Detection and Quantification Limits

Introduction

“Detected” and “measured accurately” are different claims. A faint chromatographic peak may be distinguishable from a blank but too noisy for a useful concentration estimate. Conversely, a report of “not detected” does not mean the analyte concentration is exactly zero. Detection and quantification limits express what a particular method can say about small amounts under defined sample and measurement conditions.

Core explanation

The blank has a distribution of readings, not one immutable value. Instrument noise and contamination make repeated blanks vary. A detection decision compares a sample response with that distribution and chooses a tolerated risk of falsely calling a blank “positive.” The IUPAC Gold Book definition of detection limit relates a threshold to blank mean, blank standard deviation and a factor selected for desired confidence. More elaborate frameworks distinguish the decision threshold for an observed signal from the true analyte amount that is likely to be detected. State which definition and assumptions are used rather than treating a factor such as three as universal.

The quantification limit asks a stricter question: at what level can the procedure give a numeric amount with acceptable uncertainty for its purpose? IUPAC defines LOQ through fitness for purpose, such as a specified repeatability or measurement-uncertainty requirement. Laboratories sometimes estimate it as a multiple of blank standard deviation divided by calibration slope, but the multiplier and validation must be stated. The lower bound can depend on matrix, extraction recovery and dilution as well as detector noise.

Suppose an instrument gives signal = intercept + slope × concentration . Large slope means a small concentration change creates a larger signal; smaller blank variation makes faint amounts easier to distinguish. A rough concentration-scale threshold is k × blank standard deviation / slope , with the important caveat that the correct procedure-specific threshold includes uncertainty in calibration, sample preparation and decision errors. A method validated in pure solvent may have a much higher practical limit in a soil extract because background and matrix suppression differ.

A reported value below the quantification limit may still carry information. It may be reasonable to say “analyte detected, estimated concentration below the validated LOQ” if the signal meets a defined detection decision. Do not report many precise digits as though the method has validated them. If signal does not exceed the detection threshold, report “below LOD” or a method-specific upper bound, not “zero.” In environmental monitoring, health or safety decisions, the threshold must be compared with the concentration of concern; a method with a high LOD cannot verify that a sample is below a much lower limit.

False positives and false negatives trade off. Lowering the detection threshold catches more weak signals but increases blank misclassifications. Raising it reduces false positives but misses more real low concentrations. Replicate sampling, better blanks, improved signal-to-noise and a more selective method can improve performance. Avoid repeating the same compromised sample until one run happens to cross the threshold; predefine how replicates are combined.

The NIST discussion of qualitative detection and quantitative determination emphasizes that critical level, detection limit and determination limit answer different questions. This distinction protects against the common mistake of applying one number to presence, absence and accurate concentration interchangeably.

Step-by-step reasoning

Define the analyte, matrix and preparation. Measure enough independent blanks to estimate background and its variation, and verify stability over the run. Calibrate response near the low end using appropriate standards. Choose acceptable false-positive risk and a quantification uncertainty criterion. Validate the resulting thresholds with low-level samples or spikes in the real matrix. Compare unknown signals against the correct threshold and report below-limit results using language matched to the evidence.

Visual explanation

Draw overlapping bell-shaped distributions of blank signals and low-concentration sample signals. A vertical decision line marks the detection threshold. Blank area to its right illustrates false positives; real-sample area to its left illustrates false negatives. Farther right, draw a quantification boundary where the relative uncertainty becomes acceptable. The overlap makes clear why a threshold is a risk-based convention, not a point where molecules suddenly become present.

Real-world analogy

At a noisy gathering, one may hear a faint voice well enough to know someone spoke but not well enough to transcribe the sentence. A louder voice can be both detected and measured or understood reliably. Background noise and microphone sensitivity determine the threshold. Chemical detection likewise depends on instrument, matrix and decision rules, not only the absolute amount of analyte.

Real-world example

A laboratory tests a water sample for a pesticide. The chromatogram has a small peak at the expected time, but a similar signal sometimes appears in procedural blanks. A second identifying ion ratio supports presence, yet the estimated concentration is below the validated quantification limit. The laboratory reports a qualified detection rather than a precise concentration, checks blank contamination and proposes a more sensitive confirmatory method if a regulatory decision depends on a lower threshold.

Why?

Why does a lower blank standard deviation improve detection? It narrows the range of signals expected when no analyte is present. A small true response can then stand out with less overlap between blank and sample distributions. Increasing instrument sensitivity can also help, but only if it does not increase noise or matrix interference by a similar amount.

Common misconception

“Below detection limit means no analyte exists.” It means the procedure did not resolve a signal under its chosen conditions. “Three times noise is the universally correct LOD” ignores assumptions about blank distributions, false-negative risk and matrix. “Any detected peak can be reported as an exact concentration” confuses detection with validated quantification.

Worked example

A method has blank-signal standard deviation 0.006 absorbance unit and a calibration slope of 0.12 absorbance unit per mg/L. If the laboratory uses a provisional 3s signal criterion, the net signal threshold is 3 × 0.006 = 0.018 and the corresponding approximate concentration is 0.018/0.12 = 0.15 mg/L . A provisional 10s quantification criterion would correspond to 0.060/0.12 = 0.50 mg/L . These are illustrative method estimates, not universal definitions; the lab must validate them with low-level matrix samples and state its chosen risk and precision criteria.

Quick check

1. What is wrong with replacing every non-detect by zero in a concentration data set? Answer: A non-detect can contain analyte below the method's detection capability. Replacing it by zero biases summaries and hides the method limit; use a censored-data approach suited to the analysis.

Exam focus

Distinguish a decision about presence from a reliable numeric amount. State blank mean, variation, slope and chosen threshold criterion when computing a simple limit. Carry units through the calculation. Specify whether limits apply to instrument solution or original sample after extraction and dilution. Report a result below LOQ with a qualifier, not unsupported precision.

Advanced insight

Detection limits are decision-theoretic. The cost of a false alarm and the cost of missing a hazardous analyte may differ, so a threshold appropriate for exploratory screening may not suit compliance testing. Selectivity also matters: a low numerical LOD is not useful if another compound produces the same signal. Confirmatory identification and measurement uncertainty may be more important than pushing a single threshold lower.

Summary

Detection limits describe when a signal can be distinguished from background under stated error assumptions; quantification limits describe when a number is fit for its intended use. Both depend on the complete method, including blank variability, calibration and matrix. A non-detect is not zero, and a faint detected signal is not automatically a precise concentration. State definitions, conditions and reporting qualifiers.

Practice questions

1. A method has an LOQ of 2 µg/L, and a sample is estimated at 1 µg/L with a signal above its LOD. How should it be described? Answer: It may be reported as detected below the validated quantification limit, with an explicitly qualified estimate if policy allows. It should not be reported as a confidently quantified 1.000 µg/L.

2. Blank variation doubles while calibration slope stays fixed. What happens to a simple k s/slope concentration threshold? Answer: It doubles. More variable background makes small analyte amounts harder to distinguish under the same threshold rule.

3. Why might the method LOD in a soil extract differ from the instrument LOD in pure solvent? Answer: Extraction recovery, dilution, matrix interference and procedural contamination change signal and background. The complete method must be validated in the relevant matrix.