Evaluating Treatments

Question 1

What is Treatment?

Answer 1

• In most cases - new drugs / medicines
• However, it can be much broader than drugs:
• New surgical procedure, new medical equipment
• New approach to counselling after psych. trauma
• New approach to motivate care seeking behaviour –
• Broader : vaccination effectiveness
• Screening prorammes, risk factor withdrawal
• Different organization of health care system

Question 2

What does evaluating treatments require?

Answer 2

• Requires the highest standard of research
• Requires evidence beyond reasonable doubt
• Requires sense of responsibility and high moral standards from:
• Researchers who evaluate treatment
• Editors of journals that publish this research
• Members of FDA, WHO or other advising bodies
• Governments and their committees on health care
• Corruption is a real concern at each stage

Question 3

How to randomise in RCT from best to worst

Answer 3

• Computer randomisation
• Telephone randomisation
• Identically packaged treatments
• Sealed envelopes
• Pre-randomised lists
• Coin tossing/ Card drawing
• Haphazard

Question 4

What are the two types of error needing to be handled in RCTs?

Answer 4

Random Variability- eliminated with size

Systematic bias

Question 5

What could invalidate the results of a trial?

Answer 5

=ALWAYS remain sceptical – opportunities to cheat - huge!
-=Were all the “survivors” within the group receiving chemotherapy concentrated in the same hospital or country?
=Did the analysts/statistician know which patient was allocated to which group, as he/she was receiving information on 5-year survival of individual patients? (he/she should have been BLIND to this information!)
=Who was sending him/her this information? Were they BLINDED to treatment received? (they should have been!)
=Was it possible to ensure blinding of researchers, patients?

Question 6

Why a large sample size?

Answer 6

ensures sufficient “statistical power” to detect and properly recognize true differences, even when they are relatively small between groups (e.g. true difference in survival <10% - still detectable with very narrow C.I.-s)

Question 7

Why randomization?

Answer 7

if properly conducted in a sufficiently large sample, it will ensure that there are no systematic differences between study groups, and that the difference in disease outcome is mainly attributable to treatment

Question 8

Why blinding?

Answer 8

protects the results of randomized controlled trials from all conscious and unconscious biases - BOTH can be very important

Question 9

What is the checklist for quality of randomised controlled trials?

Answer 9

• Ethics?
• Sufficient statistical power to detect clinically important difference (with estimated rate of patient drop-out accounted for)?
• Random Allocation?
• Checked comparability of groups?
• Co-interventions in different groups?
• Compliance of patients?
• Patients, researchers and analysts fully blinded?