When a patient has a negative medical outcome while using a product like a drug or vaccine, a question arises: was it caused by the product or was it a coincidence? An “adverse event” is any unfavorable medical occurrence during treatment, but it does not automatically imply the treatment was the cause. The process of determining the likelihood that a product caused an adverse event is called causality assessment.
This evaluation is a formal process that scientists and healthcare providers use to move from suspicion to a scientifically grounded conclusion. A systematic causality assessment ensures that medications are used safely and that doctors and patients can trust the information about potential risks.
Distinguishing Correlation from Causation
The first step in a causality investigation is understanding the difference between correlation and causation. Two events are correlated if they occur around the same time, but this does not mean one caused the other. For instance, ice cream sales and drowning incidents both increase during summer. They are correlated, but the summer heat is the underlying factor increasing both activities, not one causing the other.
This logic applies directly to medicine. If a patient takes a new medication and develops a headache the next day, the events are correlated. However, headaches are common, and it could be a coincidence, a symptom of an unrelated illness, or a direct effect of the drug.
Determining the true cause is the central challenge. Observing that an event followed a drug’s administration is not enough evidence for a causal link. Investigators must first rule out other factors like coincidence, other medications, or underlying illnesses. This prevents blaming a medication for an event it did not cause, which could lead a patient to stop a beneficial treatment.
The Bradford Hill Criteria for Causality
To move from correlation to a plausible case for causation, investigators use guidelines developed by epidemiologist Sir Austin Bradford Hill in 1965. These are not a rigid checklist but a framework for thinking through the evidence to determine if an observed association is likely causal.
Temporality
This criterion states that the cause must precede the effect and is the only essential one. For a drug to have caused an adverse event, the patient must have taken the drug before the event’s symptoms began. If a headache started Tuesday morning and the first dose was not taken until Tuesday afternoon, the drug cannot be the cause.
Strength of Association
This criterion examines how strong the link is between the exposure and the outcome, often measured statistically. If patients taking a drug are ten times more likely to experience an adverse event compared to those not taking it, that is a strong association. A weak association makes a causal link less certain, as it could be explained by other factors.
Dose-Response Relationship
A dose-response relationship provides evidence for causality. This means that as the drug dose increases, the frequency or severity of the adverse event also increases. For example, if a low dose causes a mild rash in a few people, but a high dose causes a severe rash in many, this strengthens the argument that the drug is responsible.
Consistency
The consistency criterion asks if the same association has been observed by different researchers, in different places, and in different patients. If doctors in multiple countries independently report the same liver injury in patients taking the same drug, the case for a causal relationship becomes much more convincing than an isolated report.
Biological Plausibility
This criterion considers if the proposed causal link makes scientific sense. Is there a known biological mechanism that could explain how the drug might cause the adverse event? For instance, if a drug is metabolized by the liver and the adverse event is liver damage, the association is biologically plausible.
Dechallenge and Rechallenge
A “positive dechallenge” occurs if the adverse event subsides after the drug is withdrawn, suggesting it was the cause. An even stronger piece of evidence is a “positive rechallenge,” where the event reappears after the drug is administered again. However, deliberately re-exposing a patient to a drug suspected of causing harm is often unethical and rarely done.
Standardized Causality Assessment Methods
To ensure consistency and objectivity, the principles of causality are applied through standardized methods. These tools formalize the assessment, transforming an expert’s judgment into a structured and reproducible evaluation. This moves the process from subjective opinion toward systematic analysis.
One widely used tool is the Naranjo Algorithm, a questionnaire of ten questions probing the likelihood of a causal link. Each question addresses an aspect of causality, like temporality or dechallenge. The evaluator answers yes, no, or “do not know,” and points are awarded or subtracted based on the answer, with the final score placing the event into a probability category.
The output of these algorithms is a classification on a standardized scale, like the World Health Organization-Uppsala Monitoring Centre (WHO-UMC) system. This system categorizes the likelihood of a drug causing an adverse event into levels: Certain, Probable/Likely, Possible, Unlikely, and Unclassifiable. A “Certain” classification requires a clear time relationship and positive dechallenge and rechallenge, while “Possible” may only have a reasonable time sequence. This provides a common language for experts to communicate certainty.
Sources of Adverse Event Data
Information for causality assessments comes from two primary sources, each with distinct strengths and limitations.
Data is first collected during pre-market clinical trials, which are highly controlled studies conducted before a drug is approved. In this setting, researchers monitor a small number of participants, collecting high-quality data. The controlled nature of these trials makes establishing causality easier, but their limited size and duration mean they are unlikely to detect very rare adverse events.
Once a product is on the market, data collection enters post-market surveillance, or pharmacovigilance. This involves monitoring a drug’s safety in a much larger and more diverse population. A primary source is spontaneous reporting systems, where patients and providers voluntarily submit reports of suspected adverse events. This system can detect rare events missed in trials, but the data is often incomplete, making it difficult to establish causality from a single report.