new england journal of medicine

original article

Hospitalization for Mental Illness among Parents after the Death of a Child

Jiong Li, M.D., Ph.D., Thomas Munk Laursen, M.Sc., Dorthe Hansen Precht, M.D., Ph.D., Jørn Olsen, M.D., Ph.D., and Preben Bo Mortensen, M.D., Dr.Med.Sc.

abstract

background From the Danish Epidemiology Science The loss of a child is considered one of the most stressful events in the life of a parent. Center, Department of Epidemiology and We hypothesized that parental bereavement increases the risk of hospital admission Social Medicine (J.L., J.O.) and the National Center for Register-Based Research (T.M.L., for a psychiatric disorder, especially for affective disorders. P.B.M.), University of Aarhus, Aarhus, Den- mark; and the Clinical Unit of Preventive methods Medicine and Health Promotion, Bispebjerg Hospital, Copenhagen (D.H.P.). Address We studied a cohort of 1,082,503 persons identified from national registers in Den- reprint requests to Dr. Olsen at the Depart- mark who were born between 1952 and 1999 and had at least one child under 18 years ment of Epidemiology, School of Public of age during the follow-up period, from 1970 to 1999. Parents who lost a child during Health, UCLA, Box 951772, Los Angeles, CA 90095-1772, or at [email protected]. follow-up were categorized as “bereaved” from the date of death of the child.

N Engl J Med 2005;352:1190-6. results Copyright © 2005 Massachusetts Medical Society. As compared with parents who did not lose a child, parents who lost a child had an overall relative risk of a first psychiatric hospitalization for any disorder of 1.67 (95 percent confidence interval, 1.53 to 1.83). Bereaved mothers had a higher relative risk of being hospitalized for any psychiatric disorder than bereaved fathers (relative risks, 1.78 [95 percent confidence interval, 1.60 to 1.98] and 1.38 [95 percent confidence interval, 1.17 to 1.63], respectively; P value for interaction, 0.01). The relative risks of hospitalization specifically for affective disorders were 1.91 (95 percent confidence interval, 1.59 to 2.30) and 1.61 (95 percent confidence interval, 1.15 to 2.27) for bereaved mothers and fathers, respectively. Among mothers, the relative risk of being hospitalized for any psychiatric disorder was highest during the first year after the death of the child but remained significantly elevated five years or more after the death.

conclusions The risk of psychiatric hospitalization was increased among parents, especially mothers, who lost a child.

1190 n engl j med 352;12 www.nejm.org march 24, 2005

mental illness after parental bereavement

he death of a child is one of the Table 1. Relative Risk of a First Psychiatric Hospitalization among Bereaved most stressful events in a parent’s life.1,2 t Parents, Stratified According to the Parent’s Sex.* Several small cross-sectional and follow- up studies have suggested high rates of symptoms Parent, Diagnosis, No. of Relative Risk and Bereavement Admissions Person-Years (95% CI)† of anxiety and depression in parents who have lost a child.2-6 However, prospective data are lacking to Mothers support an association between the death of a child All mental disorders and subsequent clinical psychiatric disease in those ≥2 children‡ 025 4,036 3.35 (2.26–4.97) parents,2,3 although it has been proposed that pa- 1 child 356 111,506 1.73 (1.55–1.92) rental bereavement is more difficult to cope with None 11,333 6,538,730 1.00§ than any other type of bereavement.2,7 The loss of a Total 11,714 6,654,272 parent during childhood or the loss of a spouse has Affective disorders been shown to increase the risk of clinical mental Bereaved 125 120,747 1.91 (1.59–2.30) illness later in life.8-12 Depression and other affec- Not bereaved 3,523 6,692,558 1.00§ tive disorders are considered to be particular risks Schizophrenia after the loss of a family member.13-15 We hypothe- Bereaved 73 121,571 1.89 (1.48–2.40) sized that the loss of a child would increase the risk Not bereaved 2,311 6,709,890 1.00§ of subsequent psychiatric illness in a parent, in Substance abuse particular for affective disorders. We expected the Bereaved 151 120,201 2.16 (1.82–2.56) effect to be strongest shortly after the death of the Not bereaved 3,584 6,687,760 1.00§ child and the risk to be affected by the sex of the Fathers parent, the age of the deceased child, and the num- All mental disorders ber of children in the family.2,3,16 ≥2 children‡ 7 1,613 2.39 (1.14–5.03) Even an extremely stressful life event affects peo- 1 child 139 58,126 1.35 (1.14–1.61) ple in different ways.2,3,16 If life events carry a risk None 8,707 4,837,699 1.00§ of severe health problems for only a small fraction Total 8,853 4,897,438 of the population, a large study is needed to detect Affective disorders such associations. We thus assessed the associa- Bereaved 35 61,954 1.61 (1.15–2.27) tion between parental bereavement and psychiatric Not bereaved 1,539 4,970,850 1.00§ hospitalization with the use of data from large na- Schizophrenia tional registries in Denmark. Bereaved 23 62,099 1.76 (1.15–2.67) Not bereaved 1,270 4,972,181 1.00§ Substance abuse methods Bereaved 88 61,124 1.43 (1.15–1.77) study design, follow-up, Not bereaved 5,233 4,919,412 1.00§ and data collection * During follow-up, we registered 1,211,634 births and a total of 11,895 deaths We conducted our study by linking records from of children who were under 18 years of age. A total of 17,033 parents were sub- 17 the Danish Civil Registration System and the Dan- sequently included in the bereaved cohort. ish Psychiatric Central Register.18 The registration † Relative risks were adjusted for calendar period, age of parents at bereavement, number of children in the family (including the child who died), and par- system included 1,082,503 persons who were born ents’ age at first birth. CI denotes confidence interval. in Denmark between January 1, 1952, and January 1, ‡ These parents lost more than one child during follow-up. 1999, and had at least one child under 18 years of § This group served as the reference group. age before January 1, 1999. The Danish Psychiat- ric Central Register has been computerized since April 1, 1969, and includes information on all psy- registration in the Danish Psychiatric Central Reg- chiatric hospitalizations in Denmark and the as- ister. We registered 1,211,634 births during fol- sociated diagnoses. For all cohort members, the low-up. The follow-up ended on the date of a first follow-up started on the date of birth of the first psychiatric diagnosis, the date of death, the date child or on April 1, 1970 (for 0.2 percent), whichev- of emigration, or January 1, 1999, whichever came er came last. We chose April 1, 1970, as the start date first. Altogether, 19,124 persons did not contribute for follow-up because some of the information we to the person-years count owing to the fact that they needed was not complete during the first year of died, left the country, disappeared, were diagnosed

n engl j med 352;12 www.nejm.org march 24, 2005 1191

The new england journal of medicine

istered a total of 11,895 deaths of children under Table 2. Relative Risk of First Psychiatric Hospitalization among Bereaved 18 years of age. For the 593 parents who lost more Parents According to the Age of the Child at the Time of Death, Stratified According to the Parent’s Sex. than one child, the date of death of the first child was set as the start of exposure. Twenty-nine parents No. of Diagnosis, Parent, and Hospital Person- Relative Risk lost two children on the same day. The bereaved Age of Deceased Child Admissions Years (95% CI)* cohort consisted of 17,033 parents. Affective disorders Outcomes of interest included all first admis- Mothers sions due to mental illness (International Classifica- Child ≥6 yr 12 6,781 2.72 (1.54–4.81) tion of Diseases, 8th Revision [ICD-8], codes 290 to Child 1–5 yr 17 16,083 1.93 (1.19–3.11) 315 for the period from 1970 through 1993, and Child ≤1 yr 96 97,883 1.84 (1.49–2.27) International Classification of Diseases, 10th Revision Trend† 1.06 (0.97–1.15) [ICD-10], codes F00 through F99 for the period Fathers from 1994 through 1999). We were particularly in- Child ≥6 yr 3 3,592 1.85 (0.59–5.75) terested in hospitalizations for affective disorders Child 1–5 yr 12 10,079 3.32 (1.88–5.89) (ICD-8 codes 296.09 through 296.99, 298.19, Child ≤1 yr 20 48,283 1.21 (0.78–1.90) 300.19, and 300.49, and ICD-10 codes F30 through Trend† 1.13 (0.98–1.30) F39), because grief may cause a clinical picture dom- Schizophrenia inated by depressive symptoms. However, we also Mothers analyzed substance abuse (ICD-8 codes 291.xx, Child ≥6 yr 3 6,839 1.09 (0.35–3.40) 303.xx, and 304.xx, and ICD-10 codes F10 through Child 1–5 yr 15 16,184 2.79 (1.67–4.65) F19) and schizophrenia and related disorders (ICD-8 Child ≤1 yr 55 98,548 1.80 (1.37–2.37) codes 295.xx, 296.8x, 297.xx, 298.39, and 301.83, Trend† 0.99 (0.87–1.13) and ICD-10 code F2x). Separate analyses were per- Fathers formed for each diagnostic group. In these analy- Child ≥6 yr 0 3,604 ses, follow-up ended on the date of admission of a Child 1–5 yr 4 10,118 1.81 (0.68–4.84) parent who was given the diagnosis of the disorder Child ≤1 yr 19 48,377 1.90 (1.20–3.01) under study. We also examined a possible modify- Trend† 0.83 (0.59–1.16) ing effect of parental sex and age, the age of the de- Substance abuse ceased child, the time since the death of the child, Mothers and the number of children in the family on the rel- Child ≥6 yr 7 6,749 1.36 (0.65–2.87) ative risks of psychiatric admission. Child 1–5 yr 19 16,024 1.99 (1.27–3.14) The study was approved by the Danish Data Pro- Child ≤1 yr 125 97,428 2.26 (1.88–2.72) tection Agency, whose role is to ensure that the pri- Trend† 0.93 (0.84–1.03) vacy and integrity of the individual subjects record- Fathers ed in the registries are protected. No informed Child ≥6 yr 3 3,497 0.78 (0.25–2.42) consent was required for the study. Child 1–5 yr 13 10,007 1.27 (0.74–2.20) statistical analysis Child ≤1 yr 72 47,620 1.51 (1.19–1.91) Trend† 0.91 (0.79–1.06) Data were analyzed in a log-linear Poisson regression model (SAS Genmod procedure, version 8.1). * Relative risks were adjusted for calendar period, age of parents at bereavement, We fitted data with the use of generalized linear number of children in the family (including the child who died), and parents’ models and by treating observations as indepen- age at first birth. The reference group was parents who did not lose a child. dent Poisson variables, with the number of person- CI denotes confidence interval. 19,20 † The trend estimate expresses the change in risk associated with an increase years serving as an offset variable. We adjusted of one year in the child’s age. for the calendar period of follow-up (1970 to 1973, 1974 to 1978, 1979 to 1983, 1984 to 1988, 1989 to 1993, or 1994 to 1999), the age of the parents at the with a mental illness before April 1, 1970, or were time of the child’s death (<20 years, 20 to 24 years, hospitalized for a mental illness before they had a 25 to 29 years, 30 to 34 years, 35 to 39 years, 40 to child. Parents who lost a live-born child under 18 45 years, or >45 years), the age of the parents when years of age contributed observation time to the be- the first child was born (<20 years, 20 to 29 years, reaved group from the date the child died. We reg- or ≥30 years), and the number of children in the

1192 n engl j med 352;12 www.nejm.org march 24, 2005

mental illness after parental bereavement family at the time of the death, including the child who died (one, two, three, or more). Table 3. Relative Risk of First Psychiatric Hospitalization among Bereaved Parents According to the Time since the Death of the Child, Stratified The parent’s age, the calendar period, the cur- According to the Parent’s Sex. rent number of children, and the number of chil- No. of dren who died were treated as time-dependent var- Diagnosis, Parent, Hospital Person- Relative Risk iables. All reported P values are two-sided. We used and Time since Death Admissions Years (95% CI)* the likelihood-ratio test to test for a trend in the Affective disorders analyses according to the age of the deceased child. Mothers We used weights of 0.5 in the group of children who <1 yr 34 10,279 6.83 (4.86–9.59) were younger than 1 year of age, 3.0 in the group 1–4 yr 33 36,640 1.86 (1.32–2.63) 1 to 5 years of age, and 7.0 in the group 6 years of ≥5 yr 58 73,828 1.33 (1.02–1.75) age or older. A trend in the time since the death of Trend† 0.82 (0.77–0.88) the child was assessed with weights of 0.5 in the Fathers group for which less than 1 year had passed since <1 yr 10 6,402 5.83 (3.13–10.88) the loss, 2.5 in the group for which 1 to 4 years had 1–4 yr 10 21,890 1.57 (0.84–2.94) passed, and 6.0 in the group for which 5 or more years had passed. A trend in the number of chil- ≥5 yr 15 33,662 1.09 (0.65–1.82) dren was assessed with weights of 1, 2, and 3. Trend† 0.82 (0.72–0.93) Schizophrenia Mothers methodsresults <1 yr 13 10,318 4.53 (2.62–7.82) During 11,551,710 person-years of follow-up 1–4 yr 20 36,836 2.02 (1.30–3.15) (4,897,438 for fathers and 6,654,272 for mothers), ≥5 yr 40 74,417 1.52 (1.10–2.10) there were 20,567 first psychiatric admissions (8853 Trend† 0.88 (0.81–0.97) among fathers and 11,714 among mothers) (Table Fathers 1). There were 11,376,429 person-years of follow- <1 yr 4 6,410 2.68 (1.00–7.16) up for parents who did not lose a child, and 175,281 1–4 yr 8 21,937 1.80 (0.89–3.61) person-years for the bereaved group (0.6 percent of ≥5 yr 11 33,751 1.53 (0.84–2.80) the person-years in the nonbereaved group came Trend† 0.94 (0.80–1.10) from the bereaved parents before they lost a child). Substance abuse A total of 9241 parents died, and their data were Mothers censored during follow-up before any psychiatric <1 yr 14 10,254 2.93 (1.73–4.95) admission. 1–4 yr 53 36,516 3.15 (2.40–4.15) The relative risk of hospitalization for any psy- ≥5 yr 84 73,431 1.72 (1.37–2.15) chiatric disease among parents who lost a child was Trend† 0.90 (0.84–0.95) 1.67 (95 percent confidence interval, 1.53 to 1.83), Fathers after adjusting for calendar period, sex and age of <1 yr 10 6,354 1.48 (0.80–2.76) the bereaved parents, number of children in the fam- 1–4 yr 28 21,701 1.34 (0.92–1.94) ily, and parents’ age at first birth. Bereaved mothers ≥5 yr 50 33,069 1.47 (1.11–1.93) had a higher relative risk of being admitted for any Trend† 0.74 (0.47–1.16) mental disorder than bereaved fathers (relative risks, 1.78 [95 percent confidence interval, 1.60 to * Relative risks were adjusted for calendar period, age of the parents at bereave- 1.98] and 1.38 [95 percent confidence interval, 1.17 ment, number of children in the family (including the child who died), and parents’ age at first birth. The reference group was parents who did not lose a child. to 1.63], respectively; P value for interaction, 0.01). CI denotes confidence interval. Mothers who had lost two or more children had a † The trend estimate expresses the change in risk associated with an increase relative risk of a first hospitalization for any psy- of one year in the time since the child’s death. chiatric disorder of 3.35 (95 percent confidence interval, 2.26 to 4.97), and fathers had a relative risk of 2.39 (95 percent confidence interval, 1.14 to chiatric diagnostic group we assessed was increased 5.03), as compared with nonbereaved mothers and among both mothers and fathers who lost a child fathers, respectively. as compared with mothers and fathers who did not The relative risk of hospitalization for each psy- (Table 1). The age of the child at the time of death

n engl j med 352;12 www.nejm.org march 24, 2005 1193

The new england journal of medicine

orders among mothers or fathers 30 years of age Table 4. Relative Risk of First Psychiatric Hospitalization among Bereaved or older was not significantly different from the Parents According to the Age of the Parents, Stratified According to the Parent’s Sex. risk among mothers or fathers less than 30 years of age (Table 4). Among mothers, the relative risk of Diagnosis, Parent, hospital admission for any of the three diagnostic and Age of Parent No. of Person- Relative Risk at Child’s Death Admissions Years (95% CI)* P Value† groups significantly decreased with an increase in the number of children in the family (Table 5). Affective disorders Among fathers, the relative risk of admission for af- Mothers 0.45 fective disorders was similarly reduced with an in- <30 yr 100 100,694 1.85 (1.51–2.27) crease in the number of children in the family, but ≥30 yr 25 20,054 2.19 (1.47–3.25) there was no significant trend toward a reduced risk Fathers 0.66 of admission for other psychiatric disorders with <30 yr 24 42,398 1.70 (1.13–2.56) an increase in the number of children in the family. ≥30 yr 11 19,556 1.45 (0.80–2.64) Schizophrenia discussion Mothers 0.23 <30 yr 57 101,323 1.76 (1.35–2.31) We found an increased risk of first psychiatric hospi- ≥30 yr 16 20,247 2.51 (1.53–3.56) talization among parents, especially mothers, who lost a child less than 18 years of age. Among moth- Fathers 0.50 ers, the relative risks of all types of psychiatric hos- <30 yr 15 42,513 1.60 (0.95–2.67) pitalization were highest during the first year after ≥30 yr 8 19,585 2.16 (1.07–4.36) bereavement, remained significantly increased five Substance abuse years or more after the loss, and decreased in accor- Mothers 0.66 dance with an increase in the number of children in <30 yr 128 100,262 2.13 (1.77–2.55) the family. Among fathers, the trend in relative risks ≥30 yr 23 19,941 2.35 (1.56–3.56) according to time since the loss and the number of Fathers 0.11 children in the family seemed similar to the trend <30 yr 72 41,872 1.56 (1.23–1.98) among mothers, but the numbers of events were ≥30 yr 16 19,252 1.02 (0.62–1.67) smaller among fathers. Previous findings on the association between * Relative risks were adjusted for calendar period, age of the parents at bereave- the loss of a first-degree relative and the risk of psy- ment, number of children in the family (including the child who died), and par- chiatric illness have been inconsistent. One hos- ents’ age at first birth. The reference group was parents who did not lose a child. CI denotes confidence interval. pital-based case–control study found no excess fre- † The P value is for the difference between age groups. quency of loss of a first-degree relative in the six months preceding the admission,14 but an association was reported in three other hospital-based did not significantly change the risk of hospital- case–control studies.8,21,22 A follow-up study found ization for any of the diagnostic groups (Table 2). that a widowed population had more psychiatric Among mothers, the relative risk of being hospital- admissions during the first year of bereavement ized for any of the psychiatric disorders was highest than nonbereaved women.9 Information on the ef- during the first year after the death of the child (Ta- fect of parental bereavement has been sparse.2,3 One ble 3). The elevated relative risk of each of these small study indicated that bereavement due to the disorders declined significantly with time since the death of a child was more frequent than other types loss but remained significantly elevated after five of bereavement in psychiatric patients.22 years or more. Among fathers, a trend toward a de- Some have suggested that bereavement may cline in the relative risk over time since the loss was be more closely associated with depression than observed for affective disorders only, and only the with other psychiatric disorders.13,15 Other psychi- risk of hospitalization for substance abuse remained atric diagnoses, however, have not been well stud- significantly elevated after five years. However, sta- ied.3 In our data, the association between the loss tistical power was limited by the smaller number of of a child and psychiatric hospitalization was not events among fathers than among mothers. restricted to affective disorders. The risk of hospitalization for psychiatric dis- Grief is often most intense shortly after bereave-

1194 n engl j med 352;12 www.nejm.org march 24, 2005

mental illness after parental bereavement ment.2,3 Thus, the observation of a greater relative risk of hospitalization among some parents during Table 5. Relative Risk of First Psychiatric Hospitalization among Bereaved Parents According to the Number of Children in the Family, Stratified the first year after a loss is not surprising. According to the Parent’s Sex. One possible explanation for these findings is that the threshold for psychiatric hospitalization Diagnosis, Parent, No. of Person- Relative Risk might be lower among persons who have lost a and No. of Children* Admissions Years (95% CI)† child. However, in this case, we would expect that Affective disorders the relative risks would decline over time to protec- Mothers tive levels (below 1.0), since parents would be hos- 1 29 13,272 3.79 (2.62–5.48) pitalized at an earlier stage of psychiatric illness and, 2 38 33,188 2.32 (1.68–3.20) therefore, removed from the population at risk. In contrast, the relative risks remained significantly ≥3 58 74,288 1.35 (1.03–1.77) increased in some analyses even five years or more Trend‡ 0.58 (0.45–0.79) after the loss. Fathers It is also possible that “reverse causation” might 1 9 7,534 4.01 (2.08–7.75) contribute to the association between the loss of a 2 11 17,486 2.29 (1.26–4.16) child and psychiatric hospitalization, particularly in bereaved mothers who lost a child less than one year ≥3 15 36,935 0.99 (0.59–1.67) of age. Infants of schizophrenic mothers, for ex- Trend‡ 0.48 (0.31–0.76) ample, have twice the risk of death by sudden in- Schizophrenia 23 fant death syndrome. However, the relative risk Mothers of hospitalization for schizophrenia (or any of the 1 18 13,372 3.22 (2.02–5.14) other mental disorders we studied) was not higher among mothers who lost a child less than one year 2 17 33,497 1.84 (1.14–2.98) of age than among those who lost a child between ≥3 38 74,702 1.56 (1.11–2.18) one and five years of age. Trend‡ 0.72 (0.52–0.99) Studies have shown that mothers often have Fathers more physical, somatic, and mental problems, in- 1 7 7,558 2.75 (1.31–5.79) cluding suicide,24 after the loss of a child than do 2 8 17,476 2.33 (1.16–4.70) fathers.16,25 The higher relative risk of psychiatric hospitalization among bereaved mothers in our ≥3 8 37,064 1.09 (0.52–2.22) study is in line with those observations, since the Trend‡ 0.68 (0.40–1.16) rate of hospitalization for mental disorders in the Substance abuse nonbereaved population was of the same magni- Mothers tude for fathers and mothers. Parents who lost their 1 38 13,232 3.96 (2.87–5.46) only child had the greatest risk of psychiatric admission. This suggests that having other children 2 45 32,962 3.00 (2.23–4.04) in the family may mitigate the effects of the loss of a ≥3 68 74,008 1.43 (1.11–1.84) child on a parent’s mental health, although we can- Trend‡ 0.59 (0.47–0.73) not rule out that having a large family may elevate Fathers 26 the threshold for hospitalization. 1 18 7,387 1.76 (1.11–2.80) Unfortunately, we have no longitudinal data on 2 22 17,200 1.51 (0.99–2.30) cohabitation and are unable to assess the risk of hospitalization among single parents as compared ≥3 48 36,537 1.29 (0.96–1.73) with married parents. Neither can we assess the ef- Trend‡ 0.85 (0.64–1.14) fect of psychiatric hospitalization in one parent on the risk of hospitalization in the other parent. * The number of children in the family includes the child who died. † Relative risks were adjusted for calendar period, age of the parents at bereave- Our study has several strengths. We used docu- ment, number of children in the family (including the child who died), and par- mented psychiatric admissions, rather than self- ents’ age at first birth. The reference group was parents who did not lose a child. reported data, as the end point. We had access to a CI denotes confidence interval. ‡ The trend estimate expresses the change in relative risk when the number large population-based sample and had a long fol- of children is increased by one. low-up period; virtually all bereaved parents in Den-

n engl j med 352;12 www.nejm.org march 24, 2005 1195

mental illness after parental bereavement

mark during the period under study were enrolled. and psychiatric hospitalization in the parent (e.g., In addition, all data used in our study were extract- if schizophrenia and sudden infant death syndrome ed from high-quality databases that were compiled share common genes). In addition, we included only independently of the research hypothesis, and we patients who were hospitalized, which would have were able to examine each of the main diagnostic underestimated incidence rates for overall psychiat- subgroups of mental illness. ric illness. Our study also has limitations. We could not ad- Our results are based on earlier hypotheses and just for a family history of psychiatric illness27-29 are not adjusted for the multiple comparisons per- or socioeconomic status.25,30 However, previous formed. Some of these comparisons were based on studies have indicated that the loss of a child affects small numbers, especially for fathers. In conclu- the various social strata with similar frequencies.16 sion, we found an increased risk of first psychiatric Even if less-educated parents (who have a higher hospitalization after the loss of a child, especially risk of psychiatric hospitalization) were overrepre- among mothers, that in many cases remained in- sented in the bereaved group, it is implausible that creased five years or more after the loss. this could account for the entire observed increase Supported by grants from the Danish National Research Founda- in relative risks for psychiatric hospitalization. Con- tion and the Danish Medical Research Council to the Danish Epide- miology Science Center and the National Center for Register-Based founding might also be possible if a shared genetic Research. Drs. Laursen and Mortensen were supported by the Stan- predisposition led to both the death of the child ley Medical Research Institute.

references 1. Diagnostic and statistical manual of 11. Kivela SL, Luukinen H, Koski K, Viramo and 40 controls. Arch Gen Psychiatry 1967; mental disorders, 4th ed.: DSM-IV. Washing- P, Pahkala K. Early loss of mother or father 16:134-45. ton, D.C.: American Psychiatric Association, predicts depression in old age. Int J Geriatr 22. Paykel ES, Myers JK, Dienelt MN, Kler- 1994. Psychiatry 1998;13:527-30. man GL, Lindenthal JJ, Pepper MP. Life 2. Rubin SS, Malkinson R. Parental re- 12. Kendler KS, Neale MC, Kessler RC, Heath events and depression: a controlled study. sponse to child loss across the life cycle: clini- AC, Eaves LJ. Childhood parental loss and Arch Gen Psychiatry 1969;21:753-60. cal and research perspectives. In: Stroebe MS, adult psychopathology in women: a twin 23. Bennedsen BE, Mortensen PB, Olesen Hansson RO, Stroebe W, eds. Handbook of study perspective. Arch Gen Psychiatry 1992; AV, Henriksen TB. Congenital malforma- bereavement research: consequences, cop- 49:109-16. tions, stillbirths, and infant deaths among ing, and care. Washington, D.C.: American 13. Brown GW, Harris TO. Life events and children of women with schizophrenia. Arch Psychological Association, 2001:219-39. illness. New York: Guilford Press, 1989. Gen Psychiatry 2001;58:674-9. 3. Osterweis M, Solomon F, Green M, eds. 14. Frost NR, Clayton PJ. Bereavement and 24. Qin P, Mortensen PB. The impact of pa- Bereavement: reactions, consequences, and psychiatric hospitalization. Arch Gen Psy- rental status on the risk of completed sui- care. Washington, D.C.: National Academy chiatry 1977;34:1172-5. cide. Arch Gen Psychiatry 2003;60:797-802. Press, 1984. 15. Finlay-Jones R, Brown GW. Types of 25. Chen JH, Bierhals AJ, Prigerson HG, 4. Clarke M, Williams AJ. Depression in stressful life event and the onset of anxiety Kasl SV, Mazure CM, Jacobs S. Gender dif- women after perinatal death. Lancet 1979;1: and depressive disorders. Psychol Med 1981; ferences in the effects of bereavement-relat- 916-7. 11:803-15. ed psychological distress in health outcomes. 5. Vance JC, Foster WJ, Najman JM, Embel- 16. Li J, Precht DH, Mortensen PB, Olsen J. Psychol Med 1999;29:367-80. ton G, Thearle MJ, Hodgen FM. Early paren- Mortality in parents after death of a child in 26. Hoyer G, Lund E. Suicide among wom- tal responses to sudden infant death, still- Denmark: a nationwide follow-up study. Lan- en related to number of children in mar- birth or neonatal death. Med J Aust 1991; cet 2003;361:363-7. riage. Arch Gen Psychiatry 1993;50:134-7. 155:292-7. 17. Malig C. The Civil Registration System 27. Jablensky AV, Kalaydjieva LV. Genetic ep- 6. Malkinson R, Bar-Tur L. The aging of in Denmark. Bethesda, Md.: International idemiology of schizophrenia: phenotypes, grief in Israel: a perspective of bereaved par- Institute for Vital Registration and Statistics, risk factors, and reproductive behavior. Am ents. Death Stud 1999;23:413-31. 1996. (IIVRS technical paper no. 66.) J Psychiatry 2003;160:425-9. 7. Cleiren M. Bereavement and adapta- 18. Munk-Jorgensen P, Mortensen PB. The 28. Johnson L, Andersson-Lundman G, tion: a comparative study of the aftermath Danish Psychiatric Central Register. Dan Med Aberg-Wistedt A, Mathe AA. Age of onset of death. Washington, D.C.: Hemisphere, Bull 1997;44:82-4. in affective disorder: its correlation with he- 1993. 19. Andersen PK, Borgen Ø, Gill RD, Keiding reditary and psychosocial factors. J Affect 8. Parkes CM. Bereavement and mental ill- N. Statistical models based on counting pro- Disord 2000;59:139-48. ness. Br J Med Psychol 1965;38:1-26. cesses. New York: Springer-Verlag, 1993. 29. Mortensen PB, Pedersen CB, Wester- 9. Clayton PJ. Mortality and morbidity in 20. Laird N, Olivier D. Covariance analysis gaard T, et al. Effects of family history and the first year of widowhood. Arch Gen Psy- of censored survival data using log-linear place and season of birth on the risk of schizo- chiatry 1974;30:747-50. analysis techniques. J Am Stat Assoc 1981; phrenia. N Engl J Med 1999;340:603-8. 10. Rozenzweig A, Prigerson H, Miller MD, 76:231-40. 30. Lewis G, Sloggett A. Suicide, deprivation, Reynolds CF III. Bereavement and late-life 21. Hudgens RW, Morrison JR, Barchha RG. and unemployment: record linkage study. depression: grief and its complications in Life events and onset of primary affective BMJ 1998;317:1283-6. the elderly. Annu Rev Med 1997;48:421-8. disorders: a study of 40 hospitalized patients Copyright © 2005 Massachusetts Medical Society.

1196 n engl j med 352;12 www.nejm.org march 24, 2005

The new england journal of medicine

original article

ACVBP versus CHOP plus Radiotherapy for Localized Aggressive Lymphoma

Félix Reyes, M.D., Eric Lepage, M.D., Gérard Ganem, M.D., Thierry J. Molina, M.D., Pauline Brice, M.D., Bertrand Coiffier, M.D., Pierre Morel, M.D., Christophe Ferme, M.D., Andre Bosly, M.D., Pierre Lederlin, M.D., Guy Laurent, M.D., and Hervé Tilly, M.D., for the Groupe d’Etude des Lymphomes de l’Adulte (GELA)*

abstract background Chemoradiotherapy is standard treatment for localized aggressive lymphoma. To deter- From the Hôpital Henri Mondor, Assis- mine the optimal therapy for nonelderly persons with low-risk localized lymphoma, tance Publique–Hôpitaux de Paris, Créteil (F.R., E.L.); Clinique Victor Hugo, Le Mans we conducted a randomized trial comparing chemoradiotherapy with chemotherapy (G.G.); Hôpital Hôtel-Dieu, Assistance alone. Publique–Hôpitaux de Paris, Paris (T.J.M.); Hôpital Saint-Louis, Assistance Publique– methods Hôpitaux de Paris, Paris (P.B.); Centre Hospitalier Lyon Sud, Pierre-Bénite (B.C.); Previously untreated patients less than 61 years old with localized stage I or II aggres- Centre Hospitalier de Lens, Lens (P.M.); sive lymphoma and no adverse prognostic factors according to the International Prog- Institut Gustave Roussy, Villejuif (C.F.); Centre Hospitalier Brabois, Vandoeuvre lès nostic Index were randomly assigned to three cycles of cyclophosphamide, doxorubicin, Nancy, Nancy (P.L.); Hôpital Purpan, Tou- vincristine, and prednisone (CHOP) plus involved-field radiotherapy (329 patients) or louse (G.L.); and Centre Henri Becquerel, chemotherapy alone with dose-intensified doxorubicin, cyclophosphamide, vindesine, Rouen (H.T.) — all in France; and the Cli- niques Universitaires de Mont Godinne, bleomycin, and prednisone (ACVBP) plus sequential consolidation (318 patients). Yvoir, Belgium (A.B.). Address reprint requests to Dr. Reyes at the Service d’Héma- results tologie Clinique, Hôpital Henri Mondor, 94010 Créteil, France, or at felix.reyes@ With a median follow-up of 7.7 years, event-free and overall survival rates were signifi- hmn.ap-hop-paris.fr. cantly higher in the group given chemotherapy alone than in the group given CHOP plus radiotherapy (P<0.001 and P=0.001, respectively). The five-year estimates of event- *Participants in the GELA study are listed in the Appendix. free survival were 82 percent (95 percent confidence interval, 78 to 87 percent) for patients receiving chemotherapy alone and 74 percent (95 percent confidence interval, 69 N Engl J Med 2005;352:1197-205. to 78 percent) for those receiving chemoradiotherapy. The respective five-year esti- Copyright © 2005 Massachusetts Medical Society. mates of overall survival were 90 percent (95 percent confidence interval, 87 to 93 percent) and 81 percent (95 percent confidence interval, 77 to 86 percent). In a multivariate analysis, event-free and overall survival rates were affected by treatment group, independently of tumor stage and the presence or absence of bulky disease. conclusions In patients under 61 years of age, chemotherapy with three cycles of ACVBP followed by sequential consolidation is superior to three cycles of CHOP plus radiotherapy for the treatment of low-risk localized lymphoma.

n engl j med 352;12 www.nejm.org march 24, 2005 1197

The new england journal of medicine

adiotherapy was the standard kidney failure, and a cardiac contraindication to r treatment for limited (stage I or II) aggres- doxorubicin. Patients with intestinal lymphoma sive lymphoma until 1980, although the that could not be safely encompassed in a radiation five-year rate of disease-free survival was less than 50 field were also excluded from the study. The trial percent.1 Subsequently, chemotherapy was added was approved by the ethics committee of Hôpital to involved-field radiotherapy (chemoradiotherapy) Saint-Louis, and all patients gave written informed with the goals of controlling occult systemic disease consent. and reducing the size of irradiation fields.2 Other investigators, however, advanced the idea that che- pathology motherapy alone could cure limited aggressive lym- The Working Formulation8 and the Kiel classifi- phoma3; as a result, either chemoradiotherapy or cation9 were used to classify lymphoma at the time chemotherapy alone was used for localized lym- of enrollment. A central review was conducted by at phoma4 until the superiority of three cycles of least two pathologists from the GELA, and tumors cyclophosphamide, doxorubicin, vincristine, and were then reclassified according to the classifica- prednisone (CHOP) followed by involved-field ration of the World Health Organization.10 diotherapy over eight cycles of CHOP alone was demonstrated.5 staging The Groupe d’Etude des Lymphomes de l’Adulte The extent of the disease was evaluated by means of (GELA) has developed a chemotherapy regimen that physical examination, computed tomography of the consists of an induction phase of intensified doxo- chest, abdomen, and pelvis, cerebrospinal fluid ex- rubicin, cyclophosphamide, vindesine, bleomycin, amination, bone marrow biopsy, and other investi- and prednisone (ACVBP) followed by sequential gational procedures depending on clinical symp- consolidation. In a previous study of two chemo- toms. The stage of lymphoma was defined on the therapy regimens for intermediate or high-grade basis of the Ann Arbor classification: stage I was de- lymphoma,6 the estimated five-year rate of overall fined as the involvement of a single lymph-node survival among patients with localized disease who region or extranodal site, and stage II as the in- received the ACVBP regimen was 80 percent, which volvement of two or more nodal regions or involve- is similar to results in other trials that used chemora- ment of an extranodal site and one or more adjacent diotherapy. We therefore launched the present ran- nodal regions on the same side of the diaphragm. domized study to compare the ACVBP regimen with Tumor measurements were obtained before biop- chemoradiotherapy in patients under 61 years of age sy, and bulky disease was defined as any mass 10 cm who had localized aggressive lymphoma and no ad- or more in maximal diameter. Performance status verse prognostic factors, as defined by the age- was assessed according to the Eastern Cooperative adjusted International Prognostic Index.7 Oncology Group scale, and the serum lactate dehydrogenase level was expressed as the ratio of the 7 methods maximal value to the normal value.

patients treatment Study patients had to be older than 15 and younger Patients were randomly assigned to treatment with than 61 years of age, to have newly diagnosed ag- chemotherapy alone or chemoradiotherapy, both gressive lymphoma (diffuse mixed, diffuse large- delivered on an ambulatory basis. Chemotherapy cell, or immunoblastic according to the Working alone consisted of three cycles of ACVBP6 (75 mg Formulation,8 and anaplastic according to the up- of doxorubicin per square meter of body-surface dated Kiel classification9), and to have no adverse area and 1200 mg of cyclophosphamide per square prognostic factors according to the age-adjusted meter on day 1, 2 mg of vindesine per square meter International Prognostic Index.7 Exclusion criteria and 10 mg of bleomycin on days 1 and 5, and 60 mg included positive serologic tests for human immu- of prednisone per square meter on days 1 through nodeficiency virus or human T-cell lymphotropic 5) given at two-week intervals and followed by se- virus type 1, transformation of a previously indolent quential consolidation consisting of two cycles of lymphoma, primary cerebral lymphoma, previous methotrexate (3 g per square meter) plus leucovorin organ transplantation, concomitant or previous rescue, four cycles of etoposide (300 mg per square cancer (except in situ cervical carcinoma), liver or meter) and ifosfamide (1500 mg per square meter),

1198 n engl j med 352;12 www.nejm.org march 24, 2005

chemotherapy for localized aggressive lymphoma and two cycles of cytarabine (100 mg per square six months, the data and safety monitoring commit- meter) subcutaneously for four days given at two- tee recommended that the sample size be calculated week intervals.6 on the assumption of an absolute improvement of Chemoradiotherapy consisted of three cycles of 7 percent in the two-year rate of event-free survival CHOP (50 mg of doxorubicin per square meter, and thus that the total enrollment be increased to 750 mg of cyclophosphamide per square meter, 600 patients. Secondary end points were the rate of 1.4 mg of vincristine per square meter [up to a max- response and overall survival. imal dose of 2 mg] on day 1 and 60 mg of predni- Analyses included all the enrolled patients and sone per square meter on days 1 through 5) repeat- followed the intention-to-treat principle. They were ed at 21-day intervals. Involved-field radiotherapy also performed on patients who met the eligibility began one month after the last cycle of CHOP. The criteria, after histologic review, and after the exclu- prescribed dose of radiation was 40 Gy in 22 frac- sion of those with an erroneous International Prog- tions of 1.8 Gy, five days per week. Irradiated vol- nostic Index score (Table 1). Patients’ characteristics umes encompassed involved nodal or extranodal and response rates were compared by means of the sites and adjacent uninvolved nodes. chi-square and Fisher’s exact tests. Event-free sur- In neither group was adjustment of the dose of vival was measured from the date of randomization chemotherapy planned in response to specific ad- to primary treatment failure, relapse, or death from verse events, but courses were postponed until leu- any cause or the stopping date, which for this analy- kocyte and platelet counts increased to greater than sis was February 1, 2004. Overall survival was mea- 2000 and 100,000 per cubic millimeter, respectively. sured from the date of randomization to the date of Patients could receive granulocyte colony-stimu- death from any cause or the stopping date. Data lating factor at the discretion of each investigator. were censored on the date of the last follow-up evaluation when the stopping date was not reached. assessment of response Survival rates were estimated according to the Kap- Response was evaluated one month after the com- lan–Meier method12 and compared with the use of pletion of treatment, according to the International the log-rank test.13 Taking into account the stratifi- Workshop criteria.11 A complete response was de- cation according to bulky-disease status, we used a fined as the disappearance of all clinical evidence stratified log-rank test to analyze event-free and of disease and of radiologic abnormalities observed overall survival. Multivariate analyses were per- at diagnosis. An unconfirmed complete response formed with the use of the Cox model for survival was defined as a complete response with some per- data.14 Differences were considered significant sisting radiologic abnormalities, which in the ag- when the two-sided P value was less than 0.05. All gregate had to be at least 75 percent smaller than statistical analyses were performed with SAS soft- the original abnormality. A partial response was de- ware, version 8.0. fined as the regression of tumor volumes by more The study was designed by the GELA scientific than 50 percent, and stable disease was defined as a committee. Data collected at participating centers lesser response. Progressive disease (growth of the were checked by GELA research assistants and sent initial lesion by more than 25 percent or the appear- to the centralized database in Creteil. One investi- ance of a new lesion) during treatment was consid- gator analyzed the data and was the principal writ- ered to indicate primary treatment failure. er of this article. statistical analysis results Randomization was stratified according to center and the presence or absence of bulky disease. The patients’ characteristics main objective of the trial was to detect an absolute Between March 1993 and June 2000, 647 patients improvement of 10 percent with chemotherapy were enrolled at 86 participating centers; 318 were alone given a two-year rate of event-free survival of assigned to chemotherapy alone with ACVBP and 75 percent in the group assigned to CHOP plus ra- 329 to chemoradiotherapy consisting of CHOP diotherapy (with a type I error of 0.05 and a type II plus involved-field radiotherapy. The main charac- error of 0.10). This design required the enrollment teristics of the patients were similar in the two of 400 patients over a period of four years. After a groups (Table 1). Diffuse large B-cell lymphoma planned interim analysis performed after 200 pa- was the most common subtype of lymphoma. Ex- tients had been enrolled and monitored for at least tranodal involvement was found in 49 percent of

n engl j med 352;12 www.nejm.org march 24, 2005 1199

The new england journal of medicine

Table 1. Characteristics of the 647 Patients.* ACVBP CHOP plus Characteristic (N=318) Radiotherapy (N=329) Median age — yr 46 47 Male sex — no. (%) 190 (60) 211 (64) Bulky disease at randomization — no. (%) 32 (10) 41 (12) HIV positive — no. (%)† 1 (<1) 0 Stage‡ I 209 (67) 215 (66) II 100 (32) 104 (32) IV† 3 (1) 7 (2) Lactate dehydrogenase level — no. (%)‡ Normal 302 (97) 317 (97) Elevated† 10 (3) 9 (3) Performance status — no. (%)‡ 0 252 (81) 251 (77) 1 59 (19) 74 (23) 2† 1 (<1) 1 (<1) Age-adjusted International Prognostic Index scores — no. (%)‡ 0 298 (96) 310 (95) 1† 14 (4) 16 (5) Extranodal involvement — no. (%)‡ 148 (47) 169 (52) With regional-node involvement 52 (17) 52 (16) Without regional-node involvement 96 (31) 117 (36) Organ involved — no. (%) Waldeyer’s ring and sinus 74 (24) 85 (26) Stomach 19 (6) 20 (6) Bone 11 (4) 12 (4) Skin 5 (2) 16 (5) Breast 9 (3) 7 (2) Thyroid 6 (2) 5 (2) Other§ 24 (8) 24 (7) Histologic findings not centrally reviewed — no. (%) 35 (11) 28 (9) Histologic findings centrally reviewed — no./total no.(%)¶ 283/318 (89) 301/329 (91) Eligible 261/283 (92) 275/301 (91) Diffuse large-B-cell lymphoma 235/283 (83) 239/301 (79) Anaplastic large-cell lymphoma 11/283 (4) 18/301 (6) Nonanaplastic T-cell or NK-cell lymphoma 8/283 (3) 15/301 (5) Unclassified aggressive lymphoma 7/283 (2) 3/301 (1) Inappropriate† 22/283 (8) 26/301 (9) Small lymphocytic lymphoma 1/283 (<1) 0 Marginal-zone lymphoma 1/283 (<1) 4/301 (1) Follicular lymphoma 7/283 (2) 8/301 (3) Mantle-cell lymphoma 1/283 (<1) 1/301 (<1) Lymphoblastic lymphoma 1/283 (<1) 1/301 (<1) Burkitt’s lymphoma 3/283 (1) 3/301 (1) Hodgkin’s lymphoma 4/283 (1) 4/301 (1) Nonmalignant 4/283 (1) 5/301 (2)

* Results are given on an intention-to-treat basis. HIV denotes human immunodeficiency virus, and NK natural killer. † This was an exclusion criterion. ‡ Information was available for 638 patients (99 percent): 312 in the ACVBP group and 326 in the group given CHOP plus radiotherapy. § Other categories include muscle, testis, orbit, kidney, pancreas, and adrenal gland. ¶The classification of the World Health Organization was used.

1200 n engl j med 352;12 www.nejm.org march 24, 2005

chemotherapy for localized aggressive lymphoma the patients. Bulky disease was present at randomization in 73 patients; mediastinal and abdominal Table 2. Outcome of Treatment at One Month in the 630 Patients Who Could Be Evaluated.* involvement accounted for 34 percent and 16 percent of these cases, respectively. ACVBP CHOP plus Radiotherapy Outcome (N=309) (N=321) response to treatment no. of patients (%) A complete or unconfirmed complete response oc- Complete or unconfirmed 288 (93) 294 (92) complete response curred in 93 percent of the patients treated with Partial response 6 (2) 4 (1) chemotherapy alone and 92 percent of those treat- Stable disease 4 (1) 6 (2) ed with chemoradiotherapy (Table 2). Eighty per- Primary treatment failure 11 (4) 17 (5) cent of the patients in the ACVBP group received at Death 0 0 least 80 percent of the planned dose of doxorubicin and cyclophosphamide, and 98 percent did so in * Response was assessed one month after the completion of treatment. Re- the group given CHOP plus radiotherapy. In the lat- sponse could not be assessed in nine patients in the ACVBP group and eight ter group, 26 patients did not receive the planned patients in the group given CHOP plus radiotherapy (data were missing for six and three patients, respectively, and treatment was stopped by the investiga- radiotherapy, as a result of a lack of response to tor without evaluation of the tumor in three and five patients, respectively). CHOP in 15, medical decision in 6, and refusal in 5. Among the remaining patients, 93 percent received a dose of 36 to 40 Gy. was 21 months and relapses involved the initial site There were no treatment-related deaths. Thirty- of disease in 41 percent of patients, a distant site in four episodes of grade 3 infection (11 percent of 38 percent, and both initial and distant sites in 21 patients) and two episodes of grade 4 infection percent. In the chemoradiotherapy group, the me- (1 percent) occurred in the chemotherapy group, dian time to relapse was 15 months and relapses as compared with four episodes of grade 3 infec- occurred in the irradiation field in 23 percent of the tion in the chemoradiotherapy group (1 percent). patients, outside the field in 72 percent, and both No life-threatening acute adverse effects of radio- within and outside the field in 5 percent. therapy were reported. There were 115 deaths, 40 in the chemotherapy group and 75 in the chemoradiotherapy group. outcome Overall survival differed significantly (P=0.001), During a median follow-up of 7.7 years, there were with five-year estimates of 90 percent in the che- 168 events (primary treatment failure, relapse, or motherapy group (95 percent confidence interval, death): 61 in the chemotherapy group and 107 in the 87 to 93 percent) and 81 percent in the chemora- chemoradiotherapy group. Event-free survival dif- diotherapy group (95 percent confidence interval, fered significantly between the groups (P<0.001), 77 to 86 percent) (Fig. 3). The difference remained with five-year estimates of 82 percent in the che- significant when the patients without bulky dis- motherapy group (95 percent confidence interval, ease (P=0.01) (Fig. 4) and those with bulky disease 78 to 87 percent) and 74 percent in the chemora- (P=0.03) were analyzed separately. In a multivari- diotherapy group (95 percent confidence interval, ate analysis of the 647 patients, survival was ad- 69 to 78 percent) (Fig. 1). The difference remained versely affected by the presence of bulky disease significant in separate analyses of the 574 patients (P<0.001; risk ratio, 3.0; 95 percent confidence in- without bulky disease (Fig. 2) and the 73 patients terval, 2.0 to 4.8) and treatment with chemoradio- with bulky disease (P=0.002 and P=0.04, respec- therapy (P=0.002; risk ratio, 1.8; 95 percent confi- tively). In a multivariate analysis of the 647 patients, dence interval, 1.3 to 2.7), but not by the presence event-free survival was independently affected by the of stage II disease (P=0.2; risk ratio, 1.3; 95 per- presence of bulky disease (P<0.001; risk ratio, 2.5; cent confidence interval, 0.9 to 1.9). 95 percent confidence interval, 1.7 to 3.7), treatment The influence of chemotherapy alone remained with chemoradiotherapy (P<0.001; risk ratio, 1.8; significant with respect to both event-free and 95 percent confidence interval, 1.3 to 2.5), and the overall survival when multivariate analysis was re- presence of stage II disease (P=0.03; risk ratio, 1.4; stricted to the patients who met eligibility criteria 95 percent confidence interval, 1 to 1.9). (Table 1) (P=0.001 and P=0.004, respectively), as There were 120 relapses, 42 in the chemothera- well as to those with diffuse large-B-cell lymphoma py group and 78 in the chemoradiotherapy group. (P=0.004 and P=0.03, respectively). In the former group, the median time to relapse Of the 115 deaths, 89 were related to progres-

n engl j med 352;12 www.nejm.org march 24, 2005 1201

The new england journal of medicine

100 90 ACVBP 80 70 60 CHOP plus radiotherapy 50 40 30 20 P<0.001 10 Probability of Event-free Survival (%) 0 0 1 2 3 4 5 6 7 8 9 10 11 Years after Randomization No. at Risk ACVBP 318 287 278 270 253 217 190 149 117 77 33 CHOP plus 329 275 262 252 243 209 179 133 94 59 22 radiotherapy

Figure 1. Event-free Survival among 647 Patients Assigned to Chemotherapy Alone with the ACVBP Regimen or to CHOP plus Involved-Field Radiotherapy.

sive lymphoma, 29 in the chemotherapy group and chemotherapy alone (three cycles of ACVBP fol- 60 in the chemoradiotherapy group. Sixteen deaths lowed by sequential consolidation) with three cy- occurred as a result of a second cancer that devel- cles of CHOP followed by involved-field radiother- oped after enrollment: seven of these were in the apy. The latter has been considered standard chemotherapy group, including one acute myelog- therapy for localized lymphoma since Miller et al. enous leukemia; nine occurred in the chemoradio- reported the results of a randomized study of 400 therapy group, including two cases of myelodyspla- patients, with a median follow-up of 4.4 years.5 Af- sia and one pancreatic cancer within the irradiation ter a median follow-up of 7.7 years, we found supe- field. At the time of death, 15 of these 16 patients rior event-free and overall survival rates among pa- with a second cancer were in a primary remission tients treated with chemotherapy alone. from their lymphoma and 1 was in a secondary re- The ACVBP regimen consists of an induction mission. Death was related to salvage treatment for phase with higher doses of doxorubicin and cyclo- relapsed lymphoma in four patients. Other causes phosphamide than those used in CHOP and a con- of death in patients in a primary remission includ- solidation phase consisting of treatment with drugs ed suicide in two, a car accident in one, myocardial not used during induction. The ACVBP regimen infarction in one (outside the irradiation field), and improves survival among elderly patients with poor- stroke in one; one patient in secondary remission risk aggressive lymphoma, as compared with eight died from peritonitis. cycles of standard CHOP.15 Two other studies have Nonfatal late effects of radiotherapy included suggested that the efficacy of CHOP can be im- persistent xerostomia in 6 percent of patients with proved by increasing the doses of doxorubicin and Waldeyer’s-ring lymphoma, hypothyroidism (five cyclophosphamide.16,17 In the present trial, the in- cases), carotid stenosis (two cases), gastritis (three creased doses and reduced intervals between the cases), and vertebral fracture (two cases). No epi- three courses of ACVBP increased the theoretical sodes of congestive heart failure were reported in dose intensity of doxorubicin and cyclophospha- either treatment group. mide by a factor of 2.25 and 2.4, respectively, as compared with three cycles of CHOP. As a consequence, discussion in 95 percent of the patients in the chemotherapy group, the theoretical dose intensity was at least 150 This randomized trial, which enrolled 647 patients percent of that delivered by three cycles of CHOP. with low-risk aggressive lymphoma in stage I or II We defined our study population of patients who were younger than 61 years of age, compared younger than 61 years with localized lymphoma on

1202 n engl j med 352;12 www.nejm.org march 24, 2005

chemotherapy for localized aggressive lymphoma

100 90 ACVBP (n=286) 80 70 60 CHOP plus radiotherapy (n=288) 50 40 30 20 P=0.002 10 Probability of Event-free Survival (%) 0 0 1 2 3 4 5 6 7 8 9 10 11 Years after Randomization

Figure 2. Event-free Survival among 574 Patients without Bulky Tumor Assigned to Chemotherapy Alone with the ACVBP Regimen or to CHOP plus Involved-Field Radiotherapy.

100 90 ACVBP 80 70 CHOP plus radiotherapy 60 50 40 30 20 P=0.001

Probability of Overall Survival (%) 10 0 0 1 2 3 4 5 6 7 8 9 10 11 Years after Randomization No. at Risk ACVBP 318 307 296 286 271 236 206 163 126 80 34 CHOP plus 329 314 292 280 265 231 199 152 113 70 28 radiotherapy

Figure 3. Overall Survival among 647 Patients Assigned to Chemotherapy Alone with the ACVBP Regimen or to CHOP plus Involved-Field Radiotherapy. the basis of the age-adjusted International Prog- In our patients with limited disease, the influ- nostic Index, the most widely used system to strat- ence of treatment group was independent of clas- ify tumors before therapy,7 thus providing a rela- sic prognostic factors such as stage II disease5 and tively homogeneous cohort of patients with regard a large tumor burden.19 Thus, the advantage of to principal prognostic factors: age, lactate dehy- chemotherapy alone over chemoradiotherapy ap- drogenase level, and performance status. By con- plied to the entire population of patients — not trast, in the study by Miller et al.,5 half the patients only to the patients with bulky disease, a condition were over 60 years of age and 20 percent had an el- in which adjuvant radiotherapy is believed to result evated lactate dehydrogenase level. An update of in optimal control of local disease,20,21 but also to that study with a longer follow-up showed conver- patients without a bulky tumor. In our population gence of survival curves as a result of an excess of of patients younger than 61 years, the ACVBP regi- relapses and deaths from lymphoma in the group men could be given on an ambulatory basis. As a given CHOP plus radiotherapy.18 consequence of the high dose intensity of ACVBP,

n engl j med 352;12 www.nejm.org march 24, 2005 1203

The new england journal of medicine

100 90 ACVBP (n=286) 80 70 CHOP plus radiotherapy (n=288) 60 50 40 30 20 P=0.01

Probability of Overall Survival (%) 10 0 0 1 2 3 4 5 6 7 8 9 10 11 Years after Randomization

Figure 4. Overall Survival among 574 Patients without Bulky Tumor Assigned to Chemotherapy Alone with the ACVBP Regimen or to CHOP plus Involved-Field Radiotherapy.

neutropenia was more frequent than with CHOP, Thus, moving from CHOP plus radiotherapy to in- but in our patients, we observed only two grade 4 tensified chemotherapy alone did not increase the life-threatening infections among 309 patients and number of deaths from second cancers. no treatment-related deaths. In our previous study In conclusion, chemotherapy with ACVBP alone of 2210 patients who received ACVBP, an age of significantly improved event-free and overall sur- more than 60 years, elevated lactate dehydrogenase vival among persons younger than 61 years with levels, and an Eastern Cooperative Oncology newly diagnosed, low-risk, localized aggressive Group score of more than 1 predicted early treat- lymphoma, as compared with the standard treatment-related deaths.22 ment with CHOP plus radiotherapy. Efforts are After a median follow-up of 7.7 years, there were now needed to further improve event-free survival a similar number of fatal second cancers in each among such patients. Given the benefit of the com- group. There were nine cases in the chemoradio- bination of rituximab and chemotherapy,27 the therapy group, which is in keeping with other re- GELA has undertaken a trial of rituximab plus the ports of a moderate risk of second cancer after brief ACVBP regimen in young adults with localized CHOP chemotherapy plus involved-field radiother- low-risk aggressive lymphoma. apy.23,24 A longer follow-up may, however, reveal Supported by a grant (AOM95061) from the Programme Hospi- additional late solid tumors related to radiothera- talier de Recherche Clinique du Ministère de la Santé and grants 25 from Amgen, Roche, Schering-Plough, and Astra-Medica. py. There were seven cases of fatal second cancers We are indebted to Isabelle Gaillard, Denis Collet, and Chafika in the ACVBP group, a number consistent with our Coppeaux for data management; to Nicolas Nio for statistical analy- retrospective analysis of 2837 patients treated with ses; and to Corinne Haioun for invaluable discussion. the ACVBP regimen in consecutive GELA trials.26

appendix The following persons and study centers participated in the GELA study (all in France, unless otherwise specified): Hôpital Saint-Antoine, Paris — M. Aoudjane; Clinique du Mail, Grenoble — D. Assouline; Hôpital Pasteur, Colmar — B. Audhuy; Centre Hospitalier de Saint Ger- main, Saint Germain en Laye — M. Azagury; Hôpital de Bayonne, Bayonne — F. Bauduer; Centre Medico-Chirurgical de Foch, Suresnes — E. Baumelou; Centre Hospitalier de Montluçon, Montluçon — M.A. Bichoffe; Centre Léon Berard, Lyon — P. Biron; Centre Hospitalier de Chambery, Chambery — M. Blanc; Hôpital Dupuytren, Limoges — D. Bordessoulle; Cliniques Universitaires de Mont Godinne, Yvoir, Bel- gium — A. Bosly; Institut Paoli Calmette, Marseille — R. Bouabdallah; Centre Hospitalier de Saint Brieuc, Saint Brieuc — P. Bourquard; Hôpital Saint-Louis, Paris — P. Brice, N. Mounier, D. Simon, C. Gisselbrecht, H. Dombret; Hôpital Beaujon, Clichy — J. Brière; Centre Hos- pitalo-Universitaire de Dijon, Dijon — D. Caillot, O. Casasnovas, M. Flesch, B. Chauffert; Centre Hospitalo-Universitaire de Nice, Nice — G.P. Cassuto; Hôpital André Mignot, Le Chesnay — S. Castaigne; Hôpital Bon Secours, Metz — B. Christian; Hôpital Lyon-Sud, Pierre- Bénite — B. Coiffier, G. Salles; Polyclinique de Courlancy, Reims — P. Colin and G. Pinon-Netter; Centre A. Vautrin, Vandoeuvre lès Nancy — T. Conroy; Centre Jean Perrin, Clermont-Ferrand — H. Curé; Hôpital Hôtel-Dieu, Paris — A. Delmer; Hôpital Jean Bernard, Poitiers — V. Del- wail; Centre Hospitalier de Corbeil, Corbeil Essonnes — A. Devidas; Hôpital Cochin, Paris — F. Dreyfus; Clinique de Chaumont, Chaumont — G. Dupont; Centre Hospitalier E. Muller, Mulhouse — J.C. Eisenmann; Centre Val d’Aurelle, Montpellier — M. Fabbro; Centre de Bligny, Briis sous Forges — C. Fermé; Centre Hospitalo-Universitaire de Liège, Liège, Belgium — G. Fillet; Hôpital J. Monod, Le Havre — C. Fru-

1204 n engl j med 352;12 www.nejm.org march 24, 2005

chemotherapy for localized aggressive lymphoma

chart; Hôpital de la Pitié-Salpêtrière, Paris — J. Gabarre; Hôpital du Creusot, Le Creusot — M. Gabez; Centre Hospitalier de Juvisy, Juvisy sur Orge — M. Gautier; Hôpital Henri Mondor, Créteil — C. Haioun, M. Diviné, K. Belhadj, F. Reyes; Hôpital de Hautepierre, Strasbourg — R. Herbrecht; Hôpital Necker, Paris — O. Hermine, B. Varet; Centre René Huguenin, Saint-Cloud — M. Janvier; Centre Hospitalo-Universitaire de Nîmes, Nîmes — E. Jourdan; Hôpital de l’Ouest Parisien, Trappes — D. Kamioner; Centre Hospitalier René Dubos, Pontoise — Y. Kerneis, F. Morvan; Centre Hospitalo-Universitaire Brabois, Vandoeuvre lès Nancy — P. Lederlin; Centre Hospitalier de Brives, Brives La Gaillarde — S. Lefort; Centre Hospitalier de la Durance, Avignon — G. Lepeu; Hôpital Tenon, Paris — J.P. Lotz, C. Bouleuc; Hôpital Paul Brousse, Villejuif — D. Machover; Hôpital de Chaneaux, Macon — F. Marechal; Centre F. Magendie, Bordeaux — G. Marit; Centre Hospitalier d’Annecy, An- necy — C. Martin; Centre Hospitalier d’Esch sur Azette, Esch sur Azette, Luxembourg — S. Meyer; Centre Hospitalier de Lens, Lens — P. Mo- rel, B. Dupriez; Institut Gustave Roussy, Villejuif — J.N. Munck, C. Fermé, P. Carde; Hôpital Universitaire de Gent, Ghent, Belgium — F. Offner; Centre Hospitalier de Troyes, Troyes — J.M. Pavlovitch; Hôpital V. Provo, Roubaix — I. Plantier; Hôpital de Fleyriat, Bourg en Bresse — H. Orfeuvre; Hôpital de Valence, Valence — P.Y. Péaud, D. Dramais; Centre François Baclesse, Caen — A.M. Peny; Hôpital de Thionville, Thionville — C. Platini; Hôpital Victor Dupuy, Argenteuil — M. Pullik; Centre Hospitalo-Universitaire de Lille, Lille — B. Quesnel, F. Mor- schausher; Centre Hospitalo-Universitaire de Caen, Caen — O. Reman, M. Leporrier; Centre Hospitalier Saint-Vincent, Lille — C. Rose; Cen- tre Hospitalier de Blois, Blois — P. Rodon; Centre Hospitalo-Universitaire de Montpellier, Montpellier — J.F. Rossi; Hôpital de Chalon, Cha- lon sur Saône — B. Salles; Hôpital de Purpan, Toulouse — D. Schlaifer, F. Huguet, A. Huyn; Hôpital Edouard Herriot, Lyon — C. Sebban; Centre Hospitalier de Valenciennes, Valenciennes — M. Simon; Clinique Victor Hugo, Le Mans — P. Solal-Céligny; Centre Hospitalo-Univer- sitaire de Grenoble, Grenoble — J.J. Sotto; Centre Hospitalier de Meaux, Meaux — C. Soussain, C. Allard; Hôpital Bicêtre, Paris — G. Tertian; Centre H. Becquerel, Rouen — H. Tilly; Centre A. Lacassagne, Nice — A. Thyss; Centre Hospitalo-Universitaire d’Amiens, Amiens — C. Traulle; Hôpital Hotel-Dieu, Clermont-Ferrand — P. Travade; Academisch Ziekenhuis Sint-Jan, Bruges, Belgium — A. Van Hoof; Hôpital Lariboisière, Paris — J.M. Zini; Pathologists — J. Brière, J. Diebold, B. Fabiani, P. Gaulard, C. Guettier, T. Molina, T. Petrella. references 1. Chen MG, Prosnitz LR, Gonzalez-Serva of World Health Organization classification Philadelphia: Lippincott Williams & Wilkins, A, Fischer DB. Results of radiotherapy in of tumours. Lyon, France: IARC Press, 2001. 2004:97-125. control of stage I and II non-Hodgkin’s lym- 11. Cheson BD, Horning SJ, Coiffier B, et al. 20. Zinzani PL, Martelli M, Magagnoli M, phoma. Cancer 1979;43:1245-54. Report of an international workshop to stan- et al. Treatment and clinical management 2. Connors JM, Klimo P, Fairey RN, Voss dardize response criteria for non-Hodgkin’s of primary mediastinal large B-cell lym- N. Brief chemotherapy and involved field ra- lymphomas. J Clin Oncol 1999;17:1244. phoma with sclerosis: MACOP-B regimen diation therapy for limited-stage, histologi- [Erratum, J Clin Oncol 2000;18:2351.] and mediastinal radiotherapy monitored cally aggressive lymphoma. Ann Intern Med 12. Kaplan EL, Meier P. Nonparametric esti- by (67)Gallium scan in 50 patients. Blood 1987;107:25-30. mation from incomplete observations. J Am 1999;94:3289-93. 3. Cabanillas F, Bodey GP, Freireich EJ. Stat Assoc 1958;53:457-81. 21. Connors J. Principles of chemotherapy Management with chemotherapy only of 13. Mantel N. Evaluation of survival data and combined modality therapy. In: Mauch stage I and II malignant lymphoma of ag- and two new rank order statistics arising in PM, Armitage J, Coiffier B, Dalla-Favera R, gressive histologic types. Cancer 1980;46: its consideration. Cancer Chemother Rep Harris N, eds. Non-Hodgkin’s lymphomas. 2356-9. 1966;50:163-70. Philadelphia: Lippincott Williams & Wilkins, 4. Jones SE, Miller TP, Connors JM. Long- 14. Cox DR. Regression model and life- 2004:201-19. term follow-up and analysis for prognostic tables. J R Stat Soc [B] 1972;34:187-220. 22. Dumontet C, Mounier N, Munck JN, et factors for patients with limited-stage dif- 15. Tilly H, Lepage E, Coiffier B, et al. Inten- al. Factors predictive of early death in pa- fuse large-cell lymphoma treated with initial sive conventional chemotherapy (ACVBP tients receiving high-dose CHOP (ACVBP chemotherapy with or without adjuvant ra- regimen) compared with standard CHOP regimen) for aggressive non-Hodgkin’s diotherapy. J Clin Oncol 1989;7:1186-91. for poor-prognosis aggressive non-Hodg- lymphoma: a GELA study. Br J Haematol 5. Miller TP, Dahlberg S, Cassady JR, et al. kin lymphoma. Blood 2003;102:4284-9. 2002;118:210-7. Chemotherapy alone compared with che- 16. Blayney DW, LeBlanc ML, Grogan T, et 23. Shenkier TN, Voss N, Fairey R, et al. motherapy plus radiotherapy for localized inal. Dose-intense chemotherapy every 2 weeks Brief chemotherapy and involved-region ir- termediate- and high-grade non-Hodgkin’s with dose-intense cyclophosphamide, dox- radiation for limited-stage diffuse large-cell lymphoma. N Engl J Med 1998;339:21-6. orubicin, vincristine, and prednisone may lymphoma: an 18-year experience from the 6. Tilly H, Mounier N, Lederlin P, et al. improve survival in intermediate- and high- British Columbia Cancer Agency. J Clin On- Randomized comparison of ACVBP and grade lymphoma: a phase II study of the col 2002;20:197-204. m-BACOD in the treatment of patients with Southwest Oncology Group (SWOG 9349). 24. Armitage JO, Carbone PP, Connors JM, low-risk aggressive lymphoma: the LNH87- J Clin Oncol 2003;21:2466-73. Levine A, Bennett JM, Kroll S. Treatment- 1 study. J Clin Oncol 2000;18:1309-15. 17. Pfreundschuh M, Truemper L, Kloess related myelodysplasia and acute leukemia 7. The International Non-Hodgkin’s Lym- M, et al. Two-weekly or 3-weekly CHOP che- in non-Hodgkin’s lymphoma patients. J Clin phoma Prognostic Factors Project. A predic- motherapy with or without etoposide for the Oncol 2003;21:897-906. tive model for aggressive non-Hodgkin’s treatment of young patients with good- 25. Aleman BM, Raemaekers JM, Tirelli U, lymphoma. N Engl J Med 1993;329:987-94. prognosis (normal LDH) aggressive lym- et al. Involved-field radiotherapy for ad- 8. The Non–Hodgkin’s Lymphoma Clas- phomas: results of the NHL-B1 trial of the vanced Hodgkin’s lymphoma. N Engl J Med sification Project. National Cancer Institute DSHNHL. Blood 2004;104:626-33. 2003;348:2396-406. sponsored study of classifications of non- 18. Miller TP, Leblanc M, Spier CM, Chase 26. Andre M, Mounier N, Leleu X, et al. Sec- Hodgkin’s lymphomas: summary and de- E, Fischer RI. CHOP alone compared to ond cancers and late toxicities after treatment scription of a working formulation for clini- CHOP plus radiotherapy for early stage ag- of aggressive non-Hodgkin lymphoma with cal usage. Cancer 1982;49:2112-35. gressive non-Hodgkin’s lymphomas: up- the ACVBP regimen: a GELA cohort study on 9. Stansfeld AG, Diebold J, Noel H, et al. date of the Southwest Oncology Group 2837 patients. Blood 2004;103:1222-8. Updated Kiel classification for lymphomas. (SWOG) Randomized Trial. Blood 2001;98: 27. Coiffier B, Lepage E, Briere J, et al. Lancet 1988;1:292-3. [Erratum, Lancet 742a-743a. abstract. CHOP chemotherapy plus rituximab com- 1988;1:372.] 19. Savage K, Shipp MA. Initial evaluation: pared with CHOP alone in elderly patients 10. Jaffe ES, Harris NL, Stein H, Vardiman J, staging and prognostic factors. In: Mauch with diffuse large-B-cell lymphoma. N Engl eds. Pathology and genetics of tumours of PM, Armitage J, Coiffier B, Dalla-Favera R, J Med 2002;346:235-42. haematopoietic and lymphoid tissues. Vol. 3 Harris N, eds. Non-Hodgkin’s lymphomas. Copyright © 2005 Massachusetts Medical Society.

n engl j med 352;12 www.nejm.org march 24, 2005 1205

The new england journal of medicine

original article

Day–Night Pattern of Sudden Death in Obstructive Sleep Apnea

Apoor S. Gami, M.D., Daniel E. Howard, B.S., Eric J. Olson, M.D., and Virend K. Somers, M.D., Ph.D.

abstract

background From the Divisions of Cardiovascular Dis- The risk of sudden death from cardiac causes in the general population peaks from eases (A.S.G., V.K.S.), Pulmonary and 6 a.m. to noon and has a nadir from midnight to 6 a.m. Obstructive sleep apnea is highly Critical Care Medicine (E.J.O.), and Hyper- tension (V.K.S.), Department of Internal prevalent and associated with neurohormonal and electrophysiological abnormalities Medicine (A.S.G., D.E.H., E.J.O., V.K.S.), that may increase the risk of sudden death from cardiac causes, especially during sleep. Mayo Clinic College of Medicine, Roches- ter, Minn. Address reprint requests to Dr. methods Somers at the Division of Cardiovascular Diseases, Mayo Clinic College of Medicine, We reviewed polysomnograms and the death certificates of 112 Minnesota residents 200 First St. SW, Rochester, MN 55905, or who had undergone polysomnography and had died suddenly from cardiac causes be- at [email protected]. tween July 1987 and July 2003. For four intervals of the day, we compared the rates of N Engl J Med 2005;352:1206-14. sudden death from cardiac causes among people with obstructive sleep apnea and the Copyright © 2005 Massachusetts Medical Society. following: the rates among people without obstructive sleep apnea, the rates in the general population, and the expectations according to chance. For each interval, we assessed the median apnea–hypopnea index and the relative risk of sudden death from cardiac causes. We similarly analyzed sudden death from cardiac causes during three time intervals that correlate with usual sleep–wake cycles.

results From midnight to 6 a.m., sudden death from cardiac causes occurred in 46 percent of people with obstructive sleep apnea, as compared with 21 percent of people without obstructive sleep apnea (P=0.01), 16 percent of the general population (P<0.001), and the 25 percent expected by chance (P<0.001). People with sudden death from cardiac causes from midnight to 6 a.m. had a significantly higher apnea–hypopnea index than those with sudden death from cardiac causes during other intervals, and the apnea– hypopnea index correlated directly with the relative risk of sudden death from cardiac causes from midnight to 6 a.m. For people with obstructive sleep apnea, the relative risk of sudden death from cardiac causes from midnight to 6 a.m. was 2.57 (95 percent confidence interval, 1.87 to 3.52). The analysis of usual sleep–wake cycles showed similar results.

conclusions People with obstructive sleep apnea have a peak in sudden death from cardiac causes during the sleeping hours, which contrasts strikingly with the nadir of sudden death from cardiac causes during this period in people without obstructive sleep apnea and in the general population.

1206 n engl j med 352;12 www.nejm.org march 24, 2005

sudden death from cardiac causes in obstructive sleep apnea

he risk of sudden death from car- board of the Mayo Foundation approved the study. t diac causes in the general population is We identified Minnesota residents who underwent significantly greater during the morning diagnostic polysomnography at the Mayo Clinic hours after waking (i.e., from 6 a.m. to noon) than Sleep Disorders Center from July 1, 1987, to July 31, during the other six-hour intervals of the day.1 Also, 2003, and confirmed their vital status (available there is a marked nadir in the risk of sudden death for 99.6 percent) with the use of records from the from cardiac causes during sleep (i.e., from mid- Minnesota Department of Health and the Mayo night to 6 a.m.).1 This same day–night pattern has Clinic. We then manually reviewed the deceased been shown for the incidence of sudden death from persons’ death certificates (available for 83.8 per- cardiac causes in patients with heart failure2 and cent) and included in our study those who died of for the incidence of acute myocardial infarction.1 sudden death from cardiac causes before July 31, The increase in risk in the morning may in part be 2003 (112 people). due to changes in sympathetic activity,3 baroreflex sensitivity,4 coagulability,5 and electrophysiologi- classification of sudden death cal abnormalities6-8 during the waking morning from cardiac causes hours, all of which may predispose persons to cardi- Manual review of death certificates and data from ac ischemia and fatal arrhythmias. the Minnesota Department of Health provided the Obstructive sleep apnea affects 17 to 24 per- immediate and underlying causes of death, time of cent of North American adults.9 In people with ob- death, and time interval from onset of symptoms structive sleep apnea, nocturnal repetitive episodes to death. A state nosologist used the methods of of acute apnea elicit hypoxemia, hypercapnia, in- the National Center for Health Statistics to assign creased sympathetic drive,10 surges in blood pres- causes of death. We classified the following causes sure,10 increases in cardiac-wall stress,11 and cardi- of death as sudden death from cardiac causes: “sud- ac arrhythmias.12-14 Obstructive sleep apnea is also den cardiac death,” “cardiac dysrhythmia,” “cardiac associated with hypercoagulability, vascular oxida- arrhythmia,” “cardiac arrest,” “cardiorespiratory ar- tive stress, systemic inflammation, and endothelial rest,” or “coronary heart disease” or “myocardial in- dysfunction.15 Thus, people with obstructive sleep farction” in which the interval from symptoms to apnea have severe perturbations of autonomic, he- death was an hour or less. modynamic, humoral, and vascular regulation dur- Persons were excluded if the time of sudden ing sleep that contrast with the physiology of nor- death from cardiac causes was not stated or if the mal sleep. It is not known whether these nocturnal death certificate provided information that explic- abnormalities are associated with an increased risk itly contradicted the following definition of sud- of sudden death from cardiac causes during the den death from cardiac causes: “natural death due night, a time when the risk of sudden death from to cardiac causes, heralded by abrupt loss of con- cardiac causes is at its lowest level for the general sciousness within one hour of the onset of acute population.1 symptoms; preexisting heart disease may have been We tested the hypothesis that people with ob- known to be present but the time and mode of structive sleep apnea are more likely to have sudden death are unexpected”16 and nontraumatic.16 The death from cardiac causes from midnight to 6 a.m. occurrence of sudden death from cardiac causes than during the other six-hour intervals of the day. during sleep was an exception to the criteria for We also tested the hypothesis that the frequency of “loss of consciousness within one hour of the acute sudden death from cardiac causes from midnight onset of symptoms,”16 since some instances of sud- to 6 a.m. is higher in people with obstructive sleep den death from cardiac causes occurring during apnea than in people without obstructive sleep ap- sleep may preclude awakening and the develop- nea, the general population, and what is expected ment of symptoms, or the witnessing of these events by chance (i.e., 25 percent per six-hour interval). by others. The death certificates of six people stated that sudden death from cardiac causes was un- methods witnessed or presumed, but their time of death was noted, and these people were included in the study. study population The death certificates of all other persons did not We included subjects who authorized their records state whether sudden death from cardiac causes to be used for research, and the institutional review was witnessed or unwitnessed. Thirty-six death

n engl j med 352;12 www.nejm.org march 24, 2005 1207

The new england journal of medicine

certificates (32 percent) provided specific infor- Fisher’s exact test. The proportion of people with mation about the time interval from symptoms to obstructive sleep apnea who had sudden death from death. cardiac causes during each time interval was compared, with the use of the binomial distribution, classification of obstructive sleep apnea with the proportion of people who had sudden death The sleep evaluations of all 112 persons were con- from cardiac causes during the same time interval ducted by a sleep specialist at the Mayo Clinic Sleep in a historical control population (from a meta- Disorder Center. We reviewed each person’s first analysis of 19,390 persons with sudden death from diagnostic polysomnogram, which included mea- cardiac causes in the general population)1 and with sures of the electroencephalogram, electro-ocu- the proportion that was expected by chance to have logram, electromyogram, electrocardiogram, tho- sudden death from cardiac causes during each quar- racoabdominal excursions, pulse oximetry, and ter of the day (i.e., 25 percent). naso-oral airflow. The apnea–hypopnea indexes of persons who An apnea was defined as cessation of airflow for had sudden death from cardiac causes during each at least 10 seconds in the presence of thoracoab- time interval were expressed as medians (with in- dominal ventilatory efforts, and a hypopnea as a re- terquartile ranges) and compared with the use of duction in airflow of at least 30 percent with a de- the Kruskal–Wallis test. For persons with and with- crease in oxygen saturation of 2 percent or more for out obstructive sleep apnea, we calculated the relat least 10 seconds in the presence of thoracoab- ative risk and 95 percent confidence interval of sud- dominal ventilatory efforts. The apnea–hypopnea den death from cardiac causes during each 6-hour index was calculated as the sum of apneas and interval as compared with the remaining 18 hours hypopneas per hour of sleep. According to Amer- of the day. ican Academy of Sleep Medicine criteria, an apnea– We also performed similar analyses of sudden hypopnea index of 5 or more established the diag- death from cardiac causes in persons with and nosis of obstructive sleep apnea.17 persons without obstructive sleep apnea for three We collected data on the demographics, coex- eight-hour intervals of the day (6 a.m. to 2 p.m., isting conditions, height, and weight of each per- 2 p.m. to 10 p.m., and 10 p.m. to 6 a.m.). These in- son at the time of polysomnography. We consid- tervals better represent usual sleep–wake cycles. ered that continuous positive airway pressure had been used if it was prescribed after the sleep study results (which was noted by the sleep-medicine physician reviewing the study results and plan of care) and patient characteristics if a subsequent note in the medical record stated Characteristics of the study population are de- that continuous positive airway pressure was be- scribed in Table 1. The diagnoses for the 34 persons ing used. The personnel who collected data from without obstructive sleep apnea were no sleep dis- polysomnograms and confirmed sleep diagnoses order (53 percent), central sleep apnea (23 percent), were different from those who collected data from periodic leg movement disorder (9 percent), obe- death certificates and confirmed death diagnoses, sity–hypoventilation syndrome (6 percent), narco- and the procedures were performed in a masked lepsy (6 percent), and hypoventilation due to neu- fashion. romuscular disease (3 percent). There was a higher proportion of men in the group with obstructive statistical analysis sleep apnea than in the group without obstructive Characteristics of the study population were de- sleep apnea, but the two groups were similar in scribed as means ±SD and counts (with percentag- terms of age, body-mass index, and coexisting con- es) and were compared according to the status of ditions. There were no differences in characteris- obstructive sleep apnea and the time of sudden tics, or reported use of continuous positive airway death from cardiac causes with the use of the two- pressure, between persons with sudden death from tailed t-test, Fisher’s exact test, and analysis of vari- cardiac causes during the various intervals of the day. ance. The frequency distribution of sudden death The use of continuous positive airway pressure was from cardiac causes for the four six-hour intervals reported in 15 percent of patients with an apnea– was compared between persons with and persons hypopnea index of 5 to 19, 39 percent of patients without obstructive sleep apnea with the use of with an apnea–hypopnea index of 20 to 39, and 58

1208 n engl j med 352;12 www.nejm.org march 24, 2005

sudden death from cardiac causes in obstructive sleep apnea

Table 1. Characteristics of the Study Population at the Time of Polysomnography, According to the Status of Obstructive Sleep Apnea (OSA) and the Time of Sudden Death from Cardiac Causes.*

Characteristic OSA Status Time of Death

Midnight– 6 a.m.– Noon– 6 p.m.– OSA No OSA P 5:59 a.m. 11:59 a.m. 5:59 p.m. 11:59 p.m. P (N=78) (N=34) Value (N=43) (N=30) (N=16) (N=23) Value† Age — yr 70±10 67±13 0.18 67±11 71±13 71±11 71±8 0.47 Male sex — no. (%) 64 (82) 21 (62) 0.02 33 (77) 24 (80) 11 (69) 17 (74) 0.85 Body-mass index‡ 34±7 33±9 0.71 35±8 32±7 33±11 33±5 0.59 Hypertension — no. (%) 51 (65) 22 (65) 0.94 27 (63) 20 (67) 8 (50) 18 (78) 0.32 Diabetes mellitus — no. (%) 32 (41) 12 (35) 0.57 20 (47) 6 (20) 9 (56) 9 (39) 0.06 Coronary artery disease — no. (%) 54 (69) 19 (56) 0.29 27 (63) 18 (60) 10 (63) 18 (78) 0.52 Congestive heart failure — no. (%) 43 (55) 18 (53) 0.83 24 (56) 13 (43) 10 (63) 14 (61) 0.51 Cerebrovascular disease — no. (%) 17 (22) 7 (21) 0.89 7 (16) 8 (27) 3 (19) 6 (26) 0.68 Dysrhythmia — no. (%)§ 14 (18) 3 (9) 0.22 6 (14) 3 (10) 5 (31) 3 (13) 0.27

* Plus–minus values are means ±SD. † The P value for time of death is for the comparison of groups according to time of death. ‡ The body-mass index is the weight in kilograms divided by the square of the height in meters. § A diagnosis of dysrhythmia includes a history of ventricular tachycardia, ventricular fibrillation, an implanted cardiac defibrillator, or resuscitation from cardiac arrest. percent of patients with an apnea–hypopnea index of 40 or more (P<0.001). OSA No OSA General population time distribution of sudden death P<0.001 P<0.001 from cardiac causes 50 P=0.01 P=0.04 P=0.02 Figure 1 shows the primary results. From midnight 46 to 6 a.m., the frequency of sudden death from car- 40 diac causes was significantly higher in persons with 41 obstructive sleep apnea than in persons without obstructive sleep apnea (46 percent vs. 21 percent, 30 30 29 P=0.01), than in the general population (46 per- 26 24 25 cent vs. 16 percent, P<0.001), and than was expect- 20 21 20 ed by chance (46 percent vs. 25 percent, P<0.001). 16 From 6 a.m. to noon, the frequency of sudden death from Cardiac Causes (%) 10 12 9 from cardiac causes was significantly lower in per- Proportion of Total Sudden Deaths sons with obstructive sleep apnea than in persons 0 without obstructive sleep apnea (20 percent vs. 41 Midnight– 6 a.m.– Noon– 6 p.m.– 5:59 a.m. 11:59 a.m. 5:59 p.m. 11:59 p.m. percent, P=0.04). From noon to 6 p.m., the frequency of sudden death from cardiac causes was Time of Day significantly lower in persons with obstructive sleep Figure 1. Day–Night Pattern of Sudden Death from Cardiac Causes apnea than in persons without obstructive sleep in 78 Persons with and 34 Persons without Obstructive Sleep Apnea (OSA) apnea (9 percent vs. 26 percent, P=0.02), than in and in the General Population. the general population (9 percent vs. 29 percent, Data for the general population were derived from Cohen et al.1 P<0.001), and than was expected by chance (9 percent vs. 25 percent, P<0.001). median apnea–hypopnea index than those with sud- severity of obstructive sleep apnea den death from cardiac causes from 6 a.m. to noon Persons with sudden death from cardiac causes (39 vs. 8, P<0.001) and noon to 6 p.m. (39 vs. 11, from midnight to 6 a.m. had a significantly higher P<0.001), and a similar median apnea–hypopnea

n engl j med 352;12 www.nejm.org march 24, 2005 1209

The new england journal of medicine

1.87 (95 percent confidence interval, 0.86 to 4.04) A for persons with mild-to-moderate obstructive sleep 140 apnea (apnea–hypopnea index, 5 to 39) and 2.61 (95 percent confidence interval, 1.27 to 5.38) for 120 persons with severe obstructive sleep apnea (apnea– 100 hypopnea index, ≥40) (Fig. 2B).

80 relative risk of sudden death from cardiac causes 60 For persons with obstructive sleep apnea, the rela- 40 39 41 tive risk of sudden death from cardiac causes from Apnea–Hypopnea Index midnight to 6 a.m. (as compared with the remain- 20 8 11 ing 18 hours of the day) was 2.57 (95 percent con- 0 fidence interval, 1.87 to 3.52), from 6 a.m. to noon Midnight– 6 a.m.– Noon– 6 p.m.– 5:59 a.m. 11:59 a.m. 5:59 p.m. 11:59 p.m. was 0.77 (95 percent confidence interval, 0.49 to 1.21), from noon to 6 p.m. was 0.30 (95 percent Time of Sudden Death from Cardiac Causes confidence interval, 0.15 to 0.61), and from 6 p.m. B to midnight was 0.97 (95 percent confidence inter- 5.00 val, 0.64 to 1.46). For persons without obstructive sleep apnea, the relative risk of sudden death from cardiac causes from midnight to 6 a.m. (as compared with the remaining 18 hours of the day) was 2.61 0.77 (95 percent confidence interval, 0.36 to 1.66), 1.87 from 6 a.m. to noon was 2.10 (95 percent confidence interval, 1.14 to 3.85), from noon to 6 p.m. was 1.08 (95 percent confidence interval, 0.54 to 1.00 2.16), and from 6 p.m. to midnight was 0.40 (95 percent confidence interval, 0.15 to 1.08) (Fig. 3). Relative Risk of Sudden Death from

Cardiac Causes from Midnight to 6 a.m. sleep–wake cycles 0.50 <5 5–39 ≥40 The sleep–wake cycles are described in Figure 4. Apnea–Hypopnea Index Included are the day–night pattern of sudden death from cardiac causes (Fig. 4A), the severity of ob- Figure 2. The Apnea–Hypopnea Index for Persons with Sudden Death from structive sleep apnea (Fig. 4B), and the relative risk Cardiac Causes during Six-Hour Intervals (Panel A) and the Relative Risk of of sudden death from cardiac causes during the Sudden Death from Cardiac Causes from Midnight to 6 a.m. for Persons three time intervals that represent usual sleep–wake with Mild-to-Moderate Obstructive Sleep Apnea and for Persons with Severe Obstructive Sleep Apnea (Panel B). cycles (Fig. 4C). In Panel A, the line within each box represents the median apnea–hypopnea index, and the box represents the interquartile range (25th percentile to 75th discussion percentile). Each black square represents one person. The figure includes data from persons with and from persons without obstructive sleep apnea (P<0.001). This study shows that persons with obstructive sleep In Panel B, the reference group consists of 34 persons without obstructive apnea have a significantly increased risk of sud- sleep apnea. There are 39 persons in the group with the apnea–hypopnea index of 5 to 39, and 39 persons in the group with the apnea–hypopnea index of den death from cardiac causes during the sleeping 40 or more. The squares represent the relative risk point estimates, and the hours, which is in striking contrast to the nadir of I bars the 95 percent confidence intervals. sudden death from cardiac causes during this time in persons without obstructive sleep apnea and in the general population. In the analysis of sudden index to that in persons with sudden death from car- death from cardiac causes from midnight to 6 a.m. diac causes from 6 p.m. to midnight (39 vs. 41) (Fig. (which allows for the comparison of our results to 2A). As compared with persons without obstruc- results of large studies in the general population) tive sleep apnea, the relative risk of sudden death and in the analysis of sudden death from cardiac from cardiac causes from midnight to 6 a.m. was causes from 10 p.m. to 6 a.m. (which is more rele-

1210 n engl j med 352;12 www.nejm.org march 24, 2005

sudden death from cardiac causes in obstructive sleep apnea vant to sleep-related pathophysiology), a marked 10.00 nocturnal peak in sudden death from cardiac causes OSA was observed in persons with obstructive sleep ap- No OSA nea. In more than half of persons with obstructive sleep apnea, sudden death from cardiac causes oc- 2.57 2.10 curred between 10 p.m. and 6 a.m. By contrast, the persons without obstructive sleep apnea in our 1.00 1.08 0.97 study sample had a day–night pattern of sudden 0.77 0.77 death from cardiac causes very similar to that in

from Cardiac Causes 0.40 the general population, with a peak in sudden death 0.30 from cardiac causes from 6 a.m. to noon. Relative Risk of Sudden Death Furthermore, our analysis showed that the sever- 0.10 ity of obstructive sleep apnea correlated directly Midnight– 6 a.m.– Noon– 6 p.m.– with the risk of nocturnal sudden death from cardi- 5:59 a.m. 11:59 a.m. 5:59 p.m. 11:59 p.m. ac causes, such that the relative risk of sudden Time of Day death from cardiac causes during the sleeping hours was 40 percent higher in persons with severe ob- Figure 3. Relative Risk of Sudden Death from Cardiac Causes during Six-Hour structive sleep apnea (apnea–hypopnea index, ≥40) Intervals as Compared with the Remaining 18 Hours of the Day, for 78 Persons than in those with mild-to-moderate obstructive with and 34 Persons without Obstructive Sleep Apnea (OSA). sleep apnea (apnea–hypopnea index, 5 to 39). The squares represent the point estimates of relative risk, and the I bars the The mean age at which sudden death from car- 95 percent confidence intervals. diac causes occurred in persons with obstructive sleep apnea was similar to the mean age of sudden death from cardiac causes in persons without ob- ably reflects the normal physiology of sleep, in structive sleep apnea and in the general popula- which sympathetic activity is decreased3 and cardi- tion.18 Our findings should not be generalized to ac dysrhythmias are uncommon.20 The unique day– younger persons with obstructive sleep apnea. Al- night pattern of sudden death from cardiac causes so, this finding suggests that obstructive sleep ap- that we observed in persons with obstructive sleep nea does not hasten sudden death from cardiac apnea is consistent with their exposure during sleep causes; however, our study cannot address wheth- to critical mechanisms that could promote sudden er obstructive sleep apnea increases the overall risk death from cardiac causes. of sudden death from cardiac causes. Obstructive sleep apnea is characterized by re- Our data may also provide insights into the con- petitive collapse of the pharyngeal airway during ventional understanding of the day–night pattern sleep, resulting in complete or partial cessation of of sudden death from cardiac causes in the general airflow, sometimes hundreds of times nightly. The population. First, obstructive sleep apnea may be resultant hypoxemia can lead to nocturnal cardiac implicated in some of the 16 percent of cases of sud- ischemia21 and ventricular arrhythmias.22 Apneic den death from cardiac causes that occur between episodes elicit increased sympathetic activity,10 midnight and 6 a.m. in the general population,1 blood-pressure elevations,10 and platelet aggrega- since obstructive sleep apnea affects at least 25 mil- tion.23 Obstructive sleep apnea is associated with lion adult Americans9 and remains undiagnosed in abnormalities in cardiac autonomic and electro- the vast majority of these people.19 Second, the risk physiological factors, including heart rate variabili- of sudden death from cardiac causes from 6 a.m. to ty,24 the duration of the QT interval,25 baroreflex noon in persons without obstructive sleep apnea function, and chemoreceptor sensitivity.26 Serious may be even greater than the risk that is currently and potentially fatal arrhythmias occur during sleep supported by data from the general population, in patients with obstructive sleep apnea12-14,22,27-31 since those data include people with and people and are attenuated by effective treatment.14,29-32 without obstructive sleep apnea, and we showed Thus, multiple pathophysiological mechanisms that the presence of obstructive sleep apnea is as- occur during sleep in persons with obstructive sociated with a lower risk during this period. sleep apnea and may explain an increased risk of The marked nocturnal nadir of sudden death nocturnal sudden death from cardiac causes. from cardiac causes in the general population prob- Two previous studies evaluated the day–night

n engl j med 352;12 www.nejm.org march 24, 2005 1211

The new england journal of medicine

Figure 4. Sudden Death from Cardiac Causes According to A Usual Sleep–Wake Cycles. OSA No OSA

Panel A shows day–night patterns of sudden death from 60 P=0.004 P=0.02 cardiac causes on the basis of eight-hour time intervals for 78 persons with and 34 persons without obstructive 50 54 sleep apnea. Panel B shows the apnea–hypopnea index 40 45 for persons with sudden death from cardiac causes dur- 30 ing eight-hour intervals. The line within each box repre- 32 sents the median apnea–hypopnea index, and the box 20 represents the interquartile range (25th percentile to 24 22 24 75th percentile). The figure includes data from persons 10 with and from persons without obstructive sleep apnea from Cardiac Causes (%) 0

(P=0.001) for the comparison of the apnea–hypopnea Proportion of Total Sudden Deaths 10 p.m.– 6 a.m.– 2 p.m.– index according to the time of sudden death. Panel C 5:59 a.m. 1:59 p.m. 9:59 p.m. shows the relative risk of sudden death from cardiac Time of Day causes during 8-hour intervals, as compared with the remaining 16 hours of the day, for 78 persons with and 34 persons without obstructive sleep apnea. The squares B represent the relative risk point estimates, and the I bars 140 the 95 percent confidence intervals. 120 100 pattern of sudden death from cardiac causes in per- 80 33,34 sons with sleep disorders. Neither study spe- 60 cifically examined obstructive sleep apnea. The first 40 41 study obtained histories of snoring from the co- 20 23 Apnea–Hypopnea Index habitants of 321 men who had sudden death from 8 0 cardiac causes. It found that habitual snorers were 10 p.m.– 6 a.m.– 2 p.m.– more likely than occasional snorers or nonsnorers 5:59 a.m. 1:59 p.m. 9:59 p.m. to have sudden death from cardiac causes during Time of Sudden Death from Cardiac Causes sleep and between 4 a.m. and 8 a.m.33 The second study reported that among 13 persons with sleep- C disordered breathing (not just obstructive sleep ap- 5.00 OSA nea) and sudden death from cardiac causes, none No OSA died during sleep.34 In contrast, our study assessed 2.33 1.58 the day–night pattern of sudden death from cardiac 1.00 causes in a large group of nonselected persons with 0.96

or without confirmed obstructive sleep apnea as 0.62 0.55 0.64 determined by the gold standard of diagnostic tests. from Cardiac Causes One limitation of the present study relates to the Relative Risk of Sudden Death 0.20 recognized difficulty of establishing the diagnosis 10 p.m.– 6 a.m.– 2 p.m.– and timing of sudden death from cardiac causes. 5:59 a.m. 1:59 p.m. 9:59 p.m. Our use of data from death certificates to identify Time of Day sudden death from cardiac causes has precedent in previous epidemiologic studies35,36 and has been validated in large populations,37-40 including our re- Our methods also improve on past studies that gional population.39,40 Furthermore, we used infor- searched databases of diagnostic codes to identify mation from the death certificates about the time in- potential cases, since we identified persons with terval from the onset of symptoms to death in order sudden death from cardiac causes by our direct re- to corroborate the diagnosis of sudden death from view of every available death record from the larger cardiac causes on the basis of current definitions.16 population. Qualifiers such as “witnessed,” “unwit- For example, we were able to exclude persons who nessed,” and “presumed” were infrequently noted were diagnosed with sudden death from cardiac on the death certificates. Therefore, we included all causes on the basis of old criteria that allowed a 24- cases of sudden death from cardiac causes regard- hour interval between symptoms and death. less of these statements. Thus, persons with sudden

1212 n engl j med 352;12 www.nejm.org march 24, 2005

sudden death from cardiac causes in obstructive sleep apnea death from cardiac causes during sleep, which may from cardiac causes increased with increasing se- preclude the development of symptoms or the wit- verity of obstructive sleep apnea despite more re- nessing of the event by others, were exceptions to ported use of continuous positive airway pressure the criteria that required timing of symptoms and with increasing severity of obstructive sleep ap- a change in consciousness, and they were included nea. Because of the lack of objective data from in the study when a time of death was noted. It is continuous-positive-airway-pressure devices, sub- important to note that we collected sleep data and jective data regarding their use were necessary. This death data independently of one another, and in a is an important limitation of retrospective assess- masked fashion, in order to avoid diagnostic suspi- ments of treatment of obstructive sleep apnea. cion bias. This study shows that persons with obstructive Another limitation of the study relates to the sleep apnea have a significant alteration in the ascertainment and interpretation of treatment with well-established day–night pattern of sudden death continuous positive airway pressure. During the from cardiac causes evident in the general popu- years of this study, continuous-positive-airway-pres- lation. Our study cannot address the question of sure devices that stored the time of use were rare- whether obstructive sleep apnea increases the overly used. The medical records lacked objective data all risk of sudden death from cardiac causes. Given regarding frequency, duration, or effectiveness of our findings and the relatively high prevalence of continuous positive airway pressure. Therefore, we obstructive sleep apnea in Western populations, could not verify whether continuous positive airway this is an important question that remains to be pressure was used in the days before sudden death answered. from cardiac causes or during the event. Supported by grants (HL61560, HL65176, HL73211, and M01 That continuous positive airway pressure may R-00585) from the National Institutes of Health, by the Dana Foun- dation, and by the Mayo Clinic College of Medicine. Dr. Gami is also have been generally ineffective is suggested by the supported by the Dr. Ralph and Marian C. Falk Medical Research finding that the risk of nocturnal sudden death Trust Fellowship for Clinical Research. references 1. Cohen MC, Rohtla KM, Lavery CE, Muller patients with coronary artery disease: ef- Cardiac rhythm disturbances in the obstruc- JE, Mittleman MA. Meta-analysis of the fects of arousal and upright posture. Circu- tive sleep apnea syndrome: effects of nasal morning excess of acute myocardial infarc- lation 1994;90:121-6. continuous positive airway pressure therapy. tion and sudden cardiac death. Am J Cardiol 8. Huikuri HH, Yli-Mayry S, Linnaluoto Chest 2000;118:591-5. 1997;79:1512-6. [Erratum, Am J Cardiol MK, Ikaheimo MJ. Diurnal fluctuations in 15. Shamsuzzaman AS, Gersh BJ, Somers 1998;81:260.] human ventricular and atrial refractori- VK. Obstructive sleep apnea: implications for 2. Moser DK, Stevenson WG, Woo MA, ness. Pacing Clin Electrophysiol 1995;18: cardiac and vascular disease. JAMA 2003; Stevenson LW. Timing of sudden death in 1362-8. 290:1906-14. patients with heart failure. J Am Coll Cardiol 9. Young T, Peppard PE, Gottlieb DJ. Epide- 16. Priori SG, Aliot E, Blomstrom-Lundqvist 1994;24:963-7. miology of obstructive sleep apnea: a popu- C, et al. Task Force on Sudden Cardiac Death 3. Somers V, Dyken ME, Mark AL, Abboud lation health perspective. Am J Respir Crit of the European Society of Cardiology. Eur FM. Sympathetic-nerve activity during sleep Care Med 2002;165:1217-39. Heart J 2001;22:1374-450. [Erratum, Eur in normal subjects. N Engl J Med 1993;328: 10. Somers VK, Dyken ME, Clary MP, Ab- Heart J 2002;23:257.] 303-7. boud FM. Sympathetic neural mechanisms in 17. Sleep-related breathing disorders in 4. Legramante JM, Marciani MG, Placidi F, obstructive sleep apnea. J Clin Invest 1995; adults: recommendations for syndrome defi- et al. Sleep-related changes in baroreflex sen- 96:1897-904. nition and measurement techniques in clin- sitivity and cardiovascular autonomic mod- 11. Virolainen J, Ventila M, Turto H, Kupari ical research: the report of the American ulation. J Hypertens 2003;21:1555-61. M. Effect of negative intrathoracic pressure Academy of Sleep Medicine Task Force. 5. Andreotti F, Davies GJ, Hackett DR, et al. on left ventricular pressure dynamics and re- Sleep 1999;22:667-89. Major circadian fluctuations in fibrinolytic laxation. J Appl Physiol 1995;79:455-60. 18. Fox CS, Evans JC, Larson MG, Kannel factors and possible relevance to time of 12. Guilleminault C, Connolly SJ, Winkle WB, Levy D. Temporal trends in coronary onset of myocardial infarction, sudden car- RA. Cardiac arrhythmia and conduction dis- heart disease mortality and sudden cardiac diac death and stroke. Am J Cardiol 1988; turbances during sleep in 400 patients with death from 1950 to 1999: the Framingham 62:635-7. sleep apnea syndrome. Am J Cardiol 1983; Heart Study. Circulation 2004;110:522-7. 6. Bexton RS, Vallin HO, Camm AJ. Diur- 52:490-4. 19. Young T, Evans L, Finn L, Palta M. Esti- nal variation of the QT interval — influence 13. Fichter J, Bauer D, Arampatzis S, Fries R, mation of the clinically diagnosed propor- of the autonomic nervous system. Br Heart Heisel A, Sybrecht GW. Sleep-related breath- tion of sleep apnea syndrome in middle-aged J 1986;55:253-8. ing disorders are associated with ventricular men and women. Sleep 1997;20:705-6. 7. Huikuri HV, Niemela MJ, Ojala S, Ranta- arrhythmias in patients with an implantable 20. Clarke JM, Hamer J, Shelton JR, Taylor la A, Ikaheimo MJ, Airaksinen KE. Circadian cardioverter-defibrillator. Chest 2002;122: S, Venning GR. The rhythm of the normal rhythms of frequency domain measures of 558-61. human heart. Lancet 1976;1:508-12. heart rate variability in healthy subjects and 14. Harbison J, O’Reilly P, McNicholas WT. 21. Schafer H, Koehler U, Ploch T, Peter JH.

n engl j med 352;12 www.nejm.org march 24, 2005 1213

sudden death from cardiac causes in obstructive sleep apnea

Sleep-related myocardial ischemia and sleep orders in patients with implantable cardio- Circadian variation in the frequency of sud- structure in patients with obstructive sleep verter defibrillators. Pacing Clin Electrophys- den cardiac death. Circulation 1987;75:131- apnea and coronary heart disease. Chest iol 1999;22:223-7. 8. 1997;111:387-93. 29. Koehler U, Fus E, Grimm W, et al. Heart 36. Goraya TY, Jacobsen SJ, Kottke TE, Frye 22. Shepard JW Jr, Garrison MW, Grither block in patients with obstructive sleep ap- RL, Weston SA, Roger VL. Coronary heart DA, Dolan GF. Relationship of ventricular noea: pathogenetic factors and effects of disease death and sudden cardiac death: ectopy to oxyhemoglobin desaturation in treatment. Eur Respir J 1998;11:434-9. a 20-year population-based study. Am J Epi- patients with obstructive sleep apnea. Chest 30. Grimm W, Hoffmann J, Menz V, et al. demiol 2003;157:763-70. 1985;88:335-40. Electrophysiologic evaluation of sinus node 37. Iribarren C, Crow RS, Hannan PJ, Ja- 23. Sanner BM, Konermann M, Tepel M, function and atrioventricular conduction in cobs DR Jr, Luepker RV. Validation of death Groetz J, Mummenhoff C, Zidek W. Platelet patients with prolonged ventricular asystole certificate diagnosis of out-of-hospital function in patients with obstructive sleep during obstructive sleep apnea. Am J Cardi- sudden cardiac death. Am J Cardiol 1998; apnoea syndrome. Eur Respir J 2000;16: ol 1996;77:1310-4. 82:50-3. 648-52. 31. Becker H, Brandenburg U, Peter JH, Von 38. Coady SA, Sorlie PD, Cooper LS, Folsom 24. Narkiewicz K, Montano N, Cogliati C, Wichert P. Reversal of sinus arrest and atrio- AR, Rosamond WD, Conwill DE. Validation van de Borne PJ, Dyken ME, Somers VK. Al- ventricular conduction block in patients with of death certificate diagnosis for coronary tered cardiovascular variability in obstruc- sleep apnea during nasal continuous posi- heart disease: the Atherosclerosis Risk in tive sleep apnea. Circulation 1998;98:1071-7. tive airway pressure. Am J Respir Crit Care Communities (ARIC) Study. J Clin Epidemiol 25. Gillis AM, Stoohs R, Guilleminault C. Med 1995;151:215-8. 2001;54:40-50. Changes in the QT interval during obstruc- 32. Tilkian AG, Guilleminault C, Schroeder 39. Folsom AR, Gomez-Marin O, Gillum tive sleep apnea. Sleep 1991;14:346-50. JS, Lehrman KL, Simmons FB, Dement WC. RF, Kottke TE, Lohman W, Jacobs DR Jr. 26. Narkiewicz K, van de Borne PJ, Pesek CA, Sleep-induced apnea syndrome: prevalence Out-of-hospital coronary death in an urban Dyken ME, Montano N, Somers VK. Selective of cardiac arrhythmias and their reversal after population — validation of death certificate potentiation of peripheral chemoreflex sen- tracheostomy. Am J Med 1977;63:348-58. diagnosis: the Minnesota Heart Survey. Am sitivity in obstructive sleep apnea. Circula- 33. Seppala T, Partinen M, Penttila A, As- J Epidemiol 1987;125:1012-8. tion 1999;99:1183-9. pholm R, Tiainen E, Kaukianen A. Sudden 40. Goraya TY, Jacobsen SJ, Belau PG, 27. Zwillich C, Devlin T, White D, Douglas death and sleeping history among Finnish Weston SA, Kottke TE, Roger VL. Validation N, Weil J, Martin R. Bradycardia during sleep men. J Intern Med 1991;229:23-8. of death certificate diagnosis of out-of-hos- apnea: characteristics and mechanism. J Clin 34. Gonzalez-Rothi RJ, Foresman GE, Block pital coronary heart disease deaths in Olm- Invest 1982;69:1286-92. AJ. Do patients with sleep apnea die in their sted County, Minnesota. Mayo Clin Proc 28. Fries R, Bauer D, Heisel A, et al. Clinical sleep? Chest 1988;94:531-8. 2000;75:681-7. significance of sleep-related breathing dis- 35. Muller JE, Ludmer PL, Willich SN, et al. Copyright © 2005 Massachusetts Medical Society.

clinical trial registration

The Journal encourages investigators to register their clinical trials in a public trials registry. The members of the International Committee of Medical Journal Editors plan to consider clinical trials for publication only if they have been registered (see N Engl J Med 2004;351:1250-1). The National Library of Medicine’s www.clinicaltrials.gov is a free registry, open to all investigators, that meets the committee’s requirements.

1214 n engl j med 352;12 www.nejm.org march 24, 2005

The new england journal of medicine

clinical practice

Pertussis — Not Just for Kids

Erik L. Hewlett, M.D., and Kathryn M. Edwards, M.D.

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 authors’ clinical recommendations.

Six weeks ago, a 45-year-old woman noticed a scratchy feeling in her throat that has now progressed to more than 20 episodes of severe, spasmodic coughing per day. Her coughing spells are worse at night and are sometimes associated with gagging and vomiting. Her adolescent son and several of his friends, who received all of their childhood immunizations on schedule, had similar illnesses involving cough several weeks before the onset of her symptoms, and they continue to cough. How should the patient be assessed for possible pertussis? Should she be treated and, if so, how? Could this illness have been prevented?

the clinical problem

Despite rates of immunization for pertussis of more than 80 percent among young From the Division of Infectious Diseases children, the number of cases of pertussis reported annually in the United States has and International Health, Departments of Medicine and Pharmacology, University of increased by a factor of six since 1980, with 11,647 cases reported in 2003 (Fig. 1). The Virginia School of Medicine, Charlottes- reported incidence rates probably substantially underestimate the true burden of dis- ville (E.L.H.); and the Division of Pediatric ease because of incomplete reporting and a lack of recognition of the illness on the part Infectious Diseases, Department of Pediat- 1-4 rics, Vanderbilt University School of Medi- of physicians. The diagnosis of pertussis is frequently missed, often because of mis- cine, Nashville (K.M.E.). Address reprint conceptions that whooping cough is solely a pediatric illness that has been controlled requests to Dr. Hewlett at P.O. Box 800419, by routine childhood immunizations and that immunity resulting from pertussis dis- University of Virginia School of Medicine, Charlottesville, VA 22908, or at eh2v@ ease or immunization is lifelong. In addition, residual immunity from prior vaccina- virginia.edu. tion may modify the clinical presentation of pertussis in adolescents and adults and make the diagnosis even more difficult. N Engl J Med 2005;352:1215-22. Copyright © 2005 Massachusetts Medical Society. clinical presentation and complications Anyone who has heard the frightening paroxysmal cough of a child with classic pertussis would question how the diagnosis of pertussis could be overlooked. The illness begins, however, less dramatically, with the catarrhal phase, which consists of nonspecific symptoms such as coryza, conjunctival irritation, and, occasionally, a slight cough, none of which suggest pertussis as the primary diagnosis. After 7 to 10 days, the characteristic cough heralds the onset of the paroxysmal phase. After several weeks, the intensity and frequency of the cough may begin to decrease, but the convalescent phase, which can include episodes of exacerbated paroxysmal coughing brought on by unrelated upper airway infections, can last for weeks to months. The typical paroxysmal cough in a child who is not immunized consists of a series of rapid, forced expirations, followed by gasping inhalation, which can result in the typical whooping sound. Post-tussive vomiting is common. These symptoms often occur in the absence of fever. Asymptomatic intervals occur between coughing spells. Very young infants may present with apnea or cyanosis in the absence of cough and are at greater risk than are older children for death or severe complications, including

n engl j med 352;12 www.nejm.org march 24, 2005 1215

The new england journal of medicine

300,000 14,000 12,000 250,000 All ages 10,000 8,000 200,000 DTP 6,000

No. of Cases 4,000 <7 yr

150,000 2,000 <7 mo 0 No. of Cases 1980 1990 2000 100,000 Year

50,000

0 1922 1930 1940 1950 1960 1970 1980 1990 2000 Year

Figure 1. Annual Reported Cases of Pertussis in the United States from 1922 through 2003. The inset shows changes in the number of reported pertussis cases since 1980 according to age group (< seven months, < seven years, and all ages). DTP denotes diphtheria, tetanus toxoids, and pertussis vaccine. Unpublished data are from the Centers for Disease Control and Prevention, by permission of Dr. Trudy Murphy and Dr. Margaret Cortese.

pneumonia (Bordetella pertussis infection or infec- 70 to 99 percent of adolescents and adults with per- tion with another pathogen), pneumothorax, severe tussis are reported to have paroxysmal cough, the pulmonary hypertension, seizures, or encephalop- reported frequencies of other symptoms are more athy.5-9 variable, with whoop in 8 to 82 percent and post- The current pertussis-related mortality rate tussive vomiting in 17 to 50 percent.6 Because ado- among infants in the United States is 2.4 deaths lescents and adults often do not seek medical care per 1 million, and fatal cases in infants account for until several weeks after the onset of their illness, more than 90 percent of all deaths from pertussis. the differential diagnosis includes other causes of These numbers highlight the need for new ap- chronic cough, such as asthma, gastroesophageal proaches to protect infants, who are too young to reflux disease, postviral bronchospasm, chronic si- be immunized according to the current vaccination nusitis with postnasal drip, tuberculosis, chlamydia schedule.2,10,11 or mycoplasma infections, other chronic lung dis- Among immunized patients, especially adoles- eases, and malignant conditions. Almost 80 percent cents and adults, a prolonged cough may be the only of adults with confirmed pertussis have an illness manifestation of pertussis.3 A number of studies involving a cough of at least 3 weeks’ duration, have documented that between 13 and 32 percent and 27 percent still had a cough after 90 days.18,19 of adolescents and adults with an illness involving Complications of pertussis, which are similar in a cough of six days’ duration or longer have sero- adolescents and adults, include pneumonia (in 2.1 logic evidence of infection with B. pertussis.12-17 The to 3.5 percent of patients), seizures (0.3 to 0.6 per- symptoms and complications of pertussis infection cent), and encephalopathy (0.1 percent). Some reported in these studies are often quite different complications, such as cough-induced urinary in- from those seen in children.14-18 For example, continence, increase with age. Unusual complica- scratchy throat and other pharyngeal symptoms oc- tions that have been reported anecdotally in older cur in about one third of adults with pertussis, and patients include a herniated intervertebral disc, episodes of sweating are reported by 40 to 50 per- the sudden onset of hearing loss, angina, and ca- cent of persons over the age of 30 years. Although rotid-artery dissection.6,18-24

1216 n engl j med 352;12 www.nejm.org march 24, 2005

clinical practice epidemiology, incidence, and burden of disease Table 1. Recommended Immunization Schedules for Pertussis in Canada, France, Germany, and the United States. Although multiple reports have highlighted the role of pertussis as an important cause of persistent Type of Ages at cough in adolescents and adults, the actual bur- Country Vaccine* Vaccination Ages at Booster† den of disease in these groups is difficult to de- Canada DTaP† 2, 4, and 6 mo 18 mo; 4–6 and 14–16 yr 4,6,12-25 termine. This seemingly simple task is France DTP‡ 2, 3, and 4 mo complicated by the biology of B. pertussis and the DTaP 16–18 mo; 11–13 yr difficulty in detecting the organism.4 Most patho- gens of the respiratory tract have short incubation Germany DTaP 3, 4, and 5 mo 11–14 mo; 9–17 yr periods, are easy to culture, cause illness for a United States DTaP 2, 4, and 6 mo 18 mo; 4–6 yr relatively short period, and are rapidly eliminated. B. pertussis is different in that the incubation peri- * DTaP denotes diphtheria, tetanus toxoids, and acellular pertussis vaccine. There are multiple formulations of this vaccine, with each formulation con- od is measured in days to weeks, the organism ex- taining one to four antigens, and multiple manufacturers; therefore, no two hibits fastidious behavior in culture, and it has the vaccines are identical. Commonly recognized side effects include swelling, ability to cause, as the Chinese term it, a “cough of redness, and tenderness at the site of injection; fever; and secondary febrile convulsions. 100 days.” The organism can be recovered from pa- † Canada, France, and Germany have initiated the use of a booster dose in tients only during the first three to four weeks of ill- adolescents at the ages indicated. In France, a DTaP booster is used after the ness and is particularly difficult to isolate in previ- primary series with DTP. In Canada and Germany, the adolescent booster is 25 dTap, reflecting a reduced dose of diphtheria and acellular pertussis. ously immunized persons. These factors result ‡ DTP denotes diphtheria, tetanus toxoids, and pertussis vaccine. Side effects in outbreaks that span months and that can be dif- include anaphylaxis; swelling, redness, and tenderness at the site of injection; ficult to track epidemiologically.26 prolonged or inconsolable crying; fever; and secondary febrile convulsions. Before vaccination was available, pertussis was responsible for more than 270,000 cases of severe illness involving cough and 10,000 deaths annually eases, are its inclusion in the differential diagnosis, in the United States.27 The introduction of whole- the availability of reliable and rapid laboratory tests cell pertussis vaccine into the general population for confirmation, and the prompt and appropriate during the 1940s was associated with a 99 percent use of the resultant information. reduction in the incidence of the disease, with a nadir of 1010 reported cases in 1976. Since that time, diagnostic methods the absolute number of reported cases has in- Three laboratory tests have been used to detect B. per- creased, with the 11,647 cases in 2003 approach- tussis or related bordetella species in the respiratory ing the highest total since 1964 (Fig. 1).1 The rou- secretions of patients suspected of having pertussis: tine immunization of young children in the United direct fluorescent antibody staining, bacterial cul- States according to the schedule shown in Table 1 ture, and polymerase chain reaction (PCR).29 Be- has markedly reduced the rates of reported pertus- cause the direct fluorescent antibody technique has sis in children. However, striking increases in rates low sensitivity and specificity in comparison with of disease have been seen among adolescents and culture and PCR, this method is not currently rec- adults during the past 10 years (Fig. 1 and 2). In the ommended.29,30 A positive culture of nasopharyn- United States, from 1997 to 2000 there were 8273 geal secretions (preferably an aspirate) on selective cases of pertussis reported in patients 10 to 19 Regan–Lowe medium is the gold standard for the years of age and 5745 cases in those older than 19 diagnosis of pertussis. The sensitivity of culture is years.1 There are probably multiple reasons for this limited by the fastidious nature of B. pertussis and increase, and they include the increased recogni- the loss of viability between the site of collection tion of the disease (with the use of serologic testing and the clinical laboratory. The yield is lower from for diagnosis) and the limited duration of protec- dilute specimens, specimens from patients who tion from vaccine.28 have had a long duration of illness, and specimens obtained after antibiotic therapy. Nonetheless, strategies and evidence when carried out appropriately, culture continues to be recommended in patients who present within The critical elements involved in the diagnosis and three weeks after the onset of cough. Although an- treatment of pertussis, as with other infectious dis- tibiotic resistance is rare, isolation of the causative

n engl j med 352;12 www.nejm.org march 24, 2005 1217

The new england journal of medicine

organism, as recommended by the Centers for Dis- ca).29,33,34 Although the period during which the ease Control and Prevention (CDC), allows for char- organism can be identified is longer with PCR than acterization of antigenic variation and antibiotic for culture, false positive PCR results have been a sensitivity.31 problem.35 For these reasons, the National Immu- The use of PCR-based diagnostic assays with one nization Program of the CDC recommends the use or more primers, available in many health depart- of culture and PCR assays during the period in ments and clinical laboratories, is an alternative ap- which patients could be infectious — at least three proach. These assays have been reported to detect weeks after the onset of cough or four weeks after fewer than 10 organisms, which do not need to be the appearance of any symptoms.36 Standardized viable to be detected, and thus the tests have signif- procedures and reagents and rigid quality control icantly greater sensitivity than does culture.29,32 PCR of PCR assays are needed to ensure accurate diag- assays with multiple primers can be used to identify noses and to minimize false positive results.35,37,38 and distinguish among several bordetella species An alternative method of diagnosing pertussis in (i.e., B. pertussis, B. parapertussis, and B. bronchisepti- adolescents and adults is the measurement of antibody to specific components of B. pertussis with the use of enzyme-linked immunosorbent assays. Since the diagnosis of pertussis is often not considered during the period when organisms can still be detected, the demonstration of increased antibody titers between the acute phase of illness (i.e., in serum obtained within the first week after the onset of symptoms) and the convalescent phase (in serum collected four to six weeks later) may be necessary. Alternatively, a single serum antibody titer (that exceeds a diagnostic cutoff point for the level of IgG against pertussis toxin or another antigen) obtained at least three weeks into the illness may be required to confirm the diagnosis.29,38-40 However, the lack of a widely available serologic test with an established cutoff point, the limited diagnostic value of commercial tests, and the delay in obtaining results limit the usefulness of serologic testing in practice.6,41 As a result, the CDC recommends that combinations of diagnostic tests be used to identify persons with pertussis more effectively.29,38 Within the first four weeks after the onset of symptoms (when cough has been present for three weeks), the use of culture and PCR is appropriate for diagnosis; when cough has been present for three to four weeks, PCR and serologic tests can both be used; and after four weeks of cough, serologic tests alone are most like- Figure 2. Epidemiologic “Life Cycles” of B. pertussis before and after the Generalized Use of Pertussis Vaccine. ly to provide a diagnosis. In the prevaccination era, the majority of pertussis cases occurred in children. treatment Adults who had had pertussis as children had their acquired immunity boost- ed by recurrent exposures in the population, and mothers then passed protec- Antibiotics should be administered to patients with tion to infants through the placental transfer of antibodies. After the use of pertussis to hasten clearance of the organism and pertussis vaccine has been established in a population, the newly immunized to limit transmission to susceptible contacts. Al- pediatric group is protected; an increasing proportion of cases occur in adolescents and adults, who have lost their vaccine-induced immunity, and in though controversial, it appears that treatment can infants, who receive fewer passive antibodies than did infants in the prevacci- sometimes reduce the duration or severity of symp- nation era and who are too young to be immunized according to the current toms, or both, but a benefit is unlikely when treat- immunization schedule (Table 1). ment is initiated after the first week of symptoms.36 Adolescents and adults with sporadic cases fre-

1218 n engl j med 352;12 www.nejm.org march 24, 2005

clinical practice quently present during the paroxysmal phase, which exposed within the previous three weeks,50 but for occurs at least a week after the onset of symptoms, persons at high risk or those who are likely to come and in such cases, antibiotics rarely affect the course into contact with high-risk persons, prophylaxis of the disease. However, viable organisms can be may be warranted for up to six to eight weeks after recovered from untreated patients for three weeks exposure. after the onset of the cough. Thus, the routine administration of antibiotics during the first four prevention weeks of illness is justified. For patients who are The mainstay for control of pertussis is vaccination, likely to be in contact with high-risk persons, such and this is recommended routinely for infants and as infants, women in the third trimester of preg- young children according to the schedules shown nancy (who might continue to be infectious after for several countries in Table 1. Whole-cell vaccines delivery), and health care workers, treatment is rec- consisting of killed organisms are highly effective ommended even six to eight weeks after the onset but have been associated with frequent local and of illness.36 systemic reactions.27 In the 1980s and early 1990s, The National Immunization Program recom- new acellular vaccines, which contain one or more mends erythromycin for the treatment of pertus- purified pertussis components, were demonstrated sis.42 However, the newer macrolides azithromycin to be immunogenic, associated with fewer local and and clarithromycin have been demonstrated in systemic reactions than whole-cell vaccines, and ef- head-to-head trials to be similar in efficacy to eryth- ficacious in the prevention of culture-confirmed romycin, with fewer side effects and more conve- pertussis disease in infants and young children.51 nient administration.43-45 Dosages, common side These acellular vaccines have replaced whole-cell effects, and contraindications are summarized in vaccines in the United States and elsewhere. Table 2.46,47 An alternative treatment for patients Given the recognition that immunity wanes after who cannot tolerate macrolides is trimethoprim– vaccination early in life, repeated vaccination in ad- sulfamethoxazole. Although B. pertussis strains are olescence or adulthood has been proposed. A large, sensitive in vitro to fluoroquinolones and ketolides, randomized clinical trial evaluated the efficacy of a there are no clinical data to support the use of these three-component acellular pertussis vaccine (con- agents.48 taining pertussis toxoids, filamentous hemaggluti- Treatment also involves supportive care, which is nin, and pertactin) in nearly 3000 healthy subjects most important for infants and small children, 15 to 65 years of age. Active surveillance for pertus- since they are vulnerable to dehydration and malnu- sis was conducted every two weeks for two years. trition from post-tussive vomiting and an inability With the use of culture or PCR and serologically to eat. None of the pharmacologic or immunologic confirmed disease as end points, preliminary data interventions that have been tested, such as cortico- from this study indicated that there was a 92 per- steroids, salbutamol, diphenhydramine, and percent efficacy rate for the vaccine (95 percent confi- tussis immune globulin, have been documented to dence interval, 32 to 99 percent).52 be effective in reducing the symptoms or control- A recently published cost-benefit analysis that ling the cough of pertussis.49 assumed that the incidence of pertussis was the same as that observed in the above-mentioned trial prophylaxis after exposure led to the conclusion that an additional booster The same agents and regimens used for the treat- dose of pertussis vaccine would be cost-effective.53 ment of patients with established pertussis are Of the scenarios considered in the analysis, vaccina- recommended for chemoprophylaxis in contacts. tion of adolescents was considered to be the least ex- These modes of treatment are expected to be effec- pensive, the easiest to implement, and the most tive in the protection of persons exposed to pa- acceptable. In light of the greater recognition of tients with active pertussis, when administered be- disease in adolescents and adults, some countries, fore the onset of symptoms in the contact. Because such as Germany, France, and Canada, now recom- infectiousness declines with the increasing dura- mend the routine vaccination of adolescents with tion of illness, prophylaxis for contacts needs to be acellular pertussis vaccine boosters (Table 1). Al- initiated only if the interval since the onset of cough though there are no formulations of pertussis vac- in the index case is three weeks or less. Prophylaxis cine currently licensed in the United States for use is generally recommended only if the contact was in persons over the age of six years, advisory groups

n engl j med 352;12 www.nejm.org march 24, 2005 1219

The new england journal of medicine

Table 2. Choice of Antibiotic Agents for the Treatment of Pertussis.

Drug Dosage Regimen Side Effects Contraindications

Erythromycin Children, 40–50 mg/kg/day 4 divided doses for Gastrointestinal irritation, Known sensitivity to any macro- Adults, 1–2 g/day 14 days* abdominal cramps, nau- lide antibiotic; should be used sea, vomiting; hypertro- with caution in infants be- phic pyloric stenosis has cause of association with hy- been reported in infants pertrophic pyloric stenosis46 Azithromycin 10 mg/kg (maximum, 500 mg) as Lower dose once Allergic reaction and Known sensitivity to any macro- a single dose on day 1; 5 mg/kg daily for 4 addi- hepatic toxicity lide antibiotic (maximum, 250 mg) thereafter43‡ tional days Clarithromycin 20 mg/kg/day (maximum, 1 g/day) Two divided doses Allergic reaction and Known sensitivity to any macro- daily for 7 days hepatic toxicity lide antibiotic Trimethoprim– Trimethoprim, 8 mg/kg/day Two divided doses Rash, kernicterus Known allergy to sulfonamides sulfameth- (maximum, 320 mg/day); daily for 7 days in newborns or trimethoprim; should not oxazole† sulfamethoxazole, 40 mg/kg/day be given to pregnant women (maximum, 1600 mg/day) shortly before delivery, breast- feeding mothers, or infants <2 mo old, because of the risk of kernicterus

* A 7-day regimen has been shown to be similar in efficacy to a 14-day regimen.54 † Trimethoprim–sulfamethoxazole should be prescribed as an alternative to the macrolides in patients who cannot tolerate them. ‡ The 2003 Red Book recommendation is for 10 to 12 mg per kilogram per day once daily for five days.

are considering the routine use of acellular vac- guidelines cines in this population once vaccines are approved and become available. The CDC recommends that for all patients with presumed pertussis, culture be performed to areas of uncertainty identify the etiologic agent during the time when patients are likely to be infectious, regardless of Despite extensive research on the toxins, adhesions, which other diagnostic tests are used.36 The CDC and other factors related to the virulence of B. per- and the Food and Drug Administration are cur- tussis, little is known about the mechanisms by rently working on standardization of PCR and se- which these factors cause the illness of pertussis. rologic methods for the diagnosis of pertussis. For example, although it is clear that pertussis Patients seen early in the course of their illness toxin causes the characteristic lymphocytosis and (i.e., during the first three weeks after cough be- other abnormalities associated with the disease, the gins) should be evaluated with the use of culture mechanism responsible for the severe and pro- and PCR; PCR and serologic tests can be used when longed cough remains unknown. cough has been present for three to four weeks; Vaccination of pregnant women has been pro- and serologic tests should be used for patients who posed as a strategy to protect infants from pertus- present with cough that has persisted for longer sis passively before they receive active vaccination. than four weeks. The potential for such a strategy to work is based on The CDC also recommends the treatment of pre- observations that maternal antibody titers to per- sumed or confirmed cases of pertussis with erythro- tussis antigens are low but that, when present, an- mycin but acknowledges the limitations of this tibodies are actively transported in cord blood.55 treatment due to side effects. In cases in which the However, data to support the safety and efficacy of patient presents after the onset of paroxysmal maternal vaccination are lacking. An alternative or cough, antibiotic treatment is unlikely to affect the complementary approach would be the active im- clinical course but will preclude transmission to sus- munization of newborns with acellular pertussis ceptible hosts beginning five days after the onset of vaccines; such approaches are being studied. therapy.

1220 n engl j med 352;12 www.nejm.org march 24, 2005

clinical practice

conclusions proach, even if her condition did progress to clini- and recommendations cal pertussis, the antibiotic treatment would have the potential to reduce the duration and severity of For the patient described in the vignette, whose her illness. symptoms started six weeks ago, single-sample se- Pertussis vaccines are highly efficacious, but in rologic testing is the only method that could yield many countries, including the United States, they the diagnosis. Given the duration of the symp- are administered only to a small subgroup of the toms, treatment with antibiotics would not affect population — namely, children younger than six the course of the patient’s illness, and we would years of age. The control of pertussis requires an not recommend it at this point. If, however, the di- increase in the immunity of all age groups. We be- agnosis of pertussis had been made two or more lieve that the vaccination of adolescents (with a weeks earlier, it would have been appropriate to suitable formulation of acellular pertussis vaccine) treat the patient in order to prevent further spread should be added to the current immunization of the infection. If pertussis had been documented schedule for pertussis in order to reduce the risk of (by culture, PCR, or both) in the patient’s son dur- the disease later in life as well as the transmission ing the first three weeks of his illness, it would have to infants. been appropriate to consider her a contact and to We are indebted to Dr. Trudy Murphy and Dr. Margaret Cortese of treat her, whether or not she became symptomatic, the National Immunization Program, CDC, for their review of the manuscript; to Dr. Emily Wong and Mr. Alan Wong for their contri- with either azithromycin or clarithromycin, ac- bution to Figure 2; and to Ms. Candace Green and Mrs. Sarah cording to the regimen in Table 2. With this ap- Baugher for their assistance in the preparation of the manuscript. references 1. Pertussis — United States, 1997–2000. gard KM, Tate JE, Murphy TV. Trends in per- Morbidity of pertussis in adolescents and MMWR Morb Mortal Wkly Rep 2002;51: tussis among infants in the United States, adults. J Infect Dis 2000;182:174-9. 73-6. 1980-1999. JAMA 2003;290:2968-75. 22. Skowronski DM, De Serres G, Mac- 2. Vitek CR, Pascual FB, Baughman AL, 12. Mink CM, Cherry JD, Christenson P, et Donald D, et al. The changing age and sea- Murphy TV. Increase in deaths from pertus- al. A search for Bordetella pertussis infec- sonal profile of pertussis in Canada. J Infect sis among young infants in the United tion in university students. Clin Infect Dis Dis 2002;185:1448-53. [Erratum, J Infect States in the 1900s. Pediatr Infect Dis J 2003; 1992;14:464-71. Dis 2002;185:1696.] 22:628-34. 13. Cattaneo LA, Reed GW, Haase DH, 23. Skowronski DM, Buxton JA, Hestrin M, 3. Deeks S, De Serres G, Boulianne N, et al. Wills MJ, Edwards KM. The seroepidemi- Keyes RD, Lynch K, Halperin SA. Carotid ar- Failure of physicians to consider the diagno- ology of Bordetella pertussis infections: tery dissection as a possible severe compli- sis of pertussis in children. Clin Infect Dis a study of persons ages 1-65 years. J Infect cation of pertussis in an adult: clinical case 1999;28:840-6. Dis 1996;173:1256-9. report and review. Clin Infect Dis 2003;36: 4. Crowcroft NS, Stein C, Duclos P, Bir- 14. Schmitt-Grohe S, Cherry JD, Heininger e1-e4. mingham M. How best to estimate the glob- U, Uberall MA, Pineda E, Stehr K. Pertussis 24. Halperin SA, Marrie TJ. Pertussis en- al burden of pertussis? Lancet Infect Dis in German adults. Clin Infect Dis 1995;21: cephalopathy in an adult: case report and re- 2003;3:413-8. 860-6. view. Rev Infect Dis 1991;13:1043-7. 5. Casano P, Odena MP, Cambra FJ, Martin 15. Birkebaek NH, Kristiansen M, Seefeldt 25. Strebel P, Nordin J, Edwards K, et al. JM, Palomeque A. Bordetella pertussis infec- T, et al. Bordetella pertussis and chronic Population-based incidence of pertussis tion causing pulmonary hypertension. Arch cough in adults. Clin Infect Dis 1999;29: among adolescents and adults, Minnesota, Dis Child 2002;86:453. 1239-42. 1995-1996. J Infect Dis 2001;183:1353-9. 6. von Konig CH, Halperin S, Riffelmann 16. Nennig ME, Shinefield HR, Edwards 26. Kurt TL, Yeager AS, Guenette S, Dunlop M, Guiso N. Pertussis of adults and infants. KM, Black SB, Fireman BH. Prevalence and S. Spread of pertussis by hospital staff. Lancet Infect Dis 2002;2:744-50. incidence of adult pertussis in an urban pop- JAMA 1972;221:264-7. 7. Smith C, Vyas H. Early infantile pertus- ulation. JAMA 1996;275:1672-4. 27. Cherry JD, Brunell PA, Golden GS, Kar- sis: increasingly prevalent and potentially 17. Wright SW, Edwards KM, Decker MD, zon D. Report of the task force on pertussis fatal. Eur J Pediatr 2000;159:898-900. Zeldin MH. Pertussis infection in adults and pertussis vaccine. Pediatrics 1988;81: 8. Goulin GD, Kaya KM, Bradley JS. Severe with persistent cough. JAMA 1995;273: Suppl:939-84. pulmonary hypertension associated with 1044-6. 28. Cherry JD. The science and fiction of the shock and death in infants infected with 18. Postels-Multani S, Schmitt HJ, Wirsing “resurgence” of pertussis. Pediatrics 2003; Bordetella pertussis. Crit Care Med 1993;21: von Konig CH, Bock HL, Bogaerts H. Symp- 112:405-6. 1791-4. toms and complications of pertussis in 29. Muller FM, Hoppe JE, Wirsing von 9. McEniery JD, Delbridge RG, Reith DM. adults. Infection 1995;23:139-42. Konig CH. Laboratory diagnosis of pertus- Infant pertussis deaths and the manage- 19. Thomas PF, McIntyre PB, Jalaludin BB. sis: state of the art in 1997. J Clin Microbiol ment of cardiovascular compromise. J Pae- Survey of pertussis morbidity in adults in 1997;35:2435-43. diatr Child Health 2004;40:230-2. western Sydney. Med J Aust 2000;173:74-6. 30. Ewanowich CA, Chui LWL, Paranchych 10. Edwards KM. Pertussis: an important 20. Trollfors B, Rabo E. Whooping cough in MG, Peppler MS, Marusyk RG, Albritton target for maternal immunization. Vaccine adults. Br Med J (Clin Res Ed) 1981;283: WL. Major outbreak of pertussis in northern 2003;21:3483-6. 696-7. Alberta, Canada: analysis of discrepant di- 11. Tanaka M, Vitek CR, Pascual FB, Bis- 21. De Serres G, Shadmani R, Duval B, et al. rect fluorescent-antibody and culture results

n engl j med 352;12 www.nejm.org march 24, 2005 1221

clinical practice

by using polymerase chain reaction meth- rological diagnosis of pertussis. Clin Infect 48. Hoppe JE, Bryskier A. In vitro suscepti- odology. J Clin Microbiol 1993;31:1715-25. Dis 1999;28:Suppl 2:S99-S106. bilities of Bordetella pertussis and Bordetella 31. Mastrantonio P, Spigaglia P, van Oir- 40. Heininger U, Schmidt-Schlapfer G, parapertussis to two ketolides (HMR 3004 schot H, et al. Antigenic variants of Bordetella Cherry JD, Stehr K. Clinical validation of and HMR 3647), four macrolides (azithro- pertussis strains isolated from vaccinated and a polymerase chain reaction assay for the mycin, clarithromycin, erythromycin A, and unvaccinated children. Microbiology 1999; diagnosis of pertussis by comparison with roxithromycin), and two ansamycins (ri- 145:2069-75. serology, culture, and symptoms during fampin and rifapentine). Antimicrob Agents 32. Edelman K, Nikkari S, Ruuskanen O, a large pertussis vaccine efficacy trial. Pedi- Chemother 1998;42:965-6. He Q, Viljanen M, Mertsola J. Detection of atrics 2000;105(3):E31. 49. Pillay V, Swingler G. Symptomatic treat- Bordetella pertussis by polymerase chain re- 41. Kosters K, Riffelmann M, Dohrn B, ment of the cough in whooping cough. Coch- action and culture in the nasopharynx of Konig CHW. Comparison of five commer- rane Database Syst Rev 2003;4:CD003257. erythromycin-treated infants with pertussis. cial enzyme-linked immunosorbent assays 50. Steketee RW, Wassilak SGF, Adkins WN Pediatr Infect Dis J 1996;15:54-7. for detection of antibodies to Bordetella pertus- Jr, et al. Evidence for a high attack rate and 33. Cloud JL, Hymas WC, Turlak A, et al. sis. Clin Diagn Lab Immunol 2000;7:422-6. efficacy of erythromycin prophylaxis in a Description of a multiplex Bordetella pertussis 42. Bergquist SO, Bernander S, Dahnsjo H, pertussis outbreak in a facility for the devel- and Bordetella parapertussis LightCycler PCR Sundelof B. Erythromycin in the treatment opmentally disabled. J Infect Dis 1988;157: assay with inhibition control. Diagn Micro- of pertussis: a study of bacteriologic and 434-40. biol Infect Dis 2003;46:189-95. clinical effects. Pediatr Infect Dis J 1987;6: 51. Edwards KM, Decker MD. Combination 34. Farrell DJ, McKeon M, Daggard G, Loef- 458-61. [Erratum, Pediatr Infect Dis J 1987; vaccines. Infect Dis Clin North Am 2001;15: felholz MJ, Thompson CJ, Mukkur TK. Rap- 6:1035.] 209-30. id-cycle PCR method to detect Bordetella 43. Langley JM, Halperin SA, Boucher FD, 52. Ward J. Acellular pertussis vaccines in pertussis that fulfills all consensus recom- Smith B. Azithromycin is as effective as and adolescents and adults. In: Program and ab- mendations for use of PCR in diagnosis of better tolerated than erythromycin estolate stracts of the 41st Interscience Conference pertussis. J Clin Microbiol 2000;38:4499- for the treatment of pertussis. Pediatrics on Antimicrobial Agents and Chemothera- 502. 2004;111:e96-e101. py, Chicago, December 16–19, 2001. Wash- 35. Lievano FA, Reynolds MA, Waring AL, et 44. Aoyama T, Sunakawa K, Iwata S, Takeu- ington, D.C.: American Society for Microbi- al. Issues associated with and recommenda- chi Y, Fujii R. Efficacy of short-term treat- ology, 2001:520. abstract. tions for using PCR to detect outbreaks of ment of pertussis with clarithromycin and 53. Purdy KW, Hay JW, Botteman MF, Ward pertussis. J Clin Microbiol 2002;40:2801-5. azithromycin. J Pediatr 1996;129:761-4. JI. Evaluation of strategies for use of acellu- 36. Guidelines for the control of pertussis 45. Summaries of infectious diseases: per- lar pertussis vaccine in adolescents and outbreaks. Atlanta: National Immunization tussis. In: Pickering L, ed. 2003 Red book: adults: a cost-benefit analysis. Clin Infect Program, 2000. (Accessed January 20, 2005, report of the Committee on Infectious Dis- Dis 2004;39:20-8. at http://www.cdc.gov/nip/publications/ eases. 26th ed. Elk Grove Village, Ill.: Amer- 54. Halperin SA, Bertolussi R, Langley JM, pertussis/guide.htm.) ican Academy of Pediatrics, 2003:472-86. Miller B, Eastwood BJ. Seven days of eryth- 37. Meade BD, Bollen A. Recommenda- 46. Mahon BE, Rosenman MB, Kleinman romycin estolate is as effective as fourteen tions for use of the polymerase chain reac- MB. Maternal and infant use of erythromy- days for the treatment of Bordetella pertussis tion in the diagnosis of Bordetella pertussis cin and other macrolide antibiotics as risk infections. Pediatrics 1997;100:65-71. infections. J Med Microbiol 1994;41:51-5. factors for infantile hypertrophic pyloric ste- 55. Healy CM, Munoz FM, Rench MA, Hala- 38. Fry NK, Tzivra O, Li YT, et al. Laboratory nosis. J Pediatr 2001;139:380-4. sa NB, Edwards KM, Baker CJ. Prevalence of diagnosis of pertussis infections: the role of 47. Ray W, Murray K, Meredith S, Narasim- pertussis antibodies in maternal delivery, PCR and serology. J Med Microbiol 2004;53: hulu S, Hall K, Stein C. Oral erythromycin cord, and infant serum. J Infect Dis 2004; 519-25. and the risk of sudden death from cardiac 190:335-40. 39. Hallander HO. Microbiological and se- causes. N Engl J Med 2004;351:1089-96. Copyright © 2005 Massachusetts Medical Society.

powerpoint slides of journal figures and tables

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

1222 n engl j med 352;12 www.nejm.org march 24, 2005

The new england journal of medicine

review article

medical progress Polycystic Ovary Syndrome

David A. Ehrmann, M.D.

n 1935, stein and leventhal published a paper on their findings From the University of Chicago, Depart- in seven women with amenorrhea, hirsutism, obesity, and a characteristic polycys- ment of Medicine, Section of Endocrinol- i ogy, Chicago. Address reprint requests to tic appearance to their ovaries — one of the first descriptions of a complex pheno- Dr. Ehrmann at 5841 S. Maryland Ave., 1 type today known as the polycystic ovary syndrome. Insight into the pathogenesis and Mail Code 1027, Chicago, IL 60637, or at treatment of the polycystic ovary syndrome has increased substantially in the decade [email protected]. 2 since this topic was last addressed in the Journal. The condition is now well recognized N Engl J Med 2005;352:1223-36. as having a major effect throughout life on the reproductive, metabolic, and cardiovas- Copyright © 2005 Massachusetts Medical Society. cular health of affected women. This review addresses current knowledge regarding the diagnosis, cause, and treatment of the polycystic ovary syndrome.

definition and differential diagnosis

Several factors contribute to difficulties in the diagnosis of the polycystic ovary syndrome. Presenting signs and symptoms are heterogeneous and vary over time; in addition, a precise and uniform definition of the syndrome has been lacking. An international consensus group3 recently proposed that the syndrome can be diagnosed after the exclusion of other medical conditions that cause irregular menstrual cycles and androgen excess (Fig. 1 and Table 1) and the determination that at least two of the following are present: oligoovulation or anovulation (usually manifested as oligomenorrhea or amenorrhea), elevated levels of circulating androgens (hyperandrogenemia) or clinical manifestations of androgen excess (hyperandrogenism), and polycystic ovaries as defined by ultrasonography.4 These criteria acknowledge the condition as functional: polycystic ovaries need not be present to make a diagnosis of the polycystic ovary syndrome,5 and conversely, their presence alone does not establish the diagnosis.6,7 Women with the polycystic ovary syndrome almost always have some aberration in gonadotropin secretion as compared with women who have normal menstrual cycles.8 However, since gonadotropin concentrations vary over the menstrual cycle and are re- leased in a pulsatile fashion into the circulation, a single measurement of luteinizing hormone and follicle-stimulating hormone provides little diagnostic sensitivity. Thus, in routine clinical practice, abnormal gonadotropin levels (an elevated level of luteinizing hormone or an elevated ratio of luteinizing hormone to follicle-stimulating hormone) need not be documented to diagnose the polycystic ovary syndrome. Chronic anovulation most often manifests as oligomenorrhea (fewer than nine menses per year) or amenorrhea. Anovulatory cycles may lead to dysfunctional uterine bleeding and decreased fertility. Cutaneous manifestations of hyperandrogenemia in the polycystic ovary syndrome include hirsutism, acne, and male-pattern hair loss (androgenic alopecia), whereas acanthosis nigricans is a cutaneous marker of hyperinsulinemia. A substantial proportion of women with the polycystic ovary syndrome are overweight; many are obese, some extremely so.9 Although obesity itself is not considered the inciting event in the development of the syndrome, excess adiposity can exacerbate associated reproductive and metabolic derangements. The symptoms of the polycystic ovary syndrome usually begin around menarche,10

n engl j med 352;12 www.nejm.org march 24, 2005 1223

The new england journal of medicine

Any 2 of the following 3 disorders confirmed: All of the following disorders ruled out: Oligomenorrhea or amenorrhea Hyperprolactinemia Hyperandrogenism (e.g., hirsutism, Nonclassic congenital adrenal hyperplasia acne, alopecia) or hyperandrogenemia Cushing’s syndrome (e.g., elevated levels of total or free Androgen-secreting neoplasm testosterone) Acromegaly Polycystic ovaries on ultrasonography

Polycystic ovary syndrome

Ancillary studies

Risk assessment for Risk assessment for Fasting cholesterol, HDL Risk assessment for endometrial carcinoma glucose intolerance cholesterol, triglycerides, obstructive sleep apnea LDL cholesterol

Endometrial biopsy Oral glucose-tolerance test Polysomnography if risk increased if risk increased if risk increased

Figure 1. Diagnostic Algorithm for the Polycystic Ovary Syndrome. Single measurements of serum prolactin and 17-hydroxyprogesterone are usually sufficient to rule out hyperprolactinemia and nonclassic congenital adrenal hyperplasia due to deficiency of 21-hydroxylase. It is important to measure the 17-hydroxyprogesterone level in a blood sample taken in the early morning, when endogenous corticotropin levels peak. As an alternate test, 17-hydroxyprogesterone can be measured in response to a single dose of exogenously administered corticotropin. Risks for glucose intolerance include an elevated body-mass index, an increased waist circumference, a history of gestational diabetes, a family history of type 2 diabetes, and certain racial or ethnic back- grounds (including black, Caribbean Hispanic, and Mexican American). HDL denotes high-density lipoprotein, and LDL low-density lipoprotein.

but onset after puberty may also occur as a result of hormone, the ovarian theca cell synthesizes andro- environmental modifiers such as weight gain. Pre- gens (Fig. 2). Androgen biosynthesis is mediated by mature pubarche, the result of early secretion of cytochrome P-450c17, an enzyme with 17a-hydrox- adrenal steroids, may be a harbinger of the syn- ylase and 17,20-lyase activities, both of which are drome.11 In addition, an aberrant intrauterine en- required to form androstenedione. The androgenic vironment has been implicated in the condition’s steroid is then converted by 17b-hydroxysteroid de- pathogenesis, particularly its metabolic compo- hydrogenase (17b-HSD) to form testosterone or is nents.12-15 aromatized by the aromatase enzyme (cytochrome P-450arom) to form estrone. Studies performed pathogenesis both in vivo and in vitro (in cultured theca cells) con- sistently suggest that ovarian theca cells in affected No single etiologic factor fully accounts for the spec- women are more efficient at converting androgen- trum of abnormalities in the polycystic ovary syn- ic precursors to testosterone than are normal theca drome. In response to stimulation by luteinizing cells.16,17

1224 n engl j med 352;12 www.nejm.org march 24, 2005

medical progress

Table 1. Conditions for Exclusion in the Diagnosis of the Polycystic Ovary Syndrome.

Hyperandro- genemia, Hyper- androgenism, Oligomenorrhea Condition or Both or Amenorrhea Distinguishing Features Clinical Hormonal or Biochemical Nonclassic congenital Yes Not often Family history of infertility, Elevated (basal) level of adrenal hyperplasia hirsutism, or both; 17-hydroxyprogester- due to deficiency of common in Ashkenazi one in the morning or 21-hydroxylase Jews on stimulation Cushing’s syndrome Yes Yes Hypertension, striae, easy Elevated 24-hr urinary free bruising cortisol level Hyperprolactinemia or None or mild Yes Galactorrhea Elevated plasma prolactin prolactinoma level Primary hypothyroidism None or mild May be present Goiter may be present Elevated plasma thyrotro- pin and subnormal plasma thyroxine level; prolactin level may also be increased Acromegaly None or mild Often Acral enlargement, coarse Increased plasma insulin- features, prognathism like growth factor I Premature ovarian None Yes May be associated with Elevated plasma follicle- failure other autoimmune stimulating hormone endocrinopathies and normal or subnormal estradiol level Simple obesity Often Not often Diagnosed by exclusion None Virilizing adrenal or Yes Yes Clitorimegaly, extreme hir- Extremely elevated plasma ovarian neoplasm sutism, or male-pattern androgen level alopecia Drug-related condition* Often Variably Evidence provided by None history

* A drug-related condition is a condition due to the use of androgens, valproic acid, cyclosporine, or other drugs.

Whereas luteinizing hormone regulates the Because women with the polycystic ovary syndrome androgenic synthesis of theca cells, follicle-stim- appear to have an increased luteinizing hormone ulating hormone is responsible for regulating the pulse frequency8 (Fig. 2), it has been inferred that aromatase activity of granulosa cells, thereby de- the pulse frequency of GnRH must be accelerated termining how much estrogen is synthesized from in the syndrome. It is not clear whether this accel- androgenic precursors. When the concentration of erated pulse frequency is due to an intrinsic abnor- luteinizing hormone increases relative to that of mality in the GnRH pulse generator or is caused by follicle-stimulating hormone, the ovaries prefer- the relatively low levels of progesterone resulting entially synthesize androgen. from infrequent ovulatory events. Since progestins The frequency of the stimulus of hypothalamic slow the GnRH pulse generator, low circulating gonadotropin-releasing hormone (GnRH) deter- progestin levels in women with the polycystic ovary mines, in part, the relative proportion of luteiniz- syndrome may lead to an acceleration in the pulsa- ing hormone and follicle-stimulating hormone syn- tility of GnRH,19 increased levels of luteinizing hor- thesized within the gonadotrope (Fig. 2). Increased mone, and overproduction of ovarian androgen. pulse frequency of hypothalamic GnRH favors tran- Insulin plays both direct and indirect roles in the scription of the b-subunit of luteinizing hormone pathogenesis of hyperandrogenemia in the polycys- over the b-subunit of follicle-stimulating hormone; tic ovary syndrome (Fig. 2). Insulin acts synergis- conversely, decreased pulse frequency of GnRH fa- tically with luteinizing hormone to enhance the vors transcription of the b-subunit of follicle-stim- androgen production of theca cells. Insulin also in- ulating hormone, which decreases the ratio of lute- hibits hepatic synthesis of sex hormone–binding inizing hormone to follicle-stimulating hormone.18 globulin, the key circulating protein that binds to

n engl j med 352;12 www.nejm.org march 24, 2005 1225

The new england journal of medicine

GnRH

Normal

Progesterone

Polycystic Polycystic ovary syndrome ovary

LH FSH

Cholesterol scc, StAR Insulin 3-HSD Pregnenolone Progesterone 17-hydroxylase Granulosa cell 17-OH-progesterone 17,20-lyase Theca cell Androstenedione 17 Estrone Aromatase Testosterone 17 Estradiol

Figure 2. The Hypothalamic–Pituitary–Ovarian Axis and the Role of Insulin. Increased ovarian androgen biosynthesis in the polycystic ovary syndrome results from abnormalities at all levels of the hypothalamic–pituitary–ovarian axis. The increased frequency of luteinizing hormone (LH) pulses in the polycystic ovary syndrome appears to result from an increased frequency of hypothalamic gonadotropin-releasing hormone (GnRH) pulses. The latter can result from an intrinsic abnormality in the hypothalamic GnRH pulse generator, favoring the production of luteinizing hormone over follicle-stimulating hormone (FSH) in patients with the polycystic ovary syndrome, in whom the administration of progesterone can restrain the rapid pulse frequency. By whatever mechanism, the relative increase in pituitary secretion of luteinizing hormone leads to an increase in androgen production by ovarian theca cells. In- creased efficiency in the conversion of androgenic precursors in theca cells leads to enhanced production of androstenedione, which is then converted by 17b-hydroxysteroid dehydrogenase (17b) to form testosterone or aromatized by the aromatase enzyme to form estrone. Within the granulosa cell, estrone is then converted into estradiol by 17b. Numerous autocrine, paracrine, and endocrine factors modulate the effects of both luteinizing hormone and insulin on the androgen production of theca cells; insulin acts synergistically with luteinizing hormone to enhance androgen production. Insulin also inhibits hepatic synthesis of sex hormone–binding globulin, the key circulating protein that binds to testosterone and thus increases the proportion of testosterone that circulates in the unbound, biologically available, or “free,” state. Testosterone inhibits and estrogen stimulates hepatic synthesis of sex hormone–binding globulin. The abbreviation scc denotes side-chain cleavage enzyme, StAR steroidogenic acute regulatory protein, and 3b-HSD 3b-hydroxysteroid dehydrogenase. Solid arrows denote a higher degree of stimulation than dashed arrows.

1226 n engl j med 352;12 www.nejm.org march 24, 2005

medical progress testosterone, and thus increases the proportion of obesity testosterone that circulates in the unbound, biolog- The cause of obesity in the polycystic ovary syn- ically available, or free, state. Because women with drome remains unknown, but obesity is present in the polycystic ovary syndrome typically have hyper- at least 30 percent of cases; in some series, the per- insulinemia, the concentration of free testosterone centage is as high as 75.62 Women in the United is often elevated when the total testosterone con- States with the polycystic ovary syndrome generally centration is at the upper range of normal or only have a higher body weight than their European modestly elevated. counterparts.62-65 This fact has been cited as an explanation for the increase in the incidence of the polycystic ovary syndrome in the U.S. population — the role of genetic 66 and environmental factors an increase that parallels the increase in obesity. Increased adiposity, particularly visceral adipos- The polycystic ovary syndrome remains one of the ity that is reflected by an elevated waist circumfer- most common hormonal disorders in women, ence (>88 cm [35 in.]) or waist-to-hip ratio, has with a prevalence estimated between 5 and 10 per- been associated with hyperandrogenemia, insulin cent.20-22 Variance in prevalence among popula- resistance, glucose intolerance, and dyslipidemia.60 tions may reflect the effect of ethnic origin, race, and Attenuation of insulin resistance, whether accom- other environmental factors on the phenotype.23,24 plished by weight loss or with medication, amelio- Several lines of evidence suggest that the poly- rates (but not necessarily normalizes) many of the cystic ovary syndrome is heritable,25-29 and various metabolic aberrations in women with the polycystic approaches have been initiated to attempt to define ovary syndrome. a specific genetic cause.30,31 In rare instances, single-gene mutations can give rise to the phenotype impaired glucose tolerance of the syndrome.32 Current understanding of the and type 2 diabetes pathogenesis of the syndrome suggests, however, Thirty to 40 percent of women with the polycystic that it is a complex multigenic disorder. Candidate ovary syndrome have impaired glucose tolerance, genes that may regulate the hypothalamic–pitu- and as many as 10 percent have type 2 diabetes by itary–ovarian axis, as well as those responsible for their fourth decade.9,67 These prevalence rates are insulin resistance and its sequelae, have been the among the highest known among women of simi- principal focus of linkage and case–control studies. lar age.68 An enhanced rate of deterioration in glu- Microarray analyses of target tissues in the polycys- cose tolerance is also evident in the polycystic ovary tic ovary syndrome31 have been used to identify nov- syndrome.9,69 Seminal studies by Dunaif et al.70-72 el candidate genes involved in the condition, and a indicated that women with the polycystic ovary syn- number of them appear to contribute modestly to drome are more insulin resistant than are unaffect- the phenotype (Table 2).30,57,58 ed counterparts matched for body-mass index, fat- free body mass, and body-fat distribution. A defect metabolic, cardiovascular, in the insulin signaling pathway appears to be and other clinical components present in both the adipocyte and skeletal muscle, the primary target tissues of insulin action.71,72 The consequences of the polycystic ovary syndrome Insulin resistance alone cannot fully account for extend beyond the reproductive axis; women with the predisposition to and development of type 2 di- the disorder are at substantial risk for the develop- abetes among patients with the polycystic ovary syn- ment of metabolic and cardiovascular abnormali- drome. In patients with normal glucose tolerance, ties similar to those that make up the metabolic syn- insulin secretion is (by definition) sufficient for the drome.59 This finding is not surprising, since both degree of insulin resistance; when the pancreatic the polycystic ovary syndrome and the metabolic b cell is no longer able to compensate sufficiently, syndrome share insulin resistance as a central glucose tolerance begins to deteriorate.73,74 Most pathogenetic feature (Fig. 2). The polycystic ovary women with the polycystic ovary syndrome are able syndrome might thus be viewed as a sex-specific to compensate fully for their insulin resistance, but form of the metabolic syndrome,60 and the term a substantial proportion (particularly those with a “syndrome XX” has been suggested as an apt term first-degree relative with type 2 diabetes27) have a to underscore this association.61 disordered and insufficient b-cell response to meals

n engl j med 352;12 www.nejm.org march 24, 2005 1227

The new england journal of medicine

hypertension and vascular dysfunction Table 2. Representative Candidate Genes with Evidence of Linkage, Hypertension develops in some women with the Association, or Both, with the Polycystic Ovary Syndrome (PCOS). polycystic ovary syndrome during their reproduc- Pathway and Protein (Gene) Comments tive years,59,80 and sustained hypertension may de- 81,82 Insulin secretion and action velop in later life in women with the disorder. 83 Insulin receptor (INSR) region D19S884, an anonymous marker 1 cM centro- Reduced vascular compliance and vascular endo- 83-86 meric to INSR; evidence for linkage and thelial dysfunction were noted in most, but not 26,33 association with PCOS all,87 studies of women with the polycystic ovary Insulin variable-number tan- Region involved in transcriptional regulation syndrome. Furthermore, the degree of impairment dem repeats (VNTR) of insulin gene; evidence for linkage and association with class III allele34-36 in vascular reactivity is significantly greater than can be explained by obesity alone.83 Insulin-lowering Insulin receptor substrate 1 Post-receptor molecule in insulin signaling (IRS-1) pathway; association with PCOS37,38 therapies appear to improve the vascular endothe- Insulin receptor substrate 2 Post-receptor molecule in insulin signaling lial dysfunction in patients with the polycystic ovary 85 (IRS-2) pathway; association with PCOS 37,38 syndrome. Calpain 10 (CAPN10) Cysteine protease with effect on insulin action and secretion; linkage and association coronary and other vascular disease 39,40 with type 2 diabetes A predisposition to macrovascular disease and Peroxisome-proliferator–acti- The Pro12Ala polymorphism in the PPARg thrombosis88,89 in women with the polycystic ovary vated receptor g (PPARg) gene is a modifier of insulin resistance in PCOS41-43 syndrome has also been described. A recent study Protein phosphatase 1 regula- Variant of regulatory subunit of the glycogen- of premenopausal women showed that those with tory subunit (PPP1R3) associated form of protein phosphatase-1 the polycystic ovary syndrome had a higher preva- derived from skeletal muscle; associated lence of coronary-artery calcification as detected by with insulin resistance44 electron-beam computed tomography.90 Increased Gonadotropin secretion and action levels of plasminogen-activator inhibitor type 1 may contribute to this risk.91-93 Levels of plasminogen- Follistatin (FST) Acts to inhibit ovarian follicular maturation and androgen production; enhances folli- activator inhibitor type 1 in patients with the poly- cle-stimulating hormone and insulin 26,45 cystic ovary syndrome may exceed those typically secretion seen in type 2 diabetes.91 A reduction in insulin lev- Androgen biosynthesis, els91 decreases levels and activity of plasminogen- secretion, transport, and metabolism activator inhibitor type 1. Androgen receptor (AR) Number of CAG repeats associated with an- Hypertriglyceridemia, increased levels of very- drogen levels in PCOS46 low-density lipoprotein and low-density lipoprotein Sex hormone–binding Association of the pentanucleotide (TAAAA)n cholesterol, and decreased levels of high-density globulin (SHBG) repeat polymorphism with PCOS47,48 lipoprotein cholesterol94 also predispose patients to Cytochrome P-450c17 (CYP17) Possible association with PCOS49-53 vascular disease in the polycystic ovary syndrome. Cytochrome P-45011a Early analyses revealed association with hy- Both insulin resistance and hyperandrogenemia 54,55 (CYP11a) perandrogenemia and PCOS ; more contribute to this atherogenic lipid profile. Testos- recently, strength of association has been questioned56 terone decreases lipoprotein lipase activity in abdominal fat cells, and insulin resistance impairs the 11b-Hydroxysteroid dehydro- Mutations in both 11b-HSD and H6PD in genase (11b-HSD) and a triallelic digenic model of inheritance ability of insulin to exert its antilipolytic effects. Al- hexose-6-phosphate de- result in low 11b-HSD expression and though these abnormalities would be expected to hydrogenase (H6PD) NADPH generation with loss of 11b-HSD 32 increase the morbidity and mortality from coronary 1 oxo-reductase activity artery disease and other vascular disorders in women with the polycystic ovary syndrome, this has been difficult to establish.95-97 or a glucose challenge.27,75-78 Before the development of frank glucose intolerance, defects in insu- obstructive sleep apnea lin secretion may be latent and revealed only in cir- Recent studies indicate that the prevalence of ob- cumstances that augment insulin resistance, as with structive sleep apnea in the polycystic ovary syn- the development of gestational diabetes in preg- drome is higher than expected and cannot be ex- nancy79 or glucose intolerance associated with glu- plained by obesity alone.98-100 In two studies,98,100 cocorticoid administration.77 the severity of sleep apnea did not correlate with

1228 n engl j med 352;12 www.nejm.org march 24, 2005 medical progress body-mass index; in a third study,99 even after con- since most progestins also possess variable andro- trolling for body-mass index, the risk of sleep-dis- genic effects. Norgestimate and desogestrel are vir- ordered breathing was increased by a factor of 30. tually nonandrogenic progestins.103 Drospirenone, Insulin resistance appears to be a stronger predic- an analogue of spironolactone with unique antimin- tor of sleep-disordered breathing than is age, body- eralocorticoid and antiandrogenic activities,104 has mass index, or the circulating testosterone concen- been approved for use in combination with ethinyl tration.98 estradiol; thus, it is potentially ideal for the treatment of women with the polycystic ovary syn- association with cancer drome.105 There is an increased prevalence of endometrial hy- Controversy persists regarding the use of oral perplasia and carcinoma in women with the poly- contraceptives as first-line therapy in women with cystic ovary syndrome.101,102 This increase has been the polycystic ovary syndrome.106 These agents attributed largely to the persistent stimulation of clearly improve hirsutism and acne and protect endometrial tissue by estrogen (mainly estrone) against unopposed estrogenic stimulation of the without the progesterone-induced inhibition of endometrium, but their potential adverse effects on proliferation and differentiation to secretory en- insulin resistance, glucose tolerance, vascular re- dometrium that occurs after ovulation. Endometrial activity, and coagulability are a concern, particular- carcinoma has also been associated with obesity and ly since insulin-lowering agents are now available type 2 diabetes, both of which are common in the (see below). polycystic ovary syndrome. Breast and ovarian cancer have been variably as- Antiandrogens sociated with the polycystic ovary syndrome102; The antiandrogen cyproterone acetate competitive- obesity, anovulation, infertility, and the hormonal ly inhibits the binding of testosterone and its more treatment of infertility are so frequent in the poly- potent conversion product, 5a-dihydrotestosterone, cystic ovary syndrome that the condition is difficult to the androgen receptor. Although not available in to isolate as an independent risk factor for these the United States, cyproterone acetate effectively types of cancer. treats hirsutism and acne107 and is used throughout Canada, Mexico, and Europe. treatment Spironolactone, typically used as an antiminer- alocorticoid, possesses moderate antiandrogenic cutaneous manifestations of androgen effects when administered in large doses (100 to excess: hirsutism and acne 200 mg daily).108 Adverse effects seem to be mini- Medical treatment of hirsutism and acne in the poly- mal, with the exception of occasional vaginal bleed- cystic ovary syndrome generally aims to reduce an- ing resulting from progestin-like properties at a drogen levels, attenuate their effects by lowering high dose. Spironolactone and oral contraceptives androgen production, augment androgen binding appear to be synergistic. For these reasons, and be- to specific plasma-binding proteins, and block an- cause antiandrogens should not be administered to drogen action at the level of the target tissue. women desiring pregnancy, a combination of estrogen and progestin is often prescribed in conjunction Oral Contraceptives with spironolactone. Estrogen–progestin combination therapy (with the Flutamide is a potent nonsteroidal antiandrogen use of the combination oral-contraceptive pill) re- that is effective in the treatment of hirsutism.109-111 mains the predominant treatment for hirsutism and Concern about inducing hepatocellular dysfunction, acne in the polycystic ovary syndrome. The estro- however, has limited its use. genic component of the oral contraceptive sup- presses luteinizing hormone and thus ovarian an- Glucocorticoids drogen production. Estrogen also enhances hepatic Some women with the polycystic ovary syndrome production of sex hormone–binding globulin (Fig. have elevated adrenal androgen levels,5,112 though 2), thereby reducing the free, or unbound, fraction the contribution to ovulatory dysfunction appears of plasma testosterone available to occupy the an- modest.113 Unless a woman with the polycystic ova- drogen receptor. ry syndrome has marked adrenal androgen excess, The choice of oral contraceptive is important, prolonged use of glucocorticoids is not advised.

n engl j med 352;12 www.nejm.org march 24, 2005 1229 The new england journal of medicine

Other Agents to be advantageous in this patient population. How- Finasteride, a 4-aza steroid and competitive inhibi- ever, several recent studies designed to address this tor of type 2 5a-reductase,114 has been reported issue have not shown a distinct benefit from calo- to treat hirsutism.115 However, the prominence of rie-restricted diets limiting carbohydrates rather type 1 5a-reductase in the pilosebaceous unit makes than fat.123,124 it unlikely to be an optimal treatment for the andro- A reduction in insulin levels pharmacologically gen-related cutaneous manifestations associated ameliorates sequelae of both hyperinsulinemia and with the condition. hyperandrogenemia. The place of insulin-reduction Eflornithine hydrochloride, an inhibitor of the therapies in treating the polycystic ovary syndrome enzyme ornithine decarboxylase in human skin, has is evolving and should be viewed in context with all been approved for topical use in treating facial hir- available therapies (Table 3). These therapies can ef- sutism. The inhibition of hair growth is its primary fectively manage the established metabolic derange- action, but clinical data are too limited to recom- ments in the polycystic ovary syndrome, but whether mend its routine use. they can prevent them is not yet established. Both metformin (a biguanide) and the thiazo- management of oligomenorrhea lidinediones pioglitazone and rosiglitazone have and amenorrhea been used to reduce insulin resistance. Although Chronic anovulation is associated with an increased metformin appears to influence ovarian steroido- risk of endometrial hyperplasia and carcinoma, as genesis directly,125,126 this effect does not appear to discussed. Thus, it is prudent to consider endome- be primarily responsible for the attenuation of ovar- trial biopsy in patients with the polycystic ovary syn- ian androgen production in women with the poly- drome who have not had menstrual bleeding for a cystic ovary syndrome. Rather, metformin inhibits year or longer. Some investigators have advocated the output of hepatic glucose, necessitating a lower the use of ultrasonography to determine endome- insulin concentration and thereby probably reduc- trial thickness in deciding whether to biopsy the ing the androgen production of theca cells. endometrium.102 Endometrial proliferation can be Subject characteristics and control measures for inhibited by administering either cyclic progestin effects of weight change, dose of metformin, and or oral contraceptives with a combination of estro- outcome vary widely among published studies of gen and progestin. The latter approach, which also metformin in the polycystic ovary syndrome. A re- reduces ovarian androgen production, may be par- cent meta-analysis of 13 studies in which metfor- ticularly beneficial in this setting. min was administered to 543 participants127 report- The induction of ovulation is a complex issue in ed that patients taking metformin had an odds ratio the polycystic ovary syndrome that is beyond the for ovulation of 3.88 (95 percent confidence inter- scope of this review but has been discussed else- val, 2.25 to 6.69) as compared with placebo and an where.116-119 It is important to note, however, that odds ratio for ovulation of 4.41 (95 percent confi- modest reductions in body weight (2 to 7 percent) dence interval, 2.37 to 8.22) for metformin plus through lifestyle modification have been associated clomiphene as compared with clomiphene alone. with reductions in androgen levels and improved Metformin also improved fasting insulin levels, ovulatory function in patients with the polycystic blood pressure, and levels of low-density lipopro- ovary syndrome.120,121 In addition, numerous stud- tein cholesterol. These effects were judged to be in- ies have shown that the lowering of insulin levels dependent of any changes in weight that were as- may increase ovulatory events, potentially restoring sociated with metformin, but controversy persists cyclic menses and fertility.122,123 The long-term out- as to whether the beneficial effects of metformin comes of pregnancies associated with the use of in- are entirely independent of the weight loss128 that is sulin-lowering medications remain unknown. typically seen early in the course of therapy. Finally, the rates of spontaneous miscarriage and gestation- insulin resistance and glucose intolerance al diabetes are reportedly lower among women with Weight reduction is important in treating over- the polycystic ovary syndrome who conceive while weight patients with the polycystic ovary syndrome. taking metformin.129-132 The long-term effects of No unique weight-loss regimen targets excess adi- metformin in pregnancy are unknown. posity specific to the syndrome. Restricting carbo- The thiazolidinediones improve the action of in- hydrates as compared with fats is generally perceived sulin in the liver, skeletal muscle, and adipose tissue

1230 n engl j med 352;12 www.nejm.org march 24, 2005 medical progress X X Insulin Lowering X X X X Induction Ovulation Uses X X X X or Amenorrhea Oligomenorrhea X X X X X X X Acne Hirsutism, Examples norgestimate (Ortho- ethinyl estra- Cyclen); diol and desogestrel (Orthocept); ethinyl estradiol and drospirenone (Yasmin) spironolactone, flutamide sone ride (Vaniqa) Glucophage XR) rosiglitazone (Avandia) Ethinyl estradiol and acetate, Cyproterone Prednisone, dexametha- Finasteride (Propecia) Eflornithine hydrochlo- Clomiphene citrate Metformin (Glucophage, Pioglitazone (Actos), -reductase in the a estrogen and progestin; effective for hirsutism and acne; may increase risk of thrombosis and metabolic abnormalities; beneficial antiandrogenic effects of drospirenone risk of hyperkalemia (spironolactone) or hepatitis (flutamide) drogen excess; long-term risks of glucose intolerance, insulin resistance, osteopenia, weight gain enzyme of 5 pilosebaceous unit topically less effective in obese pa- apy, tients; may be useful in conjunction with insulin-lowering therapies menstrual cycling, less effective for hirsutism; usually associated with initial weight loss; may have untoward gastrointestinal effects of insulin and androgens, modest effects on hirsutism; associated with weight gain Advantages or Disadvantages Cyclic exposure of endometrium to Cyclic Effective for hirsutism and acne; Attenuate adrenal component of an- Do not specifically target the iso- Minimal documented efficacy; used Moderately effective as monother- Substantial efficacy in restoration of Extremely effective at lowering levels -reductase a Mechanism of Action and FSH, suppress ovarian androgen production; progestin can act as an antiandrogen binding to the androgen receptor thus adrenal androgen production lase duction, secondarily lowering insulin levels; may have direct effects on ovarian steroidogenesis get-tissue level (adipocyte, muscle); may have direct effects on ovarian steroidogenesis Increase SHBG, suppress LH Inhibit androgens from Suppress corticotropin and Inhibit 5 Inhibit ornithine decarboxy- Induces rise in FSH, LH Reduces hepatic glucose pro- Enhance insulin action at tar- Drug Treatments.* and progestin inhibitors -Reductase inhibitors a Table 3. Table Agent of estrogen Combinations Antiandrogens Glucocorticoids 5 Ornithine decarboxylase Antiestrogen Biguanide Thiazolidinediones * SHBG denotes sex hormone–binding globulin, LH luteinizing hormone, and FSH follicle-stimulating hormone.

n engl j med 352;12 www.nejm.org march 24, 2005 1231 The new england journal of medicine

and have only a modest effect on hepatic glucose only a subgroup of women with the polycystic ova- output. As with metformin,125,126 the thiazolidine- ry syndrome go on to have glucose intolerance, just diones are reported to affect ovarian steroid synthe- that high-risk subgroup should be screened. Factors sis directly,133 although most evidence indicates that that augment the risk are increased body weight the reduction in insulin levels is responsible for de- (particularly if body fat is distributed in an android creased concentrations of circulating androgen. pattern), a history of gestational diabetes, type 2 di- Obese women with the polycystic ovary syn- abetes in a first-degree relative, and Caribbean-His- drome who took troglitazone had consistent im- panic,141 Mexican-American,23 or African-Ameri- provements in insulin resistance, hyperandro- can39 heritage. genemia, and glucose tolerance.91,134 In addition, Integral to this issue is whether the measurement troglitazone treatment was associated with a rela- of fasting glucose, with or without simultaneous tive improvement in pancreatic b-cell function and measurement of fasting insulin, is sufficient to as- a reduction in levels of the prothrombotic factor sess the risk of glucose intolerance. Although a su- plasminogen-activator inhibitor type 1.91 These pranormal fasting glucose concentration increases findings led to a double-blind, randomized, place- the likelihood that a patient will have an abnormally bo-controlled study of troglitazone in the polycys- elevated glucose concentration at two hours during tic ovary syndrome.62 Ovulation was significantly a formal oral glucose-tolerance test, a normal fast- greater for women who received troglitazone than ing glucose concentration does not necessarily pre- for those who received placebo; free testosterone dict normal glucose tolerance and is insufficient to levels decreased, and levels of sex hormone–binding distinguish between women who have normal glu- globulin increased in a dose-dependent fashion. cose tolerance and those who have impaired glucose Nearly all glycemic measures showed dose-related tolerance.9,62,67 Thus, if the goal is to detect im- decreases with troglitazone treatment. Although paired glucose tolerance for the purpose of interven- troglitazone is no longer available, subsequent ing to reduce the risk of conversion from impaired studies using rosiglitazone135,136 and pioglita- glucose tolerance to type 2 diabetes, an oral glucose- zone137 have had similar results. Because of con- tolerance test should be performed. cern about using thiazolidinediones in pregnancy, Although insulin resistance is virtually inherent the drugs have been less readily adopted for routine in the phenotype of the polycystic ovary syndrome, clinical use. there is little to support its formal assessment outside the context of a clinical study. First, insulin re- controversies sistance is not a diagnostic criterion, nor is it recommended as a factor to be used in determining Several unresolved controversies persist regarding treatment3 in the polycystic ovary syndrome. These the evaluation and treatment of women with the recommendations relate to the observation that the polycystic ovary syndrome. One issue surrounds the clinical response to insulin-lowering therapies does question of whether all women with the polycystic not appear to be related to the magnitude of insulin ovary syndrome should be screened for glucose in- resistance. Insulin resistance is virtually universal tolerance, insulin resistance, or both. Screening is and maximal once the body-mass index (the weight supported by evidence that the combined prevalence in kilograms divided by the square of the height in of impaired glucose tolerance and type 2 diabetes meters) exceeds 3073; thus, there appears to be little approaches 45 percent by the fourth decade,9,67 value gained by formal measurement of insulin sen- that both impaired glucose tolerance and type 2 di- sitivity in obese patients. Finally, it is important to abetes are associated with significant morbidity, and note that the simple and readily available methods that there is a substantial rate of conversion from proposed as tests to quantify insulin resistance — impaired glucose tolerance to diabetes in the ab- the ratio of fasting glucose to insulin142 or the sence of intervention among women with the poly- homeostatic model assessment index143 — may cystic ovary syndrome9,69 and those without the con- be misleading, since both have been shown to dition.138,139 The American Diabetes Association lack precision when compared with the gold-stan- recognizes the polycystic ovary syndrome as a risk dard method for quantifying whole-body insulin factor that justifies screening for type 2 diabetes.140 resistance (i.e., the hyperinsulinemic–euglycemic On the other hand, it may be argued that since clamp).144

1232 n engl j med 352;12 www.nejm.org march 24, 2005 medical progress

summary pathogenesis of insulin resistance that is associated with complications of the polycystic ovary syn- The polycystic ovary syndrome is one of the most drome has led to novel therapies — chiefly insulin- common hormonal disorders affecting women. As lowering medications. Research that is focused on a syndrome, it has multiple components — repro- the genetic and environmental determinants of the ductive, metabolic, and cardiovascular — with polycystic ovary syndrome may provide the basis for health implications across the life span. Androgen new treatment methods and possible prevention of excess and insulin resistance, both of which have the syndrome and its sequelae. strong genetic components, underlie much of the Supported in part by grants from the National Institutes of clinical presentation. The insulin resistance of the Health (M01-RR00055, K08-DK002315, R01-HL075079, and polycystic ovary syndrome appears to impart an in- P60-DK20595), the American Diabetes Association, and the Blum- Kovler Foundation. creased risk of glucose intolerance, diabetes, and Dr. Ehrmann reports having received a consulting fee from Merck lipid abnormalities and may enhance the develop- and lecture and grant support from Takeda. ment of macrovascular disease. Obstructive sleep I am indebted to the patients and nursing staff at the University of Chicago General Clinical Research Center and to Drs. Kenneth apnea is emerging as important in the polycystic S. Polonsky, Graeme I. Bell, Robert L. Rosenfield, and Andrea E. ovary syndrome. A better understanding of the Dunaif. references 1. Stein IF, Leventhal ML. Amenorrhea as- precocious pubarche: ontogeny of the low- that flutamide restores sensitivity of the sociated with bilateral polycystic ovaries. birthweight effect. Clin Endocrinol (Oxf ) gonadotropin-releasing hormone pulse gen- Am J Obstet Gynecol 1935;29:181-91. 2001;55:667-72. erator to inhibition by estradiol and proges- 2. Franks S. Polycystic ovary syndrome. 12. Eisner JR, Dumesic DA, Kemnitz JW, terone. J Clin Endocrinol Metab 2000;85: N Engl J Med 1995;333:853-61. [Erratum, Abbott DH. Timing of prenatal androgen 4047-52. N Engl J Med 1995;333:1435.] excess determines differential impairment 20. Knochenhauer E, Key TJ, Kahsar-Miller 3. Revised 2003 consensus on diagnostic in insulin secretion and action in adult fe- M, Waggoner W, Boots LR, Azziz R. Preva- criteria and long-term health risks related to male rhesus monkeys. J Clin Endocrinol lence of the polycystic ovary syndrome in polycystic ovary syndrome (PCOS). Hum Metab 2000;85:1206-10. unselected black and white women of the Reprod 2004;19:41-7. 13. Eisner JR, Dumesic DA, Kemnitz JW, southeastern United States: a prospective 4. Adams J, Polson DW, Franks S. Preva- Colman RJ, Abbott DH. Increased adiposity study. J Clin Endocrinol Metab 1998;83: lence of polycystic ovaries in women with in female rhesus monkeys exposed to an- 3078-82. anovulation and idiopathic hirsutism. Br drogen excess during early gestation. Obes 21. Diamanti-Kandarakis E, Kouli CR, Ber- Med J (Clin Res Ed) 1986;293:355-9. Res 2003;11:279-86. giele AT, et al. A survey of the polycystic ova- 5. Ehrmann D, Rosenfield R, Barnes RB, 14. Eisner JR, Barnett MA, Dumesic DA, Ab- ry syndrome in the Greek island of Lesbos: Brigell DF, Sheikh Z. Detection of functional bott DH. Ovarian hyperandrogenism in hormonal and metabolic profile. J Clin En- ovarian hyperandrogenism in women with adult female rhesus monkeys exposed to docrinol Metab 1999;84:4006-11. androgen excess. N Engl J Med 1992;327: prenatal androgen excess. Fertil Steril 2002; 22. Asuncion M, Calvo RM, San Millan JL, 157-62. 77:167-72. Sancho J, Avila S, Escobar-Morreale HF. 6. Polson DW, Adams J, Wadsworth J, 15. Abbott DH, Dumesic DA, Franks S. De- A prospective study of the prevalence of the Franks S. Polycystic ovaries — a common velopmental origin of polycystic ovary syn- polycystic ovary syndrome in unselected finding in normal women. Lancet 1988;1: drome — a hypothesis. J Endocrinol 2002; Caucasian women from Spain. J Clin Endo- 870-2. 174:1-5. crinol Metab 2000;85:2434-8. 7. Michelmore KF, Balen AH, Dunger DB, 16. Nelson VL, Legro RS, Strauss JF III, 23. Kauffman RP, Baker VM, DiMarino P, Vessey MP. Polycystic ovaries and associated McAllister JM. Augmented androgen pro- Gimpel T, Castracane VD. Polycystic ovarian clinical and biochemical features in young duction is a stable steroidogenic phenotype syndrome and insulin resistance in white women. Clin Endocrinol (Oxf ) 1999;51: of propagated theca cells from polycystic and Mexican American women: a compari- 779-86. ovaries. Mol Endocrinol 1999;13:946-57. son of two distinct populations. Am J Obstet 8. Waldstreicher J, Santoro NF, Hall JE, Fil- 17. Nelson VL, Qin KN, Rosenfield RL, et al. Gynecol 2002;187:1362-9. icori M, Crowley WF Jr. Hyperfunction of the The biochemical basis for increased testos- 24. Williamson K, Gunn A, Johnson N, Mil- hypothalamic-pituitary axis in women with terone production in theca cells propagated som SR. The impact of ethnicity on the pre- polycystic ovarian disease: indirect evidence from patients with polycystic ovary syn- sentation of polycystic ovarian syndrome. for partial gonadotroph desensitization. drome. J Clin Endocrinol Metab 2001;86: Aust N Z J Obstet Gynaecol 2001;41:202-6. J Clin Endocrinol Metab 1988;66:165-72. 5925-33. 25. Azziz R, Kashar-Miller MD. Family his- 9. Ehrmann DA, Barnes RB, Rosenfield 18. Haisenleder DJ, Dalkin AC, Ortolano tory as a risk factor for the polycystic ovary RL, Cavaghan MK, Imperial J. Prevalence of GA, Marshall JC, Shupnik MA. A pulsatile syndrome. J Pediatr Endocrinol Metab 2000; impaired glucose tolerance and diabetes in gonadotropin-releasing hormone stimulus 13:1303-6. women with polycystic ovary syndrome. Di- is required to increase transcription of the 26. Urbanek M, Legro R, Driscoll DA, et al. abetes Care 1999;22:141-6. gonadotropin subunit genes: evidence for Thirty-seven candidate genes for polycystic 10. Franks S. Adult polycystic ovary syn- differential regulation of transcription by ovary syndrome: strongest evidence for link- drome begins in childhood. Best Pract Res pulse frequency in vivo. Endocrinology age is with follistatin. Proc Natl Acad Sci Clin Endocrinol Metab 2002;16:263-72. 1991;128:509-17. U S A 1999;96:8573-8. 11. Ibanez L, Valls C, Potau N, Marcos MV, 19. Eagleson CA, Gingrich MB, Pastor CL, 27. Ehrmann DA, Sturis J, Byrne MM, Karri- de Zegher F. Polycystic ovary syndrome after et al. Polycystic ovarian syndrome: evidence son T, Rosenfield RL, Polonsky KS. Insulin

n engl j med 352;12 www.nejm.org march 24, 2005 1233 The new england journal of medicine

secretory defects in polycystic ovary syn- clinical profile of polycystic ovary patients. genic states. Clin Endocrinol (Oxf ) 1997; drome: relationship to insulin sensitivity J Clin Endocrinol Metab 2003;88:5529-36. 46:131-4. and family history of non-insulin-depen- 41. Orio F Jr, Matarese G, Di Biase S, et al. 53. Diamanti-Kandarakis E, Bartzis MI, dent diabetes mellitus. J Clin Invest 1995; Exon 6 and 2 peroxisome proliferator-acti- Zapanti ED, et al. Polymorphism T˚C (-34 96:520-7. vated receptor-gamma polymorphisms in bp) of gene CYP17 promoter in Greek pa- 28. Legro RS, Driscoll D, Strauss JF III, Fox polycystic ovary syndrome. J Clin Endo- tients with polycystic ovary syndrome. Fertil J, Dunaif A. Evidence for a genetic basis for crinol Metab 2003;88:5887-92. Steril 1999;71:431-5. hyperandrogenemia in polycystic ovary syn- 42. Korhonen S, Heinonen S, Hiltunen M, 54. Gharani N, Waterworth DM, Batty S, et drome. Proc Natl Acad Sci U S A 1998;95: et al. Polymorphism in the peroxisome pro- al. Association of the steroid synthesis gene 14956-60. liferator-activated receptor-gamma gene in CYP11a with polycystic ovary syndrome and 29. Kahsar-Miller MD, Nixon C, Boots LR, women with polycystic ovary syndrome. hyperandrogenism. Hum Mol Genet 1997; Go RC, Azziz R. Prevalence of polycystic Hum Reprod 2003;18:540-3. 6:397-402. ovary syndrome (PCOS) in first-degree rela- 43. San Millan JL, Corton M, Villuendas G, 55. Diamanti-Kandarakis E, Bartzis MI, tives of patients with PCOS. Fertil Steril Sancho J, Peral B, Escobar-Morreale HF. As- Bergiele AT, Tsianateli TC, Kouli CR. Micro- 2001;75:53-8. sociation of the polycystic ovary syndrome satellite polymorphism (tttta)(n) at -528 30. Urbanek M, Spielman R. Genetic analy- with genomic variants related to insulin re- base pairs of gene CYP11alpha influences sis of candidate genes for the polycystic ova- sistance, type 2 diabetes mellitus, and obesi- hyperandrogenemia in patients with poly- ry syndrome. Curr Opin Endocrinol Diabe- ty. J Clin Endocrinol Metab 2004;89:2640-6. cystic ovary syndrome. Fertil Steril 2000;73: tes 2002;9:492-501. 44. Alcoser SY, Hara M, Bell GI, Ehrmann 735-41. 31. Wood JR, Nelson VL, Ho C, et al. The DA. Association of the (AU)AT-rich element 56. Gaasenbeek M, Powell BL, Sovio U, et molecular phenotype of polycystic ovary polymorphism in PPP1R3 with hormonal al. Large-scale analysis of the relationship syndrome (PCOS) theca cells and new can- and metabolic features of polycystic ovary between CYP11A promoter variation, poly- didate PCOS genes defined by microarray syndrome. J Clin Endocrinol Metab 2004; cystic ovarian syndrome, and serum testos- analysis. J Biol Chem 2003;278:26380-90. 89:2973-6. terone. J Clin Endocrinol Metab 2004;89: 32. Draper N, Walker EA, Bujalska IJ, et al. 45. Urbanek M, Wu X, Vickery KR, et al. Al- 2408-13. Mutations in the genes encoding 11beta- lelic variants of the follistatin gene in poly- 57. Roldan B, San Millan JL, Escobar-Morre- hydroxysteroid dehydrogenase type 1 and cystic ovary syndrome. J Clin Endocrinol ale HF. Genetic basis of metabolic abnormal- hexose-6-phosphate dehydrogenase inter- Metab 2000;85:4455-61. ities in polycystic ovary syndrome: implica- act to cause cortisone reductase deficiency. 46. Hickey T, Chandy A, Norman RJ. The tions for therapy. Am J Pharmacogenomics Nat Genet 2003;34:434-9. androgen receptor CAG repeat polymor- 2004;4:93-107. 33. Tucci S, Futterweit W, Concepcion ES, et phism and X-chromosome inactivation in 58. Carmina E. Genetic and environmental al. Evidence for association of polycystic Australian Caucasian women with infertility aspect of polycystic ovary syndrome. J Endo- ovary syndrome in caucasian women with a related to polycystic ovary syndrome. J Clin crinol Invest 2003;26:1151-9. marker at the insulin receptor gene locus. Endocrinol Metab 2002;87:161-5. 59. Glueck CJ, Papanna R, Wang P, Golden- J Clin Endocrinol Metab 2001;86:446-9. 47. Cousin P, Calemard-Michel L, Lejeune berg N, Sieve-Smith L. Incidence and treat- 34. Waterworth DM, Bennett ST, Gharani H, et al. Influence of SHBG gene pentanu- ment of metabolic syndrome in newly re- N, et al. Linkage and association of insulin cleotide TAAAA repeat and D327N poly- ferred women with confirmed polycystic gene VNTR regulatory polymorphism with morphism on serum sex hormone-binding ovarian syndrome. Metabolism 2003;52: polycystic ovary syndrome. Lancet 1997; globulin concentration in hirsute women. 908-15. 349:986-90. J Clin Endocrinol Metab 2004;89:917-24. 60. National Cholesterol Education Pro- 35. Vankova M, Vrbikova J, Hill M, Cinek O, 48. Xita N, Tsatsoulis A, Chatzikyriakidou gram (NCEP) Expert Panel on Detection, Bendlova B. Association of insulin gene A, Georgiou I. Association of the (TAAAA)n Evaluation, and Treatment of High Blood VNTR polymorphism with polycystic ovary repeat polymorphism in the sex hormone- Cholesterol in Adults (Adult Treatment Pan- syndrome. Ann N Y Acad Sci 2002;967:558- binding globulin (SHBG) gene with poly- el III). Third Report of the National Choles- 65. cystic ovary syndrome and relation to SHBG terol Education Program (NCEP) Expert 36. Michelmore K, Ong K, Mason S, et al. serum levels. J Clin Endocrinol Metab 2003; Panel on Detection, Evaluation, and Treat- Clinical features in women with polycystic 88:5976-80. ment of High Blood Cholesterol in Adults ovaries: relationships to insulin sensitivity, 49. Carey AH, Waterworth D, Patel K, et al. (Adult Treatment Panel III): final report. Cir- insulin gene VNTR and birth weight. Clin Polycystic ovaries and premature male pat- culation 2002;106:3143-421. Endocrinol (Oxf ) 2001;55:439-46. tern baldness are associated with one allele 61. Sam S, Dunaif A. Polycystic ovary syn- 37. Ehrmann DA, Tang X, Yoshiuchi I, Cox of the steroid metabolism gene CYP17. drome: syndrome XX? Trends Endocrinol NJ, Bell GI. Relationship of insulin receptor Hum Mol Genet 1994;3:1873-6. Metab 2003;14:365-70. substrate-1 and -2 genotypes to phenotypic 50. Franks S, White D, Gilling-Smith C, 62. Azziz R, Ehrmann D, Legro RS, et al. features of polycystic ovary syndrome. J Clin Carey A, Waterworth D, Williamson R. Hy- Troglitazone improves ovulation and hirsut- Endocrinol Metab 2002;87:4297-300. persecretion of androgens by polycystic ova- ism in the polycystic ovary syndrome: a mul- 38. El Mkadem SA, Lautier C, Macari F, et al. ries: the role of genetic factors in the regula- ticenter, double blind, placebo-controlled Role of allelic variants Gly972Arg of IRS-1 tion of cytochrome P450c17 alpha. Baillieres trial. J Clin Endocrinol Metab 2001;86: and Gly1057Asp of IRS-2 in moderate-to- Clin Endocrinol Metab 1996;10:193-203. 1626-32. severe insulin resistance of women with 51. Liovic M, Prezelj J, Kocijancic A, Majdic 63. Franks S. Polycystic ovary syndrome: polycystic ovary syndrome. Diabetes 2001; G, Komel R. CYP17 gene analysis in hyper- a changing perspective. Clin Endocrinol 50:2164-8. androgenised women with and without ex- (Oxf ) 1989;31:87-120. 39. Ehrmann DA, Schwarz PE, Hara M, et aggerated 17-hydroxyprogesterone response 64. Conway GS, Honour JW, Jacobs HS. al. Relationship of calpain-10 genotype to to ovarian stimulation. J Endocrinol Invest Heterogeneity of the polycystic ovary syn- phenotypic features of polycystic ovary syn- 1997;20:189-93. drome: clinical, endocrine and ultrasound drome. J Clin Endocrinol Metab 2002;87: 52. Techatraisak K, Conway GS, Rumsby G. features in 556 patients. Clin Endocrinol 1669-73. Frequency of a polymorphism in the regula- (Oxf ) 1989;30:459-70. 40. Gonzalez A, Abril E, Roca A, et al. Spe- tory region of the 17 alpha-hydroxylase- 65. Carmina E, Legro RS, Stamets K, Lowell cific CAPN10 gene haplotypes influence the 17,20-lyase (CYP17) gene in hyperandro- J, Lobo RA. Difference in body weight be-

1234 n engl j med 352;12 www.nejm.org march 24, 2005 medical progress tween American and Italian women with with a history of gestational diabetes. Clin Lapidus L, Lindstedt G, Tengborn L. Hemo- polycystic ovary syndrome: influence of the Endocrinol (Oxf ) 2000;53:501-7. static and metabolic variables in women diet. Hum Reprod 2003;18:2289-93. 80. Zimmermann S, Phillips RA, Dunaif A, with polycystic ovary syndrome. Fertil Steril 66. Mokdad AH, Ford ES, Bowman BA, et et al. Polycystic ovary syndrome: lack of hy- 1994;61:455-60. al. Prevalence of obesity, diabetes, and obe- pertension despite profound insulin resis- 93. Atiomo WU, Bates SA, Condon JE, Shaw sity-related health risk factors, 2001. JAMA tance. J Clin Endocrinol Metab 1992;75: S, West JH, Prentice AG. The plasminogen 2003;289:76-9. 508-13. activator system in women with polycystic 67. Legro RS, Kunselman AR, Dodson WC, 81. Dahlgren E, Janson PO, Johansson S, ovary syndrome. Fertil Steril 1998;69:236- Dunaif A. Prevalence and predictors of risk Lapidus L, Oden A. Polycystic ovary syn- 41. for type 2 diabetes mellitus and impaired drome and risk for myocardial infarction: 94. Talbott E, Guzick D, Clerici A, et al. Cor- glucose tolerance in polycystic ovary syn- evaluated from a risk factor model based on a onary heart disease risk factors in women drome: a prospective, controlled study in prospective population study of women. Acta with polycystic ovary syndrome. Arterioscler 254 affected women. J Clin Endocrinol Obstet Gynecol Scand 1992;71:599-604. Thromb Vasc Biol 1995;15:821-6. Metab 1999;84:165-9. 82. Dahlgren E, Johansson PO, Lindstedt G, 95. Wild S, Pierpoint T, Jacobs H, McKeigue 68. Krosnick A. The diabetes and obesity et al. Women with polycystic ovary syn- P. Long-term consequences of polycystic epidemic among the Pima Indians. N J Med drome wedge resected in 1956 to 1965: ovary syndrome: results of a 31 year follow- 2000;97:31-7. a long-term follow-up focusing on natural up study. Hum Fertil (Camb) 2000;3:101-5. 69. Norman RJ, Masters L, Milner CR, history and circulating hormones. Fertil 96. Talbott EO, Zborowskii JV, Boudraux Wang JX, Davies MJ. Relative risk of conver- Steril 1992;57:505-13. MY. Do women with polycystic ovary syn- sion from normoglycaemia to impaired glu- 83. Kelly CJG, Speirs A, Gould GW, Petrie drome have an increased risk of cardiovas- cose tolerance or non-insulin dependent JR, Lyall H, Connell JMC. Altered vascular cular disease? Review of the evidence. Min- diabetes mellitus in polycystic ovarian syn- function in young women with polycystic erva Ginecol 2004;56:27-39. drome. Hum Reprod 2001;16:1995-8. ovary syndrome. J Clin Endocrinol Metab 97. Legro RS. Polycystic ovary syndrome 70. Dunaif A, Segal KR, Futterweit W, 2002;87:742-6. and cardiovascular disease: a premature as- Dobrjansky A. Profound peripheral insulin 84. Paradisi G, Steinberg HO, Hempfling A, sociation? Endocr Rev 2003;24:302-12. resistance, independent of obesity, in poly- et al. Polycystic ovary syndrome is associat- 98. Vgontzas AN, Legro RS, Bixler EO, cystic ovary syndrome. Diabetes 1989;38: ed with endothelial dysfunction. Circulation Grayev A, Kales A, Chrousos GP. Polycystic 1165-74. 2001;103:1410-5. ovary syndrome is associated with obstruc- 71. Dunaif A. Insulin resistance and the 85. Paradisi G, Steinberg HO, Shepard MK, tive sleep apnea and daytime sleepiness: role polycystic ovary syndrome: mechanism and Hook G, Baron AD. Troglitazone therapy of insulin resistance. J Clin Endocrinol implications for pathogenesis. Endocr Rev improves endothelial function to near nor- Metab 2001;86:517-20. 1997;18:774-800. mal levels in women with polycystic ovary 99. Fogel RB, Malhotra A, Pillar G, Pittman 72. Dunaif A, Wu X, Lee A, Diamanti-Kan- syndrome. J Clin Endocrinol Metab 2003; SD, Dunaif A, White DP. Increased preva- darakis E. Defects in insulin receptor signal- 88:576-80. lence of obstructive sleep apnea syndrome ing in vivo in the polycystic ovary syndrome 86. Orio F Jr, Palomba S, Cascella T, et al. in obese women with polycystic ovary syn- (PCOS). Am J Physiol Endocrinol Metab Early impairment of endothelial structure drome. J Clin Endocrinol Metab 2001;86: 2001;281:E392-E399. and function in young normal-weight wom- 1175-80. 73. Kahn SE, Prigeon RL, McCulloch DK, et en with polycystic ovary syndrome. J Clin 100. Gopal M, Duntley S, Uhles M, Attarian al. Quantification of the relationship be- Endocrinol Metab 2004;89:4588-93. H. The role of obesity in the increased prev- tween insulin sensitivity and B-cell function 87. Mather KJ, Verma S, Corenblum B, alence of obstructive sleep apnea syndrome in human subjects: evidence for a hyperbolic Anderson TJ. Normal endothelial function in patients with polycystic ovary syndrome. function. Diabetes 1993;42:1663-72. despite insulin resistance in healthy women Sleep Med 2002;3:401-4. 74. Polonsky KS, Sturis J, Bell GI. Non- with the polycystic ovary syndrome. J Clin 101. Hardiman P, Pillay OC, Atiomo W. insulin-dependent diabetes mellitus — Endocrinol Metab 2000;85:1851-6. Polycystic ovary syndrome and endometrial a genetically programmed failure of the beta 88. Yildiz BO, Haznedaroglu IC, Kirazli S, carcinoma. Lancet 2003;361:1810-2. [Erra- cell to compensate for insulin resistance. Bayraktar M. Global fibrinolytic capacity is tum, Lancet 2003;362:1082.] N Engl J Med 1996;334:777-83. decreased in polycystic ovary syndrome, 102. Balen A. Polycystic ovary syndrome 75. O’Meara NM, Blackman JD, Ehrmann suggesting a prothrombotic state. J Clin En- and cancer. Hum Reprod Update 2001;7: DA, et al. Defects in beta-cell function in docrinol Metab 2002;87:3871-5. 522-5. functional ovarian hyperandrogenism. J Clin 89. Orio F Jr, Palomba S, Spinelli L, et al. 103. Kuhl H. Comparative pharmacology of Endocrinol Metab 1993;76:1241-7. The cardiovascular risk of young women newer progestogens. Drugs 1996;51:188- 76. Ehrmann DA, Breda E, Cavaghan MK, et with polycystic ovary syndrome: an observa- 215. al. Insulin secretory responses to rising and tional, analytical, prospective case-control 104. Krattenmacher R. Drospirenone: phar- falling glucose concentrations are delayed study. J Clin Endocrinol Metab 2004;89: macology and pharmacokinetics of a unique in subjects with impaired glucose tolerance. 3696-701. progestogen. Contraception 2000;62:29-38. Diabetologia 2002;45:509-17. 90. Christian RC, Dumesic DA, Behrenbeck 105. Guido M, Romualdi D, Giuliani M, et 77. Ehrmann DA, Breda E, Corcoran MC, T, Oberg AL, Sheedy PF II, Fitzpatrick LA. al. Drospirenone for the treatment of hirsute et al. Impaired beta-cell compensation to Prevalence and predictors of coronary artery women with polycystic ovary syndrome: dexamethasone-induced hyperglycemia in calcification in women with polycystic ovary a clinical, endocrinological, metabolic pi- women with polycystic ovary syndrome. syndrome. J Clin Endocrinol Metab 2003; lot study. J Clin Endocrinol Metab 2004;89: Am J Physiol Endocrinol Metab 2004;287: 88:2562-8. 2817-23. E241-E246. 91. Ehrmann D, Schneider DJ, Sobel BE, et 106. Diamanti-Kandarakis E, Baillargeon 78. Dunaif A, Finegood DT. b-cell dysfunc- al. Troglitazone improves defects in insulin J-P, Iuorno MJ, Jakubowicz DJ, Nestler JE. tion independent of obesity and glucose in- action, insulin secretion, ovarian steroido- A modern medical quandary: polycystic ova- tolerance in the polycystic ovary syndrome. genesis, and fibrinolysis in women with ry syndrome, insulin resistance, and oral J Clin Endocrinol Metab 1996;81:942-7. polycystic ovary syndrome. J Clin Endo- contraceptive pills. J Clin Endocrinol Metab 79. Kousta E, Cela E, Lawrence N, et al. The crinol Metab 1997;82:2108-16. 2003;88:1927-32. prevalence of polycystic ovaries in women 92. Dahlgren E, Janson PO, Johansson S, 107. Golland IM, Elstein ME. Results of an

n engl j med 352;12 www.nejm.org march 24, 2005 1235 medical progress

open one-year study with Diane-35 in wom- sensitivity and luteinizing hormone. J Clin Sieve-Smith L, Wang P. Continuing metfor- en with polycystic ovarian syndrome. Ann Endocrinol Metab 1999;84:1470-4. min throughout pregnancy in women with N Y Acad Sci 1993;687:263-71. 121. Hoeger KM, Kochman L, Wixom N, polycystic ovary syndrome appears to safely 108. Spritzer PM, Lisboa KO, Mattiello S, Craig K, Miller RK, Guzick DS. A random- reduce first-trimester spontaneous abortion: Lhullier F. Spironolactone as a single agent ized, 48-week, placebo-controlled trial of a pilot study. Fertil Steril 2001;75:46-52. for long-term therapy of hirsute patients. intensive lifestyle modification and/or met- 133. Mitwally MF, Witchel SF, Casper RF. Clin Endocrinol (Oxf ) 2000;52:587-94. formin therapy in overweight women with Troglitazone: a possible modulator of ovari- 109. Muderris II, Bayram F, Guven M. Treat- polycystic ovary syndrome: a pilot study. an steroidogenesis. J Soc Gynecol Investig ment of hirsutism with lowest-dose fluta- Fertil Steril 2004;82:421-9. 2002;9:163-7. mide (62.5 mg/day). Gynecol Endocrinol 122. Nestler JE, Jakubowicz DJ, Evans WS, 134. Dunaif A, Scott D, Finegood D, Quin- 2000;14:38-41. Pasquali R. Effects of metformin on sponta- tana B, Whitcomb R. The insulin-sensitiz- 110. Moghetti P, Tosi F, Tosti A, et al. Com- neous and clomiphene-induced ovulation in ing agent troglitazone improves metabolic parison of spironolactone, flutamide, and the polycystic ovary syndrome. N Engl J Med and reproductive abnormalities in the poly- finasteride efficacy in the treatment of hir- 1998;338:1876-80. cystic ovary syndrome. J Clin Endocrinol sutism: a randomized, double blind, place- 123. Moran LJ, Noakes M, Clifton PM, Metab 1996;81:3299-306. bo-controlled trial. J Clin Endocrinol Metab Tomlinson L, Galletly C, Norman RJ. Di- 135. Ghazeeri G, Kutteh WH, Bryer-Ash 2000;85:89-94. etary composition in restoring reproductive M, Haas D, Ke RW. Effect of rosiglitazone 111. Gambineri A, Pelusi C, Genghini S, et and metabolic physiology in overweight on spontaneous and clomiphene citrate- al. Effect of flutamide and metformin ad- women with polycystic ovary syndrome. induced ovulation in women with polycystic ministered alone or in combination in diet- J Clin Endocrinol Metab 2003;88:812-9. ovary syndrome. Fertil Steril 2003;79:562-6. ing obese women with polycystic ovary 124. Stamets K, Taylor DS, Kunselman A, 136. Belli SH, Graffigna MN, Oneto A, Ote- syndrome. Clin Endocrinol (Oxf ) 2004;60: Demers LM, Pelkman CL, Legro RS. A ran- ro P, Schurman L, Levalle OA. Effect of 241-9. domized trial of the effects of two types of rosiglitazone on insulin resistance, growth 112. Azziz R, Black V, Hines GA, Fox LM, short-term hypocaloric diets on weight loss factors, and reproductive disturbances in Boots LR. Adrenal androgen excess in the in women with polycystic ovary syndrome. women with polycystic ovary syndrome. polycystic ovary syndrome: sensitivity and Fertil Steril 2004;81:630-7. Fertil Steril 2004;81:624-9. responsivity of the hypothalamic-pituitary- 125. Mansfield R, Galea R, Brincat M, Hole 137. Romualdi D, Guido M, Ciampelli M, et adrenal axis. J Clin Endocrinol Metab 1998; D, Mason H. Metformin has direct effects on al. Selective effects of pioglitazone on insu- 83:2317-23. human ovarian steroidogenesis. Fertil Steril lin and androgen abnormalities in normo- 113. Azziz R, Black VY, Knochenhauer ES, 2003;79:956-62. and hyperinsulinaemic obese patients with Hines GA, Boots LR. Ovulation after gluco- 126. Attia GR, Rainey WE, Carr BR. Met- polycystic ovary syndrome. Hum Reprod corticoid suppression of adrenal androgens formin directly inhibits androgen produc- 2003;18:1210-8. in the polycystic ovary syndrome is not pre- tion in human thecal cells. Fertil Steril 2001; 138. Knowler WC, Barrett-Connor E, Fowl- dicted by the basal dehydroepiandrosterone 76:517-24. er SE, et al. Reduction in the incidence of sulfate level. J Clin Endocrinol Metab 1999; 127. Lord JM, Flight IHK, Norman RJ. In- type 2 diabetes with lifestyle intervention or 84:946-50. sulin-sensitising drugs (metformin, trogli- metformin. N Engl J Med 2002;346:393- 114. Liang T, Heiss CE, Ostrove S, Rasmus- tazone, rosiglitazone, pioglitazone, D-chi- 403. son GH, Cheung A. Binding of a 4-methyl- ro-inositol) for polycystic ovary syndrome. 139. Tuomilehto J, Lindstrom J, Eriksson 4-aza-steroid to 5 alpha-reductase of rat liv- Cochrane Database Syst Rev 2003;3: JG, et al. Prevention of type 2 diabetes melli- er and prostate microsomes. Endocrinology CD003053. tus by changes in lifestyle among subjects 1983;112:1460-8. 128. Crave JC, Fimbel S, Lejeune H, Cug- with impaired glucose tolerance. N Engl J 115. Falsetti L, Gambera A, Legrenzi L, Ia- nardey N, Dechaud H, Pugeat M. Effects of Med 2001;344:1343-50. cobello C, Bugari G. Comparison of finas- diet and metformin administration on sex 140. American Diabetes Association. teride versus flutamide in the treatment of hormone-binding globulin, androgens, and Screening for type 2 diabetes. Diabetes Care hirsutism. Eur J Endocrinol 1999;141:361- insulin in hirsute and obese women. J Clin 2004;27:Suppl 1:S11-S14. 7. Endocrinol Metab 1995;80:2057-62. 141. Dunaif A, Sorbara L, Delson R, Green 116. Yildiz BO, Chang W, Azziz R. Polycys- 129. Glueck CJ, Goldenberg N, Pranikoff J, G. Ethnicity and polycystic ovary syndrome tic ovary syndrome and ovulation induction. Loftspring M, Sieve L, Wang P. Height, are associated with independent and additive Minerva Ginecol 2003;55:425-39. weight, and motor-social development dur- decreases in insulin action in Caribbean- 117. Barbieri RL, Gargiulo AR. Metformin ing the first 18 months of life in 126 infants Hispanic women. Diabetes 1993;42:1462-8. for the treatment of the polycystic ovary born to 109 mothers with polycystic ovary 142. Legro RS, Finegood D, Dunaif A. syndrome. Minerva Ginecol 2004;56:63- syndrome who conceived on and continued A fasting glucose to insulin ratio is a useful 79. metformin through pregnancy. Hum Re- measure of insulin sensitivity in women 118. Franks S. Assessment and manage- prod 2004;19:1323-30. with polycystic ovary syndrome. J Clin Endo- ment of anovulatory infertility in polycystic 130. Glueck CJ, Wang P, Kobayashi S, Phil- crinol Metab 1998;83:2694-8. ovary syndrome. Endocrinol Metab Clin lips H, Sieve-Smith L. Metformin therapy 143. Legro RS, Castracane VD, Kauffman North Am 2003;32:639-51. throughout pregnancy reduces the develop- RP. Detecting insulin resistance in polycys- 119. Nestler JE, Stovall D, Akhter N, Iuorno ment of gestational diabetes in women with tic ovary syndrome: purposes and pitfalls. MJ, Jakubowicz DJ. Strategies for the use of polycystic ovary syndrome. Fertil Steril Obstet Gynecol Surv 2004;59:141-54. insulin-sensitizing drugs to treat infertility 2002;77:520-5. 144. Diamanti-Kandarakis E, Kouli C, Alex- in women with polycystic ovary syndrome. 131. Glueck CJ, Wang P, Goldenberg N, andraki K, Spina G. Failure of mathematical Fertil Steril 2002;77:209-15. Sieve-Smith L. Pregnancy outcomes among indices to accurately assess insulin resis- 120. Huber-Buchholz MM, Carey DG, Nor- women with polycystic ovary syndrome tance in lean, overweight, or obese women man RJ. Restoration of reproductive po- treated with metformin. Hum Reprod 2002; with polycystic ovary syndrome. J Clin Endo- tential by lifestyle modification in obese 17:2858-64. crinol Metab 2004;89:1273-6. polycystic ovary syndrome: role of insulin 132. Glueck CJ, Phillips H, Cameron D, Copyright © 2005 Massachusetts Medical Society.

1236 n engl j med 352;12 www.nejm.org march 24, 2005

The new england journal of medicine

images in clinical medicine

Renal Papillary Necrosis Unmasking Sickle Cell Disease

No loss of cortex

Spurs Hooks

Ring sign (sloughed papilla)

26-year-old healthy greek woman was evaluated because of Michalis Voulgarelis, M.D., Ph.D. a new-onset painless hematuria. The physical examination was unremarkable. Panayiotis D. Ziakas, M.D., Ph.D. Laboratory data were notable for a hemoglobin concentration of 7.6 g per National University of Athens deciliter and a serum creatinine level of 0.7 mg per deciliter (61.9 µmol per liter ). Pye- Medical School lography that was performed with the use of contrast material revealed bilateral defor- 11527 Athens, Greece mities of the renal papillae, including hooks, spurs, and ring-shaped filling defects (sloughed papilla) of the calyxes (arrows), suggestive of papillary necrosis. Interstitial nephritides, such as pyelonephritis and nephritis due to obstructive uropathy, diabetes mellitus, or analgesic abuse, that are associated with papillary necrosis were ruled out. Evaluation of a blood specimen revealed sporadic sickle cells. Electrophoresis of the hemoglobin established the diagnosis of sickle cell–b+ thalassemia. The patient’s hematuria resolved after bed rest, vigorous hydration with hypotonic fluids, transfusions, and alkalization of the urine. Papillary necrosis due to ischemic damage to the papillae is a well-known complication of sickle cell syndromes but is uncommon at presentation. Sickle cell disease should be considered in patients with unexplained hematuria. Copyright © 2005 Massachusetts Medical Society.

Web-only Images in Clinical Medicine are published every week in the Journal. They are listed (with e page numbers) in the table of contents on the cover of the printed Journal and can be seen at www.nejm.org.

n engl j med 352;12 www.nejm.org march 24, 2005 1237

The new england journal of medicine

case records of the massachusetts general hospital

Founded by Richard C. Cabot Nancy Lee Harris, m.d., Editor Jo-Anne O. Shepard, m.d., Associate Editor Stacey M. Ellender, Assistant Editor Sally H. Ebeling, Assistant Editor Christine C. Peters, Assistant Editor

Case 9-2005: A 67-Year-Old Man with Acute Respiratory Failure

David M. Systrom, M.D., and Conrad Wittram, M.B., Ch.B.

presentation of case

From the Departments of Pulmonary and A 67-year-old man was transferred to this hospital with acute respiratory failure. Critical Care (D.M.S.) and Radiology (C.W.), Four days before admission, the patient had visited his primary care physician be- Massachusetts General Hospital; and the Departments of Medicine (D.M.S.) and Ra- cause of a three-month history of chest congestion and cough productive of sputum diology (C.W.), Harvard Medical School. that was initially clear but then became more opaque. His chest felt tight when he breathed deeply, and deep inspirations triggered the cough. During the week before N Engl J Med 2005;352:1238-46. Copyright © 2005 Massachusetts Medical Society. the physician visit, he had had slight nasal congestion, poor appetite, and decreased energy, but no fever, chills, nocturnal sweats, or postnasal drip. His physician obtained a chest radiograph, which showed increased markings at the base of the right lung; a computed tomographic (CT) scan of the chest was recommended. A combination in- haler consisting of salmeterol and fluticasone was prescribed. Over the next 24 hours, severe pleuritic chest pain developed, with productive cough and severe shortness of breath. The patient went to the emergency department of another hospital. On examination there, he was alert and afebrile, with a respiratory rate of 24 breaths per minute and somewhat labored breathing. Pulse oximetry showed an oxygen saturation of 86 percent while he was breathing ambient air. Bilateral rales were heard on examination of the lungs — although chest pain prevented deep inspiration — but no wheezing. The remainder of the physical examination was unremarkable. A chest radiograph showed opacifications that now involved the left lower lobe, in addition to those previously seen in the right middle and lower lobes. His white-cell count was 17,800 per cubic millimeter, with 74 percent neutrophils and 7 percent band forms; the hematocrit was 32.9 percent. The levels of serum electrolytes and cardiac enzymes and the results of tests of liver and renal function were all within normal ranges. He was admitted to the intensive care unit. Two chest radiographs over the next 24 hours showed progressive worsening of the opacifications to include both upper lobes. The patient’s respiratory rate increased, his dyspnea worsened, and his oxygen saturation dropped to 76 percent while he was breathing 100 percent oxygen with the use of a face mask; the trachea was intubated, and he was placed on ventilator support. Treatment with levofloxacin was begun, but as his condition deteriorated the antibiotic agents were changed to ceftriaxone and azithromycin, and he received two doses each of tetracycline and erythromycin. A spu-

1238 n engl j med 352;12 www.nejm.org march 24, 2005

case records of the massachusetts general hospital tum culture grew normal flora, and two blood cul- afebrile. Cultures of blood and urine and nasal and tures showed no growth. He was transferred to the rectal secretions were negative for bacteria, includ- intensive care unit at this institution. ing pneumocystis and legionella; chlamydia and The patient had a history of gastroesophageal re- mycoplasma were not detected by polymerase chain flux disease and chronic back pain, for which he reaction. There were no antigens of adenovirus or took esomeprazole and rofecoxib. He had had an influenza, parainfluenza, or respiratory syncytial vi- appendectomy, bilateral inguinal hernia repairs, a rus. The results of fungal and mycobacterial cul- transurethral resection of the prostate, and removal tures were pending. Serologic-test results are shown of benign colon polyps. He was married with two in Table 1. On the third day, the oxygen-saturation children. He was a retired engineer who had worked percentages improved, and the trachea was extu- in a metal machine shop. He had no known toxic ex- bated; the oxygen saturation was 99 percent while posures, although he recalled a coworker becoming the patient was breathing 40 percent oxygen by face very ill from an undefined respiratory illness in the mask. On the fourth day, he was transferred to the 1970s. He had never smoked and rarely drank alco- general medical service. On the fifth hospital day, a hol. He had not traveled outside the United States chest radiograph showed slight improvement in nor had he been in contact with anyone who had the bilateral air-space opacifications, particularly traveled recently. He swam regularly to help with in the left upper lobe. His condition continued to his back pain. He had recently rewired his daugh- improve, and he was discharged on the sixth hos- ter’s attic, but had not been exposed to excessive pital day. dust or mold. He had no pets and did not hunt. His Three days after discharge, fever up to 38.9°C, wife reported that several weeks before the onset of increased shortness of breath, and cough developed this illness, he had been exposed to bird droppings after the patient had been sitting in his hot tub. He when he was changing an outdoor bird feeder and was readmitted to the other hospital, where his a gust of wind blew the droppings into his face. symptoms rapidly abated, and he was discharged af- The temperature was 36.7°C, the blood pressure 113/54 mm Hg, and the pulse 79 beats per minute. The oxygen saturation was 97 percent with the frac- Table 1. Results of Serologic Tests. tion of inspired oxygen at 0.5. On physical exami- Test Result nation, he was a well-developed man, who was intubated and sedated. The results of an examination Varicella antibody Positive IgG of the head, eyes, ears, nose, and throat were unre- Serum cryptococcal antigen Negative markable. Auscultation of the lungs disclosed di- Histoplasma capsulatum urine antigen Negative minished breath sounds throughout both lungs Chlamydia psittaci antibody Negative with bronchial breath sounds at the bases. An ex- Q fever antibody Negative amination of the heart, abdomen, and extremities Aspergillus fumigatus 3 antibody Negative disclosed no abnormalities. A. fumigatus 6 antibody Negative The white-cell count was 13,900 per cubic milli- Thermoactinomyces sacchari antibody Negative meter; the hematocrit was 26.1 percent. A chest ra- T. candidus Negative diograph obtained with portable equipment showed T. vulgaris antibody Negative bilateral air-space consolidation, more on the left Streptococcus viridans antibody Negative than the right, and a probable left pleural effusion. An electrocardiogram showed no abnormalities. Micropolyspora faeni antibody Negative Bronchoscopy and bronchoalveolar lavage were Pigeon serum antibody Negative performed. Thin secretions were noted and the Enzyme-linked immunoassay for human Negative immunodeficiency virus sputum was not grossly purulent. The lavage fluid contained 367 white cells per cubic millimeter, with Cold agglutinins Nonreactive at 1:16 titer 53 percent neutrophils, 21 percent lymphocytes, Blastomycosis complement-fixation (CF) test Negative 4 percent monocytes, and 1 percent eosinophils. Coccidioidomycosis CF test Negative Doxycycline and levofloxacin were administered. Histoplasmosis CF test: yeast antigen Positive 1:16 titer On the second and third hospital days, chest ra- Histoplasmosis CF test: histoplasmin antigen Negative diographs showed no change. The patient remained

n engl j med 352;12 www.nejm.org march 24, 2005 1239

The new england journal of medicine

ter several days. One month later, he was seen at a follow-up appointment at the infectious disease A clinic of this hospital. He reported that he routinely treated the hot tub with bromine. He had had no further recurrence of respiratory symptoms and had resumed his exercise regimen. CT scanning of the chest showed diffuse ground-glass opacifications with a few secondary pulmonary lobules that had been spared and that may have represented a minor degree of air trapping. On high-resolution CT, centrilobular nodules were seen, which predominantly affected the upper lobes diffusely and the middle, lingular, and lower lobes more peripherally. No mediastinal lymphadenopathy or pleural disease was present. The results of a diagnostic test were reported.

differential diagnosis B

Dr. David M. Systrom: May we review the chest radiographs, please? Dr. Conrad Wittram: The first chest radiograph shows that the patient is intubated and has a nasogastric tube. There is bilateral diffuse ground-glass opacification with consolidation affecting the left upper and right lower lobes. The left hemidiaphragm is obscured, indicating collapse of the left lower lobe, and a small amount of left-sided pleural fluid is present (Fig. 1A). A radiograph obtained on the fifth hospital day shows clearing of the opacity in the left upper lobe. There are linear opacities within the right lower lobe suggesting subseg- Figure 1. Chest Radiographs Obtained on the First mental atelectasis. The left lower lobe has partially and Fifth Hospital Days. reexpanded; there appears to be some consolida- On the first hospital day, the patient is intubated and tion within the middle lobe laterally, and there are there is a nasogastric tube (Panel A). There is bilateral small bilateral pleural effusions (Fig. 1B). diffuse ground-glass opacification with consolidation The high-resolution CT scan performed one affecting the left upper and right lower lobes. The left month later shows centrilobular ground-glass opac- hemidiaphragm is obscured, indicating collapse of the left lower lobe. There is a small amount of left-sided ities within the upper lobes, some of which form a pleural fluid. On the fifth hospital day, there is improve- rosette pattern (Fig. 2). Some secondary pulmo- ment of the diffuse ground-glass opacification and the nary lobules have been spared. left upper lobe consolidation (Panel B). A new patch of Dr. Systrom: This previously healthy man’s respi- ground-glass opacification is present in the right upper ratory illness developed first in a subacute phase, lobe. A small region of ground-glass opacification persists. Subsegmental atelectasis affects both lower lobes, followed by a fulminant phase. Important clinical worse on the left. The left lower lobe collapse appears and laboratory clues to the differential diagnosis to have improved. Opacification of the middle lobe ap- include the rapid and apparently spontaneous res- pears marginally worse laterally, which may be related olution of hypoxemic respiratory failure, certain en- to consolidation or atelectasis. There are small bilateral vironmental exposures, and the results of broncho- pleural effusions. alveolar lavage. The high-resolution CT scan of the

1240 n engl j med 352;12 www.nejm.org march 24, 2005

case records of the massachusetts general hospital

pulmonary edema Noncardiogenic pulmonary edema due to acute inhalational injury deserves mention, given some of this patient’s reported exposures. At the time of his second hospitalization, he described the use of a hot tub that had been disinfected with bromine, raising the specter of inhalational injury. In 1997, Burns and Linden2 reported two cases of acute pneumonitis attributed to exposure to bromine and hydrobromic acid in a hot tub. Although I believe the hot-tub exposure may be important in this case, I shall dismiss acute inhalational injury, because the subsequent high-resolution CT image ei- Figure 2. High-Resolution CT Scan Obtained One Month ther should have been completely clear or should after Admission. have shown nonspecific changes suggesting fibro- There are centrilobular ground-glass opacities, some of sis; diffuse centrilobular nodules would not have 3 which have a rosette pattern (arrows). A few secondary been expected. pulmonary lobules appear to have been spared. pulmonary hemorrhage The reported drop in hematocrit in the absence of obvious gastrointestinal-tract blood loss could sug- chest that was obtained later and showed diffusely gest diffuse alveolar hemorrhage, even in the ab- distributed pulmonary nodules was also helpful. sence of hemoptysis. Diffuse alveolar hemorrhage Diffuse pulmonary nodules fall into four cate- can occur in disorders of hemostasis and when pul- gories.1 Nodules along pleural surfaces or abutting monary capillaries are disrupted by inflammation. the interlobular septa are designated “perilym- Examples include angiotrophic infections, such as phatic.” Ground-glass attenuation nodules along aspergillus, and vasculitis; the latter may be associ- 4 the arterial branch of the secondary lobule are des- ated with centrilobular nodules. This patient, how- ignated “centrilobular” and usually reflect disorders ever, had no known bleeding diathesis and, to our of the accompanying bronchiole. Nodules that are knowledge, he was immunocompetent. There were due to bronchiolar impaction and associated air- no reported extrapulmonary stigmata of vasculitis, space disease occur often in infection and are such as petechiae or purpura, an active urine sedi- termed “tree-in-bud” or “small airways disease– ment, or abnormal neurologic findings. The bron- associated.” Well-defined nodules that do not fit choalveolar-lavage fluid was not reported to be any of the above patterns are termed “random” and bloody. For these reasons and because the high- are seen in hematogenous metastases. resolution CT scan did not show patchy ground- An important clinical feature of this case was glass opacities or consolidation typical of diffuse rapidly reversible hypoxemic respiratory failure. alveolar hemorrhage, I will dismiss pulmonary On day 1 in the other hospital, the patient appeared hemorrhage as a unifying diagnosis. to meet the criteria for the acute respiratory distress syndrome (ARDS), with bilateral air-space acute eosinophilic pneumonia disease on chest radiography, a ratio of the partial The differential diagnosis of ARDS that has no ap- pressure of arterial oxygen to the fraction of in- parent precipitant should always include acute eo- spired oxygen that was less than 200, and no clear- sinophilic pneumonia. Accepted diagnostic criteria cut cardiogenic cause. The causes of rapidly revers- include an acute febrile illness of short duration, ible ARDS include pulmonary edema, pulmonary acute lung injury (or ARDS), and eosinophilia in hemorrhage, acute eosinophilic pneumonia, atypi- specimens obtained from bronchoalveolar lavage cal infection, and diffuse parenchymal lung dis- (>25 percent) or in lung tissue, in the absence of ease with an accelerated phase. known causes (e.g., drugs). Like this patient, many

n engl j med 352;12 www.nejm.org march 24, 2005 1241

The new england journal of medicine

patients with acute eosinophilic pneumonia have One type of viral pneumonia deserves special progressive respiratory failure, and spontaneous mention because of the timing of this case. An out- improvement (without glucocorticoids) has been break of pneumonia due to a coronavirus variant, reported.5 Although acute eosinophilic pneumonia or what came to be known as the severe acute respi- is a logical diagnosis, I will dismiss it because the ratory syndrome (SARS), was peaking as this pa- high-resolution CT scan failed to show the typical tient became ill in 2003. Although severe cases of random pattern of ground-glass opacities, and be- SARS were described in the Far East and Canada, cause the examination of the bronchoalveolar- only three patients in the United States required lavage specimen showed only 1 percent eosinophils. mechanical ventilation.6 In Massachusetts, of eight suspected cases, none was confirmed. This patient atypical infection had no travel or contacts with other sick people, in- At the time of this patient’s initial presentation cluding those who had visited areas of endemic to the other hospital, he had a peripheral leukocy- disease. I will dismiss the diagnosis of SARS be- tosis with a leftward shift and a chest radiograph cause of the poor epidemiologic fit. that showed changes that suggested community- acquired pneumonia. These findings and scanty atypical bacterial pneumonia sputum production led his treating physicians to Is it possible that this patient’s clinical and radio- consider an atypical community-acquired pneumo- graphic course can be explained by incomplete anti- nia, which in its narrowest definition is reserved for biotic treatment of atypical community-acquired conditions due to mycoplasma, legionella, and chla- bacterial pneumonia? Atypical bacteria cause pneu- mydia. A more inclusive definition of atypical pneumonia in more than a third of outpatient cases.7 monia is shown in Table 2. Negative cultures from sputum and bronchoalveolar-lavage specimens, and negative results on a poly- viral pneumonia merase-chain-reaction assay for mycoplasma and The patient had nasal congestion and nonspecific chlamydia8,9 argue against this diagnosis. The pa- constitutional symptoms over the course of the tient’s exposure to bird droppings several weeks be- week preceding the initial admission, which sug- fore the onset of acute illness might suggest psitta- gested a viral infection. Viral bronchiolitis causes cosis. Polymerase chain reaction is less sensitive centrilobular nodules, but usually with at least for Chlamydia psittaci than for C. pneumoniae, and some acinar involvement and a tree-in-bud pattern therefore a negative test does not rule out this entity. as visualized by high-resolution CT. Moreover, the Psittacosis, however, occurs only rarely in wild tests for antigens were negative and rapid resolu- birds, and spontaneous recovery would not be ex- tion of hypoxemic respiratory failure in viral bron- pected.10 chiolitis would be unusual. acute fungal pneumonia Primary histoplasma pneumonia can occur in an Table 2. Some Atypical Community-Acquired immunocompetent host, especially after exposure Pneumonias. to a high inoculum of bird or bat droppings in an en- closed space. The usual symptoms are high-grade Viral Varicella fever, headache, nonproductive cough, chills, and Respiratory syncytial virus pleuritic chest pain days to weeks after exposure, Influenza A, B and most patients recover spontaneously after two Parainfluenza Adenovirus to three weeks. Massachusetts, however, is not an Severe acute respiratory syndrome area where histoplasma is endemic. In addition, it Bacterial was assumed that the patient’s daughter’s attic was Mycoplasma pneumoniae well ventilated and we know that he cleaned the Legionella species Chlamydia pneumoniae bird feeder outdoors. Serologic testing was nega- C. psittaci tive for fungi, as was the test for the Histoplasma Fungal capsulatum urine antigen and the bronchoalveolar- Histoplasma capsulatum lavage fungal stain. His recovery from ARDS without treatment with corticosteroid agents or anti-

1242 n engl j med 352;12 www.nejm.org march 24, 2005

case records of the massachusetts general hospital fungal agents was almost certainly too quick if his in days does not occur. Respiratory bronchiolitis illness had been fulminant H. capsulatum pneumo- interstitial lung disease is associated with centri- nia. Similar arguments can be made against the lobular ground-glass opacities and patchy areas of probability of the illness having been either blasto- hyperinflation, a pattern, revealed by high-resolu- mycosis or coccidioidomycosis. tion CT, that is identical to the pattern in this case. This patient never smoked, however, which would diffuse parenchymal lung disease make both respiratory bronchiolitis interstitial lung I now turn to the genre of disease that I believe best disease and the closely related desquamative inter- accounts for this patient’s presentation: diffuse pa- stitial pneumonia extremely improbable condi- renchymal lung disease with an accelerated phase. tions for him to have. Some examples, including the newly classified id- Usual interstitial pneumonia is increasingly di- iopathic interstitial pneumonias,11 are shown in agnosed by findings on high-resolution CT images Table 3. of bibasilar honeycombing (which suggests fibro- The patient’s medication list did not suggest a sis) and upper-lung irregular lines.13 Nonspecific drug-induced lung disease, nor were there extra- interstitial pneumonia is similar to usual interstitial pulmonary stigmata of an undiagnosed connective- pneumonia, although with less fibrosis and a bet- tissue disease. Silicoproteinosis can be manifested ter prognosis. The findings on high-resolution CT as recurrent pneumonia after intense inhalation of in this case suggested neither. Lymphocytic inter- crystalline silica, and a high-resolution CT of the stitial pneumonitis is a benign polyclonal prolifer- chest can show centrilobular nodules.12 I will dis- ation of mature B cells that can diffusely involve the miss this pneumoconiosis because the patient’s lung, yielding a pattern on high-resolution CT scan- history does not report a return to the metal shop ning similar to the pattern in this case14 and the and because there was no sign of the typical ground- finding of lymphocytosis in the bronchoalveolar- glass opacities and consolidation on high-resolu- lavage fluid. Lymphocytic interstitial pneumonitis, tion CT. however, usually occurs in the presence of connective-tissue disease or an immunodeficiency condi- idiopathic interstitial pneumonias tion; furthermore, spontaneous resolution would The several types of idiopathic interstitial pneu- not be expected in association with this type of monias11 share many features of this case, and all pneumonia. of them are capable of mounting an accelerated Finally, bronchiolitis obliterans organizing pneu- phase. Acute interstitial pneumonia, formerly monia, either idiopathic cryptogenic organizing known as the Hamman–Rich syndrome, does have pneumonia or as a nonspecific reaction of the lung a fulminant course, but spontaneous recovery with- to insults such as infection and connective-tissue disease, can be recurrent, as in this case, although spontaneous resolution would be unusual. In summary, although the idiopathic interstitial pneumo- Table 3. Diffuse Parenchymal Lung Diseases. nias share some of the clinical and radiographic Known causes features of this case, none offers a reasonable uni- Drugs fying diagnosis. Connective-tissue disease Pneumoconioses granulomatous lung disease Idiopathic interstitial pneumonias Acute interstitial pneumonitis Pulmonary sarcoidosis is one of the great masquer- Respiratory bronchiolitis interstitial lung disease aders in pulmonary disease. Signs and symptoms or desquamative interstitial pneumonitis Usual interstitial pneumonia that are associated with it include hilar and medias- Nonspecific interstitial pneumonitis tinal lymphadenopathy, upper-lobe interstitial Lymphocytic interstitial pneumonitis changes, fibrosis, and less commonly, pulmonary Cryptogenic organizing pneumonia parenchymal nodules or a fulminant vasculitis. The Granulomatous diseases high-resolution CT findings usually include peri- Sarcoidosis Langerhans’-cell histiocytosis lymphatic nodules rather than centrilobular nod- Hypersensitivity pneumonia ules, however, and the vasculitis does not resolve spontaneously. Langerhans’-cell histiocytosis is a

n engl j med 352;12 www.nejm.org march 24, 2005 1243

The new england journal of medicine

diffuse parenchymal lung disease with a propen- graphic features of this case. Although the culprit sity for the upper lobes of smokers and former antigen was not defined, the patient’s relapse after smokers. An accelerated phase is extremely unusu- soaking in his hot tub strongly suggests the entity, al, and the expected cystic lesions that would ap- “hot-tub lung.” pear on high-resolution CT scans were not found Hot-tub lung15-21 is thought to be a hypersensi- in this case. tive response to aerosolized Mycobacterium avium complex, a frequent contaminant of standing water. hypersensitivity pneumonitis Chlorine is not used in hot tubs because it bubbles The spontaneous resolution of ARDS over a three- off (in most forms). Bromine is used because of its day hospitalization and then the recurrence after an lower vapor pressure, but it is highly reactive and environmental exposure strongly suggest hypersen- most likely helps form the antigens with M. avium– sitivity pneumonitis. Hypersensitivity pneumonitis intracellulare that lead to hot-tub lung. In cases of is a granulomatous immunologic response of a sen- hot-tub lung, M. avium complex is sometimes cul- sitized host to inhaled biologic aerosols. The dis- tured from lung secretions, which may lead to ease is more common in nonsmokers and includes controversy about infection versus hypersensitivity constitutional symptoms, cough, and dyspnea that as the pathogenesis. Currently, most reports favor appear four to six hours after exposure. The physi- a hypersensitivity pathway, because the high-reso- cal examination reveals rales and, rarely, wheezing. lution CT image does not typically show the tree- Symptoms subside over hours to days upon remov- in-bud appearance characteristic of infection, and al of the affected person from exposure. The hall- hot-tub lung resolves quickly without antimicrobi- mark of acute and subacute hypersensitivity pneu- al therapy. monitis is recurrence with reexposure, although Thus, hypersensitivity pneumonitis due to in- with chronic exposure to the antigens, the pattern fection with M. avium complex seems an extremely of waxing and waning symptoms may be lost. probable diagnosis in this immunocompetent pa- On radiographic examination, acute hypersen- tient who had never smoked and who had a sub- sitivity pneumonitis may be observed as a fleeting, acute course of the condition followed by a fulmi- micronodular, interstitial pattern in the lower and nant phase, spontaneous remission, and recurrent middle lung zones, whereas the subacute and symptoms with repeated exposure to his hot tub; chronic forms involve the middle to upper lung lymphocytosis in the bronchoalveolar-lavage fluid; zones. The findings on the high-resolution CT and centrilobular nodules on CT scanning. Since scan of the patient’s chest are characteristic of this the hypersensitivity panel was negative and a lung condition: diffuse centrilobular micronodules with biopsy was not done, how could a specific diagno- ground-glass attenuation and focal air trapping sis have been made? My guess is that M. avium with only mild fibrotic changes or no changes. Af- complex was cultured, either from the patient’s ter acute exposure, the complete resolution of the bronchoalveolar-lavage fluid or from a sample of findings on high-resolution CT may take weeks. his hot-tub water. Laboratory findings may include a precipitating IgG Dr. Nancy Lee Harris (Pathology): Dr. Thorner, antigen, but a positive test may merely reflect expo- you saw this patient both during and after the hos- sure, and negative results of a standard antigen pitalization; would you give us your clinical im- panel (as was obtained in this case) may be mean- pressions and then tell us about the diagnostic test ingless. The finding obtained by bronchoalveolar result that was received? lavage of lymphocytosis of greater than 30 percent Dr. Anna R. Thorner (Infectious Disease): When is characteristic; flow cytometry may show an in- we saw the patient initially at this hospital, we version of the normal ratio of CD4+ to CD8+ thought that an infection, particularly one of the T cells (which helps distinguish hypersensitivity atypical causes of pneumonia, was possible, al- pneumonitis from sarcoidosis). Typical pathologi- though the rapid resolution was unusual. When we cal findings include poorly formed, noncaseating saw him in follow-up one month after his acute ill- granulomas and mononuclear-cell inflammation ness, he reported the recurrence of his symptoms in a peribronchial distribution. after exposure to the hot tub, and by that time, we Thus, hypersensitivity pneumonitis accounts had the results of a mycobacterial culture that had for nearly all of the clinical, laboratory, and radio- been obtained from the bronchoalveolar lavage.

1244 n engl j med 352;12 www.nejm.org march 24, 2005

case records of the massachusetts general hospital

clinical diagnosis Dr. Harris: Dr. Mark, would you comment on the pathological findings in hot-tub lung? Hypersensitivity pneumonitis due to M. avium com- Dr. Eugene J. Mark (Pathology): The characteristic plex infection (hot-tub lung). pathological feature of hypersensitivity pneumonitis is the presence of poorly formed bronchiolocen- dr. david m. systrom’s tric granulomas. However, in a case of hot-tub lung diagnosis seen at this hospital and reported previously in the Case Records,22 a lung biopsy showed bronchiolo- Hypersensitivity pneumonitis due to M. avium com- centric granulomas that were more distinct than plex infection (hot-tub lung). would be typical for hypersensitivity pneumonia, and acid-fast bacilli were found within some giant pathological discussion cells. M. avium–intracellulare was cultured from the lung and from the water from the hot tub. The ear- Dr. Thorner: The mycobacterial culture yielded lier patient was treated with both corticosteroid growth three weeks after it had been obtained, and therapy and antibiotic therapy for mycobacterial a nucleic acid hybridization probe identified the iso- infection, and the patient recovered. In that case, it late as M. avium–intracellulare complex. The patient’s was not clear, therefore, whether the clinical symp- pulmonary symptoms had resolved, and there were toms were manifestations of hypersensitivity, in- no signs of recurrence. We counseled him to avoid fection, or both. hot tubs, to which he agreed; he had already drained his hot tub. We obtained another sample of sputum anatomical diagnosis for an acid-fast bacilli smear and for a mycobacterial culture, both of which were negative. A chest CT Hypersensitivity pneumonitis due to M. avium scan two months after the acute illness showed complex infection (hot-tub lung). clearing of the parenchymal abnormalities. references 1. Gruden JF, Webb WR, Naidich DP, patients with community-acquired pneu- coproteinosis: high-resolution CT and his- McGuinness G. Multinodular disease: ana- monia: the frequency of atypical agents and tologic findings. J Thorac Imaging 2001;16: tomic localization at thin-section CT — clinical course. Am J Med 1996;101:508-15. 127-9. multireader evaluation of a simple algo- 8. Blackmore TK, Reznikov M, Gordon 13. Hunninghake GW, Lynch DA, Galvin rithm. Radiology 1999;210:711-20. DL. Clinical utility of the polymerase chain JR, et al. Radiologic findings are strongly 2. Burns MJ, Linden CH. Another hot tub reaction to diagnose Mycoplasma pneu- associated with a pathologic diagnosis of hazard: toxicity secondary to bromine and moniae infection. Pathology 1995;27:177- usual interstitial pneumonia. Chest 2003; hydrobromic acid exposure. Chest 1997; 81. 124:1215-23. 111:816-9. 9. Miyashita N, Saito A, Kohno S, et al. 14. Johkoh T, Muller NL, Pickford HA, et al. 3. Nobauer-Huhmann IM, Eibenberger K, Multiplex PCR for the simultaneous detec- Lymphocytic interstitial pneumonia: thin- Schaefer-Prokop C, et al. Changes in lung tion of Chlamydia pneumoniae, Mycoplasma section CT findings in 22 patients. Radiol- parenchyma after acute respiratory distress pneumoniae and Legionella pneumophila ogy 1999;212:567-72. syndrome (ARDS): assessment with high- in community-acquired pneumonia. Respir 15. Embil J, Warren P, Yakrus M, et al. Pul- resolution computed tomography. Eur Med 2004;98:542-50. monary illness associated with exposure to Radiol 2001;11:2436-43. 10. Verweij PE, Meis JF, Eijk R, Melchers WJ, Mycobacterium-avium complex in hot tub 4. Choi YH, Im JG, Han BK, Kim JH, Lee Galama JM. Severe human psittacosis re- water: hypersensitivity pneumonitis or infec- KY, Myoung NH. Thoracic manifestation of quiring artificial ventilation: case report and tion? Chest 1997;111:813-6. Churg-Strauss syndrome: radiologic and review. Clin Infect Dis 1995;20:440-2. 16. Kahana LM, Kay JM, Yakrus MA, Waser- clinical findings. Chest 2000;117:117-24. 11. American Thoracic Society/European man S. Mycobacterium avium complex 5. Philit F, Etienne-Mastroianni B, Parrot A, Respiratory Society International Multidis- infection in an immunocompetent young Guerin C, Robert D, Cordier JF. Idiopathic ciplinary Consensus Classification of the adult related to hot tub exposure. Chest acute eosinophilic pneumonia: a study of 22 Idiopathic Interstitial Pneumonias: this 1997;111:242-5. patients. Am J Respir Crit Care Med 2002; joint statement of the American Thoracic 17. Rickman OB, Ryu JH, Fidler ME, Kalra 166:1235-9. Society (ATS), and the European Respiratory S. Hypersensitivity pneumonitis associated 6. Update: outbreak of severe acute respi- Society (ERS) was adopted by the ATS board with Mycobacterium avium complex and ratory syndrome — worldwide, 2003. of directors, June 2001 and by the ERS Exec- hot tub use. Mayo Clin Proc 2002;77:1233- MMWR Morb Mortal Wkly Rep 2003;52: utive Committee, June 2001. Am J Respir 7. 269-72. Crit Care Med 2002;165:277-304. [Erratum, 18. Aksamit TR. Hot tub lung: infection, 7. Marrie TJ, Peeling RW, Fine MJ, Singer Am J Respir Crit Care Med 2002;166:426.] inflammation, or both? Semin Respir Infect DE, Coley CM, Kapoor WN. Ambulatory 12. Marchiori E, Ferreira A, Muller NL. Sili- 2003;18:33-9.

n engl j med 352;12 www.nejm.org march 24, 2005 1245

case records of the massachusetts general hospital

19. Pham RV, Vydareny KH, Gal AA. High- ers RA, Colby TV. Diffuse pulmonary disease tent host. Arch Intern Med 2003;163:845- resolution computed tomography appear- caused by nontuberculous mycobacteria in 8. ance of pulmonary Mycobacterium avium immunocompetent people (hot tub lung). 22. Case Records of the Massachusetts complex infection after exposure to hot tub: Am J Clin Pathol 2001;115:755-62. General Hospital (Case 27-2000). N Engl J case of hot-tub lung. J Thorac Imaging 21. Cappelluti E, Fraire AE, Schaefer OP. Med 2000;343:642-9. 2003;18:48-52. A case of “hot tub lung” due to Mycobacte- Copyright © 2005 Massachusetts Medical Society. 20. Khoor A, Leslie KO, Tazelaar HD, Helm- rium avium complex in an immunocompe-

slide sets for the case records available in digital format Any reader of the Journal who uses the Case Records of the Massachusetts General Hospital as a teaching exercise or reference material is eligible to receive digital images, with identifying legends, of pertinent radiographic, neurologic, and cardiac studies, gross specimens, and photomicrographs. The images on the CD for each case are in both PowerPoint and 300 dpi jpg format. For some cases, additional images that have not been selected for publication will be included on the CD. These images, which illustrate the current cases in the Journal, are mailed from the Department of Pathology to correspond to the week of publication and may be retained by the subscriber. Each year approximately 250 images from 40 cases are sent to each subscriber. The cost of the subscription is $450 per year. 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 [email protected]. Images from individual cases may be obtained at a cost of $35 per case.

1246 n engl j med 352;12 www.nejm.org march 24, 2005

The new england journal of medicine

editorials

Concurrent Antiplatelet and Fibrinolytic Therapy Richard A. Lange, M.D., and L. David Hillis, M.D.

In patients who have myocardial infarction with ST- agents, since they block the final common pathway segment elevation, rupture of an atherosclerotic for platelet aggregation. In pilot studies, patients plaque leads to platelet adhesion, activation, and ag- who had myocardial infarction with ST-segment el- gregation, with subsequent vessel occlusion due to evation who were given aspirin and fibrinolytic ther- thrombus formation. In these circumstances, the apy had earlier and more complete coronary arterial most effective pharmacologic reperfusion regimen reperfusion when a glycoprotein IIb/IIIa inhibitor is concurrent fibrinolytic therapy and platelet inhibi- was administered concomitantly than when it was tion. The marked benefit of such a combination was not.3,4 Subsequently, several trials showed that con- first established in the Second International Study comitant administration of reduced-dose fibrino- of Infarct Survival, in which 35-day mortality among lytic therapy, aspirin, and a glycoprotein IIb/IIIa patients with suspected myocardial infarction was inhibitor decreased the incidence of in-hospital re- 13.2 percent for those receiving neither streptoki- infarction by 1 to 2 percent as compared with full- nase nor aspirin, approximately 10.5 percent for dose fibrinolytic therapy and aspirin.5,6 However, those given one or the other, and 8.0 percent for this modest reduction in the incidence of reinfarc- those receiving both agents.1 tion was offset by a concomitant increase in the inci- Although aspirin, administered with or without dence of nonintracranial bleeding. In short, the ad- fibrinolytic therapy, reduces mortality among pa- dition of very potent antiplatelet therapy (in the form tients with myocardial infarction, it has limitations. of glycoprotein IIb/IIIa inhibitors) to fibrinolytic First, it is a relatively weak antiplatelet agent. It irre- therapy (even at a reduced dose) lessened the risks of versibly inhibits platelet cyclooxygenase, thereby reocclusion and recurrent infarction after successful blocking the synthesis of thromboxane A2, a pow- reperfusion, but it did so at the expense of an in- erful promoter of platelet activation, but it exerts no crease in the frequency of bleeding complications. effect on thromboxane-independent mediators of Clopidogrel is a more potent platelet inhibitor platelet activation, such as adenosine diphosphate, than aspirin, but it is less potent than the glyco- thrombin, or serotonin. Second, in up to 30 percent protein IIb/IIIa inhibitors. It inhibits platelets by of persons with coronary artery disease, the condi- selectively and irreversibly binding to the P2Y12 tion is relatively resistant or unresponsive to aspirin, receptor, thereby blocking the adenosine diphos- as assessed by the extent to which platelet activation phate–dependent pathway of glycoprotein IIb/IIIa– and aggregation are inhibited, bleeding time is pro- receptor activation. As is the case with aspirin, the longed, or treatment is deemed to be clinically ef- response of platelets to clopidogrel is heteroge- fective.2 neous, and resistance to therapy has been reported.7 All mediators of platelet activation cause con- In patients with stable angina, dual therapy with formational changes in the platelet-surface glyco- aspirin and clopidogrel exerts greater inhibitory ef- protein IIb/IIIa receptor, allowing it to bind to cir- fects on platelet activation and aggregation than culating fibrinogen and to form platelet–fibrin therapy with either agent alone. In patients with my- aggregates. As a result, inhibitors of the glycopro- ocardial infarction without ST-segment elevation, tein IIb/IIIa receptor are the most potent antiplatelet an aspirin–clopidogrel combination improved car-

1248 n engl j med 352;12 www.nejm.org march 24, 2005

editorials diovascular outcomes, as compared with aspirin viously undergone coronary-artery bypass grafting alone.8 Unfortunately, those treated with combina- were not enrolled, it is unknown whether such pa- tion therapy had a greater risk of bleeding than those tients should receive clopidogrel in combination treated with aspirin alone, and this finding was par- with full-dose fibrinolytic therapy and aspirin. ticularly noteworthy among those in whom clopid- Third, because few patients in the study under- ogrel was discontinued within five days before cor- went coronary-artery bypass grafting while receiv- onary-artery bypass grafting. ing clopidogrel, the safety of that procedure during In this issue of the Journal, Sabatine et al.9 report clopidogrel therapy is unknown. As noted, previous that patients with myocardial infarction with ST- studies have shown that persons undergoing bypass segment elevation who were treated with fibrino- grafting who recently received clopidogrel are at in- lytic therapy and aspirin derived further benefit if creased risk for substantial perioperative bleeding. clopidogrel was administered concomitantly, in that Fourth, in many centers, coronary angiography is their coronary arterial patency improved. At the performed promptly after fibrinolytic therapy in or- same time, the addition of clopidogrel to full-dose der to evaluate the adequacy of that therapy and to fibrinolytic therapy, aspirin, and heparin did not perform revascularization, if indicated. Patients un- appear to increase the incidence of bleeding com- dergoing early angiography promptly after success- plications — a finding in clear contrast to those of ful fibrinolysis are less likely to have coronary arte- the previously mentioned studies of combination rial reocclusion, with recurrent myocardial ischemia therapy with reduced-dose fibrinolytic therapy, as- or infarction as a result, than those in whom angi- pirin, and a glycoprotein IIb/IIIa inhibitor. The fact ography is routinely delayed after fibrinolytic ther- that clopidogrel is easier to administer and is less apy, as it was in this study. Whether clopidogrel is expensive and safer than a glycoprotein IIb/IIIa in- beneficial when an early invasive strategy is routine- hibitor further adds to its attractiveness for use in ly used is unknown. patients receiving fibrinolytic therapy. Fifth, the timing of clopidogrel administration Of the individual components of the primary in the patients who underwent percutaneous coro- end point assessed by Sabatine et al. (i.e., death, re- nary revascularization may have affected the study current myocardial infarction, and occlusion of the outcome. Previous studies have shown that clopid- infarct-related artery), clopidogrel exerted its great- ogrel must be administered at least four to six hours est effect in reducing the rate of occlusion of the in- before percutaneous intervention in order for it to farct-related artery. The mechanism by which clo- be effective in preventing adverse cardiovascular pidogrel exerted this effect is unknown; it may have events.10 In the current study, treatment with pla- enhanced early reperfusion, prevented reocclusion, cebo or clopidogrel was continued until the day of or improved late reperfusion. The clinical benefit angiography, which was performed a median of 3.5 associated with the use of clopidogrel (i.e., a reduc- days after enrollment; if percutaneous coronary in- tion in the risk of recurrent myocardial infarction) tervention was performed, use of open-label clopid- suggests that its primary mechanism of action is ogrel thereafter was recommended. Thus, patients the prevention of reocclusion. who were randomly assigned to receive clopidogrel Several caveats concerning the study reported by had adequate serum concentrations of the drug at Sabatine et al. are worthy of mention. First, the pa- the time of percutaneous intervention, whereas tient population appears to be at relatively low risk: those who were assigned to placebo did not. This both treatment groups had a 30-day mortality of less aspect of the study may explain, at least in part, why than 5 percent, which is among the lowest reported the rate of the composite end point of death, recur- for any study of patients who have myocardial in- rent myocardial ischemia, or recurrent myocardial farction with ST-segment elevation. Whether un- infarction was similar in the clopidogrel and place- selected patients with myocardial infarction with bo groups before angiography (8.3 and 9.3 percent, ST-segment elevation will benefit from clopidogrel respectively; P=0.27) but favored clopidogrel ther- without having an increased incidence of bleed- apy after percutaneous intervention, at the 30-day ing is unknown. Second, because patients who follow up. were elderly or thin and treated with a standard, Because many patients are resistant to the effects non–weight-based dose of heparin (i.e., those at of a single oral antiplatelet agent, therapy with mul- increased risk of bleeding) and those who had pre- tiple agents with different mechanisms of action is

n engl j med 352;12 www.nejm.org march 24, 2005 1249

The new england journal of medicine

conceptually attractive, provided that it can be ad- ysis in myocardial infarction (TIMI) 14 trial. Circulation 1999;99: ministered without an increased risk of bleeding. 2720-32. 5. Topol EJ. Reperfusion therapy for acute myocardial infarction For patients who are receiving fibrinolytic therapy, with fibrinolytic therapy or combination reduced fibrinolytic thera- a combination of clopidogrel and aspirin appears, py and platelet glycoprotein IIb/IIIa inhibition: the GUSTO V ran- in fact, to be effective and safe. domised trial. Lancet 2001;357:1905-14. 6. Assessment of the Safety and Efficacy of a New Thrombolytic Regimen (ASSENT)-3 Investigators. Efficacy and safety of tenecte- From the Cardiology Division, Department of Internal Medicine, plase in combination with enoxaparin, abciximab, or unfractionat- Johns Hopkins Medical Institution, Baltimore (R.A.L.), and the ed heparin: the ASSENT-3 randomised trial in acute myocardial in- University of Texas Southwestern Medical Center, Dallas (L.D.H.). farction. Lancet 2001;358:605-13. This article was published at www.nejm.org on March 9, 2005. 7. Matetzky S, Shenkman B, Guetta V, et al. Clopidogrel resistance is associated with increased risk of recurrent atherothrombotic 1. ISIS-2 (Second International Study of Infarct Survival) Collabo- events in patients with acute myocardial infarction. Circulation rative Group. Randomised trial of intravenous streptokinase, oral 2004;109:3171-5. aspirin, both, or neither among 17,187 cases of suspected acute my- 8. Yusuf S, Zhao F, Mehta SR, Chrolavicius S, Tognoni G, Fox KK. ocardial infarction: ISIS-2. Lancet 1988;2:349-60. Effects of clopidogrel in addition to aspirin in patients with acute 2. Campbell CL, Steinhubl SR. Variability in response to aspirin: coronary syndromes without ST-segment elevation. N Engl J Med do we understand the clinical relevance? J Thromb Haemost (in 2001;345:494-502. [Errata, N Engl J Med 2001;345:1506, 1716.] press). 9. Sabatine MS, Cannon CP, Gibson CM, et al. Addition of clopid- 3. Strategies for Patency Enhancement in the Emergency Depart- ogrel to aspirin and fibrinolytic therapy for myocardial infarction ment (SPEED) Group. Trial of abciximab with and without low-dose with ST-segment elevation. N Engl J Med 2005;352:1179-89. reteplase for acute myocardial infarction. Circulation 2000;101: 10. Lange RA, Hillis LD. Antiplatelet therapy for ischemic heart dis- 2788-94. ease. N Engl J Med 2004;350:277-80. [Erratum, N Engl J Med 2004; 4. Antman EM, Giugliano RP, Gibson CM, et al. Abciximab facil- 351:200.] itates the rate and extent of thrombolysis: results of the thrombol- Copyright © 2005 Massachusetts Medical Society.

Defining the Stages of Aggressive Non-Hodgkin’s Lymphoma — A Work in Progress James O. Armitage, M.D.

It is well known that localized cancers, including ag- juvant involved-field radiotherapy for patients less gressive non-Hodgkin’s lymphoma, have a better than 61 years of age with apparently localized dif- prognosis than cancers that have spread. When the fuse large B-cell lymphoma.2 Assuming the ACVBP results of treatment for aggressive non-Hodgkin’s regimen is superior to CHOP, would the addition lymphoma are reported, the results for Ann Arbor of adjuvant radiotherapy further improve results? stage I disease (involvement of a single lymph-node Would the addition of rituximab (a monoclonal region) are typically reported separately from those antibody against CD20, which is expressed only on for stage III (involvement of lymph nodes on both B cells) to both regimens have changed the results? sides of the diaphragm) and stage IV (stage III plus These are important questions, because approxi- extranodal involvement). The results of treatment mately 20 percent of the patients treated with ACVBP for stage II disease (involvement of two or more and more than 25 percent of those who were treat- lymph nodes on the same side of the diaphragm) ed with CHOP plus radiotherapy were not cured. may be included in either group. One of the most important advances in the treat- Radiotherapy cures only a minority of cases of ment of cancer is the concept of “staging.” Determi- apparently localized aggressive non-Hodgkin’s lym- nation of the stage of cancer came into general use phoma. The addition of adjuvant chemotherapy im- in the mid-20th century (the American Joint Com- proves the results, and administering chemotherapy mission on Cancer was established in 1959) and ini- before radiotherapy further augments the results. tially focused on the anatomical stage (the specific A study has shown that the use of cyclophospha- sites of tumor involvement). For certain cancers, mide, doxorubicin, vincristine, and prednisone however, other factors clearly affect the prognosis. (CHOP) alone is inferior to CHOP plus adjuvant ra- Symptoms, the results of specific laboratory tests, diotherapy.1 In this issue of the Journal, Reyes et al. and the differentiation of the tumor are among the report that the very intensive regimen of doxorubi- factors now considered in establishing the stage of cin, cyclophosphamide, vindesine, bleomycin, and certain types of cancer. The anatomical stage is an prednisone (ACVBP) is superior to CHOP plus ad- imperfect surrogate for other prognostic factors.

1250 n engl j med 352;12 www.nejm.org march 24, 2005

editorials

The anatomical stage is extremely important to the lymphomas are treated similarly, the survival rate success of localized treatments, because it can be among patients with most types of peripheral T-cell used to predict the likelihood that the tumor will be lymphoma is approximately half that of patients encompassed in the resected specimen or radiation with diffuse large B-cell lymphoma.7 However, pa- field. With systemic therapy, other factors are clearly tients with anaplastic large T-cell or null-cell lym- at play. Assigning a stage to a cancer should accom- phoma have a better outcome than patients with plish three important goals: determining the opti- diffuse large B-cell lymphoma. The poor outcome mal choice of therapy, imparting an accurate prog- among patients with the more common peripheral nosis to the patient and the family, and accurately T-cell lymphomas probably reflects the fact that the stratifying patients for clinical trials and quality as- current treatment regimens were all developed in sessment. trials that enrolled mainly patients with diffuse large The sixth edition of the AJCC Cancer Staging B-cell lymphoma. Future trials should study these Manual3 recognizes that the anatomical stage is not different diseases separately if at all possible. a sufficient predictor of the outcome of aggressive Although not considered in any prognostic sys- non-Hodgkin’s lymphomas and adopts the Inter- tem, the site of origin of non-Hodgkin’s lymphomas national Prognostic Index (IPI).4 The IPI was devel- probably affects the biologic characteristics of the oped in an attempt to improve the ability to predict tumor and the outcome of treatment. Nancy Lee the outcome among patients with aggressive non- Harris, a pathologist in Boston specializing in the Hodgkin’s lymphomas. The IPI takes into account diagnosis of lymphomas, often points out that, al- not only anatomical stage but also age, perfor- though squamous-cell carcinomas originating in mance status, lactate dehydrogenase level, and the different sites of the body are not considered to be presence or absence of multiple extranodal sites of the same disease, diffuse large B-cell lymphomas disease. When the IPI was applied to patients with originating in different sites of the body are consid- aggressive non-Hodgkin’s lymphomas, five-year ered to be the same. It is clear that certain extranod- survival ranged from approximately 40 to approxi- al sites of origin have important clinical implica- mately 75 percent among patients with Ann Arbor tions. For example, testicular diffuse large B-cell stage II and from approximately 25 to approxi- lymphoma frequently metastasizes to the central mately 65 percent among those with stage IV. Al- nervous system and has a pattern of delayed re- though the IPI is a step forward in the prognostic lapse not seen with this type of lymphoma originat- process, it still does not take certain factors into ac- ing at other sites.8 count, including simple clinical observations such Although not yet ready for widespread use, the as the presence of a very large mass. The effect of patterns of gene and protein expression have recent- bulky masses was not considered in the IPI because ly been shown to be powerful predictors of the out- of the lack of uniform collection of data on the max- come of treatment for diffuse large B-cell lympho- imal tumor diameter at the sites participating in the ma. Lymphomas of this type exhibit distinctive study. However, the presence of a 10-cm mass is a gene-expression patterns: one is similar to the pat- predictor of a poor outcome of treatment, and cli- tern of normal germinal-center B cells, and the oth- nicians often take very bulky sites of disease into er resembles the gene-expression pattern of activat- account by adding radiotherapy to that site. ed circulating B cells.9 Patients whose lymphoma The development of the World Health Organiza- has the germinal-center pattern survive approxi- tion’s system of classification for non-Hodgkin’s mately twice as long as patients with a lymphoma lymphomas was an important advance in our ability of the activated B-cell type, and this difference ap- to manage these disorders.5 The system, an out- pears to be independent of previously considered growth of the revised European–American lympho- prognostic factors.10 Genes are information-stor- ma classification,6 recognizes that non-Hodgkin’s age devices, and the effect of gene expression is lymphomas are a group of illnesses, each with mediated by the proteins they produce. Since the unique biologic and clinical characteristics. The dis- relationship between gene expression and the pro- eases previously grouped in studies as “aggressive duction of specific proteins is not straightforward, non-Hodgkin’s lymphomas” are predominantly the level of expression of a specific protein, rather made up of diffuse large B-cell lymphomas (80 to than the level of gene expression, might be a better 90 percent) and peripheral T-cell lymphomas (ap- predictor of the overall outcome of treatment and proximately 10 to 15 percent). When these kinds of the likelihood of a response to specific therapies.

n engl j med 352;12 www.nejm.org march 24, 2005 1251

The new england journal of medicine

Diffuse large B-cell lymphomas that overexpress the mediate- and high-grade non-Hodgkin’s lymphoma. N Engl J Med bcl2 protein, as shown by immunohistochemical 1998;339:21-6. 2. Reyes F, Lepage E, Ganem G, et al. ACVBP versus CHOP plus analysis, but do not have the t(14;18) translocation, radiotherapy for localized aggressive lymphoma. N Engl J Med seem particularly susceptible to rituximab.11,12 2005;352:1197-205. As the concept of staging continues to evolve, 3. Greene FL, Page DL, Fleming ID, et al. AJCC cancer staging manual. 6th ed. New York: Springer-Verlag, 2002. multiple factors other than the Ann Arbor stage will 4. The International Non-Hodgkin’s Lymphoma Prognostic Fac- be important in predicting prognosis and choosing tors Project. A predictive model for aggressive non-Hodgkin’s lym- treatment for patients with aggressive non-Hodg- phoma. N Engl J Med 1993;329:987-94. 5. Jaffe ES, Harris NL, Stein H, Vardiman JW, eds. Pathology and kin’s lymphoma. In addition to the factors consid- genetics of tumours of haematopoietic and lymphoid tissues. Vol. 3 ered in the IPI, the specific type of non-Hodgkin’s of World Health Organization classification of tumors. Lyon, lymphoma, the specific extranodal primary sites, France: IARC Press, 2001. 6. Harris NL, Jaffe ES, Stein H, et al. A revised European-American and the patterns of gene and protein expression are classification of lymphoid neoplasms: a proposal from the Interna- likely eventually to be incorporated into staging sys- tional Lymphoma Study Group. Blood 1994;84:1361-92. tems. Correlating these observations with specific 7. The Non-Hodgkin’s Lymphoma Classification Project. A clinical evaluation of the International Lymphoma Study Group treatments might allow the identification of sub- classification of non-Hodgkin’s lymphoma. Blood 1997;89:3909- groups of patients most likely to benefit from a par- 18. ticular intensive regimen, such as ACVBP, or those 8. Zucca E, Conconi A, Mughal TI, et al. Patterns of outcome and prognostic factors in primary large-cell lymphoma of the testis in a for whom a simpler and less toxic treatment would survey by the International Extranodal Lymphoma Study Group. be equally efficacious. For example, ACVBP may not J Clin Oncol 2003;21:20-7. have been the best treatment for all patients in the 9. Alizadeh AA, Eisen MB, Davis RE, et al. Distinct types of diffuse large B-cell lymphoma identified by gene expression profiling. study by Reyes et al. Physicians who treat patients Nature 2000;403:503-11. with lymphomas hope that we will continue to move 10. Rosenwald A, Wright G, Chan WC, et al. The use of molecular away from the “one-size-fits-all” approach to the profiling to predict survival after chemotherapy for diffuse large- B-cell lymphoma. N Engl J Med 2002;346:1937-47. management of aggressive lymphomas. 11. Mounier N, Briere J, Gisselbrecht C, et al. Rituximab plus CHOP Dr. Armitage has reported receiving consulting fees from (R-CHOP) overcomes bcl-2–associated resistance to chemotherapy GlaxoSmithKline, Corixa, and Geneotope and lecture fees from in elderly patients with diffuse large B-cell lymphoma (DLBCL). Genentech, Corixa, GlaxoSmithKline, and Amgen. Blood 2003;101:4279-84. 12. Wilson HP, Pittaluga S, O’Connor P, et al. Rituximab may over- From the University of Nebraska Medical Center, Omaha. come Bcl-2-associated chemotherapy resistance in untreated dif- 1. Miller TP, Dahlberg S, Cassady JR, et al. Chemotherapy alone fuse large B-cell lymphoma. Blood 2001;98:343a. abstract. compared with chemotherapy plus radiotherapy for localized inter- Copyright © 2005 Massachusetts Medical Society.

Financing Health Care — Finding the Money Is Hard and Spending It Well Is Even Harder Richard Kronick, Ph.D.

In this issue of the Journal, Mongan and Lee argue The calls for universal health care in this issue of persuasively that physicians should lead the way in the Journal, including a proposal for a voucher sys- confronting the unpleasant truth that we cannot tem by Emanuel and Fuchs,2 raise at least four ques- achieve universal health care coverage in the United tions in thinking about publicly financed systems. States without tax increases.1 There is a second un- First, what role, if any, should there be for state gov- pleasant truth that we must confront: if we are going ernments? Second, what role, if any, should there be to create a sustainable system of health care financ- for health plans? Third, how should decisions be ing in which there is a balance between the costs and made about the rate of growth in expenditures? And benefits of care, then the government must become fourth, what mechanisms of accountability are built more involved in influencing how the money is into the system to ensure that resources are used spent. And although the difficulties of figuring out well? Unfortunately, there are no obvious right an- desirable ways of raising money are primarily polit- swers to any of these questions. ical, there are substantial political and technical dif- Administration by the federal government has ficulties in figuring out desirable ways of spending the advantages of increasing the chances of equita- money. ble treatment across the country and potentially pro-

1252 n engl j med 352;12 www.nejm.org march 24, 2005

editorials viding greater protection to low-income persons namic in which physicians and hospitals are re- and racial and ethnic minorities than does admin- warded, with more revenue and members, if they do istration by state governments. However, adminis- a good job.6 Health policy analysts and politicians tration by the federal government also has disadvan- have engaged in intense disputes about whether the tages. First, federal administration requires uniform government should be a single payer of providers rules, which make it difficult to respond to geo- or a single payer of health plans; these debates will graphic diversity in the preferences of patients and doubtless continue. providers.3 Second, federal administration is less Figuring out how quickly the rate of expenditures flexible than state administration.4 Third, and most should grow is a difficult problem for any method important, a federal agency will have great difficulty of health care financing. As we grow richer, and as understanding the needs of local communities. In the technological frontier of medicine expands, a nation as large and diverse as ours, it will be diffi- many of us might be willing to devote a larger part cult for a federal agency to do a good job of assess- of our income to health care in return for a better ing needs at the local level. quality and quantity of life. In a similar way, it is not clear whether payments Emanuel and Fuchs propose a value-added tax should be made directly from the government to that would serve as a “rheostat,” with the public’s providers or whether risk-bearing intermediaries demand for additional services balanced by voters’ should be used. Direct payments have the great ad- willingness to support a tax increase. If the public vantages of simplicity and low transaction costs. had well-formed and consistent preferences, and if Approximately 20 percent of the resources used in the political system were perfectly responsive to health care are devoted to making payments and those preferences, then we could be confident that collecting money, and a single-payer system would the Congress and the President would set the rheo- allow many of those resources to be used in other stat at a level that reflected the public’s willingness ways.5 Direct-payment systems also promote equi- to pay. But politics is a messy business, and we ty and uniformity in coverage policies. should not be confident that we will get the level However, it is hard for the government to deter- exactly right. In a publicly financed system, there mine whether services that physicians provide do would be ongoing debate about whether we are us- much to improve patients’ health. A fee-for-service ing too many or too few resources in health care, structure is antithetical to the development of sys- and we are unlikely to arrive at exactly the right an- tems capable of providing the kind of well-coordi- swer. Although this is an unpleasant prospect, un- nated, patient-centered care that is needed by the fortunately there are not any feasible alternatives. growing number of people with chronic illnesses. Some analysts, politicians, and physicians hope And direct payments make it difficult to reallocate that health plans with a high deductible would allow resources, because the incumbents — providers consumers to balance their willingness to pay with who have received funds in the past — are able to the value that they receive from health care. Accord- hold on to those resources even if they would be ing to this theory, if we just let patients decide better used elsewhere. whether the care they are receiving is worth their Paying health plans as intermediaries can po- money, the invisible hand of the market would de- tentially solve some of these problems. In theory, termine the proper level of resources and the effi- health plans can be more flexible in their internal cient allocation of them. However, with any reason- allocation of resources than can governmental di- able level for the deductible — for example, $3,000 rect-payment systems. If changes in technology al- per person — the preponderance of health care ex- low resources to be shifted out of the hospital into penditures will be for services above the deductible the community, it will be difficult for the govern- and will be covered by insurance.7 Thus, the rate of ment to claw funds out of hospital budgets, but it growth of total expenditures would still be deter- may be more feasible for private plans to do so. mined largely by expenditures covered by insurance, Health plans may be better able than the govern- and some combination of the heavy hand of health ment to determine which services are most valued plans and the government would be needed to get by patients and shift resources to those services. In this growth rate right. the best plans, doctors have a strong voice in the al- Underlying all discussions of reforming health location of resources. A system that is based on care financing is the desire to create a system that is competing private health plans could create a dy- accountable to patients. Accountability is elusive in

n engl j med 352;12 www.nejm.org march 24, 2005 1253

editorials

health care for two primary reasons. First, it is dif- cline within 20 years.10 This would surely be unac- ficult for anyone (patients, insurers, the govern- ceptable, and voters would demand change. ment, or providers) to measure output well, since it We do know how to ensure universal coverage, is difficult to determine the relationship between but we do not know how to ensure that our money care received and health produced. Second, because will be spent well. We should not let uncertainty insurers pay for most services for most patients, we about how best to spend money prevent us from do not use our wallets to hold providers accountable. supporting proposals for universal coverage in The proposal by Emanuel and Fuchs envisions a the context of a system that would be less chaotic variety of mechanisms to create accountability. At and more affordable. Emanuel and Fuchs offer a the public level, the proposed Federal Health Board thought-provoking suggestion for moving in that would be partially accountable to the Congress and direction. the President. To the extent that competition among plans worked as intended, health plans would be ac- From the Department of Family and Preventive Medicine, Uni- versity of California, San Diego, La Jolla. countable to both patients and providers for their 1. Mongan JJ, Lee TH. Do we really want broad access to health performance. These accountability mechanisms are care? N Engl J Med 2005;352:1260-3. far from perfect, but there is no perfect method of 2. Emanuel EJ, Fuchs VR. Health care vouchers — a proposal for creating accountability in health care. universal coverage. N Engl J Med 2005;352:1255-60. 3. Kronick R, Rice T. A state-based proposal for achieving univer- As messy and difficult as a publicly financed sys- sal coverage. In: Meyer JA, Wicks EK. Covering America: real reme- tem is likely to be, the alternatives are worse. Under dies for the uninsured. Washington, D.C.: Economic and Social Re- the status quo, expenditures are increasing at a rate search Institute, June 2001:121-34. 4. Weil A. There’s something about Medicaid. Health Aff (Mill- that is unaffordable, many resources are used in wood) 2003;22(1):13-30. clinical and administrative activities that do little to 5. Woolhandler S, Campbell T, Himmelstein DU. Costs of health improve health, and accountability mechanisms are care administration in the United States and Canada. N Engl J Med 2003;349:768-75. weak. Voluntary, employer-sponsored insurance 6. Enthoven AC. Managed competition in health care and the unwill continue to recede as health care expenditures finished agenda. Health Care Financ Rev 1986;Special No.:105-19. increase more quickly than workers’ wages or pro- 7. Berk ML, Monheit AC. The concentration of health care expen- 8 ditures, revisited. Health Aff (Millwood) 2001;20(2):9-18. ductivity. The resulting increase in the number of 8. Gilmer T, Kronick R. The calm before the storm: expected in- uninsured people will lead to additional health and crease in the number of uninsured Americans. Health Aff (Mill- financial catastrophes, increase the strain on the wood) 2001;20(6):207-10. safety net, and increase political pressure for 9. Committee on the Consequences of Uninsurance. Hidden 9 costs, value lost: uninsurance in America. Washington, D.C.: Na- change. And if health expenditures increase at tional Academies Press, 2003. “only” 2 percent per year more quickly than person- 10. Chernew ME, Hirth RA, Cutler DM. Increased spending on health care: how much can the United States afford? Health Aff al income (a rate that is lower than the historical (Millwood) 2003;22(4):15-25. record), we can expect our standard of living to de- Copyright © 2005 Massachusetts Medical Society.

1254 n engl j med 352;12 www.nejm.org march 24, 2005

The new england journal of medicine

sounding board

Health Care Vouchers — A Proposal for Universal Coverage Ezekiel J. Emanuel, M.D., Ph.D., and Victor R. Fuchs, Ph.D.

Dissatisfaction with the financing of U.S. health care does not consider the cost of technologies relative is widespread. The system is inefficient, inequitable, to their benefits. In an era of rapid technological and increasingly perceived to be unaffordable.1-4 Be- change, that is a recipe for financial disaster.4,10,11 cause only incremental reform is deemed politically Despite the availability of these programs, 15 per- feasible, inordinate attention is devoted to treating cent of Americans have no health insurance; they the institutional symptoms rather than diagnosing either cannot afford to acquire it or are unwilling to systemic problems that require major surgery. As do so.1,2 an alternative, we propose a voucher system for uni- Incremental reforms have been tried, but despite versal health care, an efficient, fair, and relatively some successes, such as the State Children’s Health simple approach that might elicit broad support. Insurance Program, the system as a whole is getting We recognize that change is not imminent, but worse, not better. Major reform is needed but will such a proposal can stimulate discussion and pro- not happen immediately. As problems mount, how- vide a readily available model when the political cli- ever, the demand for change will intensify. In antic- mate becomes hospitable for endorsing meaningful ipation of that demand, we propose a voucher sys- reform. tem with 10 fundamental features.

flaws in the current health care features of the voucher system financing system universality Most Americans obtain their health care through At the start, every American under 65 years of age employer-based insurance, Medicaid and other would receive a voucher that would guarantee and means-tested programs, or Medicare.1 Each com- pay for basic health services from a qualified insur- ponent of this 40-year-old financing system is deep- ance company or health plan. Participating health ly and irreparably flawed. plans would have to offer guaranteed enrollment Employer-based insurance, which now covers 55 and renewal for the risk-adjusted value of the vouch- percent of Americans, is inefficient and inequita- er, regardless of the patient’s medical history. Peo- ble.1,5 It distorts labor markets,6 has high adminis- ple who failed to enroll would be assigned to a trative costs, and generates discontinuous coverage. health plan. Because it is paid through pretax dollars, it is inequitable, since it provides a greater subsidy to high- free choice of health plan income people.7 Medicaid and other means-tested Individuals and families would choose which ba- programs cover about one in six Americans.1 These sic insurance program or health plan they wanted programs require costly determinations of eligibil- among several alternatives. ity, impose high marginal tax rates on recipients because the subsidies fall as income rises, and encour- freedom to purchase additional services age evasion of reported income.8 Many people who People who wanted to purchase additional services are eligible do not apply — some to avoid the admin- or amenities, such as a wider choice of hospitals and istrative hassle or stigma and others because they specialists or more comprehensive mental health expect their incomes to improve.9 The programs services, could do so with their own after-tax dollars. also generate discontinuous coverage as people move in and out of eligibility. Medicare, which cov- funding by an earmarked value-added tax ers about one in eight Americans, is a popular open- The funding for the vouchers would come from an ended entitlement but has fundamental flaws. It earmarked value-added tax. Earmarking creates a

n engl j med 352;12 www.nejm.org march 24, 2005 1255

The new england journal of medicine

direct connection between benefit levels and the tax graphic regions. The Federal Health Board would level, serving as a cost control “rheostat.” If people define and periodically modify the basic benefits want more services to be covered in the basic plan, package and through its regional boards would be they must be willing to support a tax increase. A val- an active contractor with health plans, informing ue-added tax is administratively efficient and cannot Americans about their health care options, reim- be easily evaded. The tax is based on personal con- bursing health plans, and collecting data related to sumption, which is closely related to long-term fi- patient satisfaction, the quality of care, and risk and nancial well-being, regardless of the source of in- geographic adjustments for payments. The Federal come or wealth. Health Board would regularly report to Congress on the health care system. reliance on a private delivery system This proposal does not call for government health assessment of technology and outcomes care and would not legislate changes in the current An independent Institute for Technology and Out- private delivery system. comes Assessment would be established. Its research and database would focus on assessing the end of employer-based insurance effectiveness and value of various interventions and By providing basic care for all Americans and elim- treatment strategies and disseminate information inating tax benefits for health insurance premiums, concerning outcomes of treatments delivered in reg- employer-based insurance would probably fade ular practice. Funding for the institute would come away. Critics throughout the political spectrum have from a dedicated portion of the financing tax, such noted the many shortcomings of employer-based as 0.5 percent of the total. insurance; few would mourn its passing.10,12-14 These 10 features address fundamental flaws in the current U.S. health care financing system (Ta- elimination of medicaid and other ble 1).14 Previous reform proposals have incorpo- means-tested programs rated several of these features, such as providing Since every individual and family would receive a people with choice among several plans that offer voucher, there would be no need for Medicaid (ex- similar benefits at the same price, retaining private cept for nursing home coverage), the State Chil- insurance and health plans, and removing the tax dren’s Health Insurance Program, or other means- subsidy for the purchase of additional care.15-24 This tested programs. People who are covered by such proposal, however, is a unique package marked by programs would be incorporated into the main- the use of vouchers to simplify administration, the stream health care system without means testing. creation of a financing system through an ear- Funding for long-term care, currently provided by marked value-added tax, the elimination of Medic- Medicaid, would need to be continued. aid and other means-tested programs and employer-based insurance, a beginning to the phase-out phasing out of medicare of Medicare, oversight through a Federal Health Although no one who is already enrolled in Medi- Board, and the creation of an Institute of Technolo- care would be forced to change to the voucher sys- gy and Outcomes Assessment.15,17,18,24 tem, Medicare would be phased out over time. Peo- ple turning 65 would continue to be enrolled in the universal benefits package voucher system; there would be no new enrollees in Medicare. It is important to note that current The universal benefits package covered by the Medicare benefits would be supplemented by a voucher should be sufficiently comprehensive to tiered pharmacy benefit modeled on that provided provide most Americans with most of their care as part of the basic benefits package of the voucher most of the time. It should not be designed as a system. safety net to serve only the poor.25 The benefits provided should be those typically offered by large administration employers, including inpatient and outpatient hos- Management and oversight would be the responsi- pital services, visits to physicians’ offices, well-child bility of a Federal Health Board (modeled on the care and other preventive measures, mental health structure of the Federal Reserve System), with re- care, and tiered pharmaceutical benefits, typically gional boards to manage and oversee various geo- with dollar limits. We suggest only modest deduct-

1256 n engl j med 352;12 www.nejm.org march 24, 2005

sounding board

Table 1. Problems in the Current Health Care System and Their Resolution by a Universal Voucher System.

Problem Current System Universal Health Care Voucher System

Lack of insurance coverage More than 40 million Americans are un- Every American is guaranteed basic insured because they are too young health care coverage without means for Medicare, not poor enough for testing or exclusions of any kind. Medicaid, unable to acquire insurance at an affordable price, or unwilling to purchase insurance. Increases in health care expenditures Health care expenditures have increased An earmarked tax directly links the expan- by nearly 10% annually for the past sion of the cost of a basic benefits 3 years and by nearly 75% in the past package with the public’s willingness decade. There are no firm cost-con- to increase taxes. Services beyond ba- trol mechanisms. Medicare is an sic care are paid with after-tax in- open-ended entitlement that ignores come, which ensures that users costs relative to benefits. Employer- weigh costs against benefits. A sys- based insurance typically provides tematic assessment of technology greater subsidies for more expensive and outcomes limits the use of servic- plans and insulates patients from es that have high costs relative to consideration of costs. their benefit. Discontinuity of coverage There is substantial discontinuity of cov- There is complete continuity of coverage, erage because of job loss or change, since users stay in the same plan as retirement, changes in employer- long as they wish, regardless of based health plans, or changes in changes in employment, income, means-tested eligibility. health status, or other circumstances. Inefficient labor markets Employer-based insurance distorts deci- There is no distortion of employment de- sions of employers and employees cisions, since employers and employ- about such things as outsourcing, ees are free to make decisions inde- hiring, retirement, and job changes. pendently of any considerations Means-tested programs, such as regarding health insurance. The poor Medicaid, discourage the poor from do not lose coverage if their income working because they might lose rises. their insurance eligibility. Haphazard and unfair subsidies Tax treatment of employer-based insur- Users contribute to the support of basic ance provides greater subsidies to care in proportion to their consump- high-income workers. Many low- tion of goods and services. There are income workers are ineligible for no free riders. means-tested programs, and many people who are eligible do not apply.

ibles and copayments to minimize access barriers dren’s Health Insurance Program, and other pro- for the poor. In 2004, the average annual premium grams being replaced. In 2004, excluding Medicare for such coverage in an employer-based program and nursing home coverage, the cost of personal was $9,950 for families and $3,695 for individuals.2 health care coverage for Americans under the age of Ultimately, the Federal Health Board would struc- 65 years was estimated to exceed $800 billion, in- ture the benefits package after wide consultation cluding more than $600 billion in premiums for with experts and involvement of the public through employer-based insurance and $200 billion for Med- various mechanisms.26-28 The process would be it- icaid and other means-tested programs (Levit K: erative, with modifications reflecting a balance be- personal communication).2,3 These costs have been tween the public’s desire for more health care ser- increasing by 8 to 10 percent annually for the past vices and its willingness to pay the valued-added tax. few years.2,3 How much would the voucher plan cost? Be- economic feasibility cause the cost would depend on precisely which services, deductibles, and copayment levels would be The economic feasibility of the voucher system de- incorporated into the universal benefit, the voucher pends on the cost of the publicly funded universal proposal has not been “scored.” Nevertheless, ed- benefits package, as compared with the cost of em- ucated estimates are possible. Through charity and ployer-based insurance, Medicaid, the State Chil- other mechanisms, the 15 percent of Americans

n engl j med 352;12 www.nejm.org march 24, 2005 1257

The new england journal of medicine

who are currently uninsured do get some care; the with American values than are the alternatives of voucher benefit package would probably increase employer or individual mandates with explicit sub- this group’s use of services by about one third.14 sidies or a single-payer plan. By providing publicly This would raise overall use of health care by about funded basic care for all, with free choice of a plan 5 percent. Some additional use might be expected and freedom to buy additional services, the vouch- from those currently insured with policies less gener system reconciles the distinctively American ten- erous than the voucher system’s provisions. Con- sion between equality and individual freedom more versely, people with more generous policies would directly and efficiently than do any of the alterna- probably reduce their use of the system. In addition, tives.32 some resources would be needed to eliminate de- The voucher system will appeal to both large ductibles and copayments for the very poor. Over- and small employers, who are straining under in- all, it is reasonable to expect use to rise by approxi- creasing costs and looking for an end to employer- mately 5 percent. However, expenditures need not based insurance. The growing demands of Medic- increase. aid are overwhelming state coffers, forcing choices With the voucher system, administrative costs between tax increases and cuts in education and oth- would decrease markedly. Physicians know all too er services. State governments will welcome relief well that the current system is an administrative from the financial and administrative burdens im- nightmare; the voucher system would simplify it posed by means-tested programs. substantially. For instance, the cost of screening Opposition to universal health care vouchers may and determining eligibility for each child enrolled come from employers with mostly young workers, in a State Children’s Health Insurance Program is who use little care, and from high-wage workers, the equivalent of more than two months of health who receive munificent benefit packages tax-free. care premiums, and the enrollment of adults is even Among the 1300 health insurers, the small ones that more complex and costly.8 Moreover, much of the rely on select market niches will find universal $100 billion spent on the sales and administrative health care vouchers a threat, as will health-bene- costs of private insurance would also be saved.2,29 fits consultants, who will lose business. Over time, the voucher system would foster a more The reaction of health care professionals is un- rational approach to the delivery of services. For in- certain. There will doubtless be a concern that any stance, more rational prescribing of drugs could universal-coverage system with a fixed budget will save 1 to 2 percent of health care expenditures. A inevitably threaten their income and freedom, and combination of lower administrative costs and more some practitioners will want to maintain the status efficient delivery should offset some or all of the quo — which is an impossible dream. However, 5 percent increase in use. when reform comes, many physicians will support a voucher system because it will provide the oppor- political feasibility tunity, the information, and the incentive to deliver cost-effective care to all Americans. We recognize that the voucher system is not politi- Ironically, the strongest resistance to the vouch- cally feasible at this time. Neither is any other ma- er system may come from people, who embrace jor reform in health care. Obstacles to major reform universality but might object to the elimination of arise from multiple sources — cultural, social, eco- Medicaid, the imposition of a value-added tax, or nomic, and political.10,11,14,30,31 Normally, the the use of vouchers. Such opposition would be American political system resists change; it tends misplaced. Although Medicaid provides vital servic- to enact major social programs only during times es for some people, it is grossly inequitable, offering of war, economic depression, or civil unrest. Even no coverage for many uninsured low-wage work- without such traumas, there will come a time when ers. Means testing is demeaning, an invitation to the inequities, inefficiencies, and costs of the cur- evasion, and administratively complex.8,9 Benefits rent methods of financing health care will be so in- vary greatly by state and are widely perceived to tolerable that the public will not only accept but de- constitute second-class medicine. Most important, mand comprehensive reform. Medicaid is draining resources from other state ser- At that time, the political feasibility of the vouch- vices, such as public education, that are vital to the er system will be compelling. It is more congruent overall well-being of poor people. The provision of

1258 n engl j med 352;12 www.nejm.org march 24, 2005 sounding board uniform benefits to both the middle class and the addressed. Additional study will be required of poor — even if some people forsake special bene- plan operations and other medical care issues, in- fits — seems a fair trade-off. cluding the definition of the universal benefit, the Some people reflexively reject a value-added tax development of procedures for participation in the as regressive. However, the distributional impact of plan and enrollment of beneficiaries, the establish- the voucher proposal requires looking at the bene- ment of the Institute for Technology and Outcomes fits as well as the tax burden. All things considered, Assessment, and a strategy for dealing with the ef- the program is progressive, since it implicitly sub- fects of health care vouchers on medical education sidizes the poor. It is not an accident that countries and research. Methods for calibrating the degree of such as Norway and Sweden, which provide univer- flexibility in modifying the basic benefits package sal health coverage, make substantial use of value- will also need to be developed. Limited flexibility en- added taxes to fund social programs. Furthermore, hances comparability among plans, whereas greater value-added taxes can be made more progressive by flexibility fosters innovation in the delivery of care excluding from the taxable base items such as food, and choice. which account for a declining proportion of con- Numerous legal and regulatory issues must also sumption as income rises.32,33 Finally, people of- be investigated, including the establishment of the ten reject voucher proposals as a threat to the univer- national and regional offices of the Federal Health sality of social programs. Medical care, however, is Board and a definition of the relationship between different. It has never been universally guaranteed the purview of such boards and that of numerous in the United States, and segmentation of the mar- state laws covering malpractice and mandatory ket is currently widespread. The health care voucher medical practice. Finally, considerable thought and would guarantee Americans universal coverage for study must be given to the problem of the transition the first time, and without means testing or exclu- from the current system to the voucher system. sion for medical conditions. All these issues must be addressed with data, Today, comprehensive reform seems politically analysis, and the balancing of competing values. unrealistic. During the next few years, however, as Questions about the details of health care vouchers employers continue to cut benefits or eliminate cov- are inevitable, but they are not imminently vexing. erage entirely, as states reduce Medicaid services to While pressure builds for comprehensive change, avoid deficits, and as increases in Medicare costs there is time to deliberate about them. lead to higher payroll taxes, higher Part B premi- As problems mount with the current health care ums, and cuts in reimbursements to hospitals and system, publicly funded social insurance combined physicians, there will be increasing recognition that with substantial market elements would provide a the system is irreparably broken. Support for major middle ground that could galvanize broad support reform will grow, and the combination of efficiency from businesses and states and from the uninsured and equity offered by universal health care vouchers and the general public. By making the financing of should make it the system of choice. health care in the United States considerably more efficient, fair, and simple, universal health care issues requiring additional study vouchers would also provide a framework for improving the delivery of care. This is a broad outline of the proposal for a system of We are indebted to Kenneth Arrow, David Druker, Alain En- universal health care vouchers; myriad details need thoven, John Etchemendy, Amy Finkelstein, Alan Garber, Lee Gold- man, Mary Goldstein, Hank Greeley, Judy Miller Jones, Sharon Le- careful study before such a system is implemented. vine, Hal Luft, Philip Pizzo, Antonio Rangel, Deborah Satz, Steven Economic and financial issues include developing Schroeder, and John Shoven for helpful comments and suggestions; more precise estimates of the cost of universal and to Steven Coulter, Andrea Voytko, Greg Scully, Anne Rosone- Franco, Gary Claxton, Ben Finder, and Kate Levit for assistance with vouchers, the control of costs over time, and the fi- data on coverage and rates of various health care plans and health nancing of special services. There will be a need to expenditures. develop reimbursement methods that encourage efficiency while minimizing the opportunity for health From the Posterity Project, Chicago (E.J.E.); and the Department of Economics, Stanford University, Stanford, Calif. (V.R.F.). plans to enroll only healthy patients. In a similar way, geographic variations in practice patterns, and 1. Census Bureau. Health insurance coverage in the United States: 2002. (Accessed February 23, 2005, at http://www.census.gov/prod/ thus costs, that are unjustified by differences in wag- 2003pubs/p60-223.pdf.) es, other prices, or the quality of care will need to be 2. Gabel J, Claxton G, Gil I, et al. Health benefits in 2004: four

n engl j med 352;12 www.nejm.org march 24, 2005 1259 The new england journal of medicine

years of double-digit premiums increases take their toll on cover- 19. Davis K, Schoen C, Schoenbaum SC. A 2020 vision for American age. Health Aff (Millwood) 2004;23(5):200-9. health care. Arch Intern Med 2000;160:3357-62. 3. Levit K, Smith C, Cowan C, Sensenig A, Catlin A, Health Ac- 20. Feder J, Levitt L, O’Brien E, Rowland D. Assessing the combi- counts Team. Health spending rebound continues in 2002. Health nation of public programs and tax credits. In: Meyer JE, Wicks Aff (Millwood) 2004;23(1):147-59. EK, eds. Covering America: real remedies for the uninsured. 4. Centers for Medicare & Medicaid Services. 2004 Annual report Washington, D.C.: Economic and Social Research Institute, 2001: of the board of trustees of the Federal Hospital Insurance and Federal 43-56. Supplementary Medical Insurance trust funds. (Accessed February 21. Gruber J. A private/public partnership for national health insur- 23, 2005, at http://www.cms.hhs.gov/publications/trusteesreport/ ance. In: Meyer JE, Wicks EK, eds. Covering America: real remedies 2004.) for the uninsured. Washington, D.C.: Economic and Social Re- 5. Enthoven AC. Consumer-centered vs. job-centered health in- search Institute, 2001:57-72. surance. Harv Bus Rev 1979;57:141-52. 22. Singer SJ, Garber AM, Enthoven AC. Near-universal coverage 6. Gruber J, Madrian BC. Health insurance, labor supply, and job through health plan competition. In: Meyer JE, Wicks EK, eds. Cov- mobility: a critical review of the literature. In: McLaughlin CG, ed. ering America: real remedies for the uninsured. Washington, D.C.: Health policy and the uninsured. Washington, D.C.: Urban Institute Economic and Social Research Institute, 2001:153-72. Press, 2004:97-178. 23. Wicks EK, Meyer JA, Silow-Caroll S. A plan for achieving univer- 7. Shiels J, Haught R. The cost of tax-exempt health benefits in sal health coverage: combining the new with the best of the past. In: 2004. Health Affairs Web Exclusive. February 25, 2004. (Accessed Meyer JE, Wicks EK, eds. Covering America: real remedies for the February 23, 2005, at http://content.healthaffairs.org/cgi/content/ uninsured. Washington, D.C.: Economic and Social Research Insti- full/hlthaff.w4.106v1/DC1.) tute, 2001:193-206. 8. Fairbrother G, Dutton MJ, Bachrach D, Newell KA, Boozang P, 24. Emanuel E. Health care reform: still possible. Hastings Cent Cooper R. Costs of enrolling children in Medicaid and SCHIP. Rep 2002;32:32-4. Health Aff (Millwood) 2004;23(1):237-43. 25. Pear R. Frist expects Congress to try to expand health coverage, 9. Remler DK, Glied SA. What other programs can teach us: in- but says universal insurance is impossible. New York Times. Febru- creasing participation in health insurance programs. Am J Public ary 7, 2004:A7. Health 2003;93:67-74. 26. Eddy DM. What care is ‘essential’? What services are ‘basic’? 10. Aaron HJ. Serious and unstable condition: financing America’s JAMA 1991;265:782, 786-8. health care. Washington, D.C.: Brookings Institution, 1991. 27. National Advisory Committee on Core Health and Disability 11. Peterson PG. Will America grow up before it grows old? New Support Services. Core services for 1995/96. Wellington, New York: Random House, 1996. Zealand: Ministry of Health,1994. 12. Thomas B. Vision for health care. Presented at the National 28. Danis M, Biddle AK, Goold SD. Insurance benefit preferences of Press Club, February 12, 2004. (Accessed February 23, 2005, at http: the low-income uninsured. J Gen Intern Med 2002;17:125-33. //www.ncpa.org/prs/tst/20040331bttst.htm.) 29. Centers for Medicare and Medicaid Services. Health accounts. 13. Clinton HR. Now can we talk about health care? The New York (Accessed February 23, 2005, at http://www.cms.hhs.gov/statistics/ Times Magazine. April 28, 2004:26-31, 46, 47. nhe/.) 14. Institute of Medicine. Insuring America’s health: principles and 30. Fein R. Medical care, medical costs. Cambridge, Mass.: Harvard recommendations. Washington, D.C.: National Academies Press, University Press, 1986. 2004. 31. Fuchs VR. National health insurance revisited. Health Affairs 15. Fuchs VR. Let’s make Volkswagen medicine compulsory. Med (Millwood) 1991;10(4):7-17. Econ 1969:110-9. 32. Tax reform for fairness, simplicity, and economic growth: the 16. Enthoven A, Kronick R. A consumer-choice health plan for the Treasury Department report to the President. Vol. 3. Value-added 1990s: universal health insurance in a system designed to promote tax. Washington, D.C.: Government Printing Office, 1984. (Ac- quality. N Engl J Med 1989;320:29-37, 94-101. cessed February 23, 2005, at http://www.treas.gov/offices/ 17. Emanuel EJ. The ends of human life: medical ethics in a liberal tax-policy/library/tax-reform/tres84v3All.pdf.) polity. Cambridge, Mass: Harvard University Press, 1991:155-233. 33. McLure C. The value-added tax: key to deficit reduction. Wash- 18. Fuchs VR. Health system reform: a different approach. JAMA ington, D.C.: American Enterprise Institute, 1987. 1994;272:560-3. Copyright © 2005 Massachusetts Medical Society.

Do We Really Want Broad Access to Health Care? James J. Mongan, M.D., and Thomas H. Lee, M.D.

Now that the presidential election is over, the real broader health insurance coverage should be a work on the issue of access to health care should high priority. The candidates were asked to debate begin. The work will be difficult — and probably on the basis of their values on issues such as abor- impossible to accomplish unless the health care tion, stem-cell research, and same-sex marriage. profession helps the U.S. public to confront some But their rather marked disagreement on how to in- unpleasant truths. crease access to health care was never treated as a One might have thought that an election that “values” issue, perhaps because voters were unwill- hinged on “moral values” would generate serious ing to have their own values tested on this topic. discussion of one of our most fundamental “val- Americans may say they believe that instituting ues” issues — that is, whether we really believe that broader coverage is the right thing to do, yet every-

1260 n engl j med 352;12 www.nejm.org march 24, 2005 sounding board one prefers to change the subject when the discus- concept is that the marketplace cannot directly levy sion turns to how to pay for it. We can understand taxes and is therefore seen as less threatening to in- why candidates seeking election shy away from the dividuals’ interests than government is. To date, implication that they might raise taxes. But we be- however, market-based solutions have had only lieve that the medical profession must look within modest success in controlling costs, and few peo- itself, collectively and individually, and ask whether ple even pretend that markets can help the one of we truly want broader coverage. If we do, we must every six persons who is currently uninsured and be open, honest, and direct about what will be re- cannot even afford to enter the market. quired — beginning with publicly advocating for the higher taxes or employer mandates needed to the challenge of the uninsured achieve this goal. The antitax movement should be recognized for our aim is low what it is, the main enemy of those who truly seek to meet the needs of the 45 million uninsured Amer- During the campaign, both candidates said that they icans. Admitting that we are living with two irrecon- supported the goal of access to health care for all,1 cilable values — we want to see ourselves as a caring but dollars mean more than words when it comes people, but we don’t want our taxes to go up — is to covering the uninsured. Senator John Kerry’s pro- unpleasant. For that reason, we work hard to proposals would have cost about $653 billion over the tect our self-image by means of elaborate fantasies. next 10 years and would have extended coverage to We tell ourselves that the uninsured have only about 27 million of the 45 million uninsured Amer- themselves to blame for their situation, whereas the icans. President George W. Bush’s plan includes re- truth is that about 80 percent of them are either em- fundable tax credits and encouragement of the pool- ployed or dependents of those who are employed.3 ing of purchasing power by small employers. This The uninsured work hard, often at agricultural, re- plan would increase federal spending by about tail, and service jobs, yet they fall between the cracks $90 billion over the next 10 years and extend cover- of our health insurance system. age to an estimated 2.4 million people, or about We tell ourselves that they are actually getting 5 percent of the uninsured. care despite our flawed system. And in a sense, they The absence of any real effort by the Bush ad- are receiving limited care for their more dramatic ministration to make more than a dent in the prob- crises. As a society, we don’t want to see childbirth lem of the uninsured reflects the basic ambivalence occur in the street or an obviously broken bone ig- of Americans about the role of government. On the nored. We are willing to pay enough to move the care one hand, we are quite interested in the government indoors and out of sight. But we are not willing to benefits that affect us directly, such as Social Securi- provide coverage for the chronic conditions and pre- ty, Medicare, veterans’ benefits, and certain key tax ventive care that might enable the uninsured to lead subsidies. On the other hand, we have a historical more productive and happier lives. and cultural bias against large, intrusive, regulatory From decades of health services research, we actions undertaken by the government, especially know that having health insurance makes a differ- when they involve increasing taxes. ence. The uninsured are more likely to postpone The antitax movement has only become strong- care and less likely to receive preventive care than er and more widespread since the passage of Prop- those with insurance. They are twice as likely as the osition 13 in California in 1978. Two years ago, in insured to be hospitalized for the avoidable com- the hotbed of liberalism known as Massachusetts, plications of diabetes and hypertension. The facts 45 percent of voters voted to repeal the state income are beyond dispute. The question is whether we care tax. Politicians in virtually all federal, state, and lo- enough to do what every other industrialized nation cal offices are asked to promise not to raise taxes as has done — provide coverage for all. if they were taking some kind of loyalty oath, even though our federal tax level as a percentage of the three failures gross domestic product is the lowest since 1959.2 This political environment provides fertile Three times in the past few decades we have con- ground for the notion that the marketplace should fronted the challenge of broadening access to health solve health care problems. Part of the appeal of this care, and three times we have failed to meet it. In

n engl j med 352;12 www.nejm.org march 24, 2005 1261 The new england journal of medicine

1974, President Richard Nixon made a bold, even the phrase “rugged individualism” to describe this noble proposal of broad health insurance coverage. outlook. Liberals might use a harsher, more grating Nixon had two options for financing his proposal: word, “selfishness.” We don’t know at what point an increase in taxes or — the option he chose — an rugged individualism becomes self-centeredness employer mandate (i.e., a requirement that employ- and self-centeredness, selfishness, but somewhere ers provide health insurance to their employees). in that transition health care reform died in 1974, Conservatives attacked the employer mandate as a 1979, and 1994. hidden tax, and liberals attacked the extent of coverage as not going far enough. Nixon’s proposal the costs of broad access died in the congressional cross fire. In 1979, Presi- dent Jimmy Carter put forth a similar proposal, How much would broad access to health insurance which met a similar fate. The political dynamics that cost? In 2003, the Institute of Medicine’s Committee prevented progress then were no different with a on the Consequences of Uninsurance published Democratic president than they had been under analyses in which it concluded that the incremental Nixon, a Republican. costs of health services that the uninsured could be The most resounding defeat occurred a decade expected to use, were they to gain coverage, range ago, when President Bill Clinton put forth a health- from $34 billion to $69 billion (2001 dollars) (Ta- insurance proposal that, like Nixon’s and Carter’s, ble 1).4 These new estimated costs would amount was financed by an employer mandate. Critics at the to 2.8 to 5.6 percent of national spending for per- time, and since, complained about the plan’s com- sonal health care services, or about half of the 8.7 plexity, but it was the opposition by small businesses percent increase in spending that actually occurred to the employer mandate that killed the proposal. between 2000 and 2001. The real issue was, as it always has been, how to We recognize that this “prescription” is not pay for broader coverage. Our nation and the Con- without pain. And we acknowledge that much of gress faced this issue squarely and made a decision. any new funding would flow to health care organi- We gave a higher priority to avoiding an increase in zations such as our own that do not turn anyone taxes and its closely related cousin, an employer away because of lack of insurance. We also recog- mandate, than we did to providing coverage for 45 nize that the ultimate costs of broad access to health million uninsured people. care might well be somewhat higher than these estimates, because most health insurance proposals confronting the truth have additional goals, such as seeking fairness for about ourselves the employers who currently offer coverage. Any major initiative to broaden access to health insur- How can a country as idealistic and generous as the ance can influence the coverage of those currently United States fail repeatedly to accomplish in health insured; some will have their insurance improved, care coverage what every other industrialized na- and others who now have private insurance will tion has achieved? One explanation may be that we drop such coverage in favor of any new public plan. are not so idealistic or generous as we would like to The costs could be on the same scale as the occupa- believe we are. Although there are various issues tion of Iraq. surrounding plans for broader coverage, the over- If broad access to health care is ever to be more riding concern in each of the previous defeats has than a campaign sound bite, it cannot be a casual been our reluctance to have to pay by means of tax- commitment. Americans must understand that they es or mandates for health care for others. are going to make real sacrifices for other Ameri- Our self-image was forged in crises such as the cans. Currently, the anguish caused by annual in- Depression, World War II, and the Civil Rights creases in health care costs is offset by the knowl- Movement, in times when the idea of sacrifice for edge that because of each year’s medical progress, the greater good was a political rallying cry. Over the we are “buying a better product,” potentially for our- past 30 years, however, we have become a relatively selves, should we become ill. A one-time increase affluent nation of consumers who are focused more of 50 percent in the rate of rising health care costs on the rights and desires of individuals than on the that benefits someone else is a more difficult sell. needs of the community. Many conservatives use Would there be a return on investment for these

1262 n engl j med 352;12 www.nejm.org march 24, 2005 sounding board new costs to society? The Committee on the Conse- quences of Uninsurance estimated that diminished Table 1. Estimated Annual Costs of Health Care Services for Uninsured Per- sons and the Expected Economic Value of Providing Them with Insurance.* health due to inadequate insurance leads to societal costs of $65 billion to $130 billion per year (Table 1). Current and Projected Costs Billions of Dollars

Such losses might be avoided if broad access to Current cost of care for the uninsured for all or part of the year health insurance were to reduce the rates of mor- Out-of-pocket payments 26.4 bidity and mortality among the uninsured to levels similar to those among people with insurance. As a Insurance payments (for those insured for part of the year) and workers’ compensation country, however, we get stuck, because these benefits do not directly affect the people and the em- Private sources 24.2 ployers who would be asked to pay more for the Public sources 13.8 broader access to care. Uncompensated care 34.5 Total 98.9 a defining moment Projected annual cost of additional utilization with coverage 34–69 Benefits of coverage for the uninsured — costs saved as a result 65–130 A cynical but smart bet would be that the United of improved health States will not make meaningful progress toward expanding health care coverage during the remain- * Adapted from the Institute of Medicine’s Committee on the Consequences of Unin- ing years of the Bush administration. Assuming this surance.4 Values are expressed in 2001 dollars. is the case, the number of uninsured persons can be expected to rise and access to care by the uninsured is likely to narrow, because it is increasingly difficult with a willingness to let health care continue to infor health care providers to give away care for free. crease as a percentage of the gross national prod- What should health care providers do if they re- uct, we may well find that at some point we will ally believe health care is a right of all people but still need to limit or ration services, according to market understand the political realities? At least in the forces or to budgetary allocations. near term, the United States is not likely to adopt the Our prescription — higher taxes and possibly government-based approaches of the United King- rationing — is hardly the platform of a winning dom, Canada, and many other Western countries.5 ticket for any political candidate. But now that the Instead, we think that proponents of broader ac- 2004 presidential election is over, it’s time to get cess should advocate a new balance, one that buff- serious about health care. And health care profes- ers the defects of a corporate-market model yet sionals should lead the way and begin the truth- avoids the centralized power that many fear in a telling about the “tough medicine” that will be re- government model. This new balance would have quired if we are to live up to our ideals. its own formidable political challenges. It must include a mechanism to finance care for all our peo- From Partners HealthCare System and Harvard Medical School ple, a mechanism that will inevitably include the — both in Boston. Address reprint requests to Dr. Mongan at Partners HealthCare System, Prudential Tower, 11th Fl., 800 need for new taxes, mandates, or both. We believe Boylston St., Boston, MA 02199, or at [email protected]. that a fundamental and defining question for all health care providers who support broad access to 1. Bush GW, Kerry JF. Health care coverage and drug costs — the candidates speak out. N Engl J Med 2004;351:1815-9. health care is this: Are you willing to advocate 2. Office of Management and Budget. Fiscal year 2005 budget. publicly for higher taxes? If enough of us state the (Accessed February 16, 2005, at http://www.taxpolicycenter.org/ simple truth that this is necessary, we may create a TaxFacts/TFDB/TFTemplate.cfm?Docid=205.) 3. Kaiser Commission on Medicaid and the Uninsured. Health political environment in which our leaders may insurance coverage in America: 2002 data update. Menlo Park, eventually be able to lead on this issue. Calif.: Kaiser Family Foundation, December 2003. (Accessed March At the same time, we must attempt to balance 9, 2005, at http://www.kff.org/uninsured/4154.cfm.) 4. Institute of Medicine, Committee on the Consequences of Unin- markets and regulation as a way to control health surance. Hidden costs, value lost: uninsurance in America. Wash- care costs. Health care providers must work to en- ington, D.C.: National Academies Press, 2003. sure that we provide only necessary, efficient, and 5. Budetti PP. 10 Years beyond the Health Security Act failure: subsequent developments and persistent problems. JAMA 2004;292: high-quality services. Even with our best efforts to 2000-6. reduce waste and bureaucracy, however, and even Copyright © 2005 Massachusetts Medical Society.

n engl j med 352;12 www.nejm.org march 24, 2005 1263

The new england journal of medicine

correspondence

Palifermin and Chemotherapy-Induced Oral Mucositis

to the editor: In the editorial by Garfunkel (Dec. to the editor: The study by Spielberger et al. of 16 issue),1 which accompanies the report by Spiel- palifermin is a hallmark trial for the prevention of berger et al.2 about the action of palifermin in oral severe oral mucositis due to myeloablative treat- mucositis, Garfunkel cites other approaches but ment. Stratification was performed according to the does not mention that low-level laser therapy might type of hematologic cancer and the treatment center, be useful in decreasing the severity of chemother- respectively, and treatment groups were compared apy-associated oral mucositis.3 We believe that by means of the generalized Cochrane–Mantel– low-level laser therapy should be included in a ran- Haenszel method stratified according to the study domized trial of palifermin in patients with oral center. However, there are risk factors that may in- mucositis. fluence both the incidence and the severity of oral 1 Ahmad Awada, M.D., Ph.D. mucositis. These factors include smoking, malnu- Marie-Thérèse Genot, M.D. trition, poor oral hygiene, xerostomia, and other un- Jean Klastersky, M.D., Ph.D. derlying oral disorders — none of which have been Jules Bordet Institute reported. Because these factors may have an impor- B-1000 Brussels, Belgium tant role in the development of mucositis and its [email protected] subsequent course, the analysis of the outcome 1. Garfunkel AA. Oral mucositis — the search for a solution. should have been adjusted for them. N Engl J Med 2004;351:2649-51. 2. Spielberger R, Stiff P, Bensinger W, et al. Palifermin for oral mu- Wolfgang J. Köstler, M.D. cositis after intensive therapy for hematologic cancers. N Engl J Med Michael Hejna, M.D. 2004;351:2590-8. Christoph C. Zielinski, M.D. 3. Genot M-T, Klastersky J. Low level laser for prevention and therapy of oral mucositis induced by chemotherapy and/or radiotherapy. Medical University of Vienna Curr Opin Oncol (in press). 1090 Vienna, Austria [email protected] 1. Kostler WJ, Hejna M, Wenzel C, Zielinski CC. Oral mucositis complicating chemotherapy and/or radiotherapy: options for pre- THIS WEEK’S LETTERS vention and treatment. CA Cancer J Clin 2001;51:290-315. 1264 Palifermin and Chemotherapy-Induced Oral Mucositis to the editor: A number of questions remain regarding the safe use of recombinant human ke- 1266 Care of the Potential Organ Donor ratinocyte growth factor to decrease oral mucosal 1267 Contagious Acute Gastrointestinal Infections damage after total-body irradiation and high-dose chemotherapy in patients with hematologic cancers. 1268 Cancer of the Ovary Spielberger et al. simply list the overall incidence of 1269 Case 37-2004: Postmenopausal Bleeding adverse events, without grading their severity using and a Cystic Ovarian Mass the World Health Organization (WHO) toxicity scale or Common Toxicity Criteria.1 Meropol and co- 1270 Why “Why” Matters workers previously observed severe dose-limiting 1271 Nosocomial Transmission of Cryptococcosis cutaneous toxic effects that were associated with recombinant human keratinocyte growth factor.2 Giv-

1264 n engl j med 352;12 www.nejm.org march 24, 2005

correspondence en that recombinant human keratinocyte growth gard to baseline demographic and clinical charac- factor is being used as supportive care to ameliorate teristics. The primary analysis was prospectively the adverse effects of cytotoxic drugs and radiation defined and presented in our article. Additional therapy, we believe it is vital to demonstrate that this analyses were not included because of space con- agent itself does not cause unacceptable toxic ef- straints and because the primary analysis showed fects. In addition, there is concern that recombi- unequivocal efficacy. Analysis of covariates indicat- nant human keratinocyte growth factor may stimu- ed that patients received a benefit from palifermin late the proliferation of tumor cells.3 The safety of in all subgroups analyzed. Although data on oral recombinant human keratinocyte growth factor hygiene, presence or absence of underlying oral dis- must be properly evaluated in ongoing clinical tri- orders, and smoking status were not collected, we als4 before it can be accepted into routine clinical examined the remaining potential prognostic fac- practice. tors associated with the incidence and severity of Carlo Palmieri, M.B., B.S., Ph.D. oral mucositis. The incidence of severe oral mucosi- David Vigushin, M.B., B.Ch., Ph.D. tis (WHO grade ≥3) was lower among patients who Imperial College London received palifermin, regardless of whether the pa- London W12 0NN, United Kingdom tient had undergone radiotherapy. This difference [email protected] was also seen regardless of baseline performance 1. Trotti A, Colevas AD, Setser A, et al. CTCAE v3.0: development status, age, and weight. The management of oral hy- of a comprehensive grading system for the adverse effects of cancer treatment. Semin Radiat Oncol 2003;13:176-81. giene and oral disease before the start of the treat- 2. Meropol NJ, Somer RA, Gutheil J, et al. Randomized phase I tri- ment protocol was specific to the standard of care in al of recombinant human keratinocyte growth factor plus chemo- each study center, and the results of primary analy- therapy: potential role as mucosal protectant. J Clin Oncol 2003;21: 1452-8. ses were similar across centers. In addition, the use 3. Powers CJ, McLeskey SW, Wellstein A. Fibroblast growth fac- of antiviral drugs was well balanced between the tors, their receptors and signaling. Endocr Relat Cancer 2000;7: study groups. 165-97. We agree with Awada and colleagues that low- 4. Clarke SJ, Abdi A, Davies ID, et al. Recombinant human kerati- 2 nocyte growth factor (rHuKGF) prevents chemotherapy-induced level laser therapy has not been fully evaluated, and mucositis in patients with advanced colorectal cancer: a random- we would welcome continued exploration of its use ized phase II trial. Proc Am Soc Clin Oncol 2001;20:383a. abstract. and mechanism of action. Investigators who support further development of low-level laser therapy the authors reply: Drs. Palmieri and Vigushin re- might find it helpful to compare this approach with fer to the study by Meropol et al.,1 which was a phase the standard of care, palifermin, or both in a ran- 1 trial of a fluorouracil-based regimen for metastatic domized trial. colorectal cancer.1 In that study, one patient of eight Ricardo Spielberger, M.D. treated with palifermin at a dose of 60 µg per kilo- City of Hope National Medical Center gram of body weight (the dose used in our study) Duarte, CA 91010 had grade 3 skin toxic effects that required discon- Patrick Stiff, M.D. tinuation of the drug. In our study, adverse events Loyola University Medical Center were mild to moderate (equivalent to WHO Com- Maywood, IL 60153 mon Toxicity Criteria grade 1 or 2), transient, and William Bensinger, M.D. not a cause for discontinuation of the study drug. Fred Hutchinson Cancer Research Center We found that there were no secondary cancers re- Seattle, WA 98109 lated to keratinocyte growth factor and that the rates 1. Meropol NJ, Somer RA, Gutheil J, et al. Randomized phase I trial of recombinant human keratinocyte growth factor plus chemo- of progression-free survival at approximately one therapy: potential role as mucosal protectant. J Clin Oncol 2003;21: year of follow-up were identical in the two groups. 1452-8. In regard to the comments by Köstler and col- 2. Rubenstein EB, Peterson DE, Schubert M, et al. Clinical practice guidelines for the prevention and treatment of cancer therapy- leagues, our study was randomized, and patients induced oral and gastrointestinal mucositis. Cancer 2004;100:Sup- were balanced between the study groups with re- pl 9:2026-46.

n engl j med 352;12 www.nejm.org march 24, 2005 1265

The new england journal of medicine

Care of the Potential Organ Donor

to the editor: In their review article on the care of “colloid-containing solutions are...not more the potential organ donor, Wood et al. (Dec. 23 is- effective [than crystalloids] in preserving pulmo- sue)1 emphasize the importance of the correction nary function”2 and they are much more expensive of hypovolemia and recommend the use of sodium than crystalloids. Furthermore, the recent Saline bicarbonate to treat acidosis. They also remind the versus Albumin Fluid Evaluation Study concluded reader that hypernatremia in the donor can ad- that saline and albumin are equivalent for volume versely affect the function of the transplant. resuscitation in patients in the intensive care unit.3 Fluid administration, as part of an early, goal- Finally, although the value of hormonal therapy directed therapy, is beneficial in critical care patients has not been clearly established, it has nevertheless with shock.2 However, there is no evidence to sup- been incorporated into the United Network for Or- port the use of bicarbonate therapy in the treatment gan Sharing critical pathway for the care of donors, of hypoperfusion-induced acidemia. Alkali thera- which recommends that “hormonal resuscitation” py does not improve hemodynamics, does not re- be instituted when conventional management fails.4 duce the vasopressor requirement, and may even In contrast, Wood and colleagues recommend try- be harmful.3 It can lead to hypernatremia and ex- ing other vasopressors first. tracellular fluid volume overload.4 Hemodynamic Many questions remain regarding optimal care management with the use of adequate fluid resusci- of donors. Until they are answered, practice will be tation, vasopressors, and inotropic therapy is prob- based primarily on experience, rather than hard ably the cornerstone of the treatment of the poten- science. tial organ donor. The use of sodium bicarbonate as Aaron Spital, M.D. standard therapy in the case of acidosis may be more New York Organ Donor Network deleterious than beneficial, is highly questionable, New York, NY 10001 and cannot be recommended. [email protected] Frédéric M. Jacobs, M.D. 1. Shames BD, D’Alessandro AM, Sollinger HW. Human T-cell lymphotrophic virus infection in organ donors: a need to reassess Hôpital Antoine Béclère policy? Am J Transplant 2002;2:658-63. 92140 Clamart, France 2. Rose BD, Post TW. Clinical physiology of acid-base and electro- [email protected] lyte disorders. 5th ed. New York: McGraw-Hill, 2001. 1. Wood KE, Becker BN, McCartney JG, D’Alessandro AM, Coursin 3. The SAFE Study Investigators. A comparison of albumin and sa- DB. Care of the potential organ donor. N Engl J Med 2004;351:2730- line for fluid resuscitation in the intensive care unit. N Engl J Med 9. 2004;350:2247-56. 2. Rivers E, Nguyen B, Havstad S, et al. Early goal-directed therapy 4. Rosengard BR, Feng S, Alfrey EJ, et al. Report of the Crystal City in the treatment of severe sepsis and septic shock. N Engl J Med meeting to maximize the use of organs recovered from the cadaver 2001;345:1368-77. donor. Am J Transplant 2002;2:701-11. 3. Dellinger RP, Carlet JM, Masur H, et al. Surviving Sepsis Cam- paign guidelines for management of severe sepsis and septic shock. Crit Care Med 2004;32:858-73. [Erratum, Crit Care Med 2004;32: the authors reply: The goal of our review was to 2169-70.] 4. Mathieu D, Neviere R, Billard V, Fleyfel M, Wattel F. Effects of bi- provide recommendations derived from the best carbonate therapy on hemodynamics and tissue oxygenation in pa- existing evidence and our collective experience in tients with lactic acidosis: a prospective, controlled clinical study. the care of donors at a comprehensive organ trans- Crit Care Med 1991;19:1352-6. plantation center. Unfortunately, evidence-based guidelines supported by prospective, randomized, to the editor: The review article by Wood and col- controlled trials in this population are virtually leagues contains several puzzling statements. First, nonexistent. The United Network for Organ Shar- infection with the human T-cell leukemia–lympho- ing regards the use of organs from donors infected ma virus (HTLV) is said to contraindicate donation, with HTLV-I or HTLV-II as a protocol violation, un- but an earlier report from the same institution rec- less subsequent confirmatory testing unequivocally ommended that HTLV-positive persons not be cat- indicates that the initial test result was a false posi- egorically excluded from donating organs.1 Second, tive result.1 Given the low prevalence of HTLV-I and colloid solutions are recommended for potential HTLV-II among donors and the potential for false lung donors to minimize the formation of pulmo- positive results with the use of highly sensitive nary edema. However, there are data indicating that screening assays, Shames et al. have recommended

1266 n engl j med 352;12 www.nejm.org march 24, 2005

correspondence that careful consideration be given to transplanting the complex care of this population. Its use should organs from these donors.2 be undertaken on an individualized basis. The optimal use of colloid or crystalloid in select- Kenneth E. Wood, D.O. ed populations is open to discussion. Colloid has Douglas Coursin, M.D. been recommended and has been associated with Anthony D’Alessandro, M.D. 3 higher rates of lung procurement than crystalloid. University of Wisconsin Hospital and Clinics In the absence of randomized or controlled tri- Madison, WI 53792-9988 als, the use and timing of hormonal resuscitation [email protected] remain speculative. Our algorithm recommending 1. Policy 4.0. Acquired immune deficiency syndrome (AIDS) and human pituitary derived growth hormone (HPDGH) and human early placement of a pulmonary-artery catheter is T-lymphotrophic virus type I (HTLV-I). (Accessed February 23, derived from an evolving literature and from expe- 2005, at http://www.unos.org/PoliciesandBylaws/policies/pdfs/ rience that suggests that improved fluid and hemo- policy_16.pdf.) 2. Shames BD, D’Alessandro AM, Sollinger HW. Human T-cell dynamic management result with this approach. lymphotrophic virus infection in organ donors: a need to reassess Cardiac recommendations by the Crystal City Meet- policy? Am J Transplant 2002;2:658-63. ing Group, cited by Spital, suggest that aggressive 3. Follette D, Rudich S, Bonacci C, Allen R, Hoso A, Albertson T. Importance of an aggressive multidisciplinary management ap- hemodynamic management with a pulmonary- proach to optimize lung donor procurement. Transplant Proc 1999; artery catheter be initiated simultaneously with hor- 31:169-70. monal resuscitation.4 4. Zaroff JG, Rosengard BR, Armstrong WF, et al. Consensus Conference Report: maximizing the use of organs recovered from The use of sodium bicarbonate remains contro- the cadaveric donor: cardiac recommendations. Circulation 2002; versial, and in that way is similar to many options in 106:836-41.

Contagious Acute Gastrointestinal Infections to the editor: The review of contagious acute gastrointestinal infections by Musher and Musher (Dec. 2 issue)1 is incomplete with respect to bacterial causes of acute gastrointestinal illness. With more than 50 million international travelers a year and rates of incidence of acute diarrhea among such travelers as high as 50 percent,2 enterotoxigenic and enteroaggregative Escherichia coli are important causes of acute bacterial diarrhea, accounting for 30 percent and 26 percent of cases, respectively.3 Enterotoxigenic E. coli and enteroaggregative E. coli infections are spread through water or food contaminated by bacteria from the feces of infected persons.3,4 A high inoculum (108 colony-forming units) supports the role of foodborne transmission.4 The incubation period of these bacterial infections is less than 24 hours. Methods for the diagnosis of enterotoxigenic E. coli infection rely on the detection of the heat-labile or heat-stable en- terotoxins, or both. Identification of enteroaggregative E. coli relies on a pattern of aggregative adher- 5 ence to HEp-2 cells (Fig. 1). Diarrhea caused by Figure 1. HEp-2 Cell Assay for Enteroaggregative E. coli. these bacteria is usually self-limited, with most cas- The aggregative pattern of adherence to the surface es resolving within three days in the absence of of the HEp-2 cell tissue culture, other bacteria, and the treatment. That said, enterotoxigenic E. coli and en- glass coverslip is characteristic of enteroaggregative teroaggregative E. coli are common bacterial causes E. coli (crystal violet stain). of acute gastrointestinal illness that should be con-

n engl j med 352;12 www.nejm.org march 24, 2005 1267

The new england journal of medicine

sidered, especially in travelers to developing coun- Considering the frequency of traveler’s diarrhea, tries. the absence of evidence of human-to-human spread David B. Huang, M.D., M.P.H. is all the more notable; in fact, a clinical experiment Baylor College of Medicine designed to detect transmission yielded negative re- 1 Houston, TX 77030 sults. As we emphasized, this lack of contagion is [email protected] readily understandable in light of the very high inoc- 2,3 1. Musher DM, Musher BL. Contagious acute gastrointestinal in- ulum that is needed to transmit these infections. fections. N Engl J Med 2004;351:2417-27. Although every case of acute gastroenteritis is im- 2. DuPont HL, Ericsson CD. Prevention and treatment of traveler’s diarrhea. N Engl J Med 1993;328:1821-7. portant to the person who has it, the great majority 3. Adachi JA, Jiang ZD, Mathewson JJ, et al. Enteroaggregative of cases in the United States occur in persons who Escherichia coli as a major etiologic agent in traveler’s diarrhea in have not traveled, and most of these are caused by 3 regions of the world. Clin Infect Dis 2001;32:1706-9. 4. Huang DB, Okhuysen PC, Jiang ZD, DuPont HL. Enteroaggre- viruses. gative Escherichia coli: an emerging enteric pathogen. Am J Gastro- Daniel Musher, M.D. enterol 2004;99:383-9. 5. Nataro JP, Kaper JB. Diarrheagenic Escherichia coli. Clin Micro- Michael F. DeBakey Veterans Affairs Medical Center biol Rev 1998;11:142-201. [Erratum, Clin Microbiol Rev 1998;11: Houston, TX 77030 403.] [email protected] Benjamin L. Musher, M.D. the authors reply: University of Pennsylvania School of Medicine As we stated in the introducto- Philadelphia, PA 19104 ry paragraph of our article, our focus was on infec- 1. Levine MM, Rennels MB, Cisneros L, Hughes TP, Nalin DR, tions that spread from one person to another, not Young CR. Lack of person-to-person transmission of enterotoxi- on foodborne or waterborne disease. We remain un- genic Escherichia coli despite close contact. Am J Epidemiol 1980; able to find reports of documented secondary infec- 111:347-55. 2. DuPont HL, Formal SB, Hornick RB, et al. Pathogenesis of Esch- tion — that is, acute gastroenteritis among close erichia coli diarrhea. N Engl J Med 1971;285:1-9. contacts of persons who have returned from anoth- 3. Nataro JP, Deng Y, Cookson S, et al. Heterogeneity of enteroag- er country with traveler’s diarrhea due to enterotox- gregative Escherichia coli virulence demonstrated in volunteers. J Infect Dis 1995;171:465-8. igenic or enteroaggregative E. coli — which is why we did not include them in our article.

Cancer of the Ovary

to the editor: In many women with ovarian can- 1. Cannistra SA. Cancer of the ovary. N Engl J Med 2004;351: cer, moderate-to-severe pain develops, and many 2519-29. 2. Kornblith AB, Thaler HT, Wong G, et al. Quality of life of wom- have recurrent bowel obstructions that severely im- en with ovarian cancer. Gynecol Oncol 1995;59:231-42. pair the quality of their lives. The otherwise compre- 3. Portenoy RK, Kornblith AB, Wong G, et al. Pain in ovarian can- hensive review of cancer of the ovary by Cannistra cer patients: prevalence, characteristics, and associated symptoms. 1 Cancer 1994;74:907-15. (Dec. 9 issue) would have benefited from including 4. Ripamonti C, Twycross R, Baines M, et al. Clinical-practice rec- reviews of studies of the quality of life of women ommendations for the management of bowel obstruction in pa- with ovarian cancer,2 management of their often tients with end-stage cancer. Support Care Cancer 2001;9:223-33. 3 5. Elit L, Charles C, Gold I, et al. Women’s perceptions about treat- neuropathic pain, nonsurgical options for the treatment decision making for ovarian cancer. Gynecol Oncol 2003;88: ment of bowel obstruction,4 and medical decision 89-95. making.5 Similarly, the discussion should have covered the benefits and limitations of parenteral nu- trition, hospice referral, and the psychosocial as- to the editor: Cannistra does not discuss imaging pects of care. — computed tomography, magnetic resonance imaging, and positron-emission tomography — in Ursula Matulonis, M.D. the differential diagnosis of ovarian cancer.1 De- Janet L. Abrahm, M.D. spite the difficulty of diagnosing ovarian cancer ear- Dana–Farber Cancer Institute Boston, MA 02115 ly, the benefit of measuring a combination of tumor [email protected] markers such as CA-125, carcinoembryonic anti-

1268 n engl j med 352;12 www.nejm.org march 24, 2005

correspondence gen, and squamous-cell carcinoma–associated an- surgery for definitive diagnosis is usually based on tigen might provide an alternative means of find- the clinical examination and findings on transvag- ing some early-stage disease of the ovary. Ovarian inal ultrasonography, as outlined in my review.1 Dr. cancer may mimic gastrointestinal conditions such Liu also mentions that the use of a combination of as Krukenberg’s tumor and colorectal adenocarci- serum tumor markers might be a more effective noma.2 Furthermore, although the use of laparo- screening strategy than the use of CA-125 by itself scopic surgery in treating ovarian tumors has been — a concept that has been proposed by others and increasing in recent years, this article fails to discuss is currently under investigation.2 the advantages and disadvantages of such surgery Finally, Dr. Liu raises the possibility that “sec- for ovarian tumors and does not recommend fol- ond-look” laparoscopy or laparotomy might be use- low-up (“second-look”) laparoscopy or laparotomy ful in the detection of persistent disease or early re- for early detection of recurrent ovarian cancer. lapse. For patients with advanced epithelial ovarian Ching-Ming Liu, M.D., M.P.H. cancer in whom a clinical complete response occurs Chang Gung Memorial Hospital after first-line therapy, second-look surgical tech- Tao-yuan 333, Taiwan niques are known to be more sensitive than clinical [email protected] examination or radiographic studies in detecting 1 1. Kawahara K, Yoshida Y, Kurokawa T, et al. Evaluation of posi- subclinical, persistent disease. However, the cen- tron emission tomography with tracer 18-fluorodeoxyglucose in ad- tral issue is whether early institution of second-line dition to magnetic resonance imaging in the diagnosis of ovarian cancer in selected women after ultrasonography. J Comput Assist chemotherapy for subclinical, asymptomatic dis- Tomogr 2004;28:505-16. ease confers a benefit similar to that of treatment 2. Yedema CA, Kenemans P, Wobbes T, et al. Use of serum tumor administered at the time of clinical relapse. To date, markers in the differential diagnosis between ovarian and colorectal adenocarcinomas. Tumour Biol 1992;13:18-26. there is no convincing evidence to support the use of early second-line chemotherapy,3 although in the future it may be reasonable to revisit this issue, after dr. cannistra replies: Drs. Matulonis and Abrahm more effective salvage therapy has become available. write about quality-of-life issues, which are, of Stephen A. Cannistra, M.D. course, important. Space limitations precluded in- Beth Israel Deaconess Medical Center clusion of this topic in my review, and the referenc- Boston, MA 02215 es in their letter provide helpful information for [email protected] 1. Cannistra SA. Cancer of the ovary. N Engl J Med 2004;351: readers. 2519-29. Dr. Liu asks whether imaging studies with com- 2. Skates SJ, Horick N, Yu Y, et al. Preoperative sensitivity and spec- puted tomography, magnetic resonance imaging, ificity for early-stage ovarian cancer when combining cancer antigen CA-125II, CA 15-3, CA 72-4, and macrophage colony-stimulat- and positron-emission tomography play a role in the ing factor using mixtures of multivariate normal distributions. J Clin differential diagnosis of patients with an ovarian Oncol 2004;22:4059-66. mass. In general, it is not necessary to perform these 3. Ozols RF, Bundy BN, Greer BE, et al. Phase III trial of carboplat- in and paclitaxel compared with cisplatin and paclitaxel in patients studies preoperatively in cases of suspected epithe- with optimally resected stage III ovarian cancer: a Gynecologic On- lial ovarian cancer, and the decision to proceed with cology Group study. J Clin Oncol 2003;21:3194-200.

Case 37-2004: Postmenopausal Bleeding and a Cystic Ovarian Mass to the editor: Although the words “careful stag- same issue, “biopsy of the para-aortic nodes is espe- ing,” “full staging procedure” and “optimal stag- cially important in patients with disease that other- ing” were used in the discussion of Case 37-2004 wise appears to be limited to the ovary, since such (Dec. 9 issue),1 that of a woman who appeared to patients may have more advanced disease.”2 have early-stage ovarian cancer, one point deserves The patient described in the Case Records may amplification. As Cannistra pointed out in his re- indeed have had an early-stage tumor, but she did view of the management of ovarian cancer in the not undergo the operation required to determine the

n engl j med 352;12 www.nejm.org march 24, 2005 1269 The new england journal of medicine

full extent of her disease with certainty. A positive sis of macroscopic examination were found to have para-aortic lymph-node biopsy makes the differ- a higher stage of disease when careful pathological ence between stage IA and stage IIIC disease. staging was performed.1 In a seminal study of the Surgical staging of ovarian cancer is based on incidence of advanced disease in patients with ap- an understanding of its patterns of spread. Disease parent stage I ovarian cancer, 11 percent were found that is apparently confined to the pelvis affects para- to have diaphragmatic metastases; 13 percent, aor- aortic nodes with a predictable frequency.3 The ap- tic lymph-node metastases; 8 percent, pelvic lymph- propriate management of this disease by specialists node metastases; 3 percent, microscopic omental in gynecologic oncology can improve outcome and metastases; and 33 percent, malignant peritoneal survival.4 washings.2 Fredric V. Price, M.D. In a 1993 study involving patients with ovarian Pittsburgh Gynecologic Oncology cancer who were undergoing systematic aortic and Pittsburgh, PA 15224 pelvic lymphadenectomy, positive nodes were found [email protected] in 14 percent of patients who were initially thought 3 1. Case Records of the Massachusetts General Hospital (Case 37- to have stage I disease. Lymphatic spread was ipsi- 2004). N Engl J Med 2004;351:2531-8. lateral to the tumor in all cases, suggesting that ra- 2. Cannistra SA. Cancer of the ovary. N Engl J Med 2004;351:2519- 29. tional regional dissection is appropriate and should 3. Young RC, Decker DG, Wharton JT, et al. Staging laparotomy in always include the para-aortic nodes, since the ovar- early ovarian cancer. JAMA 1983;250:3072-6. ian lymphatics drain with the ovarian veins to the 4. Junor EJ, Hole DJ, McNulty L, Mason M, Young J. Specialist gy- naecologists and survival outcome in ovarian cancer: a Scottish na- renal vein on the left and the inferior vena cava on tional study of 1866 patients. Br J Obstet Gynaecol 1999;106:1130- the right. 6. Richard T. Penson, M.D. Dushyant Sahani, M.D. Debra A. Bell, M.D. the authors reply: We completely agree that full Massachusetts General Hospital staging of epithelial ovarian cancer requires nodal Boston, MA 02114 dissection. In fact, the patient did undergo bilateral 1. Young RC, Decker DG, Wharton JT, et al. Staging laparotomy in pelvic-node dissection, as well as para-aortic node early ovarian cancer. JAMA 1983;250:3072-6. exploration as part of her staging laparotomy; no 2. Piver MS, Barlow JJ, Lele SB. Incidence of subclinical metastasis involved nodes were detected. in stage I and II ovarian carcinoma. Obstet Gynecol 1978;52:100-4. 3. Benedetti-Panici P, Greggi S, Maneschi F, et al. Anatomical and In the study cited by Price, 28 percent of patients pathological study of retroperitoneal nodes in epithelial ovarian who were thought to have stage I disease on the ba- cancer. Gynecol Oncol 1993;51:150-4.

Why “Why” Matters

to the editor: Janssen et al., in their Clinical Prob- Thus, the patient’s oxygen saturation on pulse ox- lem-Solving article (Dec. 2 issue),1 present a case of imetry should have been much higher than 85 per- glucose-6-phosphate dehydrogenase (G6PD) defi- cent and would have improved slightly with the ad- ciency with hemolysis and methemoglobinemia. ministration of 100 percent oxygen. Of course, the They report that the patient had an oxygen-satura- true oxygen-saturation value could be measured by tion value of 85 percent as measured by pulse oxim- co-oximetry. etry that did not improve with the administration of 100 percent oxygen. This would be true if the met- Edwin W. Grimsley, M.D. hemoglobin level were greater than 35 percent, but Mercer University School of Medicine Macon, GA 31201 the patient’s level was 8.8 percent. At levels up to [email protected] 35 percent, oxygen saturation on oximetry decreases by an amount proportional to the concentration 1. Janssen WJ, Dhaliwal G, Collard HR, Saint S. Why “why” mat- of methemoglobin until the latter reaches approxi- ters. N Engl J Med 2004;351:2429-34. 2. Barker SJ, Tremper KK, Hyatt J. Effects of methemoglobinemia mately 35 percent. At higher methemoglobin levels, on pulse oximetry and mixed venous oximetry. Anesthesiology 2 the oxygen saturation levels out at about 85 percent. 1989;70:112-7.

1270 n engl j med 352;12 www.nejm.org march 24, 2005 correspondence the authors reply: We appreciate Dr. Grimsley’s of these variables may have contributed to the dis- comments regarding pulse oximetry and methe- crepancy between the predicted oxygen-saturation moglobin. Pulse oximeters estimate oxygen satura- value and the one recorded in our patient. tion by comparing the absorbance of light at two We agree with Dr. Grimsley that although pulse wavelengths. Infrared light is absorbed by oxyhemo- oximetry provides a useful, noninvasive estimate globin, whereas red light is absorbed by deoxyhe- of oxygen saturation, co-oximetry should be per- moglobin. The absorbance characteristics of met- formed in all patients in whom methemoglobine- hemoglobin are similar to those of oxyhemoglobin, mia is suspected. 1,2 falsely elevating pulse-oximeter readings. William J. Janssen, M.D. Co-oximeters measure light absorbance at four Harold R. Collard, M.D. or more discrete wavelengths, providing accurate University of Colorado Health Sciences Center measurement of oxygen saturation, methemoglo- Denver, CO 80262 bin, and carboxyhemoglobin. In our patient, the [email protected] oxygen saturation as measured by co-oximetry was Gurpreet Dhaliwal, M.D. 88.9 percent, the methemoglobin level was 8.8 per- University of California, San Francisco cent, and the carboxyhemoglobin level was 1.2 San Francisco, CA 94143 percent. 1. Eisenkraft JB. Pulse oximeter desaturation due to methemoglobinemia. Anesthesiology 1988;68:279-82. A methemoglobin level of 8.8 percent would be 2. Jensen LA, Onyskiw JE, Prasad NG. Meta-analysis of arterial ox- expected to yield a pulse-oximeter reading of 92 ygen saturation monitoring by pulse oximetry in adults. Heart Lung percent,3 not 85 percent. A number of factors can 1998;27:387-408. 3. Barker SJ, Tremper KK, Hyatt J. Effects of methemoglobinemia impair the function of a pulse oximeter, including on pulse oximetry and mixed venous oximetry. Anesthesiology motion by the patient, the presence of or an excess 1989;70:112-7. of ambient light, increased skin pigmentation, ab- 4. Beall SN, Moorthy SS. Jaundice, oximetry, and spurious hemoglobin desaturation. Anesth Analg 1989;68:806-7. normal body temperature, hyperbilirubinemia, and 5. Fitzgerald RK, Johnson A. Pulse oximetry in sickle cell anemia. 2,4,5 severe anemia. It is possible that one or more Crit Care Med 2001;29:1803-6.

Nosocomial Transmission of Cryptococcosis to the editor: Human-to-human transmission of photericin B. Unfortunately, he died of septic shock cryptococcosis is very rare.1-3 An 80-year-old man and multiorgan failure associated with methicillin- (Patient A) was hospitalized because of pulmonary resistant Staphylococcus aureus bacteremia on Sep- squamous-cell carcinoma. He received mechanical tember 14, 2004. ventilation and was transferred to the medical in- The minimal inhibitory concentrations of flu- tensive care unit on July 6, 2004. A new infiltrate in conazole and amphotericin B, determined with use the right lung and pleural effusion associated with of the E Test (AB Biodisk), were identical among fever developed on August 15, 2004. Cultures of the isolates A, B1, B2, and B3 (0.125 mg per milliliter pleural effusion drawn on August 19 yielded Cryp- for fluconazole and 8 mg per milliliter for ampho- tococcus neoformans (isolate A). The patient died of re- tericin B). Molecular typing of the four isolates and fractory respiratory failure on September 19, 2004. eight other isolates (control strains) by arbitrarily A 77-year-old man (Patient B) was admitted to primed polymerase-chain-reaction analyses with the the same medical intensive care unit on July 27, use of four random oligonucleotide primers (M13, 2004, and occupied the bed next to Patient A. This OPH-15, OPH-19, and ERIC1) and karyotyping dis- patient had disseminated cryptococcal infection. closed that the isolates from the two patients were C. neoformans was isolated from specimens of his identical and were different from the eight control blood (isolate B1), cerebrospinal fluid (isolate B2), strains.4,5 sputum, bronchoalveolar-lavage fluid, and urine C. neoformans is susceptible to 1 percent sodium (isolate B3). Large amounts of the fungus were seen hypochlorite and iodine, but its susceptibility to 70 in the latter four specimens. The patient had a sat- percent ethanol and ethylene oxide is questionable. isfactory response to treatment with liposomal am- The hands of medical personnel and the instru-

n engl j med 352;12 www.nejm.org march 24, 2005 1271 correspondence

ments routinely used in the respiratory intensive Cheng-Yi Wang, M.D. care unit (such as ventilators and ultrasound equip- Huey-Dong Wu, M.D. ment) might be contaminated by secretions from Po-Ren Hsueh, M.D. an index patient. Patient A might have acquired the National Taiwan University Hospital organism through direct inhalation of airborne fun- Taipei 100, Taiwan gus as a result of the short distance between the two [email protected] beds or through a respiratory care procedure (such 1. Beyt BE Jr, Waltman WR. Cryptococcal endophthalmitis after corneal transplantation. N Engl J Med 1978;298:825-6. as sputum suction) performed with contaminated 2. Glaser JB, Garden A. Inoculation of cryptococcosis without instruments by medical personnel. Unfortunately, transmission of the acquired immunodeficiency syndrome. N Engl J cultures of the environment (floor, bed, ventilator Med 1985;313:266. 3. Sirinavin S, Intusoma U, Tuntirungsee S. Mother-to-child trans- surface, monitoring equipment, radiograph-view- mission of Cryptococcus neoformans. Pediatr Infect Dis J 2004;23:278- ing box, and nursing and emergency carts) one 9. month after the notification of this event were all 4. Chen SC, Brownlee AG, Sorrel TC, Ruma P, Ninno G. Identifica- tion by random amplification of polymorphic DNA of a common negative for C. neoformans. molecular type of Cryptococcus neoformans var neoformans in patients In conclusion, we report a case of likely noso- with AIDS or other immunosuppressive conditions. J Infect Dis comial spread of cryptococcosis. Strict adherence to 1996;173:754-8. 5. Perfect JR, Ketabchi N, Cox GM, Ingram CW, Beiser CL. Karyo- appropriate infection-control measures is manda- typing of Cryptococcus neoformans as an epidemiological tool. J Clin tory to control the spread of cryptococcosis in hos- Microbiol 1993;31:3305-9. pitals. Correspondence Copyright © 2005 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 three 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 original articles.) •Include your full mailing address, telephone number, fax number, and e-mail address with your letter.

Our Web address: 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.

1272 n engl j med 352;12 www.nejm.org march 24, 2005 The new england journal of medicine

book reviews

szasz under fire: immune to arguments on the basis of data; they are the psychiatric abolitionist premises, not conclusions. faces his critics Szasz’s central thesis is that “disease” means an abnormality of the body, and since doctors treat (The Under Fire Series.) Edited by Jeffrey A. Schaler. 450 pp. Chicago, Open Court, 2004. $36.95. ISBN 0-8126-9568-2. bodies, there may be brain diseases but not mental diseases. Corollaries are that involuntary treatment homas szasz was the enfant terrible of mental disease violates fundamental liberties, tof 20th-century American psychiatry. His 1961 that mental disease should not be considered in as- book, The Myth of Mental Illness (New York, Paul B. sessing criminal responsibility, and that physicians Hoeber), and his searing intellect and take-no-pris- should have no privileged role in the prescription oners rhetorical style defined the terms of the dis- of drugs or in assisted suicide. course. Szasz under Fire consists of 12 essays by critics, Several of his critics argue with his definition of Szasz’s replies to each, and a brief autobiographical disease. They point out that diseases happen to peo- sketch. The essays are uneven; Szasz is more inter- ple, not bodies, and review the evidence that brain esting than his critics, and, right or wrong, he al- diseases underlie major psychiatric disorders. These ways wins the debate. The book provides a déjà vu arguments have no effect on Szasz, although they experience and stimulates reflection on what we are probably the chief reason that his position seems were arguing about then, why it seemed so impor- so out of date. Szasz also seems out of step with con- tant, and how we think about it today. temporary practice. He states that “the typical men- Szasz was born in Budapest in 1920, and al- tal patient...is hospitalized and treated without though he spoke almost no English, he emigrated his consent” (which has not been true since years be- to the United States in 1938. He ranked first in his fore Szasz’s residency) and that there has been lit- medical school class, but he didn’t really want to tle progress in the diagnosis and treatment of men- practice medicine. As he put it, “My true passion was tal illness (an assertion that would receive little literature, history, philosophy, politics — or, put support). more plainly, how and why people live, suffer, and For me, the underlying issue is Szasz’s view of die.” After a year of medical residency, he shifted to psychiatric patients as competent, autonomous psychiatry in order to “be eligible for training in adults who are different and who must be protected psychoanalysis, not to practice psychiatry.” He from a society that wants to infringe on their rights sought a platform from which to attack “the im- moral practices of civil commitment and the insanity defense.” This book, some 60 years later, continues that attack. Szasz’s psychiatric residency was unusual. He never worked on an inpatient unit, and when his chairman suggested that he should have experience with “seriously ill patients,” he quit the program. Szasz makes clear that his views about mental illness, involuntary treatment, and the insanity defense were well established before his exposure to psychiatry, psychoanalysis, or even medicine and that he was unusually successful at avoiding any experience that might have been relevant to them. Szasz’s views are entirely ideological; they have Thomas Szasz in 2002. nothing to do with empirical data and are therefore Photo by Jeffrey A. Schaler.

n engl j med 352;12 www.nejm.org march 24, 2005 1273 The new england journal of medicine

and uses that difference as a justification. It is an im- tionary expansion of knowledge about teratogene- portant perspective, and one that touches on funda- sis. It was the era of the realization that “the face mental values of our society, but tragically, it is less predicts the brain,” when physicians recognized relevant to the seriously mentally ill than to almost many face–brain malformation syndromes and cor- anyone else. An alternative view — that people with related them with abnormal karyotypes or exoge- mental illness are childlike, helpless, and in need of nous teratogens. Golden points out that wide pub- our care and protection — has little appeal to him. licity about thalidomide-induced phocomelia had He even seems to question the view’s premise; he primed the profession to expect the identification speaks of the child’s relationship to his parents as of other teratogens in humans. one based on domination and submission and ar- She reviews the polarized debates among reli- gues that psychiatry rests on “a coercive pediatric gionists, feminists, and legislators as to whether to model characterized by relations of domination and consider maternal alcoholism, with its potential for subjection.” If one starts with the view that parent- harming the fetus, as a moral failure or a disease ing is domination and pediatrics is coercive, the that requires compassionate treatment. Should the conclusion is that psychiatry is evil. law punish an alcoholic mother? Is the harmed child In this book Szasz is called “the most influential justified in suing her? Is brain impairment due to the ideologist of the ‘new’ antipsychiatry of the 1960s fetal alcohol syndrome a justifiable defense for a and 1970s” and “a powerful intellectual ally of the criminal or, as attorney Alan Dershowitz contends, civil liberties movement.” He forced a sometimes an “abuse excuse” that replaces personal responsi- reluctant profession to attend to the moral and eth- bility with a diagnostic label? In the debate over ical dimensions of its work, and though he is large- “medicalizing” deviancies such as alcoholism and ly wrong, his arguments have been immensely compulsive gambling as sicknesses, I would hope valuable. that physicians would prefer medicalization to pun- Robert Michels, M.D. ishment. Because each new discovery opens a Pandora’s Weill Medical College of Cornell University New York, NY 10021 box of reactions, physicians need to find effective [email protected] means of public education that will elicit productive responses from society. In this area, Golden high- lights the shortcomings of the news media, government agencies, and the courts and points to the re- message in a bottle: the making sistance of manufacturers to publicizing warnings of fetal alcohol syndrome that raise liability concerns or that may result in con-

By Janet Golden. 232 pp. Cambridge, Mass., troversial legislation. Harvard University Press, 2005. $25.95. ISBN 0-674-01485-5. Golden writes clearly, though occasionally rep- etitiously, and provides abundant references. She he author, a history professor, re- avoids personal polemics and evangelizing. Her t views the responses of medical, political, and modus operandi is to quote opposing viewpoints legal institutions to the fetal alcohol syndrome. in their historical context and then underline con- Leaving the biomedical discussion to standard med- tradictions. At times the reader may almost wish for ical textbooks, she focuses in this book on a social recommendations, but Golden eschews easy an- context beyond the consulting room. swers. Most physicians and health workers will find Golden recounts the reluctance of physicians the book interesting and provocative and will come and society to accept alcohol as a teratogen, in spite away with a much fuller appreciation of the com- of warnings dating back centuries. For example, plex responses that medical discoveries trigger in Josef Warkany’s monumental 1971 work on con- society. These are excellent outcomes for a book. genital malformations did not indict alcohol nor William DeMyer, M.D. even include it in the index (the complete syndrome Indiana University School of Medicine includes malformations of the face, viscera, and Indianapolis, IN 46202 brain). The concept of a fetal alcohol syndrome [email protected] emerged in the 1960s and 1970s, during a revolu-

1274 n engl j med 352;12 www.nejm.org march 24, 2005 book reviews

hideous absinthe: a history Paris and London used absinthe as one of many of the devil in a bottle symbols to express their defiance of traditional middle-class values. The banning of absinthe only add- By Jad Adams. 294 pp., illustrated. Madison, University of Wisconsin Press, 2004. $24.95. ISBN 0-299-20000-0. ed to its allure, and 20th-century writers such as Ernest Hemingway found the drink a useful device bsinthe, a strongly alcoholic drink for fashioning an image of manly daring. a with a reputation of mythic proportions, is Adams is equally skeptical of the scientific and again in fashion, 90 years after it was banned in medical evidence cited by those who sought to ban France and elsewhere as a cause of madness. The the drink. He argues that such supposedly objective drink, a potent concoction of wormwood, herbs, evidence was, and in some cases continues to be, and between 55 percent and 75 percent alcohol, shaped by social or cultural influences. His prime continues to fascinate, and Hideous Absinthe is one example is the work of Valentin Magnan, France’s of several recent books on the subject. Jad Adams, a foremost psychiatrist and a leading expert on alco- writer of biographies of literary and political fig- holism in the late 19th century. On the basis of lab- ures, focuses on two aspects of absinthe’s history: oratory experiments in which guinea pigs were in- its role in the bohemian literary and artistic move- jected with high doses of thujone, Magnan claimed ments of France and Britain and the efforts made to that absinthe produced a distinct and dramatic form ban the drink. of madness called absinthism. Adams explores the By the late 19th century, absinthe was not only a dubious scientific basis of Magnan’s experiments popular aperitif but, Adams neatly suggests, a drink and their popularization in anti-absinthe cam- of “display and provocation.” Artists and writers paigns. He is equally critical of more recent scien- celebrated its hallucinatory effects and referred to it tific research on thujone. affectionately as “the green fairy” or “holy water.” Adams has read widely on his topic, from ob- Doctors, politicians, and social critics, citing med- scure poetry to the latest scholarship, and he can ical research that indicated that one component of nicely juxtapose a scientific report on the suppos- wormwood, thujone, caused hallucinations and edly devastating effects of absinthe with a minor madness, denounced the drink as “a quick coach to poet’s witty riposte. He is careful to acknowledge the madhouse” and, in the United States, “the green his scholarly sources, making this a useful book for curse of France.” Adams examines the development those who want to read further. Historians or spe- of these competing myths, which, he argues, were cialists on addiction will question some assertions, primarily the product of artistic, social, or political such as the extent to which women consumed ab- preoccupations. The book ends with an acerbic ex- sinthe or the impact of the drink’s ban on later pro- amination of the recent revival of absinthe and its hibition movements. France, after all, banned ab- myths, thanks to clever publicity, an increased pub- sinthe, but the French temperance movement was lic tolerance of drug use, and the Internet. careful to distance itself from American Prohibi- Despite its sensationalist title and colorful dust tion, which it judged excessive and puritanical. But jacket, Hideous Absinthe is a careful and considered even specialists will learn something from this bal- account of a drink whose reputation has always ex- anced and informative account. Hideous Absinthe ceeded its consumption. Adams examines the ca- might also interest public officials contemplating reers and often prodigious alcohol intake of such a ban on reputedly harmful substances or those artists and writers as Paul Verlaine, Vincent Van drinkers who, enticed by clever publicity, have tried Gogh, Henri de Toulouse-Lautrec, and Oscar Wilde. absinthe and found its taste to be vile and its effects He is, however, skeptical of claims that absinthe was disappointing. As Adams observes, absinthe will the source of these artists’ creativity or of their some- always “fill the role required of it.” times self-destructive drinking habits. Only in ex- Patricia E. Prestwich, Ph.D. ceptional cases, such as Paul Gauguin’s perceptions University of Alberta of color, does Adams discern any influence of ab- Edmonton, AB T6G 2H4, Canada sinthe. Rather, he stresses that artists and writers in [email protected]

n engl j med 352;12 www.nejm.org march 24, 2005 1275 corrections

masters of the mind: exploring count of who merits the most pages or references in the story of mental illness the index: Sigmund Freud is far ahead of the pack; from ancient times to the new following him, in chronological order, are Hippoc- millennium rates, Philippe Pinel, Charles Darwin, Ivan Pavlov, Emil Kraepelin, Alfred Adler, Carl Jung, and Carl By Theodore Millon. 641 pp., illustrated. Hoboken, N.J., John Wiley & Sons, 2004. $34.95. ISBN 0-471-46985-8. Rogers. Are we making much progress in understanding his book is a comprehensive survey ourselves? What determines our behavior — genet- t of theories about the workings of our minds. ics, learning, or experience? How do these various Theodore Millon, a well-respected psychologist and factors interact? People interested in a warm and re- prolific writer, has divided his enterprise into seven markably well informed historical discussion of kinds of “stories”: philosophical, humanitarian, these questions will enjoy this book. neuroscientific, psychoanalytic, psychoscientific, Frances R. Frankenburg, M.D. sociocultural, and “personologic.” He covers the Bedford Veterans Affairs Medical Center history of each subject area, often beginning with Bedford, MA 01730 ancient Greece or Egypt, and writes with an ency- [email protected] clopedic knowledge of all subsequent time periods. Book Reviews Copyright © 2005 Massachusetts Medical Society. These histories are followed by his commentary and reflections. corrections A striking feature of the book, as suggested by the title, is the inclusion of capsule biographies of the many people who have contributed to this field. The Effect of Air Pollution on Lung Development from 10 to 18 Years of Age (September 9, 2004;351:1057-67). On page 1059, Nearly 90 portraits, drawn by Millon or his daugh- under the heading Statistical Analysis, in the list of model adjust- ter, accompany the biographies. Millon knows many ments, lines 18 through 20 of the left-hand column should have contemporary experts in psychology, and his de- read “the square of log values for height,” rather than “body-mass scriptions of most of them are often personal and index (the weight in kilograms divided by the square of the height in meters); the square of the body-mass index,” as printed. Also, highly complimentary. These descriptions, along in the right-hand column, lines 3 and 4 should have read “ordinal with the portraits, make the book pleasant and ac- variables for the field technician and the spirometer,” rather than cessible. Millon’s steadily cheerful tone is tempered “indicator variables,” as printed. These errors had no bearing on by his gloom in the last few pages, which concern the analysis of persons with clinically low lung function. Adjust- ment for body-mass index and indicator variables as intended the post–9/11 world. changes the results slightly but does not affect the reported asso- Millon’s professional interests lie in the area of ciations between growth and pollution during an eight-year peri- the taxonomy of personality disorders, and studies od. The corrected values for Table 2 (page 1062), Table 3 (1063), of personality take up a fair amount of this book. and Table 4 (1064) and for Figure 2 (page 1062) are available at the Journal’s Web site at www.nejm.org. He also describes multiple schools of psychothera- Combination of Isosorbide Dinitrate and Hydralazine in Blacks py. Readers with a medical background may be dis- with Heart Failure (November 11, 2004;351:2049-57). On page appointed that there is relatively little information 2053, in Table 2, under the Minnesota Living with Heart Failure about the disorders of the mind that occupy so much Questionnaire score (range, 0 to 105), there should not be a foot- clinical time, such as substance abuse and demen- note indicating significant difference at baseline between trial subjects treated with placebo and those treated with isosorbide tia. Perhaps this lack reflects the relative youth of dinitrate plus hydralazine. these fields. Anti–Interleukin-12 Antibody for Active Crohn’s Disease (Novem- Readers familiar with the fields covered in this ber 11, 2004;351:2069-79). On page 2069, the list of authors survey will not find anything particularly controver- omitted the name of Zhiqiong Yang, B.S., of the Mucosal Immu- sial or provocative. The breadth of the book does not nity Section, National Institute of Allergy and Infectious Disease, allow Millon to go into detail on any topic, and he is National Institutes of Health, Bethesda, Md. too generous and appreciative a scholar to stir up controversy by deriding anyone’s theories. His di- the journal’s web and e-mail addresses: vision of the subject into seven rather arbitrarily de- For letters to the Editor: authors.nejm.org fined areas and his historical coverage of each topic For information about the status of a submitted mean that there is a certain amount of repetition. manuscript: authors.nejm.org Who are the “masters of the mind”? At the risk To submit a meeting notice: [email protected] of being invidious, but to help clarify the scope of The Journal’s Web pages: www.nejm.org this book and Millon’s interests, here is a rough

1276 n engl j med 352;12 www.nejm.org march 24, 2005

correction

The Effect of Air Pollution on Lung Development from 10 to 18 Years of Age (September 9, 2004;351:1057-67). On page 1059, under the heading Statistical Analysis, in the list of model adjustments, lines 18 through 20 of the left-hand column should have read “the square of log values for height,” rather than “body-mass index (the weight in kilograms divided by the square of the height in meters); the square of the body-mass index,” as printed. Also, in the right- hand column, lines 3 and 4 should have read “ordinal variables for the field technician and the spirometer,” rather than “indicator variables,” as printed. These errors had no bearing on the analysis of persons with clinically low lung function. Adjustment for body-mass index and indicator variables as intended changes the results slightly but does not affect the reported associations between growth and pollution during an eight-year period. The corrected values for Table 2 (page 1062), Table 3 (1063), and Table 4 (1064) and for Figure 2 (page 1062) are available at the Journal’s Web site at www.nejm.org.

Table 2. Mean Levels of Growth in Pulmonary Function during the Eight-Year Study Period, from 1993 to 2001.*

Pulmonary-Function Measure Girls Boys Average Average Age of 10 yr Age of 18 yr 8-yr growth Age of 10 yr Age of 18 yr 8-yr growth FVC (ml) 2227 3757 1530 2392 5227 2835 FEV1 (ml) 1951 3286 1335 2050 4467 2418 MMEF (ml/sec) 2284 3686 1402 2255 4676 2420

* Levels at the ages of 10 and 18 years are derived from the growth model described in the Methods section. FVC denotes forced vital capacity, FEV1 forced expiratory volume in one second, and MMEF maximal midexpiratory flow rate.

Table 3. Difference in Average Growth in Lung Function over the Eight-Year Study Period from the Least to the Most Polluted Community.*

Pollutant FVC FEV1 MMEF Difference Difference Difference (95% CI) P Value (95% CI) P Value (95% CI) P Value ml ml ml/sec O3 10 a.m.–6 p.m. ¡36.4 (¡124.5 to 51.6) 0.38 ¡17.8 (¡101.7 to 66.0) 0.65 79.7 (¡155.5 to 314.8) 0.47 1-Hour maximal level ¡51.9 (¡135.1 to 31.3) 0.20 ¡37.1 (¡117.4 to 43.3) 0.33 34.5 (¡202.1 to 271.2) 0.75 NO2 ¡70.4 (¡139.8 to ¡1.1) 0.05 ¡89.1 (¡154.0 to ¡24.2) 0.01 ¡232.7 (¡410.4 to ¡54.9) 0.01 Acid vapor ¡81.0 (¡148.5 to ¡13.6) 0.02 ¡95.7 (¡159.7 to ¡31.6) 0.007 ¡191.8 (¡382.1 to ¡1.5) 0.05 PM10 ¡57.8 (¡150.4 to 34.9) 0.20 ¡85.6 (¡168.7 to ¡2.6) 0.04 ¡175.9 (¡413.9 to 62.0) 0.13 PM2.5 ¡51.0 (¡127.1 to 25.2) 0.17 ¡78.3 (¡145.6 to ¡11.0) 0.03 ¡191.8 (¡379.8 to 3.7) 0.05 Elemental carbon ¡65.4 (¡127.1 to ¡3.8) 0.04 ¡83.5 (¡141.2 to ¡25.9) 0.009 ¡185.2 (¡353.5 to ¡16.9) 0.03 Organic carbon ¡57.3 (¡155.4 to 40.8) 0.22 ¡91.7 (¡178.3 to ¡5.1) 0.04 ¡175.9 (¡428.2 to 76.5) 0.15

* Values are the differences in the estimated rate of eight-year growth at the lowest and highest observed levels of the indicated pollutant. Dif- ferences are scaled to the range across the 12 study communities in the average level of each pollutant from 1994 through 2000 as follows: 37.5 ppb of O3 (measured from 10 a.m. to 6 p.m.), 46.0 ppb of O3 (the one-hour maximal level), 34.6 ppb of NO2, 9.6 ppb of acid vapor, 51.4 µg of PM10 per cubic meter, 22.8 µg of PM2.5 per cubic meter, 1.2 µg of elemental carbon per cubic meter, and 10.5 µg of organic carbon per cubic meter. CI denotes confidence interval. corrections

1276 n engl j med 352;12 www.nejm.org march 24, 2005 corrections

1276 n engl j med 352;12 www.nejm.org march 24, 2005

The new england journal of medicine

images in clinical medicine

Severe Fecal Impaction

Yann Loubières, M.D. 49-year-old man with a history of paraplegia due to poliomy- Olivier Chereau, M.D. a elitis, surgical treatment of a lumbar ependymoma, and chronic renal insuffi- ciency due to a neuropathic bladder was admitted to the intensive care unit be- Centre Hospitalier de Poissy–Saint Germain-en-Laye cause of abdominal pain, vomiting and diarrhea, and hyperkalemia (serum potassium, 78105 Saint Germain-en-Laye, France 7 mmol per liter). Computed tomography of the abdomen showed a severe fecal impaction with marked compression of the abdominal viscera without any sign of colonic perforation. After rehydration and correction of the hyperkalemia, the patient underwent manual disimpaction under general anesthesia. His recovery was unremarkable. Copyright © 2005 Massachusetts Medical Society.

e12 n engl j med 352;12 www.nejm.org march 24, 2005