Causality: Making a Defensible Judgment

Part of Principal Investigator (PI) Training: Adverse Event Assessment.

About 4 minutes. Free, in your language, no account needed.

Why this matters

Causality is the assessment nobody can make for you. The sponsor will make their own; the regulator will read both. Yours is the one made by the person who saw the patient, and it drives whether an event becomes a suspected adverse reaction and potentially an expedited report.

It is also genuinely difficult, made under time pressure, with incomplete information, about a single person. The goal is not certainty. It is a judgment that is honest, reasoned and recorded.

The threshold: a reasonable possibility

The Glossary's definition of an adverse drug reaction turns on whether a causal relationship between the investigational product and the event is at least a reasonable possibility — meaning the relationship cannot be ruled out.

That threshold is deliberately low, and understanding why prevents the commonest error. Safety signals are detected across thousands of participants, not within one. Your job is not to establish that the drug caused this event; it is to say whether it plausibly might have. Excluding events because you cannot prove causation removes exactly the data the system needs.

The practical consequence: when genuinely uncertain, record it as related. Uncertainty resolves toward reporting, not away from it.

What to weigh

Six considerations do most of the work. None is decisive alone.

1. Temporal relationship. Did the event follow administration in a plausible time frame? Necessary, but far from sufficient — many things follow a dose.
2. Biological plausibility. Is there a known or postulated mechanism? Consult the Investigator's Brochure; this is one of the reasons §2.1 asks you to be familiar with it.
3. Known profile. Has this been reported with this product, or with its class?
4. Dechallenge. Did the event improve when the product was stopped or reduced? A positive dechallenge is strong evidence. Note that many events resolve on their own.
5. Rechallenge. Did it recur on re-exposure? The strongest single piece of evidence, and rarely available — and rechallenge is generally not something you undertake deliberately for the sake of the assessment.
6. Alternative explanations. The participant's underlying disease, concomitant medications, intercurrent illness, procedures. The existence of an alternative explanation does not by itself exclude the product; both can be true.

The category system

Protocols use different scales — commonly a binary related / not related, or a graded scale such as not related, unlikely, possible, probable, definite. Use the one your protocol specifies. Where a graded scale is used, sponsors generally treat possible and above as a suspected reaction, so where the line falls matters.

Two habits keep the assessments consistent:

• Assess each event on its own. A participant with five events may have five different causality answers. Applying one verdict to a whole visit is a shortcut that shows.
• Assess against the investigational product specifically, and separately against the trial procedure where the protocol asks — a hematoma from a study-mandated blood draw is related to a procedure, not to the drug.

Writing it so it holds up

The assessment is a medical opinion, and an opinion with no reasoning attached cannot be defended two years later by anyone, including you. Two or three sentences are enough:

Day 12 of dosing, new onset urticaria within four hours of the morning dose. No new foods, no new concomitant medications. Rash of this type is described in the current Investigator's Brochure. Improved within 24 hours of holding the product. Assessed as possibly related.

That paragraph shows a temporal relationship, an excluded alternative, a plausibility source and a dechallenge, and it took a minute to write.

Where you later change your mind — new information, a positive dechallenge, an alternative diagnosis confirmed — update the assessment and say why. A changed causality assessment with a documented reason is good practice. A silently altered record is not.

Key takeaways

• The threshold is a reasonable possibility that cannot be ruled out — not proof.
• When genuinely uncertain, assess as related; uncertainty resolves toward reporting.
• Weigh timing, plausibility, known profile, dechallenge, rechallenge and alternative explanations — an alternative explanation does not automatically exclude the product.
• Record two or three sentences of reasoning, assess each event individually, and document any later change with its reason.