Respiratory Panel LoD: Which Strain and Matrix Does the Claim Actually Cover?

The short answer: an LoD claim has five fields, not one number

A colleague cites one limit of detection (LoD) for “influenza A” and asks two questions: does it cover the H3N2 strain circulating now, and does it still hold for a nasopharyngeal swab sent in saline instead of viral transport medium? The label-based answer is narrower than the question. An LoD is not a property of a virus name. It is a bounded statement about one result call, one specimen and diluent, one quantified isolate, one replicate rule and one study design, and it should only be repeated with all of those attached.

FDA’s 2009 respiratory viral panel special-controls guidance asks manufacturers to determine LoD for each specimen type and each analyte, using serial dilutions in pooled negative sample matrix, and to report the level detected about 95% of the time. Strain coverage is handled by a separate analytical-reactivity (inclusivity) study run at or near that LoD. Read any LoD claim as five linked fields:

1. Analyte or subtype. Is the number for an un-subtyped influenza A call or for a specific subtype result such as H1, H1-2009 or H3? Subtype calls can rely on different assays from the type-level call, so one row does not automatically speak for another.

2. Specimen type and diluent. Which specimen (for example, nasopharyngeal swab) and which liquid—transport medium, normal saline, or the pooled negative matrix used to make the dilutions?

3. Named isolate and units. Which strain was diluted, and was its concentration expressed in TCID50/mL, copies/mL or another unit? Units produced by different quantification methods are not interchangeable.

4. Replicate count and acceptance rule. How many replicates were tested at the claimed level, and what hit rate was required—for example, at least 19 of 20?

5. Evidence type. Was a strain wet-tested in the laboratory, tested only at a multiple of the LoD, or merely predicted to react by in-silico sequence analysis?

Applied to the colleague’s question: a single influenza A number supports the named isolate, matrix and replicate rule it came from. Whether a currently circulating H3N2 strain is covered depends on the reactivity table and any dated contemporary-strain results. Whether saline is covered depends on whether the label lists saline and reports supporting data. If either field is missing, narrow the sentence rather than stretch the number. The study protocols behind these figures are covered in our IVD analytical performance validation guide; this article deals only with reading the resulting claim.

What FDA actually asks a panel manufacturer to establish

Respiratory viral panel multiplex nucleic acid assays are class II devices under 21 CFR 866.3980. That regulation does not restate technical methods; it names three guidance documents as special controls: the 2009 respiratory viral panel guidance, the human metapneumovirus (hMPV) nucleic acid guidance for devices that detect hMPV, and the influenza A subtype guidance for devices that differentiate H1 and H3. Section 6.B.i of the 2009 guidance sets out the LoD recommendations. Read closely, it says:

• Definition. LoD is the lowest analyte concentration consistently detected—typically in at least 95% of samples tested under routine clinical laboratory conditions—in a defined type of sample.

• Scope. LoD should be determined for each specimen type and each analyte in the panel, applying the whole test process from sample preparation to amplicon detection.

• Material. Virus titers should be confirmed before use (for example, in TCID50 units), and serial dilutions prepared using appropriate pooled negative sample matrixes as diluents, with 3–5 replicates for each dilution.

• Reporting and confirmation. The LoD is reported as the level of virus giving a 95% detection rate, and it may be confirmed by testing at least 20 additional replicates at that concentration.

• Flexibility. Depending on the assay, a separate LoD determination across the whole concentration range may not be necessary for every specimen type, but LoD should be determined for at least the most common and most problematic ones.

Two of those points matter for reading a claim. First, the diluent is part of the evidence: an LoD made in pooled negative nasopharyngeal matrix is a statement about that matrix, not about a buffer-only dilution or a different specimen. Second, the flexibility clause means a missing per-specimen LoD number is not automatically a deficiency—but it also means the label may not give you a separate number for every specimen type it lists. Use our matrix-equivalence guide when the question is what evidence a new specimen type needs; here the question is only what the existing number already covers.

Extraction is a third hidden field. Section 6.A.iii asks manufacturers to evaluate performance using the entire pre-analytical process they recommend and, where several extraction methods are recommended, to demonstrate LoD and reproducibility for each method and each virus. A number obtained with one extraction method is therefore not automatically a number for another.

Strain diversity is a fourth, handled in Section 6.B.ii. The guidance recommends demonstrating detection of additional clinically relevant strains representing temporal and geographic diversity at viral levels at or near the LoD—for influenza, at least ten strains for each claimed subtype. That recommendation is about showing reactivity, not about assigning an LoD to each strain.

The 2009 guidance cites CLSI EP17-A (2004) for LoD principles. The current CLSI EP17-A2, Evaluation of Detection Capability for Clinical Laboratory Measurement Procedures (second edition, published 2012 and reaffirmed in 2017), covers limits of blank, detection and quantitation and the verification of manufacturers’ detection-capability claims, and CLSI states that FDA has recognized it. EP17 is a general detection-capability framework; it does not decide which respiratory strains or specimen types a panel must claim.

Why one LoD table and one strain table are not the same claim

A current manufacturer IFU shows how these pieces appear in labeling. The BIOFIRE Respiratory Panel 2.1 (RP2.1) De Novo instructions for use (document BFR0000-8579-03, September 2023) is used below as a worked reading only. It is one manufacturer’s document, not a benchmark or a statement about other panels, and its tables may have been revised since.

The IFU’s LoD section says LoD was estimated by testing dilutions of contrived samples containing known concentrations of organisms, then confirmed by testing 20 replicates on two instrument systems; LoD was confirmed when the organism was detected in at least 19 of 20 replicates (95%). The LoD table then lists one isolate per result row. For influenza A there is no single “influenza A” row. There are three subtype rows, each with its own isolate and concentration:

Two notes printed beneath that table matter as much as the numbers. The IFU says its copies/mL values come from qPCR quantification of isolate cultures that has not been compared to a reference material or other quantification methods. It also says TCID50/mL is an indirect, infectivity-based measure that varies with technique, and that it is not appropriate to judge relative sensitivity across cultures or molecular assays from TCID50 values. Even the manufacturer, in other words, warns against reading these rows as a league table.

The analytical-reactivity (inclusivity) section is a different study. It says each isolate was tested in triplicate at a concentration near LoD, that all isolates were detected within 10× LoD, and that in-silico predictions are noted for less common strains or where characterized isolates were not available. The influenza table records the multiple of LoD at which each isolate was detected: for example, A/Brisbane/10/2007 (H3N2) and A/Texas/50/2012 (H3N2) at 3×, and A/PR/8/1934 (H1N1) at 10×, with a footnote that at 3× that isolate was reported as influenza A with no subtype detected.

That is the core distinction. Detection of a strain at 3× or 10× LoD in triplicate shows that the assay reacted with that strain at that tested level. It does not assign the LoD-table concentration to that strain, and three replicates cannot demonstrate a 95% detection rate. For how inclusivity differs from cross-reactivity study design, see our comparison of respiratory panel inclusivity and cross-reactivity studies.

In-silico evidence is a third tier. The same IFU bases its SARS-CoV-2 reactivity assessment on sequence analysis of GISAID sequences available as of March 21, 2023, predicting that more than 99.99% of the sequences evaluated would be detected by one or both assays. That is useful monitoring of primer and probe matches, but it is a dated prediction rather than a wet-tested isolate result, and it should be labeled that way whenever it is quoted.

The strain question: subtype, isolate, season, and evidence type

Influenza A is the hardest case because the virus changes over time and panels can report it at more than one level. The RP2.1 IFU states that its influenza A subtyping assays target the hemagglutinin (H) gene only and do not detect or differentiate neuraminidase (N) subtypes. It also warns that its influenza A assays react variably with non-human and rarely encountered human influenza A viruses that are not H1, H1-2009 or H3, generally producing “Equivocal” or “no subtype detected” results. A strain can therefore be detected as influenza A while the subtype call—the part a laboratory may actually rely on—fails or differs.

For the colleague’s H3N2 question, check four things in order:

1. Which result row the number belongs to. In the RP2.1 example, the H3 LoD row uses A/Port Chalmers/1/73, a 1973 isolate.

2. How recent the reactivity isolates are. In the September 2023 IFU’s influenza table, the most recent seasonal human H3N2 isolate listed is A/Texas/50/2012, alongside a 2012 clinical specimen of the swine-variant H3N2v virus. Nothing in that table speaks directly to a strain that emerged later.

3. Whether dated contemporary-strain results exist. For devices classified under 21 CFR 866.4001 (below), the last three years of annual contemporary-influenza reactivity results must be made available as part of labeling. Look for that dated table, not only the static IFU.

4. The label’s own limitation. The RP2.1 IFU states that influenza A performance characteristics were established when A/H1-2009, A/H1 and A/H3 were the predominant influenza A viruses in circulation, and that performance may vary if other strains circulate or a novel influenza A virus emerges.

If none of those sources covers the strain now circulating, the accurate answer is “not shown by the label”—neither “covered” nor “not detected”.

What 21 CFR 866.4001 adds, and where it applies

A Federal Register final order published on August 21, 2025 codified 21 CFR 866.4001, the classification for a multiplex respiratory panel to detect and identify emerging respiratory pathogen(s) and common respiratory pathogens in human clinical specimens. The order is effective August 21, 2025, but it states that the classification has applied since November 24, 2017, following BioFire’s De Novo request for the FilmArray Respiratory Panel 2 plus. Its special controls relevant to LoD and strain coverage include:

• Analytical performance in labeling: analytical sensitivity (limit of detection), inclusivity, reproducibility, interference, cross-reactivity, instrument carryover/cross-contamination where applicable, and specimen stability.

• Influenza limitation statement: where applicable, a statement that influenza A performance was established when particular subtypes (for example, A/H3 and A/H1-2009) were predominant, and that performance may vary when other influenza A viruses are emerging.

• Annual contemporary-strain testing: annual analytical reactivity testing with contemporary influenza strains identified by FDA in consultation with CDC, sourced from CDC or an FDA-designated source, and tested under a standardized protocol FDA considers acceptable.

• Dated tabular results: by July 31 of each calendar year, the last three years of annual reactivity results must be included as part of labeling in tabular form—either in labeling that accompanies the device or through a prominent hyperlink to the manufacturer’s public web page.

• Post-market mismatch monitoring: a documented risk-mitigation strategy for identifying genetic mutations or novel strains, including regular review of published literature and annual in-silico analysis of target sequences to detect possible mismatches.

• Emergency influenza strains: after a qualifying emergency determination, testing within 30 days of FDA’s notice that characterized samples are available, and results added to labeling within 60 days.

Do not blend this regulation into every panel. The 866.3980 regulation names three guidance documents as its special controls; the annual-reactivity and July 31 clauses quoted above are written into 866.4001 for that device type. Before citing the annual table as a requirement for a specific product, confirm which classification regulation and product code that device is listed under, and treat any annual table as dated evidence for the strains it names.

The matrix question: specimen type, transport medium, saline, and extraction

A respiratory specimen claim has two parts: the collection site and the liquid the swab is placed in. The 2009 guidance lists respiratory tract specimens such as nasopharyngeal swab, nasal aspirate, nasal swab and nasal wash as appropriate types, and asks manufacturers to show that the device maintains acceptable performance under the transport and storage conditions recommended in labeling. Each of those conditions is a separate thing to check.

In the RP2.1 example, the intended use covers nasopharyngeal swabs only. The sample requirements specify an NPS placed in up to 3 mL of transport media or normal saline, note that phosphate-buffered saline (PBS) is not equivalent to normal saline, and give storage limits of up to 4 hours at room temperature and up to 3 days refrigerated, with frozen storage evaluated for up to 30 days. The LoD section adds one sentence about diluent: additional testing of contrived samples prepared using saline demonstrated equivalent detection near LoD relative to samples prepared in a viral transport medium.

That gives a precise answer to the colleague’s saline question for that product. Saline is a labeled transport option for NPS, and the IFU reports near-LoD equivalence with viral transport medium. What it does not report is a separate saline LoD table for every analyte, so the defensible sentence is “near-LoD detection in saline was reported as equivalent,” not “the saline LoD is X.” Because PBS is excluded by name, “saline” in a laboratory procedure must mean normal saline. For a different panel whose label lists only transport medium, the saline question is simply unanswered by the label: using saline would fall outside the labeled conditions, and the evidence needed belongs in a matrix-equivalence assessment, not in a reinterpretation of the existing LoD.

Specimen type works the same way. An LoD established in nasopharyngeal matrix says nothing direct about sputum, saliva or lower-respiratory specimens, whose composition differs. If the label does not list a specimen type, there is no LoD for it to quote. The regulatory routes for adding a claim are covered in our broader IVD regulatory guide.

Extraction completes the matrix field. The RP2.1 pouch is a closed system in which the instrument lyses the sample, extracts and purifies nucleic acids with magnetic beads, and then runs reverse transcription and nested multiplex PCR, so there is no user-selectable extraction method to bridge. For assays that recommend separate extraction kits, the 2009 guidance asks for LoD and reproducibility with each recommended method, and the reader should confirm that the quoted number came from the method actually used. False positives from instrument carryover are a separate question, covered in our guide to sample carryover versus amplification contamination.

A coverage ledger you can fill from any current label

Use the ledger below before repeating an LoD figure in a protocol, submission, specification sheet or comparison. It splits the five fields into seven rows so that diluent, date and extraction are not forgotten. The worked column is taken from the RP2.1 IFU cited above (September 2023 revision) and is illustrative only; it is not a ranking or a claim about other products.

When any row is blank, narrow the claim to what is filled in. If a discordant clinical result later raises a question about the comparator rather than the LoD, keep the two analyses apart; our article on preserving the original comparison in discordant IVD results covers that separate problem.

Five false shortcuts to avoid

1. Treating one isolate’s LoD as species-wide. The LoD table describes the isolate that was diluted. Other strains are covered, if at all, by the reactivity table at the multiple of LoD shown there. The RP2.1 example shows isolates within one subtype family detected at 1× or 3× and others only at 10×.

2. Treating in-silico prediction as wet testing. Sequence analysis is a dated prediction about primer and probe matches. Quote it with its database date and call it a prediction.

3. Substituting clinical or surveillance performance for analytical LoD. These answer different questions. A 2018 review by Diaz-Decaro and colleagues in Expert Review of Molecular Diagnostics argued that judging FDA-authorized multiplex respiratory assays for public health surveillance calls for per-target evaluation distinct from the diagnostic performance used for clinical claims, and noted that the major assays it reviewed were authorized for nasopharyngeal specimens, which limits interpretation of other sample types. Neither clinical agreement nor surveillance yield replaces the analytical LoD field, and the reverse is also true.

4. Assuming every listed specimen type has its own full LoD. The 2009 guidance allows LoD work to focus on at least the most common and most problematic specimen types. A listed specimen type without its own number is supported by the manufacturer’s overall evidence, not by a number you can quote; an unlisted specimen type is not covered at all.

5. Citing the 2009 guidance as proof of a current product’s status. The guidance sets out recommendations for a device type as of 2009. It does not show that a particular product is currently authorized, that its present label matches the guidance, or that the separate 866.4001 duties apply to it. Check the device’s own authorization and label revision.

When two targets are present in the same sample, competitive amplification is a separate evidence question; see our guide to testing multiplex respiratory panels for coinfection masking.

What to recheck before you repeat the number

An LoD citation goes stale when any of its fields changes. Before quoting one, record the following with an as-of date:

• Label revision. The document number and revision date of the IFU you read. Manufacturers revise LoD, reactivity and specimen sections.

• Analyte row. The exact result row—type-level call, subtype, or another target—that the number belongs to.

• Specimen and liquid. The labeled specimen types, transport liquids, any liquids excluded by name (such as PBS in the RP2.1 example) and storage limits.

• Isolate, units and replicate rule. The isolate name, the unit and how it was quantified, and the confirmation design (for example, at least 19 of 20 replicates).

• Reactivity evidence. The most recent isolates wet-tested for the strain family you care about, the multiple of LoD at which each was detected, and the date of any in-silico prediction.

• Annual influenza table. For devices under 21 CFR 866.4001, the current table of the last three years of contemporary-strain results, either in labeling or at the linked manufacturer web page.

• Swab and media interference. The collection materials evaluated in interference testing. The RP2.1 IFU, for example, lists rayon, nylon flocked, polyester and calcium alginate swabs and several transport media with no interference reported; do not assume another product tested the same materials.

• Regulatory references. Whether FDA has revised the respiratory panel guidance or special controls, and whether the CLSI EP17 edition you cite is still current and FDA-recognized.

Handled this way, an LoD stops being a headline number and becomes what it actually is: a dated, bounded claim that can be repeated accurately or narrowed honestly.