Chronicle of an Unforetold Death

SPECIAL ARTICLE

HEALTH CARE REFORM Chronicle of an Unforetold Death

Neil A. Holtzman, MD, MPH

riting in sorrow and anger, I express up front my potential conflict of interest in interpreting the facts surrounding the death of my wife, Barbara Starfield, W MD, MPH. Within hours after her sudden and unex- Barbara started taking low-dose aspi- pected death, I notified the dean of the rin after coronary insufficiency had been Johns Hopkins Bloomberg School of Pub- diagnosed 3 years before her death, and lic Health, on whose faculty she served, clopidogrel bisulfate (Plavix) after her right that Barbara had apparently died of a coro- main coronary artery had been stented nary occlusion. He relayed the news 6 months after the diagnosis. She reported around the world. Devoted to improving to the cardiologist that she bruised more effectiveness and equity in health care, easily while taking clopidogrel and bled Barbara received many tributes.1 longer following minor cuts. She had no Because she died while swimming alone, personal or family history of bleeding ten- an autopsy was required. The immediate dency or hypertension. cause of death was “pool drowning,” but the The autopsy findings and the official underlying condition, “cerebral hemor- lack of feedback prompted me to call at- rhage,” stunned me. The pathologist attrib- tention to deficiencies in medical care and uted the massive hemorrhage to cerebral clinical research in the United States amyloid angiopathy (CAA), listing “anti- (Table 1) reified by Barbara’s death and coagulation therapy” on the death certifi- how the deficiencies can be rectified. cate under “other significant conditions.” No Ironically, Barbara had written about all significant occlusion of any of her coro- of them. nary arteries was found. A scalp bruise ad- jacent to a larger bruise on her right tem- LACK OF COORDINATION OF CARE poralis muscle was observed, but no skull fractures. Patchy microhemorrhages were When patients die suddenly and unex- observed in the cerebral cortex. pectedly and are not in a health care fa- cility, no routine procedure is required for See also pages 1144 notifying their physicians, even if the patient is autopsied. (Unfortunately, only 8% and 1145 of deaths were autopsied in the United Cerebral amyloid angiopathy is not rare States in 2007, and only 2% in Barbara’s 5 in the elderly, estimated to occur in 8.0% age group. ) Had I, as next of kin, not in- of those aged 75 to 84 years (Barbara’s age formed Barbara’s physicians, they would group) and 12.1% in those aged 85 years not have learned of her death until she and older.3 However, a large proportion missed her next appointment, if then. of patients in whom CAA is diagnosed post Barbara strongly believed that for com- mortem do not die of cerebral hemor- mon conditions, primary care physicians rhage.3 Nonfatal microhemorrhage asso- should have primary responsibility for 6 ciated with CAA can be detected by mag- management, coordinating care with spe- netic resonance imaging.4 cialists as needed. By amending the standard death certificate to include the name Author Affiliation: Department of Pediatrics, The Johns Hopkins University of the decedent’s primary care physician School of Medicine, Baltimore, Maryland. and by requiring state departments of vi-

ARCH INTERN MED/ VOL 172 (NO. 15), AUG 13/27, 2012 WWW.ARCHINTERNMED.COM 1174

©2012 American Medical Association. All rights reserved. Downloaded From: http://archinte.jamanetwork.com/ by a Monash University Library User on 08/21/2012 tal statistics to notify the physician ming into the curved wall of the when the death certificate is filed, pool. Once she started to bleed, Table 1. Lessons From Barbara’s physicians will be made aware of clopidogrel drug regimen could have Death their patients’ deaths (and causes if made the bleeding worse. Patients Lack of coordination—providers are not the patient is autopsied). The pri- taking clopidogrel were signifi- notified of sudden and unexpected mary care physician should then be cantly more likely than patients tak- deaths responsible for notifying the spe- ing only aspirin to experience an in- Underreporting of possible adverse drug cialists involved in the patient’s care. crease in intracerebral hematoma events volume (P = .05) and possibly to Multimorbidity—best appreciated by UNDERREPORTING have twice the mortality rate (8 vs primary care physicians 14 Inadequacy of randomized controlled trials OF ADVERSE DRUG EVENTS 4 per 28 patients; P = .19). Potential bias in randomized controlled Greater awareness of the high fre- trials sponsored and supported by the In response to my request, Barba- quency of CAA in those aged 65 and pharmaceutical industry ra’s cardiologist submitted an ad- older might lead physicians caring verse drug event report to the US for elderly patients who are taking Food and Drug Administration antiplatelet drugs to be alert for tran- new drug’s safety (harm), as well as (FDA), stating that she was receiv- sient or increasing cognitive impair- its efficacy, in granting premarket ap- ing anticlotting medications that ment that might indicate microhem- proval. Important as they are, such might have contributed to her cere- orrhages and to follow up with a trials have limitations, as Barbara bral hemorrhage. The “report may thorough neurological evaluation pointed out.13 Problems of inad- be the critical action that prompts and magnetic resonance imaging if equate sample size, short duration, a modification in use or design of the indicated. Primary care physicians and comorbidity are exemplified by product, improves the understand- are better trained than cardiolo- the 2 RCTs on whose findings the ing of the safety profile of the drug gists to be on the lookout for cog- use of clopidogrel following the per- or device and leads to increased pa- nitive changes. cutaneous insertion of coronary ar- tient safety.”7 The FDA report form Multimorbidity gives reason to tery stents is based (Table 2). Both does not ask for patient identifica- question, as Barbara repeatedly did, studies were of short duration, either tion and cannot be used for mal- whether the emphasis on specialty 1 year15 or an average of 8 months16 practice litigation. care in the United States is mis- after stenting. The statistical signifi- Barbara lamented that adequate placed. “Specialty care for morbid- cance of both benefit and harm end care “is not realized when likely ad- ity that is not in the area of the phy- points is shown in Table 2. Only verse events are not systematically sician’s special competence,” Barbara with composite end points (eg, re- recorded and studied,”8(p65) which is wrote, “compromises quality of duced cardiovascular death and myo- often the case.9 She also noted that care.”13 Many of the studies on clopi- cardial infarction) is statistically sig- iatrogenic causes constitute the third dogrel following coronary stent nificant benefit attained. In one of leading cause of death in the United placement have been directed by car- these studies,15 a borderline associa- States.10 Oral antiplatelet agents are diologists and focus on the reduction of clopidogrel with harm (ma- the third most frequent category of tion of atherothrombotic out- jor bleed; P = .07) was observed. drugs implicated in hospitaliza- comes, which may be high in the Despite the short follow-up pe- tions for adverse events in patients short term, rather than on bleed- riods and the slim statistical signifi- aged 65 years or older.11 ing, which is spread throughout the cance, many cardiologists have pre- The FDA should inform the pub- duration of drug treatment. scribed clopidogrel for longer than lic that anyone can file the agency’s As our population ages and mor- 1 year after a stenting procedure, as adverse event report. Penalties (fi- bidities accumulate, the United in Barbara’s case. nancial and criminal) should be im- States needs policies that redirect Since these studies were re- posed when a drug company with- American medicine to primary ported, several others have been con- holds information from the FDA on care physicians. By providing con- ducted, of which 4 are shown in aggregated adverse events of drugs tinuing care over time, primary care Table 2. Two of these were RCTs.17,20 that it manufactures. Postmarket physicians can practice person- The other 2 were population-based surveillance of new drugs should be focused as opposed to disease- registry studies of patients being expanded. Barbara and I urged more centered care.8 They can get to know treated in practice, not in clinical stringent enforcement of the FDA’s their patients as persons and be- trials.18,19 Three of the 4 had more rules for postmarket surveillance.12 come aware of their multiple mor- patients than the first 2. None of bidities, alert to the presence or pos- these later studies found a signifi- MULTIMORBIDITY sibility of disease and/or drug cant benefit of clopidogrel admin- interactions. istered for a duration of longer than Barbara, who emphasized the im- 6 or 12 months following coronary portance of multimorbidity,13 was THE LIMITED SCOPE artery stenting. Two showed a sig- caught in its web: CAA, discovered OF CLINICAL TRIALS nificant increase in major bleeds post mortem, could have increased (P = .007 in Tsai et al18 and P Ͻ .001 her risk of cerebral hemorrhage from The FDA relies on randomized con- in Valgimigli et al20) and another a head trauma, perhaps by her swim- trolled trials (RCTs) to establish a borderline increase.19

ARCH INTERN MED/ VOL 172 (NO. 15), AUG 13/27, 2012 WWW.ARCHINTERNMED.COM 1175

©2012 American Medical Association. All rights reserved. Downloaded From: http://archinte.jamanetwork.com/ by a Monash University Library User on 08/21/2012 Table 2. Comparison of 6 Clopidogrel Studies After Coronary Artery Stenting

P Value No. of Mean Source Design Patients Duration, mo Benefit Harm Drug Company Sponsorship/Support CREDO15 RCT 1818 12a .04b .07 ϩ/ϩ PCI-CURE16 RCT 2658 8 (3-12) .047c; .03d NS ϩ/− Park et al17 RCT 2702 28-38 NS NS −/− Tsai et al18 Registry 7689 13-18 NS .007 −/− Sørensen et al19 Registry 11 680 18 vs 6e NS .06 −/− Valgimigli et al20 RCT 2013 24 vs 6e NS Ͻ.001 −/−

Abbreviations: CREDO, Clopidogrel for the Reduction of Events During Observation; NS, not significant; PCI-CURE, Percutaneous Coronary Intervention–Clopidogrel in Unstable Angina to Prevent Recurrent Events; RCT, randomized controlled trial; ϩ, received industry sponsorship/support; −, did not receive industry sponsorship/support. a From 29 days to 1 year. b Composite of death, myocardial infarction, and stroke. c Composite of cardiovascular death and myocardial infarction. d Composite of cardiovascular death, myocardial infarction, and any revascularization. e Months taking clopidogrel. manufacturer should be obliged to conduct postmarket surveillance. When a harmful effect is con- firmed, the FDA should issue a warning to physicians and require a warning in the package insert. Risks should be weighed against benefits to decide whether the drug should be removed from the market.

POTENTIAL BIAS IN RCTS SPONSORED AND SUPPORTED BY THE PHARMACEUTICAL INDUSTRY

Both of the early RCTs that reported a benefit of clopidogrel after stenting were sponsored by the drug’s original manufacturers.15,16 The title of one of these studies16 is mislead- ing, maintaining that the study was “long-term” although the average follow-up was only 8 months. In the Figure. A page from Barbara Starfield’s calendar diary written 6 weeks before her death. other of these RCTs, “medical spe- cialists employed by the sponsors Barbara reported minor bleed- gists. Neither the American Heart provided scientific input into the ing to her cardiologist following Association nor the FDA issued an study design,” and at least 2 investi- percutaneous stent insertion. Mi- alert on prolonged use of the drug. gators either received support from nor bleeding was observed in 3.5% The newsletter Worst Pills, Best Pills or consulted for the manufac- of patients taking clopidogrel in told readers that “long-term clopi- turer.15(p2419) Neither of the RCTs that one of the early RCTs, significantly dogrel may be no better than aspi- failed to show a benefit of long-term more than in the group receiving rin” and warned of its bleeding po- clopidogrel use,17,20 nor the reports only aspirin (P = .03).16 To my tential.21(p5) That newsletter, which using registry data with a similar knowledge, no study has examined should be read by more physicians, negative finding,18,19 received indus- whether patients experiencing mi- enables patients to question their try support. In a 2008 review,22(p109) nor bleeding who continue taking physicians about the safety of the pharmaceutical company sponsor- clopidogrel are more likely to sus- drugs they are taking. Barbara read ship of clinical trials was strongly as- tain a major bleed. the newsletter, but this issue was sociated with results “that favor the All but one of the later studies in published after she died. sponsors’ interests.” Table 2 were published before Bar- When a specific adverse event is Declaring that a conflict of inter- bara died. It is puzzling that their expected (eg, bleeding with clopi- est exists does not ensure that a study negative findings on clopidogrel did dogrel), adverse event reporting to is well designed, executed, and in- not reach many practicing cardiolo- FDA should be mandatory, and the terpreted. Before scientific journals

ARCH INTERN MED/ VOL 172 (NO. 15), AUG 13/27, 2012 WWW.ARCHINTERNMED.COM 1176

©2012 American Medical Association. All rights reserved. Downloaded From: http://archinte.jamanetwork.com/ by a Monash University Library User on 08/21/2012 consider an RCT for publication, ford even basic services in our ex- 11. Budnitz DS, Lovegrove MC, Shehab N, Richards they should require a declaration pensive system and suffer as a result. CL. Emergency hospitalizations for adverse drug events in older Americans. N Engl J Med. 2011; that “The funders have no role in 365(21):2002-2012. study design, study conduct, data Accepted for Publication: April 12, 12. Holtzman NA, Starfield B. Withdrawing FDA ap- management, and interpreta- 2012. proval of midodrine after marketing. JAMA. 2011; tion”23(p11) and do not provide di- Published Online: July 9, 2012. doi: 305(8):781-782. rect support to any researcher en- 10.1001/archinternmed.2012.2204 13. Starfield B. Threads and yarns: weaving the tap- estry of comorbidity. Ann Fam Med. 2006;4 gaged in the research. Correspondence: Neil A. Holtz- (2):101-103. man, MD, MPH, 2270 Eastridge Ave, 14. Campbell PG, Yadla S, Sen AN, Jallo J, Jabbour CONCLUDING COMMENTS Menlo Park, CA 94025 (nholtzma P. Emergency reversal of clopidogrel in the set- @jhsph.edu). ting of spontaneous intracerebral hemorrhage. Financial Disclosure: None re- World Neurosurg. 2011;76(1-2):100-104, dis- Could Barbara’s death have been cussion 59-60. ported. foretold? On a Saturday 6 weeks be- 15. Steinhubl SR, Berger PB, Mann JT III, et al; CREDO Additional Contributions: Iamin- Investigators. Clopidogrel for the Reduction of fore her death, Barbara wondered debted to Juan Gervas, MD, PhD, Events During Observation. Early and sustained whether she was having a “stroke Terri Haddix, MD, and Sidney dual oral antiplatelet therapy following percuta- equivalent” (Figure). The next day neous coronary intervention: a randomized con- Wolfe, MD, for critically reading she wrote “feeling better” and took trolled trial. JAMA. 2002;288(19):2411-2420. early drafts of this article. Andrew a strenuous walk. She did not re- 16. Mehta SR, Yusuf S, Peters RJ, et al; Clopidogrel Baker, MD, president of the Na- in Unstable angina to prevent Recurrent Events port this episode to her primary care tional Association of Medical Exam- trial (CURE) Investigators. Effects of pretreat- physician or cardiologist, neither of iners, kindly discussed with me how ment with clopidogrel and aspirin followed by long- whom was routinely available on term therapy in patients undergoing percutane- primary care physicians could be weekends. Had the symptoms of the ous coronary intervention: the PCI-CURE study. better informed of autopsy find- stroke equivalent continued Lancet. 2001;358(9281):527-533. ings after their patients died. Our 17. Park SJ, Park DW, Kim YH, et al. Duration of dual an- throughout the weekend, or had son, Steven C. Holtzman, sug- tiplatelet therapy after implantation of drug-eluting they occurred on another weekday, stents. NEnglJMed. 2010;362(15):1374-1382. gested changes in style and sub- Barbara might have consulted her 18. Tsai TT, Ho PM, Xu S, et al. Increased risk of bleed- stance that were in the best inter- primary care physician. And if her ing in patients on clopidogrel therapy after drug- ests of Barbara and me. eluting stents implantation: insights from the HMO primary care physician referred her Research Network-Stent Registry (HMORN-stent). for an magnetic resonance imaging, Circ Cardiovasc Interv. 2010;3(3):230-235. and if the magnetic resonance REFERENCES 19. Sørensen R, Abildstrom SZ, Weeke P, et al. Dura- imaging revealed microhemor- tion of clopidogrel treatment and risk of mortality rhage, she might have discontinued 1. Watts G. Obituary: Barbara Starfield. Lancet. 2011; and recurrent myocardial infarction among 11 680 378:564. patients with myocardial infarction treated with per- clopidogrel treatment in consulta- 2. Barbara Starfield memorial page. Johns Hopkins cutaneous coronary intervention: a cohort study. tion with her cardiologist. Whether Bloomberg School of Public Health, Department BMC Cardiovasc Disord. 2010;10:6. this would have prolonged Barba- of Health Policy and Management website. www 20. Valgimigli M, Campo G, Monti M, et al; Prolonging ra’s life, or for how long, is impos- .jhsph.edu/dept/hpm/starfield_memorial.html. Ac- Dual Antiplatelet Treatment After Grading Stent- cessed May 1, 2012. Induced Intimal Hyperplasia Study (PRODIGY) In- sible to say. 3. Biffi A, Greenberg SM. Cerebral amyloid angiopa- vestigators. Short- versus long-term duration of dual “When people pass away,” Mu- thy: a systematic review. J Clin Neurol. 2011; antiplatelet therapy after coronary stenting: a ran- rakami24 asks, “do their thoughts just 7(1):1-9. domized multicentre trial [published online ahead vanish?” Barbara’s thoughts as she 4. Greenberg SM, Eng JA, Ning M, Smith EE, Ro- of print March 21, 2012]. Circulation. 2012;125 transitioned from life to death will sand J. Hemorrhage burden predicts recurrent in- (16):2015-2026 doi:10.1161/CIRCULATIONAHA tracerebral hemorrhage after lobar hemorrhage. .111.071589. never be known, but her thoughts Stroke. 2004;35(6):1415-1420. 21. Anonymous. Inadvertent adverse reactions with while she lived will not just vanish. 5. Hoyert DL. The changing profile of autopsied commonly used drugs. Worst Pills, Best Pills They are already part of the think- deaths in the United States, 1972-2007. NCHS Data News. 2012;18:1-6. ing of many medical researchers and Brief. 2011;(67):1-8. 22. Sismondo S. Pharmaceutical company funding and 6. Starfield B, Shi L, Macinko J. Contribution of pri- its consequences: a qualitative systematic review. practitioners around the world and mary care to health systems and health. Milbank Contemp Clin Trials. 2008;29(2):109-113. have had a profound effect on health Q. 2005;83(3):457-502. 23. Collet JP, Cayla G, Cuisset T, et al. Randomized care policies in many countries, least 7. HIPAA compliance for reporters to FDA Med- comparison of platelet function monitoring to ad- of all her own. Watch. US Food and Drug Administration website. just antiplatelet therapy versus standard of care: Specialization, fragmentation, www.fda.gov/Safety/MedWatch/HowToReport rationale and design of the assessment with a /ucm085589.htm. Accessed May 1, 2012. double randomization of (1) a fixed dose versus drug-orientation, and profit- 8. Starfield B. Is patient-centered care the same as per- a monitoring-guided dose of aspirin and clopido- seeking help make American medi- son-focused care? Perm J. 2011;15(2):63-69. grel after DES implantation, and (2) treatment in- cal care the most expensive in the 9. Lopez-Gonzalez E, Herdeiro MT, Figueiras A. Deter- terruption versus continuation, 1 year after stent- world, but not the safest or most ef- minants of under-reporting of adverse drug reac- ing (ARCTIC) study. Am Heart J. 2011;161(1): tions: a systematic review. Drug Saf. 2009;32 5-12.e5. doi:10.1016/j.ahj.2010.09.029. fective. The lessons from Barbara’s (1):19-31. 24. Murakami H. What I Talk About When I Talk About death should be put in the perspec- 10. Starfield B. Is US health really the best in the world? Running. New York, NY: Vintage International; tive of the millions who cannot af- JAMA. 2000;284(4):483-485. 2009:76.

ARCH INTERN MED/ VOL 172 (NO. 15), AUG 13/27, 2012 WWW.ARCHINTERNMED.COM 1177