What does 96.4% purity on a C18 column actually tell me about a semaglutide vial?

What I am working with: 96.4% · a C18 column · semaglutide.

I have read the primary source rather than the summary, which has left me with more questions.

I understand the headline. I do not understand the footnotes, and the footnotes look important.

How should I read this, and where are the traps?

It tells you 96.4 per cent of the integrated area came off a C18 column where semaglutide comes off, and the remaining 3.6 per cent did not. That is an area statement at one wavelength, not a mass statement about the vial: 3.6 per cent of area is only 3.6 per cent of mass if every impurity absorbs exactly as strongly as the parent, which none of them do. It also says nothing about how many milligrams are in the glass — water, counter-ion and a short fill are all invisible to it. What a C18 column does add is a constraint on what could be hiding: a column that retains by hydrophobicity separates deletion sequences well and separates isomers of identical hydrophobicity not at all.

More usefully, peak shape carries as much information as peak area does, and a badly tailing peak or a shouldered peak is telling you something about the sample or the column that matters.

Reverse-phase stationary phases use C18 or C8 chains bonded to silica, and the pore size of the silica matters more for a peptide of this chain length than the ligand length does.

Mass shifts and what they usually mean

The underlying point is that sample preparation is almost always under-appreciated — a reconstituted peptide in strong solvent will distort its own peak on the gradient.

The resolving power of a separation is quantified by the resolution parameter R, defined from the heights and widths of adjacent peaks, and pharmacopoeial methods typically demand R greater than 1.5 for a method to be considered validated.

Ask for the chromatogram and the system suitability data, not just the number.

Start by understanding what the detector is measuring and what that means about how the molecule needs to be prepared and handled before injection.

System suitability checks on replicate injections of a known standard establish whether the method was in control — if the peak area varies by more than two per cent between replicates, something is wrong.

Temperature affects both the viscosity of the mobile phase and the dynamics of molecular interactions, and a method developed at 25 degrees and run at 40 degrees will not behave identically.

Peptide impurities from solid-phase synthesis include deletion sequences, truncations from premature cleavage, racemised residues from epimerisation and oxidised variants, each of which may have different chromatographic behaviour.

One qualification: the limit of detection on a reversed-phase HPLC is set by the noise on the baseline, which for these molecules is usually in the range of a tenth of one per cent or less, and anything smaller is not reproducibly detectable.

The practical summary: specify the method, run the same method on every sample you compare, and use orthogonal techniques to confirm the result.

The chemistry of peptides at low pH and high organic-solvent concentration is not something most users have intuition for, which is why published methods exist.

The 214 nanometre wavelength is chosen because it corresponds to the amide backbone absorption, making response roughly proportional to the number of peptide bonds.

It helps to be literal here: acetonitrile is the organic modifier of choice because it has a good UV cutoff, a reasonable viscosity and a refractive index that minimises baseline noise.

Inter-laboratory studies using identical methods on identical material show precision well within half a per cent when the method is locked down, pointing to method variability as the primary source of disagreement.

The caveat is that HPLC is a purity technique and says almost nothing about whether the main peak is actually your target compound — that is why identity confirmation from mass spectrometry or peptide mapping matters.

If two labs give different numbers, the method difference is the first hypothesis, not lab quality.

edited 16 Jul 2024 by Dr_Ilse_Vandenberg — updated for the 2026 guidance change