(18 days)
The ACCESS® Free T4 Assay is a paramagnetic particle, chemiluminescent immunoassay for the quantitative determination of free thyroxine levels in human or plasma (heparin) using the ACCESS® Immunoassav System. A free thyroxine test system is a device intended to measure free (not protein bound) thyroxine (thyroid hormone) in human serum or plasma (heparin). Levels of free thyroxine in plasma are thought to reflect the amount of thyroxine hormone available to the cells and may therefore determine the clinical metabolic status of thyroxine. Measurements obtained by this device are used in the diagnosis and treatment of thyroid diseases.
The ACCESS® Free T4 reagents and the ACCESS® Immunoassay Analyzer comprise the ACCESS® Immunoassay System for the quantitative determination of free thyroxine in human serum or plasma (heparin).
The provided document describes the ACCESS® Free T4 assay, an in vitro diagnostic device, and its substantial equivalence to a predicate device for the quantitative determination of free thyroxine. Since this is an in vitro diagnostic device, the acceptance criteria and study designs are different from a typical AI/ML-based medical device that processes images or other patient data. The concepts of "human readers," "AI assistance," "ground truth," "training set," and "test set" in the context of AI/ML are not directly applicable here.
However, I can extract the closest analogous information from the document to answer your questions as best as possible, interpreting the studies described for an in vitro diagnostic device.
Here's a breakdown based on the provided information:
1. Table of Acceptance Criteria and Reported Device Performance
For in vitro diagnostics, "acceptance criteria" usually refer to established performance characteristics that demonstrate the device is fit for its intended use and comparable to existing methods. "Reported device performance" refers to the results obtained from studies evaluating these characteristics.
| Acceptance Criteria Category (Analogous to AI/ML Performance Metric) | Reported Device Performance (Analogous to AI/ML Reported Metric Value) |
|---|---|
| Precision (Within-run %CV) | 3.11% CV (high control) to 4.13% CV (low control) |
| Precision (Total Imprecision %CV) | 4.56% CV (high control) to 7.42% CV (low control) |
| Correlation with Predicate Device (r value) | 0.940 |
| Correlation with Predicate Device (Deming Regression Slope) | 1.16 |
| Correlation with Predicate Device (Deming Regression Intercept) | -0.275 |
| Analytical Sensitivity (Lowest Detectable Level) | 0.15 ng/dL (distinguishable from zero with 95% confidence) |
Note: The acceptance values for these criteria (e.g., what an acceptable r-value or %CV range is) are not explicitly stated in the provided document but are implied by the FDA's clearance based on the reported performance demonstrating substantial equivalence. The predicate device's performance would serve as an implicit benchmark for these.
2. Sample Size Used for the Test Set and Data Provenance
- Test Set Sample Size: 327 samples were used for the correlation study to compare the ACCESS® Free T4 assay with the predicate device (Abbott IMx® Free T4 assay).
- Data Provenance: Not explicitly stated (e.g., country of origin). The document does not specify if the samples were retrospective or prospective, or from a specific demographic.
3. Number of Experts Used to Establish the Ground Truth for the Test Set and Qualifications of Those Experts
This concept is not directly applicable to an in vitro diagnostic device like the ACCESS® Free T4 assay, which measures a biochemical marker. The "ground truth" for these devices is typically established through a combination of:
- Reference methods or established predicate devices (as seen in the correlation study).
- Certified calibrators and controls with known analyte concentrations.
- The inherent biochemical process being measured.
There are no "experts" in the sense of human readers/interpreters establishing ground truth for the measurement itself.
4. Adjudication Method for the Test Set
Not applicable. This concept is relevant for studies where human interpretation is involved and discrepancies need to be resolved. For an IVD, the measurement output is quantitative.
5. If a Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study Was Done
No, an MRMC comparative effectiveness study was not done. These types of studies are typically performed for imaging or interpretation-based diagnostic devices where multiple human readers evaluate cases, potentially with and without AI assistance, to assess improvements in diagnostic accuracy or efficiency.
6. If a Standalone (Algorithm Only Without Human-in-the-Loop Performance) Was Done
Yes, the studies described are essentially "standalone" performance evaluations for the in vitro diagnostic device.
- Precision studies: Evaluate the inherent reproducibility of the device's measurements.
- Analytical Sensitivity: Determines the lowest quantifiable level for the device.
- Correlation study: Compares the device's quantitative outputs directly against those of a predicate device.
The ACCESS® Free T4 assay is a fully automated measurement system (immunoassay analyzer) that provides a quantitative result without direct human interpretation in the measurement process. "Human-in-the-loop" would refer to a scenario where a human modifies or interprets the device's direct output to reach a final diagnosis, which is not the primary function being evaluated here for the measurement system itself. The human-in-the-loop aspect for such devices is typically in the clinician's interpretation of the result in the context of the patient's clinical picture.
7. The Type of Ground Truth Used
For the correlation study, the "ground truth" was established by the results from the predicate device, the Abbott IMx® Free T4 assay. For precision and sensitivity, the "ground truth" relies on the known concentrations of controls and calibrators, which are themselves traceable to established reference methods or standards.
8. The Sample Size for the Training Set
The document does not explicitly mention a "training set" in the context of AI/ML development. For an in vitro diagnostic device, the development process might involve numerous samples for assay optimization, reagent formulation, and initial validation. However, this is distinct from "training data" for an AI algorithm. The 327 samples for the correlation study and the samples used for precision and sensitivity are considered "test" or "validation" samples in this context.
9. How the Ground Truth for the Training Set Was Established
Not applicable, as there is no explicitly defined "training set" in the AI/ML sense for this in vitro diagnostic device. The "ground truth" for developing such assays involves chemical and biological analyses, calibration standards, and comparison to existing validated methods during the development and optimization phases.
§ 862.1695 Free thyroxine test system.
(a)
Identification. A free thyroxine test system is a device intended to measure free (not protein bound) thyroxine (thyroid hormone) in serum or plasma. Levels of free thyroxine in plasma are thought to reflect the amount of thyroxine hormone available to the cells and may therefore determine the clinical metabolic status of thyroxine. Measurements obtained by this device are used in the diagnosis and treatment of thyroid diseases.(b)
Classification. Class II.