If you have ever tried to prove what causes a disease, you know it is rarely as simple as it appears. Previously we saw Bradford Hill criteria. On paper, these nine principles look like a clear checklist for separating causes from coincidences. In practice, the picture is messier.
Take the “strength of association” idea. You might think a stronger link means a stronger case for causation. But that assumption can trap you:
- Some causes are only one piece in a bigger chain, so their individual impact looks weak.
- The measured strength can change depending on how common other risk factors are.
- Even a strong association can be misleading if another hidden factor is driving both.
If you treat “strength” as proof, you risk ignoring real causes and chasing false ones.
Why Dose–Response and Specificity Can Mislead You
The dose–response relationship suggests that more exposure should mean more disease. It feels logical, but reality often bends the rule. Patterns like this can appear even when the exposure is not the true driver because:
- Confounding factors create a fake upward curve.
- Data collection errors skew the pattern.
- Social or environmental differences change who gets exposed and by how much.
Specificity is another principle that sounds useful but fails in practice. It claims a cause is more believable if it produces only one outcome. Yet in real life, causes often have many effects. Cigarette smoking is the perfect example:
- Lung cancer
- Heart disease
- Chronic lung conditions
- Cosmetic and social effects like stained teeth or bad breath
Rothman cuts through the noise with a blunt observation, only temporality truly defines causation. You have to prove the cause came before the effect. Even then, figuring out that sequence can be surprisingly hard.
Disease Is Not Random
One thing you should never forget, disease does not strike randomly. Some people are more likely to get sick than others because risk factors are unevenly distributed.
The job of epidemiology is to find these risk factors. When you know them, you can target prevention where it matters most. This is where models of causation come in, they give you a way to map how a disease develops.
The Epidemiologic Triangle
One of the oldest models is the epidemiologic triangle. It has three parts:
- Agent — the thing that causes harm
- Host — the person who can get the disease
- Environment — the setting that allows agent and host to interact
Disease occurs when these three elements align in the right way.
Agents, Hosts, and the Environment
An agent can be a virus, bacterium, parasite, or any other pathogen. It can also be a chemical or physical cause. But an agent alone is rarely enough. How much harm it does depends on:
- Its ability to cause illness, called pathogenicity
- The size of the dose you are exposed to
For instance, long‑term exposure to lead in drinking water has caused serious health problems in some communities. Lead is a chemical agent that can damage the nervous system, especially in children. The severity of harm depends on both the amount of lead consumed and the vulnerability of the individual.
The host, that means you or another person, brings personal factors to the mix – age, gender, genetics, immune health, and lifestyle choices all matter. The environment sets the stage, shaping exposure opportunities through factors like:
- Climate and geography
- Presence of disease‑carrying insects
- Crowding, sanitation, and access to health care
Why the Triangle Is Not Enough
The triangle works well for infectious diseases. But for conditions like heart disease or cancer, it is not enough. These illnesses often have multiple contributing factors without a single necessary cause.
To address this complexity, other models emerged. One of the most important is Rothman’s causal pies model.
The Causal Pies Model
Think of each disease pathway as a pie. Each slice is a component cause. When all slices are present, the pie is complete and the disease occurs. A complete pie is called a sufficient cause.
If a factor appears in every pie for that disease, it is a necessary cause. Without it, the disease cannot occur. But in most real cases, no single cause appears in every pie. That is why you often have several pies with different slice combinations.
When Causes Are Necessary but Not Sufficient
Consider Pneumocystis carinii. It can live harmlessly in the lungs of healthy people. But in someone with a weakened immune system, such as a person with HIV, it can cause lethal pneumonia.
In this case:
- The organism is a necessary cause — you cannot get the pneumonia without it.
- It is not sufficient — immune weakness must also be present.
Lung Cancer and Multiple Pathways
Lung cancer shows why the causal pies model is so useful. Smoking is a major risk factor, yet not all smokers develop the disease. Some people who never smoke still get lung cancer.
Imagine:
- Component cause B = smoking
- Component cause C = asbestos exposure
One pathway might have both B and C. Another might have C but not B. A third might have B but not C. And other pies could have neither. This is why no single factor explains all cases.
Acting Without Knowing Every Cause
Here is the good news: you do not need to identify every slice in the pie to act. If you remove just one slice, you block that entire pathway.
Eliminating smoking, for example, would:
- Remove it from all pies in which it appears
- Prevent all lung cancers caused by those pies
- Still leave other cases, but significantly reduce the total burden
If you work in public health, these models are not abstract. They guide real decisions about where to focus resources and which prevention strategies to use.
The Bradford Hill criteria still help you think through causation, but they are not a final verdict. Real diseases are often the product of many interacting causes. The more you understand that complexity, the better you can find the leverage points that prevent illness and save lives.