RCTs

Question 1

What is the purpose of randomisation in RCT? What does it help prevent?

Answer 1

Ensure that baseline characteristics are similar between the intervention and control groups so that any difference between them is due to chance

-minimises risk of confounding due to distributing potential confounding characteristics between the groups
=randomised KNOWN AND UNKNOWN confounders which observational studies cannot do

Possible confounding factors:

• age
• cormorbidity
• severity of illness
• genetics
• sex
• ethnicity

Question 2

What are the possible outcomes which can be measured? What criteria must outcomes in RCT meet?

Answer 2

Survival
Patient experience or QOL
Clinical measure (i.e. blood tests etc)

Criteria:

• reliable
• valid= measures what is aimed to be measured
• responsive= detects changes over time

Question 3

What is clinical equipoise?

Answer 3

Individuals in RCT must not be disadvantaged by the intervention or control
-don’t know whether the intervention or control is better

Question 4

What can be chosen as the control in RCTs?

Answer 4

Usual care

Placebos (can help with reduction of bias through blinding patients so don’t know which type have)

Sham intervention

No treatment

Question 5

Why is a large sample size important in RCTs?

Answer 5

Small sample size can lead to difference in distribution of confounders i.e. affected of differences due to chance are amplified in smaller sample size

Smaller effect needs to have larger sample size

Imprecisely measured outcomes need larger sample size

Can influence the CI i.e. precision of results
-larger population= narrower CI -> means there is increased precision as certain that results will fall into narrower range
I.e. patients less likely to differ from each other in larger population group compared with smaller group

Question 6

What does the process of allocation concealment involve?

Why is it done?

Answer 6

Process:
-Recruiters or researchers should not be involved in or know the allocation of participants to the different groups
I.e. allocation done remotely and independently of clinicians

Purpose:
-minimise allocation bias
I.e. prevents clinicians from allocating patients to groups based on whether they think they might benefit more from intervention or whether they might be able to adhere to the intervention better
-minimise sampling bias
-minimised possible confounding effects by preventing imbalance of patient characteristics between the study arms

Question 7

What is the criteria for a confounded in RCT?

Answer 7

Associated with outcome
Not on the causal pathway
Different distribution between the groups

Question 8

What are the possible consequences if the intervention and control group are not treated equally outside the intervention?

Answer 8

Performance bias can occur i.e. groups receive additional treatment/intervention/support outside the intended intervention

Control group

• contamination of control group occurs where they get treatment anyway or get alternative treatment
• seek complementary tx
• HCP provide additional care or advice (can be due to practitioners not wanting patients with condition to be disadvantaged)
• undoes the randomisation process

Intervention group
-regular contact with health services can lead to receiving additional advice or health problems being detected sooner

Question 9

When is intention to treat analysis (ITT) used? What are the adv and disadv of ITT?

Answer 9

When performance bias has occurred meaning the sample randomisation has been lost

Results analysed based on assignment to intervention at baseline regardless of whether it was received to preserve the effects of randomisation

Adv:

• preserves random allocation effects i.e. minimising confounding influence I.e. equal distribution of confounders
• mirrors real-life situation where patients might not adhere to treatment or might seek additional/supplementary treatment

Disadv:
-can lead to under-estimation of effectiveness of treatment due to not all those allocated to treatment complying with treatment

Question 10

What are the different forms of blinding?

Answer 10

Open label
- everyone aware of assignment to intervention or control

Single blinded
-participants blinded= placebo or sham intervention

Double blinded
-participant and researcher don’t know whether intervention or control

Triple blinded
-same as double but statistician interpreting the results also doesn’t know the allocation to groups

Might not use these terms in paper so need to look for description of blinding process

Question 11

Give reasons why it is important blind the different people involved in the study from whether they are allocated intervention or control?

Answer 11

Participants
-prevents disappointment at not receiving intervention which could influence behaviour I.e. seek alternative treatments that could lead to performance bias

Healthcare professionals

• avoids compensatory activities for usual care group
• avoid complementary treatment for intervention group

Researchers/statisticians
-avoids influence when interpreting data i.e. detection bias (want to ensure that outcomes from groups assessed in the same way)

Question 12

What is per-protocol analysis?

Answer 12

Used to estimate the affect of adhering to intervention when there is low adherence to intervention

Question 13

What patient factors can affect the outcomes of RCT?

Answer 13

Adherence to intervention

Loss to follow up or withdrawal

Question 14

Why might participants be lost to follow up in RCT? What is important to look for when there has been loss to follow up?

Answer 14

Patients didn’t like intervention i.e. SE or hard to carry out or comply with

People in control group disappointed about not receiving intervention (if not blinded to allocation process)

Check if there is difference between people who have withdrawn

• intervention vs control
• baseline characteristics i.e. increased severity of disease or age or cormorbity meaning unable to adhere to treatment process

Question 15

How would you interpret these results from an RCT?
Intervention= statins
Outcome= whether there is change to development of ARM

Crude odds ratio of 0.45 (95% CI 0.32-0.64)

Answer 15

People who took statins were 55% less likely to develop ARM compared with people who did not take statins.
There is a 95% certainty that people taking statins would be between a 36% to 68% less likely to develop ARM, meaning that the maximum reduction in risk would be 68% and the minimum would be 36% reduction.
The confidence interval does not contain 1 (the null value) meaning that the results are statistically significant

Question 16

What criteria do you need to look for to see in results from RCT are valid?

Answer 16

Eligibility criteria 
Randomisation method 
Allocation concealment 
Blinding 
Follow up

Question 17

What criteria does there need to be for RCT results to be applied to own patient population or general population?

Answer 17

Factors effecting generalisability (extent to which research findings can be applied to setting other than original research setting):

• patient demographic= age and sex
• severity of disease
• presence of co-morbidities
• setting
• whether bias and confounders have been fully explored

(Don’t need to assess for study type or causality in RCT)

Question 18

What are the 6 main problems to be aware of in RCTs?

Answer 18

Choice of outcome measure

• clinical effectiveness
• patient experience
• relevance to patient

Choice of control

• usual care
• no treatment
• placebo

Ethical consideration
-clinical equipoise i.e. are individuals disadvantaged by being randomised to intervention or control

Sample size

• smaller effect needs larger sample size
• significance
• power
• no systematic differences at baseline

Bias in assessing outcome

• stopping bias
• degrees of blinding
• allocation concealment
• loss to follow up

Contamination or cross-over

Question 19

What is stopping bias?

Answer 19

Trial stopped early when the desired results shown
I.e. drug might have large desired effect early on leading to trial being stopped but the effect might actually taper off or decrease over time which would not be detected if trial stopped too early

Question 20

What are the benefits of using routine/usual care as the control compared to no treatment?

Answer 20

Clinicians/GP practices might be more inclined to participate in study

Higher chance of following trial protocol if the control is routine practice

Can minimise risk of contamination/diversion (seeking the intervention or other form of treatment) compared with when control is no treatment

Question 21

What is contamination/deviation in the context of RCT? What are the advantages and disadvantages of this occurring?

Answer 21

When those allocated to control group receive the intervention anyway

Adv:
-can be representative of real life scenario which is important as intervention will need to be effective in real life scenario

Disadv:

• leads to dilution of outcomes i.e. difference between study arms is smaller than if contamination did not occur
• will have to switch to ITT analysis

Question 22

How does analysis according to treatment received differ from an intention to treat analysis? What is the importance of this?

Answer 22

Analysis according to treatment received occurs when analysis is limited to those participants who were treated according to the protocol of the branch of the trial they were originally allocated

Leads to removing of the effect of randomisation process achieved through the randomisation process
=leads to possible introduction of confounding in the interpretation of results

Question 23

Interpret the following results in terms of differing antibiotic use between the intervention and the control:
RR 0.71 (0.5-0.95) P=0.02

Answer 23

The risk ratio is 0.71 in the intervention compared to the control. This means that people in the intervention were 29% less likely to use antibiotics than those in the control group. The CI indicates that the antibiotic use could be reduce by as much as 50% or as little as 5%. The results are statistically significant because P<0.05.

Question 24

Why might RCT not always be the best study design for cause and effect?

Answer 24

When exposure not under the control of researcher due to ethical consideration that exposure is causing harm or disease

Rare outcome means that very large number of people are required to ensure that there is enough outcome to be analysed

Natural experiment

Question 25

What is the difference between allocation concealment and blinding?

Answer 25

Allocation concealment= ensures treatment to be allocated is not known before the patient enters the study i.e. person enrolling the participants will not know in advance which treatment the next person will get

Blinding= ensure patient/physician is blinded to treatment allocation after enrolment into the trial