Evidence Based Medicine &Basic Critical Appraisal David Erskine London & SE Medicines Information Service
Why do we need evidence? • Resources should be allocated to things that are EFFECTIVE • The only way of judging effectiveness is EVIDENCE • “In God we trust – all others bring data”
Why do we need evidence? • Move towards: EVIDENCE-BASED MEDICINE • Move away from: EMINENCE- BASED MEDICINE What we really, really want is EVIDENCE-INFORMED MEDICINE
Evidence-based medicine 5 steps • Converting information needs into answerable questions • Finding the best evidence with which to answer question • Critically appraising evidence for validity and usefulness • Applying the results in clinical practice • Evaluating performance
What is an answerable question? - step 1 • Should contain the following components: • The intervention you are interested in • The population you are interested in • The outcomes you are interested in
Finding evidence - step 2Workshop • What are ‘good’ sources of evidence? • Think about sources of evidence that you have used – were they useful? • What makes a ‘good’ source of evidence?
Finding evidence - step 2 • Can search many of these sources simultaneously using: • TRIP (via NLH) • OSRS (via NeLM) • NKS single search engine(via NLH)?????
Search Strategy Tertiary Secondary Primary Need to be systematic vs. time & resources Start from scratch vs. build on previous work
Appraising evidence - step 3 • Validity - closeness to the truth, i.e. do we believe it? • Usefulness - clinical applicability, i.e. is it important? Using efficacy data from clinical trials to estimate Clinical Effectiveness
Assessing validity • Results of a trial affected by: Chance Bias Effect of intervention • Judge validity by: Extent to which bias is eliminated(Randomisation, type of analysis, blinding) Must also account for chance Assessed by using questions 1-7 on checklist
Exposed Population Outcome Sample Time Outcome Not exposed Cohort study
Study Population Exposed Cases Not exposed Controls Exposed Not exposed Time Case control study
Randomisation Population Experimental intervention Outcome Sample Outcome Control intervention Time Randomised, controlled trial
Are the results of the trial valid? • Did the trial address a clearly focused issue? • Define the population studied • The intervention given • The outcomes considered • Is an RCT an appropriate method to answer this question?
How were patients assigned to treatment groups? • What is randomisation? • Why is it important? • What are acceptable methods of randomisation? • What methods of randomisation are considered dubious? • Where to find out details of the randomisation process used? Note CONSORT statement- recommend that RCTs should report how the allocation sequence was generated and concealed until the patient was randomised
How were patients assigned to treatment groups? • Could differences between groups have altered the outcome? • Are any differences reported? • Where do you check for yourself? • Was any method used to balance randomisation (stratification)?
Were patients, health workers and study personnel “blind” to treatment? • Why is this important? • When is it less important? • How is it done? Want to minimise likelihood of observer bias and Hawthorne effects
Were all the patients who entered the trial properly accounted for at its conclusion? • Was follow-up complete?- if not how was missing data handled? • Were patients analysed in the groups to which they were randomised? Intention to treat analysis vs. Per protocol or completer analysis
Aside from the experimental investigation, were the two groups treated in the same way? • Usually not such an important issue in placebo-controlled or comparative drug trials but assessing for possible bias introduced by differing intensity of follow up
Did the study have enough participants to minimise the play of chance? • Look for a power calculation • Do you agree with their definition of “important”
How are the results presented • What is being described: proportions of people, a measurement (mean or median differences), a survival curve? • How was it expressed? With proportions see terms like Relative risk reduction, absolute risk reduction, number needed to treat • Do you understand what is being expressed? • Is it clinically significant?
Expressing resultsFive years treatment with:Intervention A reduces cardiac events by 34%Intervention B causes an absolute reduction of cardiac events of 1.4%Intervention C increases the rate of event free patients from 95.9% to 97.3%Intervention D 71 patients need to be treated to avoid one cardiac eventIntervention E reduces cardiac events by 34% with a 6% relative increase in mortality(Lancet 1994; 343: 1209-1211)
Quantifying benefit OUTCOME YesNo Control interventionab Experimental interventioncd
Quantifying benefit Relative risk reduction – RRR= CER – EER [a/a+b – c/c+d] CER [ a/a+b] Absolute risk reduction – ARR= CER – EER a/a+b – c/c+d Number needed to treat – NNT = _1_ ARR Relative risk – RR = EERc/c+d CER a/a+b Odds ratio – OR = odds of event in intervention groupc/d odds of event in control group a/b
RCT example - 4S study • Stable angina or myocardial infarction more than 6 months previously • Serum cholesterol > 6.2mmol/l • Excluded patients with arrhythmia's and heart failure • All patients given 8 weeks of dietary therapy • If cholesterol still raised (>5.5) randomised to receive simvastatin (20mg > 40mg) or placebo • Outcome death or myocardial infarction (length of treatment 5.4 years ) were the outcomes
Example – the 4S study OUTCOME Dead Alive Control group (placebo) n = 2223 256 1967 Experimental group (simvastatin) n = 2221182 2039 Control event rate = 256 = 0.115 (11.5%) 2223 Control odds of event = 256 = 0.130 1967 Experimental event rate = 182 = 0.082 (8.2%) 2221 Experimental odds of event = 182 = 0.089 2039
Example – the 4S study Relative risk reduction CER – EER = 0.115 – 0.082 = 0.29 (29%) RRR CER 0.115 Absolute risk reduction CER – EER= 0.115 – 0.082 = 0.033 (3.3%)ARR Number needed to treat _1_ = _1_ = 30.1NNT ARR 0.033 Relative risk EER = 0.082 = 0.71 RR CER 0.115 Odds ratio = odds of event in experimental group = 0.089 = 0.69OR odds of event in control group 0.130
Workshop –revision • Assuming a treatment is beneficial and you are trying to prevent an unpleasant end-point. Outline what type of values you would expect for the following measures of efficacy: • Relative risk reduction • Relative risk • Odds ratio • Absolute risk reduction • NNT
NNT EXAMPLES Outcome NNT Intervention
How precise are the results? • Limitations of the p-value- what does it tell you? • Use of confidence intervals • What are confidence intervals? • What does upper and lower limits tell you? • Would your decision about whether to use this treatment be the same at the upper and lower confidence intervals?
How precise are the results? • CHANCE - p = 1 in 20 (0.05). • > 1 in 20 (0.051) = not significant • < 1 in 20 (0.049) = statistically significant • CONFIDENCE INTERVALS • the range of values between which we could be 95% confidant that this result would lie if this trial was carried out 100 times (taken as result for general population)
Confidence Interval- how to calculate in some circumstances The range of values which includes the true population value 95% of the time e.g.Confidence interval around ARR = +/- 1.96CER x (1 - CER) + __EER x (1 - EER)__ N Control group N Experimental group
Can the results be applied to the local population? • Look at inclusion and exclusion criteria- are they similar • Are there differences which make this trial irrelevant? • Be pragmatic!!
Were all the clinically important outcomes considered? • Consider outcomes reported from a range of perspectives- individual, family/ carers, policy maker, healthcare professional, wider community • Reports tend to concentrate on efficacy at the expense of harms and quality of life measures (participants and carers?)
Should policy or practice change as a result of the evidence contained in this trial? • Are the likely benefits worth the potential harms and costs? • Have you got all the data you need to make this assessment • Can you access any missing data you need • How does this intervention compare with other interventions-and the results of related RCTs (if any) • Also need to consider service implications
Applying results to local practice - step 4 • Local policies • Guideline development • Implementing clinical effectiveness and clinical governance agendas
Evaluating performance - step 5 • Clinical audit!
Evidence based medicine Formulate question Efficiently track Down best Available Evidence Evaluate Performance Implement Changes In clinical Practice Critically review the Validity and usefulness Of the evidence
Hierarchy of evidence (used in NICE Guidelines) ) Ia Evidence from systematic review of meta-analysis of randomised controlled trials Ib Evidence from at least one randomised controlled trial IIa Evidence from at least one well-designed controlled study without randomisation IIb Evidence from at least one other type of well-designed quasi- experimental study III Evidence from well-designed non-experimental descriptive studies such as comparative studies, or case studies IV Evidence from expert committee reports or opinions and/or clinical experiences of respected authorities
Strength of recommendations used in NICE Guidelines Grade A At least one randomised controlled trial as part of a body of literature of overall good quality and consistency addressing the specific recommendation (evidence levels 1a and 1b) Grade B Well conducted clinical studies but no randomised clinical trials on the topic of the recommendation (evidence levels Iia, IIb, III) Grade C Expert committee reports or opinions and/or clinical experience of respected authorities. This grading indicates that directly applicable clinical studies are absent (evidence level IV) Good practice point: recommended good practice based on the clinical experience of the Guideline Development Group
Recommendations from the BHS Guideline 1999 • Drug therapy should be started in all patients with sustained systolic BP >/= 160mmHg or sustained diastolic BP >/= 100mmHg despite non-pharmacological measures (A) • To reduce overall cardiovascular risk patients should stop smoking (B) • At least two BP measurements should be made at each visit and 4 visits to determine blood pressure (C)
Criticisms of evidence-based medicine • Effective therapies might be rejected on the basis of “absence of evidence” of efficacy rather than “evidence of absence” of efficacy. • “Cook book” medicine • Can take a long time to gather the evidence • When is it appropriate to generalise across populations? • Use of surrogate markers, class effects
Additional criteria in assessing trials claiming therapeutic equivalence • Was the active control previously shown to be effective? • Were study patients and outcome variables similar to those in the original trials that established the efficacy of the active control? • Were both regimens applied in optimal fashion? • Was the appropriate null hypothesis tested? • Was the equivalence margin specified before the study? • Was the trial large enough? • Was analysis by intention to treat AND on-treatment? • PLUS usual assessment of size and precision of treatment effect!
REFERENCES • CASP Tool • www.phru.nhs.uk/casp/resourcescasp.htm • Additional Reading • Montori V et al Users’ guide to detecting misleading claims in clinical research reports BMJ 2004; 329: 1093-1096 • Greenhalgh T. How to read a paper: Getting your bearings (deciding what a paper is about). BMJ 1997; 315: 243-246 • Greenhalgh T. How to read a paper: Assessing the methodological quality of published papers. BMJ 1997; 315: 305-308 • Guyatt GH, et al. Users’ Guides to the Medical Literature: II. How to use an article about therapy or prevention. A. Are the results of the study valid? JAMA 1993; 270: 2598-2601 • Guyatt GH, et al. Users’ Guides to the Medical Literature: II. How to use an article about therapy or prevention. B. What are the results and will they help me in caring for my patients? JAMA 1994; 271: 59-63
REFERENCES cont. Statistics for the non-statistician I: Different types of data need different statistical tests Greenhalgh T BMJ 1997; 315: 364-6 II: “Significant” relations and their pitfalls Greenhalgh T BMJ 1997; 315: 422-5 Statistics in small doses- a series of articles available on UKMI web-site (http://www.ukmi.nhs.uk/activities/Research/default.asp?pageRef=27)