Presentation on theme: "1 Health and Disease in Populations 2002 Week 9 – 2/5/02 Randomised controlled trials 2 Dr Jenny Kurinczuk."— Presentation transcript:
1 Health and Disease in Populations 2002 Week 9 – 2/5/02 Randomised controlled trials 2 Dr Jenny Kurinczuk
2 Lecture objectives You should be able to: 1.Describe suitable outcome measures for trials 2.Discuss how to deal with losses to follow-up and non-compliance
3 Lecture objectives 3.Differentiate between explanatory and pragmatic trials and be able to explain the meaning of the term ‘analysis by intention-to-treat’ 4. Explain the problems associated with non-randomised trials and the use of historical controls
4 Recap - the basic steps involved RCTs are experimental studies In contrast to observational studies the researcher has control over who receives what exposure -the exposure is the treatment A or B Is new treatment A better than old treatment B ?
5 Recap - the basic steps involved Define the disease (inclusions & exclusions) Define the treatments, A & B (issue of placebos & placebo effect) Find a suitable group of patients Invite them and consent them Randomly allocate them to A or B (double blind - Dr & patient both blind) Follow-up, define and compare the outcomes in the two groups
6 Objective 1: Suitable ‘outcome’ measures for trials Evaluation of patient progress needs to be: –Objective –Accurate –Consistent – reproducible –Blind to treatment allocation –Evaluated against a pre-defined end point
7 Outcome measures for trials Choice of primary outcome measure –eg cancer treatment: Survival time Reduction in tumour size Duration of tumour response Quality of life Choose one of these at the outset May wish to measure the others as secondary end-points together with side effects
8 Outcome measures for trials Method of assessment ( generally more rigorous than usual clinical practice ) involves: –Baseline assessment & recording –Monitoring of progress and side- effects during the treatment period –Assessment of the primary outcome (secondary and side-effects)
9 Outcome measures for trials Baseline assessment & recording: –Clinical condition using a predefined ‘measurement’ protocol eg: sitting BP measured using calibrated automated machine Size/area of wound Pain measured using validate scale –Personal characteristics (eg age, sex) –Relevant medical history –Other test results possibly eg ECG
10 Outcome measures for trials Assessment of the primary outcome (secondary and side-effects): –Pre-defined measure using predefined ‘measurement’ protocol: sitting BP measured using calibrated automated machine Size/area of wound Pain measured using validate scale –Define when and how often to measure and at what point the end of the trial is reached
11 Objective 2: Dealing with losses to follow-up and non-compliance So, assume we have randomised the patients to the two treatments and given them the drugs (tablets) We may see them at intervals to monitor progress and at the pre-defined end-point to determine the outcome
12 Losses to follow-up and non-compliance Losses to follow-up: Not everyone continues in the trial –Clinical condition requires removal from the trial (this is appropriate) –Patient opts out (this is unfortunate)
13 Losses to follow-up and non-compliance Losses to follow-up: Difficult to minimise –Make the follow-up for the patients practical and not too onerous –Make the trial requirements clear at the outset (prior to consent) –Avoid bribery and coercion –Avoid direct payment but pay travel & out-of-pocket expenses
14 Losses to follow-up and non-compliance Non-compliance: –Not everyone takes the treatment either properly or at all Difficult to minimise: - Provide clear written instructions - Tablet count –Monitor with blood or urine tests (if possible)
15 Losses to follow-up and non-compliance Need to deal with: Losses to follow-up and Non-compliance appropriately in the analysis
16 Objective 3: Explanatory and pragmatic trials Is treatment A better than treatment B? Two different ways to interpret this question: 1.Is the physiological action of drug A better than drug B ? –eg In the treatment for high blood pressure does drug A actually cause a greater fall in blood pressure than drug B ?
17 Is treatment A better than B ? 2.Would the replacement of drug B with drug A benefit patients in routine clinical practice ? eg Treatment for high blood pressure. If you prescribe A in practice, is A more effective at reducing blood pressure than B ? Will the patients take the drug and will it have the desired effect?
18 Reality of what happens Randomise ABTake Don’t take
19 Explanatory trials – ‘as treated’ analysis Is the physiological action of drug A better than drug B ? Randomise ABTake Don’t take Ignore those who don’t take the drug
20 Pragmatic (practical) trial – ‘analysis by intention to treat’ In routine clinical practice is A better than B ? Randomise ABTake Don’t take Compare by ‘intention to treat’ – regardless of whether they actually took the treatment!
21 Pragmatic trial – ‘analysis by intention to treat’ - Comparison ‘by intention to treat’ takes into account factors like the patients don’t like the drugs, forget to take them (doses too frequent) can’t swallow the drugs (too large), there are too many side effects…… and so on…… If all these are worse in A than B, although at a biological level A might work better than B, in practice A would be less effective than B
22 Example – BP treatment
23 Objective 4: Why bother with RCTs? Why not use historical controls? –Give new patients the new treatment (A) and for comparison of treatment effect use information about previous patients treated with the old treatment (B)
24 Problems with historical controls Old patient selection less well defined Old patient selection less rigorous Less information about the controls Controls may have been treated differently in other ways, apart from just the difference in the drug Control of confounding is poor The results always over-estimate the benefits of the new treatment
25 Summary 1.Suitable outcome measures 2.Losses to follow-up & non-compliance 3.Explanatory versus pragmatic trials 4.Why historical controls can’t be used instead of RCTs