icu6

Editorials

Towards changing the definition of the acute respiratory distresssyndrome: One step forward*

Having a common and reli-able definition for the acuterespiratory distress syn-drome (ARDS) or acute lung

injury is very important for the manage-ment of this severe respiratory complica-tion. At a full stage, this syndrome corre-sponds to diffuse alveolar damage, butlung pathology is usually not availableexcept at autopsy (1). There is no vali-dated biomarker, even in bronchoalveolarlavage fluid, which could improve ourability for an early recognition of in-creased lung permeability. An accuratedefinition is required for clinical trials orepidemiologic studies, with sufficientspecificity to reduce the noise-to-signalratio and good sensitivity to avoid miss-ing patients. An accurate definition canalso greatly help clinicians to definewhich patients may benefit from preciseventilatory strategies, diagnostic proce-dures, or drug therapy. Our current def-inition, coming from a consensus (2),has been extremely useful, especiallyfor clinical trials. It combines simpleparameters that are available at thebedside and a clinical assessment of thepatient. The cornerstone of this defini-tion is the PaO2/FIO2 ratio. This ratiohas become the most popular way toquantify oxygenation defects.

Several factors greatly influence thevalue of the PaO2/FIO2 ratio, indepen-dently from lung pathology, but are oftenignored at the bedside. Mixed venous ox-ygenation is a crucial one and is influ-enced by cardiac output. For a given lungstatus, varying cardiac output and arteri-al-to-venous difference in oxygen contentheavily influences the PaO2/FIO2 ratio (3).

For this reason, it was advocated in the1980s to calculate shunt using the Berg-gren equation instead of relying on PaO2

and FIO2 alone (4). The decline in the useof the pulmonary artery catheter, neces-sary to get mixed venous blood, is one ofthe reasons why this approach is not usedanymore. A second determinant is thepressure delivered by the ventilator. Forthis reason, pediatricians have intro-duced mean airway pressure in their as-sessment of oxygenation and calculationof the oxygenation index. Although thisapproach makes good sense, a clear clin-ical advantage over the PaO2/FIO2 ratioalone has not been clearly demonstrated.Along the same line, several investigatorshave argued that the level of positive end-expiratory pressure (PEEP), a major de-terminant of oxygenation, needed to betaken into account, either by introducingthe PEEP level into an oxygenation equ-ation or by imposing a fixed level of PEEPto make calculations (5, 6). It is possible,however, that the vast majority of pa-tients in whom the PaO2/FIO2 ratio is cal-culated are already on a “moderate,” av-erage level of PEEP. How frequentlyadding a requirement for a given PEEPlevel would change the definition has notbeen clearly evaluated. This was the firstquestion addressed in the work done byBritos and colleagues (7) in this issue ofCritical Care Medicine, who explored theinfluence of the initial PEEP level onmortality across different ranges of PaO2/FIO2. They used the data collected duringconduct of the ARDS Network trials,when patients were enrolled in clinicaltrials, making a total of 2,312 patientswith acute lung injury/ARDS available.They tested if, for a given range of PaO2/FIO2 values, mortality was different ac-cording to the PEEP level, which wouldsuggest that patients with very differentdisease severity were enrolled under thesame oxygenation criterion. It is remark-able, however, to see in the results ofBritos et al, how consistent the mortalityis between a PEEP at 5 cm H2O or belowand a PEEP at 11 cm H2O or above. This

suggests that adding PEEP to the defini-tion would not help in better character-izing the disease severity of the patients.Several reasons may explain this. First,when clinicians increase PEEP they prob-ably also often increase FIO2, which byitself decreases the PaO2/FIO2 ratio. Sec-ond, clinicians may vary the PEEP set-tings over a relatively narrow range. In-deed, in the ARDS Network hospitals,very few patients (1.3%) were ventilatedwith PEEP levels below 5 cm H2O. At theopposite, only 11% of patients had PEEPat 15 cm H2O or higher before enroll-ment in the trials. Therefore, their resultsmay not apply to ICUs where the PEEPlevel is frequently under 5 cm H2O orabove 14 cm H2O to calculate the initialPaO2/FIO2 ratio.

A last issue is the effect of the FIO2

level on the PaO2/FIO2 ratio itself. This isoften ignored because it sounds as coun-terintuitive. Clinicians may think thatnormalizing PaO2 to FIO2 is possible be-cause a linear relationship exists betweenthe two. This is not the case, however,and several experimental or clinical stud-ies have clearly shown that there is anonlinear relationship between PaO2 andFIO2, and consequently between PaO2 andPaO2/FIO2 (8–10). The shape of the oxyhe-moglobin dissociation curve describingthe relationship of the percentage of he-moglobin saturation to the blood PO2 andthe equation for human blood oxygen dis-sociation explains this relationship (11).This relationship is difficult to predictbecause it depends on the cardiac output,the arterial-to-venous difference in oxy-gen content, and the shunt fraction (8).In patients with a relatively high shunt,the curve will often display a U shape,with an increase in the PaO2/FIO2 ratiowhen FIO2 is raised from around 50% or60% up to 100%. This means that a givenPaO2/FIO2 value measured in a patient atFIO2 1.0 will reflect a more severe impair-ment of oxygenation than in another pa-tient with the same PaO2/FIO2 ratio at aFIO2 of only 50%. It also means that for agiven patient, changing FIO2 modifies the

*See also p. 2025.Key Words: acute lung injury; acute respiratory

distress syndrome; oxygenation; positive end-expiratory pressure; ventilation

The author has not disclosed any potential con-flicts of interest.

Copyright © 2011 by the Society of Critical CareMedicine and Lippincott Williams & Wilkins

DOI: 10.1097/CCM.0b013e318226607a

2177Crit Care Med 2011 Vol. 39, No. 9

PaO2/FIO2 value. As an illustration, a studyperformed in 14 patients with acute lunginjury or ARDS showed that decreasingFIO2 from 1.0 to 0.6 over 15 mins led to areduction of the PaO2/FIO2 ratio from amedian of 223 mm Hg to a median of 158mm Hg at the same high PEEP level; thesame change was reproduced from 171 to115 at a lower PEEP level (12). Therefore,taking into account the FIO2 in the defini-tion may help to better identify the diseaseseverity of the patients. This is what wasfound in a second part of the analysis byBritos et al (7). Contrasting with the lack ofinfluence of PEEP, they found that themortality rate significantly increased ateach range of PaO2/FIO2 values for FIO2 vary-ing from 0.50 or less to 0.70 or more.

These results are important becausethey confirm physiologic observations ona large scale and may need to be takeninto account for a revision of the currentdefinition of ARDS. Although the PEEPlevel was not found to influence the re-sults, calculating the PaO2/FIO2 ratio for aPEEP level of at least 5 cm H2O and nomore than 15 cm H2O may be reasonable.A big step forward can be to introduce the

level of FIO2 in the definition of ARDS,which could help to better stratify thedisease severity of patients.

Laurent J. Brochard, MDDepartment of Intensive CareGeneva University HospitalUniversity of GenevaGeneva, Switzerland

REFERENCES

1. Esteban A, Fernandez-Segoviano P, Frutos-VivarF, et al: Comparison of clinical criteria for theacute respiratory distress syndrome with autopsyfindings. Ann Intern Med 2004; 141:440–445

2. Bernard GR, Artigas A, Brigham KL, et al:The American-European Consensus Confer-ence on ARDS. Definitions, mechanisms, rel-evant outcomes, and clinical trial coordina-tion. Am J Respir Crit Care Med 1994; 149:818–824

3. Lemaire F, Teisseire B, Harf A: [Assessmentof acute respiratory failure: Shunt versus al-veolar arterial oxygen difference]. Ann FrAnesth Reanim 1982; 1:59–64

4. Berggren S: The oxygen deficit of arterialblood caused by non ventilating parts of thelung. Acta Physiol Scand 1942; 11:1–92

5. Villar J, Perez-Mendez L, Lopez J, et al: Anearly PEEP/FIO2 trial identifies different de-grees of lung injury in patients with acute

respiratory distress syndrome. Am J RespirCrit Care Med 2007; 176:795–804

6. Villar J, Perez-Mendez L, Kacmarek RM: Cur-rent definitions of acute lung injury and theacute respiratory distress syndrome do notreflect their true severity and outcome. In-tensive Care Med 1999; 25:930–935

7. Britos M, Smoot E, Liu KD, et al: The valueof positive end-expiratory pressure and FiO2

criteria in the definition of the acute respi-ratory distress syndrome. Crit Care Med2011; 39:3025–3030

8. Aboab J, Louis B, Jonson B, et al: Relationbetween PaO2/FIO2 ratio and FIO2: A math-ematical description. Intensive Care Med2006; 32:1494–1497

9. Gowda MS, Klocke RA: Variability of indicesof hypoxemia in adult respiratory distresssyndrome. Crit Care Med 1997; 25:41–45

10. Whiteley JP, Gavaghan DJ, Hahn CE: Varia-tion of venous admixture, SF6 shunt, PaO2,and the PaO2/FIO2 ratio with FIO2. Br JAnaesth 2002; 88:771–778

11. Severinghaus JW: Simple, accurate equa-tions for human blood O2 dissociation com-putations. J Appl Physiol 1979; 46:599–602

12. Aboab J, Jonson B, Kouatchet A, et al: Effectof inspired oxygen fraction on alveolar dere-cruitment in acute respiratory distress syn-drome. Intensive Care Med 2006; 32:1979–1986

Burn fluid resuscitation: Let the autopilot do it!*

Before the recognition of themagnitude of fluid shifts andthe massive fluid require-ments of severe burn patients,

hypovolemia, hypoperfusion, and organfailure was the leading cause of death.Investigations in the 1960s led to thewidely accepted Parkland formula (1).Despite controversy and debate overamounts, algorithms, and fluid mixes,the Parkland formula is still the mostwidely used (2). From 1970 to 2000, burnfluid resuscitation science was dominatedby the fear of hypovolemia, hypoperfu-sion, shock, and organ damage. Fluidshave been generously administered, far

beyond the original recommendations. In2000, the Seattle team showed that 58%of the patients received more fluid thanpredicted (3). A similar observationalstudy from Toronto (4) showed that pa-tients received on average 6.7 mL/kg/%burn surface area (instead of the original 4mL/kg/%), and 84% of the patients receivedmore than predicted compared to 12% inthe original study (1). Sadly, too muchfluid, a phenomenon called fluid creep, canlead to “Michelin Man” patients and organfailure (5, 6). Awareness of the side effectsof fluid over-resuscitation has increasedover the last decade. The most significanteffect of fluid over-resuscitation in burnpatients is edema of the stasis area sur-rounding the deeper burned coagulationarea. This leads to necrosis of just viabletissue yielding to a secondary deepeningand extension of the burn area. Abdominalcompartment syndrome, frequent in over-resuscitation trauma (7), is a classic prob-lem in massive burns (8). In burn patients,decreased urine output is often used to

dictate more fluids, but may also reflectover-resuscitation and the onset of abdom-inal compartment syndrome (9). Despiteawareness of the problem (10), the pendu-lum is still above Baxter’s original recom-mendations (11).

The fine balance between too little ortoo much fluid is hard to maintain andrequires clinicians with extensive burnexperience. In this context, the paper inthis issue of Critical Care Medicine bySalinas and colleagues (12) is very wel-come. They showed that using an algo-rithm programmed in Java they were ableto suggest the right amount of fluid tomaintain an adequate urine output with-out polyuria while limiting hypoperfu-sion and fluid creep in burn patients. Thesame investigators previously published afull closed-loop system in experimentalsheep burn (13). The present paper is aninteresting application, but is not yet a fullclosed-loop system. The system gives theclinician advice, which can be followed orignored (Fig. 1).

*See also p. 2031.Key Words: burn; clinical decision support sys-

tems; critical care automation; fluid creep; fluid resus-citation; open-loop system



DOI: 10.1097/CCM.0b013e31821f030d

2178 Crit Care Med 2011 Vol. 39, No. 9

The first autopilot in aviation was in-vented by Lawrence Sperry one centuryago (1912). Based on a gyroscope and analtimeter, the system was able to main-

tain a constant heading and altitude. Thesystem was connected hydraulically tothe elevators and rudder. With the evolu-tion of technology and computer science,

systems are now able to control all planesystems, including engine thrust, andperform a complete flight plan. Typicallyseveral hundred thousand lines of com-puter code are required to fly actualplanes. Flight management systems havebeen shown to provide a smoother ridewith lower fuel consumption comparedto manual flight. In a parallel manner,Salinas et al (12) showed that patientsmanaged following the advice of the de-cision support system performed better.They had a smoother ride, as shown byless wobbling in terms of both fluid ad-ministration and urine output, and theyreduced the fuel consumption, i.e., thetotal fluid needed for the same - or evenbetter - result. There are two componentsin the approach. First there is an algo-rithm, a set of rules, which define a re-sponse (fluid administration) to an input(urine production). The second is embed-ded in the software. Strategy makes thedifference in outcome, as shown in septicshock or burn resuscitation (14), and soft-ware makes it easy to apply. An algorithmicapproach to burn fluid resuscitation wasproposed 20 yrs ago by Miller et al (15). Theconcept of closed loop in medicine is notnew. A Society of Critical Care Medicinetask force pioneered the assessment ofclosed-loop technologies 16 yrs ago (16).

In aviation, the development of flightmanagement systems was welcomed bypilots, having more time and mind tofocus on the navigation and safety ratherthan the flying process. Interestingly, al-though the clinical decision system pre-sented here gives “sound” fluid resuscita-tion advice, the advice was frequently notapplied. Interestingly, when clinicians fol-lowed the advice of the decision system,they were more often in the target in termsof urine output, than when �100 mL/hr offthe computer decision support system. De-cision systems often act more quickly thanthe clinician and avoid overreacting whenlate. For example, closed-loop ventilationsystems adapt each breath based on theprevious one (17), a process that can onlybe done by a computer.

Closed-loop systems are not in theirinfancy in critical care. Several closed-loop systems managing heparin duringhemodialysis (18) or lidocaine infusion tocontrol premature ventricular contrac-tions (19) have been developed but didnot reach widespread clinical use. Clini-cians aim at personally controlling all pa-rameters of patient care. With the develop-ment of critical care and scarcity ofresources, this will not be possible. Avail-

Figure 1. Various concepts in systems. A, An open-loop system. The computer analyzes the input andmakes suggestions to the clinician. The system presented by Salinas et al in the present issue bears thisconcept. B, In closed-loop systems, the computer directly adapts the pump rates to achieve a desiredinput. In integrated systems (C) the computer combines several forms of information to program thepump rates or other types of information. In neural network type systems (D), the system adaptsthrough several hidden layer “neurons” the output from various inputs. The system “learns” fromprevious situations. BP, blood pressure; HR, heart rate; BIS, bispectral index scale.


able systems can not only manage fluidresuscitation, but also glucose level (20),anesthesia (21), or ventilation weaning(17). Future systems will be integrated andable to adapt dynamically to the changingsituation. A significant part of today’s clini-cians are reluctant to give away the controlstick to a computer. Clinicians should notreject these system (or their advice), hidingbehind the art of critical care or burn fluidmanagement. These decision support sys-tems are here to do your basic job. The newrole for clinicians will be to focus on clini-cal strategy while overlooking the systems,taking over in case of unplanned situations.Similarly, autopilots and flight manage-ment systems have not and will not replacepilots in airplanes; autopilots will never beable to land on the Hudson River with bothengines out (22). They are our assistants,our helpers, and can extend the burn in-tensive care unit expertise outside the burnunit walls to remote hospitals, austere en-vironments, transport situations, disastersituations, and less wealthy countries.

David Bracco, MD, PhD, FCCMAssociate Professor of

AnesthesiaCritical Care and TraumaMcGill UniversityMontreal General Hospital,

MontrealQuebec, Canada

REFERENCES

1. Baxter CR, Shires T: Physiological responseto crystalloid resuscitation of severe burns.Ann N Y Acad Sci 1968; 150:874–894

2. Greenhalgh DG: Burn resuscitation: The resultsof the ISBI/ABA survey. Burns 2010; 36:176–182

3. Engrav LH, Colescott PL, Kemalyan N, et al:A biopsy of the use of the Baxter formula toresuscitate burns or do we do it like Charliedid it? J Burn Care Rehabil 2000; 21:91–95

4. Cartotto RC, Innes M, Musgrave MA, et al:How well does the Parkland formula estimateactual fluid resuscitation volumes? J BurnCare Rehabil 2002; 23:258–265

5. Saffle JI: The phenomenon of “fluid creep” inacute burn resuscitation. J Burn Care Res2007; 28:382–395

6. Chung KK, Wolf SE, Cancio LC, et al: Resus-citation of severely burned military casualties:Fluid begets more fluid. J Trauma 2009; 67:231–237

7. Balogh Z, McKinley BA, Cocanour CS, et al:Supranormal trauma resuscitation causesmore cases of abdominal compartment syn-drome. Arch Surg 2003; 138:637–642; dis-cussion 642–643

8. Burke BA, Latenser BA: Defining intra-abdominal hypertension and abdominalcompartment syndrome in acute thermal in-jury: A multicenter survey. J Burn Care Res2008; 29:580–584

9. Azzopardi EA, McWilliams B, Iyer S, et al:Fluid resuscitation in adults with severeburns at risk of secondary abdominal com-partment syndrome–an evidence based sys-tematic review. Burns 2009; 35:911–920

10. Pruitt BA Jr: Protection from excessive re-suscitation: “Pushing the pendulum back.”J Trauma 2000; 49:567–568

11. Cartotto R, Zhou A: Fluid creep: The pendu-lum hasn’t swung back yet! J Burn Care Res2010; 31:551–558

12. Salinas J, Chung KK, Mann EA, et al: Comput-erized decision support system improves fluidresuscitation following severe burns: An origi-nal study. Crit Care Med 2011; 39:2031–2038

13. Salinas J, Drew G, Gallagher J, et al: Closed-loop and decision-assist resuscitation of burnpatients. J Trauma 2008; 64:S321–S332

14. Chung KK, Blackbourne LH, Wolf SE, et al:Evolution of burn resuscitation in operation Iraqifreedom. J Burn Care Res 2006; 27:606–611

15. Miller JG, Carruthers HR, Burd DA: An algo-rithmic approach to the management of cu-taneous burns. Burns 1992; 18:200–211

16. Jastremski M, Jastremski C, Shepherd M, etal: A model for technology assessment asapplied to closed loop infusion systems.Technology Assessment Task Force of theSociety of Critical Care Medicine. Crit CareMed 1995; 23:1745–1755

17. Burns KE, Lellouche F, Lessard MR: Auto-mating the weaning process with advancedclosed-loop systems. Intensive Care Med2008; 34:1757–1765

18. Jannett TC, Wise MG, Sanders PN: An adaptivecontrol system for delivering heparin to pro-vide anticoagulation during hemodialysis. ConfProc IEEE Eng Med Biol Soc 1992; 14:2297–2298

19. Jannett TC, Kay GN, Sheppard LC: Auto-mated administration of lidocaine for thetreatment of ventricular arrhythmias. MedProg Technol 1990; 16:53–59

20. Yatabe T, Yamazaki R, Kitagawa H, et al: Theevaluation of the ability of closed-loop glycemiccontrol device to maintain the blood glucoseconcentration in intensive care unit patients.Crit Care Med 2011; 39:575–578

21. Absalom AR, De Keyser R, Struys MM: Closedloop anesthesia: Are we getting close to findingthe holy grail? Anesth Analg 2011; 112:516–518

22. Federal Aviation Administration. Accident andIncident Data: USAirways 1549 (AWE1549),January 15, 2009. 2009. Available at: http://www.faa.gov/data_research/accident_incident/1549/. Accessed April 1, 2011

Acute kidney injury: Clear the kidney of apoptotic debris!*

Acute kidney injury (AKI) is amajor clinical problem in in-tensive care units. Sepsis-induced AKI is the most com-

mon form of AKI observed in critically illpatients, but ischemia-reperfusion (I/R)

injury play a crucial role in the develop-ment of AKI in all types of shock, trauma,and transplantation. Similar to sepsis, I/Rinjury is accompanied by an intense sys-temic inflammatory response, which mayinduce organ damage at a distance fromthe initial ischemic insult site. Accumu-lating evidence highlights the crucial roleof effective clearance of apoptotic cellsafter ischemia to prevent the accumula-tion of inflammatory apoptotic cells, hy-percoagulable state, and impaired tissuerepair (1, 2). Effective clearance of apo-ptotic cells requires bridging molecules,which recognize apoptotic cells and facil-itate apoptotic target-phagocyte effector

juxtaposition, thereby enabling phagocy-tosis. Milk fat globule-epidermal growthfactor 8 (MFG-E8)/lactadherin is one ofthese bridging molecules (3–6).

MFG-E8 is a glycoprotein originallyfound in milk-fat globules and mammaryepithelial cells. MFG-E8 is expressed inseveral epithelial cells and in immunecells involving macrophages and den-dritic cells. MFG-E8 has a domain struc-ture of epidermal growth factor (EGF)1-EGF2-C1-C2 in which EGF indicatesEGF homology domains, and the C do-mains share homology with the phos-phatidylserine-binding domains of bloodcoagulation factors V and VIII. The sec-

*See also p. 2039.Key Words: acute kidney injury; apoptosis; lactad-

herin; milk fat globule-epidermal growth factor 8;phagocytic clearance

The authors have not disclosed any potential con-flicts of interest.


DOI: 10.1097/CCM.0b013e318226619c


http://www.faa.gov/data_research/accident_incident/1549/



ond EGF-like domain binds to the �v�3and �v�5 integrins. The second C do-main binds to phospholipids. MFG-E8promotes phagocytosis of apoptotic cellsby establishing a bridge between phos-phatidylserine on apoptotic cells and in-tegrins on phagocytes (6, 7), modulatesdownstream inflammatory signalingpathways, promotes vascular endothelialgrowth factor and neovascularization viathe induction of integrin-dependent Aktphosphorylation in endothelial cells (8),and promotes cell/cell adhesion such asthat during sperm-oocyte interaction (9).

In this context, an interesting manu-script in the current issue of Critical CareMedicine provides useful new informa-tion. In this issue, Matsuda and col-leagues (10) described that, while the ex-pression of MGF-E8 was reduced afterrenal I/R injury in mice, intraperitonealadministration of MGF-E8 attenuates tu-bular I/R injury and improves the survivalrate of renal I/R mice. This renal-protective effect appears to be mediatedthrough anti-inflammatory and antiapo-ptotic effects of recombinant murineMFG-E8 and improvement of capillaryfunctions (recovery of vascular endothe-lial growth factor expression and suppres-sion of endothelin-1 overexpression) inthe kidney. This study confirms previousresults of the same group reporting anattenuation of inflammation and acutelung injury by MFG-E8 after mesentericI/R (11). After superior mesenteric arteryocclusion for 90 mins followed by reper-fusion for 4 hrs, MFG-E8 levels decreasedin the spleen and lungs by 50% to 60%.Treatment with recombinant murineMFG-E8 significantly suppressed inflam-mation (tumor necrosis factor-�, inter-leukin-6, interleukin-1�, and myeloper-oxidase) and injury of the lungs, liver,and kidneys, with a significant improve-ment in survival of wild-type mice.

These two studies suggest that the de-creased production of MFG-E8 in the dis-tant organ may be one of the possiblemechanisms for distant organ injury afterI/R injury.

These studies lend credence to the no-tion that inflammation and uncleared de-bris from dying cells play a major role inthe I/R injury. This role is not only lim-ited to the ischemic organ (renal I/R in-jury in the Matsuda study), but is alsoimportant in the nonischemic organs(liver injury induced by renal ischemia inthe Matsuda study). Recent studies have

clearly shown that efficient phagocytosisof apoptotic debris is critical to the main-tenance of an anti-inflammatory milieu.Phagocytic clearance of apoptotic debrisis beneficial because it removes toxic orinflammatory stimuli and avoids an exag-gerated immune response with down-stream effects, such as secretion of cyto-kines and a production of reactive oxygenspecies. Fast removal of apoptotic cells byphagocytes in tissues and circulation mayprevent a secondary (postapoptotic) ne-crosis of apoptotic cells with leakage oftoxic or inflammatory debris. Ait-Oufellaet al (2) reported a marked acceleration ofatherosclerosis in MGF-E8-deficientmice, with substantial accumulation ofapoptotic debris, both systemically andwithin the developing lipid lesions. Thisaccumulation of apoptotic debris was as-sociated with a reduction in interleu-kin-10 in the spleen and an alteration ofnatural regulatory T-cell function, but anincrease in interferon-� production inboth the spleen and the atheroscleroticarteries. Miksa et al (12), in a cecal liga-tion and puncture model, found a reduc-tion of MFG-E8 protein levels in thespleen and liver by 48% and 70%, respec-tively. Peritoneal macrophages fromMFG-E8-treated rats displayed a 2.8-foldincreased ability to phagocytose apoptoticcells with an attenuation of the systemicinflammatory response.

As mentioned by Matsuda et al (10), theobserved beneficial effect of MFG-E8 intheir renal I/R model is attributed to the factthat MFG-E8 accelerates the phagocytosis ofapoptotic cells based on several studies (3, 4,12, 13). However, the authors reported thatrecombinant murine MFG-E8 decreases apo-ptotic cells in the kidney after I/R injury, butthere is no direct data showing that clearanceof apoptotic cells is impaired after renal I/R. Itis plausible that the MFG-E8–mediated pro-tection after renal I/R results from an im-provement of the MFG-E8–mediated apopto-tic cell phagocytosis, but this hypothesis hasto be demonstrated in further investigations.

Matsuda et al (10) provide new in-sights into the field of AKI therapy. Thisstudy suggests that recombinant MFG-E8may be beneficial for the treatment oftubular I/R injury and could be a noveltreatment option for AKI. This therapeu-tic hypothesis must be confirmed byother teams, but it is obviously an excit-ing therapeutic approach that must beactively investigated.

Anatole Harrois, MDJacques Duranteau, MD, PhD

Assistance Publique - Hopitauxde Paris

Hopital BicetreDepartement d’Anesthesie-

ReanimationUniversite Paris, Sud XILe Kremlin-Bicetre, France

REFERENCES

1. Thorp EB: Mechanisms of failed apoptoticcell clearance by phagocyte subsets in car-diovascular disease. Apoptosis 2010; 15:1124–1136

2. Ait-Oufella H, Kinugawa K, Zoll J, et al: Lac-tadherin deficiency leads to apoptotic cellaccumulation and accelerated atherosclero-sis in mice. Circulation 2007; 115:2168–2177

3. Hanayama R, Tanaka M, Miyasaka K, et al:Autoimmune disease and impaired uptake ofapoptotic cells in MFG-E8-deficient mice.Science 2004; 304:1147–1150

4. Hanayama R, Tanaka M, Miwa K, et al: Iden-tification of a factor that links apoptotic cellsto phagocytes. Nature 2002; 417:182–187

5. Bu HF, Zuo XL, Wang X, et al: Milk fatglobule-EGF factor 8/lactadherin plays a cru-cial role in maintenance and repair of mu-rine intestinal epithelium. J Clin Invest 2007;117:3673–3683

6. Fens MH, Mastrobattista E, de Graaff AM, etal: Angiogenic endothelium shows lactad-herin-dependent phagocytosis of aged eryth-rocytes and apoptotic cells. Blood 2008; 111:4542–4550

7. Leonardi-Essmann F, Emig M, Kitamura Y,et al: Fractalkine-upregulated milk-fat glob-ule EGF factor-8 protein in cultured rat mi-croglia. J Neuroimmunol 2005; 160:92–101

8. Silvestre JS, Thery C, Hamard G, et al: Lac-tadherin promotes VEGF-dependent neovas-cularization. Nat Med 2005; 11:499–506

9. Ensslin MA, Shur BD: Identification ofmouse sperm SED1, a bimotif EGF repeatand discoidin-domain protein involved insperm-egg binding. Cell 2003; 114:405–417

10. Matsuda A, Wu R, Jacob A, et al: Protectiveeffect of milk fat globule-epidermal growthfactor-factor VIII after renal ischemia-reperfusion injury in mice. Crit Care Med2011; 39:2039–2047

11. Cui T, Miksa M, Wu R, et al: Milk fat globuleepidermal growth factor 8 attenuates acutelung injury in mice after intestinal ischemiaand reperfusion. Am J Respir Crit Care Med2010; 181:238–246

12. Miksa M, Wu R, Dong W, et al: Dendriticcell-derived exosomes containing milk fatglobule epidermal growth factor-factor VIIIattenuate proinflammatory responses in sep-sis. Shock 2006; 25:586–593

13. Dasgupta SK, Abdel-Monem H, Guchhait P,et al: Role of lactadherin in the clearance ofphosphatidylserine-expressing red bloodcells. Transfusion 2008; 48:2370–2376


Too much of a good thing is not necessarily better*

I n a situation of emerging multi-drug resistance among importantpathogens and a very high (toohigh?) use of very-broad-spectrum

antibiotics in the intensive care units(ICUs) worldwide, the importance of an-timicrobial stewardship is increasinglynecessary and action is urgently needed(1). At the same time, sepsis is a contin-uous risk for patients in the ICUs andbecause delay to appropriate antimicro-bial therapy in severe bacterial infectionsis related to mortality (2–4), it is recom-mended in guidelines to initiate antimi-crobial therapy within 1 hr (5). The diag-nosis of sepsis is hampered, however, byits nonspecific presentation because thesystemic inflammatory response syn-drome is common not only in severeinfection, but also in various noninfec-tious conditions of inflammation. Theclinician needs help to rational use ofpotent antimicrobial therapy if weshould not give way to blind very broad-spectrum antibiotic exposure to every-one, which will be the way directly tountreatable infections resulting frommultiresistant organisms (6).

It is also a fact that only a minority ofseptic patients will have a microbiologi-cally documented bloodstream or other-wise serious infection, so most patientsare treated on clinical suspicion based onaltered organ functions, radiologic imag-ing, biochemistry, and biomarkers suchas C-reactive protein or procalcitonin(PCT). In several infectious diseases, theapplication of PCT guidance for decisionson antibiotic therapy has proven helpfulto support antibiotic stewardship and areduction in exposure to antibiotics isusually possible without adverse out-comes (7). This principle can work in thecritically ill population of the ICU from astepdown perspective as shown in the

French Procalcitonin to Reduce Patients’Exposure to Antibiotics in Intensive CareUnits (PRORATA) trial (8) and to reduceantibiotic exposure in ventilator-associ-ated pneumonia (9), but the question re-mains if it is possible to initiate antibiotictherapy with less extensive antimicrobialspectrum and then use PCT guidance toescalate when necessary.

The study by Jensen and coworkerspublished in this issue of Critical CareMedicine (10) is so far the largest multi-center randomized clinical trial exploringa biomarker-guided strategy of antibiotictherapy in the ICU setting. The studyconclusion challenges the dogma that es-calation from target-directed to verybroad-spectrum (“blind”) antibiotics be-fore a microbiologic diagnosis increasessurvival in the ICU. The trial was con-ducted in nine multidisciplinary ICUs inDenmark with a randomization of 1200patients to daily PCT guidance or stan-dard of care. Consequently, the escalationto broad-spectrum piperacillin/tazobac-tam or meropenem was significantlymore prevalent in the PCT group, butdisappointingly, no survival benefit wasdemonstrated. The results were robustbecause several subanalyses on stratifica-tion of the cohort confirmed the outcometo be nondifferent. The study confirmsprevious observations (11) that the sen-sitivity of baseline PCT is low and thuscannot be used to rule out infection,whereas the changes in PCT concentra-tion on repeated measurements predictsprognosis.

An unexpected finding for patients onthe PCT strategy was increased time onmechanical ventilation and longer timewith renal failure. Increased morbidityassociated with a PCT strategy was alsosuggested in the PRORATA trial (8). Ad-ditionally, the length of stay in the ICUwas prolonged in the PCT-guided strat-egy. These findings are a strong reminderthat new diagnostics and new therapiesshould not be introduced to routine carewithout prior rigorous scientific evalua-tion because unexpected adverse eventsmay occur. Too much of a good thing isnot necessarily better.

The strengths of the study were thedesign, the large sample size, inclusion ofmultiple centers, rigorous statisticalanalysis, and the complete follow-up un-til the primary end point. The authorsshould be congratulated with the effortsto report 28-day mortality as the pre-ferred outcome, superior to inhospitalmortality often reported in such studies.

The limitations of the study includethe mononational design because all cen-ters in Denmark had a low prevalence ofmultiresistant organisms in contrast tomost other countries, which reduce thepredictive value of a positive PCT alert. Inthe Scandinavian countries with a rela-tively low incidence of multiresistantpathogens, the study results are reassur-ing that it is safe to continue with arestrictive antibiotic policy in the ICUswithout escalating to very broad-spec-trum agents unless supported by micro-biologic findings. The proportional expo-sure in the standard-of-care group topiperacillin/tazobactam or meropenemwas as low as 0.07 and yet the survivalwas not adversely affected. However, forcountries with a higher prevalence ofmultidrug resistance, the results may notapply, so generalization is maybe not pos-sible. Another parameter, which may dif-fer among centers, is fungal infections inthe ICU because PCT response to Candidaspecies is considerably weaker as com-pared with bacterial infections (12). Fur-thermore, the mortality in ICU patients ismultifactorial and intervention within asingle area of treatment (e.g., antibiotics)rarely results in substantial mortalityreduction.

As reported, biomarkers cannot sub-stitute for good microbiologic data in theICU (13), but admittedly, the turnaroundtime is a problem. Although we still waitfor that “superior sensitive bedside, 1-hr,broad-range pathogen test covering allrelevant organisms with a very high spec-ificity to rule out infection in the septicpatient,” the clinician is dependent onmicrobiologic findings, biochemistry ofacute-phase reactants, and careful obser-vations of patient physiology responses.The hope for PCT guidance to improveoutcome is unclear, and the study by Jen-

*See also p. 2048.Key Words: antibiotics; clinical trial; intensive care;

procalcitoninThe author has not disclosed any potential con-

flicts of interest.Copyright © 2011 by the Society of Critical Care

Medicine and Lippincott Williams & Wilkins

DOI: 10.1097/CCM.0b013e3182207c13


sen and coworkers may not apply to allcountries. We are eagerly awaiting theresults of ongoing clinical trials on PCTguidance in ICU patients (14, 15).

Henrik Nielsen, MD, DMSciDepartment of Infectious

DiseasesAalborg HospitalAarhus University HospitalsAalborg, Denmark

REFERENCES

1. Dellit TH, Owens RC, McGowan JE, et al:Infectious Diseases Society of America andthe Society for Healthcare Epidemiology ofAmerica guidelines for developing an insti-tutional program to enhance antimicrobialstewardship. Clin Infect Dis 2007; 44:159 –177

2. Leibovici L, Shraga I, Drucker M, et al: Thebenefit of appropriate empirical antibiotictreatment in patients with bloodstream in-fection. J Intern Med 1998; 244:379 –386

3. Ibrahim EH, Sherman G, Ward S, et al: Theinfluence of inadequate antimicrobialtreatment of bloodstream infections on pa-tient outcome in the ICU setting. Chest2000; 118:146 –155

4. Kumar A, Roberts D, Wood KE, et al: Dura-tion of hypotension before initiation of effec-tive antimicrobial therapy is the critical de-terminant of survival in human septic shock.Crit Care Med 2006; 34:1589–1596

5. Dellinger RP, Carlet JM, Masur H, et al: Sur-viving Sepsis Campaign guidelines for man-agement of severe sepsis and septic shock.Crit Care Med 2004; 32:858–883

6. Nordmann P, Poirei L, Toleman MA, et al:Does broad-spectrum �-lactam resistancedue to NDM-1 herald the end of the antibi-otic era for treatment of infections caused byGram-negative bacteria? J AntimicrobChemther 2011; 66:689–692

7. Simon L, Gauvin F, Amre DR, et al: Serumprocalcitonin and C-reactive protein levels asmarkers of bacterial infection: A systematicreview and meta-analysis. Clin Infect Dis2004; 39:206–217

8. Bouadma L, Luyt CE, Tubach F, et al: Use ofprocalcitonin to reduce patients’ exposure toantibiotics in intensive care units (PRORATAtrial): A multicentre randomised controlledtrial. Lancet 2010; 375:463–474

9. Stolz D, Smyrnios N, Eggimann P, et al:Procalcitonin for reduced antibiotic expo-sure in ventilator-associated pneumonia: Arandomised study. Eur Respir J 2009; 34:1364 –1375

10. Jensen JU, Hein L, Lundgren B, et al: Pro-calcitonin-guided interventions against in-fections to increase early appropriate antibi-otics and improve survival in the intensivecare unit: A randomized trial. Crit Care Med2011; 39:2048–2058

11. Jensen JU, Heslet L, Jensen TH, et al: Pro-calcitonin increase in early identificationof critically ill patients at high risk of mor-tality. Crit Care Med 2006; 34:2596 –2602

12. Charles PE, Dalle F, Aho S, et al: Serumprocalcitonin measurement contribution tothe early diagnosis of candidemia in criticallyill patients. Intensive Care Med 2006; 32:1577–1583

13. Jung B, Embriaco N, Roux F, et al: Microbio-logical data, but not procalcitonin improve theaccuracy of the clinical pulmonary infectionscore. Intensive Care Med 2010; 36:790–798

14. Placebo controlled trial of sodium selenite andprocalcitonin guided antimicrobial therapy in se-vere sepsis (SISPCT). Available at: http://www.clinicaltrials.gov/ct2/show/NCT00832039. Ac-cessed April 27, 2011

15. Safety and efficacy of procalcitonin guided an-tibiotic therapy in adult intensive care units(ICUs) (SAPS). Available at: http://www.clinicaltrials.gov/ct2/show/NCT01139489. Ac-cessed April 27, 2011

Take a deep breath…*

Patient care is an art directed byscience. New discoveries fromgood clinical research oftenprompt changes in patient care

or in the way clinicians think about dis-ease. But by shifting their focus to thelatest topic in the literature, cliniciansmay ignore more mundane problems.When an old problem comes back intofocus, there is an opportunity for newinsight.

In this issue of Critical Care Medicine,Schmidt and colleagues (1) refocus ondyspnea in mechanically ventilated pa-tients. Their findings suggest that dys-pnea is common, often severe, is associ-ated with anxiety, responds to basicventilator changes in a third of cases, and

correlates with time on mechanical ven-tilation. They conclude that more re-search is needed to evaluate whether thebenefits of modern therapies for respira-tory failure “… are worth their dyspnea-related ransom.” This conclusion reso-nates at a scientific level. At the sametime, it suggests that there are timeswhen the science applied fails an individ-ual patient.

When clinical research uncovers adramatic finding, the natural tendency isto apply it. Low tidal volumes preventlung injury in the acute respiratory dis-tress syndrome (2); they might be helpfulwith other forms of respiratory failure aswell (3). Limiting (4) or avoiding (5)some forms of sedation might reduce theincidence of delirium or the time spenton mechanical ventilation. Of course,good clinical research tends to be carriedout under tightly controlled conditionson a specific group of patients. One of thedifficult questions an astute clinicianmust ask when applying scientific data toa patient is: do the patients in the studyreflect the patient I am caring for right

now? Understanding the mechanisms ofthe disease and the therapy helps answerthis question, but no mechanistic modelis complete. Ultimately, the clinicianmust look for reasons to accept or rejectthe application of clinical science to hispatient.

In the case of respiratory failure, thefocus on therapy (mechanical ventilation)has shifted from how best to support apatient’s breathing to how to avoid harm.Changes in ventilation strategies, seda-tion regimens, and bundled-care proto-cols for ventilated patients (6) reflected adesire to minimize complications. Asventilator-induced lung injury, delirium,and ventilator-associated pneumonia be-came the focus, the discomfort of dys-pnea receded.

In their study, Schmidt et al surveyedintubated critically ill patients for dys-pnea. Dyspnea is a distressing symptom,worse than pain for some patients (7).Pain is frequently a target for interven-tion, yet dyspnea is rarely discussed out-side of the literature of palliative care.When it is described, data suggest the

*See also p. 2059.Key Words: breathing pattern; dyspnea; mechani-

cal ventilation; patient anxiety; respiratory sensationsThe author has not disclosed any potential con-



DOI: 10.1097/CCM.0b013e3182217442


http://www.clinicaltrials.gov/ct2/show/NCT00832039




sensation is severe and uncomfortable (8).Schmidt et al point out that dyspnea de-serves attention. It is common (47% of pa-tients studied) and severe enough (medianvisual analog scale of 5) to merit a discus-sion about the benefits and hazards of me-chanical ventilation and sedation strate-gies. Regarding therapies, this studysuggests some proportion of dyspnea couldbe treated by ventilator changes. Further-more, lack of improvement in dyspnea cor-related with increased time on mechanicalventilation, raising the possibility that dys-pnea-focused therapies, at least in somepatients, might improve critical outcomes.The association of dyspnea with anxietyadds to a growing literature on the psycho-logical burden of critical illness, includingthe risks of posttraumatic stress disorder(9) and depression (10). Shifting the focusfrom mechanics and complications to thesensation of breathlessness, the authorshave uncovered a new therapeutic target.

It is refreshing to think that a classicmodel of symptom and therapy mightimprove outcomes, but too much mustnot be extrapolated from this initial of-fering. The study subjects are not rep-resentative of all mechanically venti-lated patients. The study patients wereresponsive and able to answer questionsabout dyspnea. Few, if any, had delir-ium. A large number were ventilated forneurologic conditions, especially motor

weakness, and extrapolation to patientswithout these disorders will be specu-lative. Nevertheless, this study chal-lenges conventional thinking. It shoulddrive the exploration of new hypothe-ses. Not wanting to let a good answer gounquestioned, we can speculate aroundthese findings. How often do ventilatorsettings contribute to dyspnea? Whencan it be treated without doing harm?How can we use it to predict patientoutcomes?

Schmidt et al raise a hypothesis that isnot yet disproven. To help patients inrespiratory failure, dyspnea may be theprice. The possibility is a cold reality, onethat should not remain on the periphery.This paper helps refocus the discussionalong such lines.

Mark E. Nunnally, MDUniversity of Chicago,

Department of Anesthesiaand Critical Care,Chicago, IL

REFERENCES

1. Schmidt M, Demoule A, Polito A, et al: Dys-pnea in mechanically ventilated critically illpatients. Crit Care Med 2011; 39:2059–2065

2. Ventilation with lower tidal volumes as com-pared with traditional tidal volumes for acutelung injury and the acute respiratory distresssyndrome. The Acute Respiratory DistressSyndrome Network. N Engl J Med 2000; 342:1301–1308

3. Kilpatrick B, Slinger P: Lung protectivestrategies in anaesthesia. Br J Anaesth 2010;105(Suppl 1):i108–i116

4. Kress JP, Pohlman AS, O’Connor MF, et al:Daily interruption of sedative infusions incritically ill patients undergoing mechanicalventilation. N Engl J Med 2000; 342:1471–1477

5. Strøm T, Martinussen T, Toft P: A protocol ofno sedation for critically ill patients receivingmechanical ventilation: A randomised trial.Lancet 2010; 375:475–480

6. Zilberberg MD, Shorr AF, Kollef MH:Implementing quality improvementsin the intensive care unit: Ventilator bun-dle as an example. Crit Care Med 2009;37:305–309

7. Banzett RB, Pedersen SH, Schwartzstein RM,et al: The affective dimension of laboratorydyspnea: Air hunger is more unpleasant thanwork/effort. Am J Respir Crit Care Med 2008;177:1384–1390

8. Rotondi AJ, Chelluri L, Sirio C, et al: Pa-tients’ recollections of stressful experienceswhile receiving prolonged mechanical venti-lation in an intensive care unit. Crit CareMed 2002; 30:746–752

9. de Miranda S, Pochard F, Chaize M, et al:Postintensive care unit psychological burdenin patients with chronic obstructive pulmo-nary disease and informal caregivers: A mul-ticenter study. Crit Care Med 2011; 39:112–118

10. Jubran A, Lawm G, Kelly J, et al: Depressivedisorders during weaning from prolongedmechanical ventilation. Intensive Care Med2010; 36:828–835

Antibiotics in sepsis: Timing, appropriateness, and (of course)timely recognition of appropriateness*

Each year approximately 500,000severe sepsis or septic shock pa-tients will present to U.S. emer-gency departments (1). The un-

fortunate realities are that as many as150,000 of them will die during their hos-pitalization and the incidence of sepsis issteadily increasing (2, 3). Evidence suggests

that outcomes are improved with earlygoal-directed therapy (EGDT) (4) andprompt, effective antibiotic administration(5, 6) during the most proximal phase ofcritical illness.

In this issue of the Critical Care Med-icine, Puskarich et al (7) examine therelationship between time to antibioticsand mortality using data obtained fromtheir parent, randomized, multicenteredclinical trial, which compared lactateclearance to central venous oxygen satu-ration as goals of early severe sepsis ther-apy (8). Experimental data support thisinvestigation: in the murine surgicalimplantation model of septic shock ofKumar et al (9), when antibiotics were

initiated before shock onset, mortalitywas �20%, and if they were delayed �3hrs after shock onset, mortality was�85%. However, to date, only two hu-man studies have examined this impor-tant question.

In 2006, Kumar and colleagues (5)examined the relationship between theduration of hypotension before the ini-tiation of effective antimicrobial ther-apy and mortality in septic intensivecare unit patients. In this large, retro-spective, multicentered study, in-hospital mortality was 58% and the me-dian time to effective antimicrobialswas 6 hrs (interquartile range 2–15hrs). If appropriate antimicrobials were

*See also p. 2066.Key Words: appropriate antibiotics; emergency

medicine; goal-directed resuscitation; septic shock;severe sepsis; time to antibiotics



DOI: 10.1097/CCM.0b013e3182226ffa


given in the first hour of hypotension,mortality was 20.1%; for each hour’sdelay over the next 6 hrs, mortalityincreased by an average of 7.6%. Themessage to critical care clinicians wasloud and clear: do not delay effectiveantimicrobial therapy in patients withseptic shock as every hour matters.

In 2010, we presented our experiencefrom a single academic emergency de-partment with a mature EGDT protocol(6). We sought to determine whethertime to antibiotics was associated withmortality in patients receiving EGDT inthe emergency department. We studied261 patients with severe sepsis (48% ofthe cohort) or septic shock (the remain-ing 52%) over a 2-yr interval. The mediantime from triage to antibiotics was �2hrs (119 mins, interquartile range 76–192 mins), and from EGDT qualificationto antibiotics was 42 mins (interquartilerange 0–93 mins). In-hospital mortalitywas 31%, consistent with the interven-tion arm of the original EGDT trial (4).We found that time from triage to appro-priate antibiotics and time from qualifi-cation to appropriate antibiotics, notsimply time to initial antibiotics, wereassociated independently with in-hospitalmortality. Our results supported the gen-eral principle that timely, goal-directedresuscitation and timely, effective antimi-crobials save lives independent of eachother. Admittedly, our study warrantedconfirmation.

We were excited to read that Puskar-ich and colleagues took the baton to fur-ther our understanding of this compli-cated relationship. In this multicenteredcohort study, the median time from tri-age to antibiotics was 115 mins (inter-quartile range 65–175 mins). Despite81% of the patients fulfilling hemody-namic criteria for septic shock, in-hospital mortality was only 18.9%. Simi-lar to our study (6), they found nosignificant relationship between timefrom triage to initial antibiotics and in-hospital mortality; however, a delay inantibiotics until after shock recognitionwas independently associated with mor-tality. In effect, at highly experiencedcenters where the time to antibiotics hasbeen minimized and care delivery opti-mized, hourly delays in antibiotic admin-istration are not associated with mortal-ity—unless the patient either presents inshock or is initially not recognized asbeing at risk for the development ofshock and his/her antibiotics are not ad-ministered until after shock ensues. This

latter observation is consistent with Ku-mar’s murine model of septic shock andchallenges sepsis management guide-lines, which recommend administeringeffective antimicrobials within 1 hr ofrecognition of severe sepsis or septicshock (9, 10).

The strengths of the present study arenumerous. First, given the prospective,prespecified proposal to study this ques-tion within the framework of a clinicaltrial, the authors were able to accuratelyand precisely capture the time variables.Second, the exclusion criteria weresparse, permitting wide generalizabilityof the results. Third, the authors shouldbe commended for the use of institu-tion-specific antibiograms, allowing cli-nicians to select the appropriate antibi-otic in 91% of blood-culture-positivepatients. Nevertheless, while assur-ances are given that the remaining pa-tients received appropriate, antibi-ogram-based antibiotics, the potentialfor residual confounding exists.

The study’s limitations, in contrast,are few. First, while the multicentereddesign of the study is meritorious, thepotential for center effects confoundingthe relationship between time to antibi-otics and mortality exists (11) and wasnot examined in the current study. Thisis a result of the highly standardized re-suscitation protocol, which, by design,did not standardize antibiotic administra-tion but rather noted only that “antibiot-ics were administered as soon as practi-cal” (8). Second, the authors did not testfor interaction between the two resusci-tation protocols studied in the parenttrial and the relationship between timingof antibiotics and mortality. Althoughunlikely given that there were no differ-ences observed in other co-interventions,it is conceivable that behavior regardingthe timing of emergency department co-interventions (e.g., antibiotic administra-tion) could have been influenced by theunblinded treatment assignment. Third,they only report the appropriateness of an-tibiotics for blood-culture-positive patients(34.4%), whereas other studies have re-ported the results of positive cultures fromall sources, with percentages ranging from56.7% to 82.2% (5, 6, 12). This limits thegeneralizations that can be made about theoverall appropriateness of initial antibiot-ics. Finally, these three centers managed toachieve a mortality rate of 18.9% and todeliver antibiotics effectively in �2 hrs,which may have limited the power to detect

a relationship between time to antibioticsand mortality.

Despite these limitations, the study byPuskarich et al is the largest to date onthe relationship between the timing ofantibiotics and mortality in patients re-ceiving goal-directed resuscitation andfurthers our understanding of this impor-tant question. Now all we need to do islearn how to predict which septic patientswill rapidly deteriorate to fulfill criteriafor severe sepsis and septic shock andadminister appropriate antibiotics beforethey do. Let’s get to work.

Mark E. Mikkelsen, MD, MSCEPulmonary, Allergy, and

Critical Care DivisionDepartment of MedicineUniversity of Pennsylvania

School of MedicinePhiladelphia, PA

David F. Gaieski, MDDepartment of Emergency

MedicineUniversity of Pennsylvania

School of MedicinePhiladelphia, PA

REFERENCES

1. Wang HE, Shapiro NI, Angus DC, et al: Na-tional estimates of severe sepsis in UnitedStates emergency departments. Crit CareMed 2007; 35:1928–1936

2. Angus DC, Linde-Zwirble WT, Lidicker J, etal: Epidemiology of severe sepsis in theUnited States: Analysis of incidence, out-come, and associated costs of care. Crit CareMed 2001; 29:1303–1310

3. Martin GS, Mannino DM, Eaton S, et al: Theepidemiology of sepsis in the United Statesfrom 1979 through 2000. N Engl J Med 2003;348:1546–1554

4. Rivers E, Nguyen B, Havstad S, et al: Earlygoal-directed therapy in the treatment of se-vere sepsis and septic shock. N Engl J Med2001; 345:1368–1377

5. Kumar A, Roberts D, Wood KE, et al: Dura-tion of hypotension before initiation of effec-tive antimicrobial therapy is the critical de-terminant of survival in human septic shock.Crit Care Med 2006; 34:1589–1596

6. Gaieski DF, Mikkelsen ME, Band RA, et al:Impact of time to antibiotics on survival inpatients with severe sepsis or septic shock inwhom early goal-directed therapy was initi-ated in the emergency department. Crit CareMed 2010; 38:1045–1053

7. Puskarich MA, Trzeciak S, Shapiro NI, et al:Association between timing of antibiotic admin-istration and mortality from septic shock in pa-tients treated with a quantitative resuscitationprotocol. Crit Care Med 2011; 2066–2071

8. Jones AE, Shapiro NI, Trzeciak S, et al: Lactateclearance vs central venous oxygen saturation


as goals of early sepsis therapy: A randomizedclinical trial. JAMA 2010; 303:739–746

9. Kumar A, Haery C, Paladugu B, et al: Theduration of hypotension before the initiationof antibiotic treatment is a critical determi-nant of survival in a murine model of Esch-erichia coli septic shock: Association with

serum lactate and inflammatory cytokinelevels. J Infect Dis 2006; 193:251–258

10. Dellinger RP, Levy MM, Carlet JM, et al: Surviv-ing Sepsis Campaign: International guidelines formanagement of severe sepsis and septic shock:2008. Crit Care Med 2008; 36:296–327

11. Localio AR, Berlin JA, Ten Have TR, et al:

Adjustments for center in multicenter studies:An overview. Ann Intern Med 2001; 13:112–123

12. Kumar A, Ellis P, Arabi Y, et al: Initiationof inappropriate antimicrobial therapy re-sults in a fivefold reduction of survival inhuman septic shock. Chest 2009; 136:1237–1248

Myocardial infarction complicated by cardiogenic shock: A possiblerole for the Impella device?*

Cardiogenic shock (CS) is theleading cause of mortality inpatients hospitalized for myo-cardial infarction (MI). Due to

better infarct management, the incidenceof CS complicating MI has decreased andprognosis has improved in recent years(1). Nevertheless, with hospital mortalityrates up to 50%, the outcome of pa-tients with CS is still poor. Although ithas been acknowledged recently thatCS cannot be seen as a sole mechanisticmodel, and that an accompanying sys-temic inflammatory response contrib-utes to the disease entity, the severity ofthe left ventricular function depressionand the grade of mitral regurgitation atthe initial echocardiographic assessmentare significantly linked to worse prognosis(2). In most severe cases, despite successfulopening of the infarct-related artery, treat-ment with high-dose vasopressors/ino-tropes, and intra-aortic counterpulsation,hemodynamics cannot be stabilized and leftventricular assist devices are seen as thelast therapeutic option.

In this context, the retrospectiveanalysis of Engstrom and colleagues (3)in this issue of Critical Care Medicine,describes the use of the Impella 2.5/5.0devices in patients with profound CSdeemed to be refractory to conventionalinotropic and vasopressor support. Theclear strength of this study is that itprovides a real-world scenario, with de-tailed case-to-case descriptions. Someimportant clinical issues concerning

the patient population in this seriesshould be considered. Close to 60% ofpatients were resuscitated before arrivalto hospital, which implies that CS wasassociated also with the systemicpostresuscitation syndrome in manypatients. Although no clear distinctionsof these two components can be madein terms of hemodynamic sequelae, itseems obvious that a combination ofthese syndromes yields a unique andvery poor prognosis. Regarding the MIcharacteristics, it is noteworthy that al-though the total ischemic time avail-able in 27 out of 34 patients (symptomsto balloon time) was low when com-pared with the landmark SHOCK trial(4), a coronary stent was implanted inonly 73% of patients, and normal cor-onary flow was present in only 58% ofpatients at the end of the procedure!According to the case descriptions, pos-sible causes include that a substantialproportion of patients had never gainedan adequate hemodynamic response tosustain sufficient coronary perfusionpressure, either because of the severityof the shock itself or because of ongo-ing malignant arrhythmias. Interest-ingly also is the high proportion of pa-tients with triple-vessel coronary arterydisease, which is in contrast with otherseries of patients with MI complicatedby CS. In light of these specific patientcharacteristics, what conclusions uponthe feasibility, safety, and efficacy of theImpella pump system in patients withCS can be drawn from this series? Thepriming and insertion of the Impella2.5 device that can be inserted percuta-neously through a 12.5F sheet via thecommon femoral artery seems to bepossible in a dedicated catheterizationlab. However, the hemodynamic re-sponse to the 2.5 device was classified

as insufficient in 32% of patients withinthe first 48 hrs, and these patients wereupgraded to a 5.0 Impella device thatrequires a surgical cutdown of the fem-oral artery and usually the use of aDacron tube. Considering safety issuesduring the extended use of both the 2.5and 5.0 systems, several issues have tobe addressed. Three of 34 patients ex-perienced potential embolic events (twostrokes, one bowel ischemia) in thecourse of or after Impella treatment.One additional patient experienced se-vere limb ischemia after removal of the2.5 Impella device, which required sur-gical intervention. It should be alsokept in mind that, as with other extra-corporeal life support systems, patientswith the Impella system have to bedeeply sedated, prolonging the inten-sive care unit stay even after successfulweaning of the device. Although theease of use of the Impella 2.5 pump isattractive, its hemodynamic supportseems inadequate in patients with pro-found CS. Only two of 25 patients whowere solely treated with the 2.5 Impellasystem survived the hospital stay. In linewith the obvious inadequate hemodynamicsupport observed in this series, are datafrom the ISAR-SHOCK study where in-creases of the cardiac power index were notsignificantly different in comparison withpatients with conventional therapy, includ-ing an intra-aortic balloon pump, beyond 2hrs (5). The Impella 5.0 pump seems toprovide adequate hemodynamic supportand could be an attractive device in patientswith profound CS, possibly also in thosewith ventricular septal defects as a conse-quence of MI as a bridge to surgery. Thesafety issues, especially with regard tobleeding and hemolysis, seem to comparewell with other emergency assist devices.However, prospective data, in particular in

*See also p. 2072.Key Words: cardiogenic shock; Impella device; left

ventricular assist devices; myocardial infarctionThe authors have not disclosed any potential con-



DOI: 10.1097/CCM.0b013e3182217476


comparison with extracorporeal life sup-port systems, are missing.

This study adds important institutionalexperiences to the growing dataset regard-ing the use of left ventricular assist devicesin patients with CS. Nevertheless, such ex-periences are no substitute for properly de-signed prospective trials. In patients withprofound CS, such trials could adopt thegrowing experience with the Impella deviceand opt for an initial strategy with the5.0 Impella pump. Before the results ofthose trials are available, the decisionto use this device must be taken in amultiprofessional healthcare team ap-proach on an individualized and stan-dardized basis.

Lisa Krenn, MDGeorg Delle Karth, MD

Medical University of Vienna,Department of Cardiology,Vienna, Austria

REFERENCES

1. Goldberg RJ, Spencer FA, Gore JM, et al:Thirty-year trends (1975 to 2005) in the magni-tude of, management of, and hospital death ratesassociated with cardiogenic shock in patients withacute myocardial infarction: A population-basedperspective. Circulation 2009; 119:1211–1219

2. Picard MH, Davidoff R, Sleeper LA, et al: Echo-cardiographic predictors of survival and re-sponse to early revascularization in cardiogenicshock. Circulation 2003; 107:279–284

3. Engstrom AE, Cocchieri R, Driessen AH, et al:

The Impella 2.5 and 5.0 devices for ST-elevation myocardial infarction patients pre-senting with severe and profound cardio-genic shock: The Academic Medical Centerintensive care unit experience. Crit CareMed 2011; 39:2072–2079

4. Hochman JS, Sleeper LA, Webb JG, et al: Earlyrevascularization in acute myocardial infarctioncomplicated by cardiogenic shock. SHOCK In-vestigators. Should We Emergently Revascular-ize Occluded Coronaries for Cardiogenic Shock.N Engl J Med 1999; 341:625–634

5. Seyfarth M, Sibbing D, Bauer I, et al: A random-ized clinical trial to evaluate the safety and effi-cacy of a percutaneous left ventricular assistdevice versus intra-aortic balloon pumping fortreatment of cardiogenic shock caused by myo-cardial infarction. J Am Coll Cardiol 2008; 52:1584–1588

Electrocardiogram interpretation for ischemia in patients withseptic shock: A disheartening exercise*

Like many diagnostic tests, elec-trocardiogram (ECG) interpre-tation requires a combinationof knowledge, skill, and practi-

cal clinical experience. Knowledge of thepathophysiology of electrocardiographicabnormalities, skill in recognizing com-mon abnormal ECG patterns, and experi-ence in relating the result of the ECG toa patient’s clinical situation are all com-ponents of successful interpretation. Theprincipal goals of ECG monitoring in theintensive care unit (ICU) are to alert staffto changes in cardiac rhythm, which mayherald life-threatening events, to alertstaff to sudden changes in cardiacrhythm, which are themselves life-threatening, and to identify silent isch-emia. Like most clinical tests, the ECGyields both false-positive and false-negative results. For example, the routineECG surveillance in the ICU, which usu-ally includes continuous two-lead moni-toring, has been shown to have a very lowsensitivity for myocardial ischemia (1).

It is of great importance in clinicalpractice to be aware of the diagnosticlimitations of any clinical test. Reliableknowledge of test characteristics, such assensitivity and specificity, in different testsituations are essential, as well as an un-derstanding of the population character-istics in which the test is performed (e.g.,pretest probability of a certain condition).The most important questions for a diag-nostic ECG in the ICU are—and thisholds true for most tests in medicine—what are the clinical consequences forthe patient? Has an abnormal ECG prog-nostic value? What additional tests arerequired? Can we influence the abnor-mality by a therapeutic intervention thathas been shown to be effective in well-designed trials and which does not lead tofurther harm?

Test results are not absolute, and(mis-) interpretation of test values canlead to significant problems, as the his-tory of the pulmonary artery catheter hasshown us. The proper use of the pulmo-nary artery catheter requires a perfectknowledge of the numerous pitfalls anddifficulties in interpretation of its mea-surements (2). The same applies for ECGinterpretation in intensive care patients.The level of agreement in ECG interpre-tation among clinicians has to be accept-able to provide reliable diagnostic andprognostic information.

In this issue of Critical Care Medicine,Mehta and colleagues (3) present a sub-study of the Vasopressin and SepticShock Trial (VASST), in which the intra-and inter-rater agreement of ECG inter-pretation with and without knowledge oftroponin values in adult septic patients isinvestigated. A total of 373 ECGs wereobtained from 121 septic patients and in-dependently and nonsequentially inter-preted by an intensivist and a cardiologistusing a checklist to standardize interpre-tation. Appropriate statistical methodswere used to describe intra- and inter-rater agreement with and without knowl-edge of troponin values for each of sevencategories: normal, ischemia, atrial ar-rhythmias, bundle branch block, ST ele-vation, T-wave inversion, and Q waves. Ascould be expected, both intra- and inter-rater agreement for specific ECG pat-terns, such as atrial arrhythmias andbundle branch block, were good to verygood. However, intrarater agreement wasonly moderate for ischemia, and the in-ter-rater agreement for ischemia im-proved from fair to moderate when thereaders were unblinded to troponin.

These results are important and in-crease our understanding of the limita-tions of ECG interpretation, especially forthe diagnosis of ischemia in patients withseptic shock. Specific strengths of thisstudy include the large number of ECGs,

*See also p. 2080.Key Words: agreement; electrocardiogram; isch-

emia; reliability; sepsis; troponinThe author has not disclosed any potential con-



DOI: 10.1097/CCM.0b013e3182266036


the standardization of the interpretation,and the very specific cohort of patientswith septic shock. The ECGs were inter-preted nonsequentially for patient andtime, which prevents so-called couplingof images (the interpretation of the nextECG is influenced by the previous ECG).Unfortunately, at the same time thistechnique limits generalizability and ex-trapolation to real-life practice. Interpre-tation of serial or sequential ECGs andcomparison with patients’ old ECGs islikely to improve the real accuracy ofinterpretation of abnormalities (e.g., typ-ical dynamic changes over time as seen innew ischemia or infarction as opposed tofixed abnormalities in patients with pre-vious cardiac “insults”). In addition, thepretest probability for detection of isch-emia in this cohort of patients was influ-enced by the exclusion of patients withacute coronary syndrome or severe heartfailure. Another limitation of the currentstudy is the lack of a gold standard forECG interpretation. Agreement betweenboth readers does not automaticallymean their interpretation was correct.One common feature of most ECG inter-pretation studies is the use of an expertelectrocardiographer “gold standard,”typically a consensus panel of cardiolo-gists. These panels have shown to havegood intrarater agreement (4).

The results of the current study areconsistent with what has been shown pre-viously, which is that ECG detection ofischemia in intensive care patients ingeneral is not as straightforward as it maybe in a pure cardiac population. For ex-ample, in a recent observational cohortanalysis of ST elevation in critically illpatients, 85% of patients with electrocar-diographic ST-segment elevation myo-cardial infarction (STEMI) had peak tro-ponin elevation �5 ng/mL, which madeclinically “real” STEMI in these patientsunlikely (5). This means that in ICUs, theECG has a poor predictive value for acutemyocardial infarction and lacks the spec-ificity compared with non-ICU patients.Correlation with clinical symptoms andsigns is essential but unfortunately oftennot contributory in ICU patients. For ex-ample, only 14% of patients who experi-ence a perioperative myocardial infarc-tion will have chest pain, and only 53% ofpatients will have clinical signs or symp-toms (6). The ECG, therefore, which isthe gold standard for diagnosing clinicalSTEMI in patients who present with chestpain, may be a poor diagnostic modalityto indicate STEMI in critically ill patients

in ICUs. Twelve-lead ECGs in the ICU areoften obtained for reasons other thanchest pain, such as arrhythmias, hypoten-sion, agitation, pulmonary edema, a sud-den change in the vital signs, or simply asworkup for weaning failure. Under thesecircumstances, a bedside transthoracic ortransesophageal echocardiogram mayhelp increase or decrease the likelihoodof ischemia or STEMI (7). Emergent car-diac catheterization should probably bereserved for a very small number of casesin which the clinical picture and anotherimaging study, such as the echocardio-gram, independently indicates a highlikelihood of STEMI.

Another important issue raised by thecurrent study is the effect of knowledge oftroponin levels to ECG interpretation.The authors report that knowledge of tro-ponin levels improved the inter-rateragreement for myocardial ischemia intheir study. Interestingly however,knowledge of troponin levels caused anincrease in the number of ECGs reportedas normal by Reader 1, but a decrease inthis number reported by Reader 2. Bothchanges were statistically significant andthe inter-rater agreement for normalitydecreased from moderate to fair. How canwe explain these discrepancies? Bearingin mind that one reader was an intensiv-ist and the other reader a cardiologist, itis not unreasonable to hypothesize thatboth readers, working in different clinicalenvironments, interpret troponin levelsdifferently. In cardiology clinical practice,elevated troponin levels are oftenstrongly correlated with myocardial isch-emia. The new generation of sensitiveassays for troponin are highly sensitivebut not very specific for the diagnosis ofmyocardial infarction (8, 9). This repre-sents a true clinical problem in intensivecare patients, where troponin elevationscan occur in settings other than myocar-dial infarction, the most common beingtachycardia, right ventricular strain andfailure in the setting of pulmonary em-boli or exacerbation of chronic obstruc-tive pulmonary dysfunction, and myocar-dial necrosis caused by neurohumoralchanges and/or elevated left ventricularend-diastolic pressure in the setting ofcongestive heart failure, head injury, orsubarachnoidal bleeding. An increasedtroponin measurement has been shownto be an independent predictor of mortal-ity after noncardiac surgery as well asfollowing cardiac surgery (10, 11). Fi-nally, and relevant to the current study,elevated troponin concentrations have

also been reported in patients with sepsisand septic shock without indication ofsignificant coronary artery disease (12,13). It has been demonstrated that tro-ponin elevations occur far more oftenthan ischemia, even in patients withknown or high probability for coronaryartery disease in a general ICU, and thatnecrosis but not detectable ischemia isrelated to mortality (14). Unfortunately itis not possible to reliably discriminateischemic from nonischemic causes of tro-ponin elevation by simply changing thecutoff value. However, a rising or fallingpattern of troponin values could be help-ful in discriminating acute injury fromsome other causes, such as end-stage re-nal disease (15).

Where to go from here? Should allECGs be interpreted by cardiologists asdiscussed by the authors? In addition tobeing unpractical, the literature showsthat this may not improve the number offalse-positive and -negative ECGs forischemia. The interpretation by severalcardiologists reading the same ECG oftenvaries substantially. Even one cardiolo-gist reading the same ECG on separateoccasions may have substantially differ-ent interpretations (4, 16). In the currentstudy, both readers - one of whom is anexperienced cardiologist - showed onlymoderate intrarater agreement for isch-emia. In the abovementioned cohortanalysis in critically ill patients, a cardi-ologist agreed with the computer inter-pretation of ST-elevation myocardial in-farction in 39% of cases, but of thosepatients, only 33% showed a significantrise in the troponin level (5). Physiciansof all specialties and levels of training, aswell as computer software for interpret-ing ECGs, frequently make errors in in-terpreting ECGs when compared to ex-pert electrocardiographers. Adversepatient outcomes, however, occurred in-frequently when ECGs were incorrectlyinterpreted, typically in �1% of interpre-tations (17). Currently there is no evi-dence-based minimum number of ECGinterpretations that is ideal for attainingor maintaining competency in ECG inter-pretation skills (18). Since ECG monitor-ing is an integral and essential part of theintensive care environment, formal in-tensivist training on ECG interpretationshould be a standardized and compulsorypart of any intensive care curriculum.

In conclusion, the current study con-firms that, in septic shock patients, sig-nificant variability of ECG interpretationof myocardial ischemia exists. Correla-


tion with all other pieces of the diagnos-tic puzzle is essential, and this includesthe clinical context, interpreting serialECGs, the serial measurement of cardiacbiomarkers despite their known limita-tions in the ICU, and bedside echocardi-ography, to prevent both over- and un-derdiagnosis of myocardial ischemia inICU patients.

Frank van Haren, MD, PhDWaikato HospitalHamilton, New Zealand

REFERENCES

1. Martinez EA, Kim LJ, Faraday N, et al: Sen-sitivity of routine intensive care unit surveil-lance for detecting myocardial ischemia. CritCare Med 2003; 31:2302–2308

2. Richard C, Monnet X, Teboul JL: Pulmonaryartery catheter monitoring in 2011. CurrOpin Crit Care 2011; 17:296–302

3. Mehta S, Granton J, Lapinsky SE, et al:Agreement in electrocardiogram interpreta-tion in patients with septic shock. Crit CareMed 2011; 39:2080–2086

4. Holmvang L, Hasbak P, Clemmensen P, et al:Differences between local investigator andcore laboratory interpretation of the admis-sion electrocardiogram in patients with un-stable angina pectoris or non-Q-wave myo-cardial infarction (a Thrombin Inhibition in

Myocardial Ischemia [TRIM] substudy). Am JCardiol 1998; 82:54–60

5. Rennyson SL, Hunt J, Haley MW, et al: Elec-trocardiographic ST-segment elevation myo-cardial infarction in critically ill patients: Anobservational cohort analysis. Crit Care Med2010; 38:2304–2309

6. Devereaux PJ, Goldman L, Cook DJ, et al:Perioperative cardiac events in patients un-dergoing noncardiac surgery: A review of themagnitude of the problem, the pathophysiol-ogy of the events and methods to estimateand communicate risk. CMAJ 2005; 173:627–634

7. Poelaert J: Use of ultrasound in the ICU. BestPract Res Clin Anaesthesiol 2009; 23:249–261

8. Keller T, Zeller T, Peetz D, et al: Sensitivetroponin I assay in early diagnosis of acutemyocardial infarction. N Engl J Med 2009;361:868–877

9. Reichlin T, Hochholzer W, Bassetti S, et al:Early diagnosis of myocardial infarction withsensitive cardiac troponin assays. N EnglJ Med 2009; 361:858–867

10. Levy M, Heels-Ansdell D, Hiralal R, et al:Prognostic value of troponin and creatinekinase muscle and brain isoenzyme measure-ment after noncardiac surgery: A systematicreview and meta-analysis. Anesthesiology2011; 114:796–806

11. Januzzi JL: What is the role of biomarkermeasurement after cardiac surgery? MinervaAnestesiol 2011; 77:334–341

12. Ammann P, Fehr T, Minder EI, et al: Eleva-tion of troponin I in sepsis and septic shock.Intensive Care Med 2001; 27:965–969

13. Ammann P, Maggiorini M, Bertel O, et al:Troponin as a risk factor for mortality incritically ill patients without acute coronarysyndromes. J Am Coll Cardiol 2003; 41:2004–2009

14. Landesberg G, Vesselov Y, Einav S, et al:Myocardial ischemia, cardiac troponin, andlong-term survival of high-cardiac risk criti-cally ill intensive care unit patients. CritCare Med 2005; 33:1281–1287

15. Thygesen K, Alpert JS, Jaffe AS, et al: Di-agnostic application of the universal defi-nition of myocardial infarction in the in-tensive care unit. Curr Opin Crit Care2008; 14:543–548

16. Massel D, Dawdy JA, Melendez LJ: Strict re-liance on a computer algorithm or measur-able ST segment criteria may lead to errorsin thrombolytic therapy eligibility. AmHeart J 2000; 140:221–226

17. Salerno SM, Alguire PC, Waxman HS: Com-petency in interpretation of 12-lead electro-cardiograms: A summary and appraisal ofpublished evidence. Ann Intern Med 2003;138:751–760

18. Salerno SM, Alguire PC, Waxman HS: Train-ing and competency evaluation for interpre-tation of 12-lead electrocardiograms: Recom-mendations from the American College ofPhysicians. Ann Intern Med 2003; 138:747–750

Danger at the doorstep: Regulation of bacterial translocationacross the intestinal barrier by nitric oxide*

T he ability of the host to regu-late the permeability of the in-testinal barrier represents amajor determinant of disease

vs. health. This is particularly relevant tothe critically ill surgical patient, in whomintestinal obstruction – be it due to me-chanical causes or in the setting of ileus –is associated with an increase in infec-tious complications (1, 2). The fact thatsuch infections are seen to be caused

predominantly by enteric microbes hasled to the widespread notion that intesti-nal barrier dysfunction develops as a con-sequence of intestinal obstruction, re-sulting in the translocation of entericbacteria (3). It therefore stands to reasonthat an understanding of the factors thatregulate normal barrier function duringhealth and in the presence of disease mayprovide important insights into enhanc-ing care of the critically ill patient.

It has been known for some time thatthere are several pathways by which en-teric microbes or their microbial-derivedantigenic ligands can access the circula-tion. The intestinal mucosa is lined by atightly interconnected layer of epithelialcells over which a layer of mucus existsthat together represent the first line ofdefense against bacterial passage (4, 5).To transit across the normally imper-

meant intestinal mucosal barrier, bacte-ria can either transit directly through theintestinal epithelial cells (IECs, i.e., tran-scellular passage), pass between the IECs(paracellular passage), or they can be ac-tively internalized by dendritic cells andmacrophages that lie within the laminapropria with the capacity to extend cellu-lar processes between the IECs that caninternalize the bacteria and deliver themto the lymph nodes and circulation (6).While IECs have been shown to have theability to internalize whole bacteria bothin vitro and in vivo (7), the effect of in-testinal obstruction on transcellular bac-terial internalization by IECs remains un-known. By contrast, it is well known thatadjacent IECs are interconnected by tightjunctions that determine the extent towhich large molecules, and possibly bac-teria, can pass between the adjacent IECs

*See also p. 2087.Key Words: bacterial translocation; epithelial per-

meability; intestinal obstruction; nitric oxide; signaltransduction; tight junctions



DOI: 10.1097/CCM.0b013e31822661ad


(8). Recent studies have demonstratedthat circulating cytokines and other pro-inflammatory molecules can cause an in-crease in the permeability of the intesti-nal barrier by causing tight junctiondysfunction or internalization (5, 9).However, the precise role of tight junc-tion dynamics in the maintenance of bar-rier integrity in the setting of intestinalobstruction, and the molecular mecha-nisms involved in the regulation of tightjunctions in this setting, remain largelyunexplored.

In this issue of Critical Care Medicine,Wu and colleagues (10) now attempt tosolve this riddle. Using an elegant modelof mechanical intestinal obstruction inmice, this group demonstrates that thepresence of intestinal obstruction leads toan increase in permeability across theintestinal barrier and an increase in bac-terial translocation, and that thesechanges are associated with a disruptionin tight junctions (10). In seeking toidentify the molecular mechanisms in-volved, Wu et al now also show that in-testinal obstruction is associated with anincrease in inducible nitric oxide synthase,which led to disruptions in tight junctions,likely through a pathway that involved theintermediates myosin light chain, myosinphosphatase target subunit 1, and proteinkinase C � (10). Although the precise dy-namics of this pathway in vivo are un-clear, the authors show that at least invitro, there is an apparent role for nitricoxide signaling leading to a Rho-depen-dent phosphorylation in myosin lightchain that was required for the reductionin tight junctions (10). Taken together,these findings place the spotlight on therole of nitric oxide signaling in the path-ways by which tight junctions may bedisrupted in intestinal obstruction.

It is useful to consider these findingsin the broad context of bacterial translo-cation in general, and in the role of tightjunctions in the setting of intestinalobstruction in particular. In this vein,it should be mentioned that while it isaccepted that a disruption in tight junc-tions can lead to the transepithelial pas-sage of large solutes and other macro-molecules, it is unclear whetherdisruptions in tight junctions have directeffects on the translocation of entire bac-teria. In support of an argument againstthe role of tight junctions in regulatingbacterial translocation, mice lacking my-osin light chain kinase, which have abaseline instability in tight junctions, did

not demonstrate any spontaneous evi-dence of sepsis (11). However, these samemice did demonstrate an increase in mu-cosal immune activation. It is thus pos-sible that the findings of increased mu-cosal permeability that Wu et al nowobserve may in fact represent the effectsof increased local inflammation involvingincreased activity of lamina propria den-dritic cells and macrophages, whichcould then primarily mediate the trans-location of bacteria as was measured. Insupport of this possibility, nitric oxiderelease by enterocytes may directly ac-tivate subepithelial dendritic cells ormacrophages, which could potentiatethese effects, and the broad use of theinducible nitric oxide synthase globalknockout mice, as well as the pharma-cologic reagents utilized, are entirelyconsistent with this possibility (12).

It is also worth mentioning that theabove findings have direct relevance to thesituation in which the intestinal barrier islargely intact, yet may have little impactunder conditions in which the intestinalbarrier is severely damaged (13). The gutbarrier is known to be broadly susceptibleto injury in the setting of nitrosative stress,leading to cell death via apoptosis resultingin broad gaps in the intestinal epithe-lium, which would allow for direct accessto the underlying dendritic cells andmacrophages for bacterial delivery tolymph nodes and the systemic circula-tion. The degree to which epithelial lossoccurs in the model presented by Wu andcolleagues, or in the clinical setting ofintestinal obstruction, still remains anopen question.

In summary, the current findings pro-vide important insights into how the in-testinal barrier can be transformed froman intact, protective shield to a semiper-meant membrane can occur in the set-ting of intestinal obstruction. The studyalso suggests the possibility that the localdelivery of nitric oxide regulators withinthe gut epithelium may modulate the ex-tent of bacterial translocation and poten-tially prevent septic complications fromoccurring in critically ill patients withintestinal obstruction. While furtherstudies are necessary to discern in greaterdetail the precise cellular site of action atwhich bacterial translocation occurs,these studies represent an important firststep in recognizing how danger on themucosal doorstep can be kept at bay, andthe ill effects that can transpire if suchbacteria are allowed passage into the po-

tentially damaging world of the subepi-thelial immune system.

David J. Hackam, MD, PhDDivision of Pediatric SurgeryChildren’s Hospital of

PittsburghDepartment of SurgeryUniversity of Pittsburgh

School of MedicinePittsburgh, PA

REFERENCES

1. Madl C, Druml W: Gastrointestinal disordersof the critically ill. Systemic consequences ofileus. Best Pract Res Clin Gastroenterol2003; 17:445–456

2. Deitch EA, Xu D, Kaise VL: Role of the gut inthe development of injury- and shock inducedSIRS and MODS: The gut-lymph hypothesis, areview. Front Biosci 2006; 11:520–528

3. Swank GM, Deitch EA: Role of the gut inmultiple organ failure: Bacterial transloca-tion and permeability changes. World J Surg1996; 20:411–417

4. Turner HL, Turner JR: Good fences makegood neighbors: Gastrointestinal mucosalstructure. Gut Microbes 2010; 1:22–29

5. Turner JR: Intestinal mucosal barrier func-tion in health and disease. Nat Rev Immunol2009; 9:799–809

6. Artis D: Epithelial-cell recognition of com-mensal bacteria and maintenance of immunehomeostasis in the gut. Nat Rev Immunol2008; 8:411–420

7. Neal MD, Leaphart C, Levy R, et al: Entero-cyte TLR4 mediates phagocytosis and trans-location of bacteria across the intestinal bar-rier. J Immunol 2006; 176:3070–3079

8. Shen L, Weber CR, Raleigh DR, et al: Tightjunction pore and leak pathways: A dynamicduo. Annu Rev Physiol 2011; 73:283–309

9. Han X, Fink MP, Yang R, et al: IncreasediNOS activity is essential for intestinal ep-ithelial tight junction dysfunction in endo-toxemic mice. Shock 2004; 21:261–270

10. Wu L-L; Chiu H-D; Peng W-H, et al: Epithelialinducible nitric oxide synthase causes bacterialtranslocation by impairment of enterocytictight junctions via intracellular signals of Rho-associated kinase and protein kinase C zeta.Crit Care Med 2011; 39:2087–2098

11. Su L, Shen L, Clayburgh DR, et al: Targetedepithelial tight junction dysfunction causesimmune activation and contributes to devel-opment of experimental colitis. Gastroenter-ology 2009; 136:551–563

12. Anand RJ, Dai S, Rippel C, et al: Activated mac-rophages inhibit enterocyte gap junctions via therelease of nitric oxide. Am J Physiol GastrointestLiver Physiol 2008; 294:G109–G119

13. Gribar SC, Richardson WM, Sodhi CP, et al:No longer an innocent bystander: Epithelialtoll-like receptor signaling in the develop-ment of mucosal inflammation. Mol Med2008; 14:645–659


Beneficial effects of carbamylated erythropoeitin on trauma-induced brain edema: Proposed molecular mechanisms of action*

Erythropoietin (EPO), in addi-tion to its known function inerythropoiesis, is an impor-tant mediator of the adaptive

response to metabolic stress. In responseto brain trauma, as is the case in thereport of Bouzat and colleagues (1) inthis issue of Critical Care Medicine,blood-brain barrier dysfunction ensuesdue to the breakdown of various endothe-lial junction proteins/mechanisms. Thisprocess allows for the passage of immunecells and water into the cerebral paren-chyma. This type of edema is generallyreferred to as vasogenic edema. Anothertype of brain edema, termed cytotoxic,occurs from dysfunctional energy-drivensodium/potassium pumps in glial cellmembranes, resulting in edematous-swollen astrocytes. It is conceivable thatboth types of edema exist in this model ofbrain trauma and that temporal eventsare characterized by an initial vasogeniccomponent followed by a cytotoxic effect.

Carbamylation (homocitrullination)of proteins is a nonenzymatic reaction ofthe cyanate ion with the free Epsilon NH2group of lysines. This process, in vivo,can be facilitated by the presence of my-eloperoxidase (2). Carbamylation of pro-teins is known to occur in vivo in uremicpatients through the formation of theurea-derived cyanate ion. Carbamylatederythropoietin (CEPO) is a nonerythro-poietic species/modification of EPO thatdoes not bind to the EPO receptor (3).The classic homodimeric EPO receptor isnot the transducer of the tissue-protec-tive effects of CEPO. It has been sug-

gested that at least one of the EPO recep-tor chains, CD131, complexed withothers is responsible for the neuro- andtissue-protective effects of CEPO (3, 4). Apositive correlation between the expres-sion of the EPO receptor in tissue and theprotective effects of EPO and CEPO areevident and support this notion.

CEPO was also found to modulate theSonic hedgehog (Shh) pathway. Wang etal (5) elegantly demonstrated that block-age of the Shh pathway with cyclopamineor specific siRNA abolished some of CEPObeneficial effects. In the brain, Shh bindsto the transmembrane receptor proteinPatched. Patched, in the absence of Shh,has an inhibitory effect on the transmem-brane receptor Smoothened. Shh bindingto Patched blocks this inhibitory effect.Activated Smoothened results in the ini-tiation of transcription factor GLI1 sig-naling in neural cells. This mechanismmight be involved in the edema/inflam-mation reduction reported by Bouzat etal (1). The decreased extracellular signal-regulated kinase 1/2 phosphorylation ob-served appears to be a downstream event,perhaps mediated by a CEPO-induced up-regulation of the Shh pathway.

Aquaporin 4 is a water channel proteinheavily expressed in the brain, particu-larly in astrocytes. Clearance of excesswater in the brain is possibly mediatedthrough this protein. Water uptake byastrocytes through aquaporin 4 is impor-tant in the mechanism of astrocyte waterpermeability and subsequent cell swelling(cytotoxic edema) encountered in isch-emia and trauma. EPO (and potentiallyCEPO) has been shown to antagonize theeffects of group I metabotropic glutamatereceptor agonists on water permeability(6). EPO and now CEPO may directlyreduce cytotoxic edema of astrocytes viaaquaporin 4 water transport.

This therapeutic window of treatingtrauma-induced brain edema with CEPOneeds further exploration into the molec-ular mechanisms as well as the temporal

relationships of acute and repeat admin-istration effects on brain edema and otherfunctional outcomes.

The important observations of re-duced brain edema and improved func-tional recovery in an animal model oftraumatic brain injury reported by Bou-zat (1) could lead to a novel clinical pathin the treatment of this deleterious con-dition. Conferring tissue protection with-out cross talk with the hematopoietic sys-tem and its associated serious adverseevents could be ideal for treating thiscondition.

David Bar-Or, MDGregory Thomas, BSC

Swedish Medical CenterSt. Anthony Central HospitalTrauma Research DepartmentEnglewood, CO

REFERENCES

1. Bouzat P, Francony S, Thomas S, et al: Re-duced brain edema and functional deficits af-ter treatment of diffuse traumatic brain injuryby carbamylated erythropoietin derivative.Critical Care Med 2011; 39:2099–2105

2. Wang Z, Nicholls SJ, Rodriguez ER, et al:Protein carbamylation links inflammation,smoking, uremia and atherogenesis. Nat Med2007; 13:1176–1184

3. Leist M, Ghezzi P, Grasso G, et al: Derivativesof erythropoietin that are tissue protective butnot erythropoietic. Science 2004; 305:239–242

4. Brines M, Grasso G, Fiordaliso F, et al: Eryth-ropoietin mediates tissue protection throughan erythropoietin and common beta-subunitheteroreceptor. Proc Natl Acad Sci U S A2004; 101:14907–14912

5. Wang L, Zhang ZG, Gregg SR, et al: The Sonichedgehog pathway mediates carbamylatederythropoietin-enhanced proliferation and dif-ferentiation of adult neural progenitor cells.J Biol Chem 2007; 282:32462–32470

6. Gunnarson E, Song Y, Kowalewski JM, et al:Erythropoietin modulation of astrocyte waterpermeability as a component of neuroprotec-tion. Proc Natl Acad Sci U S A 2009; 106:1602–1607

*See also p. 2099.Key Words: aquaporin 4; brain edema; carbamy-

lated erythropoietin; carbamylation; erythropoietin;sonic hedgehog; trauma



DOI: 10.1097/CCM.0b013e3182204a9c


Globalization of ejection fraction: The pulmonary artery catheterhas been outsourced*

I n an elegant study by Trepte andcolleagues (1) in this issue of Crit-ical Care Medicine, we discoverhow the less invasive transpulmo-

nary measurement of cardiac physiologycompares to pulmonary arterial catheter-ization in a variety of shock states (1). Astheir reference measurement, they offerventricular pressure-time measures, longconsidered the gold standard for cardiacfunction. Due to survival concerns, theychose to measure the rate of rise in leftventricular pressure at a single preload.While this clearly limits the reader’s abil-ity to infer true cardiac contractility, itdoes reflect clinical practice in that theclinician must make decisions at a singlepoint, which in the final analysis mayactually strengthen this study’s utility.Estimating preload, global ejection frac-tion (GEF), and more controversially, ex-travascular lung water using peripheralarterial sensing of a centrally injectedmolecular tracer or thermal dilution hasgreat intuitive appeal given the ubiqui-tous nature of central venous and arterialcatheters in shock states (2–5). The keyquestions confronting clinicians are: cantranspulmonary data yield informationequivalent to that derived from pulmo-nary arterial catheterization while avoid-ing its small but significant increase inmorbidity (6), and secondly, in whom is itworth measuring?

It has long been appreciated that indi-ces of cardiac function derived from ther-modilution are inexact. Statistically, ifone uses a single measure of cardiac out-put, the minimum detectable true differ-ence must be at least 22% (3). This study(1) provides us valuable information in

that we discover that neither indices ofright heart function nor transpulmonarymeasurements correlate particularly wellwith the direct measurement of the rateof rise in left ventricular pressure. This isnot to say that transpulmonary GEF andcardiac function index (CFI) are unable todetect impaired cardiac function, butrather that they do so much more reliablyin very low output states. This is evi-denced by an increase by 40% in thecorrelation between the rate of rise in leftventricular pressure and GEF/CFI whencardiac output was highly suppressed bycalcium channel blocker infusion com-pared to normal cardiac function (Fig. 3,B and D in [1]). According to this study,the main utility of transpulmonary mea-sures would appear to be the identifica-tion of those with severely compromisedcardiac output (corresponding to a car-diac index of well under 2.0 L/min/m2).Given its performance characteristics asoutlined above, GEF/CFI should alsoserve to identify a large subsequent im-provement in output. Once in the normalrange, GEF and CFI appear rather insen-sitive to changes in cardiac function.

In whom might measurement of GEFand CFI be useful? This study (1), alongwith recent observational data (7, 8), pro-vides us some guidance. Most patientswith septic shock, regardless of their pre-senting left ventricular ejection fraction,experience an augmentation of cardiacoutput following fluid resuscitation andtreatment with noradrenaline (8). Thereare however a subset of patients in whomcardiac output declines significantly de-spite normalization of blood pressure (8).These patients are at great risk of unrec-ognized shock and subsequent organ fail-ure. Routine clinical use of a screeningGEF/CFI following stabilization of bloodpressure would serve both to identifythese patients and to guide subsequenttitration of inotropes. While the samerole could be envisioned for the pulmo-nary arterial catheter, the added morbid-ity (6) and physician time spent upon thisprocedure could only be justified if it is

obviously superior to transpulmonarymeasures. This study provides good evi-dence that this is not the case, and addspressure to continue the clinical shiftaway from pulmonary artery catheters.

John H. Boyd, MDCritical Care Research

LaboratoriesHeart � Lung InstituteSt. Paul’s HospitalUniversity of British ColumbiaVancouverBC, Canada

REFERENCES

1. Trepte CJ, Eichhorn V, Haas SA, et al: Ther-modilution-derived indices for assessment ofleft and right ventricular cardiac function innormal and impaired cardiac function. CritCare Med 2011; 39:2106–2112

2. Stetz CW, Miller RG, Kelly GE, et al: Reliabil-ity of the thermodilution method in the de-termination of cardiac output in clinical prac-tice. Am Rev Respir Dis 1982; 126:1001–1004

3. Fernandez-Mondejar E, Castano-Perez J, Ri-vera-Fernandez R, et al: Quantification oflung water by transpulmonary thermodilu-tion in normal and edematous lung. J CritCare 2003; 18:253–258

4. Combes A, Berneau JB, Luyt CE, et al: Esti-mation of left ventricular systolic function bysingle transpulmonary thermodilution. Inten-sive Care Med 2004; 30:1377–1383

5. Fernandez-Mondejar E, Rivera-Fernandez R,García-Delgado M, et al: Small increases inextravascular lung water are accurately de-tected by transpulmonary thermodilution.J Trauma 2005; 59:1420 –1423; discussion1424

6. Sandham JD, Hull RD, Brant RF, et al: Arandomized, controlled trial of the use of pul-monary-artery catheters in high-risk surgicalpatients. N Engl J Med 2003; 348:5–14

7. Ritter S, Rudiger A, Maggiorini M: Transpul-monary thermodilution-derived cardiac func-tion index identifies cardiac dysfunction inacute heart failure and septic patients: An ob-servational study. Crit Care 2009; 13:R133

8. Hamzaoui O, Georger JF, Monnet X, et al:Early administration of norepinephrine in-creases cardiac preload and cardiac output inseptic patients with life-threatening hypoten-sion. Crit Care 2010; 14:R142

*See also p. 2106.Key Words: cardiac function index; global ejection

fraction; heart failure; hemodynamics; pulmonary ar-tery catheter; shock; transpulmonary cardiac output



DOI: 10.1097/CCM.0b013e31821f02d4


Pseudomonas aeruginosa: So many virulence factors, so little time*

Despite 60 yrs of intensive an-timicrobial development,bacterial infections remainan important cause of mor-

bidity and mortality. The ever-increasingburden of resistance erodes the efficacy ofconventional antibiotics, while criticallyill patients frequently fail to respond toappropriate agents, even when infectedwith susceptible organisms. To improveupon the current situation, attempts arebeing made to leverage our impressiveknowledge of bacterial pathogenesis fordevelopment of novel therapeutics thatblock virulence mechanisms rather thansimply kill or inhibit growth of bacteria.The thought is that such agents could beused by themselves or as adjuncts thatwould boost the efficacy of conventionalantibiotics. But which virulence determi-nants should be targeted? A factor that ishighly critical to the organism’s patho-genesis, of course. In some cases, thechoice is straightforward. For example,tetanus toxin is the major virulence de-terminant of Clostridium tetani, and ad-ministration of human tetanus immuneglobulin attenuates the course of tetanus(1). In other cases, however, the choice isless clear. The problem is compounded bythe current emphasis on reductionist ap-proaches in microbiology. To test the im-portance of a particular bacterial factor inpathogenesis, scientists usually disruptthe gene encoding the factor in a well-characterized laboratory strain, and thevirulence of the mutant is compared tothat of the parental strain in an estab-lished animal model of infection. Thispowerful approach has the benefit ofdemonstrating a causal relationship be-

tween the factor and disease in the modelused, and has become the gold standardfor identification of virulence factors (2).However, it leaves unanswered two im-portant questions: Is the factor criticalfor virulence in most clinical isolates, andhow important is the factor relative toother virulence determinants made bythe same bacterium? In this issue of Crit-ical Care Medicine, Le Berre and col-leagues (3) address these questions in re-lation to Pseudomonas aeruginosa.

P. aeruginosa is an opportunisticpathogen familiar to every hospitalist andintensivist. It is a frequent cause of nos-ocomial infections, many of which areresponsible for substantial mortality evenwhen treated with appropriate antibiot-ics. Thus, novel agents that target viru-lence determinants would be a welcomeaddition to our antipseudomonal arma-mentarium. But which virulence deter-minants should be targeted? P. aerugi-nosa produces a potpourri of pathogenicfactors, any one of which might or mightnot be a worthy target for inhibition. LeBerre and colleagues focused on threeextensively studied virulence determi-nants: quorum-sensing (QS), the type IIIsecretion (TTS) system, and lipopolysac-charide (LPS) O-antigen. QS allows bac-teria to respond to their own populationdensity by coordinately regulating theirgene expression patterns (4). For exam-ple, in the presence of high bacterialnumbers, QS signaling leads to produc-tion of two P. aeruginosa toxins, elastaseand pyocyanin. TTS is a mechanism bywhich bacteria inject a set of toxins di-rectly into the cytosol of host cells. TwoP. aeruginosa proteins secreted by thismechanism are ExoU and ExoS; bothhave been previously linked to virulence(5). LPS O-antigen is a variable polysac-charide that decorates the outer surfacesof many bacteria and protects againstcomplement-mediated lysis (6); differ-ences in O-antigens are the basis for se-rotyping P. aeruginosa strains. Each ofthese three factors has been studied ex-tensively and conclusively shown to bepathogenic in animal models of infection.Importantly, P. aeruginosa strains also

differ in their production of each of thesefactors.

Le Berre and colleagues characterizedthe QS, TTS, and LPS O-antigen proper-ties of 56 nonclonal P. aeruginosa iso-lates from patients with ventilator-associated pneumonia. The severity ofpneumonia caused by each isolate wasquantified by measuring alveolar-capil-lary barrier permeability, lung wet/dryweight ratios, and bacterial dissemina-tion from the lungs in a mouse model ofpneumonia. The QS, TTS, and LPS O-an-tigen properties of each strain were thencompared to the severity of pneumonia itcaused. Univariate analysis indicated anassociation between lung injury and elas-tase production (and by inference QS),O11 O-antigen serotype, and TTS-positivephenotype. However, it had been shownpreviously that QS, TTS, and LPS O-an-tigen type are not independent variables.For example, O11 serotype strains morecommonly secrete the potent TTS toxinExoU (7), and QS regulates TTS (8). Toaccount for this, a multivariate analysis wasperformed, which showed that TTS (espe-cially secretion of ExoU) was most highly cor-related with lung injury and that secretion ofelastase was associated to a lesser degree.

The approach used by Le Berre andcolleagues has several limitations. As ac-knowledged by the authors, the produc-tion of virulence factors in vitro may notmimic expression in vivo, and virulencemeasurements at a single time point maynot adequately capture the true extent ofdisease. The measurement of QS was in-direct, using elastase and pyocyanin se-cretion as surrogate markers. Also, al-though inbred mice offer the advantageof a uniform host model, their vulnera-bility to particular virulence determi-nants may not reflect that of humans. Inthis regard, it is reassuring that severalstudies have also found associations be-tween P. aeruginosa TTS and poor out-comes in humans with acute respiratoryinfections (9–11). Perhaps the most sig-nificant limitation is that this study doesnot show a causal relationship betweensevere disease and TTS or QS, but only anassociation. However, when viewed to-gether with previously published reports

*See also p. 2113.Key Words: elastase; ExoU; O-antigen; Pseudomo-

nas aeruginosa; quorum sensing; type III secretion;virulence

This work was supported, in part, by grantsRO1AI053674 and RO1AI075191 from the NationalInstitutes of Health.

The author has an ongoing consulting agreementwith Microbiotix.


DOI: 10.1097/CCM.0b013e318221742d


describing conventional mutational ap-proaches, the paper by Le Berre and col-leagues provides a compelling argumentin favor of an important role for these twofactors in acute pneumonia.

Studies such as this one are especiallyrelevant given the current emphasis ontranslational research and the develop-ment of novel therapeutic agents. Since$800 million is required to bring a newdrug to market, it is essential that duediligence is performed to maximize thechances of success before costly clinicaltrials are begun. In this regard, the studyby Le Berre and colleagues bodes well forongoing preclinical and clinical trials ex-amining the efficacy of inhibitors of TTSand QS in infections caused by P. aerugi-nosa (12–14). When possible, this ap-proach should be considered as well forother virulence factors that are candidatedrug targets (15, 16).

Alan R. Hauser, MD, PhDDepartments of Microbiology/

Immunology and MedicineNorthwestern UniversityChicago, IL

REFERENCES1. Blake PA, Feldman RA, Buchanan TM, et al:

Serologic therapy of tetanus in the UnitedStates, 1965–1971. JAMA 1976; 235:42–44

2. Falkow S: Molecular Koch’s postulates ap-plied to microbial pathogenicity. Rev InfectDis 1988; 10:S274–S276

3. Le Berre R, Nguyen S, Nowak E, et al:Relative contribution of three main viru-lence factors in Pseudomonas aeruginosapneumonia. Crit Care Med 2011; 39:2113–2120

4. Juhas M, Eberl L, Tummler B: Quorum sens-ing: The power of cooperation in the world ofPseudomonas. Environ Microbiol 2005;7:459–471

5. Hauser AR: The type III secretion system ofPseudomonas aeruginosa: Infection by in-jection. Nat Rev Microbiol 2009; 7:654 –665

6. Pier GB: Pseudomonas aeruginosa lipopoly-saccharide: A major virulence factor, initia-tor of inflammation and target for effectiveimmunity. Int J Med Microbiol 2007; 297:277–295

7. Faure K, Shimabukuro D, Ajayi T, et al: O-antigen serotypes and type III secretory tox-ins in clinical isolates of Pseudomonasaeruginosa. J Clin Microbiol 2003; 41:2158–2160

8. Bleves S, Soscia C, Nogueira-Orlandi P, et al:Quorum sensing negatively controls type IIIsecretion regulon expression in Pseudomo-nas aeruginosa PAO1. J Bacteriol 2005; 187:3898–3902

9. El Solh AA, Akinnusi ME, Wiener-KronishJP, et al: Persistent infection with Pseudomo-nas aeruginosa in ventilator-associated

pneumonia. Am J Respir Crit Care Med 2008;178:513–519

10. Hauser AR, Cobb E, Bodi M, et al: Type IIIprotein secretion is associated with poor clini-cal outcomes in patients with ventilator-associated pneumonia caused by Pseudomonasaeruginosa. Crit Care Med 2002; 30:521–528

11. Roy-Burman A, Savel RH, Racine S, et al:Type III protein secretion is associated withdeath in lower respiratory and systemicPseudomonas aeruginosa infections. J InfectDis 2001; 183:1767–1774

12. Frank DW, Vallis A, Wiener-Kronish JP, et al:Generation and characterization of a protec-tive monoclonal antibody to Pseudomonasaeruginosa PcrV. J Infect Dis 2002; 186:64–73

13. Hentzer M, Wu H, Andersen JB, et al: Atten-uation of Pseudomonas aeruginosa virulenceby quorum sensing inhibitors. EMBO J 2003;22:3803–3815

14. Aiello D, Williams JD, Majgier-BaranowskaH, et al: Discovery and characterization ofinhibitors of Pseudomonas aeruginosa typeIII secretion. Antimicrob Agents Chemother2010; 54:1988–1999

15. Veesenmeyer JL, Hauser AR, Lisboa T, et al:Pseudomonas aeruginosa virulence andtherapy: Evolving translational strategies.Crit Care Med 2009; 37:1777–1786

16. Kipnis E, Sawa T, Wiener-Kronish J: Tar-geting mechanisms of Pseudomonasaeruginosa pathogenesis. Med Mal Infect2006; 36:78 –91

How alcohol impairs the granulocyte expansion during septicemia*

Epidemiologic studies indicatethat chronic alcohol abuse isassociated with high morbid-ity and mortality in critically

ill patients, particularly those with sepsisand septic shock (1, 2). Chronic alcoholabuse also increases the risk of develop-ing several acute insults that may lead toacute respiratory distress syndrome (1,2). Thus, many experimental and clinicalstudies have been conducted to elucidatethe mechanisms underlying the detri-mental impact of chronic alcohol abuse

on patients with sepsis (1, 2). For exam-ple, excessive alcohol consumption in-duces the dysfunction of epithelial barri-ers of the airway or gut, disrupts theantioxidant system, and impairs immu-nologic function, leading to both thepredisposition to bacterial infection andincreased severity of endotoxemia (1–4). During bacterial infection, granulo-poiesis in bone marrow is enhanced toincrease the number of circulatinggranulocytes for host defense (5). Alco-hol is known to have direct toxic effectson bone marrow, resulting in granulo-cytopenia in patients with alcohol abuse(6, 7). Currently, the exact molecularmechanisms underlying the disability ofgranulocyte expansion in patients withalcohol abuse are incompletely under-stood.

Hematopoietic stem cells are broadlyclassified into lymphoid hematopoietic

stem cells and myeloid hematopoieticstem cells, both of which have the capac-ity of self-renewal and the potential todifferentiate into in a variety of matureblood cells in bone marrow (8). Undernormal conditions, hematopoiesis inbone marrow is tightly controlled by acomplex mechanism to regulate the turn-over of different blood cell lineages. Uponbacterial infection, however, the homeo-stasis of hematopoiesis is altered and pre-dominantly shifted toward the expansionof granulocyte lineage, which serves asthe first line of phagocytic defense forinvading bacteria (9). In the regulation ofhematopoiesis, stem cell antigen-1(Sca-1) is a phosphatidylinositol-an-chored glycoprotein of the lymphocyteactivation protein-6 gene family found onthe surface of several murine marrowstem cell subtypes and it serves as animportant regulator for the proliferation

*See also p. 2121.Key Words: alcoholism; granulopoiesis; hemato-

poietic stem cell; stem cell antigen-1The authors have not disclosed any potential con-



DOI: 10.1097/CCM.0b013e318221740b


of early hematopoietic precursors (10,11). In mice, bone marrow lineage (lin)–c-kit�Sca-1� cells are highly purifiedhematopoietic stem cells (12). Uponstimulation, lin–c-kit�Sca-1– cells rap-idly transform into lin– c-kit�Sca-1�cells that contain more committed my-eloid progenitors, such as granulocyteprogenitors (12).

In this issue of Critical Care Medicine,Melvan and colleagues (13) investigatedthe involvement of Sca-1 expression ofgranulocyte lineage-committed progeni-tors in the granulopoietic response tosepticemia, and how acute alcohol treat-ment affected this response in a mousemodel with Escherichia coli insult. Twoprevious studies from this group (14, 15)reported that enhanced Sca-1 expressionpromoted the conversion of bone marrowlin–c-kit�Sca-1– cells to lin-c-kit�Sca-1�cells, which play an essential role in di-recting lineage commitment of primitiveprecursors toward myeloid lineage de-velopment in mice during septicemia.Additionally, acute alcohol intoxicationretarded this initial step of the granulo-poietic response to the challenge with E.coli (15). This elegantly designed investi-gation (13) appears to be a logical exten-sion of these two studies (14, 15), andfocuses on the functional significance ofenhanced Sca-1 expression in down-stream granulocyte lineage-committedprogenitor cells. Several important find-ings were reported. Firstly, they em-ployed an in vivo model to demonstratethat the level of Sca-1 expression of im-mature granulocyte differentiation anti-gen-1 precursors was highly linked to theexpansion of granulopoietic precursorsduring septicemia, suggesting that thesetwo events might have a cause-effect re-lationship. Indeed, they found that alco-hol treatment was able to downregulate theresponse of Sca-1 expression, maturation ofgranulocyte differentiation antigen-1 pre-cursors, and mobilization of maturegranulocytes from bone marrow into thecirculation. Secondly, it is known thatthe function of certain granulopoieticmediators, such as granulocyte colony-stimulating factor, play a crucial role inthe maturation and mobilization of bonemarrow granulocytes upon bacterial in-

fection (9). They showed that alcoholtreatment decreased the activity of gran-ulocyte colony-stimulating factor elicitedby the E. coli challenge in the Sca-1�granulocyte differentiation antigen-1 pre-cursors in mice, a result that is in linewith their finding regarding the suppres-sive effect of alcohol on granulocyte ex-pansion. Thirdly, they employed an invitro model to reveal that exposure tolipopolysaccharide could upregulate theexpression of Sca-1 in granulocyte differ-entiation antigen-1�Sca-1– cells via theactivation of c-Jun N-terminal kinase, acrucial signaling pathway for severalLPS-induced responses. Importantly, al-cohol treatment inhibited the activationof this signaling pathway. Lastly, theyfound that Sca-1-deficient mice in factexhibited a defect in the expansion ofgranulopoietic precursors in response toE. coli challenge, a result that was similarto the situation in alcohol-treated ani-mals. Thus, their in vivo and in vitro dataprovide solid evidence suggesting that al-cohol treatment attenuates the Sca-1 re-sponse in granulocyte precursors leadingto a restriction of granulocyte expansionduring septicemia.

Although data of this investigation(13) were obtained from a murine modelwith acute alcohol treatment, the find-ings may serve as one of the plausiblemechanisms to explain the impaired hostdefense in alcohol abusers. Whether thesame concept applies to an experimentalmodel with chronic alcohol ingestion (3)warrants investigation. Granulocytes arean important player for innate immuneresponses in critically ill patients withvarious etiologies. Particularly, granulo-cytopenia is an indicator of bad prognosisfor these patients. Since alcohol abuseproduces many complications in thesepatients (1, 2), the findings of this studymay also provide important informationfor potential therapeutic targets to en-hance host defense in critically ill pa-tients with alcohol abuse.

Tzong-Shyuan Lee, DVM, PhDYu Ru Kou, PhD

Department of PhysiologySchool of MedicineNational Yang-Ming UniversityTaipei, Taiwan

REFERENCES

1. Joshi PC, Guidot DM: The alcoholic lung:Epidemiology, pathophysiology, and poten-tial therapies. Am J Physiol Lung Cell MolPhysiol 2007; 292:L813–L823

2. de Wit M, Jones DG, Sessler CN, et al: Alco-hol-use disorders in the critically ill patient.Chest 2010; 138:994–1003

3. Guidot DM, Roman J: Chronic ethanol inges-tion increases susceptibility to acute lunginjury: Role of oxidative stress and tissueremodeling. Chest 2002; 122:309S–314S

4. Keshavarzian A, Farhadi A, Forsyth CB, et al:Evidence that chronic alcohol exposure pro-motes intestinal oxidative stress, intestinalhyperpermeability and endotoxemia prior todevelopment of alcoholic steatohepatitis inrats. J Hepatol 2009; 50:538–547

5. Ueda Y, Kondo M, Kelsoe G: Inflammationand the reciprocal production of granulo-cytes and lymphocytes in bone marrow.J Exp Med 2005; 201:1771–1780

6. Michot F, Gut J: Alcohol-induced bone mar-row damage. A bone marrow study in alco-hol-dependent individuals. Acta Haematol1987; 78:252–257

7. Ballard HS: Hematological complications ofalcoholism. Alcohol Clin Exp Res 1989; 13:706–720

8. Cronkite EP: Analytical review of structureand regulation of hemopoiesis. Blood Cells1988; 14:313–328

9. Terashima T, Wiggs B, English D, et al: Poly-morphonuclear leukocyte transit times inbone marrow during streptococcal pneumo-nia. Am J Physiol 1996; 271:L587–L592

10. Holmes C, Stanford WL: Concise review:Stem cell antigen-1: Expression, function,and enigma. Stem Cells 2007; 25:1339–1347

11. Trumpp A, Essers M, Wilson A: Awakeningdormant haematopoietic stem cells. Nat RevImmunol 2010; 10:201–209

12. Okada S, Nakauchi H, Nagayoshi K, et al: Invivo and in vitro stem cell function of c-kit-and Sca-1-positive murine hematopoieticcells. Blood 1992; 80:3044–3050

13. Melvan JN, Siggins RW, Bagby GJ, et al:Suppression of stem cell antigen-1 responseand granulocyte lineage expansion by alcoholduring septicemia. Crit Care Med 2011; 39:2121–2130

14. Zhang P, Nelson S, Bagby GJ, et al: Thelineage-c-Kit�Sca-1� cell response to Esch-erichia coli bacteremia in Balb/c mice. Stemcells 2008; 26:1778–1786

15. Zhang P, Welsh DA, Siggins RW 2nd, et al:Acute alcohol intoxication inhibits the lin-eage- c-kit� Sca-1� cell response to Esche-richia coli bacteremia. J Immunol 2009; 182:1568–1576


Plumbing the depths of blood pressure: Hypotensive hemorrhageand acute kidney injury*

An abrupt decline in kidneyfunction, now termed acutekidney injury (AKI), occurscommonly in hospitalized pa-

tients (1, 2). Even mild forms of AKI areindependently associated with significantmorbidity and mortality, and more severeforms are associated with mortality inexcess of 50% (1, 2). Despite advances inthe care of seriously ill patients, out-comes associated with AKI have not sig-nificantly improved over several decades.

AKI is usually diagnosed when eitheran increase in serum creatinine concen-tration (which is an insensitive and time-delayed marker for AKI) or a decrease inurine output (which often does not ac-company AKI) is observed (3, 4). Thus,diagnosis of AKI is usually delayed untilafter a substantial reduction in kidneyfunction, and perhaps nonreversible re-nal cellular damage, has been initiated.Also, despite many studies, there is cur-rently no specific treatment that has beenshown to either attenuate the severity orspeed recovery from AKI. The complexand dynamic nature of the clinical settingin which clinical AKI is encountered of-ten renders conduct and interpretation ofclinical studies difficult. Finally, manyof the animal models developed for studyof AKI may not be relevant (5). Collec-tively, these observations demonstrate aneed for better models with which tostudy AKI, potentially leading to earlierdiagnosis and tailored interventions thatattenuate AKI.

Renal ischemia occurs in several set-tings, including hemorrhagic hypoten-sion, and is one of the most commonfactors predisposing to AKI. Pioneeringwhole-animal work by Henrich et al (6,7), published more than 3 decades ago,

delineated a delicate interplay betweenneurohumoral factors to acutely regulaterenal blood flow and glomerular filtrationrate in response to hemorrhage. Removalof vasodilatory eicosanoids (which en-hance renal blood flow) and angiotensinII (which increases glomerular capillaryhydrostatic pressure by constricting theefferent arteriole), and the presence ofintact renal adrenergic neural pathways(which decrease renal blood flow), mark-edly enhance the effect of hemorrhage toinduce renal ischemia and decrease glo-merular filtration rate. These observa-tions provided a physiologic explanationfor the now well-known effects of non-steroidal anti-inflammatory drugs anddrugs that impair angiotensin II forma-tion and action to induce AKI in hemo-dynamically compromised states.

In this issue of Critical Care Medicine,Mayeur and colleagues (8) extend our in-sights into kidney responses to ischemiaby providing a comprehensive short- andlonger-term evaluation of the functionaland morphologic response of the mousekidney to hemorrhagic hypotension. Thisstudy is of interest because AKI occurringin the context of hypotension may moreclosely mimic clinical AKI than widelyused models, such as warm ischemia-reperfusion induced by short-term renalartery occlusion. It is likely that the mul-tiorgan hypoperfusion that occurs withhypotension induces systemic responsesthat contribute to altered function of sev-eral organs, including the kidney. Also,the multiple genetic alterations that arepossible to induce in mice provide apotentially rich opportunity to analyzethe pathogenetic role of several path-ways in AKI.

Mayeur et al demonstrate a time-dependent effect of controlled hypoten-sion to induce increasing degrees of AKI.The AKI is characterized by decreasedglomerular filtration rate, impaired tubu-lar resorption of sodium and water, andthe presence of tubular lesions that aremost marked in the outer medulla andcortex. Mayeur et al also find an increase

in kidney injury molecule-1 and hypoxia-inducible factor-1� mRNA, beginningshortly after induction of hypotensionand persisting for up to 6 days. The renalfunctional and histologic abnormalitiesare maximal 2–6 days after the inductionof hypotension and much less prominent21 days later. However, impaired renaltubular sodium resorption and mild peri-tubular fibrosis are observed 21 days fol-lowing induction of AKI. The finding ofmild fibrosis several days after develop-ment of AKI is of note with regard toincreasing reports of lack of complete re-covery from AKI in selected cases (9).Collectively, these results suggest thatthe model described by Mayeur and oth-ers share many characteristics with clin-ical AKI.

The hope is that with further study,relevant animal models, such as thatcharacterized by Mayeur et al, will beutilized to provide pathogenetic insightsleading to prevention, earlier diagnosis,and attenuation of AKI. While awaitingreal-time methods to diagnose and morespecific pathophysiologic-based methodsto attenuate AKI, the prudent clinicianwill use tested prophylactic methods toprotect the kidneys in response to expo-sure to potential nephrotoxins, and willidentify patients at high risk for develop-ment of AKI and monitor renal functionclosely in these patients (1, 2).

Robert J. Anderson, MDClinical Professor of MedicineDepartment of MedicineNew York University Medical

CenterNew York, NY

REFERENCES

1. Waiker SS, Liu KD, Chertow GM: Diagnosis,epidemiology and outcomes of acute kidneyinjury. Clin J Am Soc Nephrol 2008;3:844–861

2. Dennen P, Douglas IS, Anderson R: Acutekidney injury in the intensive care unit: Anupdate and primer for the intensivist. CritCare Med 2010; 38:261–275

3. Ricci Z, Cruz DN, Ronco C: Classification andstaging of acute kidney injury: Beyond the

*See also p. 2131.Key Words: acute kidney injury; hemorrhage; hem-

orrhagic shock; hypotensive hemorrhageThe author has not disclosed any potential con-



DOI: 10.1097/CCM.0b013e3182217465


RIFLE and AKIN criteria. Nat Rev Nephrol2011; 7:201–208

4. Kellum JA: Acute kidney injury. Crit CareMed 2008; 36:S141–S145

5. Lieberthal W, Nigam SK: Acute renalfailure. II. Experimental models of acute renalfailure: Imperfect but indispensable. Am JPhysiol Renal Physiol 2000; 278:F1–F12

6. Henrich WL, Anderson RJ, Berns AS, et al: The

role of renal nerves and prostaglandins in con-trol of renal hemodynamics and plasma reninactivity during hypotensive hemorrhage in thedog. J Clin Invest 1978; 61:744–750

7. Henrich WL, Berl T, McDonald KM, et al:Angiotensin II, renal nerves, and prostaglan-dins in renal hemodynamics during hemor-rhage. Am J Physiol 1978; 235:F46–F51

8. Mayeur N, Minville V, Jaafar A, et al: Morpho-

logic and functional renal impact of acutekidney injury after prolonged hemorrhagicshock in mice. Crit Care Med 2011; 39:2131–2138

9. Joanidis M, Philipp GH: Long-term outcomeof acute kidney injury. In: Management ofAcute Kidney Problems, Jorres A, Ronco C,Kellum JA Jr (Eds). Springer-Verlag BerlinHeidelburg, 2010, pp 260–278

Pediatric brain death*

T he history of brain death datesto the mid-1950s. Althoughnot using the term, PopePius XII when asked to opine

on the timing of death said in an ad-dress to an International Congress ofAnesthesiologists on November 24,1957, “[the timing of death] does notfall within the competence of theChurch … But considerations of a gen-eral nature allow us to believe that hu-man life continues for as long as itsvital functions— distinguished from thesimple life of organs—manifest them-selves spontaneously or even with thehelp of artificial processes” (1). Electro-encephalographers in France in the late1950s recognized “coma depasse” or“beyond coma”, which described a statenow consistent with brain death (2). A1964 report from the United States sug-gested the electroencephalogram beused to declare death (3), but it was theHarvard Criteria for “irreversiblecoma,” published in 1968, in which theline in the sand was drawn formallydemarcating the concept of a neurologic-based definition of death to supplementcardiorespiratory death (4). That thisera also saw the introduction of solidorgan transplants is not coincidental(5, 6).

In the work after publication of theHarvard Criteria and then incorporatedby experts from the 1981 President’sCommission (7), refinements were notadded to the conceptual definition ofbrain death (still fundamentally whole

brain death [8]), but substantialchanges were made to the operationaldefinition (e.g., more precision in theneurologic reflex examination and theapnea test requirements). With no ad-ditional data, a pediatric version of thebrain death guidelines was published in1987 (9). It is reassuring the currentguidelines change only procedural as-pects of the declaration of brain death,maintaining the conceptual and themost central aspects of the operationaldefinition (10).

Although variability in the declarationof death led to a memorable quote in an1844 story by Edgar Allen Poe (“Theboundaries between life and death are atbest shadowy and vague. Who shall saywhere one ends and where the other be-gins?”), variability in the criteria used forthe declaration of death is unacceptablein 2011. The guideline authors referencea number of studies that demonstratethis point, but the magnitude of the prob-lem must be re-emphasized. In 1995, Me-jia and Pollack (11) reported that no ap-nea testing was performed in 25% ofbrain death declarations and incompleteapnea testing practices were documentedin another 22%. A small, but real, per-centage of patients had brain death eval-uations within hours of discontinuingbarbiturate infusions. In 2003, Chang etal (12) reported a small but telling surveyin which four of 10 centers and none ofthe eight surveyed neurosurgeons an-swered “yes” to the question “Do youfollow guidelines/waiting periods ex-actly?” Finally, Mathur et al (13) reportedthe mean number of the requisite 14 ex-amination elements for 142 brain deadchildren at well �50% by neurologists,pediatric intensivists, and neurosur-geons. No apnea testing was recorded in60% of brain death confirmations. Whilepublication of consensus guidelines for

determining brain death is of great in-terest to clinicians and also has signif-icant applicability, it should be ex-pected that the revised guidelines willimprove this variability. To be purpose-fully redundant, in 2011, variability inthe declaration of brain death in chil-dren must be obliterated.

The inclusion of a checklist in therevised guidelines is a great concept andshould have widespread appeal and appli-cability (10). Similarly, clearly definedcomponents of the neurologic examina-tion for neonates, infants, and children,including a standardized procedure forapnea testing, will be of significant helpto clinicians in assuring consistent, in-clusive, and reproducible results whenperforming the examinations both be-tween physicians and over time. Giventhe complexity of the different parts of apediatric neurologic examination, the po-tential for ancillary testing, and the is-sues related to repeated testing and as-sessments, the potential for variabilitysimply due to diagnostic errors is real.Many high-risk, high-reliability profes-sions, such as airline pilots and nuclearplant operators, have reduced errors byusing checklists (14). Checklists aregaining increasing acceptance and usein medicine, especially in complex areassuch as operating rooms and intensivecare units (14). The use of a standard-ized checklist for determination of pe-diatric brain death may assist the phy-sician to optimize their cognitiveapproach, avoid diagnostic errors, andimprove evaluation of the neurologicinjury. While most checklists have notbeen subjected to rigorous evaluation,this type of organized approach may beanalogous to the implementation ofsepsis bundles in the care of patientswith septic shock to improve outcomes(15) or a multidisciplinary daily quality

*See also p. 2139.Key Words: brain death; checklist; pediatrics; registryThe authors have not disclosed any potential con-



DOI: 10.1097/CCM.0b013e3182226fc7


checklist that was shown to correlatewith decreased infection rates in atrauma intensive care unit (16). Westrongly suggest that this checklist beincorporated into the patient’s medicalrecord as it will guide clinicians duringa high-stress period and provide defin-itive documentation of the specificsteps and timeline followed for determi-nation and declaration of brain deathfor clinical and medical legal purposes.

Finally, we support the authors’ sug-gestions for the future directions that re-search and policy in this field should take(10). In particular, development of a na-tional database to track pediatric patientswho are diagnosed as brain dead couldprovide a valuable source of data forfuture investigations and queries thatmultiple users could take advantageof— clinicians, researchers, transplantorganizations, and state and federal fund-ing health agencies, to name a few. Sucha repository of data could also be re-viewed in a timely fashion for future re-visions of the pediatric guidelines forbrain death.

The Task Force, the Society of CriticalCare Medicine (Mount Prospect, IL), andthe American Academy of Pediatrics (ElkGrove Village, IL) should be congratu-lated on compiling such a timely andcomprehensive review of the 1987 guide-

lines and the evidence review of the med-ical literature.

James Fackler, MDAnesthesiology and Critical

Care MedicineThe Johns Hopkins University

School of MedicineBaltimore, MD

Brahm Goldstein, MD, MCRClinical DevelopmentIkaria, Inc.Clinton, NJ

REFERENCES

1. Pope Pius XII: To an International Congressof Anesthesiologists, Nov. 24, 1957. PopeSpeaks 1958; 4:393–398

2. Mollaret P, Bertrand I, Mollaret H: Comadepasse et necroses nerveuses centrales mas-sives. Rev Neurol 1959; 101:116–139

3. Hamlin H: Life or death by EEG. JAMA 1964;190:112–114

4. A definition of irreversible coma. Report ofthe Ad Hoc Committee of the Harvard Med-ical School to Examine the Definition ofBrain Death. JAMA 1968; 205:337–340

5. Giacomini M: A change of heart and a changeof mind? Technology and the redefinition ofdeath in 1968. Soc Sci Med 1997; 44:1465–1482

6. Spoor MT, Sutherland FR: The evolution ofthe concept of brain death. Ann R Coll Phy-sicians Surg Can 1995; 28:30–34

7. Guidelines for the determination of death.Report of the medical consultants on the

diagnosis of death to the President’s Com-mission for the Study of Ethical Problems inMedicine and Biomedical and Behavioral Re-search. JAMA 1981; 246:2184–2186

8. Truog RD, Fackler JC: Rethinking braindeath. Crit Care Med 1992; 20:1705–1713

9. Report of special Task Force. Guidelines forthe determination of brain death in children.American Academy of Pediatrics Task Forceon Brain Death in Children. Pediatrics 1987;80:298–300

10. Nakagawa TA, Ashwal S, Mathur M, et al:Guidelines for the determination of braindeath in infants and children: An update ofthe 1987 Task Force recommendations. CritCare Med 2011; 39:2139–2155

11. Mejia RE, Pollack MM: Variability in braindeath determination practices in children.JAMA 1995; 274:550–553

12. Chang MY, McBride LA, Ferguson MA: Vari-ability in brain death declaration practices inpediatric head trauma patients. Pediatr Neu-rosurg 2003; 39:7–9

13. Mathur M, Petersen L, Stadtler M, et al: Vari-ability in pediatric brain death determinationand documentation in southern California.Pediatrics 2008; 121:988–993

14. Ely JW, Graber ML, Croskerry P: Checkliststo reduce diagnostic errors. Acad Med 2011;86:307–313

15. Barochia AV, Cui X, Vitberg D, et al: Bundledcare for septic shock: An analysis of clinicaltrials. Crit Care Med 2010; 38:668–678

16. Chua C, Wisniewski T, Ramos A, et al: Mul-tidisciplinary trauma intensive care unitchecklist: Impact on infection rates.J Trauma Nurs 2010; 17:163–166

Opening our eyes to postoperative pulmonary complications*

Pulmonary complications arecommon after surgery due tothe unique stresses that areplaced on the patient during

the perioperative period, such as generalanesthesia, the surgical stress response,and postoperative pain (1). Patients withpulmonary complications experience lon-ger hospital stays, are more likely to bedischarged to a skilled nursing facility,and have increased rates of mortality (2).

Pulmonary complications have been de-fined in a diverse manner, making mea-surement and improvement difficult.Thus, improving the care of the postop-erative patient requires a valid and reli-able definition.

In this issue of Critical Care Medicine,Shander et al (3) have presented a sagediscussion of the burden postoperativepulmonary complications create throughincreasing morbidity and mortality andthus increasing resource utilization.These authors were part of a patientsafety summit consisting of a multidisci-plinary group that included health policyresearchers, patient safety and quality ex-perts, and intensivist physicians repre-senting a broad background of primaryspecialties. They believe that increasedawareness of both the scope of the prob-

lem of postoperative pulmonary compli-cations and strategies to reduce thesecomplications can lead to improved pa-tient care, better resource utilization, andoverall cost savings for the health systemas a whole. Recommendations were basedon the expert clinical experience of thegroup and evidence collected from a lim-ited literature search.

Proper application of resources is de-pendent upon accurate assessment of theprevalence of pulmonary complications.Given the heterogeneity of existing data,the authors’ have done a commendablejob of characterizing the scope and im-portance of postoperative pulmonarycomplications and have set a frameworkfor future discussions. Measuring theburden of postoperative pulmonary com-plications is contingent on standardized

*See also p. 2163.Key Words: healthcare; outcome assessment

(health care); postoperative complications; quality in-dicators; respiratory insufficiency



DOI: 10.1097/CCM.0b013e3182226f9c


definitions of disease, and there is no betterexample of this dilemma than ventilator-associated pneumonia. Ambiguities as towhat constitutes ventilator-associatedpneumonia exist internally within individ-ual published definitions and vary by defin-ing group (4). When compared to autopsyas a gold standard, our ability to accuratelydiagnose or exclude ventilator-associatedpneumonia is poor (5). Improving our un-derstanding of the scope and burden ofpostoperative pulmonary complications inthe future will require standardization ofdisease definitions for proper data collec-tion, aggregation, and analysis.

Postoperative pulmonary complica-tions encompass a spectrum of processessuch as bronchospasm, pneumonia, andrespiratory failure. As such, there is nosingle modality that is likely to reducecomplication rates, but instead a multi-faceted approach or “bundle” of interven-tions aimed at reducing complicationswill likely be required. The authors pro-pose a number of evidence-based inter-ventions that may be used to reduce com-plication rates. Given the complexity andmagnitude of the problem, eliminatingpostoperative pulmonary complication mayseem to be an impossible goal using onlystandardization of best practices, yet thisapproach has already been proven success-ful in the dramatic reductions in catheter-related bloodstream infections (6).

There is currently no standard bywhich we are able to ascertain the qualityof postoperative care as it relates to pul-monary complications. Calculating post-operative rates of mechanical ventilatorsupport in excess of 48 hrs is an enticingproposal for this new metric, but we musttread cautiously. It is true that these datashould be easy to obtain through existingcollection mechanisms; however, the va-lidity of this metric is unclear at thistime. A more rigorous analysis of currentevidence using standard principles suchas systematic literature review should beused to determine the validity of this rec-ommendation. Furthermore, rigoroustesting for reliability and accuracy needsto be performed before introduction as anaccepted quality measure.

There is the inherent danger that pub-lic reporting performance measures maydrive practice to comply with the mea-sure without actually improving deliveryof care. Emergency department bestpractice, as judged by the Centers forMedicare & Medicaid Services (Balti-more, MD), required antibiotic adminis-

tration to any patient with community-acquired pneumonia within 4 hrs ofpresentation (7). Desire to comply withthis publicly reported measure ultimatelyled to antibiotic administration, in the faceof diagnostic uncertainty, which manytimes was unnecessary (8). Similarly, man-dated reporting of the rates of postoperativemechanical ventilator use in excess of 48hrs could create incentive for early termi-nation of a necessary treatment.

Metrics that are outcome based, suchas duration of mechanical ventilation, inthe setting of external reporting forbenchmarking, can create unforeseen in-centives. To lower the reported rates ofcomplications, hospitals may “cherrypick” patients who are least likely to havecomplications, thereby improving themeasured outcome to be reported with-out ever having improved the quality ofcare. Conversely, hospitals that improvedelivery of best practices while serving ahigher risk population may not show thesame improvement (9, 10). Quality im-provement projects, especially when tiedto pay for performance, may also draw re-sources from other health initiatives, re-sulting in an overall negative health effectdespite producing improvements in the de-sired indicator (11). Thus, metrics thattrack delivery of best practices may ulti-mately have the best potential to improveoutcomes across the system as a whole.

There is a growing list of stakeholdersthat utilize quality metrics for a variety ofrequirements. They are used by individ-ual institutions for internal quality im-provement and external benchmarking,regulatory agencies for monitoring, andthird-party payers for pay-for-perfor-mance incentives. Premature adoption ofa metric has the potential for harm with-out offering improvement in care.

Reductions in postoperative pulmo-nary complications will require develop-ment and consensus acceptance of diag-nostic criteria and creation of a validstandard of measurement and ultimatelycare guidelines. Successful implementa-tion of such a quality improvement ini-tiative will require input from multiplestakeholders in the delivery process, in-cluding caregivers and their professionalsocieties, hospital systems, governmentand regulatory bodies, and third-partypayers. Inadequate consideration of anycomponent in this process may under-mine the capability to significantly re-duce postoperative pulmonary infections.We applaud the efforts of Dr. Shander and

colleagues in taking the first substantivesteps toward advancing our understand-ing of this important clinical entity.

Mark C. Romig, MDDepartment of Anesthesiology

& Critical Care MedicineJohns Hopkins University

School of MedicineBaltimore, MD

Todd Dorman, MD, FCCMDepartment of Anesthesiology

& Critical Care MedicineJoint Appointments inMedicine, Surgery, and theSchool of Nursing

Johns Hopkins UniversitySchool of Medicine

Baltimore, MD

REFERENCES

1. Ferreyra G, Long Y, Ranieri VM: Respiratorycomplications after major surgery. CurrOpin Crit Care 2009; 15:342–348

2. Thompson DA, Makary MA, Dorman T, et al:Clinical and economic outcomes of hospitalacquired pneumonia in intra-abdominal sur-gery patients. Ann Surg 2006; 243:547–552

3. Shander A, Fleisher LA, Barie PS, et al: Clinicaland economic burden of postoperative pulmo-nary complications: Patient safety summit on def-inition, risk-reducing interventions, and preven-tive strategies. Crit Care Med 2011; 39:2163–2172

4. Klompas M, Platt R: Ventilator-associatedpneumonia--The wrong quality measure forbenchmarking. Ann Intern Med 2007; 147:803–805

5. Tejerina E, Esteban A, Fernandez-Segoviano P,et al: Accuracy of clinical definitions of venti-lator-associated pneumonia: Comparison withautopsy findings. J Crit Care 2010; 25:62–68

6. Pronovost PJ, Goeschel CA, Colantuoni E, et al:Sustaining reductions in catheter related blood-stream infections in Michigan intensive careunits: Observational study. BMJ 2010; 340:c309

7. Wachter RM, Flanders SA, Fee C, et al: Publicreporting of antibiotic timing in patients withpneumonia: Lessons from a flawed performancemeasure. Ann Intern Med 2008; 149:29–32

8. Nicks BA, Manthey DE, Fitch MT: The Cen-ters for Medicare and Medicaid Services(CMS) community-acquired pneumonia coremeasures lead to unnecessary antibiotic ad-ministration by emergency physicians. AcadEmerg Med 2009; 16:184–187

9. Dranove D, Kessler D, McClellan M, et al: Ismore information better? The effects of re-port cards on health care providers. J PolitEcon 2003; 111:555–588

10. Shen Y: Selection incentives in a perfor-mance-based contracting system. HealthServ Res 2003; 38:535–552

11. Mullen KJ, Frank RG, Rosenthal MB: Canyou get what you pay for? Pay-for-perfor-mance and the quality of healthcare provid-ers. Rand J Econ 2010; 41:64–91


Just fastening the belt! Is it the future measure for assessing fluidresponsiveness?*

Observing the respiratory vari-ation of hemodynamic sig-nals has emerged as a valu-able tool for assessing

volume responsiveness in mechanicallyventilated patients (1, 2). The concept isbased on the assumption that the cyclicchanges in right ventricular preload in-duced by mechanical ventilation shouldresult in greater cyclic changes in leftventricular stroke volume when bothventricles operate on the steep ratherthan on the flat portion of the Frank-Starling curve, i.e., in the case of biven-tricular preload dependency (3). Thisconcept has gained great popularity inrecent years for at least three reasons.First, the heart-lung interaction indiceshave been reported to reliably predictfluid responsiveness in patients who arefully adapted to their ventilator and donot experience cardiac arrhythmias (4).Second, classic markers of preload, suchas cardiac filling pressures, have beendemonstrated to be poor predictors offluid responsiveness (5, 6). Third, newhemodynamic monitors that provide au-tomatic calculation and real-time displayof heart-lung interaction indices havebeen developed and are now available atthe bedside. Most of these monitors, e.g.PiCCO (Pulsion Medical Systems, Mu-nich, Germany), LiDCO (LiDCO Ltd, Lon-don, UK), FloTrac/Vigileo (Irvine, CA),use the arterial pressure waveform ob-tained from a peripheral artery catheterto derive heart-lung interaction indices.Pulse pressure variation (PPV) is probablythe most popular dynamic parameter offluid responsiveness since it is obtainedfrom a simple algorithm (7). Stroke vol-

ume variation (SVV) is another dynamicparameter derived from the arterial pres-sure waveform analysis that, unlike PPV,can only be provided by hemodynamicdevices that monitor stroke volume fromthe arterial pressure waveform. Becauseof its denomination, SVV sounds like abetter reflection of the respiratory vari-ability of stroke volume than PPV. How-ever, a recent meta-analysis in mechani-cally ventilated patients showed that PPVis actually a better predictor of fluid re-sponsiveness than SVV (4), maybe be-cause the calculation of stroke volumefrom the arterial pressure waveform anal-ysis is based on mathematical assump-tions, which are not always applicable incritically ill patients. In addition, becauseof the pulse wave amplification phenom-enon, leading to a higher pulse pressurein the periphery than in the ascendingaorta, PPV and SVV measured at the pe-riphery could differ from the correspond-ing indices, if measured in the ascendingaorta. The magnitude of this difference isvariable from patient to patient since thedegree of the pressure wave amplificationphenomenon depends on numerous fac-tors, such as age, height, arterial compli-ance, and vasomotor tone. In any case,because the femoral artery pressure iscloser to the aortic pressure than theradial artery is (8), SVV measured at thefemoral artery level should be a betterreflection of the respiratory variation ofthe left ventricular stroke volume thanSVV if measured at the radial artery level.Measuring the respiratory variability ofthe stroke volume at the level of aorticannulus is indisputably the best method,and it is feasible using ultrasound de-vices. Accordingly, the respiratory vari-ability of the aortic blood velocity mea-sured at the aortic annulus usingechocardiography (9) has been demon-strated to be a good predictor of fluidresponsiveness in mechanically venti-lated patients. The advantages of thistechnique are its low invasiveness and itsability to track the respiratory changes instroke volume at the appropriate (cen-

tral) level for minimizing the pulse waveamplification phenomenon and hence foran adequate interpretation of heart-lunginteraction indices in the perspective of as-sessing fluid responsiveness. Nevertheless,because echocardiography is an operator-dependent technique unable to provide acontinuous monitoring of the hemody-namic variables, its widespread use is re-grettably still limited in the perspective ofthe prediction of fluid responsiveness.

In this issue of the Critical Care Medi-cine, Maisch and co-workers (10) presentinteresting results of an experimental studyin which SVV was assessed neither by ul-trasound techniques nor by arterial pres-sure waveform analysis but by an innova-tive technology. The electric impedancetomography continuously tracks thechanges in intrathoracic volumes by mea-suring the changes in electric impedance ofthe thorax obtained from a thoracic beltapplied around the thorax. The authors de-veloped a new analysis of the electric aorticimpedance signal that allows assessment ofthe volume-related changes in impedanceinduced by the beating heart, which leadsto a continuous estimation of SVV. In thispreliminary validation study in animals, theauthors found a good correlation betweenSVV assessed by electric impedance tomog-raphy and the reference technique, i.e., SVVmeasured by an ultrasound probe placeddirectly around the aortic annulus. Thesepositive results must be, however, tem-pered by the fact that the limits of agree-ment between SVV assessed by electric im-pedance tomography and the referencemethod were (unacceptably?) large, sug-gesting that the technology is still imper-fect and should further be improved beforebeing applied to patients. Surprisingly, theauthors did not test whether SVV estimatedby the new technique actually predictedfluid responsiveness, i.e., whether cardiacoutput significantly increased in responseto a systematic volume challenge. Beyondthe limitations of this preliminary study,electric impedance tomography should beregarded as a promising technology. If im-proved and applied to humans, this method

*See also p. 2173.Key Words: cardiac preload; electrical impedance

tomography; fluid responsiveness; pulse pressure vari-ation; stroke volume variation

Dr. Teboul consulted for Pulsion Medical Systems.Dr. Monnet is a member of the Advisory Board forPulsion Medical Systems.


DOI: 10.1097/CCM.0b013e318226608b


would have the great advantage to be non-invasive and to assess on a real-time basisthe variations of stroke volume at the “cen-tral” level.

Jean-Louis Teboul, MD, PhDXavier Monnet, MD, PhD

AP-HPService de Reanimation

MedicaleHopitaux Universitaires Paris-

Sud andFaculte de Medecine Paris-SudUniversite Paris-SudLe Kremlin-Bicetre, France

REFERENCES

1. Marik PE, Monnet X, Teboul JL: Dynamicparameters to guide fluid therapy. Ann In-tensive Care 2011; 1:1

2. Monnet X, Teboul JL: Volume responsive-ness. Curr Opin Crit Care 2007; 13:549–553

3. Michard F, Teboul JL: Using heart-lung in-teractions to assess fluid responsiveness dur-ing mechanical ventilation. Crit Care 2000;4:282–289

4. Marik PE, Cavallazzi R, Vasu T, et al: Dy-namic changes in arterial waveform derivedvariables and fluid responsiveness in me-chanically ventilated patients: A systematicreview of the literature. Crit Care Med 2009;37:2642–2647

5. Osman D, Ridel C, Ray P, et al: Cardiac fillingpressures are not appropriate to predict he-modynamic response to volume challenge.Crit Care Med 2007; 35:64–68

6. Marik PE, Baram M, Vahid B: Does centralvenous pressure predict fluid responsive-ness? A systematic review of the literatureand the tale of seven mares. Chest 2008;134:172–178

7. Michard F, Boussat S, Chemla D, et al:Relation between respiratory changes inarterial pulse pressure and fluid respon-siveness in septic patients with acute cir-culatory failure. Am J Respir Crit Care Med2000; 162:134 –138

8. Dufour N, Chemla D, Teboul JL, et al:Changes in pulse pressure following fluidloading: A comparison between aortic root(non-invasive tonometry) and femoral artery(invasive recordings). Intensive Care Med2011; 37:942–949

9. Feissel M, Michard F, Mangin I, et al: Respi-ratory changes in aortic blood velocity as anindicator of fluid responsiveness in ventilatedpatients with septic shock. Chest 2001;119:867–873

10. Maish S, Bohm SH, Solà J, et al: Heart-lunginteractions measured by electrical imped-ance tomography. Crit Care Med 2011; 39:2173–2176


icu6

Documents

abdominal compartment syndrome

lippincott williams

electrical impedance tomography

pope pius xii

intensive care units

intensive care unit

cardiac lling pressures

electric impedance tomography