Aquatic Animal PCR: Sensitivity, Specificity and Surveillance

A positive PCR result is not automatically an outbreak count

A report says that three shrimp samples were positive by PCR. What happened on the farm? The result may signal a pathogen that requires prompt attention, but the sentence alone does not tell us the number of infected animals, the prevalence in the pond, whether every animal is clinically ill, or whether the detected target is viable. Each inference needs a defined test, sampling design, result interpretation and relevant animal-health context. Translating laboratory language responsibly means preserving what the test actually measured while avoiding the temptation to turn an uncertain biological signal into a complete disease narrative.

This guide explains polymerase chain reaction (PCR), real-time PCR or qPCR, diagnostic sensitivity, diagnostic specificity, positive and negative predictive values, cycle thresholds, sampling units, prevalence assumptions and pooling. Its purpose is educational: to help students, translators, aquaculture technicians and managers read laboratory and surveillance reports accurately. It does not diagnose an individual farm or replace laboratory, veterinary or regulatory advice. Numerical exercises use clearly hypothetical tests, populations and performance values. For real disease management, follow current disease-specific WOAH guidance and qualified aquatic animal-health professionals.

Three questions organise every case. First, what biological material did the assay target, and how reliably could it detect that target under the stated sample conditions? Second, how were animals or water samples selected from the population, and what can that selection say about the larger population? Third, what operational action is justified by the quality and scope of the evidence? A fluent translation that gets any of these wrong may cause unnecessary alarm or dangerous reassurance. The right words need an explicit denominator and the humility to state what is unknown.

1. Translate PCR as detection of a specified target, not a verdict on every animal

PCR amplifies a particular genetic sequence so that a laboratory can detect a target that may have been difficult to measure directly. Depending on the organism and assay design, laboratories may test DNA directly or convert RNA to complementary DNA before amplification, as in reverse-transcription PCR. A positive analytic signal means the assay detected its target according to its decision rule and controls. It does not by itself establish infectiousness, clinical signs, mortality cause or the geographical extent of infection. These questions depend on organism biology, sample type, validation data, confirmatory testing and the case definition used.

An aquatic animal-health report may discuss pathogens that affect finfish, crustaceans or molluscs, but a protocol validated for one host, tissue and clinical context cannot simply be transferred to another. The diagnostic meaning of a positive tissue sample differs from that of a water eDNA sample. In a tissue sample, nucleic acid is associated with the selected animal and tissue; environmental water can contain material from multiple origins and transport histories. Translation should identify sample matrix, host species, life stage, tissue, collection date and test method where those details influence interpretation.

The wording “infected” also has a technical and epidemiological meaning that can differ from “PCR-positive.” A disease-specific case definition may incorporate clinical signs, histopathology, confirmatory tests, agent isolation or other evidence. A sample can be PCR-positive without the owner observing clinical disease, especially in surveillance of apparently healthy animals. Conversely, clinically affected animals can test negative because of sampling timing, specimen quality, assay mismatch or limited target concentration. A report should not erase those distinctions in order to sound decisive. Precision is particularly important when a declaration triggers animal movement controls or trade consequences.

2. Diagnostic sensitivity and analytical sensitivity are not synonyms

Diagnostic sensitivity is the proportion of truly affected or infected units, according to the chosen reference definition, that test positive. If 95 out of 100 truly infected animals are positive, diagnostic sensitivity in that example is 95%. Analytical sensitivity instead concerns detection of the target at low concentrations under specified laboratory conditions, often expressed as a limit of detection or probability of detection at a given target amount. A test can detect tiny amounts of purified nucleic acid in the laboratory yet miss infected animals if the selected tissue contains little target or specimen handling is poor. Neither figure should be quoted without its denominator and validation context.

Diagnostic specificity is the proportion of truly unaffected or uninfected units that test negative. If 990 out of 1000 truly uninfected units test negative, specificity is 99%. That is not the same as saying that 99% of positive results are true. The latter involves positive predictive value, which depends on prevalence. The distinction becomes critical when a test is applied in a low-prevalence surveillance setting. Translating specificity as “probability that a positive is correct” changes a conditional probability and can produce a very misleading management interpretation.

A second naming issue is analytical specificity, which concerns how selectively the assay responds to its intended target rather than to cross-reacting sequences or contaminants. It can contribute to diagnostic specificity, but one number does not automatically equal the other. Diagnostic performance also depends on which units and disease definitions were used for validation. A test validated on severely affected animals may behave differently in asymptomatic carriers with lower target levels. A laboratory report should clarify which performance estimate applies to the current use and whether a local validation or published disease-specific source supports it.

3. The four cells that make a test result understandable

Imagine a two-by-two table comparing true infection status against a binary test result. True positives are infected units testing positive; false negatives are infected units testing negative. True negatives are uninfected units testing negative; false positives are uninfected units testing positive. Sensitivity equals true positives divided by all infected units. Specificity equals true negatives divided by all uninfected units. Positive predictive value equals true positives divided by all positive tests. Negative predictive value equals true negatives divided by all negative tests. Each fraction points in a different direction across the table.

This table protects readers from denominator drift. “Five false positives” is a count, not a rate until the number of truly uninfected units is known. “Two percent of positive tests were false” needs a total number of positives. “The test is 98% accurate” is generally too vague for policy unless accuracy and population composition are defined. A translated report should name sensitivity and specificity with the relevant conditional group instead of turning both into a generic “reliability” score. Where an uncertain reference standard prevents true status from being perfectly known, the validation limitations must also be reported.

Suppose a hypothetical test is evaluated against an appropriate reference standard in 100 infected units and 200 uninfected units. It identifies 92 infected units as positive and 190 uninfected units as negative. Sensitivity = 92/100 = 92%. Specificity = 190/200 = 95%. It makes eight false-negative and ten false-positive decisions in this validation sample. Those results do not imply every future farm will produce the same proportion of false positives among its positive results. Predictive values change with the infection prevalence among the animals being tested and with actual test performance under those conditions.

4. Why prevalence changes the meaning of a positive test

Consider 10,000 animals sampled hypothetically from a population with true infection prevalence of 1%. That means 100 infected and 9900 uninfected in the teaching model. A test with 95% diagnostic sensitivity detects 95 of the infected animals and misses five. With 99% specificity, it correctly calls 9801 of the uninfected negative and falsely calls 99 positive. There are 194 positive results in total, of which 95 are true positives. Positive predictive value is 95/194 ≈ 49.0%. Excellent-looking specificity does not make the proportion of true positives among positives close to 99% in this low-prevalence example.

Now keep the same test but suppose infection prevalence is 20% in another hypothetical population of 10,000. There are 2000 infected and 8000 uninfected. The test produces 1900 true positives and 100 false negatives; it produces 7920 true negatives and 80 false positives. Positive predictive value becomes 1900/1980 ≈ 96.0%. The assay’s stated sensitivity and specificity have not changed. The different prior prevalence radically changes the meaning of a positive result. Translations that report only “test specificity 99%” without the surveillance context cannot answer the operator’s actual question.

Prevalence must also be interpreted carefully. A farmer who selectively tests moribund animals is not drawing the same sample as a surveillance team randomly testing apparently healthy juveniles. The prevalence in the tested subgroup is not automatically the farm-wide prevalence. A set of convenience samples collected after a mortality event can identify a serious hazard but cannot by itself estimate unbiased population prevalence. A translation that changes “10% of sampled sick animals” into “10% of all stock” has introduced a new inference without evidence. State the sampling frame, not just the test total.

5. Positive and negative predictive values

Positive predictive value answers: among units whose test is positive, what fraction truly meet the defined infection status? Negative predictive value answers the corresponding question among negative tests. In the 1% prevalence example with 95% sensitivity and 99% specificity, NPV = 9801/(9801 + 5) ≈ 99.95%. That large figure can feel reassuring, but it applies only under the hypothetical prevalence, test performance and sampling definitions. It is not a promise of zero infected animals among negative tests. At higher infection prevalence, the negative predictive value declines even when the test itself behaves the same.

Bayes’ theorem provides a compact way to translate predictive values. PPV = Se × P / [Se × P + (1 − Sp) × (1 − P)], where P is the prevalence or prior probability for the relevant unit. The denominator combines the expected positive contributions from infected and uninfected units. If the report does not provide a credible P, the calculation cannot yield a unique PPV. It can instead show a range across plausible prevalences. This is a more honest interpretation than choosing a convenient prior without saying so, especially when the biological and economic consequences of a positive declaration are high.

The practical consequence is not to ignore positive tests. It is to integrate them with clinical context and an appropriate confirmatory strategy. In low-prevalence surveillance, even a small false-positive probability matters because uninfected animals vastly outnumber infected animals. When consequences are serious, a laboratory may recommend another assay, another sample or additional evidence. A positive screening result can therefore justify timely precaution and investigation while remaining short of a final disease classification. Translation should preserve the procedural status: screening signal, presumptive finding, confirmed detection and outbreak declaration are not interchangeable.

6. What a negative result does and does not establish

A negative result may mean that the tested specimen contained no detectable target under the assay and specimen conditions. It may also occur when infection was present but the chosen tissue, time point, sample amount or extraction did not yield sufficient target. Some assays are sensitive during particular stages or tissues and less informative at others. “Not detected” is often safer than “pathogen absent” when reporting a laboratory negative. The distinction becomes stronger when the inference extends from a few animals to an entire pond or production region.

Suppose 20 independent animals are randomly selected from a population with 5% infection prevalence, and a test has 90% sensitivity, with no false positives for this simplified detection calculation. The probability that any selected animal tests positive is 0.05 × 0.90 = 0.045. The probability of no positive detections among all 20 is (1 − 0.045)^20 ≈ 0.398. Thus a run of 20 negatives remains quite plausible despite the assumed prevalence. This calculation is a teaching approximation; clustered infection and imperfect sampling can make independence unrealistic and alter the resulting probability.

With 60 independent specimens under the same assumptions, the probability of zero detections is approximately 0.955^60 ≈ 0.063. Increasing sampling reduces the chance of missing infection at the assumed prevalence, but it does not deliver absolute freedom from disease. If animals share water or cages and infection is clustered, the effective information in 60 samples may be less than a simple independent model suggests. A surveillance report should make its confidence objective, design prevalence, population structure and test sensitivity explicit before claiming what negative results establish.

7. The unit tested may not be the unit of management

One PCR reaction may examine an aliquot from one tissue extract. That extract may come from one animal, several tissues combined, a composite from multiple animals or an environmental sample. A surveillance decision may concern a pond, cage, hatchery, shipment or region. Moving from a reaction result to a pond status requires a defined inference strategy. “Ten pools tested negative” is not synonymous with “ten animals tested negative.” If each pool contains five animals, the sampling design covers 50 animals, but the assay’s pooled detection performance may differ from testing all 50 individually.

Pooling can reduce analytical cost but may dilute target material, complicate contamination control and change diagnostic sensitivity. A pooled assay should be validated for the relevant pool size and sampling procedure. It is not enough to use individual-sample sensitivity and pretend the pool performs identically. The result “one positive pool of five animals” cannot identify which member is positive unless separate testing resolves that question. Translation should retain pool size, number of pools, number of sampled animals, protocol and whether pools were formed before or after extraction.

Environmental sampling adds another interpretation boundary. A positive water sample may be useful evidence of target nucleic acid in the sampled water, yet its origin can involve transport, shedding, dead tissue or environmental persistence. Negative water samples can occur even when affected animals exist elsewhere in the system. The sampling geometry, water flow, volume filtered, preservation and assay controls all matter. It is a scientific mistake to turn a water sample into a one-to-one census of infected animals. The same logic applies to environmental DNA methods used for ecological occupancy: detection and abundance are different quantities.

8. Reading qPCR cycle thresholds

Real-time PCR often reports a threshold cycle or quantification cycle, commonly Ct or Cq. Under a suitable assay, a lower Ct usually corresponds to more target nucleic acid at the start of amplification than a higher Ct, when the samples are otherwise comparable. However, Ct is not a standard unit of pathogen particles per gram of tissue. Extraction yields, amplification efficiency, inhibition, instrument settings, threshold algorithms and different assay targets can alter it. A value of 25 on one protocol should not be compared uncritically with 25 on another and translated into the same biological concentration.

An assay may have a validated detection cutoff, indeterminate region or confirmatory rule. A late signal near the detection threshold may carry greater stochastic uncertainty and require appropriate interpretation, not automatic dismissal. A result can be positive for target nucleic acid while indicating a lower detected load than another specimen, but viability and infection status are separate questions. “High viral load” requires a validated quantitative interpretation, usually a calibration and an appropriate specimen basis, rather than just the visual impression of a lower Ct. In translation, preserve whether the laboratory reported qualitative detection, semi-quantitative comparison or calibrated concentration.

Suppose two samples run in the same well-validated assay yield Ct 22 and Ct 28. Under an idealised amplification efficiency of 100%, a six-cycle difference suggests roughly 2^6 = 64-fold difference in initial amplifiable target, provided sample preparation, thresholding and inhibition are comparable. That is an analytical approximation, not a direct claim of 64-fold difference in viable pathogens, infectivity or clinical severity. When efficiency differs from the ideal and matrix effects are present, a simple power-of-two interpretation can mislead. Any translation should retain these conditions instead of converting Ct differences directly into disease risk ratios.

9. Why controls and sample quality belong in the interpretation

A laboratory workflow can include positive controls, negative extraction controls, no-template controls and internal amplification or extraction controls. Their purposes differ: some demonstrate that a reaction can detect the intended target, others help reveal contamination or inhibition, and others assess extraction or amplification performance. A valid negative animal result depends on the corresponding quality controls passing the laboratory’s criteria. If a control fails, the proper status may be invalid, inconclusive or requiring repeat analysis rather than negative. Translating “control failure” as “no pathogen detected” can hide the fact that the run could not support that conclusion.

Transport and storage also matter. Tissue identity, preservative, temperature, delay, freeze-thaw cycles and container labelling can influence what remains measurable. A sample compromised before testing may produce ambiguous evidence even with an excellent assay. The collection chain is not secondary paperwork; it defines whether the material analysed belongs to the animal, site and time claimed. A test result from a mislabelled tube can be analytically precise but epidemiologically wrong. Technical reports need an explicit distinction between analytic performance, specimen integrity and correct assignment of the result to a management unit.

Reference-laboratory validation of aquatic animal PCR assays illustrates why fit-for-purpose evidence matters. CSIRO has described validation work on PCR tests for white spot syndrome virus, including diagnostic performance in apparently healthy prawns. The general lesson is not that any named PCR test is perfect; it is that surveillance of apparently healthy animals requires evidence appropriate to the use case. A study conducted on heavily affected specimens does not automatically supply the same diagnostic sensitivity for low-load carriers. Use current disease-specific protocols and local competent authority guidance when interpreting consequential results.

10. Case A: a surprising positive result in a low-risk hatchery

A hatchery with no obvious clinical signs receives one positive screening result from a routine batch. Staff read “specificity 99%” and conclude that the result is 99% certain to indicate infection. That conclusion misreads specificity. The correct next steps are procedural rather than purely numerical: verify the specimen, assess test and sampling validity, review the relevant prior risk and coordinate disease-specific confirmation with qualified personnel. The result must not be ignored simply because prevalence may be low; nor should the report overstate it. A precise sentence can say “screening assay detected the target in one specimen; infection status is being evaluated under the applicable case definition.”

If the same hatchery subsequently receives a confirmed result using a validated independent procedure, the evidence status changes. The report should preserve the chronology: original screen, confirmatory sampling, laboratory methods, results and final interpretation. Combining all steps into the single phrase “PCR discovered an outbreak” erases important details and may wrongly imply that the first result independently established the outbreak. When dates are essential to trace possible animal movements, reporting them correctly is part of the biology as well as the language. A timeline can prevent unsupported assumptions about when infection began.

11. Case B: clustered infection inside cages

A facility samples 30 animals, all from a single cage because that cage is easiest to access. All test negative. A report says “the farm is free from infection.” The inference is not justified if the other cages were not sampled and disease risk differs among them. The sampling frame contains the selected cage, not the full farm. Even a perfectly performed PCR cannot compensate for systematically missing the target population. A credible surveillance plan may use stratification, risk-based sampling or other approved methods appropriate to the objective; a translation should state the selected population rather than silently expand the claim.

If a second survey samples animals from six cages using a documented random scheme, its representativeness may improve, but the number of samples, clustering, animal movement and test performance still matter. Ten animals per cage are not necessarily equivalent to sixty independent draws from a completely mixed population. Transmission within cages can make results correlated. A report should preserve the number of epidemiological units sampled as well as the number of individual animals. The unit of inference is a scientific design choice, not something a translator can infer from a final total.

12. Case C: high Ct interpreted as a small outbreak

A test report gives a late Ct near the assay’s validated cutoff, and an operator concludes that the farm has a small outbreak. Ct primarily describes an amplification signal in the tested material under the assay conditions, not outbreak size. A single animal or water sample cannot generally establish the number of affected cages or the timing of spread. The repair is to state the result class, assay limitations and needed follow-up investigation. If the result falls in an indeterminate zone defined by the laboratory, retain that classification; do not turn a technical uncertainty into a positive or negative simply to make the sentence sound clear.

In another case, staff compare Ct values from two laboratories using different targets and sample-processing methods. Even if both assays detect the same pathogen, numerical comparison without calibration may be inappropriate. One assay’s lower Ct is not necessarily proof that its farm had more infectious animals. A technical translation should preserve each lab’s method, calibration and reported interpretation, and avoid drawing a cross-laboratory quantitative trend unless validation supports it. The word “higher” must refer to a defined biological variable, not just a lower cycle number.

13. Case D: pooled samples and hidden dilution

A survey combines tissue from ten animals per pool and tests six pools. All are negative. If the pooling method has not been validated at ten specimens, the survey’s detection sensitivity may be uncertain even if the underlying PCR performs well on individual samples. Some low-load positive specimens may fall below the pooled detection limit. A responsible report says 60 animals contributed to six negative pools and gives the pooling validation or uncertainty. It does not describe the outcome as sixty independently negative PCR tests. That phrase changes both the number of laboratory observations and the sensitivity assumption.

Suppose a pool is positive. The pool result identifies a pooled target signal, not which of ten contributors supplied it. The next logical investigative stage may involve testing individual retained material or collecting additional specimens under a veterinary-approved plan. A translation must not say “ten infected animals” or “one infected animal” solely from one positive pool. The range of plausible contributors depends on pooling and assay characteristics. Explicitly naming the pool as the unit tested prevents an inference that the experiment did not resolve.

14. Case E: percentage positive is not prevalence

A diagnostic laboratory receives 100 samples from animals selected because they show disease signs. Twenty test positive. The sample positivity rate is 20% among that selected set. It does not establish that 20% of all farm animals are infected. The sample is enriched for a clinical reason, and test performance may be affected by disease stage. The correct statement identifies sample inclusion criteria, numerator and denominator. If a translation needs to say “prevalence,” it should require a defensible population sampling framework or present the estimate with explicit design and uncertainty limitations.

Even in a random survey, the observed positive proportion is not automatically true prevalence when sensitivity and specificity are imperfect. One simple adjustment, under strong assumptions, is apparent prevalence = Se × true prevalence + (1 − Sp) × (1 − true prevalence). Solving for true prevalence gives (apparent prevalence + Sp − 1)/(Se + Sp − 1), provided the denominator is positive and the assumptions are applicable. A corrected estimate can be unstable or outside feasible bounds with small samples or uncertain parameters. It is better to preserve these uncertainties than to present a precise adjusted percentage without its model.

15. Ten checked calculation exercises

Exercise 1. A validation table contains 180 truly infected units, of which 171 test positive; 820 truly uninfected units, of which 812 test negative. Sensitivity = 171/180 = 95%; specificity = 812/820 ≈ 99.02%. There are nine false negatives and eight false positives. The arithmetic is only one part of validation: the true-status reference, sample selection and intended population still govern transferability. Do not translate 99.02% specificity into “99.02% of positive results are correct.” That is a different fraction.

Exercise 2. A hypothetical population has 5% prevalence, sensitivity 90% and specificity 98%. Out of 10,000 units, 500 are infected and 9500 uninfected. True positives = 450; false negatives = 50; false positives = 190; true negatives = 9310. PPV = 450/640 = 70.31%, while NPV = 9310/9360 ≈ 99.47%. A single adjective such as “accurate” cannot preserve these two different decision probabilities. Always name the relevant result group in the denominator.

Exercise 3. In a sample of 200 animals, 14 test positive. Observed sample positivity is 14/200 = 7%. With no information on selection method, sensitivity, specificity or study population, this does not give a reliable farm-wide true prevalence. The correct translation says “7% of the 200 tested animals were positive under the assay’s decision rule.” It should not say “7% of all animals were infected.” This exercise tests inference, not arithmetic sophistication.

Exercise 4. A qPCR result shows Ct 24 versus Ct 29 in the same assay under ideal doubling per cycle. The five-cycle difference corresponds to 2^5 = 32-fold difference in initial amplifiable target under comparable processing conditions. That is not a 32-fold change in mortality risk or the number of infected animals. Sample matrix, extraction and inhibition require consideration. Keep “idealised amplification comparison” in the result and do not claim absolute pathogen concentration without appropriate calibration.

Exercise 5. With independent random selection, prevalence 2% and per-infected-unit diagnostic sensitivity 95%, approximate detection probability per selected unit is 0.02 × 0.95 = 0.019. The chance of missing all detections in 50 tests, assuming no false positives for this calculation, is (0.981)^50 ≈ 0.383. That is far from zero. The simplification excludes population clustering and sampling without replacement; a real freedom-from-disease design needs appropriate statistical methods and competent oversight.

Exercise 6. An assay has sensitivity 85% for low-load carriers but 98% for clinically affected animals in separate validated contexts. A surveillance programme using apparently healthy animals should not borrow 98% simply because it is the larger number. If a calculation uses 98% instead of 85%, it predicts a smaller probability of missed infection than may be warranted. The answer is to match test performance to the actual sampling population and specimen procedure. Changing percentages across contexts without explanation is an error in scientific translation.

Exercise 7. A farm reports 12 positive pools, each containing five animals. Without deconvolution, the number of positive pools is 12, the number of contributing animal specimens is 60 and the number of positive individuals is not identified. At least one member per positive pool may have contributed target under the intended interpretation, but contamination or pooled artefacts must be addressed by validation. The phrase “60 animals PCR positive” is unsupported; so is “12 animals infected.” State the pool-level result as measured.

Exercise 8. A laboratory tests a positive control, a negative extraction control and a specimen. The specimen and the negative extraction control both yield target signals. That may indicate contamination or another control failure depending on the validation rules. It is inappropriate to call the specimen unequivocally positive without resolving the failed control. The right translated status is the laboratory’s designated invalid, inconclusive or repeat-required class, with the problem described. An analytically compromised run is not rescued by writing a confident sentence.

Exercise 9. True negatives number 980 and false positives number 20 in a validation set. Specificity = 980/(980 + 20) = 98%. If the same test is used in a population with very few infected animals, a high fraction of positive tests may still be false. This is the base-rate effect. To calculate PPV, one also needs sensitivity and an appropriate prevalence estimate. That missing information cannot be supplied by grammatical intuition or a general marketing claim of high test quality.

Exercise 10. In a hypothetical study, sensitivity = 90%, specificity = 98%, and apparent prevalence = 10%. The simple Rogan–Gladen-type adjustment gives (0.10 + 0.98 − 1)/(0.90 + 0.98 − 1) = 0.08/0.88 ≈ 9.09% true prevalence under the model assumptions. That output is not automatically a valid population estimate: representative sampling, test independence from reference classification and uncertainty in the sensitivity and specificity estimates must be considered. Preserve the word “model-adjusted” rather than presenting the figure as directly counted animals.

16. A practical evidence and language workflow

Begin with the biological question and the test’s intended use: clinical diagnosis, screening, surveillance, movement certification or research. Record the target and sample matrix, then the organism and case definition. Identify the statistical unit—PCR reaction, specimen, animal, pooled sample, pond or shipment—at each statement. Next record the sampling frame, dates, sample size, pooling and inclusion criteria. Finally note the laboratory’s reported status, validation context, controls, test performance and uncertainty. This order makes it harder to treat a convenient fraction as a farm-wide conclusion.

Translate each outcome with its most direct evidence first: “target detected in this specimen,” “n of N selected specimens positive,” or “no target detected in sampled material under the protocol.” Then say separately what that evidence suggests and which further conditions are needed to establish disease status or population prevalence. This layered style is clearer and more ethical than either extreme—technical jargon without explanation or oversimplified certainty. It allows a veterinarian to act, a manager to understand and a translator to remain faithful without pretending to possess data that the report never supplied.

An audit checklist should include numerical denominators, positive and negative controls, interpretation thresholds, sample identifiers, dates, the population to which conclusions apply and whether the conclusion is screening or confirmation. The writer should also inspect the words “free,” “clear,” “negative,” “infected,” “diseased” and “outbreak.” Those words can have different legal and biological significance. Readers should refer to the currently applicable disease-specific Aquatic Manual and Aquatic Animal Health Code materials; an educational article cannot replace authorised surveillance, health certification or disease notification procedures.

17. FAQ: what can the test really tell us?

Does positive PCR prove the pathogen is alive? Not in general; nucleic-acid detection and viability differ. Is Ct a universal measure of pathogen load? No; assay and specimen calibration matter. Does 99% specificity mean a positive is 99% trustworthy? No; positive predictive value also depends on prevalence. Do 30 negative fish establish that a farm is free of infection? Not without a justified surveillance design and statistical inference. Can one positive pool identify every infected animal? No. Should operators disregard surprising positive screens? No; they should follow suitable confirmation and biosecurity procedures rather than jumping directly to unsupported conclusions.

Is testing more animals always the best response? Sampling design, clustering, timing and assay validation matter as well as quantity. Are apparently healthy animals safe to ignore? Not necessarily, because infection can exist without obvious signs. Can a farm compare two Ct figures from different laboratories? Only if methods and calibration support such a comparison. Does a high sensitivity value remove the need for sample-quality controls? No. Are laboratory and farm-wide prevalence the same? No: diagnostic evidence operates at a specimen or selected-unit level, while population inference depends on sampling and modelling.

18. Learning transfer: translate a technical conclusion without making it stronger

A careful conclusion might say: “In a risk-based survey of 60 pooled tissue specimens, two pools yielded target-positive signals by the validated screening procedure. The results establish detection in those pools under the specified conditions; they do not independently estimate individual-animal prevalence. Confirmatory testing and veterinary assessment will follow the relevant programme.” That sentence identifies the unit, numerator, method and limitation. It remains useful without pretending that certainty appears the moment a PCR machine prints a value. The discipline applies beyond aquaculture to food safety, public health, ecology and many kinds of environmental testing.

The central skill is translating evidence without changing its level. PCR may provide powerful molecular evidence; validation explains how reliable the test is for a defined purpose; sampling explains how far a result can be generalised; and competent veterinary assessment connects evidence to action. A good report carries all four levels, even if the opening summary is brief. When the material might affect animal welfare, disease control or trade, uncertainty should be reported explicitly and the latest competent authority procedures should govern decisions.