Incurred Sample Reanalysis (ISR) Under ICH M10

Incurred sample reanalysis (ISR) is the step in bioanalytical method validation where a subset of real study samples — not spiked calibration standards or quality control (QC) samples — is analyzed a second time, on a different day, in a separate run. ICH M10, Bioanalytical Method Validation and Study Sample Analysis, devotes a dedicated section to it because a method can pass every validation run built on spiked matrix and still produce results on actual patient or subject samples that don’t reproduce. ISR is the check that catches that gap before it becomes a study-integrity problem.

What Incurred Sample Reanalysis Actually Tests

Every other validation exercise in a bioanalytical method — the calibration curve, accuracy and precision QCs, matrix-effect assessment — is built on spiked matrix: blank plasma, serum, or urine with a known amount of analyte added back in. ICH M10 defines an “incurred sample” simply as “a sample obtained from study subjects or animals,” and defines ISR as “reanalysis of a portion of the incurred samples in a separate analytical run on a different day to determine whether the original analytical results are reproducible.”

That distinction matters because a spiked sample can never fully represent what happens inside a real biological matrix that has been through a living system. ICH M10 states the concern directly: “the performance of study samples may differ from that of the calibration standards and QCs used during method validation, which are prepared by spiking blank matrix.” The guideline names several specific mechanisms:

• Differences in protein binding — endogenous binding proteins in an incurred sample may behave differently than in freshly spiked matrix.
• Back-conversion of known and unknown metabolites — a metabolite present in the real sample can convert back to the parent analyte during sample processing, artificially inflating the measured concentration in a way a spiked QC would never show.
• Sample inhomogeneity — incurred samples are not uniformly mixed the way a prepared QC pool is.
• Concomitant medications — co-administered drugs in a real patient can interfere with the assay in ways a single-analyte spiked sample cannot reveal.
• Biological components unique to the study samples — disease state, matrix lot variability, and other patient-specific factors.

None of these are things a validation run built entirely from spiked, blank-matrix QCs can expose. ISR is the only point in the validation lifecycle where the method is tested against the actual biology it will be reporting on.

When ISR Is Required

ICH M10 does not require ISR on every study. It sets out the situations where it applies “at least”:

• For nonclinical studies within the scope of the guidance, ISR should generally be performed at least once per species.
• All pivotal comparative bioavailability (BA) / bioequivalence (BE) studies.
• The first clinical trial in subjects.
• Pivotal early patient trial(s), once per patient population.
• The first or pivotal trial in patients with impaired hepatic and/or renal function.

The pattern is deliberate: ISR is triggered at the points in a program where the bioanalytical method meets a genuinely new matrix or population — a new species, a new patient population, or organ impairment that can change protein binding and clearance — rather than repeated on every single study once a method has already demonstrated reproducibility in that context.

How Many Samples Require ISR Testing

ICH M10 sets a minimum sampling rule, while leaving room for scientific judgment above that floor: “the extent of ISR depends upon the analyte and the study samples and should be based upon an in-depth understanding of the analytical method and analyte.” The specified floor is:

• Total study samples ≤ 1,000: reanalyze 10% of the samples.
• Total study samples > 1,000: reanalyze 10% of the first 1,000 samples (i.e., 100 samples) plus 5% of every sample beyond 1,000.

Selection isn’t left to convenience. ICH M10 requires objective criteria for choosing the ISR subset to be predefined in the protocol, study plan, or an SOP, and specifies how the subset should be built:

• Subjects or animals should be picked as randomly as possible from the dosed study population.
• Adequate coverage of the concentration profile is required — samples should be chosen around C_max and some in the elimination phase.
• The chosen samples should be representative of the whole study.
• Samples must not be pooled, because pooling can mask an anomalous individual result.
• ISR samples and QCs must be processed and analyzed the same way as in the original analysis.
• ISR must be performed within the analyte’s established stability window, but not on the same day as the original analysis — the point is to test reproducibility across genuinely separate runs, not to rerun the same plate.

This is a distinct exercise from a method’s internal-standard normalization or run-acceptance QC checks, which govern whether an individual analytical run is valid at all — ISR instead asks whether a run that already passed is reproducible when the same real samples are analyzed again.

ISR Acceptance Criteria

ICH M10 defines the pass/fail threshold as a percent difference between the two results, calculated against their mean rather than against either value alone:

% difference = (repeat value − initial value) ÷ mean value × 100

Using the mean as the denominator, rather than treating the original result as the “true” value, is a meaningful design choice: it does not presume the first analysis was correct, only asks how far apart the two independent measurements are.

The acceptance criteria differ by assay platform, reflecting the inherently wider variability of ligand-binding assays (LBAs) compared with chromatographic methods such as LC-MS/MS:

• Chromatographic methods: the percent difference must be within ±20% for at least two-thirds of the repeats.
• Ligand-binding assays: the percent difference must be within ±30% for at least two-thirds of the repeats, consistent with the generally wider precision and accuracy tolerances LBAs carry elsewhere in method validation (see 4PL ELISA standard curve fitting for how that plays out in curve-fitting practice).

Both criteria are two-thirds thresholds, not 100% thresholds — ICH M10 does not require every repeat to fall inside the tolerance, only the specified majority. That is a deliberate acknowledgment that some individual-sample variability is expected biology, not automatically a method failure.

What Happens When ISR Fails

ICH M10 sets out a specific response path rather than leaving failure handling to case-by-case judgment:

• If the overall ISR results fail the acceptance criteria, an investigation must be conducted and the causes remediated, governed by a predefined SOP describing how investigations are triggered and conducted.
• If the investigation cannot identify the cause of the failure, the potential impact of the ISR failure on study validity must still be reported in the Bioanalytical Report — a failure is disclosed even when it cannot be explained.
• Even when ISR technically meets the acceptance criteria, large or systemic differences across multiple samples are still worth investigating — the guidance flags patterns such as every ISR sample from one subject failing, or every ISR sample from one analytical run failing, as trends of concern that a bare pass rate can obscure.
• Individual outliers — ICH M10 specifically mentions results more than 50% different from the original, sometimes called “flyers” — do not on their own trigger reanalysis of the original sample and do not need separate investigation, provided the overall two-thirds criterion is otherwise met.
• ISR data never replaces the original study sample result. The original bioanalytical result stands as the reported concentration regardless of the ISR outcome; ISR is a reproducibility check layered on top, not a re-measurement that supersedes the study record.

Every aspect of the ISR evaluation — the predefined selection criteria, the raw repeat data, the percent-difference calculations, and any investigation triggered by a failure — must be documented well enough to let an inspector reconstruct the study afterward. That documentation burden sits alongside the same evidentiary standard applied to a study’s broader validation master plan and its closing validation summary report.

Why ISR Sits Apart From Routine Method Validation

It’s worth being precise about what ISR is not. It is not a repeat of accuracy and precision testing, and it is not a substitute for the matrix-effect or selectivity assessments already required during method validation using spiked coefficient-of-variation (CV) and accuracy limits. Those exercises establish that the method works on matrix in principle; ISR establishes that it keeps working on the actual samples the study generated. A method can clear every spiked-QC criterion and still fail ISR if, for example, an unanticipated metabolite in a specific patient population back-converts during extraction — a failure mode spiked samples structurally cannot reveal because there is no metabolite to back-convert in blank matrix with analyte added.

This is also why ISR requirements are tied to genuinely new bioanalytical contexts — a new species in nonclinical work, a new patient population, organ-impaired subjects — rather than to every study a sponsor runs. Once a method has demonstrated ISR reproducibility in a given matrix and population, later studies in that same context don’t automatically need it repeated; the guidance is targeting the transitions where something about the sample population could plausibly behave differently from what validation already characterized.

Frequently Asked Questions

Is ISR the same as re-running a failed analytical run?

No. A failed run is rejected and reanalyzed as part of routine run-acceptance criteria. ISR is performed on samples from runs that already passed, specifically to confirm that a passing result is reproducible when the same incurred sample is measured again on a different day.

Does ISR apply to every clinical trial in a drug development program?

Not automatically. ICH M10 ties the requirement to specific transition points — the first clinical trial in subjects, pivotal comparative BA/BE studies, the first pivotal trial in each new patient population, and the first or pivotal trial in hepatically or renally impaired patients — rather than mandating it on every protocol.

What is a “flyer” in ISR, and does it need investigating?

ICH M10 uses the term informally for an individual repeat result that differs from the original by more than 50%. A single flyer does not, on its own, trigger reanalysis of the original sample or a formal investigation, provided the overall two-thirds acceptance criterion across the ISR subset is still met.

Can ISR results be used to correct or replace the originally reported concentration?

No. ICH M10 is explicit that ISR sample data should not replace the original study sample data. The original result remains the reported value; ISR only assesses whether that value is reproducible.

Why do ligand-binding assays get a wider ISR tolerance than chromatographic methods?

ICH M10 sets the LBA tolerance at ±30% versus ±20% for chromatographic methods, reflecting the generally wider inherent variability of ligand-binding platforms — the same distinction that shows up in other LBA-specific accuracy and precision criteria throughout the guideline.

Primary source: ICH M10, Bioanalytical Method Validation and Study Sample Analysis (Step 4, adopted May 2022; issued as FDA guidance November 2022), Section V, “Incurred Sample Reanalysis.”