Lecture 1 - Research questions
● Classification of medical research
○ Etiology (risk factor)
○ Diagnosis
○ Treatment
○ Prognosis
● Research question components → PICO
○ Patient (population)
○ Intervention
○ Comparator
○ Outcome
Lecture 2 - Randomized controlled trial
Outcomes
● Regression to the mean
○ If the value is at its highest point it can only go down
● Outcome model
○ Treatment (T)
○ Natural course (NC)
■ Regression to the mean
○ Extraneous factors (EF)
■ Other treatments
■ Going to the gym
■ Stop smoking
○ Error Processes (V)
■ Natural variation in the medical device used
● Possible outcomes
○ Outcome with treatment
■ T + NC + EF + V
○ Outcome without treatment
■ NC + EF + V
● Comparison
○ Compare 2 (or more) groups
○ Groups should be compara ble with respect to NC, EF, V
○ Differ only with respect to treatment
Design elements
● Randomisation
○ Randomly allocation of treatment
○ Concealment of treatment allocation
■ Physician doesn’t know which treatment he’s describing
■ Result → treatment allocation independent of patient characteristics
● Blinding
, ○ Participants don’t know which treatment they receive
■ Aims to keep the groups comparable during follow-up
■ This also applies to physicians, nurses, relatives, etc.
○ Placebo
■ Tablet which tastes/looks/smells like the active treatment, but does not
contain the active compound
■ Sometimes difficult → surgery, physiotherapy
○ Active comparator
■ Compare different types of drugs
○ Blinded outcome assessment
■ The one who ‘measures’ the outcome should not know about
treatment status
● Standardisation
○ Standardisation of intervention, concomitant care & outcome assessment
■ Minimize error processes
■ Improve interpretability of treatment effect
Comparability
● Start of treatment
○ Randomisation
○ Concealment of allocation
● Follow-up
○ Blinding of patient & physicians
● Outcome assessment
○ Blinding of outcome assessor
Equipoise → The genuine uncertainty about which treatment is better
● Ethics
○ Is it ethical to give a placebo when a treatment which is known to be effective
is also available?
○ Is it ethical to give a new treatment which is known to be uneffective?
● Relevant comparison
○ What is the clinical decision:
■ Researched drug vs. no treatment
■ Researched drug vs. other drug
Primary analysis
● Intention-to-treat
○ Purpose → include all participants in the groups to which they were originally
assigned, regardless of whether they completed the trial
● Per-protocol
○ Purpose → only analyze the participants who completed the study as per
protocol
Lecture 3 - Sample size calculations
, Why?
● Aim → comparing two treatments
○ Patients are recruited to the study, and randomized to treatment A or B
● How many patients needed?
○ Too few → not able to detect differences
○ Too many → high costs & non-ethical
Deciding sample size
● Practical
○ Number of eligible patients treated at a center
○ Number of patients willing to participate
○ Time & money
● Statistical
○ How big of an effect can be detected with a given number of patients?
Hypothesis testing
1. Decide on a null hypothesis (H0) about the population
○ Example → there is no difference between the two groups
2. Take a representative sample of the population
3. Calculate the observed difference in the sample
4. Calculate the p-value: the probability to observe at least this difference if H0 is true
5. If p-value is smaller than a prespecified value α we reject H0
○ Value α is called the significance level
Type I & II errors
● Type I → false positive
○ Rejecting a true null hypothesis
○ Controlled by the significance level (α)
○ Example: concluding a new drug works when it actually doesn’t
● Type II → false negative
○ Failing to reject a false null hypothesis
○ Controlled by ß
○ Concluding a drug doesn’t work when it actually does
● Statistical power
○ 1 - probability of a type II error = 1 - ß
Power → The probability of finding a significant effect in your sample when the effect is
really present in the population
● Depends on:
○ Relevant difference (effect size)
○ Sample size
○ Variance / standard deviation (more variation = smaller power)
○ Significance level α
● Goal
● Classification of medical research
○ Etiology (risk factor)
○ Diagnosis
○ Treatment
○ Prognosis
● Research question components → PICO
○ Patient (population)
○ Intervention
○ Comparator
○ Outcome
Lecture 2 - Randomized controlled trial
Outcomes
● Regression to the mean
○ If the value is at its highest point it can only go down
● Outcome model
○ Treatment (T)
○ Natural course (NC)
■ Regression to the mean
○ Extraneous factors (EF)
■ Other treatments
■ Going to the gym
■ Stop smoking
○ Error Processes (V)
■ Natural variation in the medical device used
● Possible outcomes
○ Outcome with treatment
■ T + NC + EF + V
○ Outcome without treatment
■ NC + EF + V
● Comparison
○ Compare 2 (or more) groups
○ Groups should be compara ble with respect to NC, EF, V
○ Differ only with respect to treatment
Design elements
● Randomisation
○ Randomly allocation of treatment
○ Concealment of treatment allocation
■ Physician doesn’t know which treatment he’s describing
■ Result → treatment allocation independent of patient characteristics
● Blinding
, ○ Participants don’t know which treatment they receive
■ Aims to keep the groups comparable during follow-up
■ This also applies to physicians, nurses, relatives, etc.
○ Placebo
■ Tablet which tastes/looks/smells like the active treatment, but does not
contain the active compound
■ Sometimes difficult → surgery, physiotherapy
○ Active comparator
■ Compare different types of drugs
○ Blinded outcome assessment
■ The one who ‘measures’ the outcome should not know about
treatment status
● Standardisation
○ Standardisation of intervention, concomitant care & outcome assessment
■ Minimize error processes
■ Improve interpretability of treatment effect
Comparability
● Start of treatment
○ Randomisation
○ Concealment of allocation
● Follow-up
○ Blinding of patient & physicians
● Outcome assessment
○ Blinding of outcome assessor
Equipoise → The genuine uncertainty about which treatment is better
● Ethics
○ Is it ethical to give a placebo when a treatment which is known to be effective
is also available?
○ Is it ethical to give a new treatment which is known to be uneffective?
● Relevant comparison
○ What is the clinical decision:
■ Researched drug vs. no treatment
■ Researched drug vs. other drug
Primary analysis
● Intention-to-treat
○ Purpose → include all participants in the groups to which they were originally
assigned, regardless of whether they completed the trial
● Per-protocol
○ Purpose → only analyze the participants who completed the study as per
protocol
Lecture 3 - Sample size calculations
, Why?
● Aim → comparing two treatments
○ Patients are recruited to the study, and randomized to treatment A or B
● How many patients needed?
○ Too few → not able to detect differences
○ Too many → high costs & non-ethical
Deciding sample size
● Practical
○ Number of eligible patients treated at a center
○ Number of patients willing to participate
○ Time & money
● Statistical
○ How big of an effect can be detected with a given number of patients?
Hypothesis testing
1. Decide on a null hypothesis (H0) about the population
○ Example → there is no difference between the two groups
2. Take a representative sample of the population
3. Calculate the observed difference in the sample
4. Calculate the p-value: the probability to observe at least this difference if H0 is true
5. If p-value is smaller than a prespecified value α we reject H0
○ Value α is called the significance level
Type I & II errors
● Type I → false positive
○ Rejecting a true null hypothesis
○ Controlled by the significance level (α)
○ Example: concluding a new drug works when it actually doesn’t
● Type II → false negative
○ Failing to reject a false null hypothesis
○ Controlled by ß
○ Concluding a drug doesn’t work when it actually does
● Statistical power
○ 1 - probability of a type II error = 1 - ß
Power → The probability of finding a significant effect in your sample when the effect is
really present in the population
● Depends on:
○ Relevant difference (effect size)
○ Sample size
○ Variance / standard deviation (more variation = smaller power)
○ Significance level α
● Goal
