Week 4: Association and Causation in Epidemiology
Learning Objectives:
Description of the Bradford Hill Causation Criteria
Critical Evaluation of Significance and Application of the Bradford Hill Criteria in
Contemporary Epidemiology
Interpretation of Cause Effect Relationships
Contents
Introduction ............................................................................................................................................ 1
Introduction to the Bradford-Hill criteria ............................................................................................... 3
Strength of Association ....................................................................................................................... 4
Consistency of data ............................................................................................................................. 5
Specificity ............................................................................................................................................ 5
Temporality ......................................................................................................................................... 5
Dose-response .................................................................................................................................... 6
Biological plausibility........................................................................................................................... 6
Coherence ........................................................................................................................................... 6
Experimental evidence........................................................................................................................ 7
Analogy ............................................................................................................................................... 7
Critical evaluation of the Bradford Hill criteria ....................................................................................... 7
Practical example of the Bradford-Hill criteria ..................................................................................... 10
References ............................................................................................................................................ 11
Introduction
Cause and effect is often the next step in science, after a discovery of patterns or events that
occur together with regularity. A search for the underlying cause of a phenomenon has sparked
some of the most compelling and productive scientific investigations. Any hypothesis that A
causes B requires an explanation mechanism for the chain of interactions that connect A and
1
, B. For example, the notion that diseases can be transmitted by a person’s touch was initially
treated with scepticism. Today it is well understood that infectious diseases can be transmitted
by the passing of bacteria or viruses between an infected person and another (San Diego County
Office of Education 2020). A principal aim of epidemiology is to uncover such causal
connections, often with the hope that understanding the mechanisms will enable predictions
and the design of preventive measures and treatments.
We have seen in Weeks 2 and 3 that epidemiological studies either observational or
experimental examine association between exposure and outcome.
Experimental studies are rarely available because it is neither ethical nor practical to design
experimental study to examine, for instance, the following associations as the outcome
measure: harms caused by environmental exposure to chemicals (e.g. carcinogenicity of
quinolone) or assessing harms caused by infectious diseases (e.g. assessing the teratogenicity
of Zika) (Wiliamson 2018).
On the other hand, most epidemiological studies are by nature observational, and, as we have
already seen, they apply estimates of association such as: prevalence for cross- sectional
studies, the odds ratio for case- control studies and incidence rate and relative risk for cohort
studies. However, because observational studies involve no control or manipulation, i.e. do not
interfere into uncontrolled conditions of everyday real-life, a number of possible explanations
for an observed (and estimated) association need to be carefully considered before one can
infer a cause-effect relationship. As such, an advantage of observational design is that people
can be observed in their natural environment, however the drawback is that it is usually difficult
to untangle the often complex situation where exposure can be difficult to define (e.g. air
pollution, socioeconomic status or duration and frequency of an exposure etc.) as well as to
2
Learning Objectives:
Description of the Bradford Hill Causation Criteria
Critical Evaluation of Significance and Application of the Bradford Hill Criteria in
Contemporary Epidemiology
Interpretation of Cause Effect Relationships
Contents
Introduction ............................................................................................................................................ 1
Introduction to the Bradford-Hill criteria ............................................................................................... 3
Strength of Association ....................................................................................................................... 4
Consistency of data ............................................................................................................................. 5
Specificity ............................................................................................................................................ 5
Temporality ......................................................................................................................................... 5
Dose-response .................................................................................................................................... 6
Biological plausibility........................................................................................................................... 6
Coherence ........................................................................................................................................... 6
Experimental evidence........................................................................................................................ 7
Analogy ............................................................................................................................................... 7
Critical evaluation of the Bradford Hill criteria ....................................................................................... 7
Practical example of the Bradford-Hill criteria ..................................................................................... 10
References ............................................................................................................................................ 11
Introduction
Cause and effect is often the next step in science, after a discovery of patterns or events that
occur together with regularity. A search for the underlying cause of a phenomenon has sparked
some of the most compelling and productive scientific investigations. Any hypothesis that A
causes B requires an explanation mechanism for the chain of interactions that connect A and
1
, B. For example, the notion that diseases can be transmitted by a person’s touch was initially
treated with scepticism. Today it is well understood that infectious diseases can be transmitted
by the passing of bacteria or viruses between an infected person and another (San Diego County
Office of Education 2020). A principal aim of epidemiology is to uncover such causal
connections, often with the hope that understanding the mechanisms will enable predictions
and the design of preventive measures and treatments.
We have seen in Weeks 2 and 3 that epidemiological studies either observational or
experimental examine association between exposure and outcome.
Experimental studies are rarely available because it is neither ethical nor practical to design
experimental study to examine, for instance, the following associations as the outcome
measure: harms caused by environmental exposure to chemicals (e.g. carcinogenicity of
quinolone) or assessing harms caused by infectious diseases (e.g. assessing the teratogenicity
of Zika) (Wiliamson 2018).
On the other hand, most epidemiological studies are by nature observational, and, as we have
already seen, they apply estimates of association such as: prevalence for cross- sectional
studies, the odds ratio for case- control studies and incidence rate and relative risk for cohort
studies. However, because observational studies involve no control or manipulation, i.e. do not
interfere into uncontrolled conditions of everyday real-life, a number of possible explanations
for an observed (and estimated) association need to be carefully considered before one can
infer a cause-effect relationship. As such, an advantage of observational design is that people
can be observed in their natural environment, however the drawback is that it is usually difficult
to untangle the often complex situation where exposure can be difficult to define (e.g. air
pollution, socioeconomic status or duration and frequency of an exposure etc.) as well as to
2
