Why the Same Method Gives Different Numbers in a Different Lab
Three failures come up again and again when a method moves house. Two impurities that the original column pulled apart arrive as one peak on the receiving column, so the impurity count falls and the purity figure climbs. A peak whose tailing was tolerable on a fresh column tails badly on a tired one, and integration stops being reproducible. Or the receiving laboratory’s water or additive lifts the baseline at 214 nm far enough that small peaks sink into it.
Notice the direction. All three make the receiving laboratory’s purity look better, and a number that has improved is a number nobody chases. Which is a useful way into the whole subject: transferring a method means demonstrating that results still mean what they meant, and the reason that takes work is that a written method has never been a complete account of what happens at the bench.
The method text is not the method
Column, gradient, flow rate, temperature, wavelength, injection volume — all of that is written down and all of it moves without difficulty. The trouble is everything the originating laboratory knew and never had occasion to record.
| Unwritten variable | What it is | What it does |
| Dwell volume | Volume between the mixing point and the column head | Varies several-fold by instrument design; delays gradient arrival and moves every retention time |
| Extra-column volume | Tubing runs, dead volume in fittings, detector cell size | Broadens bands, invisibly as far as the method text goes |
| Column batch | Production lot of a nominally identical packing | Measurable selectivity differences; larger still across manufacturers sharing a nominal chemistry |
| Mobile phase practice | Additive by weight or volume, organic added first or last, water purity | Small retention shifts |
| Integration settings | Threshold, peak width, baseline algorithm | Usually the operator’s choice, and they move the reported number directly |
Chasing retention times is the wrong target
Expect the second laboratory not to reproduce the first laboratory’s retention times. Trying to force agreement by adjusting the gradient generally makes the separation worse rather than better.
Divide each peak’s retention by that of the main peak and most of the system-to-system difference cancels out. Relative retention is the quantity that ought to match, and the reasoning behind that is developed in retention time as identity evidence.
Suitability criteria are the part that travels
This is where most of the work gets done. A system suitability test states what the chromatography must deliver before any sample result counts: resolution no worse than some value between a nominated pair of peaks, tailing no greater than a stated factor, plate count no lower than a floor, replicate injections agreeing within a defined spread.
Because those are properties of the separation and not of any particular instrument, a laboratory meeting them has shown its system performs equivalently regardless of what its dwell volume happens to be. Turn that around and a method carrying no suitability test has no transportable definition of working properly at all. The details are in system suitability in peptide analysis.
The formal exercise, briefly
What it amounts to is a comparative study: both laboratories, the same material, acceptance criteria settled before anyone collects data.
Ordinarily the receiving laboratory analyzes the same samples the originating one did, in replicate, and the two sets are set against each other on three counts — agreement in the reported value, agreement in precision, and agreement in what the method manages to detect. That third count matters, because a method that returns a matching purity while resolving fewer impurity peaks has not transferred at all.
The criteria themselves are absolute differences fixed ahead of time. Working out afterwards what ought to count as close enough is not a transfer. It is a justification written after the fact.
The version most research laboratories actually run
Formal transfer is rare in research-grade work. A laboratory picks up a published or supplied method, runs it, and the question of equivalence is never posed out loud.
A shared reference is the workable stand-in: both laboratories analyze the same retained sample and compare what they get. One such comparison will catch most of the failures described at the top of this page without any formal machinery whatsoever, and it is the strongest argument for keeping a retained reference sample in the first place.
Two certificates, two numbers, no error
Where two documents describing the same material disagree on purity, the explanation is more often a transfer that never took place than a mistake by either laboratory. Both may have executed their work correctly. Their numbers still describe separations that are not quite the same separation.
Hence a rule of thumb worth holding: purity figures compared across suppliers, or across laboratories, tell you rather little, while lots of one compound analyzed the same way tell you a great deal. Further reading: why certificates disagree on purity and measurement uncertainty.
