FDA 510(k) Search - By Innolitics (SaMD and AI Experts)

For use with Becton Dickinson FAC® flow cytometers equipped with a blue (488-nm) and red diode (635-nm) laser.
Monoclonal antibody reagents for identification and enumeration of mature human lymphocyte subsets in peripheral blood of normal individuals and patients with certain immune dysfunction.
MultiTEST CD3/CD16+56/CD45/CD19 provides percentages and absolute counts of T (CD3'), NK(CD3 CD16'CD56'), and B (CD3'CD19') lymphocytes.
MultiTEST CD3/CD8/CD45/CD4 provides percentages and absolute counts of T (CD3*), helper/inducer T (CD3CD4) and suppressor/cytotoxic T (CD3CD8) lymphocytes.
For use with erythrocyte lysed whole blood without an isotype control.
To characterize and monitor forms of auto immune diseases, such as lupus.
To characterize and monitor congenital or acquired immunodeficiences, such as SCID or AIDS.
For in vitro diagnostic use

Device Description

The Becton Dickinson CD3 fluorescein isothiocyanate (FITC)/CD16+56 phycoerythrin (PE)/CD45 peridinin chlorophyll protein (PerCP)/CD19 allophycocyanin (APC) reagent is a four-color, direct immunofluorescence reagent for identifying and enumerating percentages and absolute counts of CD3+ T lymphocytes, CD3-CD16+56+ natural killer (NK), and CD3-CD19+ B lymphocyte subsets in erythrocyte-lysed whole blood. Each vial of reagent yields 50 tests. The reagent is intended for use on the Becton Dickinson FACSort™ or FACSCalibur™ flow cytometers equipped with the FL4 Option, the Apple® Macintosh® Quadra or PowerPC computer, and CELLQuest™ or MultiSET™ software. Daily instrument set-up requires CaliBRITE™ beads (unlabeled, FITC, PE, PerCP, and APC) and FACSComp™ software. The four-color method permits the identification and enumeration of T, NK, and B lymphocyte subsets using MultiSET software's expert lymphocyte gate and automated single-tube quality control algorithms. A lymphocyte gate is drawn for the CD45 * leucocytes with low side scatter and the lymphocyte subsets are provided as an absolute count or percentage of total lymphocytes.

AI/ML Overview

This document describes the performance data for the Becton Dickinson MultiTEST CD3/CD16+56/CD45/CD19 and MultiTEST IMK Kit, which are flow cytometry reagents used for identifying and enumerating lymphocyte subsets. The goal of the studies was to demonstrate substantial equivalence to predicate devices (TriTEST CD3/CD16+56/CD45 and TriTEST CD3/CD19/CD45).

1. Table of Acceptance Criteria and Reported Device Performance

The acceptance criteria are not explicitly stated with specific numerical thresholds (e.g., a certain percentage agreement or coefficient of variation). Instead, the studies conclude that the performance is "acceptable" or "equivalent" to the predicate devices. The reported device performance is based on these qualitative assessments.

Performance Metric	Acceptance Criteria (Implicit)	Reported Device Performance
Accuracy	Equivalent to predicate devices	MultiTEST CD3/CD16+56/CD45/CD19 is equivalent to TriTEST CD3/CD16+56/CD45 and TriTEST CD3/CD19/CD45.
Within-site Reproducibility	Acceptable	Demonstrated acceptable within-site reproducibility.
Across-site Reproducibility	Acceptable	Demonstrated acceptable across-site reproducibility.
Stability	Acceptable	Demonstrated acceptable stability of samples prepared up to 48 hours after blood draw and up to 24 hours after preparation.
Linearity and Recovery	Linear response and acceptable recovery	Results indicate linear response over the tested range and acceptable recovery.

2. Sample Sizes Used for the Test Set and Data Provenance

Accuracy Study:
- Sample Size: 129 lysed whole blood (LWB) specimens.
  - 70 normal donors
  - 59 abnormal donors
- Data Provenance: The study was conducted at two clinical sites, indicating prospective collection from human subjects. The country of origin is not specified but implied to be the US given the FDA submission.
Within-site Reproducibility:
- Sample Size: LWB specimens from 3 normal and 6 abnormal donors (total 9 donors). Each specimen was divided into 10 aliquots.
- Data Provenance: Not explicitly stated but likely prospective data from human subjects.
Across-site Reproducibility:
- Sample Size: LWB specimens from a total of 16 normal and 30 abnormal donors (total 46 donors). Each specimen was divided into 5 aliquots.
- Data Provenance: Conducted at three clinical sites, indicating prospective collection from human subjects.
Stability Studies:
- Sample Size: Samples from 10 normal and 20 abnormal donors (total 30 donors).
- Data Provenance: Not explicitly stated but likely prospective data from human subjects.
Linearity and Recovery:
- Sample Size: Blood specimens from 3 normal donors.
- Data Provenance: Not explicitly stated but likely prospective data from human subjects.

3. Number of Experts Used to Establish the Ground Truth for the Test Set and Qualifications of Those Experts

The document does not describe the use of human experts to establish "ground truth" in the traditional sense (e.g., radiologists interpreting images). For flow cytometry reagents, the "ground truth" and comparison are typically established by comparing the results of the new device to a legally marketed predicate device (which itself would have established its performance against other methods or clinical outcomes), or by established laboratory reference methods. In this case, the test set's ground truth for accuracy was based on the performance of the predicate devices. The "normal" and "abnormal" classifications of the donor samples would likely be based on clinical diagnosis or established laboratory ranges, but no explicit expert involvement in defining the ground truth for each individual sample's classification is mentioned beyond the comparison to the predicate.

4. Adjudication Method for the Test Set

Not applicable. The "adjudication" in this context refers to the comparison of the new device's results with those of the predicate device, or established reference methods in a laboratory setting. There is no mention of a human adjudication process by multiple experts for discrepancies in interpretations.

5. If a multi-reader multi-case (MRMC) comparative effectiveness study was done, if so, what was the effect size of how much human readers improve with AI vs without AI assistance

Not applicable. This is not a study involving human readers interpreting medical images or data with or without AI assistance. It's a performance study of diagnostic reagents used in flow cytometry.

6. If a standalone (i.e., algorithm only without human-in-the-loop performance) was done

This study evaluates the performance of a diagnostic reagent kit in conjunction with specific flow cytometers and software (Becton Dickinson FACSort or FACSCalibur flow cytometers, Apple Macintosh computers, CELLQuest™ or MultiSET™ software). While the software includes "expert lymphocyte gate and automated single-tube quality control algorithms," the overall system involves human operation and interpretation of the flow cytometer output. Therefore, it is not a "standalone algorithm only" study in the sense of a fully automated AI system without human interaction. The performance is of the device (reagent kit + specific instrumentation/software) as an integrated system in a laboratory setting.

7. The type of ground truth used (expert consensus, pathology, outcomes data, etc.)

The primary "ground truth" for the accuracy study was the results obtained from the predicate devices (TriTEST CD3/CD16+56/CD45 and TriTEST CD3/CD19/CD45). The study aimed to demonstrate "equivalence" to these already-marketed and accepted devices. For the other studies (reproducibility, stability, linearity), the "ground truth" would be inherent in the expected biological or technical performance for those assays, often against established laboratory controls or expected ranges. The classification of samples as "normal" or "abnormal" would be based on clinical criteria and laboratory values but not necessarily a separate "expert consensus" on the specific flow cytometry results themselves for the purpose of validating the device.

8. The Sample Size for the Training Set

The document does not provide information on a "training set" as it would apply to a machine learning algorithm. This submission describes the validation of a laboratory reagent and associated instrumentation/software, where the "algorithms" (e.g., for gating) are likely rule-based or pre-programmed, rather than learned from a large, labeled training dataset.

9. How the Ground Truth for the Training Set Was Established

Not applicable, as a distinct "training set" and associated ground truth establishment process, typical for machine learning models, is not described for this device. The software's "expert lymphocyte gate and automated single-tube quality control algorithms" would have been developed based on scientific understanding of cell populations and validated through internal R&D processes, rather than a single, distinct "training set" in the AI sense.

Summary

Regulation Number and Section

§ 864.5220 Automated differential cell counter.

(a)
Identification. An automated differential cell counter is a device used to identify one or more of the formed elements of the blood. The device may also have the capability to flag, count, or classify immature or abnormal hematopoietic cells of the blood, bone marrow, or other body fluids. These devices may combine an electronic particle counting method, optical method, or a flow cytometric method utilizing monoclonal CD (cluster designation) markers. The device includes accessory CD markers.(b)
Classification. Class II (special controls). The special control for this device is the FDA document entitled “Class II Special Controls Guidance Document: Premarket Notifications for Automated Differential Cell Counters for Immature or Abnormal Blood Cells; Final Guidance for Industry and FDA.”