Reagent water: pure against what?

Federal methods define reagent water by the measurement

Method 601 in Appendix A to 40 CFR part 136 puts it in one line: “Reagent water is defined as a water in which an interferent is not observed at the MDL of the parameters of interest.” Method 625.1 in the same appendix words it slightly differently—“water in which the analytes of interest and interfering compounds are not detected at the MDLs of the analytes of interest”—and means the same thing. Purity is not an absolute in either. It is a property held relative to a named list of analytes at a named detection limit.

Method 601 goes on to offer three ways to produce that water: a carbon filter bed, a water purification system, or boiling for 15 minutes followed by an hour of inert-gas sparging at 90 °C. Any of them qualifies, because the definition is a performance test rather than a production route. That is the useful inversion for a purchasing conversation. Write down what must not be present and at what level, and the acceptable ways of getting there fall out of it; start from a type name and you have skipped the requirement.

An instrument answers only the question it asks

FDA’s inspection guide for high purity water systems states it flatly: “Typically, conductivity meters are used on water systems to monitor chemical quality and have no meaning regarding microbiological quality.” Elsewhere it adds that “Because of the similar ionic quality of distilled and deionized water, conductivity meters cannot be used to monitor microbiological quality.” Resistivity and conductivity are the same measurement reciprocated, and neither one sees an organism or an uncharged organic molecule.

The same guide records that “different products require different quality waters,” contrasting parenterals, which require water with no endotoxins, against topical and oral products, which it says have no endotoxin requirement. Three properties, three instruments, three sampling regimes—and a specification that names only one of them has left the others undefined rather than satisfied. Note the document’s own limits: it was issued in July 1993, it says on its face that it “does not bind FDA,” and it cites USP XXII, a revision superseded many times over. Its reasoning is durable; its numbers are not a current requirement and must not be quoted as one.

A grab sample is not the system

The same guide explains why a single water result is weak evidence. Organisms attach to pipe and tank walls as biofilm, “which continuously slough off organisms,” so “contamination is not uniformly distributed in a system and the sample may not be representative of the type and level of contamination.” It offers the scale of the effect directly: “A count of 10 CFU/mL in one sample and 100 or even 1000 CFU/mL in a subsequent sample would not be unrealistic.”

It also draws a distinction worth importing into any water specification: “None of the limits for water are pass/fail limits. All limits are action limits,” with an expectation that an exceedance triggers an investigation, a correction and an assessment of impact. A limit with no named owner for that investigation is not yet a control. Decide before the first result who reviews an exceedance and what happens to material already made.

• Which property would actually ruin this step—ionic, organic, microbiological, or particulate?
• At what point in the system is each property measured, and how often?
• Is each limit a specification or an action limit, and who investigates an exceedance?
• Where water is delivered rather than generated on site, which properties are known at the point of use rather than at the point of manufacture?

What this guide will not reproduce

The reagent-water type framework most laboratories quote—Types I through IV—is defined by ASTM D1193, and clinical laboratories work to a separate document published by CLSI. Both are copyrighted standards sold by their publishers. Neither was readable while this guide was written, so neither is cited here and none of their numeric limits are reproduced. Take the limits from the edition of the standard your own specification invokes and that your team actually holds.

That is not a gap to paper over with a plausible-looking table. A type designation copied from memory, or from a supplier’s marketing page, is exactly the sort of number that survives into a specification and is never checked again. Name the standard, name the edition, and record the properties you require with their units and methods beside it. Where a value is taken from a document nobody on the project has opened, write that down too.

Requirement → evidence → decision boundary

An editorial checklist for your review—not a table of product specifications.

Revision record

1.0 · September 12, 2026 — Source-based editorial edition. No original testing or product qualification is claimed.

For laboratory, pharmaceutical and research buyers specifying reagents, solvents and water. This collection does not qualify a material for a method, establish compliance with cGMP or with any pharmacopeial monograph, approve a supplier, or make a release decision for delivered material.