Randomised controlled trials

Question 1

Some professionals claim they can say if a drug/ treatment works best purely from clinical experience. What are the 5 key limitations of this?

Answer 1

• Patients may get better anyway (even without treatment)
• Placebo effect -> they expect it to work = get better
• No controls = dont know what would happen to patients without treatment
• Follow up is haphazard = difficult to tell if got better/worse
• Small numbers of patients for a single health care worker = any observation may be due to chance

Question 2

Where in the heirachy of evidence are randomised controlled trials?

Answer 2

Top = best method availiable

Question 3

What is a randomised controlled trial?

Answer 3

Randomly allocate patient to treatment or control group (patient has equal probability of ending up in either group)

If study is large enough then the groups are comparable with respect to baseline characteristics

Question 4

Which 7 things happen in the planning stage of an RCT?

Answer 4

1. Establish hypothesis/research qu
2. Define eligibility criteria
3. Define treatment/intervention
4. Define control/comparison (placebo)
5. Define baseline characteristic of interest (confounding factors)
6. Calculate sample size/power ratio
7. Specify outcome measures

Question 5

Why can eligibilty criteria not be too restrictive?

Answer 5

Results only generalisable to limited range of patients not to the population as a whole (we want)

Question 6

Why is it important to have eligibility criteria?

Answer 6

It may be too risky for an individual to have a new treatment/ deny them conventional treatmet

Unable to follow study requirements (e.g. mental illness or live far from study site)

Question 7

Can you have >1 control group?

Answer 7

Yes! e.g. conventional treatment and placebo

You can have >1 treatment group too!!

Question 8

Why must you define baseline characteristics?

Answer 8

To show that random allocation between groups was successful

Question 9

How do you work out the sample size?

Answer 9

Power clculation

= Nomogram (read size off of chart)/computer package

Question 10

Why must the sample size be big enough?

Answer 10

Must be large enough to detect the observed difference between groups

Question 11

What do you need to specify about outcome measures?

Answer 11

Which outcomes are being measured (e.g. patient satisfaction, side effects and primary & secondary outcomes)

How they are being measured

When they are being measured (time points)

Question 12

When should the power calculation for the trial be undertaken?

Answer 12

Before the trial begins so enough subjects are included to be able to answer the question but no more than neccessary are included (avoids wasting time and reasources)

Question 13

Which 5 things must be done when conducting a RCT?

Answer 13

1. obtain ethical approval
2. seek informed consent
3. generate random allocation sequence
4. ensure concealment of randomisation
5. specify subject and outcome assessor blinding

Question 14

Why can provide ethical approval?

Answer 14

Local or regional research ethics committees (include healthcare professionals/researchers/lay members)

Question 15

What can an ethics committee do?

Answer 15

Refuse permission

Request ammendments (e.g. supply more information to possible subjects about possible side effects)

Question 16

What is clinical equipoise?

Answer 16

Uncertainty that treatment is better than control

= if the control treatment is definitely better then it is unethical to give the patients a 50% chance of being allocated to the control group

Question 17

What is informed consent?

Answer 17

Risk and benefits disclosed to patients

Competant/voluntary consent

Still entitled to withdraw at any time which will in no way compromise future treatment

Question 18

How is concealment of randomisation achieved?

Answer 18

Those allocating participants are not involved in any way with the trial (do not know the individuals) = dont knowingly adjust the allocation based on prognostic factors

Question 19

What are the 3 things that should be done when analysing a RCT?

Answer 19

1. Check randomisartion was successful (compare baseline characteristics between treatment groups)
2. Check loss to follow up (those who dont complete the trial -> need to ensure similar in both groups -> may be different if side effects are worse in elderly so just the elderly drop out = may affect results)
3. Main analysis (intention to treat)

Question 20

What is intention to treat analysis?

Answer 20

Subjects are analysed in the groups they were randomly allocated to and not on basis of whether treatment was completed = only compared at point of randomisation (stops bias = no confounding in each arm)

Question 21

What is per protocol analysis?

Answer 21

Only includes the patients who did adhere to the treatment

Question 22

Why should an intention to treat analysis always be undertaken?

Answer 22

Treatment groups are only comporable if they include all subjects allocated at baseline

Has nothing to do with right to be treated and only includes the subjects who were randomised

Question 23

What is number needed to treat to benefit (NNTB)?

Answer 23

The number of subjects that will need to be treated to prevent 1 case of the outcome -> depends on how common the disease

= 1 / absolute difference in risk of outcome between groups

Decides if it is worth while to recommend treatment or not (includes cost)

Question 24

What is number needed to treat to harm (NNTH)?

Answer 24

= 1 / absolute difference in risk of adverse event between groups

Question 25

Which type of bias is minimised in a randomised controlled trial?

Answer 25

Selection, performance or detection/measurement

Question 26

What are the strengths of RCTs (5)?

Answer 26

• No confounding (if large enough and truly random as any confounding evenly distributed between groups)
• Intervention preceeds outcome (no reverse causality)
• No selection bias (random allocation)
• No performance bias (blinding)
• No detection bias (blinding)

Question 27

What are the weaknesses of RCTs (5)?

Answer 27

• Large sample size needed to detect small differences (cost)
• May need long periods of follow up to detect late outcomes (v. expensive & more loss to follow up)
• Efficacy of treatment under trial conditions may be different from normal practice (i.e. those in trial more likely to adhere to treatment)
• Loss to follow up bias
• Potential performance bias

Question 28

Why are RCTs not always possible?

Answer 28

• Unethical
• Unfeasible allocation of some kinds of exposure
• Unfeasible evaluation of very long term effects of exposure (too expensive to keep trial going too long)
• Very rare outcomes = need too many recruits

In certain situations we can use other studies to develop a hypothesis which is then tested by RCT (as only RCTs can eliminate systematic differences between treatment groups regardless of whether or not confounding variables have been measured = controlled in other study designs)