Concepts of Causation

1
Concepts of Causation

• Introduction to Epidemiology
• Fall 2002

2
Epidemiologic Reasoning

• Derive inferences regarding possible causal
  relationships
• Determine whether these relationships are
  spurious or true
• Today we
• discuss causal relationships
• introduce threats to validity
• discuss play of chance (statistical association)

3
Epidemiologic Study ofdisease etiology

• unplanned or natural experiments
• residents of Bhopal, India exposed to toxic
  chemicals
• residents of Hiroshima and Nagasaki exposed to
  atomic bomb radiation in 1945

4
Sequence of studies in humans
5
Study Designs

• Descriptive Studies
• Population
• correlation or ecologic studies
• Individual
• Case Reports
• Case Series
• Cross Sectional Survey

6
Study Designs

• Analytic Studies
• Observational Studies
• Case-Control or Case-Comparison
• Cohort Studies
• Intervention Studies
• Clinical Trials

7
APPLICATIONS

• Study of Risk to Individuals
• Observational studies
• Case-control study design
• Cohort study design
• Clinical/Policy decision criteria of causality
• Intervention study design

8
Causal Factor or Risk Factor

• A cause is a factor (or member of a set of
  factors) which results in a sequence of events
  that eventually result in an effect.
• Causation is often unknown
• disease varies by category of risk factor
• risk factor precedes onset of disease
• observed association is not due to error

9
Risk Factor or Cause
10
Risk Factor or Cause

• Is current oral contraceptive use associated with
  increased risk of myocardial infarction in
  premenopausal female nurses?

11
Risk Factor or Cause

• Does use or oral contraceptives increase the risk
  of developing urinary tract infections among
  women aged 16-49.

12
Risk Factor or Cause

• Are women who have had one MI more likely to be
  current OC users than women who have not had an
  MI?

13
Risk Factor or Cause

• Is CHD mortality (age-adjusted rates by state)
  associated with per capita cigarette sales?

14
(No Transcript)
15
The epidemiologic triangle
Host Agent Environment
Host
VECTOR
Agent
Environment
16
Henle-Koch Postulates

• First formulated by Henle and
• adapted by Robert Koch in 1877
• to described the discovery of the tubercule
  bacillus.

17
These postulates should be met

• before a causative relationship can be accepted
  between a disease agent and the disease in
  question
• The agent must be shown to be present in every
  case of the disease by isolation in a pure
  culture.
• The agent must not be found in cases of other
  disease.
• Once isolated, the agent must be capable of
  reproducing the disease in experimental animals.
• The agent must be recovered from the experimental
  disease produced

18
Difficult Issues / Inferring

• How do we prove anything
• Henle-Koch postulates used to prove causation of
  microorganisms in pathogenesis of an infectious
  disease
• Agent must be present in every case
• How to apply to cardiovascular diseaseoverweight,
  physically inactive, smoking, high blood
  pressure, elevated cholesterol
• Agent should occur in no other disease (one
  agent-one disease model)what about cigarette
  smoking?

19
Difficult Issues / Inferring

• Exposure of healthy subjects to suspected
  agentsethical?
• must rely on epidemiologic evidence from
  observational studies
• even agent, host, environmental model not
  sufficient

20
Henle-Koch Postulates

• Useful in infectious diseases
• Not useful with complicated chronic diseases
• Heart disease
• Diabetes
• Cancer
• Violence

21
Bradford Hill Criteria (1965)

• criteria for assessing causality
• ?Strength ?Consistency
• ?Specificity ?Temporality
• ?Plausibility ?Coherence
• ?Dose Response ?Analogy
• ?Experimental evidence

22
Bradford Hill Criteria

• Hill stated
• None of my criteria can bring indisputable
  evidence for or against the cause-and-effect
  hypothesis
• None can be required as sufficient alone

23
DETERMINATION OF CAUSATION

• One way of determining causation is personal
  experience by directly observing a sequence of
  events.

24
Personal Experience / Insufficient

• Long latency period between exposure and disease
• Common exposure to the risk factor
• Small risk from common or uncommon exposure
• Rare disease
• Common disease
• Multiple causes of disease

25
OBSERVATIONAL STUDIES

• The general QUESTION
• Is there a cause and effect relationship between
  the presence of factor X and the development of
  disease Y?

26
OBSERVATIONAL STUDIES

• The answer is made by inference and relies on a
• summary of all valid evidence.
• Temporal sequence
• Strength of the association
• Dose-response
• Replication of findings (consistency)
• Biologic credibility
• Consideration of alternate explanations
• Cessation of exposure (dynamics)
• Specificity

27
SMOKING AND LUNG CANCER

• 1. Strength of Association
• The relative risks for the association of smoking
  and lung cancer are in the order of
• 2. Biologic Credibility
• The burning of tobacco produces carcinogenic
  compounds which are inhaled and come into contact
  with pulmonary tissue.

28
SMOKING AND LUNG CANCER

• 3. Replication of findings
• The association of cigarette smoke and lung
  cancer is found in both sexes in all races, in
  all socioeconomic classes, etc.
• 4. Temporal Sequence
• Cohort studies clearly demonstrate that smoking
  precedes lung cancer and that lung cancer does
  not cause an individual to become a cigarette
  smoker.

29
SMOKING AND LUNG CANCER

• 5. Dose-Response
• The more cigarette smoke an individual inhales,
  over a life-time, the greater the risk of
  developing lung cancer.
• 6. Dynamics (cessation of exposure)
• Reduction in cigarette smoking reduces the risk
  of developing lung cancer.

30

• Smoking is cited as a cause of lung cancer,
  however. . .
• . . . smoking is not necessary (is not a
  prerequisite) to get lung cancer. Some people
  get lung cancer who have never smoked.
• . . . smoking alone does not cause lung cancer.
  Some smokers never get lung cancer.
• Smoking is a member of a set of factors (i.e.,
  web of causation) which cause lung cancer.
• The identity of all the other factors in the set
  are unknown. (One factor in the web of causation
  is probably genetic susceptibility.)

31
Causation

• The world is richer in associations than
  meanings, and it is part of wisdom to
  differentiate the two.
• John Barth
• All scientific work is incomplete whether it
  be observational or experimental. All scientific
  work is liable to be upset or modified by
  advancing knowledge. That does not confer upon
  us the freedom to ignore the knowledge we already
  have, or to postpone the action it appears to
  demand at a given time. Bradford Hill (1965)

32
Nature of Evidence

• 1. Temporal Sequence
• exposure precede disease
• 2. Strength of Association
• significant high risk
• 3. Dose-Response
• higher dose exposure, higher risk

33
Nature of Evidence

• 4. Replication of Findings
• consistent in populations
• 5. Biologic Credibility
• exposure linked to pathogenesis
• Consideration of alternative explanations
• the extent to which other explanations have been
  considered.

34
Nature of Evidence

• 7. Cessation of exposure (Dynamics)
• removal of exposure reduces risk
• 8. Specificity
• specific exposure is associated with only one
  disease

35
necessary / sufficient
Disease Not Present
Disease Present
Disease Present
A is necessary since it appears in each
sufficient causal complex A is not sufficient
36
necessary / sufficient

• necessary and sufficient
• the factor always causes disease and disease is
  never present without the factor
• most infectious diseases
• necessary but not sufficient
• multiple factors are required
• cancer
• sufficient but not necessary
• many factors may cause same disease
• leukemia
• neither sufficient nor necessary
• multiple cause

37
necessary / sufficient

• Component cause
• any one of a set of conditions which are
  necessary for the completion of a sufficient
  cause (piece of pie)
• Necessary cause
• a component cause that is a member of every
  sufficient cause

38
necessary / sufficient

• Few causes are necessary and sufficient
• HPV is necessary for cervical cancer but not
  sufficient because not every woman infected with
  HPV develops cervical cancer.

39
necessary / sufficient

• Few causes are necessary and sufficient
• High cholesterol is neither necessary nor
  sufficient for CVD because many individuals who
  develop CVD do not have high serum cholesterol
  levels

40
H. pylori

• Temporal relationship
• 11 of chronic gastritis patients go on the
  develop duodenal ulcers over a 10-year period.
• Strength
• H. pylori is found in at least 90 of patients
  with duodenal ulcer
• Dose response
• density of H.pylori is higher in patients with
  duodenal ulcer than in patients without
• Consistency
• association has been replicated in other studies

41
H. pylori

• Biologic plausibility
• originally no biologic plausibility
• then H. pylori binding sties were found
• know H. pylori induces inflammation
• Specificity
• prevalence of H. pylori in patients with duodenal
  ulcers is 90 to 100
• Consistency with other knowledge
• prevalence is the same in males and females

42
Establishing Causality

• trials
• randomized, double-blind, placebo-controlled with
  sufficient power and appropriate analysis
• cohort studies case-control studies
• hypothesis specified prior to analysis
• case-series
• no comparison groups

43
Accuracy and Sources of Error

• Purpose of epidemiologic study
• To estimate the effect of an exposure on an
  outcome
• Main objective
• To measure the exposure and outcome accurately
• That is, to measure without error

44
Evaluate Validity

• Absence of systematic errors
• Findings represent the study sample
• Findings are generalizable to larger populations
• Internal validity is the primary objective
• Without internal validity
• there is no reason to generalize

45
Threats to validity

• Internal validity
• do these results represent what is really
  happening in the study population.
• are the results due to
• Bias
• Confounding
• Chance

46
Evaluation of findings

• Bias
• Confounding
• Play of Chance
• Frequency Measures
• Prevalence
• Incidence
• Measures of Association
• Causal Inference

47
BIAS

• systematic errors in collecting or interpreting
  data such that there is deviation of results or
  inferences from the truth.
• selection bias noncomparable criteria used to
  enroll participants.
• information bias noncomparable information is
  obtained due to interviewer bias or due to recall
  bias

48
BIAS

• Bias has to do with research design
• Bias results from systematic flaws in
• study design
• data collection
• analysis
• interpretation

49
BIAS

• Two major types to consider
• selection bias non-comparable
• criteria used to enroll participants
• information bias non-comparable
• information obtained due to
• interviewer or recall bias

50
Confounding

• a mixing of effects
• between the exposure, the disease, and other
  factors associated with both the exposure and the
  disease
• such that the effects the effects of the two
  processes are not separated.

51
Confounding

• Confounding results when the effect of an
  exposure on the disease (or outcome) is distorted
  because of the association of exposure with other
  factor(s) that influence the outcome under study.

52
Confounding Biomedical Bestiary Michael,
Boyce Wilcox, Little Brown. 1984
Observed association, presumed causation
Gambling
Cancer
Smoking, Alcohol, other Factors
Unobserved association
True association
53
Play of Chance

• Measures of Association
• Statistical Issues

54
Statistical Issues

• The evaluation of the role of chance is done in 2
  steps
• Estimate the magnitude of the association.
• Hypothesis testing
• Calculate a test statistic,
• obtain a p value or confidence interval

55
Statistical Issues

• We have to remember that epidemiologic studies
  draw inferences about the experiences of an
  entire population based on an evaluation of only
  a sample.

56
Magnitude of Association

• Epidemiologist tend to view cause and effect as
  binary variables
• Either you are exposed (or diseased)
• Are you arent exposed (or diseased)
• How we measure these variables can have a
  profound influence on our results

57
Statistical Issues

• What do we mean by chance and how does this
  relate to determining a true association
• Where do we start?

58
Magnitude of Association

• THE 2 x 2 table
• Disease on top exposure on the left

59
Measures of Disease Association
60
COHORT STUDY

• Disease Occurrence Among Exposed
• Incidence (Ie) a / (ab)
• Disease Occurrence Among non-Exposed
• Incidence (Io) c / (cd)

61
COHORT STUDY

• Disease Occurrence Among Exposure and Non-Exposure

62
Magnitude of Association

• Risk Ratio
• Relative Risk Ie / Io
• Risk Difference excess risk of disease among
  exposures
• Attributable Risk Ie Io
• Attributable Risk (Ie Io) / Ie 100

63
Case-Control Study

• Odds of Exposed vs Non-Exposed Among Disease and
  Non-Disease Cases

64
Statistical Issues

• Evaluating chance
• This can be done by calculating a test statistic
  of the general format

65
Statistical Issues Primarily Sampling Issues

• p-value the probability of obtaining a sample
  showing an association of the observed size or
  larger by chance alone under the hypothesis that
  no association exists.
• Confidence interval a range of values that one
  can say, with a specific degree of confidence,
  contains the true population value.

66
Statistical Issues

• The p value indicates the possibility that
  findings at least as extreme as those observed
  were unlikely to have occurred by chance alone.

67
Statistical Issues

• A statistically significant finding does not mean
  that the results DID NOT occur by chance
• only that it is unlikely that the findings did
  occur by chance.
• A non-significant finding does not mean that
  there is not association
• only that it is highly unlikely that there is an
  association.

68
Statistical Issues

• More often in epidemiology we are examining
  discrete data
• the 2 x 2 table presents discrete data.
• Here we are testing whether the distribution of
  counts in the 4 cells is different than expected
  under the null hypothesis.

69
Statistical Issues

• All tests of statistical significance lead to a
• probability statement
• which is usually expressed as a p value

70
Statistical Issues

• But how do we determine the expected value for
  the cells of a
• 2 x 2 table?

71
Statistical Issues

• A probability of 0.05 is the level set for
  statistical significance
• this is the usual and arbitrary cutt-off
• If p lt0.05, we conclude that chance is an
  unlikely explanation for the finding
• these results would occur by change only very
  rarely
• The null hypothesis is rejected, and the
  statistical association is said to be significant

72
Statistical Issues

• If p gt0.05, we conclude that chance cannot be
  excluded as an explanation for the finding
• and we fail to reject the null hypothesis.
• That is
• these results are likely to occur by chance more
  than 5 of the time.

73
Statistical Issues

• No p value
• however small - completely excludes chance
• No p value
• however large - completely mandates chance

74
Statistical Issues

• p values only evaluate the role of chance
• they say nothing about other alternative
  explanations or about causality
• p values reflect the strength of the association
  and the study sample size

75
Statistical Issues

• A small difference may achieve statistical
  significance if the sample size is large
• A large difference may not achieve statistical
  significance if the sample size is too small

76
Statistical Issues

• We address these problems by calculating
  confidence intervals
• The confidence interval (CI) gives all the
  information of a p value
• PLUS the expected range of effect sizes

77
Statistical Issues

• Confidence Interval indicates the range within
  which the true magnitude of effect lies with a
  certain degree of assurance.
• The degree of assurance is defined by the p value
  you assign.

78
Statistical Issues

• If the null value is included in a 95 confidence
  interval
• then the corresponding p value is, by definition,
  greater than 0.05.
• What do I mean???

79
Statistical Issues

• If the null value is not included, the
  association is considered to be statistically
  significant.
• WHAT IS THE NULL VALUE???

80
Statistical Issues

• HOWEVER Before we get to the major result - we
  need to examine several issues
• 1. What was the question that this study
  intended to answer?
• 2. What were the methods used to answer this
  question?
• 3. Are there errors in the study design that
  might invalidate the results?

81
Statistical Issues

• Is chance a likely explanation for the results?
• Is selection bias a likely explanation for the
  results?
• Is information bias a likely explanation for the
  results?
• Are the authors conclusions reasonable in terms
  of the information presented?

82
Statistical Issues

• Test Based CI for either OR or RR
• NOTE variance for either RR or OR may be
  estimated using the chi-square test statistic.
  Miettinen, Am J Epidemiol 103226-235, 1976

83
Statistical Issues

• Taylor Series to estimate the lnOR variance
  Woolf, Ann Human Gen 19251-253, 1955

Note e is a function on you calculator. You
need a key marked ex and you enter the OR times e
raised to the power of the results between the
brackets .
84
Statistical Issues

• Taylor Series to estimate the lnRR variance
  Katz, Biometrics, 34469, 1973

Note e is a function on you calculator. You
need a key marked ex and you enter the OR times e
raised to the power of the results between the
brackets .