Lecture 4 Flashcards
Blinding
certain groups unaware which intervention they or their patiets recieved
single blind = only patient
double blind = patient and health care providers
-prevent expectations from influencing response (reduce bias)
Randomization
determines what groups patients assigned to. Helps balane baseline charactertistics that may or maynot affect the response of the intervention
-allocation concealment: cannot predit which group the next patient will be in
-want an equal balance of gender, weight, height, etc between treatement and control groups
block randomization (bioequivalence study)
equal number of patients assigned to each treatement and achieve balance
stratified randomization
by disease severity, age, BMI
-ensure balance of treatment groups with respect to all possible combinations
-two stage proedure
benefits of randomization (3)
-tends to produce comparable groups, limits effects of unknown confounding variables
-produces valid statistical tests and limits bias
-ethical - all patients have equal oppertunitiy to receive test or control
Disadvanges randomization (3)
-patients may not represnt general study population
-patients or physicians may refuse
-administrative complexity
Bias definition
deviation from the truth
-over or under estimation of the true treatment effect
-inappropriate study design or study team conduct
potential sources of bias (5)
selection, performance, detection, attrition, reporting
selection bias
allocation sequene adequatley generated
-equal liklihood for each patient to be sorted into any one group
-bias can overestimate treatement effects
-allocation concealment
-randomization sequence unknown to person entering participants until allocation has occured
-sequentially numbered drug containers or envelopes
-or use third party such as pharmacy
performance bias
-blinding of key individual from knowing allocation of patients
-proper blidnding minimizes performance and detection bias
-patients less likely to have biased psychological or physical response, comply with trial regimes, less likely to seek additional treatement during trial, less likely to leave trial
-investigators less likely to transfer inclinication to patients, differentially dose, withdraw patients
-assessors less likely to have bias affect outcome assessment, esp with subjective outcomes
Attrition bias
loss of participants from clinical trial after consent and randomization (therefoe have incomplete data for patient)
-intent to treat analysis (ITT): all participants included in their assigned arm whether or not complete the trial
-prevents disruption to baseline equivalence of randomization
-per protocol analysis: patients who complied with protocol sufficiently to ensure their data would likely exibhit the effect of treatement
-may be introduce bias if non-compliace related to treatement
reporting bias
selective outcome reporting
-original variables included in publication of study
-what is reported is different that original plan of the study
-positive outcomes reported more than negative outcomes
-before study need to determine primary and secondary outomes, measurement methods and time points, analysis methods, subsets, subgroups, sensitivity analyses
Other sources of bias (6)
-baseline imbalanes unaccounted for
-protocol deviation
-contamination
-insenstivie measurement instruments
-selective reporting of subgroups
-inappropriate influence from sponsor