bias, precision and generalisability Flashcards
what are the different types of bias?
> selection bias (introduced in stage of allocating n to treatment groups
performance bias (durign the intervention/control treatment delivery)
attrition bias
detection bias
how can we minimise selection bias?
- Use randomisation to allocate participants
o known and unknown confounders will be equally
distributed between the groups - Ensure that randomisation is well conducted
o Generation of random sequence
o Concealment of sequence
what things can contribute to performance bias?
- Surgeons/clinician may not perform allocated
treatment - Patients may not take/engage allocated treatment
- Post randomisation ‘care’ may differ
- Co-interventions received may differ
what contributes to detection bias?
open label trials with a subjective outcome measure
if you knwo you are in the placebo group you might be more likely to report flu symptoms as worse
or n recieving the trial medication may report more side effects. in open label trial n are indeed moe likely to over report events in the intervention arm
what are the different types of withdrawal in a study?
o Withdrawal from treatment (follow-up continued)
o Complete withdrawal from trial excluding notes
review (without participant involvement)
o Complete withdrawal
in what ways can attrition bias present?
o Protocol deviation
o Loss to follow up
o Withdrawal of participant
– Participant deemed ineligible after randomisation
– Safety
– Participant or clinician decision
How do low response rates impact external validity?
might be a systemic bias in those not responding
findings dont apply to population of interest as a whole
what are the implications of losses to follow up?
> attrition bias
impact on internal validity - introduces systemic error or bias in the population the treatment is being tested on
attrition bias - what is this?
- Systematic differences between groups in
withdrawals/losses to follow up from a study - systemic differences in those who do withdraw e.g., all n with high depression. bias’s the picture of those remaining - not representative of population anymore
how can we minimise eclusions after randomisation
- Maximise follow up (e.g. reminders, diary
cards, case notes review etc.) - Consider outcomes that can be collected
without participant input e.g. death - Include all randomised patients in analysis
what is detection bias?
- Systematic differences between groups in how
outcomes are determined - “systematic difference between a true value and the
value actually observed due to observer variation”
how can we deal with detection bias?
- Blinding of outcome assessors may reduce the risk
that knowledge of which intervention was received,
rather than the intervention itself, affects outcome
measurement. - Blinding of outcome assessors can be especially
important for assessment of subjective outcomes,
such as mobility. - Blinded adjudication committee e.g. to determine MI
- use objective outcome measures
ways to minimise bias: design
- Randomisation and adequate concealment
- Blinding
- Selection, definition and measurement of
outcomes - Minimise withdrawals
- Maximise follow-up
strategies to minimise bias: conduct
- Maximise follow-up
- Keep protocol deviations to a minimum
- Ensuring blinding is maintained throughout the study
o Interim analysis
o Emergency unblinding
o Unintentional unblinding
o Maintenance of blinding
Strategies to minimise bias: Analysis
- Detailed statistical Analysis Plan (SAP) written by a
statistician independent to the trial, finalised prior
to analysis of any unblinded data or prior to
datalock - Analysis by intention to treat
- All protocol deviations or exclusions to be agreed
blinded to treatment allocation
