Association and Causation

Association is a statistical relationship between two or more variables e.g. a risk factor and disease (exposure and outcome)

right

what is Association??

Association is a statistical relationship between two or more variables e.g. a risk factor and disease (exposure and outcome)

a statistical relationship between two or more variables e.g. a risk factor and disease (exposure and outcome)

Association is a statistical relationship between two or more variables e.g. a risk factor and disease (exposure and outcome)

causation is?

casual relationship between A and B so A can be the reason for B

Association does not ?

Doesn’t necessarily imply a causal relationship.

not every asoociation has causation

right

Finding an association between a risk factor and a disease is the first step in proving its causality

corrct

first step in proving the causality of assocaition?.

Finding an association between a risk factor and a disease is the first step in proving its causality

Finding an association between _________ and ________ is the first step in proving its causality

Finding an association between a risk factor and a disease is the first step in proving its causality

A statistically significant association can be either spurious (artifactual) or real association.

correct

A statistically significant association can be either

A statistically significant association can be either spurious (artifactual) or real association.

بسبب ال chance or bias حنعرفهم

هنا نعرف انو ال Association ممكن يكون حقيقي أو كاذب

زيادة في السلايد الجاي

correct

Association( Between a risk factor A and a disease B

1- Spurious (Artifactual) Due to: Chance / Bias
2- real

A- causal (direct or indirect)
- direct real causal → Direct causal → Direct causal risk factor ( A → B )
EX . DRINKING CONTMINATED WATER LEAD TO? food poisinung
- indirect real causal → Known intermediate variable in the causal chain
Iodine deficiency → Goiter → Thyroid adenoma

correct

1- Spurious (Artifactual) association Due to??

Chance / Bias

causal (direct or indirect)
- direct real causal →
- indirect real causal →

correct

Statistical association → absent?

Conclude that the Suspected factor is not Implicated in etiology

Statistical association → present?

check if bias present → present? Conclude that the Association is artifactual

bias absent? → Apply 5 criteria → Not satisfied → Conclude that the Association is non causal (Confounded)
bias absent → Apply 5 criteria → satisfied → Conclude causal association

correct

Statistical association → present?

check if bias present → present? Conclude that the Association is artifactual

bias absent ? do criteria →
1- satisfied → causal association
2- non-satisfied → confounding

correct

Statistical association → present?

check if bias present → present? Conclude that the Association is?????

bias absent ? do _________ →
1- satisfied →
2- non-satisfied →

Statistical association → present?

check if bias present → present? Conclude that the Association is artifactual

bias absent ? do criteria →
1- satisfied → causal association
2- non-satisfied → confounding

Spurious (Artifactual) Associations:
- • Epidemiological studies are usually done on samples and not on entire population
- BY spurious association we mean an association which is not present in the population but is falsely inferred from a sample as a result of:
( ASSOCIATION IN SAMPLE PRESENT BUT NOT IN THE POPULATION ) :
a.Chance

b.Bias in the study

CORRE

spurious association WHAT WE MEAN?

- BY spurious association we mean an association which is not present in the population but is falsely inferred from a sample as a result of:
( ASSOCIATION IN SAMPLE PRESENT BUT NOT IN THE POPULATION ) :
a.Chance

b.Bias in the study

How can we control chance and bias?

Chance can be reduced by increasing the sample size

How can we control chance and bias?

Chance can be reduced by increasing the sample size

How can we control chance and bias?

• Chance can be reduced by increasing the sample size
  - Bias can’t be reduced by increasing the sample size but by proper study design and control of other factors

How can we control chance and bias?

1- Chance can be reduced by increasing the sample size.

• Bias can’t be reduced by increasing the sample size but by proper study design and control of other factors

CORRECT

how can we reduce bias?

• Bias can’t be reduced by increasing the sample size but by proper study design and control of other factors

How can we reduce chance ?

1- Chance can be reduced by increasing the sample size.

How can we reduce chance and bias ?

• Bias can’t be reduced by increasing the sample size but by proper study design and control of other factors

1- Chance can be reduced by increasing the sample size.

• Bias can’t be reduced by increasing the sample size but by proper study design and control of other factors

• Bias can’t be reduced by increasing the sample size but by

• Chance can be reduced by ????
  - Bias can’t be reduced by increasing the sample size but by ????

• Chance can be reduced by increasing the sample size
  - Bias can’t be reduced by increasing the sample size but by proper study design and control of other factors

• - Bias can’t be reduced by increasing the sample size but by proper study design and control of other factors

• - Bias can’t be reduced by increasing the sample size but by

• Chance can be reduced by

• Chance can be reduced by increasing the sample size

f

مثال على ال Association by chance

ابا أسوي دراسة عن مستوى الطالبات في الدفعة فأخذت 5 طالبات عشوائيا وأعطيتهم

اختبار وبالصدفة طلع انو ال 5 طالبات متفوقات ومن الأوائل وكلهم جابوا علامات كاملة

هل ممكن أعتمد هذي الدراسة و أقول انو الدفعة كلها حتجيب علامات كاملة في نفس

الاختبار ؟

لا لأنو في الحقيقة الدفعة في منها المتفوقين والمتعثرين وأنا كباحثة ما ينفع العينة تكون

بس من فئة معينة ولكن هنا صار بالصدفة أنا ما كنت مخططة انو كل عينتي تكون طالبات

متفوقات

طيب كيف أحل هذي المشكلة

ببساطة أكبر العينة وأزيد عدد البنات عشان أتأكد انها تتضمن طالبات متفوقين وال

مستواهم متوسط وكمان المتعثرين

corrct

مثال على ال Bias

حناخد نفس المثال الأول ابا أسوي دراسة عن مستوى

الدفعة وأخذت 5 طالبات بس هذي المرة ما أخذتهم

عشوائيًا ولكن قررت اخذ على مزاجي البنات الشطار الي

أحبهم وكلهم جابوا علامات كاملة ونجي نسأل مرة ثانية

هل ممكن أعتمد هذي الدراسة ؟

أكيد لأ لانو العينة حقتي اخترتها بشكل خاطئ وخليت

رأيي الشخصي ومشاعري تتدخل في النتائج

Bias:

• Systematic deviation from truth

Types of Bias:

1- Selection bias:

a- Self selection

b-Selection by researcher

2- Measurement or Information bias:

a- Interviewer bias

b-Recall bias

correct

1- Selection bias:

a- Self selection

b-Selection by researcher

2- Measurement or Information bias:

2- Measurement or Information bias:

a- Interviewer bias

b-Recall bias

Real Association:

• Both chance and bias have been ruled out

• Real association means an association which is present in a population from which the sample is drawn. ( in sample + population)

• Real association could be:

• a.Causal b.Non causal ( Confounding)

correct

Non causal association (Confounded) or secondary association

→ A common factor is associated with both the potential risk factor and disease and can overestimate or underestimate the magnitude between exposure and disease.

correct

1) Studies found that smoking is

a risk factor for ischemic heart

disease. However, most smokers

are coffee drinkers so I can't tell

if IHD is caused by smoking or

coffee

3rd factor confounding

• Real association could be:

• a.Causal b.Non causal ( Confounding)

• Real association means

• Real association means an association which is present in a population from which the sample is drawn. ( in sample + population)

A common factor is associated with both the potential risk factor and disease and can overestimate or underestimate the magnitude between exposure and disease.

confounding factor

2) Studies found an association between

contraceptive pills and myocardial

3infarction. An elderly woman on

contraceptive pills was diagnosed with

MI but it's not clear if it was caused by

the medication or her age

confounding

3) Studies found an

association between alcoho

consumption and

esophageal cancer, but

most alcohol drinkers are

Prof.

smokers

confofnding facor

Criteria of Causality

• Criteria of causality is attributed to Austin Bradford Hill:
  - 1- Strength of association

• 2- Consistency of results

• 3-Time sequence or temporal relationship

• 4-Dose response relationship

• 5- Specificity

• 6- Biological plausibility (causal mechanism)

• 7- Coherence (with other evidence)

• 8- Experimental evidence

• 9- Analogy (effect of similar factors)

correct

1- Strength of association has to be?

significant and strong

The stronger the association (i.e., larger the effect size – RR, OR .. Etc) between exposure and disease, the higher the likelihood of presence of a causal relationship.

corrrect

The stronger the association (i.e., larger the effect size – RR, OR .. Etc) between exposure and disease, the _________ the likelihood of presence of a causal relationship.

The stronger the association (i.e., larger the effect size – RR, OR .. Etc) between exposure and disease, the higher the likelihood of presence of a causal relationship.

Relative risk (RR) of > 2

or (Odds ratio) OR > 4

implies a strong association

correct

Relative risk (RR) of > ____

or (Odds ratio) OR > _____

implies a strong association

Relative risk (RR) of > 2

or (Odds ratio) OR > 4

implies a strong association

Relative risk (RR) of > 2

or (Odds ratio) OR > 4

implies a ___________ association

Relative risk (RR) of > 2

or (Odds ratio) OR > 4

implies a strong association

Consistency of results

→ all studies have similar or same results

• The same result would be observed within the same or between different types of studies

Example: all studies found that there is a direct association between smoking and lung cancer (cohort and case control studies in different populations).

Consistency of results:

Consistency of results:
all results in same study → are same

Example: all studies found that there is a direct association between smoking and lung cancer (cohort and case control studies in different populations).

f 10 scientist did a research on how can smoking cause IHD and all of them had the same result, then I can call it a causal relationship

Consistency of results:

Time sequence or temporal relationship:

• Risk factor or cause must (before)precede the disease.

• Exposure → Disease E → D

• This can be determined in cohort or experimental studies

• This can't be determined from cross-sectional or case-control studies (problem of temporality).

CORREC

Time sequence or temporal relationship: MEANS?

• Risk factor or cause must (before)precede the disease.

• Exposure → Disease E → D

CORREDT

Time sequence or temporal relationship CAN BE DTERMINED BY?

• This can be determined in cohort or experimental studies

Time sequence or temporal relationship CAN NOT BE DETERMINED BY?

• This can't be determined from cross-sectional or case-control studies (problem of temporality).

Dose-response relationship:

The risk of disease is greater with exposure to higher dosis of the risk factor

The risk of disease is greater with exposure to higher dosis of the risk factor

Dose-response relationship:

relationship between Number of packs of cigarettes per day and death from lung cancer per 100,000

Dose-response relationship:

In this graph we can see that every time we increase the number of cigarettes ( risk factor ) the risk of lung cancer ( disease ) also increases

Dose-response relationship:

Specificity:

• • The outcome (disease) should be specific to a particular risk factor.

• • The specificity is more applicable in infectious diseases than in non-infectious diseases.

• • E.g. specificity in infectious disease: tubercle bacilli is a specific cause of tuberculosis.

• • Not applicable for non-communicable diseases because there are usually multiple risk factors (causes) of the disease (multi-factorial).

CORRECT

Coherence with existing knowledge:

• Should make biological sense and should not conflict with other theories

: 1.Evidence supported by cellular and sub-cellular mechanisms.

2.Evidence from correlational studies.

3.Evidence obtained through exclusion of other hypotheses.

RIGHT

Example: Cigarette Smoking and Lung Cancer

• Strength of association:

• • In 3 studies the death rates from lung cancer were higher among smokers than non smokers.

• RR is about 20 – 40 times higher among smokers than among non-smokers

In 3 studies the death rates from lung cancer were higher among smokers than non smokers. RR is about 20 – 40 times higher among smokers than among non-smokers

Q) Which causal criteria does this represent ?

Strength of association:

Dose response relationship :

Studies revealed that the risk of death from lung cancer increases with larger amounts of cigarettes smoked.

Studies revealed that the risk of death from lung cancer increases with larger amounts of cigarettes smoked.

Dose response relationship

The three different studies that has been conducted show similar results. Which causal criteria does this represent?

Consistency of association.

) Consistency of association.

The three different studies that has been conducted show similar results. Which causal criteria does this represent?

Consistency of association:

• The association between cigarettes smoking and development of lung cancer were found in several countries and different groups.

• Different study designs found the same relation:

1- Prospective cohort studies 2- Case-control studies

orret

Consistency of association:

Consistency of association:

• The association between cigarettes smoking and development of lung cancer were found in several countries and different groups.

• Different study designs found the same relation:

like what studies?

• Different study designs found the same relation:

1- Prospective cohort studies 2- Case-control studies

Temporal relationship:

• Lung cancer tends to develop in life many years after starting of smoking.

Exposure precedes the disease. E → D

• Other studies reported that death rate for lung cancer decrease with the increase in the duration of smoking cessation.

• Lung cancer tends to develop in life many years after starting of smoking.

Exposure precedes the disease. E → D

Temporal relationship:

Coherence with existing knowledge:

• Cigarette smoke contains a large number of carcinogens.

• • The increase in use of cigarettes preceded the onset of lung cancer by about 30 years.

• • Male/ female difference in lung cancer death rate are due to differences in smoking.

right