Sensitivity

For DNA tests, two questions need to be distinguished:

• Analytical sensitivity: does the test detect the variant being tested for when it is present? For tests for a known mutation, it is usually close to 100 %.
• Clinical sensitivity: how many of the animals that are actually affected are detected by the test? It can be lower if the same clinical picture in a breed can also be caused by other mutations or by non-hereditary causes. An animal with a negative test may then still become affected.

Clear-cut case: monogenic disease caused by one mutation. In Osteogenesis Imperfecta in the Dachshund, a single mutation in the SERPINH1 gene causes the disease, which is inherited recessively. Animals with two copies become affected, animals with one or no copy do not. Here the analytical and the clinical statement coincide: the test is almost 100 % sensitive and specific, and the result reliably predicts the clinical picture.

Analytically correct, clinically not conclusive: PRA. A Golden Retriever has gone blind from progressive retinal atrophy (PRA), but the test for GR-PRA 1 shows n/n. The result is correct; for its mutation the test is 100 % specific. The animal's PRA, however, has a different cause, for example GR-PRA 2, prcd-PRA or a mutation not yet known. In relation to the clinical picture of PRA as a whole, the clinical sensitivity of a single test is therefore less than 100 %. A clear test result only rules out the form tested, not the disease as such.

In multifactorial diseases, several genes and environmental factors interact. A single genetic variant is a risk factor here: it increases the probability of becoming affected, but it is neither necessary nor sufficient. Animals with the variant may remain healthy, animals without it may become affected. Sensitivity and specificity can therefore not be read, as for monogenic diseases, as a measure of whether an animal will become affected.

For such tests, a different approach is taken:

• Assess analytical validity separately: whether the test detects the variants correctly is evaluated as for any DNA test.
• Report the result as a risk: what matters is how much a variant increases the risk of disease compared with animals without this variant. This is determined in studies of affected and healthy animals (relative risk, odds ratio).
• Evaluate variants together: several risk variants are considered jointly, for example via the number of risk alleles or a risk score.
• Take the environment into account: the result is interpreted together with breed, husbandry, feeding and workload. These factors in particular can often be influenced in a targeted way.
• Proceed gradually in breeding: the proportion of risk variants in the population is reduced over generations without jeopardising genetic diversity through strict exclusions from breeding.

Example horse: In Muscle Integrity Myopathy (MIM), variants in six genes are analysed as risk factors. Whether and how strongly a horse shows symptoms depends largely on feeding, training and husbandry. The test shows the genetic susceptibility, not certain disease.

Example human: In type 2 diabetes or hypertension, many gene variants, each with a small effect, contribute to the risk. Lifestyle, diet, physical activity and body weight have a major influence. Genetic tests provide a risk profile here, not a diagnosis.

Sensitivity and specificity are the key criteria used to measure the informative value of a test procedure. How reliable an individual positive or negative result is (predictive value) also depends on how common the disease is (see Bayes Theorem).

Lexicon:

• Specificity
• Bayes Theorem

Literature:

• Trevethan, R.: Sensitivity, specificity, and predictive values: foundations, pliabilities, and pitfalls in research and practice.Front Public Health 5:307, 2017. DOI: 10.3389/fpubh.2017.00307