CLINICAL TRIALS
• Trial protocol: group 9 – Routine coronary function testing in patient with angina with non-
obstructive coronary arteries
• Journal clubs: group 3
WEEK 1
LECTURE: INTRODUCTION TO THE COURSE
Why a course about clinical trials?
• Make a new intervention (drug or other) available or implemented in daily care, efficacy and
effectiveness needs to be proved in several steps.
• Refine/ optimize treatments in a changing practice, also clinical trials are needed to provide
the necessary evidence.
• RCT characteristics: designed experiments… (1) randomisation, (2) control group, (3)
controlled setting
• There are many forms of trials, and some are more suitable given a specific research
questions than others.
• Specific challenges related to specific kind of interventions.
o Drug trials across the life cycle
o Appropriate care (= often doing less)
o Surgical interventions and medical devices
o Complex interventions
Course overview
Please submit a draft proposal to Brightspace before 17th of November 12:00 PM
LECTURE: TRAILS OF NEW DRUG TREATMENTS
Introduction to (randomised) clinical trials
• Trial participants are randomly assigned to experimental group which get experimental
treatment and control group which get standard treatment
• Pre-registration is important because:
o It ensures that negative results are also published, preventing unnecessary exposure of
patients to ineffective treatments
o Allows transparency about the planned primary analyses, so you have to think carefully
about your analyses in advance
o Helps patients find trials they can participate in
,RCT in ALS SOD1 patients
• Randomized controlled trial of Tofersen for SOD1 ALS
• Primary endpoint (change in ALSFRS-R total score) showed no significant difference
between Tofersen and placebo, as the confidence interval included zero, meaning the result
is not convincing in terms of efficacy
• Plasma neurofilament light chain (NfL) levels, a biomarker of neuronal injury, decreased with
Tofersen, indicating a biological effect.
• After 52 weeks, although active treatment stopped, a difference in survival emerged,
suggesting a possible long-term benefit for early treatment.
→ Although the trial did not meet its primary endpoint, and the clinical benefit remains
uncertain, the treatment showed no major safety concerns. Despite limited evidence of
efficacy, Tofersen was approved under exceptional circumstance
→ Clinical trials are not straightforward. Results can be unexpected, biomarkers should be
interpreted within a theoretical framework, careful planning of primary endpoints is needed, and
adequate study duration is needed to fully understand treatment’s potential benefit.
Clinical trials are a complicated journey
• Clinical trial awareness: 85% fail to retain enough patients and 80% fail to finish on time,
50% of sites enroll one or no patients in their studies, 40% of total US pharmaceutical
clinical trial budget goes toward recruitment, and 30% of patients drop out of a clinical trial
• Many trials require multiple sites to participate.
• Ethical and patient protection considerations set high standards for conduct &
documentation (ICH - Good Clinical Practice).
• Trials constitute additional burden to patients.
Knowledge… a few questions…
In a controlled trial to compare two treatments, the main purposes of randomisation are so
that:
• The two groups will be similar in prognostic factors.
Yes. Randomisation helps ensure that known and unknown prognostic factors are balanced
between treatment groups. This prevents confounding and selection bias, so any observed
difference in outcome can be attributed to the treatment rather than baseline differences.
• The clinician does not know which treatments subjects are receiving.
That’s blinding, not randomization. Randomisation determines how treatments are
assigned; blinding prevents bias in treatment administration and assessment by keeping
clinicians or participants unaware of the allocation.
• The sample may be referred to a known population
, Randomisation occurs within the trial, not in sampling from the general population. It
ensures internal validity (fair comparison between groups), not external validity
(generalizability to the population). Representativeness depends on how participants are
recruited, not on randomisation.
• The clinician cannot predict in advance which treatment subjects will receive.
Proper randomisation methods prevent allocation prediction, which avoids selection bias. If
clinicians can guess the next assignment, they might (consciously or unconsciously)
influence which patients get which treatment.
• The number of subjects in each treatment arm are (+/-) the same.
Although simple randomisation often results in roughly equal group sizes,
randomisation does not guarantee equal numbers. Sometimes, unequal allocation
ratios (e.g. 2:1) are used intentionally for ethical or practical reasons.
Randomisation and inference
Randomisation ensures that differences
between study groups arise by chance rather
than bias, strengthening internal validity and
making the study’s findings more reliable.
Inference then allows us to generalise from the
observed study results to the broader
population, while recognising that each step
involves potential error and uncertainty.
Randomisation
• The process of assigning trial subjects to treatment or control groups using an element of
chance to determine the assignments in order to reduce bias (ICH).
• “...having used a random allocation, the sternest critic is unable to say when we eventually
dash into print that quite probably the groups were differentially biased through our
predilections or through our stupidity.”
• Generation of allocation sequences
Blinding
• Randomisation facilliates blinding
• Blinding of participant, treating physician, researcher, data analyst, …
• Balances observation errors & prevents bias
• Balances ‘non-specific’ effects
• Placebo (only makes sense if blinded?)
• Randomization & blinding allow comparison against standard of care
Example: COVID19 Vaccines
• Four vaccines approved: Pfizer/BIONTech, Moderna, Oxford/AZ (not in US), Jansen
• All approvals based on very large clinical trials and interim analysis of results
• Statistical issues for the trials
o Which VE (vaccine efficacy) is exactly estimated?
o Endpoint was disease (“cases”), not infection
o Interim decision making
• Estimating VE: Moderna
o Case definition (global): Confirmed COVID-19) presence of at least 2 of the symptoms
OR 1 of the symptoms AND SARS-CoV-2 positive NT-PCR test during symptomatic
period.
, o Primary endpoint: VE to prevent the first
occurrence of COVID 19 in baseline
seronegative subjects. (including cases
starting after 14 days after second
injection)
o Placebo-Controlled, Randomized,
Observer-Blind. Sample size planned to
30,000, randomized 1:1
▪ Primary analysis at 151 COVID 19
cases
▪ Two interim analyses, at 35% and 70%
of events.
• Key results: Data strongly support protection in
adults, but the certainty of the estimate
decreases with age, mainly because of limited numbers of infections among the oldest
participants
Another set of questions
Assume an anti-depressant clinical trial was conducted in Nijmegen. Efficacy (e.g., based on
HAMD-17) was significantly better for A compared to B. Based on the results of the trial alone
(and its statistical analysis), we can conclude:
• A is better than B for the population of patients that would fulfill the entry criteria of the trial.
-> This refers to the target population (all patients who meet the entry criteria). The trial
provides evidence for inference about that group, but not proof — generalisation requires
external validity.
• B. It is better to prescribe A than B for Mrs. Pieterse, who actually does fulfill these criteria.
-> You cannot conclude efficacy for a single individual (Mrs. Pieterse), because individual
response may differ even if the group effect is significant.
• A is better than B on average for the group of patients that entered the trial.
-> This is what the statistical test actually evaluates: the average treatment effect in the
sample studied.
• The psychiatry clinic in Brussels would treat patients better if it would choose A instead of B
for all its patients that would match the entry criteria.
-> You cannot assume the same results will hold in another clinic (Brussels), since context,
population, and practice may differ — that’s a question of generalisability, not something
proven by the trial alone.
Randomisation
• How?
o Set of envelopes in desk drawer
o Random number generator at desk/computer
o Dial to random number generator / web based / external treatment allocation.
• Simple randomization: each participant is randomly assigned to a treatment group (like
flipping a coin)
o Pro: Easy to implement and fully unpredictable
o Con: May lead to unequal group sizes, especially in smaller trials
• Block randomization: Participants are randomized within smaller blocks (e.g. blocks of 4 or
6; not the same block size for every block but different in random order to make
unpredictable) to ensure balance between treatment groups
o Pro: Keeps numbers of participants in each group roughly equal throughout the trial.
o Con: If block size is known, allocation may become predictable, risking selection bias.
• Trial protocol: group 9 – Routine coronary function testing in patient with angina with non-
obstructive coronary arteries
• Journal clubs: group 3
WEEK 1
LECTURE: INTRODUCTION TO THE COURSE
Why a course about clinical trials?
• Make a new intervention (drug or other) available or implemented in daily care, efficacy and
effectiveness needs to be proved in several steps.
• Refine/ optimize treatments in a changing practice, also clinical trials are needed to provide
the necessary evidence.
• RCT characteristics: designed experiments… (1) randomisation, (2) control group, (3)
controlled setting
• There are many forms of trials, and some are more suitable given a specific research
questions than others.
• Specific challenges related to specific kind of interventions.
o Drug trials across the life cycle
o Appropriate care (= often doing less)
o Surgical interventions and medical devices
o Complex interventions
Course overview
Please submit a draft proposal to Brightspace before 17th of November 12:00 PM
LECTURE: TRAILS OF NEW DRUG TREATMENTS
Introduction to (randomised) clinical trials
• Trial participants are randomly assigned to experimental group which get experimental
treatment and control group which get standard treatment
• Pre-registration is important because:
o It ensures that negative results are also published, preventing unnecessary exposure of
patients to ineffective treatments
o Allows transparency about the planned primary analyses, so you have to think carefully
about your analyses in advance
o Helps patients find trials they can participate in
,RCT in ALS SOD1 patients
• Randomized controlled trial of Tofersen for SOD1 ALS
• Primary endpoint (change in ALSFRS-R total score) showed no significant difference
between Tofersen and placebo, as the confidence interval included zero, meaning the result
is not convincing in terms of efficacy
• Plasma neurofilament light chain (NfL) levels, a biomarker of neuronal injury, decreased with
Tofersen, indicating a biological effect.
• After 52 weeks, although active treatment stopped, a difference in survival emerged,
suggesting a possible long-term benefit for early treatment.
→ Although the trial did not meet its primary endpoint, and the clinical benefit remains
uncertain, the treatment showed no major safety concerns. Despite limited evidence of
efficacy, Tofersen was approved under exceptional circumstance
→ Clinical trials are not straightforward. Results can be unexpected, biomarkers should be
interpreted within a theoretical framework, careful planning of primary endpoints is needed, and
adequate study duration is needed to fully understand treatment’s potential benefit.
Clinical trials are a complicated journey
• Clinical trial awareness: 85% fail to retain enough patients and 80% fail to finish on time,
50% of sites enroll one or no patients in their studies, 40% of total US pharmaceutical
clinical trial budget goes toward recruitment, and 30% of patients drop out of a clinical trial
• Many trials require multiple sites to participate.
• Ethical and patient protection considerations set high standards for conduct &
documentation (ICH - Good Clinical Practice).
• Trials constitute additional burden to patients.
Knowledge… a few questions…
In a controlled trial to compare two treatments, the main purposes of randomisation are so
that:
• The two groups will be similar in prognostic factors.
Yes. Randomisation helps ensure that known and unknown prognostic factors are balanced
between treatment groups. This prevents confounding and selection bias, so any observed
difference in outcome can be attributed to the treatment rather than baseline differences.
• The clinician does not know which treatments subjects are receiving.
That’s blinding, not randomization. Randomisation determines how treatments are
assigned; blinding prevents bias in treatment administration and assessment by keeping
clinicians or participants unaware of the allocation.
• The sample may be referred to a known population
, Randomisation occurs within the trial, not in sampling from the general population. It
ensures internal validity (fair comparison between groups), not external validity
(generalizability to the population). Representativeness depends on how participants are
recruited, not on randomisation.
• The clinician cannot predict in advance which treatment subjects will receive.
Proper randomisation methods prevent allocation prediction, which avoids selection bias. If
clinicians can guess the next assignment, they might (consciously or unconsciously)
influence which patients get which treatment.
• The number of subjects in each treatment arm are (+/-) the same.
Although simple randomisation often results in roughly equal group sizes,
randomisation does not guarantee equal numbers. Sometimes, unequal allocation
ratios (e.g. 2:1) are used intentionally for ethical or practical reasons.
Randomisation and inference
Randomisation ensures that differences
between study groups arise by chance rather
than bias, strengthening internal validity and
making the study’s findings more reliable.
Inference then allows us to generalise from the
observed study results to the broader
population, while recognising that each step
involves potential error and uncertainty.
Randomisation
• The process of assigning trial subjects to treatment or control groups using an element of
chance to determine the assignments in order to reduce bias (ICH).
• “...having used a random allocation, the sternest critic is unable to say when we eventually
dash into print that quite probably the groups were differentially biased through our
predilections or through our stupidity.”
• Generation of allocation sequences
Blinding
• Randomisation facilliates blinding
• Blinding of participant, treating physician, researcher, data analyst, …
• Balances observation errors & prevents bias
• Balances ‘non-specific’ effects
• Placebo (only makes sense if blinded?)
• Randomization & blinding allow comparison against standard of care
Example: COVID19 Vaccines
• Four vaccines approved: Pfizer/BIONTech, Moderna, Oxford/AZ (not in US), Jansen
• All approvals based on very large clinical trials and interim analysis of results
• Statistical issues for the trials
o Which VE (vaccine efficacy) is exactly estimated?
o Endpoint was disease (“cases”), not infection
o Interim decision making
• Estimating VE: Moderna
o Case definition (global): Confirmed COVID-19) presence of at least 2 of the symptoms
OR 1 of the symptoms AND SARS-CoV-2 positive NT-PCR test during symptomatic
period.
, o Primary endpoint: VE to prevent the first
occurrence of COVID 19 in baseline
seronegative subjects. (including cases
starting after 14 days after second
injection)
o Placebo-Controlled, Randomized,
Observer-Blind. Sample size planned to
30,000, randomized 1:1
▪ Primary analysis at 151 COVID 19
cases
▪ Two interim analyses, at 35% and 70%
of events.
• Key results: Data strongly support protection in
adults, but the certainty of the estimate
decreases with age, mainly because of limited numbers of infections among the oldest
participants
Another set of questions
Assume an anti-depressant clinical trial was conducted in Nijmegen. Efficacy (e.g., based on
HAMD-17) was significantly better for A compared to B. Based on the results of the trial alone
(and its statistical analysis), we can conclude:
• A is better than B for the population of patients that would fulfill the entry criteria of the trial.
-> This refers to the target population (all patients who meet the entry criteria). The trial
provides evidence for inference about that group, but not proof — generalisation requires
external validity.
• B. It is better to prescribe A than B for Mrs. Pieterse, who actually does fulfill these criteria.
-> You cannot conclude efficacy for a single individual (Mrs. Pieterse), because individual
response may differ even if the group effect is significant.
• A is better than B on average for the group of patients that entered the trial.
-> This is what the statistical test actually evaluates: the average treatment effect in the
sample studied.
• The psychiatry clinic in Brussels would treat patients better if it would choose A instead of B
for all its patients that would match the entry criteria.
-> You cannot assume the same results will hold in another clinic (Brussels), since context,
population, and practice may differ — that’s a question of generalisability, not something
proven by the trial alone.
Randomisation
• How?
o Set of envelopes in desk drawer
o Random number generator at desk/computer
o Dial to random number generator / web based / external treatment allocation.
• Simple randomization: each participant is randomly assigned to a treatment group (like
flipping a coin)
o Pro: Easy to implement and fully unpredictable
o Con: May lead to unequal group sizes, especially in smaller trials
• Block randomization: Participants are randomized within smaller blocks (e.g. blocks of 4 or
6; not the same block size for every block but different in random order to make
unpredictable) to ensure balance between treatment groups
o Pro: Keeps numbers of participants in each group roughly equal throughout the trial.
o Con: If block size is known, allocation may become predictable, risking selection bias.
