Causality Assessment and Attribution

Causality assessment is the structured process by which pharmacovigilance judges how likely it is that a particular drug caused a particular adverse event in a particular patient or report. Because a single case rarely proves causation, the field relies on transparent criteria — timing, dechallenge and rechallenge, alternative explanations, and prior knowledge of the drug — to grade the strength of the suspected drug-event relationship.

Definition

Causality assessment is the evaluation of the probability that a suspected medicinal product is responsible for an observed adverse event, expressed as a graded likelihood (for example certain, probable, possible, unlikely) on the basis of temporal, pharmacological, and clinical evidence in a single case or report.

Scope

This area orients the reader to the logic of attributing adverse events to drugs at the level of the individual case report. It covers the major families of methods (expert global introspection, structured algorithms, and probabilistic approaches), the criteria they share, and the standard likelihood categories used to express the verdict. It treats causality assessment as a methodological reference topic in pharmacovigilance and not as clinical guidance for managing any individual patient.

Sub-topics

• Alternative Causes and Differential Diagnosis
• Dechallenge and Rechallenge Assessment
• Naranjo Causality Algorithm
• Temporal Relationship and Event Timing

Core questions

• How can the contribution of a drug to an adverse event be judged from a single case rather than from a population study?
• What criteria distinguish expert global introspection, algorithmic, and probabilistic (Bayesian) approaches to causality assessment?
• How do timing, dechallenge, rechallenge, and exclusion of alternative causes combine into a likelihood category?
• Why do different assessment methods often disagree on the same case, and what does that imply for reproducibility?

Key concepts

• Imputability and likelihood categories (certain, probable, possible, unlikely, unclassifiable)
• Expert global introspection
• Structured algorithms (e.g., the Naranjo algorithm)
• Probabilistic and Bayesian causality assessment
• Temporal relationship between exposure and event
• Dechallenge and rechallenge
• Exclusion of alternative causes
• Inter-rater reproducibility of assessment
• Individual case safety report (ICSR)

Mechanisms

Methods of causality assessment fall into three broad families. Expert global introspection relies on the clinical judgement of one or more assessors weighing all available information without a fixed scoring scheme, which is flexible but poorly reproducible. Structured algorithms convert recurring criteria — temporal plausibility, response to drug withdrawal (dechallenge), response to readministration (rechallenge), the existence of alternative causes, and prior reports of the reaction — into explicit questions with weighted answers that yield a likelihood category; the Naranjo algorithm is the most widely used example. Probabilistic methods, including Bayesian approaches, express the assessment as the posterior odds that the drug rather than an alternative caused the event, combining a prior based on background epidemiology with the likelihood of the observed case features. Across all three families the shared building blocks are timing, dechallenge and rechallenge, and the exclusion of competing explanations.

Clinical relevance

Causality assessment underpins signal detection, regulatory reporting, and the labelling of medicines, so understanding its logic is part of evidence appraisal in the pharmaceutical and health sciences. It describes how the drug-relatedness of an adverse event is judged and recorded; it characterises how safety evidence is generated and is not a basis for individual diagnostic or treatment decisions.

Evidence & guidelines

Systematic comparison of the published methods has found that no single technique can be regarded as the gold standard, that the methods frequently disagree when applied to the same case, and that reproducibility and validity remain limited; structured algorithms improve consistency over unstructured expert judgement but do not resolve the underlying uncertainty of single-case inference (Agbabiaka 2008; Hutchinson & Lane 1989). The World Health Organization Uppsala Monitoring Centre likelihood categories and structured tools such as the Naranjo algorithm are the conventional reference frameworks for expressing and standardising the verdict.

History

Concern with attributing adverse events to drugs grew after the thalidomide tragedy of the early 1960s and the consolidation of national and international spontaneous-reporting systems. Karch and Lasagna's 1977 call for an operational definition of adverse drug reactions framed the problem of moving from impression to explicit criteria, and Naranjo and colleagues' 1981 algorithm offered a reproducible scoring scheme that became a standard reference. Subsequent decades produced numerous structured and probabilistic methods, and systematic reviews later documented both their proliferation and their persistent disagreement.

Debates

Is there a gold-standard method for causality assessment?

Systematic review found that the many available methods — global introspection, algorithms, and Bayesian approaches — differ in their verdicts on the same case and that none can be considered definitively valid, leaving the choice of method a continuing methodological judgement.

Key figures

• Cesar A. Naranjo
• I. Ralph Edwards
• Jeffrey K. Aronson
• Louis Lasagna
• Fred E. Karch
• Thomas A. Hutchinson

Seminal works

• naranjo-1981
• karch-lasagna-1977
• agbabiaka-2008

Frequently asked questions

What is causality assessment in pharmacovigilance?

It is the structured judgement of how likely it is that a specific drug caused a specific adverse event in an individual report, expressed as a graded likelihood such as certain, probable, possible, or unlikely.

Why do causality assessment methods sometimes disagree?

Different methods weight the shared criteria — timing, dechallenge and rechallenge, and alternative causes — differently, and a single case rarely contains enough information to be decisive, so systematic reviews have found no single method to be a gold standard.