what is a cause? epidemiology matters: a new introduction to methodological foundations chapter 7

What is a cause?

Epidemiology matters: a new introduction to methodological foundationsChapter 7

2Epidemiology Matters – Chapter 1

Seven steps

1. Define the population of interest2. Conceptualize and create measures of exposures and health

indicators3. Take a sample of the population4. Estimate measures of association between exposures and health

indicators of interest

5. Rigorously evaluate whether the association observed suggests a causal association

6. Assess the evidence for causes working together7. Assess the extent to which the result matters, is externally valid, to

other populations

Epidemiology matters – Chapter 7 3

“for now, I will stretch out mine hand, that I may smite thee and thy people with pestilence”

God, from Exodus (9:14)

Early conceptions of cause

4Epidemiology matters – Chapter 7

1. A motivating example

2. What is a cause?

3. Disease causation – marble game

4. Disease causation – time and space

5. Public health implications

6. Disease causation in a non-deterministic world

7. Epidemiology and probability

8. Summary



2. What is a cause?






8. Summary


Motivating examples The Uncle Joe example “My mother smoked in pregnancy - no one knew the

health dangers back then - and I’m just fine. All of these warnings about smoking during pregnancy are overdone.”

If one person smoked in pregnancy with no adverse consequences for their offspring, can we conclude that smoking in pregnancy is not harmful?

ExposedUnexposed


ExposedUnexposed Unexposed with disease

Exposed with disease



Epidemiologists understand disease causationto be a multifactorial process



2. What is a cause?






8. Summary


Example: A man’s depression

Man developed depression

Personal history: Born Boise, Idaho 1965

Parents working class

Youngest of 5 children

Trouble with the law growing up

Married at 25, divorced 4 years ago

Lost job 2 years ago

What are the causes of the man’s depression?


Would the depression have occurred if we kept everything

about his life the same, but changed one detail?

1. If everything about the man stayed the same, but the plant

had not closed, would the depression have occurred?

2. If everything about the man had stayed the same, but the

divorce had not occurred, would the depression have

occurred?

3. If everything about the man had stayed the same, but his

parents had more resources in his childhood, would the

depression have occurred?


Counterfactual

The counterfactual is the condition that is counter to the fact

A factor is a cause if the outcome would not have occurred in the absence of that factor, holding all other things constant, including space and time.


What is a cause?


2. What is a cause?







Non-diseased Diseased


Example: Individual diabetes

Combination of causes in individuals

Person 1. Obese weight, lack of preventive care, diabetes family history, 20 pack-years smoking

Person 2. Poor nutritional education, diabetes family history, high blood pressure, advanced age

Person 3. Obese weight, diabetes family history, high blood pressure


Example: Individual diabetes

Is each factor necessary? Is each factor sufficient?

For each individual case, all component causes are necessary for that individual, but none are sufficient.


Causes and the individual marble game

Each marble is a component cause

But, one marble is rarely sufficient to cause disease

A particular marble set can be a sufficient cause for disease

There can be more than one marble set that become a

sufficient cause

Causes are rooted in a counterfactual definition, each

marble is a necessary cause of disease for that particular

sufficient cause set


Necessary if all cases of disease require the cause in order

for disease to onset

Sufficient if all individuals exposed to the cause will acquire

the disease

Causes can be necessary and sufficient

unnecessary but sufficient

necessary but insufficient

unnecessary and insufficient

Necessary and sufficient causes for populations

Epidemiology matters – Chapter 7

Necessary and sufficient causesTrisomy 21 and Down Syndrome

All individuals with three copies of the 21st chromosome will evidence Down Syndrome. Trisomy 21 is thus sufficient for DS.All individuals with Down syndrome have three copies of the 21st chromosome. Trisomy 21 is thus necessary for DS.

21


Necessary but insufficient causesAlcohol consumption and alcoholism

Not all individuals who consume alcohol will develop alcoholism. Alcohol consumption is thus insufficient for alcoholism.However, all individuals with alcoholism will have consumed alcohol. Alcohol consumption is thus necessary for alcoholism.

22

Unnecessary but sufficient causeHysterectomy and pregnancy prevention

All women who have a hysterectomy are unable to become pregnant. Hysterectomy is thus sufficient for pregnancy prevention.Not all pregnancies are prevented through hysterectomy. Hysterectomy is thus unnecessary for pregnancy prevention.


Unnecessary and insufficient causeSmoking and lung cancer

Not all individuals who smoke will develop lung cancer. Smoking is thus insufficient to cause lung cancer.Not all lung cancer cases are smokers. Smoking is thus unnecessary to cause lung cancer.


25


2. What is a cause?






8. Summary


26

Collecting marbles across the life course

People can accumulate marbles At birth In adolescence During young adulthood At older adulthood At one point vs. slow accumulation over time


Collecting marbles across the life course: Example

Collecting throughout life tobacco smoke in utero chronic poverty, chaotic

home environment in childhood

cigarette smoking starts in adolescence

poor nutrition in adulthood

Infancy

Childhood

Adolescence

Adulthood]


28

Marbles are not independent

Shared across individuals One person’s marble collection may influence

another person’s marble collection Example, person-to-person infectious disease

transmission


Examples of shared marbles

Unhealthy food environment Community violence Social norms around substance use and cigarette

smoking; e.g., adolescents are more likely to begin smoking if an influential peer begins smoking

Policies and laws managing access to quality health care


Person 1 Person 2

Person 3 Person 4

Sharing marble exposures


Person 1 Person 2

Person 3 Person 4



34

Person 1 Person 2

Person 3 Person 4



35

Person 1 Person 2

Person 3 Person 4



36

Summary: marble analogy

Individuals share marbles and transmit marbles from

one space to another

Within each person’s space, there remains a

complete set of marbles that is necessary to cause

disease.

One person’s complete set of marbles may differ

from another persons


37

Summary: causes in time and space

At population level need to understand exposure to

unhealthy environments and transmission of disease

to understand and intervene to prevent adverse

health conditions

By identifying marbles that are common across many

marble spaces, we can identify the exposures and

environments for intervention and prevention

efforts.


38


2. What is a cause?






8. Summary


39

Public health implications

Marble example, individual

Each individual’s set of marbles that caused disease may be unique,

with or without overlap across individuals

Each marble was necessary for that person to develop the disease

when and how he or she did

Marble example, population

Epidemiologists look for the ‘marbles’ that are most common across

individuals with disease compared to those without disease

Preventing any one of the marbles can prevent disease in that

individual

Prevent common marbles can prevent more disease in more

individualsEpidemiology matters – Chapter 7

40

Exposure combinations and disease causation

Public health target?

Which cause should we try and prevent?Epidemiology matters – Chapter 7

41

Public health target?

Remove causeW

Person 1 and 3 savedX

Person 1 and 2 savedY

Person 1, 2, and 3 savedZ

Person 2 and 3 savedTarget cause Y for prevention to save most people



Summary: public health implications

At the population level causes may be necessary

and/or sufficient, but need not be either

Multifactorial and complex diseases are often caused

by many factors necessary in at least one person

Identifying factors, i.e., component causes, common

to most individuals has the greatest impact on

reducing disease for largest amount of people



2. What is a cause?






8. Summary


From marble space to probability Process of disease development may begin in utero

and continue until the moment that the disease occurs

Often many causes must align for a disease to occur in an individual


Example: Smoking and lung cancer Smoking is not sufficient to cause lung cancer; smoking

must act with other causes – i.e. an individual smokes, works in a occupation with a high degree of exposure to asbestos, has a genetic predisposition to develop lung cancer

Another person who develops lung cancer may have a different constellation of causes

Causes can be shared across people or be unique to a certain person


From marble space to probability

The idea that many causes must accumulate through

the life course before the disease manifests is, in

epidemiology, expressed as the concept of interaction

That is, if seven marbles are all necessary to cause

disease in an individual, then all of these marbles

interact with each other

By preventing exposure to even one marble, the

disease will not occur


Example A. Diet and phenylketonuria (PKU)

Disorder: PKU, an inability to process phenylalanine (amino acid); if untreated results in altered appearance, hyperactivity, mental retardation, and seizures

All PKU patients have specific maternal and paternal genetic sequence alone will not cause PKU

What type of cause is genetic sequence?


Diet and phenylketonuria (PKU)

Disorder: PKU, an inability to process phenylalanine (amino acid); if untreated results in altered appearance, hyperactivity, mental retardation, and seizures

All PKU patients have specific maternal and paternal genetic sequence alone will not cause PKU

What type of cause is a PA-rich diet?


Diet and phenylketonuria (PKU)

Only genetic + PA-rich diet

= PKU manifestation

Example B. Causes of obesity

Genetic variants involved in the process of increasing and maintaining high weight

In utero environment on obesity in childhood and adulthood

Childhood factors including food insecurity, socio-economic position, availability of healthy food and food cost

Health behaviors including high consumption of sugar-sweetened beverages

Example B. Causes of obesity

Potential causal mechanism may be

High consumption of sugar sweetened beverages + low

physical activity + genetic predisposition = obesity

There is no single cause of obesity, we need to

conceptualize causes as interacting


Interaction: many causes must accumulate through the life course before the disease manifests

Necessary but insufficient causes interact for disease to manifest in an individual

Multifaceted disease causation requires many component causes interacting

Summary: Causation in nondeterministic world



2. What is a cause?






8. Summary


Epidemiology is probabilistic

Deterministic “… occurrences … are causally determined by preceding events or natural laws”*

Probabilistic “of, relating to, or based on probability”*

Epidemiology is probabilistic Considering component causes Often have not identified all of the causal partners for a

sufficient cause*Definitions: http://www.merriam-webster.com/dictionary/determinism and http://www.merriam-webster.com/dictionary/probabilistic accessed on 9/10/2013

http://www.merriam-webster.com/dictionary/determinism

http://www.merriam-webster.com/dictionary/determinism

http://www.merriam-webster.com/dictionary/probabilistic

http://www.merriam-webster.com/dictionary/probabilistic


Hypothetical example X,Y, ZReality

Component Causes # People with component causes Probability of disease

X Y Z

✔ ✔ ✔ 10 1

✔ ✔ 8 0

✔ ✔ 11 0

✔ ✔ 5 0

✔ 9 0

✔ 3 0

✔ 12 0

42 0

X, Y, and Z are all necessary and insufficient causes of disease in an individual


Hypothetical example X,Y, Z

Component Causes

# People with component causes

Probability of disease

X Y Z

✔ ✔ ✔ 10 1

✔ ✔ 8 0

✔ ✔ 11 0

✔ ✔ 5 0

✔ 9 0

✔ 3 0

✔ 12 0

42 0

X, Y, and Z are all necessary and insufficient causes of disease in an individualWhat is prevalence of disease, given exposure to X?N exposed to X = 38 (10+8+11+9)N exposed to X with disease = 10Prevalence of disease : P(D|X)=10/38 = 0.26

Component Causes

# People with component causes


X Y Z

✔ 10 1

✔ 8 0

✔ 11 0

5 0

✔ 9 0

3 0

12 0

42 0

RealityMeasured


Transition from knowing all component causes to just those measured takes epidemiologist from

deterministic to probabilistic


Hypothetical example A,B,C, X,Y, ZReality

Component Causes # People with component causes


A B C X Y Z

✔ ✔ ✔ ✔ ✔ ✔ 5 1

✔ ✔ ✔ 6 1

✔ ✔ ✔ 10 1

✔ ✔ 20 0

✔ ✔ 15 0

✔ ✔ 14 0

✔ ✔ 12 0

✔ ✔ ✔ ✔ 18 0

A,B, and C OR X, Y, and Z are all necessary and insufficient causes of disease in an individual


Hypothetical example A,B,C, X,Y, Z



A B C X Y Z

✔ ✔ ✔ ✔ ✔ ✔ 5 1

✔ ✔ ✔ 6 1

✔ ✔ ✔ 10 1

✔ ✔ 20 0

✔ ✔ 15 0

✔ ✔ 14 0

✔ ✔ 12 0

✔ ✔ ✔ ✔ 18 0

RealityMeasured

A,B, and C OR X, Y, and Z are all necessary and insufficient causes of disease in an individualWhat is probability of disease, given X?P(D|X): 15/47, or 0.32 or 32%



A B C X Y Z

✔ 5 1

6 1

✔ 10 1

✔ 20 0

15 0

14 0

✔ 12 0

18 0

ExposedUnexposed Unaffected with disease

Affected with disease

Remember this?


Higher ‘risk’ of disease given exposure to X

Probabilistic or deterministic?


1. Disease causation is a multifactorial process

2. A cause, necessary or sufficient

3. Disease causation, the marble game

4. Disease causation, time and space

5. Public health implications, identifying component causes for most individuals

6. Disease causation in a non-deterministic world - interaction

7. Epidemiology is probabilistic

8. Summary


Seven steps

1. Define the population of interest2. Conceptualize and create measures of exposures and health

indicators3. Take a sample of the population4. Estimate measures of association between exposures and health

indicators of interest

5. Rigorously evaluate whether the association observed suggests a causal association

6. Assess the evidence for causes working together7. Assess the extent to which the result matters, is externally valid, to

other populations


epidemiologymatters.org

what is a cause? epidemiology matters: a new introduction to methodological foundations chapter 7

Documents

disease causation time

disease causation marble

result matters

depression man

public health implications

health indicators

health dangers

deterministic world