To save a month, the trial batch is dropped from the transfer plan.
That makes the first batch the receiving site ever produces the one whose stability data goes into the dossier. The month is only an example. The cut itself is tempting, because a trial batch uses API, line time and people, and its tablets are not sold.
The trial batch is the one run where surprises are allowed: a blender that discharges slower than the sending site's, a granulation end-point that arrives early on a different mixer, operators who have only read the batch record. There, each of these ends as a note in a report. Met for the first time on a registration batch, the same event is a deviation in the file behind the application.
ICH Q1A(R2) (2003) says the manufacturing process used for primary stability batches should simulate the one to be applied to production batches and provide product of the same quality.
The WHO guideline on technology transfer (TRS 1044, Annex 4, 2022) mentions trial batches once: the transfer document should include the timing of trial production batches and of validation batches. It sets no number of them.
Our advice is to run the trial batch, compare its results with the sending unit's data, close what it shows, and then make the batches that count.




