Upload
others
View
9
Download
2
Embed Size (px)
Citation preview
Faculty of Medicine Epidemiology and Biostatistics
واإلحصاء الحيوي الوبائيات (31505204)
Lecture 15 Bias and Confounding
By
Hatim Jaber MD MPH JBCM PhD
30-7-2018
1
1. Basic epidemiological concepts/ Epidemiological study types
1. Association and causation
2. Bias and confounding 3. Screening tests and result interpretation 1. Communicable diseases Epidemiology 2. Transmission of infectious diseases 3. Chronic Non-communicable Diseases Epidemiology 4. Risk factors of NCD 1. Workplace Hazards – Radiation and Noise at workplace 2. Current global environmental problems, their causes, effects, and
prevention measures.(1) 3. Current global environmental problems, their causes, effects, and
prevention measures.(2) 1. Food contamination and food borne diseases(1) 2. Food contamination and food borne diseases (2)
Presentation outline
Time
Measurements of the variables 10:30 to 10:40
Bias 10:40 to 10:50
Types of bias 10: 50 to 11:10
Cofounding
11:10 to 11:25
Types and ways to control them in various types of biases.
11:25 to 11:35
3
Analytical Methods
1. Measures of association /strength of association
2. Testing hypothesis of association
3. Controlling confounders
To Show a Valid Statistical Association We need to assess:
1. Bias: whether systematic error has been built into the study design
2. Confounding: whether an extraneous factor is related to both the disease and the exposure
3. Role of chance: how likely is it that what we found is a true finding
Measurements
• At every step in our medical practice – whether asking the history of a symptom, looking for a sign or reviewing the investigations, we are in fact making ‘measurements’.
Measurements of the variables listed in any study has 3 categories:
1. exposure 2. outcome and 3. confounders
Research Variables
In general we should list out the following categories of variables related to our research objectives :
• (a) The exposure variable (s)
• (b) The outcome variable (s)
• (c) The potential confounding variables and probable effect modifiers.
For example
• in a study proceeding with the research objective of studying the association between alcohol use and oral cancer, the variables of study would be :
• Exposure variable : Alcohol use
• Outcome variable : Oral cancer
• Potential Confounding variables and possible effect modifying variables : Age, Sex, tobacco use, oral hygiene, dental malformations and genetic background
Three reasons due to which our measurement process may become incorrect
1. ● The basic technique of measurement may be incorrect, due to defective instruments, wrong techniques, inadequately trained data collectors, etc. This is the situation of “measurement error”.
2. ● While selecting the two groups or else while making measurements on the two groups of subjects, which are to be compared, we start treating the two groups in a differential manner. This is known as systematic error or Bias.
3. ● The observed association which we have shown between the exposure and the outcome variables, is actually due to a third, indirectly acting variable and not really due to an association between the exposure and outcome variables. This is the situation of confounding error
Internal validity
Any one or more of the above erroneous situations:
1. (measurement error,
2. systematic error or Bias,
3. confounding error),
if present, will lead to “lack of internal validity” of the epidemiological or research work.
Threats to Internal & External Validity
• Keep in mind that in science it is always important to balance multiple concerns. – Ethics: Harm versus Benefit
– Sampling: Representativeness versus Practicality • When choosing a research design, it is important to address
whether:
– You can address or eliminate alternative explanations for your results (Internal Validity).
– You can generalize your results (External Validity).
Basic Measurement Technique
• (a) The measurement process should be valid, i.e. the measurements which we are making and recording should correctly measure what we really intend to measure
• (b) Secondly, the measurement process should have “Reliability” (Syn : Precision, consistency, replicability, repeatability)
Types of Association
Association may be grouped into following three types;
1. Spurious Association : When the observed association
between suspected cause and effect may not be real. Example- Perinatal mortality being high in hospital deliveries than home deliveries implying hospital is unsafe. The cause
of spurious association is poor control of Biases in study.
2. Indirect Association : It is a statistical association between a factor of interest and a disease due to presence of another
factor known as Confounding Factor. Example-Iodine deficiency and Altitude association with Endemic Goitre.
3. Direct Causal Association : One to one and multifactorial 13
THE DIFFERENCE BETWEEN BIAS AND CONFOUNDING
Bias creates an association that is
not true,
but confounding describes an
association that is true, but
potentially misleading.
Chance vs Bias
• Chance is caused by random error
• Errors from chance will cancel each other out in the long run (large sample size)
• Chance leads to imprecise results
• Bias is caused by systematic error
• Errors from bias will not cancel each other out whatever the sample size
• Bias leads to inaccurate results
Bias is one of the three major threats to internal validity:
1. Bias
2. Confounding
3. Random error / chance
What is Bias?
Any trend in the collection, analysis, interpretation, publication or review of data that can lead to conclusions that are systematically different from the truth (Last, 2001)
A process at any state of inference tending to produce results that depart systematically from the true values (Fletcher et al, 1988)
Systematic error in design or conduct of a study (Szklo et al, 2000)
What is Bias?
Errors can be differential (systematic) or non-differential (random)
Random error: use of invalid outcome measure that equally misclassifies cases and controls Differential error: use of an invalid measures that misclassifies cases in one direction and misclassifies controls in another
Term 'bias' should be reserved for differential or systematic error
Bias is systematic error
BIAS
Systematic error built into the study design
Selection Bias
Information Bias
Selection bias Unrepresentative nature of sample Information (misclassification) bias Errors in measurement of exposure of disease ** Confounding bias ** Distortion of exposure - disease relation by some other factor Types of bias not mutually exclusive (effect modification is not bias)
Types of Bias
Types of Selection Bias
• Berksonian bias – There may be a spurious
association between diseases or between a characteristic and a disease because of the different probabilities of admission to a hospital for those with the disease, without the disease and with the characteristic of interest Berkson J. Limitations of the application of fourfold table analysis to
hospital data. Biometrics 1946;2:47-53
Types of Selection Bias (cont.)
• Response Bias – those who agree to be in a study
may be in some way different from those who
refuse to participate
– Volunteers may be different from those who are
enlisted
Types of Information Bias
• Interviewer Bias – an interviewer’s knowledge may
influence the structure of questions and the manner of
presentation, which may influence responses
• Recall Bias – those with a particular outcome or
exposure may remember events more clearly or
amplify their recollections
Types of Information Bias (cont.)
• Observer Bias – observers may have preconceived
expectations of what they should find in an
examination
• Loss to follow-up – those that are lost to follow-up
or who withdraw from the study may be different from
those who are followed for the entire study
Information Bias (cont.)
• Hawthorne effect – an effect first documented at
a Hawthorne manufacturing plant; people act
differently if they know they are being watched
• Surveillance bias – the group with the known
exposure or outcome may be followed more closely
or longer than the comparison group
Information Bias (cont.)
• Misclassification bias – errors are made in
classifying either disease or exposure status
Types of Misclassification Bias
• Differential misclassification – Errors in
measurement are one way only
– Example: Measurement bias – instrumentation may
be inaccurate, such as using only one size blood
pressure cuff to take measurements on both adults
and children
Misclassification Bias (cont.)
• Nondifferential (random) misclassification –
errors in assignment of group happens in more than
one direction
– This will dilute the study findings -
BIAS TOWARD THE NULL
Controls for Bias
• Be purposeful in the study design to minimize the chance for bias – Example: use more than one control group
• Define, a priori, who is a case or what constitutes exposure so that there is no overlap – Define categories within groups clearly (age
groups, aggregates of person years)
• Set up strict guidelines for data collection – Train observers or interviewers to obtain data in
the same fashion – It is preferable to use more than one observer or
interviewer, but not so many that they cannot be trained in an identical manner
• Randomly allocate observers/interviewer data collection assignments
• Institute a masking process if appropriate
– Single masked study – subjects are unaware of whether they are in the experimental or control group
– Double masked study – the subject and the observer are unaware of the subject’s group allocation
– Triple masked study – the subject, observer and data analyst are unaware of the subject’s group allocation
• Build in methods to minimize loss to follow-up
Controls for Bias (cont)
Controlling for Information Bias
- Blinding prevents investigators and interviewers from knowing case/control or exposed/non-exposed status of a given participant
- Form of survey mail may impose less “white coat tension” than a phone or face-to-face interview - Questionnaire
use multiple questions that ask same information acts as a built in double-check
- Accuracy multiple checks in medical records
gathering diagnosis data from multiple sources
Available at: http://ebp.lib.uic.edu/applied_health/files/images/random_cry.gif. Accessed on Oct 18, 2011.
Confounding
• Definition : A confounding variable is one which throws into confusion, an observed association between an exposure and an outcome variable
A confounder variable should have the following properties
• (i) Be associated with the exposure of interest.
• (ii) Be (independent of the exposure), related to the
outcome of the interest.
• (iii) It should not be in the direct chain or link between the exposure and outcome; its associations with exposure and outcome are indirect and independent.
• (iv) It exerts its effect because it is differentially distributed in the two groups
Confounding: A Fundamental Problem of
Causal Inference
Confounding is bias due to inherent (unobservable)
differences in risk between exposed and unexposed
populations, i.e., a lack of comparability.
Confounding is usually not a major source of bias in
randomized trials (assuming sample size is large enough)
because randomization tends to equalize inherent risks
between treatment groups
(treated group = exposed, untreated = unexposed)
37
Confounding
May lead to observation of association when
none exists
May obscure an association that exists
Information on potential confounders
should be collected in the study and used in
analysis, otherwise they cannot be excluded
as alternate explanations for findings
Confounding factors must be considered
during study design
38
Controlling confounders
At time designing of epidemiological study or while carrying study
1. Randomization 2. Restriction 3. Matching
At analysis stage 1. Stratification 2. Adjustment 3. Statistical modeling
The most important step is to be aware of the phenomena of confounding and to identify all Potential Confounding Variables (PCV) right at the time when the research question is being developed. Once all PCV have been identified, action may be taken to control them either in planning stage or during analysis, by following methods :
A case-control study was conducted to investigate the association between artificial sweetener and bladder cancer.
Controls were selected from a group of people diagnosed with obesity related conditions.
It is well known that obesity related conditions are associated with an increased likelihood of using artificial sweetener.
Could the association between artificial sweetener and bladder cancer be confounded by any external factors?
Artificial Sweetener and Bladder Cancer
Obesity related conditions
Artificial sweetener --- Bladder Cancer
Unhealthy lifestyle (consumption of other artificial preservatives & carcinogens)
Confounding
Can you think of any other factors?
• A study was done to explore the association between birth order and Down syndrome. It was found with increasing birth order, there was also an increase in the occurrence of Down syndrome.
• The prevalence of Down syndrome was 6/1000 live births at the first birth and 16/1000 live births, for birth of 5 or greater.
Birth Order and Down Syndrome
0
2
4
6
8
10
12
14
16
18
Birth Order
1st Child
2nd Child
3rd Child
4th Child
5th Child
Birth Order and Down syndrome A
ffecte
d b
abie
s p
er
10
00
liv
e b
irth
s
Kennith J. Rothman, Epidemiology and introduction, p 102
• Mother’s age
• Birth order ------------ Down syndrome
What do you think could affect this trend?
Because mother age and birth order are highly correlated, we expect that mothers who give birth to their fifth baby might be considerably older than mothers giving birth to first baby. We also know that the risk of Down syndrome increases with maternal age.
• A study was done to examine the association between Caffeine and Breast cancer. The following data was obtained:
• What are the odds of caffiene intake in cases compared to controls?
Exercise on Confounding
Caffeine Breast Cancer No breast cancer
Yes 30 18
No 70 82
Total 100 100
OR= 30x82 = 1.95 70x18
• The investigators thought that the calculated OR was confounded by the effect of age. They stratified participants according to age:
• Age <40 Age ≥ 40
• OR=? OR=?
• What do you conclude?
Is age a confounder?
Caffeine Ca No Ca
Yes 5 8
No 45 72
Total 50 80
Caffeine Ca No Ca
Yes 25 10
No 25 10
Total 50 20
OR= 5x 72 = 1 45x8
OR= 25x10 = 1 25x10
Age confounds the association between caffeine intake and breast cancer
• Nurse’s Health study; a cohort study was done to determine the association between oral contraceptive use and ovarian cancer. The following data were obtained:
• What is the crude RR?
Exercise 2
Oral contraceptive Ovarian Ca No ovarian Ca Total
Yes 350 200 550
No 125 200 325
Total 475 400 875
RR= 350÷550 = 1.65 125÷325
• The investigators thought that the calculated Risk ratio might be confounded by the effect of smoking. Data were stratified according to smoking status and relationship was studied :
• Smokers Non-smokers
• RR=? RR=?
Is smoking a confounder?
OCP Ovarian Ca No ovarian Ca
Total
Yes 298 152 450
No 100 50 150
OCP Ovarian Ca No ovarian Ca
Total
Yes 95 5 100
No 150 25 175
RR= 298÷450 = 0.99 100÷150
RR= 95÷100 = 1.1 150÷175
• A third factor which is related to both exposure and outcome, and which accounts for some/all of the observed relationship between the two
• Confounder not a result of the exposure – e.g., association between child’s birth rank
(exposure) and Down syndrome (outcome); mother’s age a confounder?
– e.g., association between mother’s age (exposure) and Down syndrome (outcome); birth rank a confounder?
Confounding
Exposure Outcome
Third variable
To be a confounding factor, two conditions must be met:
Be associated with exposure
- without being the consequence of exposure
Be associated with outcome
- independently of exposure (not an intermediary)
Confounding
Birth Order Down Syndrome
Maternal Age
Confounding
Maternal age is correlated with birth order and a risk factor even if birth order is low
Birth Order
Down Syndrome Maternal Age
Confounding ?
Birth order is correlated with maternal age but not a risk factor in younger mothers
Coffee CHD
Smoking
Confounding
Smoking is correlated with coffee drinking and a risk factor even for those who do not drink coffee
Coffee
CHD Smoking
Confounding ?
Coffee drinking may be correlated with smoking but is not a risk factor in non-smokers
Alcohol Lung Cancer
Smoking
Confounding
Smoking is correlated with alcohol consumption and a risk factor even for those who do not drink alcohol
Smoking CHD
Yellow fingers
Not related to the outcome
Not an independent risk factor
Confounding ?
Diet CHD
Cholesterol
Confounding ?
On the causal pathway
Confounding
(www)
If each case is matched with a same-age control, there will be no association (OR for old age = 2.6, P = 0.0001)
Confounding or Effect Modification
Birth Weight Leukaemia
Sex
Can sex be responsible for the birth weight association in leukaemia?
- Is it correlated with birth weight?
- Is it correlated with leukaemia independently of birth weight?
- Is it on the causal pathway?
- Can it be associated with leukaemia even if birth weight is low?
- Is sex distribution uneven in comparison groups?
Confounding or Effect Modification
Birth Weight Leukaemia
Sex
Does birth weight association differ in strength according to sex?
Birth Weight Leukaemia
Birth Weight Leukaemia / /
BOYS
GIRLS
OR = 1.8
OR = 0.9
OR = 1.5
Effect Modification
In an association study, if the strength of the association varies over different categories of a third variable, this is called effect modification.
The third variable is changing the effect of the exposure.
The effect modifier may be sex, age, an environmental exposure or a genetic effect.
Effect modification is similar to interaction in statistics.
There is no adjustment for effect modification. Once it is detected, stratified analysis can be used to obtain stratum-specific odds ratios.
Effect modifier Belongs to nature Different effects in different strata Simple Useful Increases knowledge of biological mechanism Allows targeting of public health action
Confounding factor Belongs to study Adjusted OR/RR different from crude OR/RR Distortion of effect Creates confusion in data Prevent (design) Control (analysis)
HOW TO CONTROL FOR CONFOUNDERS?
• IN STUDY DESIGN…
1. RESTRICTION of subjects according to potential confounders (i.e. simply don’t include confounder in study)
2. RANDOM ALLOCATION of subjects to study groups to attempt to even out unknown confounders
3. MATCHING subjects on potential confounder thus assuring even distribution among study groups
HOW TO CONTROL FOR CONFOUNDERS?
• IN DATA ANALYSIS…
1. STRATIFIED ANALYSIS using the Mantel Haenszel method to adjust for confounders
2. IMPLEMENT A MATCHED-DESIGN after you have collected data (frequency or group)
3. RESTRICTION is still possible at the analysis stage but it means throwing away data
4. MODEL FITTING using regression techniques