Written and maintained by CASRAI Editorial Board
Published · Last updated
On this page
The limit of detection (LOD) and limit of quantitation (LOQ) are the two thresholds that define how low a concentration an analytical method can reliably see. They are often quoted together and just as often confused. LOD is the lowest amount of an analyte a method can distinguish from background noise with reasonable confidence — it tells you a substance is present. LOQ is the lowest amount the same method can measure with acceptable precision and accuracy — it tells you how much is present, reliably enough to report a number. A result between LOD and LOQ is real but not quantifiable; below LOD, you cannot distinguish signal from noise at all.
Both figures are method-specific, not properties of the analyte alone. The same compound will have a different LOD/LOQ on a different instrument, column, detector, or sample matrix. This guide compares the two terms directly, then works through both standard calculation routes — the blank/calibration-curve standard-deviation method (3σ/10σ) and the signal-to-noise ratio method — on a single shared dataset so the resulting numbers can be compared side by side.
LOD vs LOQ: side-by-side comparison
| Dimension | Limit of Detection (LOD) | Limit of Quantitation (LOQ) |
|---|---|---|
| What it answers | Is the analyte present at all? | How much of the analyte is present, reliably? |
| Typical multiplier (blank-SD / regression method) | 3.3 × σ / S | 10 × σ / S |
| Typical multiplier (signal-to-noise method) | Signal-to-noise ratio of 3:1 (generally acceptable, ICH Q2(R2)) | Signal-to-noise ratio of at least 10:1 (ICH Q2(R2)) |
| Precision/accuracy requirement | None — detection only, no quantitative claim | Demonstrated acceptable precision (%RSD) and accuracy (%recovery) at that concentration |
| Primary regulatory/standards basis | ICH Q2(R2) Validation of Analytical Procedures, read alongside ICH Q14 Analytical Procedure Development | ICH Q2(R2); USP General Chapter <1225> Validation of Compendial Procedures |
| Environmental/regulatory variant | The MDL is the nearest equivalent; see the section on the method detection limit below | US EPA 40 CFR Part 136, Appendix B defines a related but distinct Method Detection Limit (MDL), calculated from replicate low-level spikes and a Student’s t multiplier rather than a calibration-curve slope |
| Relationship between the two | LOQ is always numerically higher than LOD for the same method and dataset — typically LOQ ≈ 3 × LOD when both are derived from the same σ/S ratio, because 10/3.3 ≈ 3.03 | |
Two ways LOD and LOQ get calculated
In practice, laboratories use one of two calculation routes, and the two do not always agree with each other on the same data:
- Blank/calibration-curve standard-deviation method (the “3σ/10σ” or “3.3σ/10σ” method). Run a set of low-concentration standards (or replicate blanks), fit a linear calibration curve, and use the standard deviation of the response together with the slope of that curve. This is the method ICH Q2(R2) describes as the regression-based approach and is the most common route for chromatographic and spectroscopic methods being formally validated.
- Signal-to-noise ratio method. Inject or measure a low-concentration sample, compare the height of the analyte’s signal to the peak-to-peak (or RMS) baseline noise measured in a blank region near the same retention time or wavelength, and find the concentration at which that ratio reaches 3:1 (LOD) or 10:1 (LOQ). This route is common for chromatographic detectors with a visible, measurable baseline (UV, fluorescence, most mass spec scan modes) and is often faster to apply during method development, before a full validation-grade calibration curve is available.
ICH Q2(R2) lists both among its approaches (alongside visual evaluation and direct accuracy-and-precision validation, covered below); neither is universally “more correct,” and the approach used should be stated when the limit is reported. They frequently produce different numbers on the same underlying data, for reasons explained after the worked calculations below.
LOD and LOQ formulas at a glance
The most common calibration-based formulas, from ICH Q2(R2), are LOD = 3.3 × σ / S and LOQ = 10 × σ / S, where σ is the standard deviation of the response and S is the slope of the calibration curve. For the signal-to-noise route, LOD is the concentration giving a ratio of about 3:1 and LOQ the concentration giving at least 10:1. Report both in concentration units and state which approach you used.
| Approach | LOD | LOQ | What you need |
|---|---|---|---|
| Standard deviation and slope | 3.3 × σ / S | 10 × σ / S | A calibration curve and σ (blank SD, residual SD of the regression, or SD of the y-intercepts) |
| Signal-to-noise | S/N of 3:1 | S/N of at least 10:1 | A chromatogram or trace with a measurable baseline |
How to calculate LOD and LOQ from a calibration curve
Using the residual standard deviation of the regression line as σ needs only the calibration data you already have. The figures below are an illustrative dataset constructed to show the arithmetic, not a published measurement.
- Fit the line. Six standards at 0, 10, 20, 30, 40 and 50 ng/mL gave responses of 0.02, 1.55, 3.08, 4.46, 6.07 and 7.52 units. Least-squares regression gives a slope S = 0.1498 units per ng/mL and an intercept of 0.038.
- Find the residual standard deviation. Subtract each fitted value from its measured response, square the differences, sum them, and divide by n − 2 (here 4 degrees of freedom) before taking the square root. This gives σ = 0.0487 units. In Excel this is the standard error of the y estimate from LINEST; in R it is the sigma value in the summary of the fitted model.
- Apply the formulas. LOD = 3.3 × 0.0487 / 0.1498 = 1.07 ng/mL. LOQ = 10 × 0.0487 / 0.1498 = 3.25 ng/mL.
- Check the result. Both limits sit below the lowest non-zero standard (10 ng/mL), so the curve does not bracket them. Prepare samples at or near the calculated limits and confirm they behave as expected, as ICH Q2(R2) describes, ideally with calibration standards in the low range. See building an analytical calibration curve for choosing standards and checking linearity.
Worked example: shared dataset
The figures below are an illustrative worked example built to demonstrate the arithmetic, not measurements from a specific published study or a specific commercial instrument — treat the units as generic response units for a chromatographic detector (e.g. peak area or peak height in absorbance units).
- Six replicate injections of a low-concentration analyte standard, run across the range 0–50 ng/mL, produce a linear calibration curve with slope S = 0.15 response units per ng/mL.
- Six replicate blank injections give a standard deviation of the blank response, σ = 0.42 response units.
- On the same instrument, baseline peak-to-peak noise measured over a blank stretch near the analyte’s retention time is 0.06 mAU.
- An injected standard at 10 ng/mL produces a peak height of 0.60 mAU, and the peak height response is linear (proportional to concentration) across this low range.
Calculation 1: blank-standard-deviation / regression method
Formula: LOD = 3.3 × (σ / S); LOQ = 10 × (σ / S)
Working:
- LOD = 3.3 × (0.42 / 0.15) = 3.3 × 2.8 = 9.24 ng/mL
- LOQ = 10 × (0.42 / 0.15) = 10 × 2.8 = 28.0 ng/mL
Calculation 2: signal-to-noise ratio method
Formula: find the concentration C at which (peak height at C) / (baseline noise) equals the target ratio.
Working:
- At 10 ng/mL, S/N = 0.60 mAU / 0.06 mAU = 10.0 — this concentration is the S/N = 10 point, so LOQ (S/N method) = 10 ng/mL.
- Because response is linear in this range, the concentration giving S/N = 3 scales proportionally: 10 ng/mL × (3 / 10) = LOD (S/N method) = 3.0 ng/mL (peak height 0.18 mAU ÷ 0.06 mAU noise = 3.0, confirming the extrapolation).
Why the two routes gave different numbers
On this shared dataset, the blank-SD/regression method gives LOD = 9.24 ng/mL and LOQ = 28.0 ng/mL; the signal-to-noise method gives LOD = 3.0 ng/mL and LOQ = 10.0 ng/mL. The two routes are not measuring identical things, so a mismatch like this is expected rather than a sign either calculation is wrong:
- The blank-SD method captures the full run-to-run variability of the blank response across six separate injections — a broader, noisier estimate of “noise” that includes injection-to-injection variation, not just the instantaneous baseline.
- The signal-to-noise method captures only the instantaneous baseline ripple in a single trace — a narrower, usually smaller estimate of noise, which is why it tends to produce lower (more optimistic) LOD/LOQ values.
- Because of this gap, a calculated value should be checked against real samples. ICH Q2(R2) says an estimated detection limit can subsequently be validated, and that a quantitation limit should be subsequently validated by analysing a suitable number of samples near or at the QL. Neither formula is a final answer on its own.
What ICH Q2(R2) says about detection and quantitation limits
ICH Q2(R2) Validation of Analytical Procedures reached Step 4 on 1 November 2023, and the US FDA published its final Q2(R2) and Q14 guidances on 7 March 2024 (docket FDA-2022-D-1503). In section 3.2.3 the guideline groups the detection limit (DL) and quantitation limit (QL) under “lower range limits” and describes four approaches:
- Visual evaluation (3.2.3.1): analysing samples with known concentrations and establishing the level at which the analyte can be reliably detected or quantified.
- Signal-to-noise (3.2.3.2): a ratio of 3:1 is generally acceptable for the DL and a ratio of at least 10:1 for the QL. The region used to measure noise should be defined and, where possible, centred on the position of the analyte peak.
- Standard deviation of the response and the slope (3.2.3.3): DL = 3.3σ/S and QL = 10σ/S, where S is the slope of the calibration curve.
- Accuracy and precision at the lower range limits (3.2.3.4): directly demonstrating acceptable accuracy and precision at the proposed limit.
The guideline treats confirmation differently for the two limits. An estimated DL can subsequently be validated; a QL should be subsequently validated by analysing a suitable number of samples near or at the QL. That confirmation can be omitted, with justification, where the QL is about ten times below the reporting limit. For impurity tests the QL should be equal to or below the reporting threshold. Section 3.2.3.5 lists what should be recorded: the approach used, the data and the resulting limits.
How this differs from ICH Q2(R1)
Q2(R1) said a signal-to-noise ratio of “3 or 2:1” was generally acceptable for the detection limit and described 10:1 as a typical ratio for the quantitation limit. Q2(R2) drops the 2:1 option, states 3:1 for the DL and “at least 10:1” for the QL, and uses the “lower range limits” heading. The data table below the article compares these factors with the EPA and clinical conventions.
Which sigma should you use?
Q2(R2) allows the standard deviation of the response to be estimated from the standard deviation of blank measurements, the residual standard deviation of the regression line, or the standard deviation of y-intercepts of regression lines. The worked example above uses the blank standard deviation. Whichever is chosen, σ must be in response units and S in response units per unit concentration, so that the ratio σ/S has concentration units. Report which estimate was used, because the three can give different limits on the same data.
Why the factor is 3.3 and where 10 comes from
The 3.3 follows from hypothesis-testing logic described by Boqué and Vander Heyden in LCGC International. If the false-positive and false-negative risks are both set at 5% (α = β = 0.05), the critical level is about 1.64σ0 and the detection limit is about 3.3σ0, that is, 1.64 + 1.64. Setting the decision threshold at zero instead would mean that half of true blanks exceed it, a 50% false-positive risk. The same article notes that when σ is itself estimated from a small number of replicates, a Student’s t-value should replace the normal-distribution factor, and that at least ten portions of a low-level test sample should be carried through the whole procedure. The factor of 10 for quantitation is a convention: it places the limit far enough above the noise that the relative uncertainty of a result is acceptably small. The article also advises stating the approach used when an LOD is reported.
Limit of blank, limit of detection and limit of quantitation in clinical laboratories
Clinical laboratory practice, following CLSI EP17, adds a third term. As summarised by Armbruster and Pry (Clinical Biochemist Reviews, 2008):
- Limit of blank (LoB) = mean of blank results + 1.645 × SD of blank results. It is the highest result expected from a sample containing no analyte, at 95% confidence.
- Limit of detection (LoD) = LoB + 1.645 × SD of a low-level sample. It is the lowest concentration reliably distinguished from the LoB.
- Limit of quantitation (LoQ) = the lowest concentration at which predefined goals for bias and imprecision are met. It may equal the LoD or be much higher.
The suggested replicate numbers are 60 to establish a limit (a manufacturer’s task) and 20 to verify a claimed limit (a laboratory’s task). So the LoB is the answer to “what does a blank look like?”, and the LoD answers “how far above that must a real sample sit?” This differs from ICH’s 3.3σ/S, which works from calibration-curve behaviour rather than from replicate blank and low-level samples.
The EPA method detection limit, step by step
40 CFR Part 136 Appendix B (Revision 2, 82 FR 40939, 28 August 2017) defines the MDL as the minimum measured concentration that can be reported with 99% confidence that the result is distinguishable from method blanks. The procedure is:
- Estimate an MDL and spike reagent water or matrix, typically at 2 to 10 times that estimate.
- Analyse at least seven spiked samples and at least seven method blanks, spread over at least three batches on three dates.
- Calculate MDLs = t(n−1, 0.99) × Ss, where Ss is the standard deviation of the spiked results (t = 3.143 for seven replicates, six degrees of freedom).
- Calculate MDLb = mean of the method blanks + t × Sb, where Sb is the standard deviation of the blanks.
- The initial MDL is the greater of MDLs and MDLb.
Note that, unlike ICH, this procedure needs no calibration slope: variability is measured directly on spiked samples that pass through the full method.
How to report a detection or quantitation limit
- State the approach (visual, signal-to-noise, σ/S, or accuracy and precision) and, for σ/S, which sigma estimate was used.
- State the number of replicates and the matrix of the samples.
- For signal-to-noise, say how and where noise was measured, ideally in a defined region centred on the peak position.
- Give the confirmation data (samples at or near the limit), or the justification for omitting it.
- For impurity methods, show that the QL is at or below the reporting threshold.
Regulatory and standards context
| Standard / body | What it governs | Notes |
|---|---|---|
| ICH Q2(R2), Validation of Analytical Procedures | Pharmaceutical method validation, internationally harmonised (ICH member regulators including FDA, EMA) | Revised and finalised alongside the new ICH Q14 (Analytical Procedure Development), which introduces the Analytical Target Profile concept for defining method performance requirements, including LOD/LOQ, up front |
| USP General Chapter <1225> | Validation of compendial (pharmacopeial) analytical procedures in the US | Compendial validation chapter; check the current USP text for which performance characteristics it requires for a given test category |
| US EPA 40 CFR Part 136, Appendix B | Method Detection Limit (MDL) procedure for environmental/wastewater testing | A related but distinct calculation: at least seven spiked samples and seven method blanks, in at least three batches on three dates; the standard deviation of the spiked results is multiplied by a one-sided Student’s t-value at 99% confidence (3.143 for seven replicates) — not the same formula as the ICH 3.3σ/S approach, and the two should not be used interchangeably |
| CLSI EP17 | Detection capability (LoB, LoD, LoQ) for clinical laboratory (diagnostic) methods | The guideline is Evaluation of Detection Capability for Clinical Laboratory Measurement Procedures, EP17-A2 (2nd edition, 18 June 2012, reaffirmed June 2017). Clinical chemistry uses a three-tier framework: Limit of Blank (LoB), Limit of Detection (LoD), and Limit of Quantitation (LoQ), each with its own statistical derivation from blank and low-level replicate samples |
Common pitfalls and troubleshooting
| Symptom | Likely cause | Fix |
|---|---|---|
| Calculated LOD/LOQ looks unrealistically low | Signal-to-noise ratio calculated from a single, unusually clean trace, or noise measured too far from the analyte’s retention time/wavelength | Measure baseline noise directly adjacent to the analyte peak, across multiple traces, and cross-check against the blank-SD/regression method |
| LOD/LOQ from the calibration curve doesn’t hold up when a real low-level sample is injected | The value was extrapolated mathematically but never confirmed experimentally | Prepare and inject samples at or near the calculated LOD/LOQ and confirm they are detectable/quantifiable, as ICH Q2(R2) describes (see the Q2(R2) section below for when this can be justified as unnecessary) |
| Blank standard deviation (σ) is unstable between replicate runs | Too few replicates (n=3 or fewer) used to estimate σ | Use enough replicate blank or low-level measurements; report the number of replicates used alongside the LOD/LOQ value |
| LOQ is reported but the method fails precision/accuracy checks at that concentration | LOQ was set purely from the S/N = 10 or 10σ/S formula without verifying %RSD or %recovery | LOQ requires a demonstrated acceptable precision and accuracy at that concentration, not just a signal threshold — run replicate quantitation at the proposed LOQ and confirm both criteria before finalizing it |
| Instrument detection limit (IDL) is quoted as if it were the method LOD | IDL is measured on a clean standard in pure solvent, with no sample preparation or matrix | Report a method-based LOD/LOQ, derived from samples that have gone through the full extraction/preparation/matrix workflow — matrix effects almost always raise the real-world LOD/LOQ above the solvent-only IDL |
| An LOD/LOQ value from one instrument is carried over unchanged to a different instrument, column, or reagent lot | LOD/LOQ are treated as if they were a fixed property of the analyte | Re-verify LOD/LOQ whenever the instrument, column, detector, reagent lot, or sample matrix changes — they are properties of the whole method, not the compound alone |
| Sigma and slope in mismatched units | σ taken in response units but the slope S expressed per a different concentration unit (or vice versa) | Express σ in response units and S in response units per unit concentration, so that σ/S comes out in concentration units, as in the worked example (0.42 ÷ 0.15 = 2.8 ng/mL) |
| Peak height and peak area are mixed between the calibration curve and the noise measurement | Inconsistent response metric used across the two inputs to the calculation | Use the same response metric (height or area) consistently throughout both the calibration curve and the noise/S/N measurement |
Frequently asked questions
Is LOQ always higher than LOD?
Yes. LOQ is, by definition, a stricter threshold than LOD — it requires not just detecting the analyte but measuring it with acceptable precision and accuracy. When both are derived from the same σ/S ratio, LOQ works out to roughly three times LOD, since 10 ÷ 3.3 ≈ 3.03.
What is limit of quantitation, specifically?
The limit of quantitation (LOQ) is the lowest concentration of an analyte in a sample that can be quantitatively determined with acceptable precision (repeatability, expressed as %RSD) and accuracy (expressed as %recovery), under the stated experimental conditions of the method. It is distinct from LOD, which only requires that the analyte’s presence be distinguishable from background noise, with no requirement that the amount be measured reliably.
Can I use the 3.3σ/S formula and the signal-to-noise method interchangeably?
They will usually give different numbers on the same data, for the reasons set out above, so they should not be mixed within a single validation report. Pick one method, document why, apply it consistently, and confirm the resulting LOD/LOQ experimentally rather than relying on either formula alone.
Does a higher signal-to-noise ratio always mean a better method?
Not directly — S/N at a given concentration reflects how far above baseline noise that concentration’s signal sits, which is one input into LOD/LOQ, but overall method quality also depends on selectivity, linearity across the full working range, and precision/accuracy at LOQ and above, not S/N alone.
Is Method Detection Limit (MDL) the same thing as LOD?
No, not exactly, though the two are related and sometimes used loosely as synonyms. MDL, as defined in US EPA 40 CFR Part 136 Appendix B, is calculated from replicate low-level spiked samples using a Student’s t-value at 99% confidence, whereas the ICH-style LOD is typically calculated from the standard deviation of the response and the slope of a calibration curve (3.3σ/S) or from a signal-to-noise ratio. Both estimate a similar concept — the lowest reliably detectable amount — but via different statistics, and they are not numerically interchangeable.
Why is the LOD formula 3.3σ/S instead of 3σ/S?
With 5% false-positive and 5% false-negative risks, the critical level is about 1.64σ and the detection limit is about twice that, 3.3σ (Boqué and Vander Heyden, LCGC International). A bare 3 is a rounded rule of thumb.
What signal-to-noise ratio does ICH Q2(R2) accept?
3:1 is generally acceptable for the detection limit and at least 10:1 for the quantitation limit. The older Q2(R1) text also allowed 2:1 for detection.
Do I have to confirm a calculated LOD or LOQ experimentally?
Q2(R2) says a DL estimate can subsequently be validated and a QL should be subsequently validated with a suitable number of samples near or at the QL; confirmation may be omitted with justification if the QL is about ten times below the reporting limit.
How many replicates do I need?
ICH Q2(R2) does not set a fixed number in the sections cited here. The EPA MDL needs at least seven spiked and seven blank samples in at least three batches; clinical LoB/LoD protocols suggest 60 replicates to establish and 20 to verify; the LCGC derivation suggests at least ten portions of a low-level sample.
What is the limit of blank?
The LoB is the mean blank result plus 1.645 standard deviations of the blank. The LoD is then LoB plus 1.645 standard deviations of a low-level sample, so LoD always sits above LoB.
Which sigma should I use?
Q2(R2) permits the standard deviation of blank responses, the residual standard deviation of the regression, or the standard deviation of y-intercepts. State which one you used.
Sources
Sources (accessed 2026-11-13):
- ICH Q2(R2) Validation of Analytical Procedures, Step 4, 1 November 2023: database.ich.org
- ICH Q2(R1) Validation of Analytical Procedures: Text and Methodology: database.ich.org
- 40 CFR Part 136 Appendix B, Revision 2: ecfr.gov
- Armbruster DA, Pry T. Limit of blank, limit of detection and limit of quantitation. Clin Biochem Rev 2008: PMC2556583
- Boqué R, Vander Heyden Y. The Limit of Detection. LCGC International: chromatographyonline.com
- CLSI EP17 product page: clsi.org
Related CASRAI guides
LOD/LOQ calculations sit alongside several other analytical and instrument-operation topics covered elsewhere on CASRAI:
- Atomic Absorption Spectroscopy: Flame vs Graphite Furnace Detection Limits — a worked comparison of detection-limit ranges across AAS modes
- HPLC Method Development: A Stepwise Workflow – where LOD and LOQ targets are set during development
- Beer’s Law Calibration Plot: Guide & Standards – the calibration line behind a spectrophotometric LOD
- Gas Chromatography: Columns, Carrier Gases & Detectors
- LC-MS Explained: Liquid Chromatography-Mass Spectrometry Coupling
- UV-Vis Spectrophotometer Basics: How It Works and Measures Concentration
- Spectrophotometer Calibration: Wavelength and Photometric Accuracy Checks
- Analytical Balance Calibration and Proper Weighing Technique
- Serial Dilution Technique: How to Perform and Calculate
- Molarity and Solution Calculations for the Lab
- ELISA Protocol Basics: Step-by-Step Overview
- Western Blot Protocol Basics: Step-by-Step Walkthrough
Further reading: ICH Stability — ICH Q1 is ICH’s stability-testing guideline series — how shelf-life, retest periods, and storage conditions are established, the still-in-progress consolidated Q1 revision, and its relevance to investigational product dating in clinical trials.
Data: LOD and LOQ estimation approaches compared: factors, replicates and confirmation rules by guideline
Compares how ICH Q2(R1), ICH Q2(R2), US EPA 40 CFR 136 Appendix B (MDL), CLSI-style LoB/LoD/LoQ protocols and the statistical derivation in LCGC International set the detection and quantitation limit factors, the minimum number of replicates and the confirmation expectations. Compiled by hand from the guideline texts and one peer-reviewed or trade-journal explanation of each statistical approach; factors are numeric multipliers or ratios as stated in the sources.
Period covered: 1994/2023.
| Approach (source) | Detection factor (multiplier or S/N ratio) | Quantitation factor (multiplier or S/N ratio) | Minimum replicates stated (replicates) | How the value is confirmed or reported |
|---|---|---|---|---|
| ICH Q2(R1) signal-to-noise | 3 | 10 | not stated | Q2(R1) says an S/N between 3 or 2:1 is generally acceptable for DL and a typical ratio is 10:1 for QL; the estimate is later validated by analysing samples near the limit |
| ICH Q2(R2) signal-to-noise, section 3.2.3.2 | 3 | 10 | not stated | 3:1 generally acceptable for DL; at least 10:1 acceptable for QL; S/N region should be defined and, if possible, centred on the peak position |
| ICH Q2(R2) standard deviation and slope, section 3.2.3.3 | 3.3 | 10 | not stated | DL = 3.3 sigma/S and QL = 10 sigma/S; sigma from blank SD, residual SD of regression, or SD of y-intercepts; estimates validated on samples at or near the limit; QL confirmation may be omitted with justification if QL is about 10 times below the reporting limit |
| US EPA 40 CFR 136 Appendix B (MDL), Revision 2 | 3.143 | not defined | 7 | MDL = t(n-1, 0.99) x SD of spiked replicates (3.143 for 7 replicates, 6 degrees of freedom); at least 7 spiked and 7 method blanks in at least 3 batches on 3 dates; the greater of MDLs and MDLb is the initial MDL |
| CLSI-style LoB/LoD/LoQ (Armbruster and Pry 2008, following CLSI EP17) | 1.645 | not defined | 20 | LoB = mean blank + 1.645 SD blank; LoD = LoB + 1.645 SD of low-level sample; 60 replicates to establish (manufacturers), 20 to verify (laboratories); LoQ is the lowest level meeting predefined bias and imprecision goals |
| Statistical derivation with alpha = beta = 0.05 (Boque and Vander Heyden, LCGC 2009) | 3.3 | not defined | 10 | Critical level 1.64 sigma0, detection limit about 3.3 sigma0; at least 10 portions of a low-level test sample through the full procedure; state the approach used when reporting an LOD |
Sources
- ICH Q2(R2) Validation of Analytical Procedures (Step 4, adopted 1 November 2023) (accessed 2026-11-13)
- ICH Q2(R1) Validation of Analytical Procedures: Text and Methodology (accessed 2026-11-13)
- 40 CFR Part 136 Appendix B, Definition and Procedure for the Determination of the Method Detection Limit, Revision 2 (accessed 2026-11-13)
- Armbruster DA, Pry T. Limit of blank, limit of detection and limit of quantitation. Clin Biochem Rev 2008 (accessed 2026-11-13)
- Boque R, Vander Heyden Y. The Limit of Detection. LCGC International (accessed 2026-11-13)
Compiled by CASRAI from the sources above. Free to reuse under CC BY 4.0 with attribution.







