Intervention studies Flashcards
Clinical trial - template
- Define objectives
- Define study population and units
- Allocate subjects
- Specify intervention
- Masking/blinding
- Follow up and compliance
- Specifying and measuring outcome
- Analysis
- Ethical considerations
Clinical trial - phases (4)
I Healthy animals to evaluate safety
II Target population (small number), document activity of drugs with no control group
III Large scale study comparing intervention with placebo or existing treatments, evaluating safety
IV post-registration trials to inform most effective way of using product
Clinical trial - design, key concepts (3)
Gold standard for assessing clinical effectiveness of therapeutics and preventive measures. Animals assigned to 1 or 2 groups: intervention or no intervention (e.g. placebo) groups. Animals are followed for a period of time to determine if/when they develop an outcome of interest.
Key concepts:
- Randomization
- Masking
- Follow up
Clinical trial - factorial design - uses
Used when assessing 2 or more interventions; all possible combinations assigned to study subjects (e.g. no treatment, 1 only, 2 only, 1 and 2)
Clinical trials - cross-over design - definition, when used (2)
Each subject gets both interventions; first assignment is randomly assigned, followed by period of washout, then alternate intervention is provided.
Suitable when:
- Condition of subject is stable
- Intervention has short-lasting effect
Clinical trial - eligibility criteria considerations (3)
- Evaluation of therapeutic agents: need clear case definitions
- Evaluation of prophylactic agents: need healthy subjects
- Avoid subjects with high risk for adverse events
Clinical trial - randomization - rationale, method, impact of sample size
Goals of randomization: (1) unbiased treatment allocation (removes influence of investigator/participant on treatment choice); (b) ensure that all factors associated with the outcome (other than the intervention) are equally distributed between groups, including both known and unknown confounders. Basic procedure is to allocate through simple randomization (e.g. using random number generator).
With large sample sizes randomization is usually sufficient to balance groups. With smaller sample sizes matching may also be needed (e.g. match on sex).
Clinical trials - cluster randomization - definition, when used (3)
Groups of animals allocated to receive the intervention. Less statistically efficient and more complicated to analyse. Best to follow up all individuals in cluster; otherwise follow random selection within cluster.
Used when:
- no alternative (e.g. treatment in water, no ability to track individuals);
- there is potential of spread of the treatment to the control group (e.g ectoparasitocides); or
- treatment affects the control group (e.g. herd immunity).
Clinical trial - alternatives to randomization
Uncontrolled trial: before/after comparison with changes in disease subsequent to treatment assumed to be related to the treatment (rarely the case). Requires that outcome is measured in same way, there have been no changes in management, complete database available.
Clinical trial - masking (blinding) - rationale, method, types (3)
Participant/investigator/statistician unaware of group assignment. Reduces bias stemming from knowledge of treatment assignment (e.g. non-compliance/LTF/change in management practices by owner, non-random misclassification by person measuring outcome). Usually achieved through use of placebo. Note that placemebo may still have some effects (e.g. adjuvant may induce immunity) and outcome may be so noticeable that subjects become unblinded.
Single blinding: Participant unaware of assignment (enables equal follow-up/management)
Double blinding: Participant and investigator unaware of assignment (enables equal assessment of outcome)
Triple blinding: Participant, investigator and statistician unaware of assignment (enables unbiased analysis)
Clinical trials - follow-up/compliance (4)
Must ensure all groups are followed rigorously and equally.
Consderations:
- Regular communication
- Incentives (financial, educational, public recognition)
- Evaluate reasons for withdrawal
- Interviews/testing to assess compliance (e.g. collect empty containers)
Clinical trials - measuring the outcome - considerations (2)
- Objective
- Masking to reduce non-random misclassification of outcome
Clinical trial - analysis - approaches (2), other considerations
Approaches:
- Intention to treat: data analyzed according to treatment assignment to preserve randomization (regardless of what actually happened) - dilutes treatment effect but tends to be more reflective of real world (non-compliance etc)
- Per protocol analysis: only subjects which complied with and completed study are included in the analysis
Baseline comparisons (adequacy of randomization) usually presented first. Results reported as incidence (incidence risk or incidence rate) in each group, with incidence compared using relative risk. May use logistic regression if interested in overall outcome (disease did or did not occur) or survival analysis methods if interested in time to outcome (disease occurred sooner rather than later). Other considerations in analysis: repeat measures, cluster randomization, multi-center trials (need to account for variance structures - greater similarity of measurements taken from one participant/site).
Clinical trial - bias
- Selection bias: unlikely since treatment is randomized and outcome hasn’t occurred yet (so can’t select on the basis of exposure/outcome status); LTF bias can arise if owers withdraw from the study for reasons related to the treatment and outcome (e.g. seek treatment for animal elsewhere because animal gets worse while on placebo - only healthiest remain in placebo group)
- Information bias: masking of person measuring outcome reduces likelihood of non-random misclassification of outcome
- Confounding bias: reduced by randomization, otherwise managed in design and analysis phases
Vaccine trial - considerations (3)
Protection influenced by:
- Effectiveness of vaccine (what we want to measure)
- Percentage of population we chose to vaccine
- Level of transmission* (differs by site and manipulated by vaccine)
