IVD Clinical Study Terms, Defined.

CLIA categorization is FDA's assignment of each commercially marketed test to waived, moderate complexity, or high complexity, which determines the certificate a facility needs to run it.

After clearing or approving an IVD, FDA assigns a CLIA complexity category based on criteria such as the knowledge, training, reagent preparation, interpretation, and judgment required to perform the test. Most newly cleared tests are categorized as moderate complexity by default. High complexity applies to tests needing substantial technical skill or interpretation, for example many manual microscopy and molecular procedures. Waived status is granted only on a separate application, or automatically for tests FDA has cleared for home use.

A CLIA waiver is FDA's determination that a test is simple and carries insignificant risk of erroneous results, allowing use in facilities holding only a CLIA Certificate of Waiver.

A CLIA waiver by application asks FDA to categorize a cleared test as waived. The sponsor must show that the test is simple and that it has an insignificant risk of an erroneous result. Simplicity is argued from the device design and labeling. Insignificant risk is demonstrated with two kinds of data: flex studies that stress the test under conditions untrained users might create, and a clinical study in which intended use operators at waived-type sites run the test and their results are compared to a reference method or to results from trained laboratory personnel.

A composite comparator is a reference result built from a predefined combination of two or more methods, used when no single test is accepted as the reference standard.

For some analytes there is no single trusted method. A sponsor may define the comparator as, for example, two cleared molecular assays with a third method or sequencing to adjudicate disagreements, or a combination of culture and PCR. The rule for combining them is written into the protocol before any testing, so that a specimen's reference status is determined by algorithm rather than by judgment after the fact.

A confidence interval is the range of values consistent with the observed data; IVD submissions often judge performance by the lower bound of the two-sided 95 percent interval.

An observed PPA of 95 percent from 40 positives and the same figure from 400 positives carry very different weight. The confidence interval expresses that difference: with 40 positives the interval is wide and the lower bound may fall well below 90 percent, while with 400 it is narrow. FDA expects performance estimates to be reported with two-sided 95 percent confidence intervals, usually computed with an exact or score method appropriate for proportions.

Discrepant resolution is the practice of retesting only specimens where the investigational test and comparator disagree, a design FDA considers biased when used to recalculate performance.

When a new test and its comparator disagree, it is natural to want to know which was right. Discrepant resolution retests only the discordant specimens with a third method and then reclassifies them. The problem is statistical: concordant specimens are never rechecked, so errors shared by both methods are never discovered, and only errors that favor the comparator are corrected. The result is an upward bias in apparent agreement.

A dual submission combines a 510(k) and a CLIA waiver by application in one FDA submission, supported by a single clinical study run in waived-type settings with intended use operators.

FDA allows sponsors to seek 510(k) clearance and CLIA waiver together. The clinical study is designed so that one set of data satisfies both purposes: the test is run by untrained intended use operators at waived-type sites, and results are compared to a comparator method. Because the operators are already the waived population, the agreement data support the 510(k) claim and the waiver claim at once. Flex studies and the other analytical work are included in the same file.

Flex studies are analytical stress tests supporting a CLIA waiver, showing that a test gives correct results or clearly fails when users deviate from instructions or face adverse conditions.

The waiver standard requires an insignificant risk of erroneous results. Flex studies, sometimes called robustness studies, support that claim by deliberately introducing the variations a non-laboratory user might create. Typical stressors include sample volume above or below specification, reading the result early or late, incorrect timing of reagent steps, temperature and humidity extremes for storage and operation, light exposure, device tilt or vibration, and hemolyzed or viscous specimens. The sponsor tests specimens near the cutoff under each condition.

An intended use operator is someone representative of the people who will actually run the test in its labeled setting, such as untrained clinic staff or a lay user.

IVD performance depends on the person performing the test, so FDA expects the clinical study to use operators who match the labeling. For a laboratory test that means trained technologists. For a CLIA waiver study it means staff at physician offices, urgent care clinics, and pharmacies who have no formal laboratory training and who receive only the instructions a purchaser would get. For an over-the-counter product it means lay users testing themselves or a family member.

ISO 20916 is the international good study practice standard for IVD clinical performance studies using human specimens; ISO 14155 governs clinical investigations of other medical devices and excludes IVDs.

ISO 14155 is the widely cited standard for medical device clinical investigations, but its scope explicitly excludes in vitro diagnostic devices. IVD clinical performance studies differ in kind: subjects provide specimens rather than receiving an intervention, the result is usually not used for their care, and the risk is informational rather than physical. ISO 20916, first published in 2019, was written to address these studies and describes planning, conduct, recording, and reporting of clinical performance studies for IVDs.

A lay user study has untrained members of the public perform an IVD on themselves or a family member using only the package instructions, supporting an over-the-counter or home-use claim.

The lay user study is the central evidence for a home-use IVD. Subjects representing the intended consumer population, spanning ages, education levels, and familiarity with health products, are given the retail test and its instructions and asked to perform it without help. Study staff observe without intervening, record errors and difficulties, and collect a comparator specimen. Performance of the test in lay hands is reported alongside the rate of invalid results and user errors.

The limit of detection is the lowest analyte concentration that a test can reliably distinguish from a blank, typically defined as the level detected in about 95 percent of replicates.

LoD is the central analytical sensitivity claim for most qualitative IVDs. It is established by testing serial dilutions of a characterized material, such as quantified virus or a reference standard, in the claimed specimen matrix, with enough replicates at each level to identify the concentration at which nearly all results are positive. The CLSI EP17 document describes the accepted approach, including the related concepts of limit of blank and limit of quantitation for quantitative assays.

Negative percent agreement is the proportion of comparator-negative specimens the investigational test also calls negative, used instead of specificity when the comparator is not a true reference standard.

NPA is the negative-side counterpart to PPA. It is the number of specimens negative by both the investigational test and the comparator divided by the number negative by the comparator, reported with a 95 percent confidence interval. In a prospective all-comers study, negatives are usually plentiful, so NPA tends to have a tight confidence interval while PPA does not. For that reason NPA is rarely the binding constraint on sample size, although it can become one when the sponsor claims very high specificity and the margin above the criterion is small.

An over-the-counter or home-use IVD is a test sold directly to consumers for self-testing without a prescription, which FDA evaluates for lay-user performance and treats as CLIA waived.

Home-use IVDs include pregnancy tests, glucose meters, and, since the COVID-19 pandemic, a growing range of rapid infectious disease tests. Because the user has no training and no clinician present, FDA evaluates the device as a complete system of test, instructions, and packaging, and asks whether an untrained person can collect the specimen, run the test, read the result, and understand what to do next. Tests cleared or approved for home use are categorized as CLIA waived.

Point-of-care testing is diagnostic testing performed at or near the site of patient care, outside a central laboratory, with results available during the same encounter.

Point-of-care tests are run in urgent care clinics, physician offices, emergency departments, pharmacies, and similar settings by staff whose main job is patient care rather than laboratory work. The value is immediate results that change management during the visit, for example prescribing an antiviral or antibiotic. In the United States the setting is governed by CLIA, and most point-of-care tests sold to non-laboratory facilities must be CLIA waived.

Positive percent agreement is the proportion of comparator-positive specimens the investigational test also calls positive, used instead of sensitivity when the comparator is not a true reference standard.

FDA draws a distinction between sensitivity and agreement. Sensitivity is measured against a reference standard that is treated as truth. When the comparator is itself an imperfect test, for example another cleared assay, the correct term is positive percent agreement, because disagreements may be errors of either test. Most IVD 510(k) studies report PPA and NPA rather than sensitivity and specificity for exactly this reason.

Prevalence is the proportion of tested subjects who have the target condition; because IVD sample size is driven by positives required, lower prevalence means more subjects must be enrolled.

The precision of a sensitivity or PPA estimate depends on the number of true or comparator positives, not on total enrollment. If a study needs a given number of positives and the condition is present in one in ten tested subjects, it must enroll roughly ten times that number; at one in fifty, fifty times. Negatives accumulate as a by-product and are rarely the constraint. This arithmetic, more than any other factor, determines the cost and duration of an IVD clinical study.

Prospective specimen collection is the enrollment of consenting subjects to provide fresh specimens, with associated clinical data, under an IRB-approved protocol for use in IVD development or validation.

IVD developers need characterized human specimens long before a pivotal study, for feasibility, cutoff setting, analytical studies in real matrix, and verification. Purchasing banked specimens is one route; prospective collection is the other. In a prospective collection, sites enroll subjects who meet defined criteria, collect the specimen types the sponsor needs, record the requested data such as demographics, symptoms, onset, and medications, and ship the specimens fresh or process them to the sponsor's specification.

Substantial equivalence is the 510(k) standard: a new device shares the predicate's intended use and has the same technological characteristics, or differences that raise no new safety or effectiveness questions.

Substantial equivalence is a comparative standard, not an absolute one. FDA does not ask whether the new IVD is good in isolation; it asks whether the data show it performs at least as well as a device already on the market for the same intended use. In practice this means the predicate sets the bar. If the predicate reports a given clinical sensitivity, a new test with materially lower sensitivity will have difficulty establishing equivalence even if it is adequate for clinical purposes.

An umbrella IRB protocol is a broad, already-approved protocol covering collection of defined specimen types and data from consenting subjects, so new collections start without a study-specific IRB submission.

Rather than writing and submitting a new protocol for each sponsor's specimen needs, a site network can hold a general IRB-approved protocol that describes the specimen types, volumes, populations, consent process, and data fields it is permitted to collect. When a sponsor needs specimens, the collection is set up within the approved scope, sites are notified, and enrollment begins. This typically compresses the startup from months to about a week for common specimen types such as nasal swabs, throat swabs, blood, urine, or stool.

FDA and the Common Rule allow certain IVD studies using leftover, de-identified specimens to proceed without individual informed consent, under documented conditions typically confirmed by an IRB.

Many IVD analytical and some clinical studies use remnant specimens that were collected for routine clinical care and would otherwise be discarded. FDA has issued guidance describing when it will exercise enforcement discretion and not require informed consent for studies using leftover specimens that are not individually identifiable. The conditions generally include that specimens were not collected for the study, that they are de-identified before the investigator receives them, and that the study is otherwise compliant with device regulations.