Naranjo Causality Algorithm

The Naranjo causality algorithm is a structured questionnaire, published by Naranjo and colleagues in 1981, that estimates the probability that a drug caused an adverse event. It poses a fixed set of yes/no/unknown questions, each carrying a weight, and sums the answers into a score that maps onto the likelihood categories definite, probable, possible, and doubtful.

Definition

The Naranjo algorithm is a ten-item weighted questionnaire that converts answers about the temporal and clinical features of a suspected adverse drug reaction into a total score, which is then classified into one of four probability categories of drug causation.

Scope

This entry describes the algorithm's purpose, the kinds of criteria its questions capture (timing, dechallenge, rechallenge, alternative causes, prior reports, and objective confirmation), and its strengths and limitations as a reproducibility tool. It is a methodological reference on a named assessment instrument and is not clinical guidance for evaluating an individual patient's reaction.

Core questions

• What criteria does the Naranjo questionnaire encode, and how are its items weighted into a total score?
• How does the resulting score map onto the definite, probable, possible, and doubtful categories?
• How well does the algorithm reproduce expert judgement and improve inter-rater agreement?
• What are its main limitations relative to global introspection and Bayesian methods?

Key concepts

• Weighted questionnaire scoring
• Probability categories: definite, probable, possible, doubtful
• Temporal relationship item
• Dechallenge and rechallenge items
• Alternative-cause item
• Prior-report and dose-response items
• Inter-rater reproducibility
• Operational definition of an adverse drug reaction

Mechanisms

The algorithm presents a fixed list of questions that capture the standard causality criteria: whether there were previous conclusive reports of the reaction, whether the event appeared after the drug was given, whether it improved on dechallenge, whether it reappeared on rechallenge, whether alternative causes could explain it, whether it recurred with placebo, whether the drug was detected in toxic concentrations, whether severity varied with dose, whether the patient had a similar reaction before, and whether the event was confirmed by objective evidence. Each answer (yes, no, or do-not-know) carries a predefined weight, and the weighted answers are summed; the total score is then read off against thresholds that assign the case to definite, probable, possible, or doubtful drug causation. By fixing the questions and weights, the instrument aims to make assessments explicit and repeatable across different raters (Naranjo 1981).

Clinical relevance

The Naranjo algorithm is one of the most frequently cited tools for standardising how the drug-relatedness of an adverse event is recorded in case reports and studies, so familiarity with it supports critical reading of pharmacovigilance literature. It documents how an attribution is reached and is not a basis for individual diagnostic or treatment decisions.

Evidence & guidelines

When introduced, the algorithm was reported to improve inter-rater agreement and validity compared with unaided clinical judgement (Naranjo 1981). Later systematic comparison placed it among the structured algorithms that increase consistency over global introspection but found that, like all available methods, it can disagree with other tools on the same case and is not a definitive gold standard (Agbabiaka 2008; Hutchinson & Lane 1989). It remains a widely used reference instrument in case reporting and research.

History

The algorithm emerged from late-1970s efforts to make adverse drug reaction assessment operational and reproducible, following Karch and Lasagna's call for explicit criteria. Naranjo and colleagues published the probability scale in 1981, and it was rapidly adopted as a convenient, transparent scoring scheme that could be applied without specialist statistical training.

Debates

Does the Naranjo algorithm settle causality, or only standardise it?

The algorithm improves reproducibility relative to unstructured judgement, but systematic reviews note that structured algorithms still disagree with other methods on individual cases and cannot be treated as a validated gold standard, so the score is best read as a structured opinion rather than a definitive verdict.

Key figures

• Cesar A. Naranjo
• Usoa Busto
• Edward M. Sellers

Seminal works

• naranjo-1981

Frequently asked questions

What does the Naranjo algorithm produce?

It produces a numeric score from a weighted ten-item questionnaire that classifies a suspected adverse drug reaction as definite, probable, possible, or doubtful in its relationship to the drug.

Is the Naranjo algorithm a definitive test of causation?

No. It standardises and makes the reasoning explicit, improving agreement between assessors, but systematic reviews report that no causality method, including this one, is a validated gold standard.