Study Bias Flashcards
What are the factors that can lead to study bias?
-not randomized
-retrospective
-observational
-no blinding
-lack of strict interventional methods
Why are low-evident study designs important?
They help to determine if it is reasonable to use stronger study designs to investigate a hypothesis
What is Selection Bias?
-participants selected for the study are not representative for the target population
-> can’t be generalized f.e. exclusion of women -> low external validity
-Participants are not randomly selected f.e. choosing patients from an area with more healthy persons -> BIAS
What is Performance Bias and how is it avoided?
-Different treatment of groups bc researchers are not blinded and deliver more care to people in one group (often experimental group) compared to another group (Placebo group)
-can be avoided by double-blinding
What is Observational Bias?
-experimenter bias
-researches expectations influence their observation and interpretation
-they see what they want to see
-to minimize -> double-blinding
Recall Bias
-Participants inaccurately recall their past events or exposure due to their current condition or study goals
-F.e. The study is about saturated fat intake, the participant may underestimate his fat intake to claim he eats healthier than he actually does
-Concerning when events are far in the past!
What is Attrition Bias?
Different dropouts of patients in treatment and control groups -> affect the accuracy of your study results
-> f.e. it can worsen the outcome of the drug in the drug arm bc there wasn’t enough time to see the effect of the drug in those who dropped out
-often in the treatment group due to side effects
What type of Validity is affected by Attrition?
Internal and external Validity
How can the impact of Attrition be minimized?
Intent-to-treat Analysis
-Once randomized always analyzed -> as long as participants receive any dose their data are being recorded in the outcomes, no matter if and when they drop out (the intention was to treat everybody so we analyze everybody)
-carefully selecting and recruiting participants, maintaining open and effective communication with participants, providing incentives for participation,
What is the Hawthorne Effect?
Participants modify their behavior because they are aware that they are being observed (studied)
-> f.e. taking more care in following protocols -> better outcomes compared to a real-world setting w/o close monitoring
OR general change in behavior -> affecting the validity
How are valid results ensured despite the Hawthorne Effect?
-Use a Placebo Group
The Hawthorne Effect is observed in the Placebo group, as well as in the experimental group; to number of affected people can be subtracted in the experimental group to determine the real number of people affected by the drug rather than by the Hawthorne Effect + BLINDING
f.e.: students logging their study time -> they may study more bc they know are observed and they should study appropriately
Ways the Hawthorne Effect affects studies
-Increased compliance
-Placebo effect enhancement
-Exaggerated Reporting
-Temporary Behavior Change
What is a placebo effect?
- when a person’s belief in the efficacy of treatment leads to actual physiological or psychological changes -> while the treatment itself is inactive
-Simply the thought of getting a drug can cause an effect -> in the placebo group as well as in the experimental group
-the number of pt with the placebo effect on the placebo group can be subtracted from the experimental group -> ensuring validity
Characteristics of a Well-performed RCTs
-Well-Defined Study Population
-Random Allocation
-Placebo-controlled
-Appropriate Blinding
-Good adherence/compliance
-Minimal attrition (lost to follow-up)
Strength of RCTs
RCTs minimize bias – maximize internal validity
-Measure a primary outcome
-Highly selective group of patients
-Randomly assigned to receive one or more
intervention
In which phase of a trial are RCTs used?
phase III trials and a New Drug Application (NDA)
What is Simple Randomization
Simply using a number generator and allocating pt to either drug or placebo group
What determines if a study has a well-defined population?
Inclusion and Exclusion -> The question of the study has to be specific -> f.e. investigation of DIABETES 2 in 60-70 yr old
Inclusion of 60-70yr w/ Diab Type 2
Exclusion of patients to limit confounding variables f.e. pt with serious disease bc it may interfere with the ability to answer the question + they might die during the study and affect the results
-It also has to be generalizable (external validity), so don’t exclude smokers bc a high percentage of people smoke
Why is a reasonably homogenous group important for a well-defined population?
an overly heterogeneous group (20-80 y) may dilute the results of the study -> maybe the older pt benefits more than the younger pt
an overly homogenous group may not be generalizable to the population with the disease (f.e. all male)
What is Random Allocation?
Randomize groups to equalize both groups (placebo and drug group) -> neutralizing potential confounding variables
f.e. if you have a potential confounding variable -> distribute them randomized and equally between two groups -> identify the confounding variable and subtract it from the drug group
What is implied by the Table Baseline characteristics in a study?
It displays all characteristics in both groups (placebo and drug group)
-> to check if there is equality in all of the characteristics
What are the types of Randomization?
-Simple randomization
-Block randomization
-Stratified Randomization
-Adaptive Randomization
What is a potential issue in Simple Randomization?
Using a number generator can lead to unequal numbers of pt between the groups WHEN the number of pt is LOW (f.e. coin flip of 10 -> you could get 7-3 distribution)
-> thereby influences the distribution of baseline characteristics across two groups
What are the goals of Randomization?
-should be reproducible and unpredictable
-allocates treatments (drugs) randomly
-should be tamperproof (it shouldn’t be easy to find to recognize the placebo drug)
What should be avoided in Randomization?
-Avoid every other patient allocation (pt may recognize the scheme)
-days of the week - giving out the drug/placebo on a specific day (pt may recognize the scheme OR different quality of care bc inexperienced or experienced providers cover specific days of the week)
-odd/even schemes, don’t use MRN (medical record numbers or date of birth on drugs/placebo)
What is Stratified Randomization?
Participants are stratified based on a potential confounding variable and then allocated to the placebo or drug group
When is Stratified Randomization used?
Used when a confounding variable (age, gender, race, condition) is potentially unequally distributed in two groups
-more common in small studies, the risk of imbalance in larger studies is unlikely
What is Block Randomization?
Randomizing pt in blocks at the beginning of the study f.e. blocks of 10 (5 in placebo and 5 in drug group)
-so that in case of early termination of the study, there are equal groups (early termination -> low numbers -> could lead to unequal groups)
When are Block and Stratified Randomization used?
Block Randomization: to ensure equal numbers in the groups along the way of the study
Stratified Randomization: Maintain equal numbers relative to a confounder
What is Adaptive Randomization?
-UNUSUAL STUDY DESIGN
-used to allocate participants to different treatment groups in a way that allows for adjustments during the course of the trial
-enhance the efficiency of the trial and potentially improve the overall quality of the research
-Treatment assignment is not entirely random but rather influenced by the evolving data
Example of Adaptive Randomization
-Randomized Play-the-Winner
-Allocates more participants to the treatment arm that appears to be performing better based on observed outcomes
What is a placebo control?
-a substance resembling a treatment/drug
-standard for well-designed drug studies
-important to sort out the placebo effect
When would it be unethical to use a placebo?
Studies investigating drugs for pt with severe diseases -> pt would have a 50% chance to receive no treatment (placebo)
-> More appropriate: use active control instead of a placebo
What is an active control?
-Standard treatment (with proven efficacy) used as a control in the control group of a study
-the active control should be superior or equivalent in efficacy and safety to the investigational drug
What is historical control?
-using records of the control group from another time period
-data from previously conducted trials compared with the current study population
-> The comparison is used to estimate future outcomes without a treatment
What are the issues with historical controls?
-Selection BIAS
-Lack of Randomization
-No placebo control - (has got a historical control, but it is not good as a placebo or active control -> bc they can be blinded)
-Unblinded
What are the different types of Blinding?
-Single Blind: either the investigator or the pt is blinded (if only pt is blinded, there could still be room for observational bias for investigators)
-Double-blind:
-Triple-blind: Patient, investigator, and third party (evaluator) is blinded
-Open-label: Unblinded
What type of Blinding is required by the FDA for drug approval?
At least two double-blinded trials
When are open-labeled studies conducted?
-f.e. Phase I of a study, with a small group to test for side effects, toxicity, Pharmacokinetics
-No BIAS expected bc you just look for objective findings
When are Study Blindings required?
-to avoid BIAS
-researcher has a strong desire for a study to go well
-hope that the new drug will be better
-subjective outcomes
Reasons for blinding based on the groups
Blinding subject (Patient): to avoid the Placebo effect (if they knew, they would experience the placebo effect), to avoid dropout - if they knew they are getting Placebo, biased reporting
Blinding investigator team: to avoid Observational bias, performance bias (the way they treat pt)
Blinding evaluator: to avoid Observational BIAS
What is the double-dummy technique?
-used to blind studies with 2 dosage forms (capsule, IV)
-Each patient gets both: the active agent of one dosage form and a placebo of the other dosage form
What are other reasons to use Open-labeled studies?
-When blinded studies are impractical or unethical
-f.e. surgery: where you put 2 subjects under the same condition, but in one subject you don’t perform the intervention (don’t take out the organs f.e.) to see if there are different outcomes -> possible in animals - not in human
