Why should an out-of-trend result and an out-of-specification result trigger different workflows?

Why

The FDA's OOS guidance applies to results outside specifications or acceptance criteria. It sets a phased investigation, rules for retesting and resampling, and the principle that an OOS result cannot be discarded without an assignable cause. The guidance notes that much of it may also be useful for out-of-trend results, but OOT is not a specification failure and has no registered limit.

Its value is foresight. Imagine stability assay results of 100.1, 99.8, 99.6 and then 97.9 at 12 months for one lot while other lots sit near 99.5. The 97.9 passes a limit of 95.0, yet it predicts the lot may fail before expiry.

Treating it like an OOS would mean holding a passing batch on a judgement call. Ignoring it would mean learning about the failure from a stability OOS later, possibly after a field alert. An OOT procedure therefore defines how trends are detected, such as a regression prediction interval for stability or control chart rules for release data, and what the response is: confirm the data, check the method and the lab, assess the lot and related lots, and decide whether a formal investigation is needed.

EU Annex 16 lists out-of-trend investigations alongside OOS as things a QP must see completed enough before certifying. OOT detection depends on enough history to know the trend, so new products have weak OOT power.