Randomisation
The process of assigning trial subjects to treatment or control groups using an element of chance, so that each participant has an equal chance of being allocated to either group.
This is to reduce allocation bias and balance known and unknown possible confounding factors, such that any difference between the two groups is purely by chance.
Stratified randomisation
Prevents imbalance between treatment groups for known factors that influence prognosis or treatment responsiveness. As a result, stratification may prevent type I error and improve power for small trials (<400 patients), but only when the stratification factors have a large effect on the prognosis.
Allocation concealment
Refers to masking the randomisation code or sequence before patients are recruited so that investigators don’t know what group the next patient will be randomised to, thus avoiding selection bias
Blinding
Involves masking the allocation of the treatment group after randomisation, so one or more parties involved in the trial are not aware whether the participants are in the intervention or control group. This aims to remove bias, including performance and observer bias.
Efficacy
The capacity for the therapeutic effect of a given intervention, e.g. whether a drug demonstrates a health benefit over a placebo or other intervention when tested in an ideal situation. These are called explanatory trials.
Effectiveness
How well an intervention works in practice, i.e. in the real world outside a clinical trial. This tends to be lower than the efficacy. There are called pragmatic trials.
Placebo
Traditionally refers to a substance or treatment with no active therapeutic effect. In medical research, it is unethical to give a substance without therapeutic effect when one is available. Hence, a placebo refers to the control used, for example, the gold-standard intervention in which the intervention of interest is being measured against.
Surrogate endpoint
A variable that is relatively easily measured and which predicts a rare or distant outcome of the intervention tested. It is not a direct measure of either harm or clinical benefit.
Surrogate outcome
Markers that are substituted in for a clinical outcome e.g. ejection fraction as an indicator of disease severity in heart failure
Composite outcome
A combination of variables e.g. a scoring tool. Useful when each variable in the scoring tool is rare as it gives power to the study. Difficult to interpret.
Internal validity
the degree to which the causal relationships found in a study can be trusted (with bias reducing this).
Secondary research
E.g. systematic review/literature review. This means research that amalgamates the results of many studies, largely negating the influence of bias on the results.
Systematic review definition
Assesses all the published literature on a
research question, using a structured protocol to review each paper. This is useful for establishing the current evidence base for a particular research question.
Meta-analysis definition
A meta-analysis is the statistical analysis of the results of several trials, which are combined in order to minimise bias, reach a more accurate ‘true’ population effect, and increase the statistical power of the evidence.
(May accompany a systematic review and involves the
statistical analysis of the results of many research trials. By combining the results of several trials together and increasing the population size, it may be possible to detect statistical differences that previously could not be found. Results of a meta-analysis are presented as a forest plot.)
Advantages of SR and MA
- More accurate assessment of the ‘true’ population effect.
- Larger patient numbers lead to increased power and reduce type II error (which is the erroneous conclusion that an intervention has no effect).
- Less bias (funnel plot → publication bias assessment, heterogeneity analysis → method and data collection)
Disadvantages of SR and MA
- Direct comparisons between studies are difficult, due to differing methodology.
- It can introduce types of bias (selection, publication (consider grey literature), and language bias (non-English articles), selective reporting bias)
Steps followed in a SR
- Check for existing reviews/protocols
- Identify your research question
- Define inclusion/exclusion criteria
- Search for studies
- Select studies for inclusion based on pre-defined criteria
- Extract data from included studies
- Evaluated risk of bias of included studies
- Present results and assess quality of evidence
Funnel plot:
Scatter plot of the intervention effect estimates from
individual studies against some measure of each study’s size or precision.
Steps followed in a MA
- Frame a question (based on a theory):
PICOK method. - Run a search
- Screen
- Extract data
- Determine the quality of information in the articles (assess internal validity but also use GRADE criteria). Critically appraise the information contained within each study. You must decide if the study meets the internal validity criteria. A method to ascertain the quality of each outcome within an article is to use the GRADE criteria (grading recommendations assessment,
development, and evaluation). - Determine the extent to which these articles are heterogeneous.
- Estimate the summary effect size in the form of ORs and using both fixed and random effects models.
- Construct a forest plot
- Determine the extent to which these articles have publication bias and run a funnel plot.
- Conduct subgroup analyses and meta regression to test if there are subsets of research that capture the summary effects.
What is the problem with surrogate outcomes?
May not reflect outcomes that are important to patients
RCT
Study where participants are randomly allocated to an experimental or comparison group. Outcomes in each group are compared to determine the effect of the intervention
RCT pros
- Gold standard for studying treatment effects
- Random allocation reduces selection bias and equally distributes confounding factors between arms
- Prospective: allows one to conclude causation between intervention and outcome (and outcomes determined a priori)
- Reliably measures efficacy
- Allows for meta-analyses
RCT cons
- Difficult
- Time-consuming
- Expensive
- May be prone to underpowering
- Ethical problems in giving different treatments to the groups
Prospective cohort study definition
Observational, prospective. Two groups selected on basis of differences in their exposure to a particular agent and followed up to see how many in each group develop a particular disease.
