GIMD

Imtetsh

D

R2FDSaCI5

©P

lucose and Surgical Sepsis: A Study of Underlyingmmunologic Mechanismsotaz Qadan, MD, PhD, MRCS(Ed), E Brooks Weller, BS, Sarah A Gardner, BS, Claudio Maldonado, PhD,onald E Fry, MD, FACS, Hiram C Polk, Jr, MD, FACS

BACKGROUND: Early clinical trials investigating the role of tightly controlled glucose levels showed markedbenefit in survival of critically ill patients. However, a recent meta-analysis and large random-ized controlled trial have failed to reproduce the benefit, showing instead substantially increasedrisk of dangerous hypoglycemia. We sought to investigate the effects of varying glucose con-centrations on previously tested, prognostically significant, innate immune parameters, todefine any potential effects of glucose at the cellular level.

STUDY DESIGN: After formal approval and informed consent, venous blood samples were collected from younghealthy volunteers. Up to 11 corresponding (same-subject) samples were incubated at 100, 350,or 600 mg/dL glucose concentrations and analyzed to determine human leukocyte antigen-DRsurface receptor expression, cytokine release, phagocytic capacity, and formation of reactiveoxygen species. Data are presented as mean � SEM.

RESULTS: After incubation, the change in human leukocyte antigen-DR mean channel fluorescence fromresting baseline values in lipopolysaccharide-stimulated monocytes was not significantly differ-ent between 100, 350, and 600 mg/dL (1,749 � 110; 1,748 � 120; and 1,725 � 96,respectively; p � 0.89). Tumor necrosis factor-� concentrations were significantly lower forsamples incubated at higher glucose concentrations (179 � 50 pg/mL, 125 � 30 pg/mL, and107 � 29 pg/mL; p � 0.05). The phagocytic capacity of the innate immune system wasmarginally enhanced by glucose. However, the formation of reactive oxygen species was mark-edly impaired by rising glucose (55% to 66% impairment; p � 0.05).

CONCLUSIONS: Increasing glucose concentrations exert considerable opposing effects on several well-established innate immunologic processes. The opposing findings might contribute to recentclinical controversies. Physician judgment and experience are essential to imminent treatmentof critically ill and perioperative surgical patients. ( J Am Coll Surg 2010;210:966–974. © 2010by the American College of Surgeons)

sp

wlctptctwtepicc

n recent years, attempts to understand innate host defenseechanisms have provided a specific focus for the reduc-

ion of surgical site infection and associated sepsis, whichxtends beyond prophylactic or therapeutic antimicrobialherapy.1 Manipulating or even failing to control factorsuch as temperature, oxygen, and glucose, among others,as been shown to substantially influence development of

isclosure Information: Nothing to disclose.

eceived October 21, 2009; Revised January 31, 2010; Accepted February 1,010.rom The Price Institute of Surgical Research (Qadan, Weller, Gardner),epartment of Physiology and Biophysics (Maldonado), and Department of

urgery (Polk), University of Louisville School of Medicine, Louisville, KY,nd Michael Pine and Associates, Chicago, IL (Fry).orrespondence address: Motaz Qadan, MD, PhD, MRCS(Ed), The Price

nstitute of Surgical Research, Medical-Dental Research Building, 3rd Floor,11 South Floyd St, Louisville, KY 40202. email: m0qada01@louisville.edu

9662010 by the American College of Surgeons

ublished by Elsevier Inc.

urgical site infection and sepsis after surgery, despite ap-ropriate use of prophylactic antibiotics.2-5

The importance of more tightly controlled glucose levelsas once assumed to be clear-cut at both clinical and cel-

ular levels.6 In reality, that conclusion appears to be radi-ally changing. The early, large, randomized controlledrial by Van den Berghe and colleagues6 in 2001 rapidlyrovided the clinical standard of practice, having shownhat tight glucose control reduced deaths considerably inritically ill patients.6 Contemporary acceptance in hospi-als led to development of closely adhered to protocols,hich were aimed at tightly controlling glucose levels

hrough use of intensive insulin regimens. The practice wasxtrapolated and expanded to include progressively largeratient populations, including surgical patients. However,n 2008, a meta-analysis that combined all randomizedontrolled trials investigating the effect of tight glucoseontrol in critically ill medical and surgical patients failed

ISSN 1072-7515/10/$36.00doi:10.1016/j.jamcollsurg.2010.02.001

tpscgcsdos

igntwnAsrpp3rltsemocdta

MWAR

vtvmEspbmjmtanr

SFwem�h�p(fceS��6wr

lDamm

lwpanbtttt

967Vol. 210, No. 6, June 2010 Qadan et al Glucose and Surgical Sepsis

o reproduce the distinct benefit on mortality describedreviously.7 In addition, the meta-analysis highlighted aubstantially increased risk of hypoglycemia, with an asso-iated increase in mortality secondary to this overtly dan-erous complication. The clinical benefit of tight glucoseontrol was suddenly thrown into question. A large pro-pective analysis published in 2009 further confirmed theetrimental effect of tightly controlling glucose levels, seen,nce again, as a substantially increased mortality rate re-ulting from associated hypoglycemia.8

Because of this current difference of opinion at the clin-cal level, the benefit and risks from tightly controllinglucose are no longer known and represent a clinical co-undrum without a solution in sight. We sought to inves-igate the effect of increasing glucose concentrations on 4ell-recognized, previously tested, and prognostically sig-ificant innate immune parameters at the cellular level.lthough acknowledging the limitations of in vitro re-

earch, we hoped that the translation of such bench labo-atory findings to the bedside might aid in uncoveringotential effects, if any, from tight glucose control in hos-italized patients. We used glucose concentrations of 100,50, and 600 mg/dL, which were chosen to respectivelyepresent a wide spectrum that ranges from normal fastingevels through poorly controlled diabetic levels, and extendo a grossly elevated ketoacidotic glucose level. By usinguch a wide-ranging spectrum, we hoped that any in vitroffects of varying glucose concentrations would be un-asked. We specifically investigated the effects of glucose

n monocyte presentation through surface human leuko-yte antigen (HLA)-DR receptor expression, cytokine pro-uction, neutrophil phagocytosis, and formation of reac-ive oxygen species (ROS) by neutrophils, using a consistentnd reproducible whole-blood model.

ETHODShole-blood samplingfter approval by the University of Louisville Institutionaleview Board and written, informed consent, up to 15 mL

Abbreviations and Acronyms

DHR � dihydrorhodamine 123FITC � fluorescein isothiocyanateHLA � human leukocyte antigenIL-10 � interleukin-10LPS � lipopolysaccharideMCF � mean channel fluorescencePMA � phorbol myristate acetateROS � reactive oxygen speciesTNF-�� tumor necrosis factor-�

enous blood samples were collected from a random selec-ion of 15 healthy (8 male, 7 female), 18- to 42-year-oldolunteers. Volunteer body mass indices (calculated as kg/2) were all within normal limits (range 18 to 25 kg/m2).xclusion criteria included any history of immunosuppres-

ive disorders, diabetes mellitus, chronic medication, orregnancy. All volunteers fasted for a minimum of 6 hoursefore venipuncture and blood glucose levels were deter-ined using a Glucometer Elite (Bayer Corporation). Sub-

ects with fasting blood glucose levels �90 mg/dL or �110g/dL were excluded. Sample identities and concentra-

ions were blinded throughout the experimental procedurend data analysis and equal or near equal (if odd-umbered) numbers of male and female volunteers wereecruited for all experiments conducted.

ample preparationor HLA-DR receptor and cytokine determinations, bloodas collected in ethylenediamine tetraacetic acid Vacutain-

rs (Becton-Dickinson and Co.). Samples were supple-ented with 200 mM L-glutamine at a concentration of 10L/mL whole blood because incubation periods were �1our (Sigma Chemical Co.). Aliquots of whole blood (900L) were subsequently transferred into 5-mL Falconolypropylene culture tubes (VWR). A lipopolysaccharideLPS; Sigma-Aldrich) concentration of 1 ng/mL,5 as con-irmed by our pilot experiments, provided the endotoxinhallenge immediately before incubation and was used tomulate inevitable surgical contamination intraoperatively.amples were supplemented with 100 �L normal saline, 50L normal saline, and 50 �L 5% dextrose solution, or 100L 5% dextrose solution, to provide 1 mL of 100, 350, and00 mg/dL glucose solutions, respectively. Glucose levelsere rechecked using the glucometer to confirm the accu-

acy of the solutions prepared.For phagocytosis assays, venous blood samples were col-

ected in sodium heparin BD Vacutainers (Becton-ickinson and Co.). LPS (1 ng/mL) was supplemented

nd 2-mL whole-blood aliquots were preincubated for 15inutes at the constituted glucose solutions before com-encement of quantification of phagocytosis.For ROS quantification experiments, blood was also col-

ected in sodium heparin BD Vacutainers. Aliquots ofhole blood (1 mL) were transferred into 5-mL Falconolypropylene culture tubes. Optimal phorbol myristatecetate (PMA; Sigma-Aldrich) concentrations of 100g/mL were used to provide cellular stimulus before incu-ation, as determined by previous studies,9 and by our ownime and dose�response pilot experiments. Working solu-ions of PMA were prepared by adding 5 �L (5 �g) PMAo 0.5 mL PBS. Ten microliters working solution wereransferred into Falcon polypropylene culture tubes. Whole-

bccn

PEC1ambAazaevh

MWcaBSrqwu

mcCtCs

FMaf2oi

PETfwb

wasdciTcAs

CP1c(wtwewpsLp

PPEAuFacirt1Pfbamtv

EAsag6

968 Qadan et al Glucose and Surgical Sepsis J Am Coll Surg

lood samples (940 �L) with the appropriate glucose con-entrations were then transferred into the polypropyleneulture tubes that contained the added PMA to give a 100g/mL whole-blood volume.

art 1: HLA-DR receptor expressionxperimental designorresponding LPS-treated samples with glucose levels of00, 350, and 600 mg/dL were incubated at 21% (roomir) oxygen with 5% carbon dioxide, at 37°C. To determineonocyte HLA-DR surface expression, samples were incu-

ated for 2 hours, as determined by early pilot experiments.liquots of cultured samples (50 �L) from the HLA-DRssay were stained immediately before incubation (timeero) and after incubation, and were then subsequentlynalyzed to determine baseline and final monocyte HLA-DRxpression, respectively. Throughout incubation, gentleortex was applied at 45-minute intervals to ensure cellularomogeneity.

onocyte CD 14�/HLA-DR staining and extractionhole-blood samples (50 �L) were stained with fluores-

ein isothiocyanate (FITC)-labeled antihuman CD14�

nd phycoerythrin-labeled anti�HLA-DR antibodies (BDiosciences) to determine monocyte HLA-DR expression.taining was carried out for 25 minutes in the culture envi-onment to prevent down-/upregulation of HLA-DR beforeuantitative binding. Appropriately matched isotype controlsere used to determine nonspecific binding thresholds. Man-facturer instructions were followed closely.

After staining, red blood cell lysis was carried out for 6inutes using ice-cold ammonium chloride, potassium bi-

arbonate, and ethylenediamine tetraacetic acid (Sigmahemical Co.) solution. Monocytes were pelleted by cen-

rifugation, washed with 1 mL Dulbecco’s PBS (Sigmahemical Co.), and fixed in 250 �L 1% paraformaldehyde

olution (Polyscience Inc.).

low cytometric analysisonocyte CD14� and HLA-DR surface expression were

nalyzed within 4 hours of cell culture using a FACSCaliburlow cytometer (Becton-Dickinson and Co.). A total of0,000 events were acquired. HLA-DR mean channel flu-rescence (MCF) was analyzed in CD14� monocytes us-ng Cell Quest software (Becton-Dickinson and Co.).

art 2: Cytokine assaysxperimental designwo series of cytokine experiments were conducted. Theirst series of experiments were conducted simultaneouslyith HLA-DR assay experiments where, after 2-hour incu-ation and staining for HLA-DR, whole-blood samples

ere centrifuged at 3,000 rpm for 12 minutes to obtain ancellular supernatant, which was stored at �81°C for sub-equent analysis. The second series of experiments was con-ucted to provide a time-course analysis of anti-inflammatoryytokine release after endotoxin challenge. Samples werencubated for intervals of 30, 60, 120, and 240 minutes.ime intervals were selected to allow characterization ofytokine trends, as determined by early pilot experiments.fter incubation, acellular supernatant was obtained and

tored as described.

ytokine determinationslasma tumor necrosis factor-� (TNF-�) and interleukin-0 (IL-10) concentrations were quantified using commer-ially available enzyme-linked immunosorbent assay kitse-Biosciences). All enzyme-linked immunosorbent assaysere carried out in 96-well plates according to manufac-

urer’s instructions. Samples were assayed in duplicate,ith either recombinant human TNF-�, or IL-10, to gen-

rate a standard curve. Enzyme activity was measured at aavelength of 450 nm on a SpectraMax Plus384 spectro-hotometer and data were generated using Softmax Prooftware (Molecular Devices) and expressed in pg/mL.ower limits of detection for TNF-� and IL-10 were 4g/mL and 2 pg/mL, respectively.

art 3: Phagocytosis assaysreparation and opsonization of FITC-labeledscherichia colistock solution of FITC-labeled E coli (Invitrogen Molec-

lar Probes) was prepared by diluting 5 mg lyophilizedITC-labeled E coli with Hanks Balanced Salt Solution forfinal concentration of 1 mg/mL, where 1 mg bacteria

onsisted of 3 � 108 particles/mg. The bacteria were son-cated and stored at �81°C until required for use. Whenequired, the labeled organisms were thawed and washedwice with PBS by centrifugation. To opsonize bacteria, a0% pooled human serum solution was freshly prepared inBS and was mixed with equal volumes of bacteria for a

inal opsonin concentration of 5%. The mixture was incu-ated for 25 minutes at 37°C with 5% carbon dioxide tollow for opsonization to occur. Excess opsonins were re-oved by washing bacteria twice with PBS and centrifuga-

ion, followed by resuspension to the original bacterialolume.

xperimental designfinal volume of 40 �L/mL whole blood (40 �g) of op-

onized bacteria was quickly added to the preincubatedliquots. Samples were incubated at 21% (room air) oxy-en with 5% carbon dioxide, at 37°C. Sampling times of 2,, 12, 20, and 45 minutes were selected, which adequately

ct�mwl

FBe�tMsw

PERItD��sdtm(3sa

FSFap

PEgivrA1stwfa

cs

peSBS

RW(cbu1wM0

cccp3cp

it(

Fga

969Vol. 210, No. 6, June 2010 Qadan et al Glucose and Surgical Sepsis

haracterized the hyperbolic nature of the phagocytic reac-ion in early pilot experiments and were sufficient to attain

95% of cells contributing to phagocytosis within 45inutes. At these times, 50-�L samples were extractedith subsequent red cells lysis, white blood cell pellet iso-

ation, and fixation as described previously.

low cytometric analysisefore acquisition using a FACSCalibur flow cytometer,xtracellular fluorescent bacteria were quenched using 75L trypan blue reagent (Invitrogen Molecular Probes). A

otal of 20,000 events were acquired for each sample.onocytes and neutrophils were gated according to light

cattering properties and the proportion of fluorescent cellsas recorded.

art 4: ROS assaysxperimental designOS were quantified using dihydrorhodamine 123 (DHR;

nvitrogen-Molecular Probes). A 60-�g/mL working solu-ion of DHR was freshly prepared by adding 8 �L stockHR to 1,312 �L PBS to obtain a final volume of 3g/mL whole blood. Immediately after its preparation, 50L DHR working solution was added to the combined

olution of whole blood, glucose, and PMA. After the ad-ition of DHR, culture tubes were promptly placed withinhe incubation chambers. Samples were incubated for 30inutes, as determined by early pilot experiments, at 21%

room air) oxygen with 5% carbon dioxide, at 37°C. After0 minutes incubation, 50-�L samples were extracted withubsequent red cells lysis, white blood cell pellet isolation,nd fixation as described previously.

low cytometric analysisamples were analyzed within 1 hour of cell culture using aACSCalibur flow cytometer. A total of 10,000 events werecquired. MCF (ROS quantity) was recorded in gated neutro-hils using Cell Quest software (Becton-Dickinson and Co.).

art 5: Statistical analysisach subject’s blood was incubated across the 3 differentlucose concentrations, referred to herein as “correspond-ng samples” (ie, same-subject). This reduces inter-subjectariability and allows use of more specialized tests, such asepeated-measures ANOVA testing. Repeated-measuresNOVA was used to detect significant differences between00, 350, and 600 mg/dL (3 groups). In the event that aignificant difference was detected, multiple-comparisonesting to compare all pairwise groups was used to assesshether any pairwise comparisons were significantly dif-

erent (ie, 100 versus 350 mg/dL, 100 versus 600 mg/dL,nd 350 versus 600 mg/dL). Multiple comparisons were

arried out using the Holm-Sidak test. The p values pre-ented were adjusted for multiple comparisons.

For each experiment, subjects were recruited from theool of 15 volunteers, ensuring that no predisposition toither gender had occurred. Data are presented as mean �EM. Statistical analyses were performed using Primer ofiostatistics software (version 6.0, McGraw Hill, 2005).ignificance was assigned at the 5% level.

ESULTShole-blood cells were separated using flow cytometry

Fig. 1). After 2 hours of incubation at the varying glucoseoncentrations, the change in HLA-DR MCF from restingaseline HLA-DR values at time 0 (provided in arbitrarynits) in LPS-stimulated monocytes at 100 mg/dL was,749 � 110. When corresponding samples (same-subject)ere incubated at 350 mg/dL and 600 mg/dL, recordedCF was 1,748 � 120 and 1,725 � 96, respectively (p �

.89) (Fig. 2).At 2 hours, approximate peak values of TNF-� signifi-

antly and progressively decreased with increasing glucoseoncentrations (Fig. 3A). Compared with 100 mg/dL glu-ose, where mean TNF-� concentration was 179 � 50g/mL, corresponding samples (same-subject) incubated at50 mg/dL and 600 mg/dL contained mean TNF-� con-entrations of 125 � 30 pg/mL (p � 0.05) and 107 � 29g/mL (p � 0.05), respectively.The anti-inflammatory cytokine response, assessed us-

ng IL-10 concentrations, demonstrated a similar trend be-ween 100, 350, and 600 mg/dL glucose concentrationsFig. 3B). All concentrations demonstrated a small early

igure 1. A typical flow cytometric analysis of whole-blood cells withated regions A1, A2, and A3 representing lymphocytes, monocytes,nd neutrophils, respectively. FSC, forward scatter; SSC, side scatter.

rcric

atuaappsgccrtmqca(

Rgcoe

iabtfd

Fp�chmpcfmstandg

Fhnivgr0

970 Qadan et al Glucose and Surgical Sepsis J Am Coll Surg

ise in IL-10 concentrations, which indicated satisfactoryell viability and function during the entire incubation pe-iod. There were no substantial differences in the early anti-nflammatory response between the different glucose con-entrations at any point in the first 4 hours.

Figure 4 depicts the hyperbolic response, which is char-cteristic of neutrophil phagocytosis. Using the selectedime periods, approximately all neutrophils had contrib-ted to the process of E coli phagocytosis, indicating andequate experimental response, technique, and, oncegain, with near complete cell viability and function. Therocess is characteristically hyperbolic as early ingestionroceeds at an exponential rate followed by a slower re-ponse as neutrophils become saturated with ingested or-anisms. In the initial exponential phase, 600 mg/dL glu-ose provided consistently faster rates of phagocytosis, asompared with 350 and 100 mg/dL. Phagocytosis rates,epresented by Km50 (time for 50% of neutrophils to con-ribute to phagocytosis) values, were 22% quicker for 600g/dL as compared with 350 mg/dL (p � 0.05), and 28%

uicker as compared with 100 mg/dL (p � 0.05). Phago-ytosis at 350 mg/dL was 8% faster than at 100 mg/dL,lthough the difference was not statistically significantp � 0.05).

In order to provide a functional aspect to phagocytosis,OS formation was measured and compared between the 3lucose concentrations. Although these values do not spe-ifically represent bacterial killing, they represent the amountf substrate available for the process and have been repeat-dly shown to positively correlate with intracellular kill-

igure 2. The effect of 100, 350, and 600 mg/dL glucose onuman leukocyte antigen�DR receptor surface expression in 1g/mL lipopolysaccharide-stimulated monocytes after 2 hours of

ncubation. Repeated-measures ANOVA was used to generate the palue of 0.89, which compares 100 versus 350 versus 600 mg/dLlucose. Because p � 0.05, multiple comparison testing was notequired to detect the source of significant differences. n � 11; p �.89.

ng.10 ROS formation was quantitatively assessed using therbitrary-valued MCF (Fig. 5). When samples were incu-ated at 600 mg/dL glucose as compared with 100 mg/dL,here was a statistically significant decrease of 66% in ROSormation (p � 0.05). At 350 mg/dL, there was a 55%ecrease compared with 100 mg/dL (p � 0.05). ROS for-

igure 3. (A) The effect of 100, 350, and 600 mg/dL glucose onroinflammatory cytokine production (tumor necrosis factor-� [TNF-]) in 1 ng/mL lipopolysaccharide (LPS)-stimulated whole-bloodells after 2 hours incubation, and obtained during simultaneouslyuman leukocyte antigen (HLA)-DR experiments. Because repeated-easures ANOVA revealed a p � 0.05, Holm-Sidak testing for allairwise comparisons was used. Significance was detected whenomparing (100 versus 350 mg/dL) and (100 versus 600 mg/dL),or which a p value is shown. The p value presented is adjusted forultiple comparisons. There was no difference between (350 ver-us 600 mg/dL). n � 11 ; *p � 0.05. (B) A time-course analysis ofhe effect of 100, 350, and 600 mg/dL glucose on the earlynti-inflammatory cytokine production (interleukin-10 [IL-10]) in 1g/mL LPS-stimulated whole-blood cells. There were no differencesetected at any sampling point between 100, 350, and 600 mg/dLlucose. n � 9.

ma

DHaawmasqectpmcmsctiaTps

brcpftlHrtd

edvnhRrsimautisml

Frmwa(pv0

FrpRmded(p*

971Vol. 210, No. 6, June 2010 Qadan et al Glucose and Surgical Sepsis

ation was not significantly different between 600 mg/dLnd 350 mg/dL glucose concentrations (p � 0.05).

ISCUSSIONow tightly glucose levels must be controlled in surgical

nd critically ill patients is no longer known. Tight controlchieved using intensive insulin regimens as comparedith conventional regimens was once thought to reduceortality and septic sequelae in hospitalized patients. The

bility to adhere to protocols employing the intensive in-ulin regimens in time came to represent a marker of theuality of care provided by physicians and hospitals. How-ver, recent prospective trials and meta-analyses, whichonstitute level I clinical evidence, have failed to reproducehese beneficial outcomes. Using clinical studies that em-loyed tight insulin protocols at blood glucose levels �150g/dL in an attempt to maintain levels �150 mg/dl, re-

ent level I evidence has instead showed increased detri-ental outcomes, including higher mortality rates as a re-

ult of aggressive glucose control. By resorting to in vitroellular studies, the effects of varying glucose concentra-ions could be studied on early, previously tested, innatemmune parameters, which provide the first line of defensegainst inevitable contamination after surgery or trauma.hese parameters have repeatedly been shown to be ofrognostic value in the course of patient recovery afterurgery and trauma.

Using an unchanged and reproducible in vitro whole-

igure 4. The effect of 100, 350, and 600 mg/dL glucose on theate of neutrophil phagocytosis, shown by the proportion of neutro-hils (Km50 � 50%) contributing to phagocytosis with given time.epeated-measures ANOVA revealed p � 0.05 at 2-, 6-, and 12-inute sampling points. Holm-Sidak tests were used to detectifferences between all pairwise comparisons. *Statistical differ-nce between (600 versus 100 mg/dL) and (600 versus 350 mg/L), for which a p value is shown. �Statistical difference between600 versus 350 mg/dL) only, also for which a p value is shown. Thevalues presented are adjusted for multiple comparisons. n � 9;p � 0.05; � p � 0.05.

lood model, our results show several important effectsesulting from increasing glucose concentrations. Specifi-ally, increasing glucose concentrations substantially im-aired formation of reactive oxygen intermediates availableor intracellular killing, but enhanced neutrophil phagocy-osis rates and attenuated proinflammatory cytokine re-ease. However, there were no substantial differences in

LA-DR surface receptor expression, despite the extremeange of glucose concentrations used. One would thinkhat such extremes of glycemia would unmask even subtleifferences in the host defense process.ROS formation represents a functional assessment of the

fficacy of the early components of the immune system toestroy pathogens immediately after contamination, pro-iding an estimate of patients’ capacity to potently elimi-ate early infection before gross dissemination within theost, ie, during the well-described decisive period.”11,12

OS formation might arguably be the most important pa-ameter that we have investigated in this setting, becauseeveral anesthetic drugs have been shown to impair thennate immune response and specifically blunt ROS for-

ation. In addition, because the ability of cells to presentntigen, produce cytokines, and contribute to phagocytosisltimately culminate in the process of intracellular killing,his parameter might arguably offset all other neutral find-ngs described at the cellular level. The lethality of diseasesuch as Chediak-Higashi syndrome, characterized by re-arkably effective phagocytosis but ineffective intracellu-

ar killing, provides evidence of the importance of effective

igure 5. The effect of 100, 350, and 600 mg/dL glucose oneactive oxygen species formation by neutrophils. Repeated-easures ANOVA revealed p � 0.05. Therefore, Holm-Sidak testsere used to detect differences between all pairwise comparisons,nd revealed statistical significance in the differences between100 versus 350 mg/dL) and (100 versus 600 mg/dL), for which avalue is shown, but not between (350 versus 600 mg/dL). The p

alue presented is adjusted for multiple comparisons. n � 8; *p �.05.

RgRutp

Rtefn

atsWiaTwrniacaehsblhfosmaomasspuccetrf

dttcm2e

0woptmlcgwc

Rtsscfcfcgolgb

tbsci

cdmwnsltmm

972 Qadan et al Glucose and Surgical Sepsis J Am Coll Surg

OS generation. Our results clearly show that increasinglucose concentrations substantially impair formation ofOS and the potential for intracellular killing. This partic-lar cellular finding correlates with early studies favoringight glucose control to reduce mortality and septic com-lications in critically ill patients.6

From a technical point of view, we used DHR to detectOS because it has been shown to detect all available in-

ermediates more sensitively than other stains. It is consid-red by many to be the most sensitive technique availableor accurately quantifying ROS formation and, with it,eutrophil oxidative killing potential.12,13

Impairment of ROS formation was not reproducedmong other immune parameters, however. In stark con-rast, increasing glucose concentrations were found to sub-tantially attenuate proinflammatory cytokine production.

e have previously attempted to explain the clinical signif-cance of cytokine production and benefit derived fromttenuation of release of proinflammatory proteins such asNF-�, an early proinflammatory cytokine released inhat seems to be a cascade of mediators. It plays a major

ole in activation and enhancement of immune mecha-isms, but has also been arguably linked to host tissue

njury and multiorgan dysfunction from many disparatend disagreeing reports. Some evidence for this pathologi-al role began to emerge in 1985, when pretreatment withntiserum to TNF-� was shown to protect mice againstndotoxin lethality in vivo.14,15 Van der Poll and colleaguesypothesized that secretion of proinflammatory cytokines,uch as TNF-�, occurred in relatively short-lived and rapidursts immediately after an insult.16,17 Exaggerated or pro-

onged early responses, however, were associated withigher morbidity and mortality rates, and were shown to beollowed by pathologically increased compensatory levelsf anti-inflammatory cytokines, such as IL-10, with a re-ultant “stunning” of the innate immune processes. Re-arkably, IL-10 levels were subsequently shown to serve asprognostic marker of impending mortality.18 Although

ur overall view remains that an integrated, sequential, andodulated release of both proinflammatory and, later,

nti-inflammatory cytokines are more important than ab-olute values, the beneficial progressive decrease of TNF-�hown with increasing glucose concentrations was unex-ected. This cellular finding coincided with a previouslyndeclared fatal complication of intensive lowering of glu-ose levels in the ACCORD (Action to Control Cardiovas-ular Risk in Diabetes) cardiovascular trial, and resulted inarly discontinuation of the study. The authors describedhe peculiar finding that increased mortality rate was not aesult of hypoglycemia in the enrolled patients.19,20 An un-avorable modification of the cytokine profile, as eluci-

ated by our findings, might have accounted, in part, forhe heightened mortality, particularly considering the po-ential detrimental effects of TNF-� levels on cardiac myo-yte function.21,22 After all, TNF-� levels are highest at 100g/dL glucose and the ACCORD study consisted of typediabetic patients with a propensity for cardiovascular dis-

ase as a selection criterion.Also unexpected was the statistically significant (p �

.05) dose-dependent enhancement in phagocytosis ratesith increasing glucose concentrations. The mere presencef additional substrate (glucose) for this energy-dependentrocess might have accounted for the rate enhancement ofhis cellular process. Interestingly, because the enhance-ent was dose-dependent and persisted at 600 mg/dL, it is

ikely that the Km of the enzymatic processes that use glu-ose as an energy source might be able to do so even at theselucose ranges. However, although the small enhancementas shown to be statistically significant, clinical signifi-

ance of that difference is not known.Our cytokine and phagocytosis findings clearly oppose

OS results. When translated into clinically meaningfulerms, ROS findings are in keeping with early clinical trials,uch as the Van den Berghe and colleagues’ trial,6 whichupport the use of aggressive glucose control to lower glu-ose serum levels. However, cytokine and phagocytosisindings are more in keeping with the most recent contrarylinical body of evidence. In addition, hyperglycemic ef-ects at the cellular level appear to be comparable past aertain threshold level. Although ROS findings can be ar-ued to be of more functional value and importance, thesepposing findings at the cellular level represent the under-ying scientific complexity of this debate, which might be-in to explain conflicting findings seen at the clinical level,ut ultimately contribute to heighten the disagreement.Differences in early anti-inflammatory cytokine produc-

ion and HLA-DR surface receptor expression, which haveoth been repeatedly shown to possess prognostic value,23

urprisingly, failed to demonstrate any differences with in-reasing glucose concentrations at the early stages of thenflammatory response seen in our model.

Although the specific mechanisms by which varying glu-ose levels might have accounted for the cellular changesescribed here have not been elucidated by these experi-ents, studies are underway to investigate specific path-ays that might be enhanced or inhibited by glucose. Ofote, however, is that previous studies that attempted topecifically isolate the effects of increasing solution osmo-arity failed even to reproduce the cellular findings encoun-ered with increasing glucose concentrations.24 Potentialechanisms to explain the effects of glucose in infectionodels have included the nuclear factor��B pathway.24-26

�ript

pctsiiiust

isfcwjbrtdmtctpmpgr

spMiRtftcrctic

A

SAA

DC

ApMafad

EF

R

1

1

1

973Vol. 210, No. 6, June 2010 Qadan et al Glucose and Surgical Sepsis

-O-linkage of N-acetylglucosamine, which is a glucose-elated post-translational modification,27 might also provemportant in this context.The intricacy and density of suchathways truly serve to demonstrate the complexity of thisopic and the associated findings in the clinical setting.

Although these studies have attempted to shed light onotential processes that might be affected by rising glucoseoncentrations, limitations related to in vitro experimenta-ion must be kept in mind. These include use of artificialtimulants, such as LPS and PMA; inability to prolongnfection models to periods that might allow more complexnteractions between various proinflammatory and anti-nflammatory processes to surface, if at all possible; and thenderstanding that clinical significance of mathematicallyignificant findings might not be truly known outside ofhe complex interactions that exist in vivo.

In view of the complexity and nature of opposing find-ngs now shown at both the cellular and clinical level, whathould treating physicians do to ensure the best outcomesor their patients? We believe that our scientific findings,ombined with recent clinical findings, provide an examplehere rigid protocols do not substitute for expert physician

udgment and experience, especially when the availableody of evidence suddenly changes. Conventional insulinegimens, which traditionally lower glucose serum concen-rations in a moderate, yet timely fashion, largely avoidangerous complications, such as hypoglycemia, whichight not be detected in over-rapid glucose-lowering pro-

ocols. Conventional regimens thereby avoid “roller-oaster” glucose levels and provide more controlled main-enance regimens.28 As a final point, glucose levels shoulderhaps be adjusted to patients’ individual baseline require-ents, rather than to generalized set values for the entire

opulation. Additional clinical and scientific studies sug-ested here are warranted to help answer one of the mostapidly evolving and debated topics of this century.

Increasing glucose concentrations exert multiple sub-tantial and opposing effects on several well-recognized andreviously tested cellular and immunologic parameters.ost importantly, the functional capacity of the innate

mmune response, reflected by formation of potent killingOS, is substantially impaired by high glucose concentra-

ions in vitro, but with a paradoxical attenuation of proin-lammatory cytokine release and enhancement of phagocy-osis. The opposing findings at the cellular level mightontribute to the clinical controversies that have thrownecent practice and protocols into disarray. Expert physi-ian judgment and experience are essential in the imminentreatment of one of medicine’s most important recent clin-cal debates, as additional cellular mechanistic studies andlinical trials continue to emerge.

uthor Contributions

tudy conception and design: Qadan, Fry, Polkcquisition of data: Qadan, Weller, Gardnernalysis and interpretation of data: Qadan, Maldonado, Fry,

Polkrafting of manuscript: Qadan, Weller, Gardnerritical revision: Maldonado, Fry, Polk

cknowledgment: A substantial degree of guidance and su-ervision were provided by our laboratory manager, the later James D Pietsch, whose contributions included the design

nd implementation of our in vitro model. We are also grate-ul to Drs Ozan Akca and Susan Galandiuk for their directionnd guidance in the study design process. Finally, we are in-ebted to the volunteers who donated blood for these studies.Motaz Qadan holds the Joint Royal College of Surgeons of

dinburgh (RCSEd)/James and Emmeline Ferguson Researchellowship.

EFERENCES

1. Sessler DI. Non-pharmacologic prevention of surgical woundinfection. Anesthesiol Clin 2006;24:279–297.

2. Bratzler DW, Houck PM, Richards C, et al. Use of antimicrobialprophylaxis for major surgery: baseline results from the NationalSurgical Infection Prevention Project. Arch Surg 2005;140:174–182.

3. Classen DC, Evans RS, Pestotnik SL, et al. The timing of pro-phylactic administration of antibiotics and the risk of surgical-wound infection. N Engl J Med 1992;326:281–286.

4. Polk HC Jr, Lopez-Mayor JF. Postoperative wound infection: aprospective study of determinant factors and prevention. Sur-gery 1969;66:97–103.

5. Qadan M, Gardner SA, Vitale DS, et al. Hypothermia and sur-gery: immunologic mechanisms for current practice. Ann Surg2009;250:134–140.

6. Van den Berghe G, Wouters P, Weekers F, et al. Intensive insulintherapy in the critically ill patients. N Engl J Med 2001;345:1359–1367.

7. Wiener RS, Wiener DC, Larson RJ. Benefits and risks of tightglucose control in critically ill adults: a meta-analysis. JAMA2008;300:933–944.

8. Finfer S, Chittock DR, Su SY, et al. Intensive versus conven-tional glucose control in critically ill patients. N Engl J Med2009;360:1283–1297.

9. Vowells SJ, Sekhsaria S, Malech HL, et al. Flow cytometric anal-ysis of the granulocyte respiratory burst: a comparison study offluorescent probes. J Immunol Methods 1995;178:89–97.

0. Hopf HW, Hunt TK, West JM, et al. Wound tissue oxygentension predicts the risk of wound infection in surgical patients.Arch Surg 1997;132:997–1004.

1. Miles AA, Miles EM, Burke J. The value and duration of defencereactions of the skin to the primary lodgement of bacteria. Br JExp Pathol 1957;38:79–96.

2. Wenisch C, Narzt E, Sessler DI, et al. Mild intraoperative hy-pothermia reduces production of reactive oxygen intermediates

1

1

1

1

1

1

1

2

2

2

2

2

2

2

2

2

974 Qadan et al Glucose and Surgical Sepsis J Am Coll Surg

by polymorphonuclear leukocytes. Anesth Analg 1996;82:810–816.

3. Smith JA, Weidemann MJ. Further characterization of the neu-trophil oxidative burst by flow cytometry. J Immunol Methods1993;162:261–268.

4. Beutler B, Milsark IW, Cerami AC. Passive immunizationagainst cachectin/tumor necrosis factor protects mice from le-thal effect of endotoxin. Science 1985;229:869–871.

5. Tracey KJ, Fong Y, Hesse DG, et al. Anti-cachectin/TNF mono-clonal antibodies prevent septic shock during lethal bacterae-mia. Nature 1987;330:662–664.

6. Van der Poll T. Immunotherapy of sepsis. Lancet Infect Dis2001;1:165–174.

7. Arndt P, Abraham E. Immunological therapy of sepsis: experimentaltherapies. Intensive Care Med 2001;27[Suppl 1]:S104�S115.

8. Lekkou A, Karakantza M, Mouzaki A, et al. Cytokine produc-tion and monocyte HLA-DR expression as predictors of out-come for patients with community-acquired severe infections.Clin Diagn Lab Immunol 2004;11:161–167.

9. Dluhy RG, McMahon GT. Intensive glycemic control in theACCORD and ADVANCE trials. N Engl J Med 2008;358:2630–2633.

0. Gerstein HC, Miller ME, Byington RP, et al. Effects of intensiveglucose lowering in type 2 diabetes. N Engl J Med 2008;358:2545–2559.

1. Bozkurt B, Kribbs SB, Clubb FJ Jr, et al. Pathophysiologically

relevant concentrations of tumor necrosis factor-alpha promote

progressive left ventricular dysfunction and remodeling in rats.Circulation 1998;97:1382–1391.

2. Kubota T, McTiernan CF, Frye CS, et al. Dilated cardiomy-opathy in transgenic mice with cardiac-specific overexpres-sion of tumor necrosis factor-alpha. Circ Res 1997;81:627–635.

3. Cheadle WG, Hershman MJ, Wellhausen SR, Polk HC Jr.HLA-DR antigen expression on peripheral blood monocytescorrelates with surgical infection. Am J Surg 1991;161:639–645.

4. Turina M, Miller FN, Tucker CF, Polk HC. Short-term hyper-glycemia in surgical patients and a study of related cellular mech-anisms. Ann Surg 2006;243:845–851.

5. Ardeshna KM, Pizzey AR, Devereux S, Khwaja A. The PI3kinase, p38 SAP kinase, and NF-kappaB signal transductionpathways are involved in the survival and maturation oflipopolysaccharide-stimulated human monocyte-derived den-dritic cells. Blood 2000;96:1039–1046.

6. Dandona P, Aljada A, Mohanty P, et al. Insulin inhibits intranu-clear nuclear factor kappaB and stimulates IkappaB in mononu-clear cells in obese subjects: evidence for an anti-inflammatoryeffect? J Clin Endocrinol Metab 2001;86:3257–3265.

7. Ngoh GA, Jones SP. New insights into metabolic signaling andcell survival: the role of beta-O-linkage of N-acetylglucosamine.J Pharmacol Exp Ther 2008;327:602–609.

8. Hirsch IB. Sliding scale insulin—time to stop sliding. JAMA

2009;301:213–214.