7-8: Experimental studies (RCT) Flashcards
RCT
Motivation
best study designs
Motivation
Does an intervention really help, is there a causaility between intervention an outcome?
Best study design in reality
Randomization regarding control and intervention group
Definition
- epidemiological experiment in which subjects in an population are randomly allocated into groups., usually called study and control group, to receive or not to receive an exp. preventive or therapeutic procedure or intervention
- results assessed by rigorous comparison of outcomes in both groups
- RCT most scientifically rigorous method of hypothesis testing
Structure of RCT
Control group
- confirm causality because they allow to distinguish between outcomes induced by intervention or natural course of disease or other factors
- usually standard care, if there is no standard care, then placebo
Key points for RCTs
Structural equality
- comparibiliry of groups in regard to baselinew characteristics (randomization)
Operational and observational equality
avoidance of different behavior and expectations of physicians / study personal and participants (blinding)
Randomization
Why?
Why?
- prevent systematic differences between study groups = structural equality
- by increasing sample size, random differences become smaller
3 methods of randomization
- name
- aim
- how
- limitations
- …
Simple unrestricted randomization
Aim:
- securing unbiased comparison groups
How:
- Allocation by rolling the dice, computer aided, …
Caution:
- methods like date of presentstion, birth date are not random, dont use such methods
Limitations:
- unequal sample size relevant in small sample sizes
Advantages
- complete unpredictability
Blocked randomization
Aim:
- securing unbiased comparison groups with sample groups size
How:
- form blocks and randomly order the interventions / control group with that block
Limitations:
- predictability possible
Recommendation
- use different block sizes to make allocation less predictable
Stratified randomization
Aim:
- avert imbalances by use of stratification on important factors, such as age or disease severity
How:
- e.g. Ration of overweight / normal weight is 1:10, stratify by weight category and apply e.g. blocked randomization
Limitations:
- high sample size should already be balanced, complex, imbalances can be adjusted statistically
Blinding methods
Open
patient and physicians knows about treatment
Single blinded
only the physician knows about group
Double blinded
physicians and participant
Triple blinded
Double blinded + the people who analyse the data do not know about group
Why blinding?
- avoid influence of expectations, privileges for single patients, chance of manipulation in treatment or outcome assessment
- achieve operational equality (involved in treatment) and observational equalty (involved in observing the outcome)
Strategy
- identical format of control and intervention
- similar presentations
- person, who assesses the outcome is independent
Protocol review by ethics committee
uncertainty principle
it is unclear which intervention is better when patient enters one of the study groups
ethical prerequisites
informed consent, voluntary, withdrawal possible, no consequences for standard care if no participation or withdrawal, contact person, criteria for end of clinical trial
Benefit risk ratio
Different approaches for statistical analysis
Intention to treat
- includes all study participants that have been randomized to maintain streuctural equality / same sample size, but effect will be understimated
- prefered
Per protocol
- includes only those that have been comliang and were data on outcomes are available
- can lose structural equality and create biased estimate of treatment effect possible
Outcomes or endpoints of RCTs
Clinically relevant endpoint
melanoma diagnosis and death
Proxy or surrogate parameters
referral to dermatologist because of suspected melanoma
Intermediary outcome
sund burns, happens on the way to clinically relevant outcome
Possible measures of risk
Absolute Risk (Incidence)
e.g. AR for Control: C / (C+D)
The risk to have … when receiving the placebo is …%.
Relative Risk
e.g. AR(I) / AR(C)
When using the intervention, the risk to have … is only …. of the risk when using the placebo
Absolute risk reduction
ARR = AR(I) - AR(C)
ARI = (AR(I) + AR(C) (Absolute risk increase)
ARR: The risk of having …. is ….% lower/higher when receving intervention instead of standard care / placebo
Relative Risk reduction
RRR = ( AR(I) - AR(C) ) / AR(C) = RR - 1
RRI = ( AR(I) + AR(C) ) / AR(C) = RR + 1
RRR: The risk to have … is by …% higher/lower when receing intervention and not the placebo / standard care
Numbers needed tp treat
NNT = 1 / |ARR|
On average, X person with diseae must be treated with intervention to eliminate disease in one person more than with treatment by placebo / standard care
Pros and Cons of RCTs
Cons
- complex and costly
- relationship between patient and physicians affceted?
- ethical questions
- study participants need to represent population of interest
Pros (theroretical)
- Intervention and Control are as similar as possible in their basic structure, even unknwon confounders should be distributed equally -> structural equality
- no chance to influence exposure status, nobody can select to be in control or intervention group (no selection bias)
- strictes methods to confirm causality
Placebo and nocebo effect
Placebo
Effect of intervention is reported, but person only had placebo
Nocebo effect
adverse effcet of intervention is reported, but person only had placebo
