Design of Clinical Trials (Statistics) Flashcards
What is bias? What are the types at each stage of research?
Bias: Any systematic error in design, conduct or analysis that results in a mistaken estimate of the exposures effect on risk of disease
Selection bias: Study sample not representative of the population of interest, due to non-response, loss of follow up etc
Allocation bias: systematic error in intervention/control allocation. E.g. without random allocation doctors may choose to give new treatments to sicker patients
Ascertainment bias - systematic distortion of the assessment of outcome by researchers (assessor bias) or study participants (response bias). E.g. a participant who is aware they are in the placebo arm may not admit to/recognise a change in symptoms due to knowledge that they have taken the placebo.
Analytic bias - bias in analysing data
HAWTHORNE EFFECT- the alteration of behaviour by the subjects of a study due to the awareness of being observed
Define confounding
The effect of the exposure of interest with the effect of another variable, leading to an incorrect effect estimate e.g. shark attacks, ice cream sales
Solution to Selection Bias
Careful recruitment to ensure sample represents population of interest
Using random methods when selecting subgroups from populations
Solution to Allocation bias
Random allocation, usually number generated
Minimises known confound and unknown confounders
Usually shares out characteristics evenly between control and intervention
Types of control
Positive/active control = standard treatment
Negative control = placebo or sham treatment with no active therapeutic ingredient or effect.
MUST BE indistinguishable from the active treatment
Effect must be as close to zero as possible
Issues with use of Negative control
May not be possible/practical
When a certain “standard of care” is routine for a condition, it is not ethical to give nothing to the patient
Types of blinding
Single-blind - Participants only
Double blind - study personnel and participants
Blinding not always achievable or practicable, particularly in clinical settings
PROS: Less response bias, less assessor bias, may reduce loss to follow up
Solution to Analytic Bias
Data analysis of Primary/Secondary outcomes
Triple Blinding
What is treatment response and its components?
Treatment response: observed outcome for intervention group
TREATMENT RESPONSE = TREATMENT EFFECT + NATURAL EPIDEMIOLOGY + PLACEBO EFFECT
What is placebo response and its components?
Observed outcome for the placebo group
PLACEBO RESPONSE = NATURAL EPIDEMIOLOGY + PLACEBO EFFECT
Define natural epidemiology
Outcome response that would occur without treatment or observation (outside of the trial). Natural tendency towards recovery, biological healing
Define placebo effect
A real psychological effect, therapeutic response to investigation, observation/assessment and doctor patient interaction
Hawthorne effect - Behavioural change as a motivational response to interest, care and attention received from observation and assessment. Part of placebo effect
Define treatment effect
Pharmacological effect
TREATMENT EFFECT = TREATMENT RESPONSE - PLACEBO GROUP RESPONSE
What is a Cohort Study and Pros/Cons
- Investigator selects a disease free cohort of exposed or unexposed individuals
- Follows up both groups to compare incidence of disease (or rate of death from disease)
- Outcome = RELATIVE risk, ATTRIBUTABLE risk, ABSOLUTE risk
Prospective cohort study - exposure and non exposure are ascertained as they occur during the study.
Groups are followed up over time and incidence of the outcome is measured
Retrospective cohort study - Exposure is ascertained from past records and outcomes ascertained at the time the study began
What is a Cohort Study and Pros/Cons
- Investigator selects a disease free cohort of exposed or unexposed individuals
- Follows up both groups to compare incidence of disease (or rate of death from disease)
- Outcome = RELATIVE risk, ATTRIBUTABLE risk, ABSOLUTE risk
Prospective cohort study - exposure and non exposure are ascertained as they occur during the study.
Groups are followed up over time and incidence of the outcome is measured
Retrospective cohort study - Exposure is ascertained from past records and outcomes ascertained at the time the study began
