The Naranjo Algorithm

The Naranjo algorithm takes a different structural approach to the same underlying problem WHO-UMC solves. Instead of walking through criteria to land on a qualitative category, it asks ten specific yes/no/unknown questions, each with its own point weighting, and sums the answers into a numeric score. That score then maps to one of four categories: Definite (9 or above), Probable (5 to 8), Possible (1 to 4), or Doubtful (0 or below).

The ten questions cover ground that will feel familiar from the WHO-UMC discussion: prior conclusive reports of the reaction, timing relative to drug administration, response to dechallenge, response to rechallenge, alternative explanations, response to placebo, drug detected in a body fluid at a toxic concentration, dose-response relationship, prior history of a similar reaction to the same or a similar drug, and whether the event was confirmed by objective evidence. Each answer contributes points in a specific direction — notably, "did the event reappear on rechallenge" carries a +2 for yes and a −1 for no, giving positive rechallenge outsized weight in the score, mirroring exactly why WHO-UMC reserves "Certain" for cases with a positive rechallenge.

The practical difference from WHO-UMC is that Naranjo produces a single number rather than a category reached through branching logic, which has real advantages: it’s mechanically easier for two different assessors to apply identically, since there’s less room for judgment about which "branch" a case falls into. The tradeoff is that a numeric sum can obscure exactly which specific piece of evidence is driving the score — two cases can both land at "Probable" with the same total, one because of strong timing and dechallenge evidence, the other because of a dose-response relationship and prior similar reaction history, and the resulting category alone doesn’t tell you which.

In practice, many PV organisations use WHO-UMC as the primary company causality method (it’s explicitly recommended in several national programmes, including India’s PvPI) while Naranjo remains popular in clinical and academic settings for its simplicity and reproducibility. Research comparing inter-rater agreement between the two scales has found meaningful disagreement between assessors on both — a reminder that structured scales reduce inconsistency, they don’t eliminate the underlying judgment calls entirely.