Not Detected, Below Limit, Zero: Three Different Claims
Share
Three phrases appear on certificates where a number could be, and they are not interchangeable. "Not detected" means the method saw nothing above its threshold. "Below the limit of quantitation" means the method saw something it cannot count. A zero means the method asserts an absence, which is the one claim an analytical method cannot actually make.
The difference between the three is the limit of quantitation, and it is the number most readers skip.
Two thresholds, two different verbs
The limit of detection answers a presence question: is there anything there at all? Around three times the noise, the answer becomes yes with reasonable confidence. The limit of quantitation answers a measurement question: can a number be attached? Around ten times the noise, with the method's accuracy and precision demonstrated at that level, the answer becomes yes.
Between the two sits a band where the substance is visible but not countable. A result reported in that band should say so, and the honest report marks it as an estimate or states it as below the quantitation limit rather than omitting it.
| Reported as | What it means | What it does not mean |
|---|---|---|
| Not detected | No signal above the detection threshold | The substance is absent |
| Below LOQ | Signal present, too small to quantify | The value is zero |
| Value with digits | Quantified within the method's demonstrated range | The value is exact to the digits shown |
| Zero | Not a method output; at best a policy statement | Anything measurable |
Why the LOQ has to sit below the limit
A specification limit that cannot be verified is a statement of intent. If the limit for a named impurity is 0.05 per cent and the method's quantitation limit is 0.08 per cent, then every batch will read either not detected or a number, and the region between 0.05 and 0.08 is unverifiable. A batch at 0.07 per cent passes for the wrong reason: the method cannot see it well enough to fail it.
The general expectation is that the quantitation limit should be no higher than half the specification limit, so that a result at the limit is a measurement rather than an extrapolation. That expectation is part of the ICH Q2 framework, and it is the first thing to check when a customer's limit is tighter than a supplier's method can support.
Where the mismatch exists, the conversation is not about the certificate. It is about the method, and the honest options are a more sensitive method or a documented justification for the limit as written.
How not detected gets misused
The phrase migrates. A method that detects nothing above its threshold becomes, in a summary table, "impurities: none". From there it is a short step to "free from related substances", which is a claim no method supports and no pharmacopoeia makes. The EDQM and USP monograph formats report impurities against named limits precisely because absence is not a testable claim.
The misreading has commercial consequences. A buyer comparing two certificates, one reporting 0.03 per cent and one reporting not detected, will rank the second as cleaner. If the first method's LOQ is 0.02 per cent and the second's is 0.10 per cent, the ranking is backwards: the first result demonstrates more control, because it measured further into the region where impurities live.
This is the same distinction that governs how impurity limits are read, and it is the reason a comparison between suppliers should start with the methods rather than the numbers.
What to establish about a method's LOQ
Three facts. The value itself, in percentage terms. How it was established, from signal to noise or from the calibration statistics. And whether it was verified at the level claimed, with spiked samples at the LOQ rather than extrapolated from a linearity study at higher concentrations.
The third is where shortcuts hide. An LOQ extrapolated from data collected at one per cent is a hypothesis. One verified by spiking at the claimed level, with recovery inside an agreed window, is a measurement. The validation summary states which one it was, and the difference matters the first time a result is disputed.
Reading the impurity line on a certificate
Four questions, in order. What is the reporting threshold of this method? Is the limit in the specification verifiable by it? Is the result a number, an estimate or an absence? And has anything changed in the method since the previous batch?
The fourth question is the quiet one. A method tightened to a lower LOQ mid-year makes the earlier "not detected" results incomparable with the later numeric ones, and the family of related substances is exactly where that comparison matters, because the individual impurities move in different directions.
Where this leaves a supply-chain partner
We state the reporting threshold wherever an impurity is reported as not detected, and we keep the LOQ below half the specification limit for every limit we agree to. Where a customer's limit is tighter than our method supports, we say so at quotation rather than at rejection.
The question worth asking of any certificate: what is the LOQ of this method, and where did the number come from? If the answer is a validation section, the certificate is complete. If the answer is "not detected", the certificate has told you less than it appears to.
How the two limits are actually established
Two routes exist, and they should agree within reason. The signal-to-noise route compares the analyte's response with the baseline fluctuation: around three to one for detection, around ten to one for quantitation. The statistical route uses the standard deviation of the response and the calibration slope, and it produces the same answers when the data behind it is honest.
When the two routes disagree substantially, the disagreement is diagnostic. Usually the noise was measured on a clean day and the slope on a used column, or the standard deviation came from one injection instead of several. Either way the method's low end is shakier than either number suggests, and the fix is a small experiment: replicate injections at a concentration near the claimed LOQ, with recovery calculated against the known spike.
That experiment is also what separates a verified limit from an extrapolated one, and the ICH Q2 framework is explicit that quantitation limits are demonstrated rather than derived. The compendial methods published by the EDQM and the USP carry their own reporting thresholds, which is why a monograph method's "not detected" is a defined statement while an in-house method's is only as good as its documentation.
The comparison trap between suppliers
Two certificates, same lot type, one reports an impurity at 0.03 per cent, the other reports not detected. The instinctive ranking favours the second. The ranking is only valid if the second method could have seen 0.03 per cent, which depends entirely on its LOQ.
The corrective question is therefore not about the result but about the method: what is the smallest value this method can report? Once that number is on the table, the comparison becomes meaningful, and the supplier whose method measures deeper has demonstrated more control even if their number is higher. A buyer who normalises for the LOQ before comparing is comparing methods rather than marketing.
The same logic runs inside a single product family, where the related substances each sit at different levels and the method's floor decides which of them are measurements and which are absences. A limit that only one impurity approaches is doing most of its work in the dark.
Writing the threshold into the specification
The cleanest solution is documentary: the specification or the method states the reporting threshold, and the certificate uses it consistently. "Below 0.05 per cent, not detected" is a complete statement. "ND" is a habit, and habits are what audits are for.
One line in the specification converts every future conversation from "what does not detected mean" to a number both parties already share. It costs one revision cycle, and it is the single highest-leverage edit available on most impurity specifications.
Reporting thresholds across a family
A product family creates a specific problem: the same specification template is applied to compounds whose impurities behave differently, and the reporting threshold that suits one may hide the most important impurity in another. For bile acids the relevant impurities are structurally close, so a method with a high threshold does not merely miss material, it misses the material most likely to be present.
The correction is to set the threshold from the impurity that matters rather than from the main peak's convenience. If a specification carries a 0.10 per cent limit for a named related substance, the method's quantitation limit needs to be at or below 0.05 per cent so that a result at the limit is measured rather than inferred.
The second correction is documentary: the threshold appears in the method, and the certificate's "not detected" refers to it. Two suppliers reporting "not detected" against different thresholds are not reporting the same thing, and a buyer comparing them without that context is comparing two absences.
What zero means when it appears
Some certificates print 0.00 where a method found nothing. It is a formatting convention and a misleading one, because it reads as a measurement while asserting an absence. The honest formats are the three described above: not detected, below the quantitation limit, or a number.
A certificate that prints 0.00 for every impurity and a comfortable assay is worth a second look, not because it is necessarily wrong, but because the format has removed the information the reader needs to judge what the method could see.
Frequently asked questions
What is the difference between detection and quantitation limits?
The detection limit is the level at which the presence of a substance can be inferred but not measured. The quantitation limit is the level at which a value can be reported with defined accuracy and precision. Between them, a substance can be seen but not counted.
How are the two limits established?
Either from the signal-to-noise ratio, conventionally three for detection and ten for quantitation, or from the standard deviation of response and the slope of the calibration. Both routes should give a similar answer, and a large disagreement between them is itself a finding.
Does not detected mean the impurity is absent?
No. It means no signal above the method's detection threshold was observed on that run. Absence cannot be demonstrated by an analytical method; it can only be bounded by the method's capability.
Why does the relationship between LOQ and the limit matter?
Because a specification limit that sits below the method's quantitation limit cannot be verified by that method. A limit of 0.05 per cent with an LOQ of 0.08 per cent produces results that are either not detected or quantified, with nothing measurable in between.
What does below the limit of quantitation mean on a report?
That a signal was seen but is too small to report as a number with confidence. It is a stronger statement than not detected and a weaker one than a value. Some reports give it as an estimate marked with a qualifier, which is the most honest option.
Should the LOQ be on the certificate?
Not necessarily on every certificate, but it should be stated in the method or the specification so that the reader knows what not detected means. A method whose LOQ is not stated has left the reader to assume it, and assumptions do not survive audits.