LOD vs LOQ: How to Calculate Both

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 ≈ 3:1	Signal-to-noise ratio ≈ 10:1
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	Not typically reported alone in this context	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.

Both are accepted by ICH Q2(R2) provided the choice is justified and applied consistently; neither is universally “more correct.” They frequently produce different numbers on the same underlying data, for reasons explained after the worked calculations below.

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 systematic gap, ICH Q2(R2) and most method-validation SOPs require that whichever LOD/LOQ value is proposed, it must be confirmed experimentally by actually preparing and injecting a sample at that concentration and demonstrating the analyte is genuinely detectable (at LOD) or quantifiable within acceptable precision and accuracy (at LOQ) — neither formula is accepted as a final answer on its own.

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	Defines LOD and LOQ as validation characteristics required for impurity/limit tests and quantitative assays respectively
US EPA 40 CFR Part 136, Appendix B	Method Detection Limit (MDL) procedure for environmental/wastewater testing	A related but distinct calculation: replicate (commonly n=7) low-level spiked samples, standard deviation of results, multiplied by a Student’s t-value for n−1 degrees of freedom at 99% confidence — 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	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	Always prepare and inject an actual standard at the calculated LOD/LOQ concentration and confirm it is genuinely detectable/quantifiable, per ICH Q2(R2)
Blank standard deviation (σ) is unstable between replicate runs	Too few replicates (n=3 or fewer) used to estimate σ	Use at least 6–10 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
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.

Related CASRAI guides

LOD/LOQ calculations sit alongside several other analytical and instrument-operation topics covered elsewhere on CASRAI:

Limit of Detection vs Limit of Quantitation: How to Calculate LOD and LOQ

Ask about Limit of Detection vs Limit of Quantitation: How to Calculate LOD and LOQ