Samenvatting AWV
HC Research question
- Medical research
o Ultimate goal of medical research: to improve medical practice
o Classification of medical research
▪ Prevention / risk factor (etiology)
▪ Diagnosis
▪ Treatment
▪ Prognosis
o Examples:
▪ Etiology: is a high caloric diet a risk factor for cardiovascular disease?
▪ Diagnosis: what is the probability of having a hip fracture (= diagnosis) if the
affected leg is shorter and in exorotation?
▪ Treatment: does chloroquine treatment reduce risk of mortality among
COVID-19 patients admitted to the ICU?
▪ Prognosis: what is the probability of dying within 5 years (= prognosis) after
breast cancer diagnosis?
- Research questions
o Research questions:
▪ Arise in practice
▪ Starting point when designing a study (or when reading a paper)
▪ Answerable
▪ Standard elements
o Components of a research question
▪ PICO
• Patient / (population)
• Intervention
• Comparator
• Outcome
▪ DDO
• Domain
• Determinant
• Outcome
o Example
▪ Research question: to what extent does alemtuzumab compared to
interferon beta improve the time until relapse occurs in patients with
relapsing remitting MS?
• Classification: treatment
• Patients: patients with relapsing remitting MS
• Intervention: alemtuzumab
• Comparator: interferon beta
• Outcome: time until relapse occurs
▪ Research question: what is the predictive value of a second ultrasound, in
case a first was inconclusive, in patients suspected of having an acute
appendicitis?
• Classification: diagnosis
, • Patients: patients suspected of having an acute appendicitis, in
whom a first ultrasound was inconclusive
• Intervention: second ultrasound
• Comparator: -
• Outcome: appendicitis
▪ Research question: is smoking a risk factor for lung cancer?
• Classification: etiology
• Patients: humans
• Intervention: smoking
• Comparator: no smoking
• Outcome: lung cancer
HC Randomized Controlled Trial (RCT)
- Outcome model
o Outcome = treatment (T) + natural course (NC) + extraneous factors (EF) + error
processes (V)
o The outcome is a combination of different phenomena → it could be due to the
treatment effect, but it could also be due to the natural course of the disease,
extraneous factors (i.e. starting/quitting smoking, going to the gym etc.), error
processes (= natural variation, i.e. due to the measurement device you use)
o Obviously, our interest lies in the treatment
o Possible outcomes
▪ Outcome with treatment = T + NC + EF + V
▪ Outcome without treatment = NC + EF + V
o Comparison
▪ To identify treatment effect:
• Compare 2 (or more) groups
• Those groups should be comparable with respect to NC, EF, and V
• And differ only with respect to treatment!
▪ In that case, an observed difference in the outcome between the groups, can
be attributed to the only aspects that the two groups differ on: the
treatment
▪ → we need comparability (with respect to NC, EF and V)
- Design elements of a randomized controlled trial
o 1. Randomisation
▪ Randomisation = the treatment of interest is randomly allocated to
participants in an RCT → so it is a chance process whether a participant
receives the treatment or an alternative treatment
▪ Randomisation can happen in a lot of different ways
▪ Concealment of treatment allocation
• Physician who asked patients to participate does not know what
treatment the next patient will receive, nor does the patient
him/herself
• As a result, treatment allocation is independent of patients
characteristics
o 2. Blinding
, ▪ a. Participants should not know which treatment they receive, because that
could influence their ‘behaviour’
• This also applies to treating physicians / nurses / relatives / etc.
o Blinding aims to keep the groups comparable during follow-
up
▪ b. Placebo vs. active comparator
• 1. Placebo
o It tastes / looks / smells like the active treatment but does
not contain the active compound
o Sometimes difficult → e.g. surgery, physiotherapy
• 2. Active comparator
o E.g. a new diuretic treatment vs. hydrochlorothiazide (= a
different, already existing diuretic treatment)
▪ c. Blinded outcome assessment
• The one who ‘measures’ the outcome should not know about the
treatment status, because that could influence the measurement
o 3. Standardisation
▪ Standardization of intervention
▪ Standardization of concomitant care
▪ Standardization outcome assessment
▪ Reason for including randomization:
• Minimize error processes
• Improve the interpretability of treatment effect
- Comparability
o Comparability is key when making a comparison between two groups
o Comparability is relevant at different stages of an RCT:
- Summary
o Estimation of ‘effect’ of treatment requires comparability
o In a randomised trial, several design elements aim at improving comparability:
▪ Concealment of allocation
▪ Randomisation
▪ Blinding
▪ Standardisation
HC Sample size calculations
- Randomized controlled trial (RCT)
o Aim: compare two treatments
o Patients are recruited to the study, and randomized to treatment A or B
o How many patients are needed (= sample size)?
, ▪ Too few: not able to detect differences between the 2 groups
▪ Too many: costs and not ethical
o Factors for deciding sample size
▪ Practical
• Number of eligible patients treated at centre
• 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 (quick overview)
o 1. Decide on a null hypothesis H0 about the population
▪ H0: there is no difference between the two groups
o 2. Take a representative sample of the population
o 3. Calculate the observed difference in the sample
o 4. Calculate the p-value, the probability to observe at least this difference if H0 is
true. This is done by a statistical test.
o 5. If p-value is small, smaller than a prespecified value α, we reject H0
▪ The value α is called the significance level
o Type 1 and type 2 errors
▪ Type 1 error (α): rejection of the H0, while it
is true → so rejection of the H0, thus saying that
the two populations differ, while there in fact is no
significant different between the two groups
▪ Type 2 error (β): acceptance of the H0, while
it should have been rejected → so no rejection of
the H0, thus saying that there is a difference
between the two populations, while in fact there is
a significant difference between the two groups
▪ Power = 1 – the probability of a type II error
- Power
o Power = the probability of finding a significant effect in your sample when the effect
is really present in the population
o Depends on:
▪ Relevant difference (effect size)
▪ Sample size → power increases, as sample size increases
▪ Variance / standard deviation → more variation, leads to smaller power
▪ Significance level α
o We want to have a study with a large power!!
- Example study
o RCT on patients with high blood pressure → comparison between intervention (new)
and active comparator (old); outcome: blood pressure after 6 weeks of treatment
o Trial with two groups of 30 patients, H0 is true!
▪ Distribution of mean difference (between the 2 groups) in blood pressure
HC Research question
- Medical research
o Ultimate goal of medical research: to improve medical practice
o Classification of medical research
▪ Prevention / risk factor (etiology)
▪ Diagnosis
▪ Treatment
▪ Prognosis
o Examples:
▪ Etiology: is a high caloric diet a risk factor for cardiovascular disease?
▪ Diagnosis: what is the probability of having a hip fracture (= diagnosis) if the
affected leg is shorter and in exorotation?
▪ Treatment: does chloroquine treatment reduce risk of mortality among
COVID-19 patients admitted to the ICU?
▪ Prognosis: what is the probability of dying within 5 years (= prognosis) after
breast cancer diagnosis?
- Research questions
o Research questions:
▪ Arise in practice
▪ Starting point when designing a study (or when reading a paper)
▪ Answerable
▪ Standard elements
o Components of a research question
▪ PICO
• Patient / (population)
• Intervention
• Comparator
• Outcome
▪ DDO
• Domain
• Determinant
• Outcome
o Example
▪ Research question: to what extent does alemtuzumab compared to
interferon beta improve the time until relapse occurs in patients with
relapsing remitting MS?
• Classification: treatment
• Patients: patients with relapsing remitting MS
• Intervention: alemtuzumab
• Comparator: interferon beta
• Outcome: time until relapse occurs
▪ Research question: what is the predictive value of a second ultrasound, in
case a first was inconclusive, in patients suspected of having an acute
appendicitis?
• Classification: diagnosis
, • Patients: patients suspected of having an acute appendicitis, in
whom a first ultrasound was inconclusive
• Intervention: second ultrasound
• Comparator: -
• Outcome: appendicitis
▪ Research question: is smoking a risk factor for lung cancer?
• Classification: etiology
• Patients: humans
• Intervention: smoking
• Comparator: no smoking
• Outcome: lung cancer
HC Randomized Controlled Trial (RCT)
- Outcome model
o Outcome = treatment (T) + natural course (NC) + extraneous factors (EF) + error
processes (V)
o The outcome is a combination of different phenomena → it could be due to the
treatment effect, but it could also be due to the natural course of the disease,
extraneous factors (i.e. starting/quitting smoking, going to the gym etc.), error
processes (= natural variation, i.e. due to the measurement device you use)
o Obviously, our interest lies in the treatment
o Possible outcomes
▪ Outcome with treatment = T + NC + EF + V
▪ Outcome without treatment = NC + EF + V
o Comparison
▪ To identify treatment effect:
• Compare 2 (or more) groups
• Those groups should be comparable with respect to NC, EF, and V
• And differ only with respect to treatment!
▪ In that case, an observed difference in the outcome between the groups, can
be attributed to the only aspects that the two groups differ on: the
treatment
▪ → we need comparability (with respect to NC, EF and V)
- Design elements of a randomized controlled trial
o 1. Randomisation
▪ Randomisation = the treatment of interest is randomly allocated to
participants in an RCT → so it is a chance process whether a participant
receives the treatment or an alternative treatment
▪ Randomisation can happen in a lot of different ways
▪ Concealment of treatment allocation
• Physician who asked patients to participate does not know what
treatment the next patient will receive, nor does the patient
him/herself
• As a result, treatment allocation is independent of patients
characteristics
o 2. Blinding
, ▪ a. Participants should not know which treatment they receive, because that
could influence their ‘behaviour’
• This also applies to treating physicians / nurses / relatives / etc.
o Blinding aims to keep the groups comparable during follow-
up
▪ b. Placebo vs. active comparator
• 1. Placebo
o It tastes / looks / smells like the active treatment but does
not contain the active compound
o Sometimes difficult → e.g. surgery, physiotherapy
• 2. Active comparator
o E.g. a new diuretic treatment vs. hydrochlorothiazide (= a
different, already existing diuretic treatment)
▪ c. Blinded outcome assessment
• The one who ‘measures’ the outcome should not know about the
treatment status, because that could influence the measurement
o 3. Standardisation
▪ Standardization of intervention
▪ Standardization of concomitant care
▪ Standardization outcome assessment
▪ Reason for including randomization:
• Minimize error processes
• Improve the interpretability of treatment effect
- Comparability
o Comparability is key when making a comparison between two groups
o Comparability is relevant at different stages of an RCT:
- Summary
o Estimation of ‘effect’ of treatment requires comparability
o In a randomised trial, several design elements aim at improving comparability:
▪ Concealment of allocation
▪ Randomisation
▪ Blinding
▪ Standardisation
HC Sample size calculations
- Randomized controlled trial (RCT)
o Aim: compare two treatments
o Patients are recruited to the study, and randomized to treatment A or B
o How many patients are needed (= sample size)?
, ▪ Too few: not able to detect differences between the 2 groups
▪ Too many: costs and not ethical
o Factors for deciding sample size
▪ Practical
• Number of eligible patients treated at centre
• 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 (quick overview)
o 1. Decide on a null hypothesis H0 about the population
▪ H0: there is no difference between the two groups
o 2. Take a representative sample of the population
o 3. Calculate the observed difference in the sample
o 4. Calculate the p-value, the probability to observe at least this difference if H0 is
true. This is done by a statistical test.
o 5. If p-value is small, smaller than a prespecified value α, we reject H0
▪ The value α is called the significance level
o Type 1 and type 2 errors
▪ Type 1 error (α): rejection of the H0, while it
is true → so rejection of the H0, thus saying that
the two populations differ, while there in fact is no
significant different between the two groups
▪ Type 2 error (β): acceptance of the H0, while
it should have been rejected → so no rejection of
the H0, thus saying that there is a difference
between the two populations, while in fact there is
a significant difference between the two groups
▪ Power = 1 – the probability of a type II error
- Power
o Power = the probability of finding a significant effect in your sample when the effect
is really present in the population
o Depends on:
▪ Relevant difference (effect size)
▪ Sample size → power increases, as sample size increases
▪ Variance / standard deviation → more variation, leads to smaller power
▪ Significance level α
o We want to have a study with a large power!!
- Example study
o RCT on patients with high blood pressure → comparison between intervention (new)
and active comparator (old); outcome: blood pressure after 6 weeks of treatment
o Trial with two groups of 30 patients, H0 is true!
▪ Distribution of mean difference (between the 2 groups) in blood pressure
