• 
• 

Three Part Question

In [adult patients undergoing cardiac surgery], does [Aprotinin use] decrease [risk of stroke and neurological dysfunction].

Clinical Scenario

You are about to perform CABG and aortic valve replacement surgery on a 75-year-old man with a history of diabetes, hypertension, carotid disease and transient ischaemic attacks. Your colleague suggests that you should give your patient aprotinin to minimise the risk of cerebrovascular complications but given recent controversies you decide to review the literature to investigate what evidence there is to suggest aprotinin has neuroprotective properties.

Search Strategy

Medline 1966 to May Week 4 2006 using OVID interface EMBASE 1980 to 2006 Week 21
Cardiopulmonary Bypass/OR CABG.mp. OR exp Thoracic Surgery/OR Coronary art$ bypass.mp. OR Cardiopulmonary bypass.mp. OR exp Cardiopulmonary Bypass/OR exp Cardiovascular Surgical Procedures/OR exp Thoracic Surgical Procedures/OR exp Coronary Artery Bypass/OR cardiac transplantation.mp. OR exp Heart Transplantation/] AND [Aprotinin.mp. OR exp APROTININ/OR Trasylol.mp. OR exp Aprotinin/OR Serine protease inhibitors.mp. OR exp Serine Proteinase Inhibitors/] AND [cerebrovascular disorders.mp. OR exp Cerebrovascular Disorders/OR brain ischemia.mp. OR Brain Ischemia/OR brain hypoxia.mp. OR Hypoxia, Brain/OR confusional state.mp. OR Confusion/OR postoperative complications.mp. OR Postoperative Complications/OR postoperative neurological complications.mp. OR coma.mp. OR exp Coma/OR encephalopathy.mp. OR cerebrovascular accident.mp. OR Cerebrovascular Accident/OR stroke.mp.].

Search Outcome

Using the reported search 224 papers were identified on Medline, 722 on Embase and 12 by hand-searching reference lists. Eight papers provided the best evidence to answer the question

Relevant Paper(s)

Author, date and country	Patient group	Study type (level of evidence)	Outcomes	Key results	Study Weaknesses
Murkin et al, 2002, Canada	Data from 1867 patients in 7 PRCTs comparing placebo to 'full dose' aprotinin Aprotinin 2x10 (to the power of 6) kIU Loading dose; 2x10 (to the power of 6) kIU in prime; 0.5x10 (to the power of 6) kIU/h by infusion	Meta-analysis (level 1a)	Incidence of postoperative stroke	9/935 (1.0%) Aprotinin vs. 24/932 (2.6%) had postoperative stroke; P<0.009	Abstract form only No information on diagnosis of stroke Includes data from 2 unpublished Bayer studies
Smith et al, 1996, USA	Data from 2283 patients in 6 studies Placebo vs. Pump prime Aprotinin (PPA) vs. Half-dose Aprotinin (HDA) vs. Full-dose Aprotinin (FDA) FDA: 2x10 (to the power of 6) kIU Loading dose; 2x10 (to the power of 6) kIU in prime; 0.5x10(to the power of 6) kIU/h by infusion HDA: 10 (to the power of 6) kIU Loading dose; 10 (to the power of 6) kIU in prime; 0.25x10 (to the power of 6) kIU/h by infusion PPA: 2x10 (to the power of 6) kIU in prime only	Meta-analysis (level 1a)	Mortality	23/861 (2.7%) Placebo vs. 7/245 (2.9%) PPA vs. 17/317 (5.4%) HDA vs. 24/860 (2.8%) FDA	Meta-analysis produced as editorial to accompany publication of paper by Lemmer et al (1996) Includes data from 2 unpublished Bayer studies Recognises insufficient power to identify differences in mortality and morbidity Also looked at rate of myocardial infarction Same 'published' as Murkin
			Incidence of stroke	Stroke incidence Placebo vs. PPA & Placebo vs. HDA P=ns Stroke incidence decreases in FDA group (1%) vs. Placebo group (2.4%); P=0.027
Sedrakyan et al, 2004, UK	Data from 3879 patients in 35 CABG studies between 1988 and 2001 identified using Medline, Embase and Pharmline databases Only randomised placebo controlled studies in which aprotinin given prophylactically & continuously intraoperatively included in meta-analysis (i.e. no pump prime only aprotonin patients) Personal communication with investigators of each study to confirm/clarify study details and gain additional information of interest not in original paper prior to inclusion in meta-analysis Stroke incidence data evaluated from 2976 patients in 18 trials (incl 2179 in 9 studies in which stroke data 'Investigator reported' only)	Meta-analysis (level 1a)	Evauation of stroke frequency based on clinical diagnosis of stroke, severe neurological deficit, cerebrovascular accident, cerebral embolism, cerebral haemorrhage, cerebral ischaemia	19/1714 (1.10%) Aprotinin vs. 28/1262. (2.22%) Placebo patients had stroke	Wide ranging definition of stroke Personal communication with investigators maximised data for meta-analysis although not always possible to confirm this additional data in original papers Does not appear to differentiate between full-dose' and 'half- dose' aprotinin data Does not include any pump-prime only' aprotinin data Meta-analysis also identified – decreased transfusion requirements (RR 0.61; 95% C.I. 0.58–0.66) no signif difference in increase/decrease incidence of mortality (RR 0.96; 95% C.I. 0.64–1.40); myocardial infarction (RR 0.85; 95% C.I. 0.63–1.14); or renal failure (RR 1.01; 95% C.I. 0.55–1.83); trend towards decrease in risk of AF (OR 0.90; 95% C.I.0.78–1.03)
			Relative risk calculation (RR)	Incidence of stroke decreased by 47%–RR 0.53; 95% C.I. 0.31–0.90
			Risk difference calculation (RD)	RD – 10 event decreased per 1000 patients treated (95% C.I. -20,0)
Levy et al, 1995	287 patients requiring repeat sternotomy randomised to high dose aprotinin (HDA), low-dose aprotinin (LDA), pump-prime only aprotinin (PPO) or placebo in 11 participating centres HDA–2x10 (to the power of 6) kIU bolus + 2x10 (to the power of 6) kIU in prime + 5x10 (to the power of 5) kIU/h during surgery LDA – 1x10 (to the power of 6) kIU bolus + 1x10 (to the power of 6) kIU in prime + 2.5x10 (to the power of 5) kIU/h during surgery PPO – 2x10 (to the power of 6) in pump-prime only Standardised anticoagulation blood conservation and transfusion protocol	PRCT (level 1b)	Stroke Incidence	0/61 HDA vs. 0/59 LDA vs. 1/68 (1.47%) PPO vs. 5/65 (7.69%) Placebo group had stroke (P=0.01)	Study specifically designed to look at bleeding and transfusion requirements and cost of aprotinin therapy rather than neurological morbidity Adverse neurological outcome not strictly defined Incidence of stroke in placebo group 7.69%) seems rather high in comparison to rate of similar studies
			Incidence of MI	No difference
			Incidence of renal dysfunction	Renal dysfunction in 19/215 aprotinin patients vs. 6/72 placebo
			Mortality	15/215 (7%) Aprotinin groups vs. 5/72 (7%) placebo group died
Dignan et al, 2001, Australia	202 patients undergoing first time CABG with LIMA use randomised to aprotinin (n=101) or placebo (n=99) Aprotinin group – 0.5 million kIU before incision; 0.5 million kIU during initiation of CPB Standard surgical, anticoagulation & transfusion protocols	PRCT (level 1b)	Neurological impairment	Temporary neurological dysfunction Aprotinin 1% Placebo 0%. Permanent neurological dysfunction – 1% vs. 1%	Study specifically designed to look at bleeding and transfusion requirements and cost of aprotinin therapy rather than neurological morbidity Inadequately powered to identify difference in gross neurological outcome Absolute numbers for adverse sequelae not quoted Adverse neurological outcome not strictly defined
			MI	Aprotinin 3% Placebo 5%
			Renal impairment	Aprotinin 2% Placebo 3%
			Cost benefit analysis of aprotin/blood product use aprotinin vs. Placebo	Unit cost per patient $463 vs. $343 Placebo vs. Aprotinin
Frumento et al, 2003, USA	1524 patients undergoing cardiac surgery with CPB in 15-month period 149 identified as being at high risk stroke (age 70 + history + increased BP + diabetes + previous stroke/TIA; + aortic atheroma) No aprotinin (Control) vs Full dose aprotinin (FDA - 2X10 to the power of 6 kIU pre-CPB + 2X10 to the power of 6 kIU in prime + 0.5X10 to the power of 6 kIU/h infusion) vs Half-dose aprotinin (HDA - 10 to the power of 6 kIU pre-CPB + 10 to the power of 6 kIU in prime + 0.25X10 to the power of 6 kIU/h infusion) 56 Control vs 67 HDA vs 26 FDA patients	Retrospective cohort study (level 1b)	Preoperative stroke risk index (McSPI score)	No signif difference in stroke risk index between groups	Retrospective involving 4 surgeons with inherent risk of bias Decision to give aprotinin, and dosage used not governed by protocol - clinician dependent Only assessed those patients who developed gross neurological dysfunction FDA group smallest
			Postoperative stroke (diagnosed clinically and confirmed by CT or MRI scanning)	24/149 (16%) patients experienced postoperative stroke overall - 9/56 (16%) Control vs 15/93 (16%) all-aprotinin group (p=ns) 9/56 (16%) Control vs 15/67 (22%) HDA vs 0/26 (0%) FDA had postop stroke (P=0.03 FDA vs HDA or control; 95% CI 0-11.5%)
Harman et al, 2004, Ireland	36 ASA III-IV patients undergoing elective CABG surgery, standard anaesthetic/surgical/perfusion management Aprotinin (2X10 to the power of 6 kIU loading dose + 2X10 to the power of 6 kIU in prime + 0.5X10 to the power of 6 kIU/h) vs placebo Cognitive tests performed on patients pre-op, 4 days and 6 weeks postop (and on 18 spouses as control group)	Single-blinded PRCT (level 1b)	Neuropsychological assessment	No patient experienced overt stroke	Pilot study Small numbers One death in aprotinin group 3 days post-op - excluded from analysis
			Mood assessment - hospital depression and anxiety score Delirium assessment - DSM-III-R criteria + minimental score	Surgical parameters comparable (except for signif lower intraop and postop blood loss in aprotinin group Preop neuropsychological and anxiety/depression scores comparable between aprotinin and placebo groups Preop depression scores higher in placebo vs control group P<0.05 (aprotinin vs control P=ns)
			Cognitive function assessment - Rey Auditory Verbal Language Test RAVLT (verbal memory) - Trail Marking Test TMT A&B (attention and mental flexibility) - Purdue Pegboard test (motor speed, coordination and manual dexterity)- Controlled Oral Word Association Test - COWAT (word fluency) - Digit Symbol Test - Dig Symb (psychomotor speed, attention and concentration)	10/17 (58%) aprotinin vs 17/18 (94%) placebo patients displayed cognitive dysfunction 4 days postop (95% CI 0.10-0.62;P=0.01) 4/17 (23%) aprotinin vs 10/18 (55%) placebo patients displayed cognitive dysfunction 6 weeks postop (95% CI 00.80-0.16;P=0.05) Anxiety and depression scores at 4 days and 6 weeks comparable between aprotinin and placebo groups
Mangano et al, 2006, UK	Cohort of 4374 patients undergoing CABG surgery systematically sampled from 69 centres in N and S America, Europe, Asia and Middle East Effects of drug interventions assessed using multivariate logistic regression and propensity score adjustment to minimise selection bias 3013 primary (elective CABG or angioplasty) and 1361 complex procedures Adverse events pre-specified and defined by protocol Aprotinin (Apro): -796 primary, 499 complex = 1295 total Aminocaproic acid (ACA): -597 primary, 286 complex = 883 total Tranexamic acid (TXA): -598 primary, 224 complex = 822 total No anti-fibrinolytic (Cont): -1022 primary, 352 complex = 1374 total	Prospective cohort study (level 1b)	Incidence of neurological complication: - cerebrovascular event	Overall incidence of neurological complication Aprotinin vs Aminocaproic acid vs Tranexamic acid vs control - CVA event - 6% vs 3% vs 4% vs 3% (P<0.001)	Study designed to look at risk of various adverse outcomes (all comparisons Aprotinin vs Aminocaproic acid vs Tranexamic acid vs control) Increased risk of renal composite outcome event (8 vs 3 vs 4 vs 3% P<0.001). Increased risk of renal dysfunction (5 vs 3 vs 3 vs 2%; P<0.001) Increased risk of renal dialysis (5 vs 1 vs 1 vs 1%; P<0.001) Increased risk of non-renal composite outcome event (29 vs 20 vs 21 vs 19%;P<0.001) Increased risk of vascular event (25 vs 17 vs 19 vs 17%; P<0.001) Increased risk of cardiovascular event (22 vs 16 vs 17 vs 16%;P<0.001) No study control of intervention - choice dictated by contributing centre - subject to bias Systematic sampling - more samples from bigger units - ?risk of bias and institutional differences Several relevant risk factors significantly higher in aprotinin group than control group - hypertension, insulin-dependant diabetes, carotid disease, valve and valve+graft surgery (all P<0.001). Similar trends in ACA but not TXA group Raw data presented as cerbrovac disease, stroke or encephalopathy; propensity adjusted as cerebrovac disase which included stroke, encephaopathy and coma Use of any anit-fibrinolytic associated with decreased blood loss vs control
			Stroke	3% vs 1% vs 1% vs 2% (P<0.001)
			Encephalopathy	4% vs 2% vs 3% vs 2% (P<0.001) Incidence of cerebrovascular events in primary surgery only Aprotinin vs Aminocaproic or Tranexamic acid 4.5 vs 1.6% (P<0.001) Risk of adverse cerebrovasc event (stroke, coma or encephalopathy) signif higher in Aprotinin vs control in primary surgery - OR 2.15; 95% CI 1.14-4.06; P=0.02 (multivariate logistic regression) and OR 1.19; 95% CI 1.08-1.30; P<0.001 (propensity score). No signif difference in complex surgery

Comment(s)

In the past 10 years, three meta-analyses have reported decreased postoperative stroke incidence in patients receiving perioperative aprotinin by bolus and infusion (Murkin, Smith, Sedrakyan). Those by Murkin et al. and Smith et al. analyse essentially the same studies (Cosgrove, Lemmer 1994, 1996, Levy, Alderman, Bidstrup) and include unpublished data from the manufacturers Bayer. However, whilst Murkin et al. only analysed studies using the so-called 'Hammer-smith' or 'full-dose' aprotinin regime, Smith et al. also studied data from 'half-dose' and 'pump-prime only' studies. Whilst both showed a lower incidence of stroke in aprotinin-treated patients, Smith's group demonstrated that this only occurred with 'full-dose' techniques and raised safety concerns about use of 'half-dose' and 'prime only' techniques. The most recent meta-analysis by Sedrakyan et al. assessed 2976 patients in 18 trials. However, this included 2179 patients from 9 trials in which additional stroke/neurological dysfunction data were obtained from personal communication with the primary investigator and so cannot always be verified in the original paper (Cosgrove, Lemmer 1994, 1996, Alderman, Bidstrup, Harig, Kalangos, Klein, Wahba). The remaining patients came from 9 trials in which 'neurological outcomes' were reported (Levy, Asimakopoulos, Bidstrup, Blauhut). Unfortunately most are small studies underpowered to identify significant differences in stroke incidence with only two of the latter group having study populations of over 50 in each limb. Nevertheless, Levy et al. did find a significantly lower incidence of stroke in patients receiving 'full-dose' or 'half-dose' aprotinin compared to placebo although the stroke incidence in their placebo group seems extraordinarily high at 7.69% compared to similar studies. In contrast, Dignan et al. noted comparable incidences of temporary and permanent neurological dysfunction in 101 patients receiving a non-standard ultra-low dose of aprotinin (Dignan). Even larger 'Investigator Reported' studies by Lemmer et al. and Alderman et al. with 704 and 850 patients, respectively, failed to show significant differences in stroke incidence between aprotinin and placebo groups with stroke incidences of 0.85% vs.1.12% (Lemmer et al. 1996) and 1.15% vs. 1.84% (Alderman et al.). In addition, Lemmer's 1996 group could find no difference in stroke incidence between 'full-dose', 'half-dose' and 'pump-prime only' regimes. However, in their meta-analysis, Sedrakayan et al. concluded that the use of intraoperative prophylactic aprotinin decreased the incidence of postoperative stroke by 47% with a 10-event risk difference per 1000 patients treated although they did not differentiate between 'full-dose' and 'half-dose' data. They also confirmed the decreased transfusion requirements seen in many other aprotinin studies with no excess of myocardial infarction, renal failure or mortality. Subsequent more recent studies support aprotinin's neuroprotective actions. Frumento et al. recognised a lower stroke risk with 'full-dose' rather than 'half-dose' aprotinin in a patient group at high risk of stroke (Frumento) although their cohort was not randomised and may contain inherent surgical bias. Harmon et al. demonstrated lower incidence of cognitive dysfunction 4 days and 6 weeks postoperatively in a small group of CABG patients given high dose aprotinin (Harmon). In contrast, Mangano et al. reported significantly increased adverse outcomes in 1295 patients who received aprotinin within a cohort of 4374 patients undergoing 'primary' (CABG or angioplasty only) or 'complex' (all other) surgery. Using logistic regression analysis and propensity scoring techniques they reported that the risk of stroke was increased by 181% and the risk of MI by 55% in 'primary' surgery, and the incidence of renal failure doubled in both 'primary' and 'complex' surgery. They also noted dose-response aprotinin effects and commented that as other anti-fibrinolytic such as tranexamic acid and aminocaproic acid had the same blood-sparing effects without adverse effects, continued use of aprotinin was 'not prudent'. However, whilst this study has several weaknesses, including a risk of bias from systemic sampling across multiple institutions with inherently embedded practices, and higher risk factors for some adverse outcomes within the aprotinin group, it has resulted in considerable debate and may lead to some reappraisal of aprotinin's use particularly in uncomplicated 'primary' surgery.

Clinical Bottom Line

There is evidence from three meta-analyses and two more recent RCTs that the use of aprotinin is associated with decreased incidence of stroke and neurological complications in some patients undergoing cardiac surgery. However many single studies within these meta-analyses are small, designed to look at other outcomes, and underpowered to assess neurological differences. In contrast, a recent cohort study has suggested that aprotinin use is associated with a significant increased risk of stroke in uncomplicated CABG surgery.

References

1. Murkin JM, Maurer J, Niemcryk S. Full dose aprotinin is associated with a significant decrease in perioperative stroke in patients undergoing elective cardiac surgery. A meta-analysis. Ann Thorac Surg 2002; 73:S374.
2. Smith PK, Muhlbaier LH. Aprotinin: safe and effective only with the full-dose regimen. Ann Thorac Surg 1996; 62:1575–1577.
3. Sedrakyan A, Treasure T, Elefteriades JA. Effect of aprotinin on clinical outcomes in coronary artery bypass graft surgery: a systematic review and meta-analysis of randomized controlled trials. J Thorac Cardiovasc Surg 2004; 128:442–448.
4. Cosgrove DM 3rd, Heric B, Lytle BW, Taylor PC, Novoa R, Golding LA, Stewart RW, McCarthy PM, Loop FD. Aprotinin therapy for reoperative myocardial revascularization: a placebo-controlled study. Ann Thorac Surg 1992; 54:1031–1036.
5. Lemmer JH Jr, Stanford W, Bonney SL, Breen JF, Chomka EV, Eldredge WJ, Holt WW, Karp RB, Laub GW, Lipton MJ, et al. Aprotinin for coronary bypass operations: efficacy, safety, and influence on early saphenous vein graft patency. A multicenter, randomized, double-blind, placebo-controlled study. J Thorac Cardiovasc Surg 1994; 107:543–551.
6. Levy JH, Pifarre R, Schaff HV, Horrow JC, Albus R, Spiess B, Rosengart TK, Murray J, Clark RE, Smith PA. Multicenter, double-blind, placebo-controlled trial of aprotinin for reducing blood loss and the requirement for donor-blood transfusion in patients undergoing repeat coronary artery bypass grafting. Circulation 1995; 92:2236–2244.
7. Lemmer JH Jr, Dilling EW, Morton JR, Rich JB, Robicsek F, Bricker DL, Hantler CB, Copeland JG 3rd, Ochsner JL, Daily PO, Whitten CW, Noon GP, Maddi R. Aprotinin for primary coronary artery bypass grafting: a multicenter trial of three dose regimens. Ann Thorac Surg 1996; 62:1659–1667.
8. Alderman EL, Levy JH, Rich JB, Nili M, Vidne B, Schaff H, Uretzky G, Pettersson G, Thiis JJ, Hantler CB, Chaitman B, Nadel A. Analyses of coronary graft patency after aprotinin use: results from the International Multicenter Aprotinin Graft Patency Experience (IMAGE) trial. J Thorac Cardiovasc Surg 1998; 116:716–730.
9. Bidstrup BP, Royston D, Sapsford RN, Taylor KM. Reduction in blood loss and blood use after cardiopulmonary bypass with high dose aprotinin (Trasylol). J Thorac Cardiovasc Surg 1989; 97:364–372.
10. Harig F, Feyrer R, Mahmoud FO, Blum U, von der Emde J. Reducing the post-pump syndrome by using heparin-coated circuits, steroids, or aprotinin. Thorac Cardiovasc Surg 1999; 47:111–118.
11. Kalangos A, Tayyareci G, Pretre R, Di Dio P, Sezerman O. Influence of aprotinin on early graft thrombosis in patients undergoing myocardial revascularization. Eur J Cardio-Thorac 1994; 8:651–656.
12. Klein M, Keith PR, Dauben HP, Schulte HD, Beckmann H, Mayer G, Elert O, Gams E. Aprotinin counterbalances an increased risk of peri-operative hemorrhage in CABG patients pre-treated with aspirin. Eur J Cardio-Thorac 1998; 14:360–366.
13. Wahba A, Black G, Koksch M, Rothe G, Preuner J, Schmitz G, Birnbaum DE. Aprotinin has no effect on platelet activation and adhesion during cardiopulmonary bypass. Thromb Haemost 1996; 75:844–848.
14. Asimakopoulos G, Kohn A, Stefanou DC, Haskard DO, Landis RC, Taylor KM. Leukocyte integrin expression in patients undergoing cardiopulmonary bypass. Ann Thorac Surg 2000; 69:1192–1197.
15. Bidstrup BP, Underwood SR, Sapsford RN, Streets EM. Effect of aprotinin (Trasylol) on aorta-coronary bypass graft patency. J Thorac Cardiovasc Surg 1993; 105:147–152.
16. Bidstrup BP, Hunt BJ, Sheikh S, Parratt RN, Bidstrup JM, Sapsford RN. Amelioration of the bleeding tendency of preoperative aspirin after aortocoronary bypass grafting. Ann Thorac Surg 2000; 69:541–547.
17. Blauhut B, Harringer W, Bettelheim P, Doran JE, Spath P, Lundsgaard-Hansen P. Comparison of the effects of aprotinin and tranexamic acid on blood loss and related variables after cardiopulmonary bypass. J Thorac Cardiovasc Surg 1994; 108:1083–1091.
18. Dignan RJ, Law DW, Seah PW, Manganas CW, Newman DC, Grant PW, Wolfenden HD. Ultra-low dose aprotinin decreases transfusion requirements and is cost effective in coronary operations. Ann Thorac Surg 2001; 71:158–163.
19. Moran SV, Lema G, Medel J, Irarrazaval MJ, Zalaquett R, Garayar B, Flaskamp R. Comparison of two doses of aprotinin in patients receiving aspirin before coronary bypass surgery. Perfusion 2000; 15:105–110.
20. Murkin JM, Lux J, Shannon NA, Guiraudon GM, Menkis AH, McKenzie FN, Novick RJ. Aprotinin significantly decreases bleeding and transfusion requirements in patients receiving aspirin and undergoing cardiac operations. J Thorac Cardiovasc Surg 1994; 107:554–561.
21. Santamaria A, Mateo J, Oliver A, Litvan H, Murillo J, Souto JC, Fontcuberta J. The effect of two different doses of aprotinin on haemostasis in cardiopulmonary bypass surgery: similar transfusion requirements and blood loss. Haematologica 2000; 85:1277–1284.
22. Frumento RJ, O'Malley CM, Bennett-Guerrero E. Stroke after cardiac surgery: a retrospective analysis of the effect of aprotinin dosing regimens. Ann Thorac Surg 2003; 75:479–483.
23. Harmon DC, Ghori KG, Eustace NP, O'Callaghan SJ, O'Donnell AP, Shorten GD. Aprotinin decreases the incidence of cognitive deficit following CABG and cardiopulmonary bypass: a pilot randomized controlled study. Can J Anaesth 2004; 51:1002–1009.
24. Mangano DT, Tudor IC, Dietzel C. The risk associated with aprotinin in cardiac surgery. New Engl J Med 2006; 354:353–365.