GOOD CLINICAL PRACTICE
QUALITY MANAGEMENT SYSTEM
3 pillars:
- conform GCP and international legislation
o need for quality standard (international conference on harmonization 1996, after
softenon drama) => protect patients and improve quality
- conform protocol (only deviate for the safety of the patient!)
- validity of results
quality management system includes:
- quality manual: overarching document with structure, personnel responsibilities, procedures (on
equipment, subjects, archiving) and control of activities that ensure that clinical studies meet the
3 pillars
procedures : SOP (standard operating procedure, more general topics) STP (standard
technical procedures, more technical, eg. how a centrifuge works)
- documents, forms, examples…
Quality control:
- quality control itself is not required for GCP, but it does require quality so controls are needed
- internal audits by 5 employees from the same unit
- approval of documents and processes
Quality assurance:
- independent external GCP audits: once every two years => supervision of implementation of
quality processes and evaluation of the quality management system
Improvement: PDCA = plan, do, check, act
Training employees
- SOP trainings (5y, communal individual SOP and STP training)
- Test for GCP, national and international legislation
- BLS/AED: Basic Life Support & Automated External Defibrillation => accreditation for phase 1
studies
- IATA: international air transport accreditation => transport and storage of different samples (no
shipping, no IATA certificate needed)
- Study specific trainings
- All staff should be registered in a log + their training! (always make sure that all the staff is trained
for the study)
Document management system => Zenya system (has version control, documents first get approved and
is readily available for staff)
Infrastructure: temperature management and control of devices (install alarm when deviating)
+ calibration (eg. ECG every 2y)
,Client satisfaction (surveys to provide feedback from sponsors and subjects) ; register complaints and
evaluate the suppliers (internal eg. laboratory or external eg. other companies)
involvement of management: 1x a year, the head of department goes to advisory board and explain all the
activities + discuss issues about financial, quality, staff, sponsors… => managers not only provide
supervision, but also contribute to quality management involvement of team: each employee
individually ensures compliance with the GCP, SOP and quality guidelines
inspection: first party: internal audits ➔ second party audits: from sponsor, specific for study ➔ third
party audits: FAHMP (Federal Agency for health and medicinal products in Belgium) or FDA (food&drug
administration USA) inspection
before clinical trial directive: all studies submitted to ethics committee (not independent! and FAGG in 15
days => now: 1 European portal (independent EC evaluation), evaluated in 1month
drug research unit Ghent: sponsors (majority is pharmaceutical industry ; university or university hospital)
patient recruitment:
- VCT: verified clinical trials => prevents that people can participate multiple studies at same time!
- Database with healthy active subjects to recruit for studies
- Patients: internally: UZghent doctors recruit patients from various (oncology) departments
externally: recruitment of patients across Flanders (referrals and newsletter) (e.g. for oncology
patients for whom therapy does not work)
,CLINICAL TRIAL DESIGN
Different types of study design
DESCRIPTIVE STUDIES
Cross-sectional: 1 fixed timepoints (eg diagnosis) => describe for each patient what you see
Case report: describe 1 patient over time (what diagnosis, what subtype, which medication and how he
responds)
Case series: number of patients with the same mutation and condition => test therapy and describe how
the patients react
SURVEY RESEARCH
= an organized method of obtaining data from research participants via written, electronic or oral
questioning; analysis can give conclusions about the question in a sample of the population => eg how
patients feel about therapy
Convenience sampling: sample that you take is representative for the type of the population that you
screen => sufficient participants of the right subgroup!
But bias (interviewer, misinterpretation, respondent bias, recall bias, sample bias)
ANALYTIC STUDIES
CASE CONTROL STUDY
Cohort with or without disease => look back whether or not they have been exposed to a certain condition
(cases: exposed controls: not exposed)
RETROSPECTIVE COHORT STUDY
Analyze specific exposures/risk factors which contribute to the disease of the presence => different from
case control bcs there might be other factors influencing the cancer that you overlook!
Population selected with a common starting point, outcome or intervention (information already available
=> reduced chance of bias!) ; used for rare diseases
,PROSPECTIVE COHORT STUDIES
Sample with common exposure (bv smoking) followed over time to assess an outcome (bv cancer) =>
after completion of follow-up period, data is analyzed to determine what risk factors contribute to the
given disease (or other outcome)
More accurate and high quality data: customization of follow up; but time inefficient and more bias
HISTORICAL CONTROL
External control group of previously studied persons with similar characteristics to the present sample
(mostly used when control group is impossible or unethical) => new therapy of today comparing to control
group of earlier study => often used inappropriately! (bcs of population deviations or different number of
participants in both groups => bias!)
Problem: drift (difference between true unknown current control parameter and the observed historical
data ; bv different standard care to controls in past) => bias!, exaggerate the value of new treatment
- Requirements: recent study for same indication, samples with similar demographics and
prognostic factors, similar study procedures, similar outcome measurements…
- From literature or database(= more powerful, but GDPR!)
EXPERIMENTAL STUDIES
PHASE 1
Is the medication safe in humans? : test different doses in healthy volunteers => highest dose without
toxicity goes to phase 2
PHASE 2
Testing the efficacy of the drug : test in patients with certain condition => improvement or not? bv tumor
shrinking ipv growing => goes to phase 3
PHASE 3
Compare new drug to something already on market : better results => get market authorization (more
participants than phase 2 + more rare adverse events will come up)
CONTROL GROUPS
Discriminate outcomes caused by the experimental treatment from outcomes caused by other (non)
specific factors (unrelated or related to the experimental treatment!
Randomized, double-blind trials using random selected control groups provide the strongest scientific
evidence for the treatment efficacy
Placebo: prevent bias by making it exactly the same as the treatment
- Placebo effect: control group performed better than expected, just bcs they are in a study (and
think they got the treatment)
- Other nonspecific factors like the natural course of the disease, feeling special (extra attention
bcs of joining a study)
, - Great for drug trial, not experience based interventions + ethical issues bcs only half of the
participants will only receive treatment
No treatment: participants and investigators are not blinded, participants receive usual care
- Less reliable data compared to other control groups, high risk of bias => large effect sizes
reflecting (non)specific effects and biases
- used for behavior and rehabilitation research
Different treatment dose: difference in dose and frequency => drug trials mostly blinded (working with
partly placebo, but not possible in behavioral research)
Different active ingredient: randomize with experimental product/intervention or a product that is already
available on the market => show that the treatment is better or as good as known treatment
- if they know upfront that the drug will not be better, check if it has less side effects or better in
certain conditions/populations => non-inferiority tests
- difficult to blind participants and investigators so bias ; mostly in drug trials
RANDOMIZATION
= all participants are equally likely to be assigned to the treatment or control group => increases the
study’s internal validity bcs it reduces bias (investigator and selection)
FIXED
Probability of being assigned to a study group does not change as the study progresses
1 Simple randomization: allocate participants following a generated list (order of entering the study, eg.
treatment A is even numbers, treatment B odd numbers)
No selection bias
But premature ending of the study can lead to imbalanced groups!
2 Block randomization : allows balance in the distribution of the treatment
Subdivide into blocks, within each block, the assignments are balanced
Eg. block of 4 participants: the get classified in treatment A or B, with
randomized sequence of these treatments! (AABB, ABAB, ABBA…)
The larger the block, the more permutations of sequence so less
predictable which treatment
3 Stratified randomization : study population is divided into strata according to important variables (eg.
prognostic factors, demographics) => then randomize
(simple or block) within each stratum
To exclude effects of covariants on the outcome of the
study (eg. you know upfront that age or HPV infection will
influence the outcome)
QUALITY MANAGEMENT SYSTEM
3 pillars:
- conform GCP and international legislation
o need for quality standard (international conference on harmonization 1996, after
softenon drama) => protect patients and improve quality
- conform protocol (only deviate for the safety of the patient!)
- validity of results
quality management system includes:
- quality manual: overarching document with structure, personnel responsibilities, procedures (on
equipment, subjects, archiving) and control of activities that ensure that clinical studies meet the
3 pillars
procedures : SOP (standard operating procedure, more general topics) STP (standard
technical procedures, more technical, eg. how a centrifuge works)
- documents, forms, examples…
Quality control:
- quality control itself is not required for GCP, but it does require quality so controls are needed
- internal audits by 5 employees from the same unit
- approval of documents and processes
Quality assurance:
- independent external GCP audits: once every two years => supervision of implementation of
quality processes and evaluation of the quality management system
Improvement: PDCA = plan, do, check, act
Training employees
- SOP trainings (5y, communal individual SOP and STP training)
- Test for GCP, national and international legislation
- BLS/AED: Basic Life Support & Automated External Defibrillation => accreditation for phase 1
studies
- IATA: international air transport accreditation => transport and storage of different samples (no
shipping, no IATA certificate needed)
- Study specific trainings
- All staff should be registered in a log + their training! (always make sure that all the staff is trained
for the study)
Document management system => Zenya system (has version control, documents first get approved and
is readily available for staff)
Infrastructure: temperature management and control of devices (install alarm when deviating)
+ calibration (eg. ECG every 2y)
,Client satisfaction (surveys to provide feedback from sponsors and subjects) ; register complaints and
evaluate the suppliers (internal eg. laboratory or external eg. other companies)
involvement of management: 1x a year, the head of department goes to advisory board and explain all the
activities + discuss issues about financial, quality, staff, sponsors… => managers not only provide
supervision, but also contribute to quality management involvement of team: each employee
individually ensures compliance with the GCP, SOP and quality guidelines
inspection: first party: internal audits ➔ second party audits: from sponsor, specific for study ➔ third
party audits: FAHMP (Federal Agency for health and medicinal products in Belgium) or FDA (food&drug
administration USA) inspection
before clinical trial directive: all studies submitted to ethics committee (not independent! and FAGG in 15
days => now: 1 European portal (independent EC evaluation), evaluated in 1month
drug research unit Ghent: sponsors (majority is pharmaceutical industry ; university or university hospital)
patient recruitment:
- VCT: verified clinical trials => prevents that people can participate multiple studies at same time!
- Database with healthy active subjects to recruit for studies
- Patients: internally: UZghent doctors recruit patients from various (oncology) departments
externally: recruitment of patients across Flanders (referrals and newsletter) (e.g. for oncology
patients for whom therapy does not work)
,CLINICAL TRIAL DESIGN
Different types of study design
DESCRIPTIVE STUDIES
Cross-sectional: 1 fixed timepoints (eg diagnosis) => describe for each patient what you see
Case report: describe 1 patient over time (what diagnosis, what subtype, which medication and how he
responds)
Case series: number of patients with the same mutation and condition => test therapy and describe how
the patients react
SURVEY RESEARCH
= an organized method of obtaining data from research participants via written, electronic or oral
questioning; analysis can give conclusions about the question in a sample of the population => eg how
patients feel about therapy
Convenience sampling: sample that you take is representative for the type of the population that you
screen => sufficient participants of the right subgroup!
But bias (interviewer, misinterpretation, respondent bias, recall bias, sample bias)
ANALYTIC STUDIES
CASE CONTROL STUDY
Cohort with or without disease => look back whether or not they have been exposed to a certain condition
(cases: exposed controls: not exposed)
RETROSPECTIVE COHORT STUDY
Analyze specific exposures/risk factors which contribute to the disease of the presence => different from
case control bcs there might be other factors influencing the cancer that you overlook!
Population selected with a common starting point, outcome or intervention (information already available
=> reduced chance of bias!) ; used for rare diseases
,PROSPECTIVE COHORT STUDIES
Sample with common exposure (bv smoking) followed over time to assess an outcome (bv cancer) =>
after completion of follow-up period, data is analyzed to determine what risk factors contribute to the
given disease (or other outcome)
More accurate and high quality data: customization of follow up; but time inefficient and more bias
HISTORICAL CONTROL
External control group of previously studied persons with similar characteristics to the present sample
(mostly used when control group is impossible or unethical) => new therapy of today comparing to control
group of earlier study => often used inappropriately! (bcs of population deviations or different number of
participants in both groups => bias!)
Problem: drift (difference between true unknown current control parameter and the observed historical
data ; bv different standard care to controls in past) => bias!, exaggerate the value of new treatment
- Requirements: recent study for same indication, samples with similar demographics and
prognostic factors, similar study procedures, similar outcome measurements…
- From literature or database(= more powerful, but GDPR!)
EXPERIMENTAL STUDIES
PHASE 1
Is the medication safe in humans? : test different doses in healthy volunteers => highest dose without
toxicity goes to phase 2
PHASE 2
Testing the efficacy of the drug : test in patients with certain condition => improvement or not? bv tumor
shrinking ipv growing => goes to phase 3
PHASE 3
Compare new drug to something already on market : better results => get market authorization (more
participants than phase 2 + more rare adverse events will come up)
CONTROL GROUPS
Discriminate outcomes caused by the experimental treatment from outcomes caused by other (non)
specific factors (unrelated or related to the experimental treatment!
Randomized, double-blind trials using random selected control groups provide the strongest scientific
evidence for the treatment efficacy
Placebo: prevent bias by making it exactly the same as the treatment
- Placebo effect: control group performed better than expected, just bcs they are in a study (and
think they got the treatment)
- Other nonspecific factors like the natural course of the disease, feeling special (extra attention
bcs of joining a study)
, - Great for drug trial, not experience based interventions + ethical issues bcs only half of the
participants will only receive treatment
No treatment: participants and investigators are not blinded, participants receive usual care
- Less reliable data compared to other control groups, high risk of bias => large effect sizes
reflecting (non)specific effects and biases
- used for behavior and rehabilitation research
Different treatment dose: difference in dose and frequency => drug trials mostly blinded (working with
partly placebo, but not possible in behavioral research)
Different active ingredient: randomize with experimental product/intervention or a product that is already
available on the market => show that the treatment is better or as good as known treatment
- if they know upfront that the drug will not be better, check if it has less side effects or better in
certain conditions/populations => non-inferiority tests
- difficult to blind participants and investigators so bias ; mostly in drug trials
RANDOMIZATION
= all participants are equally likely to be assigned to the treatment or control group => increases the
study’s internal validity bcs it reduces bias (investigator and selection)
FIXED
Probability of being assigned to a study group does not change as the study progresses
1 Simple randomization: allocate participants following a generated list (order of entering the study, eg.
treatment A is even numbers, treatment B odd numbers)
No selection bias
But premature ending of the study can lead to imbalanced groups!
2 Block randomization : allows balance in the distribution of the treatment
Subdivide into blocks, within each block, the assignments are balanced
Eg. block of 4 participants: the get classified in treatment A or B, with
randomized sequence of these treatments! (AABB, ABAB, ABBA…)
The larger the block, the more permutations of sequence so less
predictable which treatment
3 Stratified randomization : study population is divided into strata according to important variables (eg.
prognostic factors, demographics) => then randomize
(simple or block) within each stratum
To exclude effects of covariants on the outcome of the
study (eg. you know upfront that age or HPV infection will
influence the outcome)
