Specification Limits: Who Sets Them and What They Mean

A specification limit is the boundary a test result is judged against before the result is known. It is written down in advance as an acceptance criterion, such as "not less than" a stated purity or "not more than" a stated impurity level. A result without its specification is a number with no yardstick. A result with a specification can be read as inside, outside or close to the line.

Most readers look at the measured value and move on. The interesting part is the column beside it, or the absence of one. A certificate that reports a figure and a verdict, but never states the criterion, asks the reader to trust a judgement they cannot check.

What a specification actually is

ICH Q6A, the guideline on specifications for new drug substances and products, defines a specification as a list of tests, references to analytical procedures, and appropriate acceptance criteria. That is three components, not one. A limit with no named method is incomplete. A method with no limit is only a measurement.

Three terms are routinely confused, so it helps to separate them.

• Specification: the pre-set criterion. "Not less than" a stated value, or "not more than" a stated value, or a range.
• Result: what the method measured in the sample submitted, with its units and, ideally, its uncertainty.
• Verdict: the comparison of the two, written as "conforms", "complies" or "pass".

The verdict is the least informative of the three. It is a conclusion, and a conclusion is only as good as the premises printed above it.

Who sets specifications

There are three common sources, and they carry different weight.

Compendial specifications. Pharmacopoeias such as the USP, the Ph. Eur. and the BP publish monographs for substances they cover. A monograph states the tests, the procedures and the limits, and it is public. Anyone can check a result against it. Not every peptide has a monograph, and most research peptides do not.

In-house specifications. A manufacturer proposes limits for its own material, supported by development data, batch history and validated methods. ICH Q6A and, for biotechnological and biological products, ICH Q6B set out the principles for doing this. The limits are justified, but they belong to that manufacturer and that process.

Agreed or reporting criteria. Where no monograph exists and no manufacturer file is involved, the criterion may be one the submitter and laboratory agree on, or a laboratory's own stated threshold. This is legitimate. It simply has to be stated, because nothing external defines it.

The source matters because it tells the reader who is accountable for the limit. A compendial limit has been published for scrutiny. An in-house limit rests on the manufacturer's justification. An unstated limit rests on nothing the reader can see.

How limits are set in the first place

A sensible limit is not chosen by picking an attractive number. Under the ICH framework it is set from several kinds of evidence together.

• Process capability. What the manufacturing process can consistently deliver across batches.
• Method capability. What the analytical procedure can reliably distinguish. ICH Q2 describes the validation characteristics involved, including accuracy, precision, specificity, detection limit, quantitation limit and range.
• Stability data. How the material changes over time under defined storage conditions, in line with the principles of ICH Q1A.
• Relevant requirements. Any clinical, safety or regulatory consideration that bears on the attribute.

A limit tighter than the method can resolve is not a real limit. If a method cannot tell two levels apart, a specification between them cannot be enforced by that method. This is why specification and method validation are developed together and not separately.

Release limits and shelf-life limits

ICH Q6A recognises that a material can have different acceptance criteria at release and at the end of its shelf life. Some attributes change with time, and the shelf-life criterion has to allow for that change. The release criterion is often tighter so that material still meets the shelf-life criterion after storage.

For a reader, the consequence is simple. A result on a certificate is a snapshot taken at one point. It describes the sample as tested on the day it was tested. It does not say how the material will behave in a month or a year, and it does not say how the sample was stored before it arrived. A certificate that states a single criterion without saying whether it is a release or shelf-life criterion has left a question open.

Limits, method capability and uncertainty

Every measurement carries uncertainty. A reported purity of a given percentage is an estimate with a spread around it, produced by a method with a defined precision. None of this makes the measurement a bad one. It means the number is a range in practice, even when it is printed as a single value.

Two method properties are worth knowing.

• Limit of detection (LOD): the lowest amount the method can detect, though not necessarily quantify.
• Limit of quantitation (LOQ): the lowest amount that can be measured with acceptable precision and accuracy.

A statement such as "not detected" or "below the limit of quantitation" is only meaningful alongside the actual LOQ. "Not detected" by a method with a coarse LOQ is a weaker statement than the same words from a sensitive method. The words are identical. The information is not.

A pass near the line versus a pass well inside it

Two results can both be marked "conforms" and still tell different stories. Suppose a criterion of not less than a stated purity. One sample sits comfortably above it. Another sits just above it, within the range of the method's own variability.

The verdict on the certificate is identical. The confidence is not. A result close to the limit could fall on the other side if the measurement were repeated. A result well inside the limit would not. This is not a flaw in the process of comparison. It is a property of measurement, and it is why the number, the criterion and the method precision should be read together and not reduced to a single word.

The same logic applies in reverse. A result slightly outside a limit is not automatically a different material from one slightly inside. It is a result that did not meet the criterion, and the criterion exists precisely so that the decision does not depend on interpretation after the fact.

Impurity thresholds as an example

Impurity control shows how layered specifications can be. ICH Q3A, for impurities in new drug substances, describes three separate thresholds. Each answers a different question.

• Reporting threshold: the level above which an impurity must be reported.
• Identification threshold: the level above which its structure should be established.
• Qualification threshold: the level above which its biological safety needs to be justified.

For a maximum daily dose of up to 2 g per day, Q3A lists a reporting threshold of 0.05 percent, and identification and qualification thresholds of 0.10 percent and 0.15 percent respectively (or a stated daily intake, whichever is lower). ICH Q3B applies similar concepts to drug products. These figures belong to those guidelines and to the material they cover. ICH Q3A in fact places peptides outside its own scope, so the thresholds are shown here only to illustrate how layered an impurity specification can be. They are not a universal standard for every peptide, and applying them to a research material without a defined context would be a mistake.

The point is structural. A single "total impurities" figure hides the fact that a result can be reported, identified or qualified. An impurity can be present and quantified without being identified, and identified without being assessed for safety. See peptide impurities for how these show up in practice.

How to read a pass

Before accepting a verdict, work through a short set of questions.

• Is the acceptance criterion printed beside the result, with its units?
• Is the source of the criterion stated: a monograph, an in-house specification or an agreed limit?
• Is the method named, and does it fit the attribute being measured?
• If a result is "not detected" or "below LOQ", is the actual limit given?
• Does the certificate identify the sample, the batch and the date, so that the result is tied to a specific material?

A "conforms" that survives all five is informative. One that fails the first is a label. The wider reading of the verdict is covered in what conforms means on a COA.

One boundary applies to every case. A certificate describes the sample submitted and nothing beyond it. A pass against a specification means that sample met that criterion by that method on that day. It does not extend to other vials, other batches, or questions the specification never asked. Purity is composition. It is not safety.

Novagen is a third-party analytical laboratory and reports what was measured in the sample submitted. Certificates are listed in a public register at novagenlaboratories.com, where a reader can confirm that a document matches a record. Current quoted turnaround is 21 days from receipt of samples, and questions can be sent to admin@novagenlaboratories.com.