Secondary Data/Meta Analysis

11

Using Secondary Data in Using Secondary Data in Statistical AnalysisStatistical Analysis

Charles Natanson M.D.Charles Natanson M.D.Critical Care Medicine DepartmentCritical Care Medicine Department

Clinical CenterClinical CenterNational Institutes of HealthNational Institutes of Health

Bethesda, MDBethesda, MD

MetaMeta--analysis Definitionanalysis Definition•• Glass 1976, “the statistical analysis of a large Glass 1976, “the statistical analysis of a large

collection of results from individual literature collection of results from individual literature for the purpose of integrating their respective for the purpose of integrating their respective findings.”findings.”

•• Two basic purposes: Two basic purposes: p pp p1.1. Determine if similar treatment effects exist for a Determine if similar treatment effects exist for a

therapy in independent studies to estimate a net therapy in independent studies to estimate a net effect for this therapyeffect for this therapy

2.2. Alternatively, if treatment effects differ Alternatively, if treatment effects differ substantially for a therapy among independent substantially for a therapy among independent studies, to examine factors that may explain studies, to examine factors that may explain these differing effectsthese differing effects

Education Research 1976Education Research 1976

Techniques of MetaTechniques of Meta--AnalysisAnalysis

Step One: Formulating the QuestionStep One: Formulating the Question•• Validity and importance are contingent on Validity and importance are contingent on

this stepthis step•• Poorly conceived research hypothesis will Poorly conceived research hypothesis will

usually lead to an analysis of dubious valueusually lead to an analysis of dubious value

22


Step Two: Defining Eligibility CriteriaStep Two: Defining Eligibility Criteria•• Protocols for study inclusion should be Protocols for study inclusion should be

prospective, systematic, and explicitprospective, systematic, and explicit•• Ideally, randomized trials similar in diagnosis, Ideally, randomized trials similar in diagnosis,

outcome, patient characteristics, and outcome, patient characteristics, and treatment groupstreatment groups


Step Two: Defining Eligibility Criteria Cont’dStep Two: Defining Eligibility Criteria Cont’d•• Including all available studies, regardless of Including all available studies, regardless of

size, design, or quality results in an analysis size, design, or quality results in an analysis that is broadly representative but maythat is broadly representative but maythat is broadly representative but may that is broadly representative but may compromise accuracycompromise accuracy

•• Alternatively, exclusion of poorly done Alternatively, exclusion of poorly done studies may increase the statistical validity studies may increase the statistical validity but limit the ability to generalize findingsbut limit the ability to generalize findings


Step Three: Identifying Studies and Data Step Three: Identifying Studies and Data AbstractionAbstraction

•• Usually begins with a search of online Usually begins with a search of online databases such as MEDLINE Currentdatabases such as MEDLINE Currentdatabases such as MEDLINE, Current databases such as MEDLINE, Current Contents, Best Evidence, Cochrane, and Contents, Best Evidence, Cochrane, and HealthSTARHealthSTAR

•• Title and abstract perused to exclude studiesTitle and abstract perused to exclude studies

33


Step Three: Identifying Studies and Data Step Three: Identifying Studies and Data Abstraction Cont’dAbstraction Cont’d

•• Full texts of the remaining articles retrieved Full texts of the remaining articles retrieved and thoroughly studiedand thoroughly studiedand thoroughly studiedand thoroughly studied

•• Reference lists of these articles are reviewedReference lists of these articles are reviewed•• Once a study selected for inclusion, data Once a study selected for inclusion, data

should be extracted by more than one should be extracted by more than one reviewer onto structured formsreviewer onto structured forms


Step Four: AnalysisStep Four: Analysis•• A common measure of treatment effect must A common measure of treatment effect must

be determinedbe determined•• Fixed versus random effect model used to Fixed versus random effect model used to

combine datacombine data


Step Four: Analysis Cont’dStep Four: Analysis Cont’d•• Cochran’s Q statistic and ICochran’s Q statistic and I22 calculatedcalculated•• Consider metaregression when IConsider metaregression when I22 > 30% and > 30% and

P < 0.10P < 0.10•• Publication bias examinedPublication bias examined

–– Funnel PlotFunnel Plot

44


Step Five: Reporting and Interpreting ResultsStep Five: Reporting and Interpreting Results•• To improve overall quality of reporting for To improve overall quality of reporting for

metameta--analysis, a checklist and a flow chart analysis, a checklist and a flow chart should be constructedshould be constructedshould be constructedshould be constructed

•• Quality of Reporting of MetaQuality of Reporting of Meta--analyses analyses (QUOROM) conference provides guidelines (QUOROM) conference provides guidelines for reporting searches, study selection, for reporting searches, study selection, validity assessment, data abstraction, study validity assessment, data abstraction, study characteristics and data synthesischaracteristics and data synthesis

Lancet, 1999.Lancet, 1999.

MetaMeta--analysis of Clinical Trials in Sepsisanalysis of Clinical Trials in Sepsis

Septic Shock ManagementSeptic Shock Management

Early recognitionEarly recognition

The right antibiotics right awayThe right antibiotics right awayg g yg g y

Rapid fluid resuscitationRapid fluid resuscitation

Judicious use of vasopressorsJudicious use of vasopressors

Promptly address removablePromptly address removable nidinidi

55

McCabeMcCabe 19621962FriedFried 19681968BryantBryant 19711971KrogerKroger 19801980IshpahaniIshpahani 19871987ChowChow 19911991WeinsteinWeinstein 19931993Vid lVid l 19961996

Benefits of Starting Appropriate Benefits of Starting Appropriate Antibiotic TherapyAntibiotic Therapy

Odds Ratio of Survival (95% CI)Odds Ratio of Survival (95% CI)0.0010.001 0.010.01 0.10.1 11 1010 100100 10001000

VidalVidal 19961996SchiappaSchiappa 19961996Leibovici Leibovici 19971997LeiboviciLeibovici 19971997LeiboviciLeibovici 19981998KulleyKulley 19991999IbrahimIbrahim 20002000FaragofaFaragofa 20032003GarnachoGarnacho--MonteroMontero 20032003HorbarthHorbarth 20032003LeoneLeone 20032003

Benefit of Early Benefit of Early versusversus Late AntibioticsLate AntibioticsAuthorAuthor YearYear NN DiagnosisDiagnosisMinerMiner 20012001 171171 MeningitisMeningitisLarcheLarche 20022002 8888 Bact/pneumoniaBact/pneumonia**HouckHouck 20042004 13,77113,771 PneumoniaPneumoniaProulxProulx 20052005 118118 MeningitisMeningitisM hM h 19971997 14 06914 069 P iP i

Odds Ratio of Survival (95% CI)Odds Ratio of Survival (95% CI)

0.010.01 0.10.1 11 1010 100100BenefitBenefitHarmHarm

* Patients with cancer * Patients with cancer

MeehanMeehan 19971997 14,06914,069 PneumoniaPneumoniaGacouinGacouin 20022002 213213 LegionellaLegionellaIreguiIregui 20062006 107107 VAPVAPLodisLodis 20032003 167167 S. aureusS. aureusKangKang 20032003 123123 P. aeruginosaP. aeruginosa

The Host Inflammatory The Host Inflammatory Response Hypothesis:Response Hypothesis:Response Hypothesis: Response Hypothesis:

What went wrong?What went wrong?

66

Pathogenesis of Septic ShockPathogenesis of Septic ShockPathogensPathogens

andandToxinsToxins

Host Defenses Host Defenses ReleaseRelease

Inflammatory Inflammatory MediatorsMediators

InfectionInfection

Shock and Shock and InjuryInjury

Multiorgan FailureMultiorgan Failureand Deathand Death

Number of Patients EnrolledNumber of Patients Enrolled00 10001000 20002000

Human Clinical Trials of AntiHuman Clinical Trials of Anti--Inflammatory Inflammatory Therapies in Sepsis Therapies in Sepsis Ranked by SizeRanked by Size

Increasing HarmIncreasing Harm Increasing BenefitIncreasing BenefitNo EffectNo Effect

0.10.1 0.30.3 0.50.5 11 2.02.0 4.04.0 8.08.0

Crit Care Med 1998Crit Care Med 1998

Odds Ratio of SurvivalOdds Ratio of Survival

95% CI95% CI



BAY x 1351BAY x 1351



AntiAnti--TNF AntibodiesTNF Antibodies


95% CI95% CI


0.10.1 0.30.3 0.50.5 11 2.02.0 4.04.0 8.08.0

MAK 195FMAK 195FBN52021BN52021

MAK 195FMAK 195F

CB0006CB0006CDP571CDP571

MAK 195FMAK 195F


77


AntiAnti--TNF AntibodiesTNF AntibodiesBAY x 1351BAY x 1351PP--5555



PP--5555

Soluble TNF ReceptorsSoluble TNF Receptors



0.10.1 0.30.3 0.50.5 11 2.02.0 4.04.0 8.08.0

MAK 195FMAK 195F

PP--8080MAK 195FMAK 195F

CB0006CB0006CDP571CDP571

MAK 195FMAK 195F



95% CI95% CI


BAY x 1351BAY x 1351PP--5555

AntrilAntrilBAY x 1351BAY x 1351


AntrilAntril

PP--5555



ILIL--1ra1raSoluble TNF ReceptorsSoluble TNF Receptors


0.10.1 0.30.3 0.50.5 11 2.02.0 4.04.0 8.08.0

MAK 195FMAK 195F


AntrilAntrilCB0006CB0006

CDP571CDP571MAK 195FMAK 195F



95% CI95% CI


BAY x 1351BAY x 1351PP--5555


BN52021BN52021BAY x 1351BAY x 1351

AntrilAntril

PP--5555



PAFraPAFraILIL--1ra1raSoluble TNF ReceptorsSoluble TNF Receptors


0.10.1 0.30.3 0.50.5 11 2.02.0 4.04.0 8.08.0

MAK 195FMAK 195FBN52021BN52021






95% CI95% CI

88


BAY x 1351BAY x 1351PP--5555


BN52021BN52021BAY x 1351BAY x 1351

IbuprofenIbuprofen

AntrilAntril

PP--5555



AntiAnti--ProstaglandinProstaglandinPAFraPAFraILIL--1ra1raSoluble TNF ReceptorsSoluble TNF Receptors


0.10.1 0.30.3 0.50.5 11 2.02.0 4.04.0 8.08.0

IbuprofenIbuprofenMAK 195FMAK 195F

BN52021BN52021




IbuprofenIbuprofenIbuprofenIbuprofen



95% CI95% CI


BAY x 1351BAY x 1351PP--5555


BN52021BN52021BAY x 1351BAY x 1351

IbuprofenIbuprofen

CPOCPO--127127

AntrilAntril

PP--5555



AntiAnti--BradykininBradykininAntiAnti--ProstaglandinProstaglandinPAFraPAFraILIL--1ra1raSoluble TNF ReceptorsSoluble TNF Receptors


0.10.1 0.30.3 0.50.5 11 2.02.0 4.04.0 8.08.0


BN52021BN52021CPOCPO--127127







95% CI95% CI


BAY x 1351BAY x 1351PP--5555


BN52021BN52021BAY x 1351BAY x 1351

IbuprofenIbuprofen

CPOCPO--127127

AntrilAntril

PP--5555



AntiAnti--BradykininBradykininAntiAnti--ProstaglandinProstaglandinPAFraPAFraILIL--1ra1raSoluble TNF ReceptorsSoluble TNF Receptors


0.10.1 0.30.3 0.50.5 11 2.02.0 4.04.0 8.08.0


BN52021BN52021CPOCPO--127127







95% CI95% CI

99

Control Arm Mortality Rates by Control Arm Mortality Rates by Type ofType of AntiAnti--inflammatory inflammatory AgentAgent

nt M

orta

lity

(nt

Mor

talit

y (±±

SEM

)SE

M) 100100

9090

8080

7070

6060

5050

Mea

n Pe

rcen

Mea

n Pe

rcen 4040

3030

2020

1010

00# of Trials# of Trials

# of# ofDeathsDeaths

# of# ofPatientsPatients 102/258102/258

N = 3N = 3256/853256/853

N = 3N = 3602/1493602/1493

N = 7N = 7218/438218/438

N = 2N = 2276/783276/783

N = 3N = 376/21076/210N = 2N = 2

Crit. Care Med 1997Crit. Care Med 1997

Preclinical Studies of Preclinical Studies of AntiAnti--inflammatory Agentsinflammatory Agents

P < 0.0001P < 0.0001

Favo

rs T

hera

pyFa

vors

The

rapy

ds R

atio

ds R

atio

10001000

100100

1010

11

BeneficialBeneficial

Am J Respir Crit Care Med 2002;166:1197 Am J Respir Crit Care Med 2002;166:1197

25255050

100100

AnimalsAnimals

Favo

rs C

ontr

olFa

vors

Con

trol

Odd

Odd 0.10.1

0.010.01

0.0010.001

0.00010.00010.010.01 0.10.1 11 1010 100100

Control Odds of DyingControl Odds of Dying

PatientsPatientsHarmfulHarmful

Prospective Animal Studies of Prospective Animal Studies of Selected AgentsSelected Agents

P < 0.0001P < 0.000110001000

100100

1010

11

ololFa

vors

The

rapy

Favo

rs T

hera

pys

Rat

ios

Rat

io

PP--80 TNF80 TNFαα soluble receptorsoluble receptorTyrosine kinase inhibitorTyrosine kinase inhibitor


0.10.1

0.010.01

0.0010.001

0.00010.0001

Favo

rs C

ontr

oFa

vors

Con

tro

Odd

Odd

0.010.01 0.10.1 11 1010 100100Control Group Odds of DyingControl Group Odds of Dying

25255050100100

RatsRatsEndotoxinEndotoxinGramGram--negativenegativeGramGram--positivepositivei.v.i.v. challengechallengePneumoniaPneumonia

DogsDogs PatientsPatients

1010

Randomized Clinical Trials in Sepsis Randomized Clinical Trials in Sepsis Divided by Severity of Illness ScoresDivided by Severity of Illness Scores

Favo

rs T

hera

pyFa

vors

The

rapy

of S

urvi

val

of S

urvi

val 1010

44

P = 0.0002P = 0.0002


Favo

rs C

ontr

olFa

vors

Con

trol

FF

Odd

s R

atio

oO

dds

Rat

io o

11

0.250.25

0.10.1

0.10.1 0.250.25 11 101044Control Odds of DyingControl Odds of Dying

Activated Protein CActivated Protein CILIL--1ra1ra

pp--5555--TNFsrTNFsr

SummarySummaryAntiAnti--Inflammatory Agents in Sepsis Inflammatory Agents in Sepsis

•• MetaMeta--analysis:analysis:–– treatment effects are small (3%), treatment effects are small (3%),

but statistically significantbut statistically significanty gy g•• MetaMeta--regression analysis:regression analysis:

–– Efficacy dependent on risk of Efficacy dependent on risk of deathdeath»»Beneficial at high risks of death, Beneficial at high risks of death, »»ineffective or harmful when risk ineffective or harmful when risk

was moderate or lowwas moderate or low

Paradox of Paradox of Corticosteroids in SepsisCorticosteroids in Sepsis

Less may have benefits, Less may have benefits, but only in sickest patientsbut only in sickest patients

1111

Corticosteroids in SepsisCorticosteroids in Sepsis

•• Investigated since the 1960sInvestigated since the 1960s

•• By early 1990s, shown to be By early 1990s, shown to be ineffective or possibly harmfulineffective or possibly harmful

•• Renewed interest and new Renewed interest and new trials over the last decadetrials over the last decade

Trials of Corticosteroids in SepsisTrials of Corticosteroids in SepsisAuthorAuthor Year PublishedYear Published Odds ratioOdds ratio

95% CI95% CI

Ann Intern Med 2004; 141:47Ann Intern Med 2004; 141:47Clin Microbiol Infect 2009; 15: 308Clin Microbiol Infect 2009; 15: 308 Odds Ratio of SurvivalOdds Ratio of Survival

0.140.14 1.001.00 7.407.40 55.0055.00Favors SteroidsFavors SteroidsFavors ControlFavors Control

BennettBennettKlasterskyKlasterskySchumerSchumerThompsonThompsonLucasLucasSprungSprungBoneBoneVAVALuceLuce

196319631971197119761976197619761984198419841984198719871987198719881988


95% CI95% CI

II22 = 70%= 70%


0.140.14 1.001.00 7.407.40 55.0055.00Favors SteroidsFavors SteroidsFavors ControlFavors ControlAnn Intern Med 2004; 141:47Ann Intern Med 2004; 141:47

Clin Microbiol Infect 2009; 15: 308Clin Microbiol Infect 2009; 15: 308

1212


95% CI95% CI

II22 = 0%= 0%

Short course (24 h) high dose Short course (24 h) high dose


196319631971197119761976197619761984198419841984198719871987198719881988




corticosteroids (median corticosteroids (median --24,000 mg of hydrocortisone 24,000 mg of hydrocortisone equivalents) worsened survivalequivalents) worsened survival

SummarySummaryPre 1989 studies (N = 8)Pre 1989 studies (N = 8) P = 0.008P = 0.008


95% CI95% CI

II22 = 0%= 0%


196319631971197119761976197619761984198419841984198719871987198719881988

BollaertBollaertBriegelBriegelChawlaChawlaYildizYildiz

19981998199919991999199920022002

Stress dose corticosteroids Stress dose corticosteroids (median (median -- 1200 mg of 1200 mg of hydrocortisone equivalents)hydrocortisone equivalents)




YildizYildizAnnaneAnnaneConfalonieriConfalonieriMussackMussackOppertOppertTandanTandanRinaldiRinaldiCicarelliCicarelliSprungSprungSummarySummary

200220022002200220052005200520052005200520052005200620062007200720082008

hydrocortisone equivalents) hydrocortisone equivalents) tapered over 6 days were tapered over 6 days were associated with improved associated with improved survivalsurvival

Pre 1989 studies (N = 8)Pre 1989 studies (N = 8) P = 0.008P = 0.008Post 1997 studies (N = 12)Post 1997 studies (N = 12) P < 0.001P < 0.001


95% CI95% CI

II22 = 25%= 25%

II22 = 0%= 0%


196319631971197119761976197619761984198419841984198719871987198719881988


19981998199919991999199920022002




II 25% 25%YildizYildizAnnaneAnnaneConfalonieriConfalonieriMussackMussackOppertOppertTandanTandanRinaldiRinaldiCicarelliCicarelliSprungSprungSummarySummary

200220022002200220052005200520052005200520052005200620062007200720082008

Post 1997 studies (N = 12)Post 1997 studies (N = 12) P < 0.001P < 0.001Pre 1989 studies (N = 8)Pre 1989 studies (N = 8) P = 0.008P = 0.008

Corticus TrialCorticus Trial

1313


95% CI95% CI

II22 = 0%= 0%


196319631971197119761976197619761984198419841984198719871987198719881988


19981998199919991999199920022002 II22 = 0%= 0%




YildizYildizAnnaneAnnaneConfalonieriConfalonieriMussackMussackOppertOppertTandanTandanRinaldiRinaldiCicarelliCicarelliSprungSprungSummarySummary

200220022002200220052005200520052005200520052005200620062007200720082008

Post 1997 studies (N = 12)Post 1997 studies (N = 12) P < 0.001P < 0.001

II 0% 0%

Post 1997 studies (N = 11)Post 1997 studies (N = 11) P < 0.001P < 0.001

Pre 1989 studies (N = 8)Pre 1989 studies (N = 8) P = 0.008P = 0.008

How is Corticus Different How is Corticus Different from the 11 Other Trials offrom the 11 Other Trials offrom the 11 Other Trials of from the 11 Other Trials of

LowLow--Dose Steroids?Dose Steroids?

P = 0.03P = 0.03

of S

urvi

val

of S

urvi

val

2 702 70

7.407.40

20.0020.00

vors

Ste

roid

svo

rs S

tero

ids

Effect of Corticosteroids During Sepsis Effect of Corticosteroids During Sepsis Dependent on the Severity of IllnessDependent on the Severity of Illness

Control Odds of DeathControl Odds of Death0.10.1 1.01.0 10.010.0

Odd

s R

atio

oO

dds

Rat

io o

0.140.14

0.370.37

1.001.00

2.702.70

Favo

rs C

ontr

olFa

vors

Con

trol

Fav

Fav

Corticus TrialCorticus Trial

1414

Funnel Plot of Sepsis Trials of Funnel Plot of Sepsis Trials of Low Dose SteroidsLow Dose Steroids

1/St

d Er

r)1/

Std

Err)

66

44

55

Prec

isio

n (1

Prec

isio

n (1

Log OR of deathLog OR of death00--11--22--33

11

22

33

00332211

Favors SteroidsFavors Steroids Favors ControlFavors Control

Corticosteroid Effect on ShockCorticosteroid Effect on Shock

Low Dose Corticosteroids Trials Low Dose Corticosteroids Trials Reporting Shock ReversalReporting Shock Reversal

BriegelBriegelBollaertBollaertChawlaChawlaAnnaneAnnane

AuthorsAuthors II22 = 0%= 0%

Odds Ratio of Shock ReversalOdds Ratio of Shock Reversal

0.020.02 0.140.14 1.001.00 7.397.39 54.6054.60

AnnaneAnnaneMussackMussackTandanTandanCORTICUSCORTICUS

Overall (N = 7)Overall (N = 7)

Favors CorticosteroidsFavors CorticosteroidsFavors ControlFavors Control

P < 0.001P < 0.001

1515

SummarySummary

•• Corticosteroid effects during Corticosteroid effects during sepsis depend on sepsis depend on dosedose andandseverityseverity of illnessof illness

•• Low dose steroids decrease Low dose steroids decrease vasopressor requirements vasopressor requirements and enhance shock reversaland enhance shock reversal

High dose corticosteroids High dose corticosteroids increase mortalityincrease mortality

Low dose corticosteroids Low dose corticosteroids improve survival in improve survival in severely ill patientsseverely ill patients

•• At present, the beneficial effects of low At present, the beneficial effects of low doses of corticosteroids are based on doses of corticosteroids are based on small trials (median 40 patients, IQR 41small trials (median 40 patients, IQR 41--44) 44) confounded by publication biasconfounded by publication bias

LimitationsLimitations

•• The largest trial of low dose corticoThe largest trial of low dose cortico--steroids (CORTICUS, n = 499) studied a steroids (CORTICUS, n = 499) studied a relatively low risk populationrelatively low risk population

•• Benefit from low dose corticosteroids has Benefit from low dose corticosteroids has not been confirmed in a large multicenter not been confirmed in a large multicenter trial of high risk patientstrial of high risk patients

•• Until new data are available, the Until new data are available, the decision to administer low dose decision to administer low dose steroids for septic shock shouldsteroids for septic shock should

ConclusionsConclusions

steroids for septic shock should steroids for septic shock should be individualized:be individualized:

–– Severity of illnessSeverity of illness–– Assessment of riskAssessment of risk

1616

Intensive Insulin Therapy Intensive Insulin Therapy in Patients with Sepsisin Patients with Sepsis

How much risk and how much benefit?How much risk and how much benefit?

Endorsement of Glycemic Control as Endorsement of Glycemic Control as Standard of Care for the Critically IllStandard of Care for the Critically Ill

•• JCAHOJCAHO–– Core quality of care Core quality of care -- all Medicare hospitalsall Medicare hospitals

•• American College of EndocrinologyAmerican College of Endocrinology•• Volunteer Hospital AssociationVolunteer Hospital Association•• Volunteer Hospital AssociationVolunteer Hospital Association

–– Care bundleCare bundle

•• Institute for Healthcare ImprovementInstitute for Healthcare Improvement–– Sepsis bundleSepsis bundle–– Post cardiac surgeryPost cardiac surgery

•• Surviving Sepsis CampaignSurviving Sepsis Campaign–– Sepsis bundleSepsis bundle

Am J Resp Crit Care Med 2005; 172:1358Am J Resp Crit Care Med 2005; 172:1358

Selected Baseline CharacteristicsSelected Baseline CharacteristicsConventionalConventional

Insulin: glucoseInsulin: glucose180 180 -- 200 mg/dl200 mg/dl

N = 783N = 783

Intensive Intensive Insulin: glucose Insulin: glucose 80 80 -- 110 mg/dl110 mg/dl

N = 765N = 765

MenMen 557 (71%)557 (71%) 544 (71%)544 (71%)

Age (yr)Age (yr) 62.2 62.2 ±± 13.913.9 63.4 63.4 ±± 4.44.4

Reason for ICU care:Reason for ICU care:Cardiac SurgeryCardiac SurgeryNonNon--cardiac indicationscardiac indications

493 (63%)493 (63%)290 (37%)290 (37%)

477 (62%)477 (62%)288 (38%)288 (38%)

Apache II (median, IQR)Apache II (median, IQR) 9 (7 9 (7 -- 13)13) 9 (7 9 (7 -- 13)13)N Engl J Med 2001;345:1359 N Engl J Med 2001;345:1359

1717

Mortality Associated with Mortality Associated with Conventional Conventional versusversus Intensive InsulinIntensive Insulin

Death in ICUDeath in ICU # of # of patientspatients

ConventionalConventionalInsulin Insulin

Intensive Intensive Insulin Insulin

∆ ∆ deathsdeaths

Cardiac SurgeryCardiac Surgery 970970 25 (5%)25 (5%) 10 (2%)10 (2%) 1515Th iTh i 122122 10 (18%)10 (18%) 5 (7 6%)5 (7 6%) 55ThoracicThoracic 122122 10 (18%)10 (18%) 5 (7.6%)5 (7.6%) 55OtherOther 7070 6 (17%)6 (17%) 0 (0%)0 (0%) 66Neuro, Vascular, Neuro, Vascular, Trauma,TransplantTrauma,Transplant 386386 22 (11%)22 (11%) 20 (11%)20 (11%) 2 2

All patientsAll patients 15481548 63 (8%)63 (8%) 35 (5%)35 (5%) 28*28*

** P < 0.04P < 0.04N Engl J Med 2001;345:1359 N Engl J Med 2001;345:1359

LimitationsLimitations

•• Single center, unblinded study Single center, unblinded study •• Relatively high mortality among cardiac Relatively high mortality among cardiac

surgery patients in control group (5.1%)surgery patients in control group (5.1%)•• Immediate postImmediate post--operative operative i.v.i.v. glucose glucose

(200(200--300 g per day: ~ 2 300 g per day: ~ 2 -- 3 L D10 or D20) 3 L D10 or D20) and early feeding (enteral or parenteral)and early feeding (enteral or parenteral)

–– Not routine care for cardiothoracic Not routine care for cardiothoracic surgery patientssurgery patients

N Engl J Med 2001;345:1359 N Engl J Med 2001;345:1359

MetaMeta--analysis of Tight Glucose Control in Critically Illanalysis of Tight Glucose Control in Critically IllVan den BergheVan den BergheStecherStecherKiaKiaGreyGreyBilottaBilottaBilottaBilottaChanChanVan den BergheVan den BergheFernandezFernandezBlandBlandOksanenOksanenDaviesDaviesWaltersWaltersGray/GISTGray/GIST--UKUKBruno/THISBruno/THISBrunkhorst/VISEPBrunkhorst/VISEPDevos/GLUCONTROLDevos/GLUCONTROL

20012001200620062005200520042004200720072008200820082008

2005200520062006

2005200520072007199119912006200620072007200820082008200820072007

Favors Tight GlucoseFavors Tight Glucose Favors Usual CareFavors Usual Care

RR and 95% CIRR and 95% CI

27 trials of tight 27 trials of tight

Modified from Wiener RS, Modified from Wiener RS, et al. JAMA 2008; 300:933et al. JAMA 2008; 300:933 Relative Risk of Hospital MortalityRelative Risk of Hospital Mortality

Mackenzie/GLYCOGENICMackenzie/GLYCOGENICArabiArabiWangWangYuYuMitchellMitchellDeLaRosaDeLaRosaFarahFarahMcMullin/LOGICMcMullin/LOGICHenderson/SUGARHenderson/SUGARAzevedoAzevedo

20052005200620062006200620052005200620062006200620072007200720072005200520082008

1001000.010.01 0.10.1 11 1010

glucose controlglucose control(N = 8315; I(N = 8315; I22 = 17%)= 17%)

OverallOverall

Sensitivity Sensitivity AnalysisAnalysis

1818


20012001200620062005200520042004200720072008200820082008

2005200520062006

2005200520072007199119912006200620072007200820082008200820072007




Devos/GLUCONTROLDevos/GLUCONTROLMackenzie/GLYCOGENICMackenzie/GLYCOGENICArabiArabiWangWangYuYuMitchellMitchellDeLaRosaDeLaRosaFarahFarahMcMullin/LOGICMcMullin/LOGICHenderson/SUGARHenderson/SUGARAzevedoAzevedo

2007200720052005200620062006200620052005200620062006200620072007200720072005200520082008

1001000.010.01 0.10.1 11 1010

Glucose < 110 mg/dlGlucose < 110 mg/dl 14 trials very tight control14 trials very tight control

OverallOverall



20012001200620062005200520042004200720072008200820082008

2005200520062006

2005200520072007199119912006200620072007200820082008200820072007





20052005200620062006200620052005200620062006200620072007200720072005200520082008

1001000.010.01 0.10.1 11 1010

Glucose < 110 mg/dlGlucose < 110 mg/dlGlucose < 150 mg/dlGlucose < 150 mg/dl 13 trials moderate control13 trials moderate control

OverallOverall



20012001200620062005200520042004200720072008200820082008

2005200520062006

2005200520072007199119912006200620072007200820082008200820072007





20052005200620062006200620052005200620062006200620072007200720072005200520082008

1001000.010.01 0.10.1 11 1010

SurgicalSurgical

Glucose < 110 mg/dlGlucose < 110 mg/dlGlucose < 150 mg/dlGlucose < 150 mg/dl

7 trials in SICUs7 trials in SICUs

OverallOverall


1919


20012001200620062005200520042004200720072008200820082008

2005200520062006

2005200520072007199119912006200620072007200820082008200820072007




Devos/GLUCONTROLDevos/GLUCONTROLMackenzie/GLYCOGENICMackenzie/GLYCOGENICArabiArabiWangWangYuYuMitchellMitchellDeLaRosaDeLaRosaFarahFarahMcMullin/LOGICMcMullin/LOGICHenderson/SUGARHenderson/SUGARAzevedoAzevedo

2007200720052005200620062006200620052005200620062006200620072007200720072005200520082008

1001000.010.01 0.10.1 11 1010

MedicalMedicalSurgicalSurgical


8 trials in MICUs8 trials in MICUs

OverallOverall



20012001200620062005200520042004200720072008200820082008

2005200520062006

2005200520072007199119912006200620072007200820082008200820072007





20052005200620062006200620052005200620062006200620072007200720072005200520082008

1001000.010.01 0.10.1 11 1010

MedicalMedicalMed / SurgMed / Surg

SurgicalSurgical


12 trials in mixed ICUs12 trials in mixed ICUsOverallOverall


MetaMeta--analysis of Tight Glucose Control in Critically Illanalysis of Tight Glucose Control in Critically IllVan den BergheVan den BergheStecherStecherKiaKiaGreyGreyBilottaBilottaBilottaBilottaChanChanVan den BergheVan den BergheFernandezFernandezBlandBlandOksanenOksanenDaviesDaviesWaltersWaltersGray/GISTGray/GIST--UKUKBruno/THISBruno/THISBrunkhorst/VISEPBrunkhorst/VISEPDevosDevos/GLUCONTROL/GLUCONTROL

20012001200620062005200520042004200720072008200820082008

2005200520062006

2005200520072007199119912006200620072007200820082008200820072007





20052005200620062006200620052005200620062006200620072007200720072005200520082008

1001000.010.01 0.10.1 11 1010

MedicalMedicalMed / SurgMed / Surg

SurgicalSurgical


OverallOverall

Sensitivity Sensitivity AnalysisAnalysis P = NS for allP = NS for all

2020

Tight Glucose Control and Tight Glucose Control and the Risk of Hypoglycemiathe Risk of Hypoglycemia

RR and 95% CIRR and 95% CIHypoglycemiaHypoglycemia

Favors Tight ControlFavors Tight Control Favors Usual CareFavors Usual Care

Glucose goal Glucose goal << 110 mg/dL110 mg/dL

Tight glucose controlTight glucose controlincreasedincreased 7 7 -- 8 fold the risk of 8 fold the risk of

Modified from Wiener RS, Modified from Wiener RS, et al. JAMA 2008; 300:933et al. JAMA 2008; 300:933

OverallOverall

0.010.01 0.10.1 11 1010 100100Relative RiskRelative Risk

Glucose goal Glucose goal << 150 mg/dL150 mg/dLSurgical ICUSurgical ICUMedical ICUMedical ICUMedicalMedical--Surgical ICUSurgical ICU

II22 = 0%= 0%N = 15N = 15

hypoglycemia (< 40 mg/dl) hypoglycemia (< 40 mg/dl) independent of target glucose independent of target glucose ((<< 150 or 150 or << 110 mg/dL) or type 110 mg/dL) or type of ICU (medical, surgical or of ICU (medical, surgical or combined)combined)

NICENICE--Sugar TrialSugar TrialBaseline CharacteristicsBaseline Characteristics

Intensive Intensive InsulinInsulin

ConventionaConventional Insulinl Insulin

Enrolled (N)Enrolled (N) 30543054 30503050Enrolled (N)Enrolled (N) 30543054 30503050SurgicalSurgical 37%37% 37%37%Apache II > 25Apache II > 25 31%31% 31%31%Severe SepsisSevere Sepsis 22%22% 21%21%MechMech VentilatorVentilator 94%94% 94%94%

N Engl J Med 2009; 360:1283N Engl J Med 2009; 360:1283

2121

Intensive Intensive InsulinInsulin

ConvenConven--tional tional

OROR(95% CI)(95% CI)

PP--ValueValue

NICENICE--Sugar TrialSugar TrialOutcomesOutcomes

In 6014 critically ill patients, tight glucose In 6014 critically ill patients, tight glucose control was associated with hypoglycemia control was associated with hypoglycemia

Mortality (90d)Mortality (90d) 27.5%27.5% 24.9%24.9% 1.141.14(1.02 (1.02 -- 1.28)1.28)

0.020.02

HypoglycemiaHypoglycemia 6.8%6.8% 0.5%0.5% 14.714.7(9 (9 -- 25.9)25.9)

<0.001<0.001

N Engl J Med 2009; 360:1283N Engl J Med 2009; 360:1283

and increased mortality at 90 days. and increased mortality at 90 days.

“On the basis of [these] results we do not “On the basis of [these] results we do not recommend use of the lower target (81 recommend use of the lower target (81

-- 110 mg/dL) in critically ill patients.”110 mg/dL) in critically ill patients.”

MetaMeta--analyses of Sepsis Trials analyses of Sepsis Trials with at Leastwith at Least

One Significant Benefical TrialOne Significant Benefical Trial

SummarySummaryAntiAnti--EndotoxinEndotoxin

J5 AntiserumJ5 Antiserum MonoclonalsMonoclonals

** **

p = 0.59p = 0.59 p = 0.60p = 0.60

High DoseHigh DoseCorticosteroidsCorticosteroids

**p = 0.009p = 0.009

Shift from beneficial to last trial, p = 0.003Shift from beneficial to last trial, p = 0.003**

ILIL--1RA1RA

**

p = 0.14p = 0.14

Activated Protein CActivated Protein C

**

p = 0.55p = 0.55

Intensive InsulinIntensive Insulin

**p = 0.32p = 0.32

2222

The randomized control trial minimizes The randomized control trial minimizes bias but does not eliminate the need for bias but does not eliminate the need for reproducibility which is thereproducibility which is the sine qua nonsine qua nonreproducibility which is the reproducibility which is the sine qua nonsine qua non(i.e. the indispensable and essential (i.e. the indispensable and essential condition) of scientific evidencecondition) of scientific evidence

Documents

Secondary Data/Meta Analysis