Holger Schünemann Professor of Clinical Epidemiology, Biostatistics and Medicine McMaster University, Hamilton, Canada Italian NCI „Regina Elena“, Rome, Italy

Content  Describe the grade of recommendation and what each category means: strong/weak and optional language  How the quality of evidence can be upgraded/down-graded  What happens when you’re recommending something not be done?  Maybe provide some ID-type examples if possible - I’m attaching our clinical questions that may be used as examples?  Provide a quick tutorial of GRADEPro  Questions

The GRADE approach Clear separation of 2 issues: 1) 4 categories of quality of evidence: very low, low, moderate, or high quality?  methodological considerations  likelihood of systematic deviation from truth  by outcome 2) Recommendation: 2 grades – weak/conditional or strong (for or against)?  Quality of evidence only one factor  Influenced by magnitude of effect(s) – balance of benefits and harms, values and preferences, cost *

Content  Describe the grade of recommendation and what each category means: strong/weak and optional language  How the quality of evidence can be upgraded/down-graded  What happens when you’re recommending something not be done?  Maybe provide some ID-type examples if possible - I’m attaching our clinical questions that may be used as examples?  Provide a quick tutorial of GRADEPro  Questions

Implications of a strong recommendation  Patients: Most people in this situation would want the recommended course of action and only a small proportion would not  Clinicians: Most patients should receive the recommended course of action  Policy makers: The recommendation can be adapted as a policy in most situations

Implications of a weak recommendation  Patients: The majority of people in this situation would want the recommended course of action, but many would not  Clinicians: Be prepared to help patients to make a decision that is consistent with their own values/decision aids and shared decision making  Policy makers: There is a need for substantial debate and involvement of stakeholders

Content  Describe the grade of recommendation and what each category means: strong/weak and optional language  How the quality of evidence can be upgraded/down-graded  What happens when you’re recommending something not be done?  Maybe provide some ID-type examples if possible - I’m attaching our clinical questions that may be used as examples?  Provide a quick tutorial of GRADEPro  Questions

Answer Same type of interpretation

Content  Describe the grade of recommendation and what each category means: strong/weak and optional language  How the quality of evidence can be upgraded/down-graded  What happens when you’re recommending something not be done?  Maybe provide some ID-type examples if possible - I’m attaching our clinical questions that may be used as examples?  Provide a quick tutorial of GRADEPro  Questions

Determinants of quality - For body of evidence -  RCTs start high  observational studies start low  5 factors that can lower quality 1. limitations of detailed design and execution 2. inconsistency 3. Indirectness/applicability 4. publication bias 5. Imprecision  3 factors can increase quality 1. large magnitude of effect 2. all plausible confounding may be working to reduce the demonstrated effect or increase the effect if no effect was observed 3. dose-response gradient

Assessing the quality of evidence 11

1. Design and Execution  limitations  lack of concealment  intention to treat principle violated  inadequate blinding  loss to follow-up  early stopping for benefit  selective outcome reporting  Example: RCT suggests that danaparoid sodium is of benefit in treating HIT complicated by thrombosis  Key outcome: clinicians’ assessment of when the thromboembolism had resolved  Not blinded – subjective judgement

2. Inconsistency of results (Heterogeneity)  if inconsistency, look for explanation  patients, intervention, outcome, methods  unexplained inconsistency downgrade quality  Bleeding in thrombosis-prophylaxed hospitalized patients  seven RCTs  4 lower, 3 higher risk

Example: Bleeding in the hospital Dentali et al. Ann Int Med, 2007

 Judgment  variation in size of effect  overlap in confidence intervals  statistical significance of heterogeneity I2I2

Akl E, Barba M, Rohilla S, Terrenato I, Sperati F, Schünemann HJ. “Anticoagulation for the long term treatment of venous thromboembolism in patients with cancer”. Cochrane Database Syst Rev Apr 16;(2):CD Heparin or vitamin K antagonists for survival in patients with cancer

Non-steroidal drug use and risk of pancreatic cancer Capurso G, Schünemann HJ, Terrenato I, Moretti A, Koch M, Muti P, Capurso L, Delle Fave G. Meta-analysis: the use of non-steroidal anti-inflammatory drugs and pancreatic cancer risk for different exposure categories. Aliment Pharmacol Ther Oct 15;26(8):

3. Directness of Evidence  differences in  populations/patients (mild versus severe COPD, older, sicker or more co-morbidity)  interventions (all inhaled steroids, new vs. old)  outcomes (important vs. surrogate; long-term health- related quality of life, short –term functional capacity, laboratory exercise, spirometry)  indirect comparisons  interested in A versus B  have A versus C and B versus C  formoterol versus salmeterol versus tiotropium

Alendronate Risedronate Placebo Directness interested in A versus B available data A vs C, B vs C

4. Publication Bias & Imprecision  Publication bias  number of small studies

Egger M, Smith DS. BMJ 1995;310: I.V. Mg in acute myocardial infarction Publication bias Meta-analysis Yusuf S.Circulation 1993 ISIS-4 Lancet 1995

Egger M, Cochrane Colloquium Lyon Funnel plot Standard Error Odds ratio Symmetrical: No publication bias

Egger M, Cochrane Colloquium Lyon Funnel plot Standard Error Odds ratio Asymmetrical: Publication bias?

Egger M, Smith DS. BMJ 1995;310: I.V. Mg in acute myocardial infarction Publication bias Meta-analysis Yusuf S.Circulation 1993 ISIS-4 Lancet 1995

Egger M, Smith DS. BMJ 1995;310: Meta- analysis confirme d by mega- trials

26 Publication bias (File Drawer Problem)  Faster and multiple publication of “positive” trials  Fewer and slower publication of “negative” trials

5. Imprecision  small sample size  small number of events  wide confidence intervals  uncertainty about magnitude of effect  how to decide if CI too wide?  grade down one level?  grade down two levels?  extent to which confidence in estimate of effect adequate to support decision

Example: Bleeding in the hospital Dentali et al. Ann Int Med, 2007

Offer all effective treatments?  atrial fib at risk of stroke  warfarin increases serious gi bleeding  3% per year  1,000 patients 1 less stroke  30 more bleeds for each stroke prevented  1,000 patients 100 less strokes  3 strokes prevented for each bleed  where is your threshold?  how many strokes in 100 with 3% bleeding?

01.0%

0

0

0

What can raise quality? 1. large magnitude can upgrade (RRR 50%)  very large two levels (RRR 80%)  common criteria  everyone used to do badly  almost everyone does well  oral anticoagulation for mechanical heart valves  insulin for diabetic ketoacidosis  hip replacement for severe osteoarthritis 2. dose response relation (higher INR – increased bleeding) 3. all plausible confounding may be working to reduce the demonstrated effect or increase the effect if no effect was observed

All plausible confounding would result in an underestimate of the treatment effect  Higher death rates in private for-profit versus private not-for-profit hospitals  patients in the not-for-profit hospitals likely sicker than those in the for-profit hospitals  for-profit hospitals are likely to admit a larger proportion of well-insured patients than not-for- profit hospitals (and thus have more resources with a spill over effect)

All plausible biases would result in an overestimate of effect  Hypoglycaemic drug phenformin causes lactic acidosis  The related agent metformin is under suspicion for the same toxicity.  Large observational studies have failed to demonstrate an association  Clinicians would be more alert to lactic acidosis in the presence of the agent

Content  Describe the grade of recommendation and what each category means: strong/weak and optional language  How the quality of evidence can be upgraded/down-graded  What happens when you’re recommending something not be done?  Maybe provide some ID-type examples if possible - I’m attaching our clinical questions that may be used as examples?  Provide a quick tutorial of GRADEPro  Questions

Relevant clinical question? Example from a not so common disease Clinical question: Population: Avian Flu/influenza A (H5N1) patients Intervention: Oseltamivir (or Zanamivir) Comparison: No pharmacological intervention Outcomes: Mortality, hospitalizations, resource use, adverse outcomes, antimicrobial resistance Schunemann et al. The Lancet ID, 2007

Methods – WHO Rapid Advice Guidelines for management of Avian Flu  Applied findings of a recent systematic evaluation of guideline development for WHO/ACHR  Group composition (including panel of 13 voting members):  clinicians who treated influenza A(H5N1) patients  infectious disease experts  basic scientists  public health officers  methodologists  Independent scientific reviewers:  Identified systematic reviews, recent RCTs, case series, animal studies related to H5N1 infection

Evidence Profile Oseltamivir for treatment of H5N1 infection: - -

Oseltamivir for Avian Flu Summary of findings:  No clinical trial of oseltamivir for treatment of H5N1 patients.  4 systematic reviews and health technology assessments (HTA) reporting on 5 studies of oseltamivir in seasonal influenza.  Hospitalization: OR 0.22 (0.02 – 2.16)  Pneumonia: OR 0.15 ( )  3 published case series.  Many in vitro and animal studies.  No alternative that is more promising at present.  Cost: ~ 40$ per treatment course Schunemann et al. Lancet ID, 2007 & PLOS Medicine 2007

Determinants of the strength of recommendation

Example: Oseltamivir for Avian Flu Recommendation: In patients with confirmed or strongly suspected infection with avian influenza A (H5N1) virus, clinicians should administer oseltamivir treatment as soon as possible (????? recommendation, very low quality evidence). Schunemann et al. The Lancet ID, 2007

Example: Oseltamivir for Avian Flu Recommendation: In patients with confirmed or strongly suspected infection with avian influenza A (H5N1) virus, clinicians should administer oseltamivir treatment as soon as possible (strong recommendation, very low quality evidence). Values and Preferences Remarks: This recommendation places a high value on the prevention of death in an illness with a high case fatality. It places relatively low values on adverse reactions, the development of resistance and costs of treatment. Schunemann et al. The Lancet ID, 2007

Other explanations Remarks: Despite the lack of controlled treatment data for H5N1, this is a strong recommendation, in part, because there is a lack of known effective alternative pharmacological interventions at this time. The panel voted on whether this recommendation should be strong or weak and there was one abstention and one dissenting vote.

Strength of recommendation  “The strength of a recommendation reflects the extent to which we can, across the range of patients for whom the recommendations are intended, be confident that desirable effects of a management strategy outweigh undesirable effects.”  Strong or weak/conditional

Quality of evidence & strength of recommendation  Linked but no automatism  Other factors beyond the quality of evidence influence our confidence that adherence to a recommendation causes more benefit than harm  Systems/approaches failed to make this explicit  GRADE separates quality of evidence from strength of recommendation

Content  Describe the grade of recommendation and what each category means: strong/weak and optional language  How the quality of evidence can be upgraded/down-graded  What happens when you’re recommending something not be done?  Maybe provide some ID-type examples if possible - I’m attaching our clinical questions that may be used as examples?  Provide a quick tutorial of GRADEPro  Questions

Creating a new GRADEpro file

Profile groups Profiles

Profiles: Questions

Importing a RevMan 5 file of a systematic review Imported data from RevMan 5 file: outcomes meta-analyses results bibliographic information

Managing outcomes to include a maximum of 7

Entering/editing information for dichotomous outcomes

Entering/editing information to grade the quality of the evidence

Content  Describe the grade of recommendation and what each category means: strong/weak and optional language  How the quality of evidence can be upgraded/down-graded  What happens when you’re recommending something not be done?  Maybe provide some ID-type examples if possible - I’m attaching our clinical questions that may be used as examples?  Provide a quick tutorial of GRADEPro  Questions

Example: Oseltamivir for Avian Flu Recommendation: In patients with confirmed or strongly suspected infection with avian influenza A (H5N1) virus, clinicians should administer oseltamivir treatment as soon as possible (strong recommendation, very low quality evidence). Values and Preferences This recommendation places a high value on the prevention of death in an illness with a high case fatality. It places relatively low values on adverse reactions, the development of resistance and costs of treatment. Remarks Despite the lack of controlled treatment data for H5N1, this is a strong recommendation, in part, because there is a lack of known effective alternative pharmacological interventions at this time. Schunemann et al. The Lancet ID, 2007

Confidence in evidence  There always is evidence  “When there is a question there is evidence”  Research evidence alone is never sufficient to make a clinical decision  Better research  greater confidence in the evidence and decisions

77 Sequence gen. Allocation concealment Blinding/Masking Intention-to-treat analysis Blinding/Masking Baseline Allocation A B InterventionNo interv. Follow up Outcome Method Random? Selection bias? Performance bias? Attrition bias? Detection bias? Question Factors leading to bias? Can you explain them? P-values and confidence intervals important? CONSENSUS ALWAYS REQUIRED

Limitations of existing systems  confuse quality of evidence with strength of recommendations  lack well-articulated conceptual framework  criteria not comprehensive or transparent  GRADE unique  breadth, intensity of development process  wide endorsement and use  conceptual framework  comprehensive, transparent criteria

G rades of R ecommendation A ssessment, D evelopment and E valuation *Grade Working Group. CMAJ 2003, BMJ 2004, BMC 2004, BMC 2005, AJRCCM 2006, BMJ 2008

GRADE Working Group David Atkins, chief medical officer a Dana Best, assistant professor b Martin Eccles, professor d Francoise Cluzeau, lecturer x Yngve Falck-Ytter, associate director e Signe Flottorp, researcher f Gordon H Guyatt, professor g Robin T Harbour, quality and information director h Margaret C Haugh, methodologist i David Henry, professor j Suzanne Hill, senior lecturer j Roman Jaeschke, clinical professor k Regina Kunx, Associate Professor Gillian Leng, guidelines programme director l Alessandro Liberati, professor m Nicola Magrini, director n James Mason, professor d Philippa Middleton, honorary research fellow o Jacek Mrukowicz, executive director p Dianne O ’ Connell, senior epidemiologist q Andrew D Oxman, director f Bob Phillips, associate fellow r Holger J Sch ü nemann, professor g,s Tessa Tan-Torres Edejer, medical officer t David Tovey, Editor y Jane Thomas, Lecturer, UK Helena Varonen, associate editor u Gunn E Vist, researcher f John W Williams Jr, professor v Stephanie Zaza, project director w a) Agency for Healthcare Research and Quality, USA b) Children's National Medical Center, USA c) Centers for Disease Control and Prevention, USA d) University of Newcastle upon Tyne, UK e) German Cochrane Centre, Germany f) Norwegian Centre for Health Services, Norway g) McMaster University, Canada h) Scottish Intercollegiate Guidelines Network, UK i) F é d é ration Nationale des Centres de Lutte Contre le Cancer, France j) University of Newcastle, Australia k) McMaster University, Canada l) National Institute for Clinical Excellence, UK m) Universit à di Modena e Reggio Emilia, Italy n) Centro per la Valutazione della Efficacia della Assistenza Sanitaria, Italy o) Australasian Cochrane Centre, Australia p) Polish Institute for Evidence Based Medicine, Poland q) The Cancer Council, Australia r) Centre for Evidence-based Medicine, UK s) National Cancer Institute, Italy t) World Health Organisation, Switzerland u) Finnish Medical Society Duodecim, Finland v) Duke University Medical Center, USA w) Centers for Disease Control and Prevention, USA x) University of London, UK Y) BMJ Clinical Evidence, UK

GRADE Uptake  World Health Organization  National Institute Clinical Excellence (NICE)  Allergic Rhinitis in Asthma Guidelines (ARIA)  American Thoracic Society  British Medical Journal  Infectious Disease Society of America  American College of Chest Physicians  UpToDate  American College of Physicians  Cochrane Collaboration  Infectious Disease Society of America  European Society of Thoracic Surgeons  Clinical Evidence  Agency for Health Care Research and Quality (AHRQ)  Over 20 major organizations

The GRADE approach Clear separation of 2 issues: 1) 4 categories of quality of evidence: very low, low, moderate, or high quality?  methodological quality of evidence  likelihood of systematic deviation from truth  by outcome 2) Recommendation: 2 grades – weak/conditional or strong (for or against)?  Quality of evidence only one factor  Influenced by magnitude of effect(s) – balance of benefits and harms, values and preferences, cost *

GRADE Quality of Evidence “Extend of confidence on how adequate the estimate of effect is to support decision”  high: considerable confidence in estimate of effect.  moderate: further research likely to have impact on confidence in estimate, may change estimate.  low: further research is very likely to impact on confidence, likely to change the estimate.  very low: any estimate of effect is very uncertain

Developing recommendations

Conclusion  clinicians, policy makers need summaries that separate:  quality of evidence  strength of recommendations  explicit rules  transparent, informative  GRADE  four categories of quality of evidence  two grades for strength of recommendations  transparent, systematic by and across outcomes  applicable to diagnosis  wide adoption

Consistency of results  consistency of results  if inconsistency, look for explanation  patients, intervention, outcome, methods  unexplained inconsistency downgrade quality  oxygen for day-to-day dyspnea in COPD with exercise hypoxemia  five cross-over RCTs oxygen versus placebo  4 no benefit, 1 substantial benefit

Evidence profiles

Directness of Evidence  indirect comparisons  interested in A versus B  have A versus C and B versus C  formoterol versus salmeterol versus tiotropium  Acetylcysteine alone for Pulmonary Fibrosis  (all that is available is Acetylcysteine + Prednisone + Azathioprine vs. Prednisone + Azathioprine)

Directness of Evidence  differences in  patients (inhalers for mild versus moderate to severe COPD)  interventions (all inhaled steroids versus those used in clinical trials – drug class effect)  outcomes (long-term health-related quality of life, short – term functional capacity, laboratory exercise, spirometry)

Reporting Bias & Imprecision  reporting bias  reporting of studies  publication bias  number of small studies  reporting of outcomes  small sample size  small number of events  wide confidence intervals  uncertainty about magnitude of effect

Differences in exercise capacity in short-term randomized trials of oxygen in COPD patients.

What can raise quality?  large magnitude can upgrade (RRR 50%)  very large two levels (RRR 80%)  common criteria  everyone used to do badly  almost everyone does well  Insulin in diabetic ketoacidosis  dose response relation (smoking - cancer)

The clinical scenario A 68 year old male long-term patient of yours. He suffers from COPD but is unable to stop smoking after over 30 years of tobacco use. He has been taking beta-carotene supplements for several months because someone in the “healthy food” store recommended it to prevent cancer. He wants to know whether this will prevent him from getting cancer and whether he should use beta-carotene.

The clinical question Population: In patients with COPD Intervention: does beta-carotene suppl Comparison: compared to no suppl. Outcomes: reduce the risk of lung cancer?

Where do you look for an answer?

Clinical Practice Guidelines Systematically developed statements to assist practitioner and patient decisions about appropriate health care for specific clinical circumstances Institute of Medicine, 1992

Determinants of quality  RCTs start high  observational studies start low  5 Factors that lower quality (bias)  3 Factors that increase quality (bias is unlikely to explain observed effect)  Final quality by outcome:  High  Moderate  Low  Very low

Design and Execution  limitations  lack of concealment  intention to treat principle violated  inadequate blinding  loss to follow-up  early stopping for benefit  13 RCTs bacterial extract (immunomodulation) for preventing exacerbation  unclear concealment of randomization  questionable intention to treat  inadequate attention to loss to follow-up

Consistency of results  consistency of results  if inconsistency, look for explanation  patients, intervention, outcome, methods  unexplained inconsistency downgrade quality  oxygen for day-to-day dyspnea in COPD with exercise hypoxemia  five cross-over RCTs oxygen versus placebo  4 no benefit, 1 substantial benefit

Directness of Evidence  indirect comparisons  interested in A versus B  have A versus C and B versus C  formoterol versus salmeterol versus tiotropium  differences in  patients (mild versus severe COPD)  interventions (all inhaled steroids)  outcomes (long-term health-related quality of life, short – term functional capacity, laboratory exercise, spirometry)

How should recommendations be formulated and presented?  Few written standards exist  For strong recommendations, the GRADE working group has suggested adopting terminology such as, “We recommend…” or “Clinicians should…”.  For weak recommendation, they should use less definitive wording, “We suggest…” or “Clinicians might…”.

Clinicians and patients want to know!  1) UpToDate ® Users  2) Mini Medical School attendees*: Participants preferred to know about the uncertainty relating to outcomes of a treatment or a test more interested in knowing about uncertainty relating to benefits than harms (96% vs. 90%; P<0.001). strong preference to be informed about the quality of evidence that supports a recommendation. *Akl et al. J Clin Epi, 2007, in press

GRADE Quality of Evidence Extent to which confidence in estimate of effect adequate to support decision  high: considerable confidence in estimate of effect.  moderate: further research likely to have impact on confidence in estimate, may change estimate.  low: further research is very likely to impact on confidence, likely to change the estimate.  very low: any estimate of effect is very uncertain

There always is evidence The better the research and the evidence, the more confident the decision Evidence alone is never sufficient to make a clinical decision

Do evidence based guidelines make a difference? Non-rigorous guidelines: Create noise & bias Make more aggressive recommendations Can harm patients and impair research efforts Can reduce credibility of professional societies Evidence-based clinical practice guidelines can: reduce delivery of inappropriate care support introduction of new knowledge into clinical practice Grimshaw et al (1992); Woolf et al (1999); Fretheim et al (2002)