Accuracy and Recovery: Three Ways to Find Out Whether a Method Reads True
A method that gives the same answer six times is precise. Whether that answer is correct is a separate question, and no amount of repetition will settle it. A systematically low reading is low every time — consistently, reproducibly, convincingly wrong.
Quick answer
What is accuracy? ICH Q2(R2) defines it as the closeness of agreement between the value accepted as a conventional true value or accepted reference value and the value or set of values measured. The guideline requires it to be established across the reportable range, demonstrated under regular test conditions — in the presence of sample matrix and using the described sample preparation steps — and reported as the mean percent recovery of a known added amount, or as the difference between the mean and the accepted true value, together with an appropriate confidence interval.
The phrase doing the most work is under regular test conditions. Accuracy measured on a clean standard says nothing about accuracy on your actual sample, because the losses live in the preparation.
The three routes to establishing it
| Route | What is done | What limits it |
|---|---|---|
| Reference material comparison | The procedure is applied to an analyte of known purity — a reference material, a well characterised impurity or a related substance — and measured results are compared with theoretically expected results | Inherits whatever uncertainty the reference material carries |
| Spiking study | A known amount of analyte is added to a matrix containing all components except the analyte, and spiked and unspiked results are compared | Requires the matrix to be reproducible without the analyte |
| Orthogonal procedure comparison | Results are compared against a procedure using a different measurement principle, whose own accuracy must be reported | Only as good as the comparator, which must state its own accuracy |
Every route is a comparison against something already assumed known. There is no fourth option in which a method simply reveals the truth unaided — which is the practical meaning of the chain described in metrological traceability.
The guideline adds one shortcut worth knowing: in certain cases, accuracy can be inferred once precision, response within the range and specificity have been established. Accuracy is not always measured directly; sometimes it is argued from the other three.
How much work it actually takes
ICH Q2(R2) is specific about the minimum. Accuracy should be assessed using an appropriate number of determinations and concentration levels covering the reportable range — the guideline's own example being three concentrations with three replicates each of the full analytical procedure.
Note of the full analytical procedure. Not nine injections from one prepared vial. Nine complete runs through weighing, dissolution, dilution and measurement, because that is where accuracy is lost. Nine injections of one solution measure the instrument; nine preparations measure the method.
This is the single most useful thing on this page for someone working with research material: if you want to know whether your measurement is right rather than merely repeatable, the replicates have to start at the powder, not at the autosampler.
What recovery figures look like in practice
Published validations give a sense of the range, and the contrast between them is the lesson.
| Context | Reported recovery | Why |
|---|---|---|
| Small molecules in a tablet matrix (2022 stability-indicating method) | 98.51–100.80% | Soluble, stable analytes in a well characterised formulation |
| A protein quantified in a food matrix via a signature peptide (2017) | 83.1–91.6% | Extraction and enzymatic digestion steps, each incomplete |
A method recovering 85% of what is present is not broken. It is a method with a known, quantified, declared bias. The two legitimate responses are to apply a correction factor — which carries its own uncertainty into the result, as discussed in measurement uncertainty — or to accept the bias and state it. The illegitimate response is to report the number as if recovery were 100%.
Why peptides lose material before the instrument sees them
Recovery below 100% in peptide work usually has physical causes rather than chemical ones, and most of them happen in the first five minutes:
- Adsorption to surfaces. Peptides bind to glass and plastic, and at low concentrations the fraction lost is not small — see adsorption losses.
- Incomplete dissolution. Material still on the vial wall is not in the aliquot. The cake behaviour that causes this is covered in cake, powder and film.
- Weighing error on small masses. A few hundred micrograms weighed on an unsuitable balance carries a percentage error that dwarfs everything downstream.
- Water and counterion. The weighed mass is not all peptide, which is the correction described in net peptide content versus purity.
The last one is a systematic offset rather than a loss, and it runs in the direction that makes material look like there is more of it than there is.
The confidence interval requirement
ICH Q2(R2) does not accept a bare recovery figure. Accuracy should be reported together with an appropriate 100(1−&alpha)% confidence interval, or a justified alternative statistical interval, and the observed interval should be compatible with the accuracy acceptance criteria.
In other words: the criterion is not whether the mean recovery looks acceptable, but whether the interval around it is. A mean of 99% from three scattered determinations is a weaker claim than a mean of 95% from nine tight ones, and the interval is what makes the difference visible.
What a research buyer can do
- Run a spike into your own prepared sample once. Add a known amount to material you have already prepared, measure, and compare. It is the single most informative experiment available without specialist equipment.
- Make replicates start at the powder. Replicate injections measure the instrument. Replicate preparations measure what you actually want to know.
- Record recovery alongside results. A value measured by a method with 88% recovery is a different number from one measured at 99%, and in six months nothing else will record that.
- Do not correct silently. If you apply a recovery factor, record that you did and what it was.
- Expect lower recovery at lower concentration. Adsorptive losses are proportionally worse the less material is present.
Key takeaways
- Accuracy is closeness to an accepted true value; precision is closeness between replicates. Repetition cannot reveal a systematic error.
- ICH Q2(R2) gives three routes: reference material comparison, spiking study, and orthogonal procedure comparison.
- The guideline's example minimum is three concentrations with three replicates each, of the full analytical procedure.
- Accuracy must be demonstrated under regular test conditions, including sample matrix and the described preparation steps.
- Results are reported as mean percent recovery, or difference from the accepted true value, with a confidence interval — not as a bare number.
- Published recoveries range from 98.51–100.80% for small molecules in a tablet matrix to 83.1–91.6% for a protein measured through a digestion step.
- Accuracy can sometimes be inferred once precision, response and specificity are established, rather than measured directly.
Frequently asked questions
If recovery is 85%, should I multiply my result by 1.18?
Only if you have established that recovery is 85% for your material, your preparation and your concentration, and only if you record that you did it. An uncorrected result with a declared recovery is more honest than a corrected one with an undocumented factor.
Can I test accuracy without a reference material?
Yes — a spiking study needs a known added amount rather than a certified material, and comparison against a method using a different principle needs no reference material at all. Both are weaker than a certified comparison, and both are far better than nothing.
Why does the guideline insist on the full procedure?
Because the dominant losses are in sample preparation. A validation that starts at the instrument has measured the part that was never in doubt.
Does high purity imply accurate quantification?
No. Purity is a ratio within the detected material; accuracy concerns an absolute quantity. A 99% pure material can still be quantified 15% low if the preparation loses material — the distinction set out in what 99% purity actually means.
Is a supplier's accuracy claim verifiable?
Not from the certificate. What is verifiable is whether your own measurement agrees with theirs, and a disagreement tells you something is different between the two methods — which is worth knowing even when you cannot say which is closer to the truth.
Sources
- International Council for Harmonisation. ICH Harmonised Guideline Q2(R2): Validation of Analytical Procedures. Final version adopted 1 November 2023. ICH guideline
- Vankalapati KR, Alegete P, Boodida S. Stability-indicating HPLC method development and validation for simultaneous estimation of metformin, dapagliflozin, and saxagliptin in bulk drug and pharmaceutical dosage form. Biomed Chromatogr. 2022;36(7):e5384. DOI
- Yuan M, Feng C, Wang S, Zhang W, Chen M, Jiang H, Feng X. Selection of possible signature peptides for the detection of bovine lactoferrin in infant formulas by LC-MS/MS. PLoS One. 2017;12(9):e0184152. DOI
Journal citations retrieved via PubMed. The small-molecule and food-matrix studies are quoted as reference points for what recovery figures look like; neither material is a research peptide.
This article describes analytical methodology only. It contains no usage instructions, no dosing information and no protocols of any kind.
For laboratory and in-vitro research use only. Not for human or veterinary use.
For laboratory and in-vitro research use only. Not for human or veterinary use. See our research-use disclaimer.
