I

O

A

CB

a

b

I

Tcn(ebc

HT

0d

ntensive and Critical Care Nursing (2007) 23, 137—144

RIGINAL ARTICLE

chieving tight glycaemic control

arol Ball a,b,∗, Karen de Beera, Amanda Gomma,arbara Hickmana, Peta Collinsa

Royal Free Hampstead NHS Trust, London, UKCity University London, UK

Accepted 17 November 2006

KEYWORDSICU;Blood glucose;Insulin;Hypoglycaemia

Summary The implementation of tight glycaemic control (TGC) is becomingaccepted best practice within intensive care units throughout the world. It is recom-mended by the Surviving Sepsis Campaign and is included in the sepsis care bundle.The major impact of TGC is currently thought to be associated with reduced mor-bidity and mortality. The process of achieving TGC is, however, not without risk. Inparticular, the need for frequent, accurate blood glucose measurement and the pos-sibility of prolonged, unrecognised hypoglycaemia are of concern. There is also thepotential for patients who exhibit significant insulin resistance to require the admin-istration of large amounts of insulin. The transfer of patients from the intensive care

unit to the operating theatre or for computerised tomography during intensive insulintherapy is also hazardous. The purpose of this paper is to describe a series of nurseled pilot studies which aimed to introduce the process of TGC whilst maintainingpatient safety. The results demonstrate the effectiveness of a staged approach andthe achievement of TGC.

sevie

2riaId

© 2007 Published by El

ntroduction

ight glycaemic control (TGC), within the intensiveare unit (ICU), has been defined as the mainte-ance of blood glucose between 4.4 and 6.1 mmol/l80—110 mg/dl) (Van den Berghe et al., 2001). Its

ffect in reducing mortality and improving mor-idity has been demonstrated in surgical intensiveare unit (ICU) patients (Van den Berghe et al.,

∗ Corresponding author at: Intensive Care Unit, Royal Freeampstead NHS Trust, Pond St., London NW3 2PQ, UK.el.: +44 20 7472 6137.

E-mail address: Carol.Ball@royalfree.nhs.uk (C. Ball).

fD

L

To

964-3397/$ — see front matter © 2007 Published by Elsevier Ltd.oi:10.1016/j.iccn.2006.11.007

r Ltd.

001), reduced morbidity in medical patients, andeduced mortality, if stay in the intensive care units for 3 days or more (Van den Berghe et al., 2006a)nd reduced morbidity in mixed medical—surgicalCU patients (Krinsley and Jones, 2006). It has alsoemonstrated substantial cost savings. These rangerom D 1232/patient (Krinsley and Jones, 2006) to2638 (Van den Berghe et al., 2006b).

iterature review

he ability of TGC to produce these effects is basedn a number of theories. TGC is thought to be

dobfiidr1IldtwtbOgptati(aats

cc(eeaoanartratc(pt

sts

138

influential in combating stress-induced hypergly-caemia prevalent in the critically ill. The stressresponse involves an increase in counter-regulatoryhormones, e.g. growth hormone, cortisol and cat-echolamines which produce a rise in availableglucose through the process of gluconeogenesis.The production of counter-regulatory hormonesalso enhances insulin resistance in cells whereinsulin is responsible for the transport of glucoseinto the cell. There are however cells which takeup glucose independent of insulin relying on otherglucose transporters. Glucose Transporter 2 (GLUT2) allows glucose to enter the cell directly untilequilibrium is reached with the level of extra-cellular glucose. Cells are then prone to glucosetoxicity if extracellular levels of glucose are high(Van den Berghe, 2004). Those cells most likelyto be affected are hepatocytes, renal tubularcells, pancreatic b cells and the gastro-intestinalmucosa. The transport of glucose into some cellsis also up-regulated by GLUT 3 and GLUT 4 in thepresence of hypoxia and cytokine release, in par-ticular, angiotensin 2 and endothelin 1 (Quinn andMcCumbee, 1998; Sanchez-Alvarez et al., 2004).The cells likely to be affected in this instance arethose of the endothelium, alveolar epithelial cellsand vascular smooth muscle. There is therefore asymbiosis of insulin resistance and glucose toxic-ity disabling normal cellular growth and metabolicprocesses. Disturbed cellular energy metabolismhas been associated with multiple organ failureand complications such as nosocomial infection andnewly acquired renal injury (Van den Berghe et al.,2001, 2006a).

A number of studies have been undertaken since2001 which have addressed the implementation andeffectiveness of TGC within the ICU. A system-atic review undertaken by Meijering et al. (2006),published prior to Van den Berghe et al.’s (2006)second study, identified 24 papers reporting workin this area. Of these nine were undertaken in theICU. Blood glucose control ranges were from 4 to8.3 mmol/l within the ICU population possibly rep-resenting the difficulty those working in ICU had inattaining the very tight control advocated by Vanden Berghe et al. (2001). Two were performed incardiothoracic units and a further two in only crit-ically ill diabetic patients. Five papers remainedwhich were relevant to our mixed patient popu-lation (Van den Berghe et al., 2001; Laver et al.,2004; Krinsley, 2004; Goldberg et al., 2004; Kanjiet al., 2004). A further paper was included from

our review of the literature, that of Orford et al.(2004).

The papers were utilised in the following man-ner. First as a guide in the development and use of

lmpt

C. Ball et al.

ynamic insulin protocols and to establish a rangef glycaemic control which was achievable in ausy ICU with no external funding, e.g. researchunding. Meijering et al. (2006) found that dynamicnsulin protocols were more effective in achiev-ng TGC than sliding scales. In a sliding scale theosage of insulin is the same and dependent on theange, i.e. if the blood glucose is between 8 and0 mmol/l then 2 units of insulin are administered.n a dynamic protocol the previous blood glucoseevels influence the rate, i.e. the amount of insulinelivered is dependent on the difference betweenhese two levels. Second to identify the criteriahich would indicate TGC had been achieved. In

he main four criteria were found, those of meanlood glucose values (Van den Berghe et al., 2001;rford et al., 2004; Krinsley, 2004) median bloodlucose values (Laver et al., 2004), the extent (inercentage terms) to which patients were withinhe targeted blood glucose range (Goldberg etl., 2004; Krinsley, 2004) and finally, the timeaken to capture TGC from the onset of intensivensulin therapy (Goldberg et al., 2004). Kanji et al.2004) identified the number of hours/day TGC wasttained but did not identify the mean. This did notppear an appropriate measure of overall control sohis measure was not used as a comparator for ourtudy.

The process of achieving TGC is complex andarries associated risk (Orford et al., 2004). Theomplexity lies in educating all intensive care unitICU) staff in the use of clinical guidelines tonsure uniformity in commencing TGC and adher-nce to the process of insulin delivery (Laver etl., 2004). Risk is associated with the incidencef hypoglycaemia, particularly if this is prolongednd unrecognised as it has the potential to causeeurological damage (Finney et al., 2003). Largemounts of insulin may also be required if cellularesistance to insulin is high and there is the poten-ial for blood glucose levels to fall precipitouslyesulting in profound hypoglycaemia. Problems maylso arise when patients require transfer fromhe ICU and the close regulation of blood glu-ose is not possible due to radiological hazardcoronary angiography) or the ICU nurse cannot beresent (e.g. surgical debridement in the operatingheatre).

The introduction of TGC involved three pilottudies. Key areas of development in the firstwo were the refinement of the dynamic insulincale, the appropriateness of the clinical guide-

ines addressing inclusion/exclusion criteria and theanagement of delayed gastric emptying. The thirdilot study represents an evaluation of our effec-iveness in achieving TGC.

A

T

STtstBn

DT(imst

PTtittgLcttbat(agp

btaBfWccsftbfgtcttr

(mp

iostbaittmsb

lpatwothnnape

T

Heig

dash(g1ndt(ivb

chieving tight glycaemic control

he pilot studies

ettinghe context was a general ICU of 24 beds admit-ing both surgical and medical patients. Regionalpecialties include liver transplantation and ter-iary referral of traumatic brain injury patients.ed occupancy is in the region of 92% and theurse:patient ratio is predominantly 1:1.

esignhe pilot studies took place over an 8-month periodOctober 2005—May 2006). The introduction andmplementation of TGC were nurse-led. Regulareetings were held with consultant intensivists and

enior nurses to provide an update on progress ando serve as a platform for discussion.

reparationhe first stage involved a practice developmenteam of nurses analysing the evidence base, review-ng published clinical guidelines and algorithms forhe titration of insulin and contact with other ICUso ascertain local methods of implementation. Theuidelines and dynamic insulin scale developed byaver et al. (2004) were judged to have the mostlinical relevance as the visual representation ofhe dynamic insulin scale was easier to follow thanextual guidance. In this study, TGC was defined aslood glucose between 4 and 7 mmol/l and this wasdopted. Like many ICUs the resources available inhe seminal research studies associated with TGCVan den Berghe et al., 2001, 2006a) were not avail-ble to us and therefore the chosen range of bloodlucose appeared achievable without increasing theotential risk of hypoglycaemia.

The main addition to the guidelines developedy Laver et al. (2004) at this time was to includehe delivery of 20% glucose (50 ml/h), to maintainconstant glucose load as advocated by Van den

erghe et al. (2001). The action to take when trans-erring a patient out of the ICU was also included.e were concerned that if insulin administration

ontinued when glucose, in the form of 20% glu-ose, parenteral or enteral nutrition, had beentopped then hypoglycaemia could occur. There-ore TGC was discontinued during the period ofransfer. Internal transfer from our ICU is commonoth for Computerised Tomography, tracheostomyormation and coronary angiography. Arterial bloodlucose samples were to be tested each hour unlesshe patient had been within range for 2 h when it

ould be reduced to 2 hourly. An algorithm for con-inuous enteral feeding was also developed, againo reduce the possibility of hypoglycaemia duringest periods. A standard concentration of insulin

ie

g

139

Actrapid, Novo Nordisk) of 1unit/ml diluted in nor-al saline was used and delivered via a syringeump.

The second stage of preparation involved thendividual training of all nursing staff in the usef the glucometer. Point of care testing was cho-en as it would have been impossible for sampleso be tested on the blood gas analyser on an hourlyasis due to the geography of the ICU. This, and thepplication of the clinical guidelines and dynamicnsulin scale, was undertaken by four members ofhe practice development team to reduce varia-ion in teaching methods. An assessment was alsoade of nurses’ ability to use the dynamic insulin

cale through the administration of a scenario-ased test.

The inclusion and exclusion criteria were estab-ished and remained the same throughout theilot study periods. Inclusion criteria includedll patients who would require invasive ven-ilation for more than 24 h. Exclusion criteriaere mask continuous positive airways pressurer bi-level positive airways pressure, ventila-ion for less than 24 h, diabetic ketoacidosis oryperglycaemic hyperosmolar syndrome, requiringaso-gastric administration of phenytoin due to theeed to stop enteral feeding and children under thege of 16. Neurosurgical patients were excludedending the neurosurgeons agreement to the deliv-ry of intravenous 20% glucose.

he first pilot study

ypoglycaemia was defined as <2.2 mmol/l (Osoriot al., 1999). The key factors evaluated were thencidence of hypoglycaemia and adherence to theuidelines.

The pilot study was stopped after only two weeksue to the erratic swings in blood glucose associ-ted with hypoglycaemia observed in patients witheverely impaired hepatic function. Liver failureas now been identified by Van den Berghe et al.2006a) as an independent risk factor for hypo-lycaemia in medical ICU patients (odds ratio OR.62; 95% confidence interval 1.01—2.60). This isot completely understood as a review of the pre-isposing factors for hypoglycaemia demonstratedhat liver failure was not an independent factorVriesendorp et al., 2006). However, the same studyndicated an association with continuous veno-enous haemofiltration (CVVH) with bicarbonateased substitution fluid (OR 14; 95% confidence

nterval 1.8—1.06) and this may have been theffect we observed.

It was also found that occasionally the bloodlucose would reduce to within range without any

aeiatAaur

T

Ivsw

DBnttdowrwp

DDfbem2prfia

R

Ttmftg

140

reduction in the amount of insulin being deliv-ered. This resulted in hypoglycaemia (n = 3) andfrequent resuscitation with 50% glucose for patientswith a blood glucose below 3.9 mmol/l. It was alsonoted that though enteral feeding was now deliv-ered continuously nutritional delivery was impairedif the patient subsequently developed delayed gas-tric emptying. Delayed gastric emptying appearedto be associated with deterioration in the patient’scondition, e.g. hypotension requiring inotropic sup-port.

The second pilot study

All patients with impaired hepatic function wereexcluded and an algorithm for the proactivemanagement of delayed gastric emptying was intro-duced to avoid the potential for hypoglycaemia.The dynamic insulin scale was changed to state‘halve the amount of insulin being given if bloodglucose drops to 4—5.5 mmol again to avoid hypo-glycaemia’. The key factors evaluated were theincidence of hypoglycaemia, the daily haemoglobin(due to the number of blood samples required),adherence to the guidelines and the impact on nurs-ing practice.

Sixteen patients were recruited into the sec-ond pilot study (03.01.06—30.01.06) with no morethan four patients being in the study at any onetime. Overall there was only one episode of hypo-glycaemia. There were however 32 episodes ofblood glucose being <3.9 requiring resuscitationwith 50% glucose revealing an incidence in theregion of 5%. These episodes were associated withtotal parenteral nutrition bag changes and glu-cose 20% being stopped when enteral feed wasestablished and not reduced by 10ml/h as per theguideline. Blood glucose results over 7 mmol/l andhigh insulin requirements (17 units/h) were relatedto the administration of Linizolid (due to the highglucose content of the diluent) and stopping TGCfor internal transfers. Labile blood glucose resultswere associated with:

• Worsening condition of patient (i.e. if haemo-dynmically unstable blood glucose was alsounstable).

• Delayed gastric emptying.• Noradrenaline and hydrocortisone administra-

tion.• User error, halving the insulin rate when this was

not indicated and not halving it when it was.

There was no increase in blood product usedespite the increase in arterial blood samples forglucose measurement. Evaluation of workload bythe nurses involved (n = 32) indicated that the

aeia

C. Ball et al.

dministration of TGC was manageable. Experi-nced senior nurses did report some difficultiesn administering TGC if the patient was unstablend required multiple organ support. As a result ofhese findings TGC was introduced to the whole ICU.fter 1 month a final evaluation was undertaken toscertain the degree of glycaemic control achieved.sing the criteria identified in the literatureeview.

he third pilot study

nclusion criteria were all patients likely to requireentilation for more than 24 h, now including neuro-urgical patients but continuing to exclude patientsith severe hepatic impairment.

ata collectionlood glucose, units of insulin delivered, mcg/minoradrenaline, administration of hydrocortisone,ype of feed delivered and presence of delayed gas-ric emptying were collected for each hour on aaily basis by two members of the practice devel-pment team. Quality assessment of data collectionas undertaken by a different member of the team

etrospectively to ensure accuracy. Fifty patientsere enrolled in to the evaluation over a 2-montheriod (15.03.06—15.05.06).

ata analysisata were transcribed onto an excel spreadsheetor each patient. This was then analysed for meanlood glucose values over time (Van den Berghet al., 2001; Orford et al., 2004; Krinsley, 2004)edian blood glucose values over time (Laver et al.,

004), the extent (in percentage terms) to whichatients were within the targeted blood glucoseange (Goldberg et al., 2004; Krinsley, 2004) andnally, the time taken to capture TGC (Goldberg etl., 2004).

esults

he average age of the patients participating inhe evaluation was 50, 31 (62%) were male and theean APACHE II was 17.5. TGC was administered

or a total of 335. 7 patient days which equatedo 7189 h of insulin administration and 6424 bloodlucose measurements.

Table 1 indicates the mean and median results

chieved in comparison to those in the published lit-rature. Van den Berghe et al. (2001) assumed TGCf the mean 6am blood glucose was between 4.4nd 6.1 mmol/l for the sample group and a mean

Achieving tight glycaemic control 141

Table 1 Comparison of measures of tight glycaemic control

Study Royal FreeHospital(2006)

Van denBerghe etal. (2001)

Orford et al.(2004)

Krinsley(2004)

Laver et al.(2004)

Mean 06.00 (a.m.) bloodglucose

6.86 5.7(+/− 1.1 mmol/l)

Lowest mean blood glucosebetween 06.00 and 09.00a.m.

6.06 6.45 (+/− 2.1 mmol/l)

oca(0t6gdtBtaobwive

wwtwrrrietb

i8r

scmweaadre

bco

S

Ieoa(

Overall mean blood glucose 7.08

Median blood glucose 6.5

f 5.7 mmol/l was achieved (n = 1548). Blood glu-ose control was not reported in Van den Berghe etl’s 2006 study of medical patients. Orford et al.2004) selected the lowest blood glucose between6.00 and 09.00 a.m. and calculated the mean forhe sample group (n = 148) and achieved control at.45 mmol/l. Laver et al. (2004) established a bloodlucose/time curve, calculated the mean and thenivided this by the total time spent in intensive careo give the mean blood glucose for that admission.ecause the whole patient stay was chosen andime to starting TGC was not normally distributedmedian result of 6.2 mmol/l was achieved. The

verall mean (i.e. not related to a specific timeut over time) achieved in Krinsley’s (2004) studyas 7.2 mmol/l. In comparison with our data (RFH)

t can be seen that TGC was achieved using thesearious criteria, although the overall mean was inxcess of our target range.

Fig. 1 describes the percentage of time TGCas achieved. 3956 (61.6%) blood glucose measuresere within the 4—7 mmol range. This compares

o the results achieved by Goldberg et al. (2004)here 66% samples were between an established

ange of 4.4—7.7 mmol/l. Also demonstrated in ouresults was an incidence of 484 (7.5%) blood glucoseeadings greater than 10 mmol/l. In comparison,

ncidences of 3 and 7% have been reported by Orfordt al. (2004) and Goldberg et al. (2004), respec-ively. However, better results have been achievedy Lonergan et al. (2006) using specialised relative

Figure 1 Glycaemic control.

s(frooarciAa

dae

7.2

6.2

nsulin and nutrition tables (SPRINT). In this study,9% of measurements were within the 4—7 mmol/lange.

Fig. 1also demonstrates 170 (2.6%) blood glucoseamples were below 3.9 mmol/l requiring 50% glu-ose resuscitation. The number of blood glucoseeasurements below 2.2 mmol/l in our evaluationas 3 (0.05%). This is considerably lower than thatxperienced by other studies. Van den Berghe etl. (2001, 2006a) reported an incidence of 5.1nd 18.7%, respectively whilst Orford et al. (2004)escribed one of 2.7%. Lonergan et al. (2006)eported no hypoglycaemic episodes with the low-st recording being 3.2 mmol/l.

The median time to achieve TGC was reportedy one other study (Goldberg et al. (2004) at 9 h. Inomparison we achieved TGC within a median timef 5 h (mean = 6.48).

ummary—–third time lucky?

n comparison to the majority of the published lit-rature reflecting patient populations similar to ourwn we achieved TGC in terms of the percent-ge of time patients were within our target rangeFig. 1), the mean blood glucose based on cho-en time frames, the overall median blood glucoseTable 1) and the time to capturing the target rangeor blood glucose. The overall mean blood sugaresult was just under one decimal point in excessf our target range but in relation to an incidencef hypoglycaemia at 0.05% this was thought to becceptable. The use of the SPRINT protocol hasevealed interesting results and considerable suc-ess in achieving TGC but the number of patientsncluded in the study, whilst having a high meanPACHE score of 22, was small (n = 12) (Lonergan etl., 2006).

Our experience of implementing TGC howeveremonstrated a number of issues not previouslyddressed in the literature. Once patients werextubated, indicating recovery from critical illness,

142 C. Ball et al.

aem

o

Figure 2 Labile nature of glyc

and TGC discontinued blood glucose readings were

stable and easily controlled even in diabetic (type1 and type 2) patients. This is important because ifthe entire period of insulin administration is usedto estimate TGC (Laver et al., 2004), rather than

ta

l

Figure 3 High insulin resista

ic control in the ‘critically ill’.

nly the period of intensive insulin therapy, con-

rol may be assumed to be more robust than itctually is.

We also found blood glucose measures wereabile and difficult to control despite hourly titra-

nce in type 2 diabetes.

A

tttdngsl4ibmt

rraettph1Fmhtc(

D

OaltupAIbiwduh(coe(fdd

rF

e0lscdnaaso

mGrsehss

C

Ttwrcwecwtcrtr

ssdcGiocac

chieving tight glycaemic control

ion of insulin when ventilation was combined withhe need for nor-adrenaline and hydrocortisone,he period of time when the patient might beefined as being in the ‘critical’ phase of their ill-ess. The lack of control and diversity of bloodlucose results is demonstrated in Fig. 2. Thequares represent blood glucose and reveal howittle time was spent within the target range of—7 mmol/l (y axis Figs. 2 and 3) despite the admin-stration of 17—18 units of insulin as determinedy the dynamic insulin scale (indicated by the dia-ond shapes). The overall mean blood glucose for

his patient was 8.57 mmol/l.Precipitously high levels of insulin were also

equired in some patients. Fig. 3 demonstrates aequirement for insulin in excess of 50units/h for

period of 10 h and then for a further 3 h. Thisntailed a syringe change every hour together withhe administration of other drugs with observa-ions and other therapeutic interventions to beerformed. The maximum amount of insulin whichad to be delivered during the third pilot study was30 units/h. This was in a non-diabetic patient.or this reason, the safe implementation of TGCandates a 1:1 nurse:patient ratio. An extremely

igh level of vigilance had to be maintainedo avoid the potential for profound hypogly-aemia which, as it can be seen, was avoidedFig. 3).

iscussion

ur results demonstrate TGC can be achieved safelynd effectively utilising criteria established in theiterature. However, controversy still surrounds theherapy (Watkinson et al., 2006). Debate contin-es regarding the preponderance of cardiac surgicalatients in the main surgical group and the lowPACHE score of 9 (Van den Berghe et al., 2001).

t has been argued that had the surgical studyeen carried out at the same time as the med-cal cohort (Van den Berghe et al., 2006a) thereould be no difference in sepsis related mortalityue to the increase in statin use which may mod-late sepsis (Wise, 2006; Kruger et al., 2006). Postoc analysis of the data from Van den Berghe et al’s2001) study has also demonstrated that improvedontrol of lipids may explain the reduction in multi-rgan failure and improved survival to a greaterxtent than that of glycaemic control and insulin

Mesotten et al., 2004). Insulin itself is a power-ul anabolic hormone and may be more therapeuticuring recovery from critical illness rather thanuring the ‘critical’ phase of critical illness, e.g.

ocdp

143

equiring nor-adrenaline and hydrocortisone (seeig. 2) (Nunnally, 2005).

The time of day chosen by both Van den Berghet al. (2001) and of Orford et al. (2004) of 6 and6.00—09.00 a.m., respectively can also be chal-enged. It is usual for early morning fasting bloodugars to be used to assess the degree of glycaemicontrol in diabetic patients. However, the need toeliver continuous enteral feed, total parenteralutrition or 20% glucose negates the choice of times patients will not be fasting. Possibly the over-ll mean blood glucose or the percentage of timepent within the target range are better measuresf glycaemic control.

Hypoglycaemia is also seen as a major impedi-ent to the implementation of TGC. In 2005 theerman SepNet group suspended a multi-centre

andomised controlled trial in patients with sep-is due to the incidence of hypoglycaemia (12.1%,xperimental group versus 2.1%, control group). Itas also been argued that the restriction of glucoseolutions in achieving glycaemic control may be farafer than the use of insulin (Nunnally, 2005).

onclusion

he control achieved in our study was comparableo that described in the literature. In particular,e had a very low incidence of hypoglycaemia, a

educed percentage of patients with a blood glu-ose in excess of 10 mmol/l and 60% of patientsere in the control range of 4—7 mmol/l. In ourxperience, it was difficult to maintain blood glu-ose measurements within the range of 4—7 mmol/lhen patients were in the ‘critical’ phase of

heir illness, i.e. requiring nor-adrenaline, hydro-ortisone. Some patients presented considerableisk in relation to the degree of insulin resis-ance demonstrated and the amount of insulinequired.

The future of TGC, and its therapeutic benefit, istill far from certain. It is unusual for a single centretudy to have the impact of that achieved by Vanen Berghe et al. (2001). In 2007, two large multi-entre studies will be completed (NICE-SUGAR andLUControl). Hopefully on their publication the

mportant questions surrounding the intensive usef insulin will be answered. In particular those asso-iated with lipid versus glucose control, the mostppropriate method of assessing the degree of gly-aemic control achieved, decreasing the incidence

f hypoglycaemia and the period of time when gly-aemic control may be most effective, that is to say,uring the critical phase of illness or in the recoveryeriod.

144

Acknowledgements

The authors would like to thank the members ofthe practice development team who participatedin the implementation of TGC: Kanta Velani, SamLemmon, Fiona Donaldson, Vivienne Reddy, AndrewThomlinson, Aileen Ferrais, Anna Valmores, ElsaPla-Canalda and all nursing staff who worked sohard to achieve glycaemic control.

References

Finney SJ, Zekveld C, Elia A, Evans TW. Glucose control andmortality in critically ill patients. JAMA 2003:2041—7.

Goldberg PA, Siegal MD, Sherwin RS, Halickman JI, Lee M, BaileyVA, Lee SL, Dziura JD, Inzucchi SE. Implementation of a safeand effective insulin infusion protocol in a medical intensivecare unit. Diabetes Care 2004;27:461—7.

Kanji S, Singh A, Tierney M, Meggison H, McIntyre L, Herbert PC.Standardization of intravenous insulin therapy improves theefficiency and safety of blood glucose control in critically illadults. Intensive Care Medcine 2004;30:804—10.

Krinsley JS. Effect of an intensive glucose management protocolon the mortality of critically ill adult patients. Mayo Clin Proc2004;79:992—1000.

Krinsley JS, Jones RL. Cost analysis of glycaemic control in crit-ically ill adult patients. Chest 2006;129:644—50.

Kruger P, Fitzsimmons K, Cook D, Jones M, Nimmo G. Statintherapy is associated with fewer deaths in patients with bac-teraemia. Intensive Care Med 2006;32:75—9.

Laver S, Preston S, Turner D, McKinstry C, Padkin A. Implement-ing intensive insulin therapy: development and audit of thebath insulin protocol. Anaesth Intensive Care 2004;32:311—6.

Lonergan T, Compte AL, Willacy M, Chase JG, Shaw GM, HannCE, et al. A pilot study of the SPRINT protocol for tight gly-

caemic control in critically ill patients. Diab Technol Ther2006;8:449—62.

Meijering S, Corstjens AM, Tulleken JE, Meertens JHJM, Zijl-stra JG, Ligtenberg JJM. Towards a feasible algorithm fortight glycaemic control in critically ill patients: a systematic

C. Ball et al.

review of the literature. Critical Care 2006;10:1/R19 OpenAccess (http://ccforum.com).

Mesotten D, Swinnene JV, Vanderhoydonc F, Wouters PJ, Van denBerghe G. Contribution of circulating lipids to the improvedoutcome of critical illness by glycaemic control with inten-sive insulin therapy. J Clin Endocrinol Metab 2004;89:219—26.

Nunnally ME. Con: tight perioperative glycaemic control:poorly supported and risky. J Cardiothorac Vasc Anaesth2005:689—90.

Orford N, Stow P, Green D, Corke C. Safety and feasibility of aninsulin adjustment protocol to maintain blood glucose con-centrations within a narrow range in critically ill patientsin an Australian Level 3 adult intensive care unit. Crit CareResuscitation 2004;6:92—8.

Osorio I, Arafah BM, Mayor C, Troster AI. Plasma glucose alonedoes not predict neurologic dysfunction in hypoglycaemicnon-diabetic patients. Ann Emerg Med 1999;33:291—8.

Quinn LA, McCumbee WD. Regulation of glucose transport byangiotensin 2 and glucose in cultured vascular smooth musclecells. J Cell Physiol 1998:94—102.

Sanchez-Alvarez R, Tabernero A, medina JM. Endothelin-1 stim-ulates the translocation of both glucose transporter andhexokinase in astrocytes: relationship with gap junction com-munication. J Neurochem 2004;89:703—14.

Van den Berghe G, Wouters P, Weekers F, Verwaest C, BruyninckxF, Schetz M, et al. Intensive Insulin therapy in critically illpatients New England. J Med 2001;345:1359—67.

Van den Berghe G. How does blood glucose control with insulinsave lives in intensive care? J Clin Invest 2004;114:1187—95.

Van den Berghe G, Wilmer A, Hermans G, Meersserman W,Wouters PJ, Milants I, et al. Intensive insulin therapy in themedical ICU New England. J Med 2006a;354:449—61.

Van den Berghe G, Wouters PJ, Kesteloot K, Hilleman DE. Analy-sis of healthcare utilisation with intensive insulin therapy incritically ill patients. Crit Care Med 2006b;34:612—6.

Vriesendorp TM, an Santen S, deVries JH, de Jonge E, RosendaalFR, Schultz MJ, et al. Predisposing factors for hypoglycaemiain the intensive care unit. Crit Care Med 2006;34:246—8.

Watkinson P, Barber VS, Young JD. Strict glucose control in thecritically ill (editorial). BMJ 2006;332:865—6.

Wise M. Strict glucose control reduces morbidity and costs. BMJ2006;332(April):865—6 [rapid responses for Watkinson et al.,2006].