The CellSearch™ Epithelial Cell Kit is intended for the enumeration of circulating tumor cells (CTC) of epithelial origin (CD45-, EpCAM+, and cytokeratins 8, 18+ and/or 19+) in whole blood. The presence of CTC in the peripheral blood, as detected by the CellSearch™ Epithelial Cell Kit, is associated with decreased progression free survival and decreased overall survival in patients treated for metastatic breast cancer.

Device Description

The CellSearch™ Epithelial Cell Kit analyzed on the CellSpotter™ Analyzer is called the CellSearch Assay. The CellSearch Assay is a semi-automated in vitro diagnostic device. Epithelial cells are immunomagnetically labeled by targeting the Epithelial Cell Adhesion Molecule (EpCAM) antigen. Anti-EpCAM monoclonal antibodies conjugated to ferrofluid particles are colloidal and, when mixed with a sample containing the target epithelial cells, bind to the EpCAM antigen associated with the epithelial cells. After immunomagnetic selection of epithelial cells from 7.5 mL of blood, fluorescent reagents are added at this time to discriminate between the immunomagnetically selected cells. Anti-Cytokeratin - Phycoerythrin (CK-PE) stains the intracellular cytoskeleton cytokeratin proteins expressed in cells of epithelial origin, anti-CD45-Allophycocyan (CD45-APC) stains leukocytes and DAPI stains DNA present in the cell nucleus. A strong magnetic field is applied to the processed reagent/sample mixture that causes the labeled target cells to move to the cartridge surface. The cartridge is then placed on the CellSpotter™ Analyzer for data acquisition and analysis. The CellSpotter™ Analyzer acquires images of PE, APC and DAPI fluorescence staining of the entire viewing surface. After data acquisition is completed, the images are analyzed for any event where cytokeratin-PE and DAPI are within a specified space in the CellSpotter™ Cartridge, i.e. indicating the possible presence of a cell with a nucleus that expresses cytokeratin. Images from each fluorescent color as well as a composite image of the cytokeratin staining (green) and the nuclear staining (purple) are presented to the user in a gallery for final cell classification. A cell is classified as a tumor cell when it its EpCAM+ (i.e., it is captured), CK+, DAPI+ and CD45-. A check mark placed by the operator next to the composite images classifies the event as a Circulating Tumor Cell (CTC) and the software tallies all the checked boxes to obtain the CTC count.

AI/ML Overview

Here's a breakdown of the acceptance criteria and the study details for the CellSearch™ Epithelial Cell Kit and CellSpotter™ Analyzer, based on the provided document:

1. Table of Acceptance Criteria and Reported Device Performance

The acceptance criteria are implicitly derived from the performance goals demonstrated in the clinical and analytical studies, particularly concerning the predictive power of the CTC count cutoff for patient outcomes. The primary clinical acceptance is based on the device's ability to identify patients with decreased Progression-Free Survival (PFS) and Overall Survival (OS) using a defined cutoff.

Performance Metric	Acceptance Criteria (Implicit from Study)	Reported Device Performance
Analytical Performance
Precision (Low Spike)	Consistent and reproducible cell counts	Mean CTC Count per 7.5mL: 47, Total Precision Standard Deviation (ST) % CV: 15.8% (N=80)
Precision (High Spike)	Consistent and reproducible cell counts	Mean CTC Count per 7.5mL: 258, Total Precision Standard Deviation (ST) % CV: 9.4% (N=80)
Reproducibility (Patient Samples <5CTC)	Reproducible CTC counts across different systems/sites	Number of Duplicates: 123, Mean CTC Count of Duplicates: 0.7, Avg. Duplicate Standard Deviation: 0.5, Avg. %CV of Duplicates: 60.0% (Note: High %CV due to very low counts, common in this range)
Reproducibility (Patient Samples ≥5CTC)	Reproducible CTC counts across different systems/sites	Number of Duplicates: 40, Mean CTC Count of Duplicates: 210.5, Avg. Duplicate Standard Deviation: 12.0, Avg. %CV of Duplicates: 20.0%
Linearity/Reportable Range	Linear response over a broad range of cell counts	Linear over 4 to 1022 cells per 7.5 mL, Regression equation: y=0.99x +5.71, r-=0.9912 (Intercept 95% CI -0.29 to 11.4, overlapping zero).
Detection Limit	Ability to detect low numbers of CTCs	1 CTC per 7.5 mL of blood can be detected. Analytical sensitivity: 1 CTC in a CellSpotter™ Cartridge.
Analytical Specificity (Interference)	No significant interference from common substances/conditions	No significant interference from various cancer drugs, OTC drugs, other exogenous substances (except doxorubicin), lipemia, hemolysis, icterus, HAMA 1/2, and hematocrit 18-60%. Doxorubicin causes identifiable aberrant staining if drawn within recommended washout period.
Normal Background CTCT	CTC counts in healthy/non-malignant individuals are low	Of 345 control subjects (healthy, non-malignant breast/other disease), only 1 had >5 CTC/7.5mL (Table 3 shows means of 0.1-0.2, max 1-12, all single cases were below 5 for healthy and non-malignant other disease, one false positive for non-malignant breast disease).
Cell Recovery	Efficient recovery of spiked cells	Mean recovery of spiked cells approximately 85% (regression equation Y=0.85x +5.64, R=0.9973).
Clinical Performance
PFS Prediction (Baseline)	CTC count of ≥5 should predict decreased PFS	Baseline CTC ≥5 group (N=87) had median PFS of 11.7 weeks vs. <5 CTC group (N=90) with 30.3 weeks. Log-rank p=0.0001, Cox Hazards Ratio=1.9547.
PFS Prediction (1st Follow-up)	CTC count of ≥5 should predict decreased PFS post-treatment	1st Follow-up CTC ≥5 group (N=43) had median PFS of 5.7 weeks vs. <5 CTC group (N=111) with 26.4 weeks. Log-rank p<0.0001, Cox Hazards Ratio=2.4842.
OS Prediction (Baseline)	CTC count of ≥5 should predict decreased OS	Baseline CTC ≥5 group (N=87) had median OS of 43.3 weeks vs. <5 CTC group (N=90) with >80 weeks. Log-rank p<0.0001, Cox Hazards Ratio=4.3865.
OS Prediction (1st Follow-up)	CTC count of ≥5 should predict decreased OS post-treatment	1st Follow-up CTC ≥5 group (N=49) had median OS of 30.0 weeks vs. <5 CTC group (N=114) with >80 weeks. Log-rank p<0.0001, Cox Hazards Ratio=5.4537.
Predictive Value of CTC Reduction	Decrease in CTC count to <5 should predict improved PFS and OS	Patients with <5 CTC at both time points or a decrease to <5 CTC at 1st follow-up showed significantly improved PFS (median 30.3/32.9 weeks) and OS (>80/62.6 weeks) compared to those with ≥5 CTC at 1st follow-up (PFS 8.9 weeks, OS 35.4 weeks), all with high statistical significance (p<0.0006 for PFS, p<0.0007 for OS).

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

Analytical Test Sets (Precision, Linearity, Recovery, Specificity):
- Spiked Specimens (SKBr-3 cells): 80 samples for low spike and 80 for high spike in precision studies (over 20 days, using blood from 20 different normal donors). 25 samples for linearity (serial dilutions of SKBr-3 cells spanning 4 to 1022 cells). 25 samples for recovery (serial dilutions of SKBr-3 cells spanning 4 to 1142 cells).
- Patient Specimens (Reproducibility): 163 duplicate samples from 47 patients.
- Control Subjects (Normal Background): 145 healthy volunteers, 101 women with non-malignant breast disease, 99 women with other non-malignant diseases (Total N=345).
- Data Provenance: The document does not explicitly state the country of origin. The clinical trial involving metastatic breast cancer patients was a "multi-center prospective, longitudinal clinical trial," implying multiple sites within a country or countries (likely the US, given FDA submission). The analytical studies involve healthy donors and patient samples, also without explicit geographic origin. The clinical trial data is prospective.
Clinical Test Set:
- Metastatic Breast Cancer Patients: N=177 patients for the main clinical trial.
- Training Set (for CTC Cutoff determination): 90 patients (subset of the 177).
- PFS using Baseline CTC: N=177 (Split into <5 CTC group N=90, ≥5 CTC group N=87)
- PFS using 1st Follow-up CTC: N=154 (177 - 23 non-evaluable). Split into <5 CTC group N=111, ≥5 CTC group N=43.
- Predictive Value of CTC on PFS (3 groups): N=163. Negative (N=81), Decrease to <5 (N=33), Positive (N=49).
- OS using Baseline CTC: N=177 (Split into <5 CTC group N=90, ≥5 CTC group N=87)
- OS using 1st Follow-up CTC: N=163 (177 - 14 deaths/lost to follow-up, based on text: "56 patients died of the 163 patients who were evaluable at the first follow-up"; implies 177-14=163 eval at FU - this number seems to fluctuate slightly with PFS for FU analyses). Split into <5 CTC group N=114, ≥5 CTC group N=49.
- Predictive Value of CTC on OS (3 groups): N=163. Negative (N=81), Decrease to <5 (N=33), Positive (N=49).
- Data Provenance: Multi-center prospective, longitudinal clinical trial.

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

The document does not mention experts being used to establish a ground truth for individual CTC counts in the test set.

Ground Truth for Clinical Performance: The ground truth for the clinical studies (PFS and OS) was established by clinical outcomes data (disease progression determined by CT scans and/or clinical signs/symptoms, and patient death). The determination of CTC counts by the CellSearch™ system itself (which involves an operator to classify cells) serves as the "device output" that is then correlated with these clinical outcomes.
CTC Final Classification: The CellSpotter Analyzer presents images to a user "for final cell classification." "A check mark placed by the operator next to the composite images classifies the event as a Circulating Tumor Cell (CTC) and the software tallies all the checked boxes to obtain the CTC count." This indicates human involvement in the final count, but not "experts" establishing a "ground truth" independent of the device's output and operator.

4. Adjudication Method for the Test Set

Not applicable in the usual sense of expert adjudication of device output. The device itself involves a human operator to perform the final classification of potential CTCs identified by the analyzer. This is inherent to the "semi-automated" nature of the device. There's no mention of multiple operators adjudicating discrepant counts or external expert review of the operator's classifications.

5. If a Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study was Done

No, a multi-reader multi-case (MRMC) comparative effectiveness study comparing human readers with and without AI assistance was not done. This device is a semi-automated system where the "human reader" (operator) is an integral part of its final output (classifying the identified events as CTCs). The study focuses on the system's ability to count CTCs and correlate those counts with clinical outcomes of metastatic breast cancer patients.

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

No, a standalone algorithm-only performance study was not done. The device description explicitly states it is a "semi-automated" system. The CellSpotter™ Analyzer "acquires images and displays any event to the user where CK-PE and DAPI are co-located. Images are presented to the user in a gallery format for final classification of the magnetically captured cells. An event is classified as a tumor cell when its morphological features are consistent with that of a cell and it exhibits the correct phenotype, i.e., EpCAM+, CK+, DAPI+ and CD45-. A check mark placed by the operator... classifies the event as a Circulating Tumor Cell (CTC)." This clearly indicates a human-in-the-loop process for obtaining the final CTC count.

7. The Type of Ground Truth Used

Analytical Performance:
- For precision, linearity, and recovery, the ground truth was spiked cells of a known concentration (SKBr-3 cell line).
- For analytical specificity, the ground truth was controlled exposure to known interfering substances or physiological conditions.
- For normal background, the ground truth was blood samples from presumptively healthy/non-malignant individuals.
Clinical Performance:
- For the determination of the clinical cutoff (5 CTCs) and the clinical utility studies (PFS and OS), the ground truth was clinical outcomes data: disease progression (diagnosed by CT scans and/or clinical signs and symptoms) and overall survival (time to death).

8. The Sample Size for the Training Set

CTC Cutoff Determination Training Set: 90 patients. This was a subset of the 177 patients from the multi-center prospective clinical trial.

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

The "ground truth" for the training set (in the context of determining the optimal CTC cutoff) was established by correlating the device's CTC counts with clinical progression-free survival (PFS) outcomes.

The training set of 90 patients (who were not diagnosed with disease progression before or at first follow-up) was used.
PFS was calculated from the date of the 1st follow-up.
The system analyzed various CTC cutoffs (from 1 to 82, and up to 10,000) for their ability to differentiate median PFS for positive and negative patient groups.
The optimal cutoff (5 CTCs) was determined based on three criteria:
1. Above normal background levels.
2. PFS of positive patients reached an initial plateau.
3. Highest Cox hazard's ratio at or adjacent to the plateau.
This process involved statistical analysis of the relationship between CTC counts obtained by the semi-automated device (with human operator classification) and the observed clinical PFS data. The FDA statistician, Harry Bushar, Ph. D., reviewed and agreed with the chosen optimal cutoff.

Summary

{0}------------------------------------------------

510(k) SUBSTANTIAL EQUIVALENCE DETERMINATION DECISION SUMMARY

A. 510(k) Number:

K031588

B. Analyte:
EpCam, Cytokeratins 8, 18 and/or 19, and CD45
C. Type of Test:
Semi-automated Immunomagnetic and immunofluorescent

D. Applicant:

Veridex, LLC, A Johnson and Johnson Company

E. Proprietary and Established Names:

Cell Search Epithelial Cell Kit and CellSpotter Analyzer

F. Regulatory Information:

1. Regulation section: 21 CFR 866.6020 Immunomagnetic circulating cancer cell selection and enumeration system
Classification: 2. Class II
1. Product Code:

NQI System, immunomagnetic, circulating cancer cell, enumeration

1. Panel:
  Pathology 88

G. Intended Use:

The CellSearch™ Epithelial Cell Kit is intended for the enumeration of circulating tumor cells (CTC) of epithelial origin (CD45-, EpCAM+, and cytokeratins 8, 18+ and/or 19+) in whole blood.

1. Indication(s) for use:
  The presence of CTC in the peripheral blood, as detected by the CellSearch™ Epithelial Cell Kit, is associated with decreased progression free survival and decreased overall survival in patients treated for metastatic breast cancer.
1. Special condition for use statement(s): For prescription use only.
1. Special instrument Requirements:

The CellPrep™ Semi-Automated Cell Preparation System and the CellSpotter™ Analyzer. The CellSpotter™ Analyzer is a semi-automated fluorescence microscope intended to enumerate fluorescently labeled cells that are immunomagnetically selected and distributed over a viewing surface

H. Device Description:

The CellSearch™ Epithelial Cell Kit analyzed on the CellSpotter™ Analyzer is called the CellSearch Assay. The CellSearch Assay is a semi-automated in vitro diagnostic device.

Epithelial cells are immunomagnetically labeled by targeting the Epithelial Cell Adhesion Molecule (EpCAM) antigen. Anti-EpCAM monoclonal antibodies

{1}------------------------------------------------

conjugated to ferrofluid particles are colloidal and, when mixed with a sample containing the target epithelial cells, bind to the EpCAM antigen associated with the epithelial cells. After immunomagnetic selection of epithelial cells from 7.5 mL of blood, fluorescent reagents are added at this time to discriminate between the immunomagnetically selected cells. Anti-Cytokeratin - Phycoerythrin (CK-PE) stains the intracellular cytoskeleton cytokeratin proteins expressed in cells of epithelial origin, anti-CD45-Allophycocyan (CD45-APC) stains leukocytes and DAPI stains DNA present in the cell nucleus. A strong magnetic field is applied to the processed reagent/sample mixture that causes the labeled target cells to move to the cartridge surface. The cartridge is then placed on the CellSpotter™ Analyzer for data acquisition and analysis. The CellSpotter™ Analyzer acquires images of PE, APC and DAPI fluorescence staining of the entire viewing surface.

After data acquisition is completed, the images are analyzed for any event where cytokeratin-PE and DAPI are within a specified space in the CellSpotter™ Cartridge, i.e. indicating the possible presence of a cell with a nucleus that expresses cytokeratin. Images from each fluorescent color as well as a composite image of the cytokeratin staining (green) and the nuclear staining (purple) are presented to the user in a gallery for final cell classification. A cell is classified as a tumor cell when it its EpCAM+ (i.e., it is captured), CK+, DAPI+ and CD45-. A check mark placed by the operator next to the composite images classifies the event as a Circulating Tumor Cell (CTC) and the software tallies all the checked boxes to obtain the CTC count.

Substantial Equivalence Information: I.

1. Predicate device name(s): None.
1. Predicate K number(s): None.
1. Comparison with predicate:

DEVICE	PREDICATE
A. Similarities
Not Applicable	Not Applicable
B. Differences
Not Applicable	Not Applicable

J. Standard/Guidance Document Referenced (if applicable):

1. NCCLS Approved Guideline EP5A, "Evaluation of Precision Performance of Clinical Chemistry Devices".
1. NCCLS Approved Guideline C28-A2, "How to Define and Determine Reference Intervals in the Clinical Laboratory".

K. Test Principle:

The CellSearch™ Kit contains a ferrofluid-based capture reagent and immunofluorescent reagents. The ferrofluid reagent consists of nano-particles with a magnetic core surrounded by a polymeric layer coated with antibodies targeting the Epithelial Cell Adhesion Molecule (EpCAM) antigen for capturing CTC. After immunomagnetic capture and enrichment, fluorescent reagents are added for

{2}------------------------------------------------

identification and enumeration of CTC. Anti-CK-PE is specific for the intracellular protein cytokeratin (characteristic of epithelial cells), DAPI stains the cell nucleus, and anti-CD45-APC is specific for leukocytes.

The reagent/sample mixture is dispensed by the CellPrep™ Semi-Automated Cell Preparation System into a CellSpotter™ Cartridge that is inserted into a MagNest™ fixture, a device of two magnets held together by steel. The strong magnetic field of the MagNest™ fixture causes the magnetically-labeled epithelial cells to move to the surface of the cartridge. The CellSpotter™ Analyzer automatically scans the entire surface of the CellSpotter™ Cartridge, acquires images and displays any event to the user where CK-PE and DAPI are co-located. Images are presented to the user in a gallery format for final classification of the magnetically captured cells. An event is classified as a tumor cell when its morphological features are consistent with that of a cell and it exhibits the correct phenotype, i.e., EpCAM+, CK+, DAPI+ and CD45-.

L. Performance Characteristics (if/when applicable):

1. Analytical performance:
- a. Precision/Reproducibility:
  - System Reproducibility with Spiked Specimens i. Each day, blood from a single healthy donor was pooled and 7.5 mL samples were spiked with SKBr-3 cells. Low cell spikes (~58 cells/7.5mL) and high cell spikes (~319 cells/7.5mL) were prepared on CellPrep™ instruments and run in duplicate twice each day. This process was repeated each day for a period of 20 days as per NCCLS guideline EP5-A using blood from a total of twenty different normal donors. Over the course of the study, three operators, two different CellPrep™ Systems and two different CellSpotter™ Analyzers were used to generate the cell count data. Summary statistics for cell counts of the spiked samples are presented in Table 1.

Table 1. Summary of Precision Analyses

	Low	High
N	80	80
Mean CTC Count per 7.5mL	47	258
Total Precision Standard
Deviation (ST) % CV	15.8%	9.4%

ii. System Reproducibility with Patient Specimens A total of 163 duplicate samples were collected from 47 patients over the course of the clinical study. These samples were processed separately on multiple systems at different sites (including different CellPrepTM instruments) to determine the reproducibility of CTC measurements. The regression equation for the comparison of these 163 duplicate samples was y=0.98x + 0.67, r2=0.9978. Table 2 shows the summary of the data for

{3}------------------------------------------------

replicates where the average of the two CTC results was <5 compared to those where the average was >5.

Table 2. Reproducibility of CTC Counts in Duplicate Samples (n=163) with an Average of <5 or >5 CTC per 7.5 mL of blood

	CTC <5	CTC ≥5
Number of Duplicates	123	40
Mean CTC Count of Duplicates	0.7	210.5
Avg. Duplicate Standard Deviation	0.5	12.0
Avg. %CV of Duplicates	60.0%	20.0%

b. Linearity/assay reportable range:

Blood from a single healthy donor was pooled and five 7.5 mL samples were spiked with SKBr-3 cells. Serial dilutions were performed resulting in twenty-five samples with expected cell counts ranging from 4 to 1022 cells per 7.5 mL. These samples were processed and analyzed over five consecutive days and the observed cell count results were plotted against the expected cell counts. To determine the overall, or least squares fit, for the comparison of the observed and expected cell counts across all the data, linear regression analysis was performed. The regression equation for these 25 samples was y=0.99x +5.71, r-=0.9912. The 95% confidence interval of the intercept of 5.71 was -0.29 to 11.4, which overlaps zero. In this analytical study, the CellSearch™ Kit was shown to be linear over the tested range from 4 to 1022 cells.

Traceability (controls, calibrators, or method): c. No recognized reference material or methods.
Detection limit (functional sensitivity): d. One CTC per 7.5 mL of blood can be detected by the CellSpotter™ Analyzer resulting in a limit of detection of 1 CTC in a CellSpotter™ Cartridge. An average of 85% of CTC are recovered through the sample 7.5 mL processing (see Recovery section), the CTC loss of approximately 15% is not sufficient to reduce the analytical sensitivity of 1 CTC in a 7.5 mL blood sample. This CTC loss however does indicate that no CTC will be detected in approximately 15 of 100 samples in which only one EpCAM+, cytokeratins 8, 18, and/or 19+, CD45- circulating tumor cell is present in 7.5 mL of blood.
Analytical specificity: e.

Interfering Substances - SKBr-3 cells spiked into blood samples were exposed to potential interfering substances and compared to untreated controls. Toxic levels (5 times therapeutic index) of the following cancer drugs, over-the-counter drugs, and other exogenous substances were tested: cyclophosphamide, Mitomycin C, Procrit,

{4}------------------------------------------------

biotin, 5-fluorouracil, methotrexate, Tamoxifen Citrate, paclitaxel, Arimidex, acetaminophen, acetylsalicylic acid, caffeine, dextromethorphan, Aredia, Human Anti-Mouse Antibody (HAMA) type 1, HAMA type 2, Herceptin, and ibuprofen. No significant differences in SKBr-3 cell numbers were detected, indicating that these substances do not interfere with the CellSearch™ kit.

Samples spiked with toxic levels of doxorubicin resulted in aberrant staining of leukocytes as cytokeratin and CD45 dual positive cells, due to the doxorubicin being a fluorescence compound that is incorporated into nucleated cells. If seen, the staining pattern, of all cells being CD45 positive and cytokeratin positive, is obvious and easily identified by the operator as a known interference staining profile. If blood is drawn outside of the recommended 7 day washout period following doxorubicin infusion, this interference is unlikely to be observed in clinical practice, given controlled therapeutic levels and rapid drug clearance.

Potential interference from lipemia was studied by adding Intralipid to samples to a concentration of 2.6%, which is greater than 1000 mg/dl triglyceride. Samples were lysed to simulate total hemolysis. Bilirubin at 7.4 mg/dL, HAMA 1/HAMA 2 and hematocrit from 18-60% were studied. Lipemia, hemolysis, icterus and a broad range of hematocrit values do not interfere with the CellSearch™ assay. HAMA 1 and HAMA 2 also do not interfere. indicating that individuals receiving mouse Ig by parenteral routes can be tested successfully with the CellSearch 110 assay.

f. Assay cut-off:
A CTC count of 5 or more per 7.5mL of blood is predictive of shorter progression free survival and overall survival.

Median progression free survival (PFS) times and Cox hazard's ratios over a range of circulating tumor cell (CTC) cutoffs in the Training Set using 1st follow-up data were used to determine the optimal CTC cutoff for the prediction of PFS. The PFS time for each patient used in this analysis was calculated from the date of the 1st follow-up. Ninety (90) of the 102 patients in the Training Set were not diagnosed with disease progression before or at first follow-up (Table 2). For every possible cutoff from 1 to 82 CTC, and at selected cutoffs up to 10,000 CTC, the median PFS was calculated for the positive and negative patient groups. In addition, the percentage of positive patients at each CTC cutoff was calculated. The following criteria, listed according to priority, were used to determine the optimal CTC cutoff:

{5}------------------------------------------------

The CTC cutoff must be above normal background (See Table 3)

The CTC cutoff at which the median PFS initially reaches a 2. plateau in the positive patients (red line, Figure 1).

The CTC cutoff at or adjacent to the plateau defined in #2 3. above, which has the highest Cox hazard's ratio (black line, Figure 1A).

The determination of the CTC cutoff in the 90 patient Training Set is illustrated in Figure 1. In Figure 1 the median PFS for positive patients (red line), negative patients (green line), and the Cox hazard ratio (black line) are plotted against the various CTC cutoffs. Only cutoffs with a value change are shown. For example, the CTC cutoffs of 16 to 19 are not shown on the graph because the median PFS of positive and negative patients and the Cox hazard's ratio at these cutoffs are the same as the cutoff of 15 CTC. Figure 1 clearly demonstrates that the presence of CTC has important prognostic implications for the patients. The median PFS of all 90 patients is 5.0 months and is represented in the figure by the 0 CTC cutoff value (blue dot). As an example, at a CTC cutoff of > 1, patients with < 1 (green line) CTC had a median PFS of 6.3 months whereas patients with > 1 CTC (red line) had a PSF time of 3.4 months.

The median PFS reached an initial plateau in the positive patient group at a PSF of 1.4 months for a cutoff of 5 or more CTC (criterion 2 above). This 5 CTC cutoff is well above the normal CTC background of 0 to 2 CTC (criterion 1 above). The third criterion used to define the cutoff was the Cox hazard's ratio; at 5 CTC the Cox hazard's ratio was 2.41, at 4 CTC it was 2.36 and at 6 CTC it also was 2.34. Thus, the 5 CTC cutoff provided the highest Cox hazard's ratio at or adjacent to the plateau defined by the median PFS. Based on the criteria outlined above, a cutoff of 5 or more CTC was chosen, and is indicated by a vellow line in Figure 1.

The cutoff of 5 CTC was determined using only results from the 1st follow-up and PFS as the outcome. Additional analyses have shown that the cutoff for OS at the 1st blood draw, as well as for PFS and OS at baseline, may differ from 5 CTC. For purposes of uniformity and simplicity in the interpretation of test results (by the clinician), a CTC cutoff of 5 was used for all analyses.

{6}------------------------------------------------

Image /page/6/Figure/1 description: This figure shows three different plots. The first plot shows the Cox Hazard's Ratio, the second plot shows the median PFS for patients with less than CTC# at the 1st follow-up blood draw, and the third plot shows the median PFS for patients with greater than or equal to CTC# at the 1st follow-up blood draw. The x-axis shows the 1st follow-up blood draw CTC cutoff, and the y-axis shows the median PFS in months and the Cox Hazard's Ratio.

Figure 1. Determination of CTC Cutoff Using Median PFS in Training Set (n=90)

Figure 2 shows the median PFS for the positive patients (red line) and the percentage of positive patients (black line) for the Training Set at the cutoffs where either of these values changed. At the 5 CTC cutoff, 30% of patients are within the positive group and have a median PFS of 1.4 months.

Figure 2

Image /page/6/Figure/5 description: The image shows two line graphs that compare the percentage of patients with a circulating tumor cell (CTC) number at the first follow-up blood draw and the median progression-free survival (PFS) for patients with a CTC number at the first follow-up blood draw. The x-axis represents the CTC cutoff number, while the left y-axis represents the percentage of patients, and the right y-axis represents the median PFS in months. As the CTC cutoff number increases, both the percentage of patients and the median PFS generally decrease. For example, at a CTC cutoff of 1, approximately 54% of patients have a CTC number at the first follow-up blood draw, with a median PFS of 3.4 months.

The cutoff chosen also corresponded well with expected values determined with populations of healthy volunteers, persons with non-malignant breast disease, and persons with non-malignant other

{7}------------------------------------------------

disease. (See Table 3). Single point CTC analyses were performed on control groups of 145 healthy volunteers, 101 women with non-malignant breast disease, and 99 women with other non-malignant diseases.

Epithelial cells are not expected to be in the peripheral blood. Of the 345 total samples from healthy volunteers and women with nonmalignant disease, only one subject had more than 5 CTC/7.5 mL. The results are presented in Table 3.

Table 3. Control Subjects

Category	N	Mean# CTC	SD	# Patientswith > 5 CTC	Max.*
Healthy	145	0.1	0.2	0	1
Non-malignantbreast disease	101	0.2	1.2	1	12
Non-malignantother disease	99	0.1	0.4	0	3

NCCLS Guideline C28-A2

Based on the above data and analyses, the optimal CTC cutoff was determined to be at 5 CTC per 7.5 mL of blood. The FDA statistician, Harry Bushar, Ph. D., checked all of the statistical claims and agreed that the cutoff chosen by the sponsor was optimal.

2. Comparison studies:

Method comparison with predicate device: a. Since the claimed predicate device was so different from the new device, no comparative studies were able to be performed.
b. Matrix comparison: Since the only matrix claimed was blood, no matrix comparison studies were necessary nor performed.
1. Clinical studies:
- Clinical sensitivity: a.

Not done. Kaplan-Meier analysis.

b. Clinical specificity: Not done. Kaplan-Meier analysis
Other clinical supportive data (when a and b are not applicable) C.

Summary of Clinical Trial Results

Metastatic Breast Cancer Patients

A multi-center prospective, longitudinal clinical trial was conducted to determine whether trends in the number of CTCs correlates with disease progression. Only patients with measurable disease and who were starting therapy were enrolled (N=177). Patients were included on an intent to treat basis.

Clinical Trial Results

{8}------------------------------------------------

Baseline CTC count was determined prior to initiation of a new line of therapy. A first follow-up CTC count was determined after the initiation of therapy. For the baseline analyses. PFS was measured from the time of the baseline blood draw to the diagnosis of progression by CT scans and/or clinical signs and symptoms, and OS was measured from the time of baseline blood draw to the time of death. For the first follow-up analyses, PFS was measured from the time of 1st follow-up blood draw (mean 4.5 + 2.4 weeks following enrollment) to diagnosis of progression or death, and OS was measured from the time of 1st follow-up blood draw to the time of death.

Progression Free Survival (PFS) Analysis

PFS Using Baseline CTC Results

All 177 patients had a baseline CTC test performed. For Kaplan-Meier analysis, patients were segmented into two groups based upon their CTC count at baseline:

Negative group (N=90), represented in green, was those patients with <5 CTC, ●
Positive group (N=87), represented in red, was those patients with >5 CTC.

Median PFS was 30.3 weeks (~7.0 months) for the Negative group and 11.7 weeks (~2.7 months) for the Positive group. The difference in PFS between the two groups is highly significant (Log-rank p=0.0001, Cox Hazards Ratio=1.9547, chisquare=15.33, p = 0.0001). These results are illustrated in Figure 1

Image /page/8/Figure/8 description: This image is a Kaplan-Meier plot showing the probability of progression-free survival over time. There are two curves, one for patients with less than 5 CTCs at baseline and another for patients with 5 or more CTCs at baseline. The median progression-free survival time for patients with less than 5 CTCs at baseline is 30.3 weeks, while for patients with 5 or more CTCs at baseline, it is 11.7 weeks. The log-rank p-value is 0.0001, the Cox hazards ratio is 1.9547, and the chi-square value is 15.33.

Figure 1. PFS of Patients with < 5 or > 5 CTC at Baseline (N=177).

PFS Using 1st Follow-up CTC Results

Of the 177 patients, 23 were not evaluable at first follow-up. Of these 23 patients, ten patients died before a follow-up blood draw could be obtained, nine patients progressed prior to the 1st follow-up blood draw, and four were lost to follow-up. Additionally, the ten patients who died had high to extremely high CTC counts at baseline (CTC counts 9. 11, 15, 24, 111, 126, 301, 1143, 4648 and 23618). For Kaplan-Meier analysis, patients were segmented into two groups based upon their CTC count at 1st follow-up:

Negative group (N=111), represented in green, was those patients with <5 CTC, .
Positive group (N=43), represented in red, was those patients with >5 CTC. .

{9}------------------------------------------------

Median PFS was 26.4 weeks (~6.1 months) for the Negative group and 5.7 weeks (~1.3 months) for the Positive group. The difference in PFS between the two groups is highly significant (Log-rank p<0.0001, Cox Hazards Ratio=2.4842, chi-square=18.83, p< 0.0001). These results are illustrated in Figure 2.

Image /page/9/Figure/11 description: This image is a graph showing the probability of progression-free survival over time. The x-axis represents time from the first follow-up in weeks, while the y-axis represents the probability of progression-free survival in percentage. There are two curves on the graph, one for patients with less than 5 CTCs at the first follow-up and another for patients with more than 5 CTCs at the first follow-up. The median PFS time for patients with less than 5 CTCs is 26.4 weeks, while for patients with more than 5 CTCs, it is 5.7 weeks.

Figure 2. PFS of Patients with < 5 or > 5 CTC at 1st Follow-Up (N=154).

Predictive Value of CTC on PFS

For Kaplan-Meier analysis, patients were segmented into three groups based on their CTC counts at baseline and 1st follow-up:

Negative group (N=81), represented in green, was those patients with <5 CTC at both ● time points.
. Patients with 5 or more CTC at baseline that decreased to below 5 CTC at 1st followup are represented in olive green (N=33),
Positive group (N=49), represented in red, was those patients with >5 CTCs at 1st follow-up.

Elapsed PFS time was calculated from the baseline blood draw. Three groups were plotted in Figure 3. The Negative group (N=81, green line) had a median PFS of 30.3 weeks (~7.0 months) and the patients represented by the olive green line (N=33) had a median PFS of 32.9 weeks (~7.6 months). The Positive group (N=49, red line) had a median PFS of 8.9 weeks (~2.1 months). The difference in the PFS of the patients in the Negative and olive green groups compared to the PFS of the patients in the Positive group is highly significant (Log-rank p<0.0006, Cox Hazards Ratio=1.6600, chisquare=22.08. p< 0.0001).

Figure 3. A Reduction in CTC Count to Below 5 at the 1st Follow-Up Time Point After the Initiation of Therapy Predicts Improved PFS (N=163)

{10}------------------------------------------------

Image /page/10/Figure/1 description: This image is a Kaplan-Meier plot showing the probability of progression-free survival over time, measured in weeks. There are three survival curves, each representing a different group of patients based on their circulating tumor cell (CTC) counts at baseline and first follow-up. The red curve represents patients with ≥ 5 CTC at 1st Follow-Up, the green curve represents a decrease in CTCs to < 5 at 1st Follow-Up, and the yellow curve represents < 5 CTC at Baseline & at 1st Follow-Up. The median progression-free survival times for each group are 8.9 weeks, 32.9 weeks, and 30.3 weeks, respectively.

Overall Survival (OS) Analysis

OS Analysis Using Baseline CTC Results

At the time of these analyses, 66 (37%) of the 177 patients had died. Seventeen (19%) of 90 patients from Negative group (<5 CTC at baseline) compared to 49 (56%) of 87 from Positive Group (>5 CTC at baseline) died. Median OS was greater than 80 weeks (>18 months) for the Negative group and 43.3 weeks (~10.1 months) for the Positive group. The OS difference between the two groups is highly significant (Log-rank p<0.0001, Cox Hazards Ratio=4.3865, chi-square=31.73, p< 0.0001). These results are illustrated in Figure 4.

Figure 4. OS of Patients with < 5 or > 5 CTC at Baseline (N=177).

Image /page/10/Figure/6 description: This image is a survival plot showing the probability of survival over time. The x-axis represents time from baseline in weeks, and the y-axis represents the probability of survival. There are two curves on the plot, one for patients with less than 5 CTCs at baseline and one for patients with greater than or equal to 5 CTCs at baseline. The median survival time for patients with less than 5 CTCs at baseline is greater than 80 weeks, while the median survival time for patients with greater than or equal to 5 CTCs at baseline is 43.3 weeks.

OS Using 1st Follow-up CTC Results

For Kaplan-Meier analysis, patients were segmented into two groups based upon their CTC count 1st follow-up:

Negative group (N=114), represented in green, was those patients with <5 CTC, ●
. Positive group (N=49), represented in red, was those patients with >5 CTC.

{11}------------------------------------------------

Fifty-six patients died of the 163 patients who were evaluable at the first follow-up. Twenty-three (20%) from the Negative group died with a median OS greater than 80 weeks (>18 months). Thirty-three (67%) from the Positive group died, with a median OS of 30.0 weeks (~7.0 months). The difference in OS between the two groups is highly significant (Log-rank p<0.0001, Cox Hazards Ratio=5.4537, chi-square=37.52, p< 0.0001). Results are summarized in Figure 5.

Image /page/11/Figure/2 description: This image is a survival plot showing the probability of survival over time. The x-axis represents time from the first follow-up in weeks, with an approximate conversion of 4.3 weeks per month. There are two survival curves: one for patients with less than 5 CTCs at the first follow-up, and another for patients with 5 or more CTCs at the first follow-up. The median survival time for patients with less than 5 CTCs is greater than 80 weeks, while the median survival time for patients with 5 or more CTCs is 30 weeks.

Image /page/11/Figure/3 description: This image is titled "Figure 5. OS of Patients with < 5 or ≥ 5 CTC at 1st Follow-Up (N=163)". The figure is about the overall survival (OS) of patients with circulating tumor cells (CTC) at the first follow-up. The sample size is 163 patients.

Predictive Value of CTC on OS

For Kaplan-Meier analysis, patients were segmented into three groups based upon their CTC counts at baseline and 1st follow-up:

Negative group (N=81), represented in green, was those patients with <5 CTC at both ● time points.
Patients with 5 or more CTC at baseline that decreased below 5 CTC at 1st follow-up ● are represented by the olive green line (N=33),
Positive group (N=49), represented in red, was those patients with >5 CTC at 1st ● follow-up,

Figure 6 illustrates that a decrease to <5 CTC after the initiation of therapy significantly impacts OS. Elapsed OS time was calculated from the baseline blood draw. The three groups were plotted in Figure 6. The Negative group (green line) had a median OS of >80 weeks (18 months) and the patients represented by the olive green line (N=33) had a median OS of 62.6 weeks (~14.6 months). The Positive group (red line) had a median OS of 35.4 weeks (~8.2 months). This difference in the OS of the patients in the Negative and olive green groups compared to the OS of the patients in the Positive group is highly significant (Log-rank p<0.0007, Cox Hazards Ratio=2.7462, chi-square=40.51, p<0.0001). These data suggests that baseline and follow-up samples support the concept that repeated evaluation of CTC levels may provide ongoing information on the patients overall survival.

{12}------------------------------------------------

Figure 6. A Reduction in CTC Count to Below 5 at the 1st Follow-Up Time Point After the Initiation of Therapy Predicts Improved OS (N=163)

Image /page/12/Figure/2 description: This image is a survival plot showing the probability of survival over time in weeks. There are three groups of patients: those with less than 5 CTC at baseline and at the first follow-up, those with a decrease in CTC to less than 5 at the first follow-up, and those with greater than or equal to 5 CTC at the first follow-up. The median survival time for each group is also shown, with the first group having a median survival time of greater than 80 weeks, the second group having a median survival time of 62.6 weeks, and the third group having a median survival time of 35.4 weeks.

1. Clinical cut-off: Please see 1f. above.

Expected values/Reference range:

Single point CTC analyses were performed on control groups of 145 healthy volunteers, 101 women with non-malignant breast disease, and 99 women with other non-malignant diseases.

Epithelial cells are not expected to be in the peripheral blood. Of the 345 total samples from healthy volunteers and women with non-malignant disease, only one subject had more than 5 CTC/7.5 mL. The results are presented in Table 3.

Category	N	Mean# CTC	SD	# Patientswith > 5 CTC	Max.*
Healthy	145	0.1	0.2	0	1
Non-malignantbreast disease	101	0.2	1.2	1	12
Non-malignantother disease	99	0.1	0.4	0	3

Table 3. Control Subiect

M. Instrument Name:

CellSpotter™ Analyzer

N. System Descriptions:

Modes of Operation: 1.
- semi-automated

2. Software:

Operating System -Windows NT 4.0

User Interface -
Primarily written in Visual Basic; Graphic User Interface (GUI)

* NCCLS Guideline C28-A23

{13}------------------------------------------------

Communication -

No communication in or out of the system;

No LIMs or LIS System interface (no standards exist for this device) Data Management -

DB2 database from IBM is used for data management

FDA has reviewed applicant's Hazard Analysis and software development processes for this line of product types: Yes X

1. Sample Identification: Sample information is manually entered by user from printout from CellPrep
1. Specimen Sampling and Handling: Sample prepared on the CellPrep™ Analyzer and manually placed into instrument
1. Assay Types: Cytochemistry image analysis
Reaction Types: 6. Fluorescence microscopy
1. Calibration: None
1. Quality Control: The CellSearch™ Control Cell Kit is provided by the manufacturer to control the system.

O. Other Supportive Instrument Performance Characteristics Data Not Covered In The "L. Performance Characteristics" Section Of The SE Determination Decision Summary.

Recovery

Blood from a single healthy donor was pooled and five 7.5 mL samples were spiked with cells of a breast cancer cell line (SKBr-3). Serial dilutions were performed resulting in twenty-five samples with expected cell counts ranging from 4 to 1142 cells per 7.5 mL. These samples were processed on CellPrep™ instruments, analyzed over five consecutive days, and the results of the observed cell counts were plotted against the results of the expected cell counts. The mean recovery of cells ranged from 85 to 123%. However, because of the low cell numbers and the inherent variation in spiking low numbers of cells, recovery at the low end of the range was imprecise. A difference of only 2 cells in the lowest dilution represents a 50% difference. To determine the overall, or least squares fit, for the comparison of the observed and expected cell counts across all the data, linear regression analysis was performed. The regression equation for these 25 samples was Y=0.85x +5.64, R =0.9973. Analysis of the regression data showed that the 95% confidence interval of the intercept of 5.64 was -4.421 to 15.69, which overlaps zero and therefore is not statistically different from zero. This study suggests that the average recovery of spiked cells using the CellPrep™ instrument and the CellSearch™ Kit is approximately 85%.

{14}------------------------------------------------

P. Proposed Labeling:

The labeling is sufficient and it satisfies the requirements of 21 CFR Part 809.10.

Q. Conclusion:

The petition for Evaluation of Automatic Class III Designation for this device is accepted. The device is classified as Class II under regulation 21 CFR 866.6020 with special controls. The special control guidance document "Immunomagnetic Circulating Cancer Cell Selection and Enumeration System" is available at WWW.fda.gov/cdrh/oivd/guidance/1531.pdf

Regulation Number and Section

§ 866.6020 Immunomagnetic circulating cancer cell selection and enumeration system.

(a)
Identification. An immunomagnetic circulating cancer cell selection and enumeration system is a device that consists of biological probes, fluorochromes, and other reagents; preservation and preparation devices; and a semiautomated analytical instrument to select and count circulating cancer cells in a prepared sample of whole blood. This device is intended for adjunctive use in monitoring or predicting cancer disease progression, response to therapy, and for the detection of recurrent disease.(b)
Classification. Class II (special controls). The special control for this device is FDA's guidance document entitled “Class II Special Controls Guidance Document: Immunomagnetic Circulating Cancer Cell Selection and Enumeration System.” See § 866.1(e) for availability of this guidance document.