Karl Fischer Titration: Measuring Water in a Lyophilized Peptide

A certificate of analysis for a lyophilized peptide usually carries a line called water content or residual moisture, given as a percentage by weight. It is on the document because the quantity is not negligible. A freeze-dried cake is not anhydrous, and the water it holds is weighed along with everything else in the vial. Writing about freeze-dried biological products, authors at the FDA's Center for Biologics Evaluation and Research stated that for most such products residual moisture levels should be low, usually from less than 1.0% to 3.0%; the same paper describes dropping a fixed 1.0% limit and a required gravimetric method from the regulation, a change the authors attribute to changes in testing methods and to the results obtained once newer methods — gravimetric, coulometric Karl Fischer, thermogravimetric and thermogravimetric/mass spectrometric — were applied to the vaccines they report ^[8]. Those are biologicals, not peptide salts. The order of magnitude is what transfers: water in a dried cake is a mass fraction of the solid, and which method produced the figure is part of the figure.

The reference method behind that line is Karl Fischer titration. What follows is what it measures, how the two instrument formats differ, what makes a result wrong in a specific direction, and how the number propagates into a mass balance.

The reaction, and two ways of delivering iodine

Karl Fischer chemistry consumes iodine in a reaction with water carried out in a medium containing sulfur dioxide, an alcohol and a base. The stoichiometry between iodine and water is fixed, so counting iodine counts water. The two commercial formats differ only in how the iodine reaches the sample.

• Volumetric. Iodine is delivered from a burette as a titrant of known titer, expressed as milligrams of water consumed per milliliter of titrant.
• Coulometric. Iodine is generated in the cell itself by electrolysis of iodide in the reagent. The mass of iodine produced follows from the charge passed, by Faraday's law, so the instrument computes water from an electrical measurement rather than from a delivered volume — a calculation that assumes 100% current efficiency with no side reactions. The coulometers named in one published protocol are specified over a determination range of 10 µg to 200 mg of water, and the format is described there as primarily for determining small amounts of water in samples such as drug substances and drug products ^[1].

That range explains why coulometry is the usual choice for a lyophilized peptide. Take a 5.00 mg portion of a cake containing 4.8% w/w water: it holds 5.00 mg × 0.048 = 0.240 mg, or 240 µg of water — comfortably inside the working range and roughly 24 times the 10 µg floor. A volumetric titration would need either a much larger sample, which a research-scale vial may not contain, or a titrant delivery near the resolution of the burette.

Neither format is intrinsically more correct. Neither format is intrinsically more correct. A study of systematic bias in both examined six sources — the moisture standard, two kinds of instrumental bias, solvent composition, cell design and the sample itself: in oil and water-saturated octanol test systems, adjustable instrument parameters set incorrectly may cause a small but significant negative bias, reported as under 7%, solvent composition together with vessel design can introduce a negative bias of up to 10%, and the nature of the sample introduces a negative bias of its own. Under conditions the author describes as optimum, different amounts of water were measured in each of four oils using different Karl Fischer procedures and five different instruments ^[2].

Getting the sample into the cell

The cake is a solid and the titration happens in a liquid. There are two routes across that gap.

Direct introduction. The weighed solid goes into the titration vessel and dissolves or disperses, releasing its water to the reagent. Fast, and no extra hardware — but it puts the entire matrix, peptide and counterion and any excipient, in contact with the reagent, where it may react.

Oven, or vaporizer, introduction. The weighed solid is sealed in a vial, heated, and the liberated water is swept into the titration cell by a dry carrier gas. The matrix never enters the cell. This approach has been characterized with coulometric detection across matrices including coal, cement and refined oils, using parallel determinations with reagents that respond differently to interfering species in order to establish whether the volatilized material really is water. The coal and cement reference materials contained no material that reacted with methanol and very little that did not volatilize at 107 °C, from which the authors concluded that what was measured in those solids was water. The refined oils behaved differently: they released material volatile at 107 °C and at 160 °C, none of which reacted with the aldehyde–ketone reagent — which the authors read as a substance other than water that reacts with methanol to form water ^[4].

The oven route trades one problem for another. Choosing the temperature is a judgment: too low and tightly held water is never released, too high and the sample decomposes into something the reagent will titrate. That choice is method-development work for each solid, not a setting that transfers between materials.

Where the number goes wrong, and in which direction

Karl Fischer failures are not random noise. Each mechanism pushes the answer one way.

Co-titration inflates the result. Anything in the sample that consumes iodine, or that reacts with the alcohol in the reagent to liberate water, is counted as water. This is a documented pharmaceutical failure mode rather than a theoretical one: in penicillins, decomposition products — penicilloic acid among them — were shown to be partly co-titrated with conventional reagents, and the differences between results from slow and fast reagents were compared against penicilloic acid content determined independently by a mercurimetric method ^[5]. The chemistry does not transfer, but the lesson does: the titration answers "how much iodine was consumed", and assigns that to water by assumption.

Inaccessible water deflates the result. Water that never reaches the reagent is never counted. In hydrocarbons, the disparity between volumetric and coulometric results was traced to exactly this: when the sample was incompletely dissolved and the vessel contents were heterogeneous, water residing in the non-aqueous phase appeared to be completely inaccessible to either reagent ^[3]. A peptide cake that clumps, films, or fails to wet in the working medium is the same hazard in a different matrix.

Atmospheric ingress inflates the result, and it is the easiest to cause. A lyophilized cake is hygroscopic. A review of moisture in protein solids identifies sample handling that keeps atmospheric moisture out as requiring particular attention, because accurate determination depends on it ^[9]. Work on stopper drying for a lyophilized vaccine with a plug weight of about 3 mg noted that microgram quantities of absorbed water change the product's moisture content drastically, which is why the closure was treated as a moisture reservoir worth controlling ^[10]. Run the arithmetic on that geometry: a 3.00 mg plug at 1.0% w/w holds 30 µg of water; taking up another 30 µg gives 60 µg in a now 3.03 mg solid, or 60/3030 = 1.98% w/w. The number has nearly doubled, and nothing changed except how long the vial was open.

Precision degrades exactly where you need it

The lower the water content, the worse the relative precision — a general property of the measurement rather than a defect of one instrument. In a study of residual moisture in freeze-dried biological products, the gravimetric procedure gave relative standard deviations of 3.6% to 9.1% on samples containing 1.12% to 4.4% residual moisture, but 16.7% to 47.0% on samples below 1.0%. The authors report that Karl Fischer and thermogravimetric determinations on comparable samples showed similar variability ^[7].

Those figures come from one laboratory's work on biological products and are not a specification for a peptide salt. The structure of the result is what transfers: a limit written low is a limit written where the method is least reproducible, and a single determination reported to two decimal places somewhere below that 1.0% boundary claims more than the measurement supports.

One vial is not the batch

Karl Fischer is destructive. The vial that was titrated is gone, and the number describes it rather than the lot. Whether that matters depends on how uniformly water is distributed across a batch, which depends on the formulation. Frequency modulation spectroscopy applied non-invasively to every vial in freeze-dried batches showed a clear linear relationship between headspace moisture and Karl Fischer values for freeze-dried sucrose, but a more complex correlation for mannitol, where different crystalline forms complicated the relationship; mapping across the dryer related intra-batch variability to formulation, process conditions and equipment geometry ^[11]. Those results are specific to the materials and equipment studied. The safer implication is narrower: a water content figure is a lot-level attribute derived from a sample, and how well it represents the lot depends on a sampling plan that a certificate rarely states.

How the number feeds a mass balance

Water content is not an isolated quality attribute. It is a subtraction in the calculation that converts a weighed mass into a peptide mass. In the mass balance approach used to assign purity to synthetic peptide reference materials, peptide content is what remains after every non-peptide component is measured separately and removed. For synthetic glucagon, water and trifluoroacetic acid contents were determined by Karl Fischer titration and ion chromatography respectively, inorganic ion content by ICP-MS, and peptide impurity sequences by high-resolution mass spectrometry; subtracting all impurity masses from the sample mass gave 896.36 ± 0.68 mg/g, traceable to SI units ^[12]. The same architecture applied to synthetic oxytocin used coulometric Karl Fischer for water alongside ion chromatography for non-volatile counterions, ICP-MS for inorganic elements and headspace GC-MS for volatile organics, with twenty-one structurally related impurities quantified individually, yielding an assigned value of 796.5 mg/g with a measurement uncertainty of 6.5 mg/g at k = 2 ^[13] — 6.5/796.5 = 0.82% in relative terms.

A worked example makes the leverage visible. Suppose a certificate for a lot reports, on a 10.0 mg gross fill:

• Water by coulometric Karl Fischer: 4.8% w/w
• Trifluoroacetate by ion chromatography: 7.6% w/w
• Inorganic residue: 0.3% w/w
• Residual solvent: 0.2% w/w
• Peptide-related impurities: 1.8% w/w

The non-peptide mass fractions sum to 4.8 + 7.6 + 0.3 + 0.2 + 1.8 = 14.7%, so peptide content is 100 − 14.7 = 85.3% w/w, or 853 mg/g. The peptide in the vial is 10.0 mg × 0.853 = 8.53 mg. Dissolved in 2.00 mL, that is 8.53 / 2.00 = 4.265 mg/mL. Anyone who used the label mass instead would have written down 10.0 / 2.00 = 5.00 mg/mL. The error, taking the corrected value as the denominator, is (5.00 − 4.265) / 4.265 = 0.735 / 4.265 = 17.2% high.

Now isolate the water term. Drop it to zero and leave everything else alone: peptide content becomes 90.1%, the peptide mass 9.01 mg, the concentration 4.505 mg/mL, and the error against 4.265 mg/mL is 0.240 / 4.265 = 5.63% high. Ignoring water alone accounts for roughly a third of the total overstatement here. Now suppose instead that the titration is wrong by one percentage point and reports 3.8%. Peptide content rises to 86.3%, the concentration to 4.315 mg/mL, and the error against 4.265 mg/mL is 0.050 / 4.265 = 1.17%. A one-point error in the water figure moves the stock by about 1.2% — small next to the 17.2% penalty for skipping the correction, which is how to rank the two problems.

One check is free on any certificate that reports the components: confirm that the stated peptide content and the non-peptide fractions sum to 100%. When they do not, either a component is unmeasured or the peptide content came from somewhere other than a mass balance — most often an HPLC area percent, which is not a mass fraction at all.

What the titration does not tell you

Karl Fischer reports total water. It does not report where that water is or how tightly it is held. In a comparison of four methods on a pharmaceutical polymer, the titration was judged to determine total water content accurately while thermal methods underestimated it — but the authors state that it does not indicate the type of water present, and recommend complementary methods ^[6]. That distinction is not academic. In protein solids, degradation is described as minimal at or below the monolayer level of hydration, while moisture beyond a monolayer generally raises decomposition rates, both through greater conformational flexibility and because the less tightly bound water can mobilize reactants; water also acts as a plasticizer lowering the glass transition temperature ^[9]. Whether those thresholds transfer to a short synthetic peptide salt is not something that review addresses.

Two further limits follow from the method itself: the titration cannot distinguish water present at manufacture from water that arrived while the vial sat open, and it says nothing about the vials you did not titrate.

Measured, convention, and unknown

Measured. The relationship between charge and iodine in the coulometric cell ^[1]. Directional biases from instrument settings, solvent and vessel design ^[2]. Inaccessibility of water in a heterogeneous vessel ^[3]. Co-titration of a matrix decomposition product ^[5]. The collapse of relative precision below 1% water ^[7]. Headspace moisture against titrated water in specific formulations ^[11]. Peptide purity assigned by mass balance with stated uncertainties ^{[12, 13]}.

Convention. Much of what surrounds the measurement is accepted practice with thin primary literature behind it, at least for short synthetic peptides. Reporting water as a percentage by weight rather than an absolute mass. Standardizing a volumetric titrant against a water standard before use and repeating it as the titer drifts. Running a blank or drift determination on the cell and subtracting it. Weighing hygroscopic solids in a glove bag or dry box, transferring through a septum, and equilibrating a cold vial to ambient temperature before opening it. Duplicate or triplicate determinations. Treating the figure as lot-specific. These are sensible and near-universal; they are uncited here because the literature supporting them for peptide salts is thin.

Unstated. Most certificates report a water content without saying which format produced it, whether the sample was introduced directly or through an oven, at what temperature, how many determinations were averaged, or how the vials were sampled. None of those omissions makes the number wrong. All of them make it harder to know how much of your own error budget to allocate to it — which is the only reason to read the line at all.