STAR
Bias and Confounding
Knut Borch-Johnsen
STAR - bias and confounding 1
Epidemiology
Natural History
Birth Genetically determined (often unknown) variability
| of susceptibility|
|
Exposure Environment
| (family, social environment, macro environment
| occupation etc.) (rarely objective)
|
Disease Diagnostic threshold
| (not objective)
|
Death Heterogeneity with respect to lethality (unknown)
|
Cause of death Heterogeneity
|
Autopsy Heterogeneity
STAR - bias and confounding 2
Where do we get the
evidence from?
Epidemiology/observational studies
Intervention studies
– Structured
– unstructured experiments
STAR - bias and confounding 3
INTERVENTION STUDIES
Clinical trials
RANDOMIZED (DOBBELT BLIND)
CONTROLED CLINICAL TRIAL
|
RANDOMIZED SINGLE-BLIND CONTRILED CLINICAL TRIAL
|
OPEN TRIAL
AIM:
To Compare the effect of different treatment regiments
STAR - bias and confounding 4
RCCT
Blinding with respect to exposure
Allocation by chance
Control for unknown prognostic
factors (versus stratification)
"Simple" interpretation of results
STAR - bias and confounding 5
RCCT vs. Epidemiology
RCCT Epidemiology
Exposure Controlled Yes No
Objective exposure Yes ?
Comparable groups Yes (randomized) ?
Controlled behaviour randomized No
Case Ascertainment Complete (In-)Complete
Conclusions Strong but Weak but
restricted generalized
Possibility Planned Analytical
Interventions
STAR - bias and confounding 6
Some key problems in
epidemiology
Study population
Methods
Measurements
Multifactorial diseases
Can all risk factors be measured
STAR - bias and confounding 7
Validity of the association
Due to chance
Bias
Confounding
STAR - bias and confounding 8
Did this occur by chance
Statistical test; p-value/confidence intervals
Retinopathy in sample – 60% in males
40% in females
How to test this – what do you need ?
STAR - bias and confounding 9
Did this occur by chance
Statistical test p-value/Confidence Intervals
Retinopathy in sample – 60% in males
40% in females
The probability of observing en effect
at least as extreme as the observed
effect, provided that the nill-
hypothesis (no effect) is true
STAR - bias and confounding 10
Sample size
Each collor is one socioeconomic group
How many needed in the sample to obtain a valid estimate?
STAR - bias and confounding 11
2 minutes
STAR - bias and confounding 12
Types of problems
Bias
– Any systematic error in data collection an
epidemiological study that results in an incorrect
estimate of the association between exposure
and risk of disease
Confounding
– Mixing of the effect of the exposure under study
on the disease with that of a third factor –
associated with the exposure and independent
of that exposure be a risk factor for the disease
STAR - bias and confounding 13
Bias
Any systematic error in en
epidemiological study that results in an
incorrect estimate of the association
between exposure and risk of disease
STAR - bias and confounding 14
Bias
Selection bias
Observation or information bias
STAR - bias and confounding 15
Selection bias
Relates to selection of study-
population
Descriptive studies
– Sample representative for the population
Analytical studies/C-C studies
– Study populations from the same
populations
STAR - bias and confounding 16
Descriptive studies
Sampling strategy
Representative sample
STAR - bias and confounding 17
Descriptive studies
Sampling strategy
Representative sample
HOW TO MAKE A
REPRESENTATIVE
SAMPLE
STAR - bias and confounding 18
2 minutes
STAR - bias and confounding 19
Descriptive studies
Prevalence of complications among
patients with type 2 diabetes
– Screened population
– Population based sample
– Primary care
– Secondary care
– Tertiary care
STAR - bias and confounding 20
Descriptive studies
Representative ness of sample
Population based sampling
– Responders
– Non-responders
Non-responders differs with respect to
– Socioeconomic status
– Morbidity
– Mortality
– Life style
STAR - bias and confounding 21
Selection bias
probability of sampling
Will all with the disease have the same
probability of being diagnosed/sampled
– Women and gallstones
– Men and gastric/duodenal ulcers
– Asbestos exposure and COLD/Cancer
STAR - bias and confounding 22
Selection bias
probability of sampling
Will all with the disease have the same
probability of being diagnosed/sampled
– Women and gallstones
– Men and gastric/duodenal ulcers
Difference in diagnostic threshold
– Asbestos exposure and COLD/Cancer
Economic incentive for diagnosis
Organic solvents and dementia
Asbestos and COLD/Cancer
STAR - bias and confounding 23
Selection bias; analytical
studies
Case-control studies
– Oral contraceptives and
thromboembolism
Hospital based C-C studies
Doctors aware of the possible link
Women with symptoms and using OC more
likely to be hospitalised
Leads to selection bias
STAR - bias and confounding 24
Observation/information
bias
Is a consequence of systematic
differences in the way data on
exposure or outcome are obtained
from the various study groups
Recall bias
Interviewer bias
STAR - bias and confounding 25
Recall bias
The diseased individual remember and
report their previous exposure
experience differently from non-
diseased
Or
The exposed individual reports events
differently from unexposed
STAR - bias and confounding 26
Recall bias
Birth defects among laboratory technicians
working with organic solvents OS
Case-control study
Case = birth defect, control = normal child
Exposure: self reported exposure to OS
C-C-study OR > 1.5 (p<0.01)
Cohort study RR = 1.02 (ns)
WHY
STAR - bias and confounding 27
2 minutes
STAR - bias and confounding 28
Interviewer bias
Soliciting, recording or interpretation
of information may differ between
cases and controls
STAR - bias and confounding 29
Interviewer bias -
solutions
Blinding of interviewer
Structured interviews
Interview guides
”dummy questions” (exposures known
to be unrelated to condition under
study)
STAR - bias and confounding 30
Question
Compare and contrast the likelihood of
selection and observation/information
bias in case-control and cohort study
STAR - bias and confounding 31
2 minutes
STAR - bias and confounding 32
Confounding
Mixing of the effect of the exposure
under study on the disease with that
of a third factor – associated with the
exposure and independent of that
exposure be a risk factor for the
disease
STAR - bias and confounding 33
CONFOUNDING
CONFOUNDER
STUDY FACTOR DISEASE
STAR - bias and confounding 34
Confounder
characteristics
1. Associated to exposure
2. Risk factor in it self
If 1 not 2: Intermediate variable
If 2 not 1: Independent Risk Factor
STAR - bias and confounding 35
Confounders
Diabetes and macrovascular disease
– Hypertension
– Dyslipidaemia
– Smoking
– Low physical activity
STAR - bias and confounding 36
Bias and Confounding
what to do ?
Bias Confounding
– In general terms error in – In general terms
data leading to incorrect incorrect estimation of
estimation of association between
association between exposure and outcome
exposure and outcome due to a third factor
– Solution: improve data associated to exposure
collection and disease
– Solution: restriction;
matching; stratification;
multivariate analysis
Bias: design problem; Confounding: analysis problem
STAR - bias and confounding 37
Other important terms
Misclassification
– By exposure
– By event
– Systematic
– Random
STAR - bias and confounding 38
Other important terms
Misclassification
– By exposure often systematic (recall bias)
– By event often systematic (by
exposure)
– Systematic any result possible
– Random underestimates the effect
STAR - bias and confounding 39
STAR - bias and confounding 40