Accepted answer
Twelve months is 52 weekly administrations across 365 days, and on a four-week ladder from the bottom of the range about 5 steps — so roughly 20 of the 52 doses are escalation doses and 32 are at maintenance. Model it in that order and the routes become comparable: doses per year first, milligrams per dose second, cost per milligram third. Anything quoted per vial hides the second of those, which is the one that changes most between the first 20 doses and the last 32. Then add what each route charges that the other does not. A prescription route carries consultation and dispensing fees, spread across the 52 doses rather than paid once. A research route carries testing, shipping, and the material lost between them. Testing is the line most sheets omit. At one lot a quarter, a test-every-lot policy is 4 assays a year; at one lot a month it is 12. That difference is usually larger than any difference in price per milligram, and it is a policy you choose rather than a cost you are quoted. Put doses per year in the top row and derive everything under it, and the twelve-month totals compare on arithmetic instead of on presentation.
The honest answer is that the cheapest headline price is frequently not the cheapest outcome.
Independent testing costs roughly the price of one to two vials at the services this community uses. On a two-vial order that is a fifty to a hundred per cent surcharge; on a twenty-vial order it is five per cent.
Carriage amortises across the order. Twenty-five pounds of carriage on one vial is £2.50/mg on a 10 mg vial; on ten vials it is £0.25/mg. That single term explains most of the case for larger, less frequent orders.
Published content assay results across the independent services show nominal and measured content differing by one to ten per cent, which is the term that makes label-price comparisons unreliable.
A spreadsheet built on label claim rather than measured content is precise about the wrong number.
Divide by measured content, not by label claim. That is the whole correction.