Grading evidence and recommendations Holger Schünemann Gunn Vist Gordon Guyatt for the GRADE Working Group.

Grading evidence and Grading evidence and recommendationsrecommendations

Holger Schünemann

Gunn Vist

Gordon Guyatt

for the GRADE Working Group

Introduction to GRADE Example of applying GRADE Demonstration of the GRADE profiler

sofware

Today’s talk

Intro ExerciseIntro Exercise0

Where would you prefer to Where would you prefer to live?live?

← Option 1

Option 2 →

← Option 1 (pink card)

Option 2 → (green card)

Introduction to GRADEIntroduction to GRADE1

When to make a recommendation?When to make a recommendation? never

patient values differ guidelines should just lay out

benefits and risks

when evidence strong enough if very weak, too uncertain

clinicians need guidance• intense study demands decision

What type of recommendations?What type of recommendations?

strong recommendations– high quality methods – large precise effect – few down sides of therapy

weak recommendations– low quality methods– imprecise estimate– small overall effect– substantial down sides– people react differently to same

outcomes/circumstances

Should we grade Should we grade recommendations?recommendations?

People draw conclusions about the– quality of evidence– strength of recommendations

Systematic and explicit approaches can help– protect against errors– resolve disagreements– facilitate critical appraisal– communicate information

Why grade Why grade recommendations?recommendations?

Change practitioner behavior

Strong: apply uniformly – just do it

Weak: think about it– examine evidence yourself– consider patient circumstances very carefully– explore with the patient

However, there is wide variation in currently used approaches

Which grading system? Which grading system?

Evidence Recommendation

II-2 B C+ 1 Strong Strongly

recommended

Organization

USPSTF ACCP GCPS

Still not confused?Still not confused?

EvidenceRecommendation

B Class I C+ 1 IV C

Organization

AHA ACCP SIGN

Recommendation for use of oral anticoagulation in patients with atrial fibrillation and rheumatic mitral valve disease

Grading SystemGrading System

Current profusion: can there be consensus?

GRADEGRADE

Grades of Recommendation Assessment, Development

and Evaluation

What do you know about What do you know about GRADE?GRADE?

Have prepared a guidelineHave prepared a guideline Read the BMJ paper Read the BMJ paper

Have prepared a systematic review and a Have prepared a systematic review and a summary of findings tablesummary of findings table

Have attended a GRADE meeting, workshop or talkHave attended a GRADE meeting, workshop or talk

About GRADEAbout GRADE

o Began as informal working group in 2000Began as informal working group in 2000o Researchers/guideline developers with Researchers/guideline developers with

interest in methodologyinterest in methodologyo Aim: to develop a Aim: to develop a commoncommon system for system for

grading the quality of evidence and the grading the quality of evidence and the strength of recommendations that is sensible strength of recommendations that is sensible and to explore the range of interventions and and to explore the range of interventions and contexts for which it might be useful*contexts for which it might be useful*

o 13 meetings (~10 – 35 attendants)13 meetings (~10 – 35 attendants)o Evaluation of existing systems and reliability*Evaluation of existing systems and reliability*o Workshops at various places including Workshops at various places including

Cochrane Colloquia, WHO and GIN since 2000Cochrane Colloquia, WHO and GIN since 2000

*Grade Working Group. CMAJ 2003, BMJ 2004, BMC 2004, BMC 2005

GRADE Working GroupGRADE Working GroupDavid Atkins, chief medical officerDavid Atkins, chief medical officeraa Dana Best, assistant professorDana Best, assistant professorbb Peter A Briss, chiefPeter A Briss, chiefcc Martin Eccles, professorMartin Eccles, professordd Yngve Falck-Ytter, associate directorYngve Falck-Ytter, associate directoree Signe Flottorp, researcherSigne Flottorp, researcherff Gordon H Guyatt, professorGordon H Guyatt, professorgg Robin T Harbour, Robin T Harbour, quality and information quality and information

directordirector h h Margaret C Haugh, methodologistMargaret C Haugh, methodologistii David Henry, professorDavid Henry, professorjj Suzanne Hill, senior lecturerSuzanne Hill, senior lecturer jj Roman Jaeschke, clinical professorRoman Jaeschke, clinical professorkk Gillian Leng, guidelines programme directorGillian Leng, guidelines programme director ll Alessandro Liberati, professorAlessandro Liberati, professormm Nicola Magrini, directorNicola Magrini, directornn

James Mason, professorJames Mason, professordd Philippa Middleton, honorary research fellowPhilippa Middleton, honorary research fellowoo Jacek Mrukowicz, executive directorJacek Mrukowicz, executive directorpp Dianne O’Connell, senior epidemiologistDianne O’Connell, senior epidemiologistqq Andrew D Oxman, directorAndrew D Oxman, directorff Bob Phillips, associate fellowBob Phillips, associate fellowrr Holger J Schünemann, associate professorHolger J Schünemann, associate professorg,sg,s Tessa Tan-Torres Edejer, medical Tessa Tan-Torres Edejer, medical

officer/scientistofficer/scientisttt Helena Varonen, associate editorHelena Varonen, associate editoruu Gunn E Vist, researcherGunn E Vist, researcherff John W Williams Jr, associate professorJohn W Williams Jr, associate professorvv Stephanie Zaza, Stephanie Zaza, project directorproject directorww

a)a) Agency for Healthcare Research and Quality, Agency for Healthcare Research and Quality, USAUSA b)b) Children's National Medical Center, Children's National Medical Center, USA USAc) Centers for Disease Control and Prevention, c) Centers for Disease Control and Prevention, USAUSAd) University of Newcastle upon Tyne, d) University of Newcastle upon Tyne, UKUKe) German Cochrane Centre, e) German Cochrane Centre, GermanyGermanyf) Norwegian Centre for Health Services, f) Norwegian Centre for Health Services, NorwayNorwayg) McMaster University, g) McMaster University, CanadaCanadah) Scottish Intercollegiate Guidelines Network, h) Scottish Intercollegiate Guidelines Network, UKUKi) Fédération Nationale des Centres de Lutte i) Fédération Nationale des Centres de Lutte Contre le Cancer, Contre le Cancer, FranceFrancej) University of Newcastle, j) University of Newcastle, AustraliaAustraliak) McMaster University, k) McMaster University, CanadaCanadal) National Institute for Clinical Excellence, l) National Institute for Clinical Excellence, UKUKm) m) Università di Modena e Reggio Emilia, Università di Modena e Reggio Emilia, ItalyItalyn)n) Centro per la Valutazione della Efficacia della Centro per la Valutazione della Efficacia della Assistenza Sanitaria, Assistenza Sanitaria, ItalyItalyo) Australasian Cochrane Centre, o) Australasian Cochrane Centre, AustraliaAustralia p) Polish Institute for Evidence Based Medicine, p) Polish Institute for Evidence Based Medicine, PolandPolandq) The Cancer Council, q) The Cancer Council, AustraliaAustraliar) r) Centre for Evidence-based Medicine, Centre for Evidence-based Medicine, UKUKs)s) National Cancer Institute, National Cancer Institute, ItalyItalyt) World Health Organisation, t) World Health Organisation, SwitzerlandSwitzerland u) Finnish Medical Society Duodecim, u) Finnish Medical Society Duodecim, FinlandFinland v) Duke University Medical Center, v) Duke University Medical Center, USA USA w) w) Centers for Disease Control and Prevention, Centers for Disease Control and Prevention, USAUSA

How can we judge the How can we judge the extent of our confidence extent of our confidence

that adherence to that adherence to aa recommendation will do recommendation will do more good than harm?more good than harm?

Grading SystemGrading System

Strength of the recommendation• do it (or don’t do it)/recommend• probably do it (or probably don’t do it)/suggest

Quality of underlying evidence• high quality (well done RCT)• moderate (quasi-RCT)• low (well done observational)• very low (anything else)

Moving quality downMoving quality down• poor (RCT) design, implementation

→randomization, blinding, concealment, follow-up, intention to treat principle, early stopping for benefit

• inconsistency• Indirect evidence

→patients, interventions, outcomes→A vs B, but have A to C, B to C

• sparse or imprecise data• reporting bias

Reporting biasReporting bias

• high likelihood of reporting bias can lower quality

• reporting of outcomes

• reporting of studies• publication bias

Moving quality upMoving quality upObservational studies – high or moderate Observational studies – high or moderate

quality?quality?

Strong association→strong association: RR > 2 or RR < 0.5→very strong association: RR > 5 or RR < 0.2

Dose response relationship– bleeding risk associated with increasing INR (blood

thinning with warfarin)

Plausible confounders would have reduced the effect

For example, plausible explanatoryfactors that were not adjusted for in

studies comparing mortalityrates of for-profit and not-for-profit

hospitals would havereduced the observed effect. Thus, the evidence showing that for-profit

hospitals have a higher risk of mortality is more convincing

Quality assessment criteriaQuality assessment criteria

Quality of evidence Study design Lower if Higher if

High Randomised trial

Moderate

Low Observational

study

Very low Any other evidence

Study quality:

-1 Serious

limitations

-2 Very serious

limitations

-1 Important

inconsistency

Directness:

-1 Some

uncertainty

-2 Major

uncertainty

-1 Sparse or

imprecise data

-1 High probability

of reporting bias

Strong association:

+1 Strong, no

plausible

confounders

+2 Very strong,

no major

threats to

validity

+1 Evidence of a

Dose response

gradient

+1 All plausible

confounders

would have

reduced the

effect

Categories of qualityCategories of quality

High: Further research is very unlikely to change our confidence in the estimate of effect.

Moderate: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.

Low: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.

Very low: Any estimate of effect is very uncertain.

“The extent to which one can be confident that an estimate of effect or association is correct.”

Strength of recommendationStrength of recommendation

“The extent to which one can be confident that adherence to a recommendation will do more good than harm.”

quality of the evidence translation of the evidence into practice

in a specific setting uncertainty about baseline risk trade-offs (the relative value attached to

the expected benefits, harms and costs)

More practice using the More practice using the voting instrument….voting instrument….



Remember

You are hiking.You are hiking.

Which of the following animals Which of the following animals would you prefer to would you prefer to

encounter?encounter?



You are buying an ice cream.You are buying an ice cream.

Which flavor do you prefer?Which flavor do you prefer?



Chocolate

Strawberry

You are buying a new car.You are buying a new car.

Which one would you buy?Which one would you buy?



Yellow fox

Red Ferrari

What determines your What determines your choices?choices?

•

Values and preferencesValues and preferences

underlying values and preferences always present

sometimes crucial

important to make explicit

Judgements about the balance Judgements about the balance between benefits and harmsbetween benefits and harms

Before considering cost and making a recommendation

Judgment: Benefits vs Risks/CostsJudgment: Benefits vs Risks/Costs

quality of evidence seriousness of outcome magnitude of effect precision of treatment effect risk of target event risk of adverse events cost of therapy values

Judgements about Judgements about recommendationsrecommendations

1. Benefit and downside evaluation

? ?

2. Recommendation (wording) STRONG WEAK WEAK STRONG Recommend suggest suggest recommend don’t do it probably don’t do it probably do it do it should not do it might not do it might do it should do it

Benefits<< Downsides

Benefits ?≤ Downsides

Benefits ?≥ Downsides

Benefits >> Downsides

Strong recommendation when Strong recommendation when evidence is weak?evidence is weak?

Balance of benefits and downsides clearly on one side

Not frequent if quality is low or very low

Comparison of GRADE and other Comparison of GRADE and other systemssystems

• Explicit definitions• Explicit, sequential judgements• Components of quality• Overall quality• Relative importance of outcomes• Balance between health benefits and harms• Balance between incremental health benefits

and costs• Consideration of equity• Evidence profiles• International collaboration• Software• Consistent judgements?• Communication?

Who is interested in GRADEWho is interested in GRADE

• American College of Chest Physicians (ACCP)

• WHO• American Endocrine Society• UpToDate• Clinical Evidence• American Society of Clinical Oncology

(ASCO)• American Thoracic Society (ATS)• Urologists worldwide (EBUro)• NICE

ConclusionConclusion Challenges in grading

– judgment always required

Must consider study design, execution, consistency, directness, reporting bias– magnitude, precision

Balance of benefits and risks/cost– magnitude of effects; precision of effects; values and preferences

Separation of recommendation from quality of the evidence

GRADE working group active in obtaining feedback and dissemination

Making judgments using GRADEMaking judgments using GRADE2

The clinical questionThe clinical question

Population: In patients with chronic atrial fibrillation and no prior history

of stroke

Intervention: does oral anticoagulation (comparison) compared with no therapy

Outcome: reduce the risk for embolic stroke, hemorrhage

and death?

The evidenceThe evidence Systematic Review* 5 RCTs 2,313 Patients randomised Warfarin in all studies Studien 1.5 years mean follow-up Outcomes: Ischemic Stroke, hemorrhage

(major, including intracranial), death (vascular and all cause) and dependency

*Systematic Review: Aguilar & Hart. Cochrane Database of Systematic Reviews 2005, Issue 3.

All disabling or fatal stroke (isch. and hemorrh.)

Major hemorrhage (non IC) All cause mortality Minor bleeding (hematoma, prolonged

bleeding of minor wounds)

*Systematic Review: Aguilar & Hart. Cochrane Database of Systematic Reviews 2005,

Issue 3.

Outcomes/endpointsOutcomes/endpoints

How important is the endpoint for decision making?

Judgment about the relative importance for each endpoint on a scale from 9 (most important) to 1 (least important): • 7 – 9: the endpoint is critical for decision making.• 4 – 6: the endpoint is important but not critical.• 1 – 3: the endpoint is not important.


All disabling or fatal stroke (isch. and hemorrh.)

Major hemorrhage (non IC) All cause mortality Minor bleeding (hematoma, prolonged

bleeding of minor wounds)

*Systematic Review: Aguilar & Hart. Cochrane Database of Systematic Reviews 2005,

Issue 3.


9

7

9

5

Quality assessment criteriaQuality assessment criteria

Quality of evidence Study design Lower if Higher if

High Randomised trial

Moderate

Low Observational

study

Very low Any other evidence

Study quality:

-1 Serious

limitations

-2 Very serious

limitations

-1 Important

inconsistency

Directness:

-1 Some

uncertainty

-2 Major

uncertainty

-1 Sparse or

imprecise data

-1 High probability

of reporting bias

Strong association:

+1 Strong, no

plausible

confounders

+2 Very strong,

no major

threats to

validity

+1 Evidence of a

Dose response

gradient

+1 All plausible

confounders

would have

reduced the

effect

Disabling or fatal Disabling or fatal strokestroke

Study design: 5 RCTs

Quality of evidence for this endpoint: High


Detailed design and execution Concealment Follow-up In two studies (CAFA; SPINAF) both

patients and outcome assessors were blinded; in the other studies only outcomes assessors.

Quality of evidence for this endpoint now:

High (or -1 Moderate)


Consistency:


Consistency: No inconsistency


High

Directness of evidenceDirectness of evidence

indirect treatment comparisons– interested in A versus B– have A versus C and B versus C

alendronate vs risedronate (biphosponates)– both versus placebo, no head-to-head

Directness - patientsDirectness - patients

patients meet trials’ eligibility criteria

not included, but no reason to question– slight age difference, comorbidity, race

some question, bottom line applicable– valvular atrial fibrillation

serious question about biology– heart failure trials applicability to aortic stenosis

Directness - interventionsDirectness - interventions

same drugs and doses– captopril 100 mg. tid in heart failure

similar drugs and doses– captopril in lower doses

same class and biology– other ACEI in heart failure

questionable class and biology– ARB in heart failure

Directness - outcomesDirectness - outcomes

same outcomes– alendronate over 3 years on fracture

similar but questionable– alendronate over long-term

serious question– surrogate outcomes

– bone density; arrhythmia suppression; cholesterol levels (clofibrate)


Directness of the evidence: Population, Intervention, Outcomes Direct


High


Imprecise or sparse data: Would few additional events or

larger studies likely alter the results?


Imprecise or sparse data: No imprecise or sparse data


High


Reporting bias: Not present


High


Strong association? present (RR = 0.46)


High [or +1 High (from moderate)]

strong, no plausible confounder, consistent

and direkt evidence

Endpoint: Major extracranial Endpoint: Major extracranial hemorhagehemorhage

• Study design: 4 RCTs → Quality: High• Study details and execution:

No serious limitations• No inconsistency and direct• Imprecise or sparse data?

Imprecise or sparse dataImprecise or sparse data

There is not an empirical basis for defining imprecise or sparse data. Two possible definitions are:

Data are sparse if the results include just a few events or observations and they are uninformative

Data are imprecise if the confidence intervals are sufficiently wide that an estimate is consistent with either important harms or important benefits.

These different definitions can result in different judgments.

Major extracranial hemorhageMajor extracranial hemorhage

• Study design: 4 RCTs → Quality: High• Study details and execution:

no serious limitations• No inconsistency and direct• Imprecise or sparse data?

• Imprecise data (wide confidence intervals benefit and harm uncertain)

GRADE Evidence Profile Author(s): Schunemann Date: 8/28/2005 Question: Should warfarin vs placebo or no treatment be used for patients with non-valvular atrial fibrillation? Patient or population: Patients with non-valvular atrial fibrillation Settings: Long term outpatient management Systematic review: Aguilar MI, Hart R. Oral anticoagulants for preventing stroke in patients with non-valvular atrial fibrillation and no previous history of stroke or transient ischemic attacks. The Cochrane Database of Systematic Reviews 2005, Issue 3. Art. No.: CD001927.pub2. DOI: 10.1002/14651858.CD001927.pub2.

Summary of findings Quality assessment

No of patients Effect

No of studies

Design Limitations Consistency Directness Other

considerations warfarin

placebo or no treatment

Relative (95% CI)

Absolute (95% CI)

Quality Importance

Disabling or Fatal Stroke (ischemic and hemorrhagic) (Neuroimaging or autopsy6 Follow up: Mean: 1.5 years)

5 Randomised trials

No limitation8,12

No important inconsistency

No uncertainty

Strong association (+1)13

18/1154 (1.6%)

39/1159 (3.4%)

RR 0.46 (0.27 to 0.81)

20/1 000 (30 fewer/1000 to 10 fewer/1000)

High

914

Extracranial hemorrhage (Transfusion or invasive procedure requirement1 Follow up: 1.5 years)

4 Randomised trials

No limitations 8


No uncertainty

Imprecise or sparse data (-1)11

17/1154 (1.5%)

16/1159 (1.4%)

RR 1.06 (0.54 to 2.09)

0/1 000 (10 fewer/1000 to 10 more/1000)

Moderate

7

All cause mortality3,5 (Direct patient follow-up Follow up: 1.5 years)

5 Randomised trials

No limitations15


No uncertainty

None 69/1225 (5.6%)

99/1236 (8%)

RR 0.70 (0.52 to 0.95)

20/1 000 (40 fewer/1000 to 1 fewer/1000)

High

9

Footnotes:

1. required transfusion of two or more units of red blood cells, hospitalization, or invasive procedures to control bleeding and those that resulted in death or permanent functional impairment (e.g. blindness) were included.

2. These consisted of death due to stroke, heart disease, hemorrhage and sudden deaths of unknown cause.

3. All cause mortality: death from any cause (vascular and non- vascular) within 30 days from onset of stroke symptoms. For this outcome, results of published data, which included about 6% of patients with prior stroke or TIA, were used.

4. The diagnosis of MI was usually based upon electrocardiographic changes, elevation of enzymes or post-mortem examination.

5. From Fig 10

6. Follow-up for this outcome was less than 100%.

7. Ischemic stroke was an identified outcome in all trials, with the ischemic nature conrmed by neuroimaging or autopsy in the majority of cases.

8. In two studies (CAFA; SPINAF) patients and outcome assessors were blind to OAC administration, while in the remaining trials treatment was given open label with outcomes verified by those unaware of treatment assignment.

9. Methodological quality was not downgraded because the lack of blinding in some studies did not have important impact on the results

10. Only 5 events in the OAC group and 2 events in the control group, confidence intervals wide

11. Only 17 events in the OAC and 16 events in the control group

12. Loss to follow-up not reported in AFASAK I and CAFA, ranged from 0 to 3% in the other studies

13. Strong association present: RR 0.46

14. Importance is rated on a scale from 1 to 9. 1 represents least important (not important for decision making) and 9 most important (for decision making).

15. Lack of blinding in two trials of lesser concern


Judgements about the overall Judgements about the overall quality of evidencequality of evidence

Most systems not explicit

Options:– strongest outcome– primary outcome– benefits– weighted– separate grades for benefits and harms– no overall grade– weakest outcome

Based on lowest of all the critical outcomes Beyond the scope of a systematic review




No of studies




Relative (95% CI)

Absolute (95% CI)

Quality Importance


5 Randomised trials

No limitation8,12


No uncertainty


18/1154 (1.6%)

39/1159 (3.4%)

RR 0.46 (0.27 to 0.81)

20/1 000 (30 fewer/1000 to 10 fewer/1000)

High

914


4 Randomised trials

No limitations 8


No uncertainty


17/1154 (1.5%)

16/1159 (1.4%)

RR 1.06 (0.54 to 2.09)

0/1 000 (10 fewer/1000 to 10 more/1000)

Moderate

7


5 Randomised trials

No limitations15


No uncertainty

None 69/1225 (5.6%)

99/1236 (8%)

RR 0.70 (0.52 to 0.95)

20/1 000 (40 fewer/1000 to 1 fewer/1000)

High

9

Quality across all endpoints

Risk groupsRisk groupsRisk for cardio-embolic stroke: High (prior TIA or stroke*, > 75 yrs,

LVEF/CHF, HTN or DM): 10%/year

Moderate risk (65 to 75 years) or one risk factor: 3 to 4%/year

Low risk (< 65 years): 0.5%/year




No of studies




Relative (95% CI)

Absolute (95% CI)

Low risk (0.5%/year) Mean/moderate risk (3.5%/year)

High risk (10%/year)

Quality Importance


5 Randomised trials

No limitation8,12


No uncertainty


18/1154 (1.6%)

39/1159 (3.4%)

RR 0.46 (0.27 to 0.81)

Low: 2 fewer/1000 (1 fewer/1000 to 3 fewer/1000)

Mean (moderate): 20 fewer/1 000 (30 fewer/1000 to 10 fewer/1000)

High: 50 fewer (20 fewer to 75 fewer)

High

914


4 Randomised trials

No limitations 8


No uncertainty


17/1154 (1.5%)

16/1159 (1.4%)

RR 1.06 (0.54 to 2.09)

0/1 000 (10 fewer/1000 to 10 more/1000)

Moderate

7


5 Randomised trials

No limitations15


No uncertainty

None 69/1225 (5.6%)

99/1236 (8%)

RR 0.70 (0.52 to 0.95)

20 fewer/1 000 (40 fewer/1000 to 1 fewer/1000)

High

9

Risk groupsRisk groupsRisk for cardio-embolic stroke: High (prior TIA or stroke*, > 75 yrs,

LVEF/CHF, HTN or DM): 10%/year– Benefits greater downsides: do it, high

Moderate risk (65 to 75 years) or one risk factor: 3 to 4%/year – Benefits greater downsides: do it, high

Low risk (< 65 years): 0.5%/year– Benefits smaller than downsides: values:

probably do not do it, high

Value and preference Value and preference statementsstatements

underlying values and preferences always present

sometimes crucial

important to make explicit

Observational studies – Observational studies – high or moderate quality?high or moderate quality?

Strong association

Dose response relationship– bleeding risk associated with increasing INR

(blood thinning with warfarin)

Plausible confounders would have reduced the effect

For example, plausible explanatoryfactors that were not adjusted for in

studies comparing mortalityrates of for-profit and not-for-profit

hospitals would havereduced the observed effect. Thus,

the evidence that for-profithospitals have a higher risk of mortality is more convincing

GRADEproGRADEpro3

Guideline development processGuideline development process

Prioritise Problems, establish panelPrioritise Problems, establish panel

Systematic ReviewSystematic Review

Evidence ProfileEvidence Profile

Relative importance of outcomesRelative importance of outcomes

Overall quality of evidenceOverall quality of evidence

Benefit – downside evaluationBenefit – downside evaluation


Implementation and evaluation of guidelinesImplementation and evaluation of guidelines

GRADE

Guideline development processGuideline development process

Prioritise Problems, establish panelPrioritise Problems, establish panel

Systematic ReviewSystematic Review

Evidence ProfileEvidence Profile

Relative importance of outcomesRelative importance of outcomes

Overall quality of evidenceOverall quality of evidence

Benefit – downside evaluationBenefit – downside evaluation


Implementation and evaluation of guidelinesImplementation and evaluation of guidelines

GRADE

Summary of Findings

GRADE ProfilerGRADE Profiler

GRADEpro©GRADEpro©

Visual studio.net Windows based (Mac version

coming) Simple installation Help file Will be integrated with Revman

(trial) Free availability Beta version

Development of GRADE Development of GRADE profilesprofiles




No of studies




Relative (95% CI)

Absolute (95% CI)

Quality Importance


5 Randomised trials

Serious limitations (-1)8,12


No uncertainty


18/1154 (1.6%)

39/1159 (3.4%)

RR 0.46 (0.27 to 0.81)

20/1 000 (30 fewer/1000 to 10 fewer/1000)

High

914

Intracranial Hemorrhage (Clinical diagnosis confirmed by CT or post-mortem Follow up: Mean follow-up: 1.5 years)

5 Randomised trials

No limitations


No uncertainty


5/1244 (0.4%)

2/1159 (0.2%)

RR 1.75 (0.47 to 6.41)

3/1 000 (0 more/1000 to 10 more/1000)

Moderate

8

Extracranial hemorrhage (Transfusion or invasive procedure requirement1 Follow up: )

4 Randomised trials

Serious limitations (-1)8


No uncertainty


17/1244 (1.4%)

16/1159 (1.4%)

RR 0.93 (0.48 to 1.79)

0/1 000 (10 fewer/1000 to 10 more/1000)

Low

7

All cause mortality3,5 (Direct patient follow-up Follow up: )

5 Randomised trials

No limitations15


No uncertainty

None 69/1225 (5.6%)

99/1236 (8%)

RR 0.70 (0.52 to 0.95)

20/1 000 (40 fewer/1000 to 1 fewer/1000)

High

9

Vascular death2 (Death due to stroke, heart disease, hemorrhage and sudden death Follow up: )

5 Randomised trials

No limitations15


No uncertainty

None 43/1244 (3.5%)

51/1159 (4.4%)

RR 0.79 (0.53 to 1.17)

1/1 000 (3 fewer/1000 to 1 more/1000)

High

9

All ischemic stroke (Neuroimaging or autopsy6,7 Follow up: 1.5 years)

5 Randomised trials

No limitations8,9


No uncertainty

None 22/1154 (1.9%)

69/1159 (6%)

RR 0.32 (0.20 to 0.51)

40/1 000 (60 fewer/1000 to 20 fewer/1000)

High

7

8. In two studies (CAFA; SPINAF) patients and outcome assessors were blind to OAC administration, while in the

remaining trials treatment was given open label with

outcomes verified by those unaware of treatment

assignment.




No of studies




Relative (95% CI)

Absolute (95% CI)

Quality Importance


5 Randomised trials

Serious limitations (-1)8,12


No uncertainty


18/1154 (1.6%)

39/1159 (3.4%)

RR 0.46 (0.27 to 0.81)

20/1 000 (30 fewer/1000 to 10 fewer/1000)

High

914

Intracranial Hemorrhage (Clinical diagnosis confirmed by CT or post-mortem Follow up: Mean follow-up: 1.5 years)

5 Randomised trials

No limitations


No uncertainty


5/1154 (0.4%)

2/1159 (0.2%)

RR 1.87 (0.51 to 6.82)

3/1 000 (0 more/1000 to 10 more/1000)

Moderate

8


4 Randomised trials

Serious limitations (-1)8


No uncertainty


17/1154 (1.5%)

16/1159 (1.4%)

RR 1.06 (0.54 to 2.09)

0/1 000 (10 fewer/1000 to 10 more/1000)

Low

7


5 Randomised trials

No limitations15


No uncertainty

None 69/1225 (5.6%)

99/1236 (8%)

RR 0.70 (0.52 to 0.95)

20/1 000 (40 fewer/1000 to 1 fewer/1000)

High

9

Vascular death2 (Death due to stroke, heart disease, hemorrhage and sudden death Follow up: 1.5 years)

5 Randomised trials

No limitations15


No uncertainty

None 43/1154 (3.7%)

51/1159 (4.4%)

RR 0.85 (0.57 to 1.26)

1/1 000 (3 fewer/1000 to 1 more/1000)

High

9

All ischemic stroke (Neuroimaging or autopsy6,7 Follow up: 1.5 years)

5 Randomised trials

No limitations8,9


No uncertainty


22/1154 (1.9%)

69/1159 (6%)

RR 0.32 (0.20 to 0.51)

40/1 000 (60 fewer/1000 to 20 fewer/1000)

High

7

Footnotes:

1. required transfusion of two or more units of red blood cells, hospitalization, or invasive procedures to control bleeding and those that resulted in death or permanent functional impairment (e.g. blindness) were included.

2. These consisted of death due to stroke, heart disease, hemorrhage and sudden deaths of unknown cause.

3. All cause mortality: death from any cause (vascular and non- vascular) within 30 days from onset of stroke symptoms. For this outcome, results of published data, which included about 6% of patients with prior stroke or TIA, were used.

4. The diagnosis of MI was usually based upon electrocardiographic changes, elevation of enzymes or post-mortem examination.

5. From Fig 10

6. Follow-up for this outcome was less than 100%.

7. Ischemic stroke was an identified outcome in all trials, with the ischemic nature conrmed by neuroimaging or autopsy in the majority of cases.

8. In two studies (CAFA; SPINAF) patients and outcome assessors were blind to OAC administration, while in the remaining trials treatment was given open label with outcomes verified by those unaware of treatment assignment.

9. Methodological quality was not downgraded because the lack of blinding in some studies did not have important impact on the results

10. Only 5 events in the OAC group and 2 events in the control group, confidence intervals wide

11. Only 17 events in the OAC and 16 events in the control group

12. Loss to follow-up not reported in AFASAK I and CAFA, ranged from 0 to 3% in the other studies


14. Importance is rated on a scale from 1 to 9. 1 represents least important (not important for decision making) and 9 most important (for decision making).

15. Lack of blinding in two trials of lesser concern


GRADEproGRADEpro

Reproducible Transparent

– Footnotes– Judgments

GRADE profiles– Summary

Integration with Revman Real time

Judgements about Judgements about recommendationsrecommendations

• “We recommend”…”should” …“Do it”• “We suggest”…”may” … “Probably do it” • “We suggest not”… “may not” …“Probably

don’t do it”• “We recommend not”…”should not”… “Don’t

do it”

No recommendation

This could include considerations of costs; i.e. “Is the net gain (benefits-downsides) worth the costs?”

Questions?Questions?

What are we concerned about?What are we concerned about?

methodological quality of evidence– likelihood of bias– high, moderate, low and very low

quality

strength of recommendations– must do to might do– strong recommendation or weak


recommendations against– uncertainty of benefit– confidence in down sides

whole body CT or MRI screening– maybe benefit, maybe not– true positives some harm– false positive some harm


known benefit, strong recommendation for one of two alternatives– CABG for left main stem disease– but venous or mammary artery graft

benefit: weak evidence suggests appreciable benefit for mammary artery

harm: strong evidence that little difference

Population: In patients with chronic atrial fibrillation and no prior history

of stroke

Intervention: does oral anticoagulation (comparison) compared with no therapy

Outcome: reduce the risk for embolic stroke, hemorrhage

and death?

Different risk groups:Low, moderate, high

Other outcomes:Inconvenience, quality of

life


known benefit, strong recommendation for one of two alternatives

benefit: strong evidence of equivalence

harm: weak evidence that harm differs appreciably

Grading evidence and recommendations Holger Schünemann Gunn Vist Gordon Guyatt for the GRADE Working Group.

Documents

grade example

grade meeting

grading evidence

evidence recommendationii

strength of recommendations

informal working group

bmj paper

common system