1
Q
How do we know what is effective?
A
1, Authority figures tell us - but they may be wrong
- There is an underlying theory that explains why something should work - theories often don’t work in practice - underlying theory from the pathophysiology of the disease supports the use of a treatment (involves applying logic)
- Intervention studies - use science (obs evidence through studies,
2
Q
What are the diff types of intervention studies?
A
- Uncontrolled before and after study
- Controlled before and after study
- Randomised controlled trial
- > Indiv randominsation
- > Cluster randomisation
3
Q
What are the key features of a randomised controlled trial?
A
- Randomisation
- Concealment of allocation
- Blinding
- Intention to treat analysis
4
Q
What are the different sections of CASP checklist for RCT?
A
Section A: Are the results of the trial valid?
Section B: What are the results?
Section C: Will the results help locally?
5
Q
What do outcome measures in RCTs need to be?
A
- Reliable - ability to produce consistent, reproducible estimates of true effect
- Valid - measures the construct that it claims to measure
- Responsive - can detect changes in the construct to be measured over time/ the extent to which changes in the scores are ass. with changes in the underlying clinical status of the subject
6
Q
What needs to be measured to make sure that the choice of outcome measure is relevant to patients
A
- Survival
- Patient experience
- Clinical measure (QoL)
-> If it is a proxy measure (indirect measure of outcome), are we confident that it is linked to the outcome of interest?
7
Q
What are the ethical issues that can arise in RCTs?
A
- Indivs should not be disadvantages by being randomised to intervention or control
- > We can only ethically randomise if we genuinely don’t know whether intervention or control is better - EQUIPOISE?
- > If we do not know which is better, is it ethical NOT to randomise?
8
Q
What are the diff options for control in RCTs?
A
- Usual care
- No treatment (only if there is no usual care)
- Placebo - enabled us to distinguish placebo effects/ helps blind pts and researchers/ clinicians
- Sham intervention (An inactive treatment or procedure that is intended to mimic as closely as possible a therapy in a clinical trial. aka placebo therapy)
9
Q
How can we randomise in RCTs?
A
- Allocation by chance of indivs or groups
- > Need to specify the method (block/ stratified/ minimisation)
- Allocation at random (indiv or cluster randomisation)
10
Q
Why do we randomise in RCTs?
A
To reduce the risk of bias and confounding
-> Randomisation should lead to known and unknown confounders being equally distributed (provided that the sample size is large enough)
11
Q
What is the definition of a ‘confounder’?
A
“A variable, other than the one studies, that can cause or prevent the outcome of interest - it influences both the dependent and independent variable)
- Confounding variable must be related to both the exposure/ intervention and independently with the outcome
12
Q
Requirements of confounding in RCTs?
A
In an RCT:
- A confounding variable has to be ass. with the outcome
- It must NOT be on the causal pathway
- Its distribution must differ between the groups
13
Q
What is allocation concealment?
A
The person recruiting a participant to a trial does not know which group the participants will be allocated to
14
Q
Why is allocation concealment important?
A
- Imbalance in characteristics of participants between the study arms (confounding)
- Selection bias
15
Q
What are the possible problems that might occur with intervention groups in RCTs?
A
- Non-adherence (important outcome)
- > Didn’t like it/ SE/ req effort and commitment (hard to do) - Comparator group get extra (performance bias)
- > They get the treatment anyway (contamination) / they seek some other complementary treatment/ HCPs provide some additional care - Intervention group gets extra (performance bias)
- > due to more regular contact with health service intervention group could have health problems detected sooner -> given additional advice
16
Q
Why do you need to blind the participants in an RCT?
A
- Avoid disappointment for ‘usual care’ groups and means they do not change their behaviour
17
Q
Why do you need to blind the HCPs providing their care?
A
- Avoids any compensatory activities for the usual care group
- Avoids any complementary treatments for intervention group
- Avoiding performance bias?
18
Q
Why do you need to blinds the researchers and statisticians?
A
- Avoids any bias in measuring or interpreting the outcomes
19
Q
What are the diff degrees of blinding?
A
- Unblinded/ open
- Single blinded - participant does not know what they get (placebo/ sham intervention)
- Double blinded - partici and researcher do not know whether treat or control
- Triple blinded - partici, researcher/ HCPs and statisticians do not know
20
Q
How do we deal with contamination/ crossover in RCTs?
A
ANALYSE BY INTENTION TO TREAT (ITT)
- Involves analysing according to the group they were originally in
- Reduces risk of allocation bias
- Gives you the effect of assignment to the interventions at baseline , regardless of whether the interventions are received as intended
- ITT may under-estimate the effect (those who shouldnt have taken smth, took it, therefore control group better)?
(if analyse according to treatment received, the sample is no longer randomised - allocation bias)
21
Q
What is per protocol anaylsis
A
For an intervention where adherence has been low -> sometimes it is useful to estimate the effect of adhering to the interventions as specified in the trial protocol (the ‘per-protocol effect’)
- a comparison of treatment groups that includes only those patients who completed the treatment originally allocated. If done alone, this analysis leads to bias
- interpretation of randomized clinical trial results that removes data from patients who didn’t comply with the protocol.
- Only the participants who adhered to their randomisation assignment are included in the analysis
- This should not be the main anaylss
- It is usually exploratory?
22
Q
What are the possible biases that could arise when assessing outcome ?
A
- Influenced by partici’s expectations - particularly is patient reported outcome like health related QoL
- Influenced by researcher’s expectations
- Influenced by statistical analyst’s expectations
(Reduced by blinding is possible - outcomes must be assessed the same way for both study arms otherwise you get detection bias)
23
Q
What is detection bias?
A
Systematic differences between groups in how outcomes are determined
24
Q
What might affect follow up rates?
A
- Participants lost to follow up nay be diff (older/ younger/ BAME group - attrition bias)
- Losses may be related to outcome - sicker people (attrition bias)
- Losses may relate to allocation - inter partici lost - didn’t like inter/ control group lower -disappointed that in control group (if unblinded)
- Losses may relate to treatment allocation and outcome (attrition bias)
25
How is relative risk calculated?
Incidence or risk of event in exposed/ incidence/ risk of event in control
26
How can outcomes in RCTs be interpreted?
1. Mean difference between groups (may be adjusted for important possible confounders)
2. Relative risk
3. Odds ratio
4. Survival analysis (time to event)
27
How do we look at precision of results?
- Look at the 95% CI
| - > Large sample = greater precision
28
How can we determine whether the results can be applied locally?
1. Likely to see a similar effect if the intervention was delivered locally?
2. What was the popn of the trial?
3. What is usual care ilke?
29
What biases could come up in RCTs?
1. Selection bias
2. Performance bias
3. Allocation
4. Attrition bias
5. Detection bias
6. Reporting bias?
- Placebo effect
30
How can we reduce selection bias in RCTs
1. Random sequence generation
2. Allocation concealment
3. Baseline characteristics comparable
4. Intention to treat analysis
31
How can we reduce performance bias in RCTs
1. Blinding of participants and personnel
32
How can we reduce measurement bias in RCTs
1. Blinding of outcome assessment
33
How can we identify attrition bias in RCTs
1. Look out for incomplete outcome data
34
What is a meta analysis?
A statistical summary of the results of more than one primary study
- Prods one summary outcome estimate
35
What is a forest plot?
A forest plot provides a representation of the amount of variation between the results of studies included in a meta analysis
36
How can we assess heterogeneity?
1. Eyeball test (look for overlap of chi2 and spread of effect estimates)
2. Statistical test (using chi-square (Cochran Q) or Tau2)
3. I2 test - ranging from 0-100% and measure the degree of inconsistency across studies in meta analysis
37
What is internal validity?
Do I believe the results?
- Consider METHODOLOYU/ BIAS and CONFOUNDING/ CHANCE
- the extent to which a piece of evidence supports a claim about cause and effect, within the context of a particular study.
- the extent to which the observed difference between groups can be correctly attributed to the intervention under investigation.
38
What is external validity?
Can I apply the results to my popn and setting (external validity and generalisability)?
- Consider APPLICABILITY
- Was there clear study q?
- Popn chara, prior testing of pts
- Cost and acceptability to pts and HCPs
- Definition of target condition match?
- CANNOT EXT VALIDITY WITHOUT GOOD INT VALIDITY as the findings would not even apply to the study popn let alone your own lolz
39
What are RCTs?
- An intervention study where parts are allocated into two or more groups, to which the groups will then receive diff interventions (one may be control)
- These groups are then followed up after randomisations and the outcomes are assessed
40
Why do we use randomisation?
- Ensures known and unknown chara. that might affect the outcome (confounders) are distributed by chance
- Any diffs between groups at the start are therefore due to chance
- Minimises confounding
41
What are the 3 broad types of data which all outcomes will fall under?
1. Nominal (categorial, i.e. dead or alive)
2. Ordinal (natural, ordered categories)
3. Interval (measured on a scale, can calculate a mean value)
42
What are the issues with RCTs?
1. Choice of outcome measure
2. Ethics
3. Choice of control
4. Contamination (crossover)
5. Bias in assessment of outcome
6. Losses to follow up
7. Sample size
43
What are the different choices of control groups we could use in an RCT?
1. Usual care - e.g. structured care vs standard care
2. No treatment - if no usual care
3. Placebo - distinguishes intervention effects from placebo effects, also helps to blind pts and researchers which minimises confounders
44
Why is analyse according to treatment received no bueno when there is contamination in an RCT?
- Groups no longer randomised
- ?Allocation bias due to self-selection or by Drs
- No longer technically an RCT as diffs between the two groups is no longer purely due to chance
45
How do we analyse categorical results?
1. Calculate the risk of outcomes in each group and then calculate the relative risk and absolute risks as well as the NNT
- But, this doesn't tell us whether to treat someone or not but it DOES provide us with a SCALE OF BENEFIT
2. Survival curves or curves showing the proportion of parts. in each group who have an event
46
How can we reduce contamination/ crossover in studies + DISADVANTAGE??
CLUSTER RANDOMISATION
- Randomise groups or clusters rather than indivs - e.g. schools, classes, hospitals or clinics or villages
- Stats analysis is more complex and it needs a BIGGER study
47
When might crossover/contamination be a problem/ occur?
1. If the intervention is information or education then it is hard to confine to one person as it gets passed on to other ppl
2. If we cannot treat pts differently in the same clinic
48
What can influence bias when assessing outcome?
1. Influenced by pt's expectations?
2. Influenced by researchers' expectations
3. Influenced by analysts/ expectations
-> THEREFORE BLIND ALL 3 NERDS
49
What is generalisability?
Describes the extent to which research findings can be applied to settings and patients other than that in which they were originally tested
50
What sort of issues can affect generalisability?
1. Age
2. Gender
3. severity of disease
4. Presence of other co-morbidities
5. The setting (e.g. pts from primary or secondary care)
-> If the parts are not REPRESENTATIVE of the profile of pts with a condition, then the trial results are unlikely to be broadly generalisable
51
How to determine whether the results of the trial is valid?
1. What TYPE of study is this?
2. What is the RESEARCH QUESTION?
3. Bias: may affect study
- Definitions/ naming - DESCRIBE and APPLY
- Application to study
- How have they tried to minimise
3. Confounding
4. Method
52
What needs to be considered when deciding whether the findings of an RCT are valid?
1. Eligibility criteria
2. Randomisation method
3. Allocation concealment
4. Blinding
5. Follow up
