Causation in Medicine

Emory University - Rollins School of Public Health
Atlanta, Georgia

Instructor(s): Frumkin, Howard
Subject area: Health / Medicine
Department: Public Health
Level: Undergraduate Medical
Number of participants: 90
Duration of exercise: 90 minutes
Cost/equipment needed: Overhead Projector
Learning objective: Provide Information
Teaching style: Passive Learning

Please note that the copyright for this course project is retained by the instructor.

This lecture is presented to first year medical students at Emory University as part of their Patient/Doctor Course. The lecture introduces the concept of causation in medicine beginning with a discussion of the Henle-Koch postulates and the Bradford-Hill criteria, then moving to a discussion of the relationship between causation, probability, and certainty in medicine. The application of these concepts to the individual patient is discussed followed by the patient interview and case presentation. The class is divided into small groups to discuss the case and whether or not the illness was caused by the exposure in question (monoethanolamine in soldering). A list of suggested readings on causation, occupational asthma, and asthma in soldering is included.

The Patient/Doctor Course
Emory University School of Medicine
Spring, 1997

The overall goal of this session is to introduce the student to the concept of causation in medicine. By the end of the case presentation and discussion the students should be able to:

1. Discuss and critique two important schemes for inferring causation: the Henle-Koch postulates and the Hill criteria.

2. Discuss the relationship between causation and probability in medicine.

3. Discuss the application of these concepts to individual patients.

The Henle-Koch postulates and the Hill criteria

Often in medicine we are called on to make judgments as to whether an exposure causes a disease. Even in ancient times the concept of specific causes existed; disease-ridden ships were quarantined in ports and people with obvious communicable diseases were banned from cities. In the mid-19th century, as the germ theory of disease arose, Henle and his pupil Koch formulated postulates from which the inference could be made that a specific living organism caused a particular disease. In simplified form there were three:

1. The organism is always found with the disease.

2. The organism is not found with any other disease.

3. The organism, cultured from one with the disease and cultured through several generations, produces the disease.

In 1876 Koch demonstrated that anthrax met these criteria, and many other infectious diseases followed. The idea that a specific, identifiable agent could cause a specific disease was revolutionary, and paved the way for interventions such as vaccines.

Almost a century later, in 1964, the Surgeon General's report on smoking implicated tobacco as a cause of lung cancer. This ignited a tremendous controversy about whether such causal thinking could be applied to chronic diseases. The Surgeon General's report, and a soon published paper by Sir Austin Bradford Hill, advanced standards that could be used to judge when an association might be causal. Here they are, along with brief comments:

1. Strength. Hill meant that a strong association, such as a five- or tenfold increase in risk, is more likely to be causal than a weak association, such as a 10% increase in risk, because a weak association is more likely to be spurious, arising from bias, confounding, or chance. (However, a weak association does not rule out causality!)

2. Consistency. If the association is repeatedly observed in different populations in different settings, it is more likely to be causal than an isolated observation. (However, lack of consistency does not rule out a causal connection; some causes only work in certain circumstances, say in the presence of cofactors.)

3. Specificity. The idea here is that a cause should lead to a single effect, not multiple effects. This may have been a holdover from infectious disease thinking; it is clearly wrong in many other circumstances. For instance, smoking causes lung cancer, bladder cancer, emphysema, and heart disease.

4. Temporality. A cause must precede an effect in time. This one is true!

5. Biological gradient. This is the presence of a dose-response gradient; if more of a dose leads to more of an effect, this supports the idea of causality. A dose-response gradient is helpful, but its absence doesn't rule out causation (DES and vaginal adenocarcinoma, asbestos and mesothelioma) and its presence doesn't prove causation (since it may result from confounding or bias).

6. Plausibility. The idea of causation must be biologically plausible. This may be elusive because we hold many fixed ideas; many people doubted for years that peptic ulcer disease cold be infectious in origin!

7. Coherence. The idea of causation must accord with other observations. For example, as Hill wrote, a causal relationship between smoking and lung cancer was coherent with the observations that smokers had dysplasia of the bronchial epithelium, or that lung cancer was a predominantly male disease. However, the absence of coherent information does not rule out a causal relationship.

8. Experimental evidence. Supporting data from human or animals experiments, such a lung cancer in animals exposed to cigarette smoke, helps establish a causal relationship.

9. Analogy. For example, if thalidomide can cause birth defects, perhaps other drugs taken during pregnancy can also cause birth defects. Analogy can be helpful, although the help seems limited since anybody with a little creativity can probably dream up an analogy!

A more complex view of causation

Actually, the concept of causation is far more complex than suggested by these schemes. Consider just a couple of familiar examples (or develop your own in the class discussion).

An outcome may have many component "causes." Jim has a bad day at work and stops at the bar for a couple of drinks on the way home. By the time he finishes, a couple of hours later, he has had five drinks. He gets into his car, drives too fast, slips on an icy patch as he rounds a curve on the way home, spins into oncoming traffic, and collides with Mary's car. Mary is pregnant. In the collision she suffers an abdominal injury, and loses the baby. What caused Mary's pregnancy termination? Was it Joe's bad day? Was it Joe's irresponsible excess drinking? Was it the bartender who continued to serve Joe after he appeared drunk? Was it the weather, which left an icy patch on the road? Was it Joe's car, which did not have anti-lock brakes? Was it Mary's car, which did not have air bags? Was it Mary, who was not wearing her seat belt? If any one of these antecedent conditions had not been met, the tragic outcome might have been averted. Maybe your view of the cause depends on whether you are the highway engineer, the car designer, the police officer, or Mary's lawyer.

There may be several causal "pathways" to a disease. If there are many causal factors, there may be different combinations, each sufficient to cause a disease. For instance, what causes tuberculosis? A particular microorganism, immune deficiency, crowded living conditions, malnutrition, silicosis, and perhaps some genetic factors. In one case two of these causes -- the mycobacterium and immune deficiency -- may operate to cause a case of TB. In another case three of these causes -- the mycobacterium, silicosis, and malnutrition-may operate. One cause, the mycobacterium is necessary; various combinations that include the mycobacterium may be sufficient.

Thinking probabilistically about causation

In reality, what we often mean by causation is an increased probability. Cigarette smoking "causes" lung cancer, but many people who smoke do not develop lung cancer, and some lung cancer patients have never smoked. Cigarette smoking simply increases the probability that a person will develop lung cancer. Here, the probability is fairly high. Aspirin "causes" Reye's syndrome in children and certain tampons "cause" toxic shock syndrome , but here the probabilities are quite low.

The term risk factor is worth discussing here. We recognize obesity, hypercholesterolemia, diabetes, genetic predisposition, and other risk factors for myocardial infarction. Perhaps "risk factor" is a better term to use than "cause," especially if we use it carefully. Earlobe creases seem to be a risk factor for myocardial infarction, but are they causal?

The students should be reminded of how we identify risk factors. While the Henle-Koch postulates are fairly intuitive, epidemiologic studies of chronic diseases are not. An exposure is a risk factor for a disease if the relative risk associated with that exposure is elevated; this should be familiar from their epidemiology instruction. This is not an abstraction. Your mother will call to ask if she should stop drinking coffee, after hearing on the news that it causes pancreatic cancer. Your patients will call you to ask if they should change their blood pressure medicines, after hearing on the news that they cause heart attacks. You will find yourself constantly evaluating causal associations as a clinician.

Causation and the individual patient

One of the most challenging situations is deciding on causation in an individual case of disease. A patient with a disease may ask what caused the disease, a natural response to a life crisis ("Why me?). A patient may want to know what food caused her skin rash, so she can modify her diet accordingly. A patient may ask whether a workplace exposure caused her disease, wanting to know whether to change jobs or perhaps file a workers compensation insurance claim.

Here the concept of certainty should be discussed. We often operate with uncertainty in medicine, and although this is almost never comfortable, it need not stop us from reaching conclusions. We hospitalize patients when we think they are having heart attacks and we operate on patients when we think they have appendicitis, knowing that we will be wrong in a certain proportion of cases. In fact, reasonable medical certainty as defined by the courts means "more certain than not," which amounts to only 51% certainty!

Imagine a risk factor that triples the probability of developing a disease. If the risk factor is absent, two people out of 100 will become ill. If the risk factor is present, six people out of 100 will become ill. If one of those six people consults you, you won't know if he is one of the two who would have developed the disease anyway, or one of the four excess cases caused by the exposure. How do you decide what to tell the patient?

Today's case presentation

Today's patient developed asthma after working with chemicals known to cause asthma. Of course, asthma can develop without any particular exposures. The challenge is to answer the patient's questions: Did my job cause my asthma? Should I change my job? Should I collect workers compensation insurance?

Linda Phrampus, who has graciously agreed to participate in class, is a 51-year-old woman. Until the age of 45 she was healthy except for hypertension, an ovarian cyst, and allergy to enalapril. She smoked one half pack of cigarettes daily from age 30 to age 47. At age 40 she started working in the electronics industry, assembling computer components. She worked at three different companies, advancing to lead assembler and then to supervisor, during her first five years in the industry. Just before her 45th birthday she took a new position at a company in Roswell, where she was responsible for operating a wave soldering machine. She will describe the machine in some detail in class. Within weeks she developed watery eyes, a runny nose, a cough productive of clear sputum, slight wheezing, and shortness of breath. The symptoms were more severe during the day than at night, worsened from Monday through Friday and improved over weekends, and improved during a one-week vacation two years after she began the job. After three years at the job with continuing symptoms, she presented to Emory's Environmental and Occupational Medicine Program for evaluation.

Physical examination was unremarkable; her chest was clear. Pulmonary function testing showed low normal airflow and volume (FEV1 87% of predicted, FVC 83% of predicted) and hyperresponsive airways (FEV1 decreased 14% with saline inhalation, and returned to baseline promptly with bronchodilator). Specific challenge testing was not performed. Review of the chemicals she used at work revealed monoethanolamine (a reported cause of asthma among people who solder), several macromolecules such as guanidine HCl, and several respiratory irritants. The diagnostic question was whether she had occupational asthma.

A brief primer on occupational asthma

Occupational asthma is associated with a wide variety of workplace exposures, including:
1. metal salts, e.g. platinum, chrome, and nickel
2. wood and vegetable dusts, e.g. western red cedar, flour, coffee beans
3. pharmaceutical agents, e.g. antibiotics, cimetidine
4. industrial chemicals and plastics, e.g. toluene diisocyanate (TDI), acid anhydrides used in epoxy resins, ethylenediamine
5. biological enzymes, e.g. laundry detergents, pancreatic enzymes
6. animal and insect products

There are thought to be three distinct mechanisms responsible for occupational asthma:
1. Immunologic: The patient forms specific IgE antibodies against high-molecular weight agents such as proteins, polysaccharides, and peptides. Low-molecular weight agents can also elicit an immunologic reaction by forming haptens; platinum salts and acid anhydrides act this way. Atopic history is a risk factor. Skin testing is usually positive.
2. Pharmacologic: Other agents, mostly low-molecular weight compounds, directly trigger the release of bronchoconstrictors. Examples include TDI and plicatic acid,m the active component of western red cedar.
3. Irritant: Severe irritant exposures, such as to chlorine gas or acid mists, can lead to acute inflammatory changes of the airways with asthmatic symptoms, and a persisting clinical picture indistinguishable from asthma. This is now known as "reactive airways dysfunction syndrome" or RADS.

It is important to remember that non-occupational exposures, in the home environment, in recreational activities, and elsewhere, can also cause asthma. New-onset asthma in an adult should always raise the question of an occupational or environmental cause.

References and suggested reading

On causation:

Evans AS. Causation and disease: a chronological journey. Am J Epidemiol 1978; 108: 249-58.

Feinstein AR. Scientific standards vs. statistical associations and biological logic in the analysis of causation. Clin Pharmacol Therap 1979; 25: 481-2.

Hill AB. The environment and disease: association or causation? Proc Royal Soc Med 1965; 58: 295-300

Rothman KJ, Ed. Causal Inference. Chestnut Hill MA: Epidemiology Resources, 1988.

Susser M. Causal Thinking in the Health Sciences. New York: Oxford University Press, 1973.

On occupational asthma in general:

Chan-Yeung M, Malo J-L. Occupational asthma. New Eng J Med 1995; 333: 107-12.

On asthma in soldering:

Pepys J, Pickering CJ. Asthma due to inhaled chemical fumes -- aminoethyl ethanolamine in aluminum soldering flux. Clin Allergy 1972; 2: 197-204.

Vallieres M, Cockroft DW, Taylor DM, et al. Dimethyl ethanolamine-induced asthma. Am Rev Resp Dis 1977; 115: 867-71.

The following three lists are overheads for this lecture:

Cigarette smoking causes lung cancer.
The HIV virus causes AIDS.
Certain tampons cause toxic shock syndrome.
Breast implants cause connective tissue disease.
Service in the Persian Gulf War caused many diseases.
Electromagnetic fields cause brain cancer.
Air pollution causes asthma.
BSE causes CJD in the UK.

1. The organism is always found with the disease.
2. The organism is not found with any other disease.
3. The organism, cultured from one with the disease and cultured through several generations, produces the disease.

1. Strength.
2. Consistency.
3. Specificity.
4. Temporality.
5. Biological gradient.
6. Plausibility.
7. Coherence.
8. Experimental evidence.
9. Analogy.

This document was last modified on 06/14/2000 03:07:50 PM

This resource was acquired by CEEM (Consortium for Environmental Education in Medicine), a program of Second Nature, under the auspices of a NIEHS grant to gather and disseminate environmental health educational resources over the internet in order to help medical and allied health sciences faculty identify, locate and use resources for incorporating environment and health perspectives into their curricula. CEEM has authorized the use of these materials on this website for archival purposes. Please note that the copyright for this material is retained by the instructor and/or contributing institution.