Spike recovery (assay accuracy)Spike and Recovery / Accuracy by Standard Addition

A calibration curve built from pure reference standard dissolved in clean solvent tells you how the detector responds to the analyte in isolation. It says nothing about whether the full sample-preparation procedure — extraction, filtration, dilution, cleanup, chromatography — actually recovers that analyte intact out of a real product matrix: a gummy base, a softgel fill, a film-coated tablet, a fatty botanical extract, or a fermentation broth. Spike-and-recovery (also called standard addition when done directly in the sample rather than in a placebo) closes that gap.

The design is simple in concept: a known, accurately measured amount of reference standard is added ('spiked') into a matrix — ideally a representative placebo that contains every excipient except the active, at label-equivalent proportions — at several concentration levels spanning the assay's working range. The spiked sample is then carried through the entire validated procedure exactly as a routine sample would be, and the amount measured is compared against the amount added. Percent recovery = (measured amount / amount spiked) × 100.

ICH Q2(R2) (the 2023 revision, harmonized with ICH Q14) defines accuracy as the closeness of agreement between the value found and the value accepted as either a conventional true value or an accepted reference value. For an assay of a drug substance or drug product, Q2(R2) recommends accuracy be assessed across a minimum of nine determinations covering the specified range of the procedure — commonly executed as three concentration levels with three replicates each, i.e. the classic 80/100/120% design.

Precision (the spread across replicate determinations) and accuracy (recovery) are reported together because a method can be tightly reproducible and still systematically biased — precise but not accurate — and recovery is the only characteristic that catches a consistent offset.

1. Select three spike levels bracketing the routine assay range — typically 80%, 100%, and 120% of the target/label-claim concentration, though wider brackets (e.g., 50–150%) are used for methods that must also support blend uniformity or content-uniformity testing across a broader dose range.
2. Prepare a representative placebo matching every non-active component of the formulation at its labelled proportion — the single hardest step for complex matrices, since a placebo missing a minor excipient (a coating polymer, a flow aid, a flavor system) will understate any matrix interference that excipient causes.
3. Spike each level in triplicate (nine total preparations minimum) using an independently weighed/pipetted amount of reference standard traceable to a qualified reference material.
4. Carry every spiked preparation through the complete method — extraction time, temperature, filtration, dilution, chromatographic run — exactly as written in the SOP; no shortcuts, because the whole point is to challenge the procedure end to end.
5. Calculate percent recovery at each level and the %RSD across the nine (or more) results; assess trend across levels (recovery that declines at high concentration suggests saturation of an extraction or a detector nonlinearity; recovery that declines at low concentration suggests adsorptive loss or a limit-of-quantitation issue).

A matrix effect is any way the sample's non-analyte components change the measured response for a given amount of analyte, relative to how that analyte behaves in a clean solvent. In chromatography-mass spectrometry methods, ion suppression or enhancement from co-eluting lipids, salts, or matrix background is the classic mechanism; in UV/HPLC methods, co-eluting excipient peaks, baseline shift, or incomplete extraction from a coating or a fill matrix are more typical.

The standard-addition technique is the specific tool for matrices too complex to blank out entirely: instead of comparing a spiked placebo to a clean standard curve, the analyst spikes increasing known amounts of reference standard directly into aliquots of the real, unmodified sample, measures each, and extrapolates back to the x-intercept to determine the native concentration — a method that inherently compensates for matrix-driven signal suppression or enhancement because every spiked point sits in the same matrix background as the unspiked sample.

There is no single universal recovery acceptance criterion in ICH Q2(R2) — the guideline requires the acceptance criteria to be justified and pre-specified in the validation protocol, scaled to the difficulty of the matrix and the concentration level being tested. Widely used, defensible starting points by matrix class are shown below; they must still be justified against the specific method's demonstrated performance, not simply asserted from a table.

The AOAC INTERNATIONAL Appendix F framework is the widely referenced source for concentration-dependent recovery and precision expectations in food and supplement testing, since expected recovery and RSD legitimately widen as analyte concentration drops — a low-ppm micronutrient assay simply cannot hit the same tight recovery window as a 100 mg/tablet potency assay, and forcing a pharma-grade tolerance onto a trace analysis produces a method that looks like it's failing when it is behaving exactly as physically expected.

The most common source of systematic low recovery is incomplete extraction, and it shows up differently depending on the dosage form.

• Hard-shell or film-coated tablets — a coating polymer or a slow-dissolving binder can trap active inside a matrix that the extraction solvent and sonication/agitation time do not fully penetrate; recovery studies must challenge realistic (not idealized, pre-crushed) sample presentation.
• Softgel/liquid-fill capsules — the fill matrix (often oil-based) can partition the active away from an aqueous or polar extraction solvent; recovery studies need a solvent system validated specifically against the fill's polarity, and a documented equilibration/mixing time.
• Sustained/controlled-release formulations — by design these resist dissolution; a potency-assay extraction method (which needs full, immediate recovery) must use a different solvent system and mechanical energy than the dissolution test (which deliberately measures a rate), and conflating the two is a frequent validation error.
• Botanical extracts and whole-herb matrices — plant cell walls, waxes, and co-extracted lipids or tannins can bind the marker compound or interfere with its detection; extraction efficiency for botanicals is typically the single biggest driver of recovery variability and should be characterized across multiple raw-material lots, not just one reference lot.

A consistently low or high recovery is not a one-time data point — it is a property of the method against a given matrix, and it will bias every routine result in the same direction until it is found and corrected. That makes recovery data one of the first places an OOS investigation should look when a batch trends unexpectedly low (or a customer complaint alleges under-potency) across multiple lots rather than a single outlier lot.

1. Confirm whether the method's validated recovery range at the relevant concentration actually covers 100% recovery, or whether it has a documented, accepted bias that the reported result is expected to be corrected against.
2. Check whether a formulation, supplier, or process change occurred since the recovery study was last performed — a new excipient supplier, a coating formulation change, or a new botanical extraction ratio can silently invalidate an old recovery justification.
3. Rule out an extraction-step failure specific to the OOS sample (incomplete sonication, wrong solvent lot, filter adsorption) before concluding the product itself is truly sub-potent.
4. Where recovery is suspect, re-run a targeted recovery check alongside the retest, spiking the actual OOS sample matrix if possible, to distinguish a true product failure from a method artifact.