Shield

P230009 · Guardant Health, Inc. · PHP · Jul 26, 2024 · Medical Genetics

Device Facts

• Record ID: P230009
• Device Name: Shield
• Applicant: Guardant Health, Inc.
• Product Code: PHP · Medical Genetics
• Decision Date: Jul 26, 2024
• Decision: APPR
• Device Class: Class 3
• Attributes: AI/ML, Real-World Evidence

Real-World Evidence

Submission	Device	Sponsor	RWD Sources	RWE Use Summary	Key Tags
P230009 · Jul 26, 2024	Shield	Guardant Health, Inc.	ECLIPSE study interim analysis dataset (co-morbidity analysis); Post-approval study (longitudinal real-world performance data)	Real-world data from the ECLIPSE interim analysis was used to assess the impact of non-cancer co-morbidities on test specificity. A future post-approval study will utilize RWE to evaluate the longitudinal performance of the device in an average-risk population.	Post-approval study; Co-morbidity analysis; Longitudinal performance

Clinical Evidence

Study Design	Population	Comparator	Key Endpoints
Shield Post-Approval Study (PAS); Prospective, longitudinal study supplemented with Real World Evidence (RWE); Follow-up/Duration: 3 years of longitudinal follow-up; Study Period: 5 years total	Average risk population, age 45-84; Sample Size: At least 1000 evaluable subjects	Not applicable for this study	Sensitivity for CRC, sensitivity for AA, specificity for non-advanced neoplasia, PPV, and NPV at T3
ECLIPSE interim analysis co-morbidity evaluation; Observational analysis of interim clinical study data	2440 subjects from the interim analysis dataset in the ECLIPSE study; Sample Size: 2440	Not applicable for this study	Positivity rates for observed co-morbidities

AI Performance

Output	Algorithm	Acceptance	Observed	Dev DS	Dev Readers	Test DS	Test Readers
Colorectal Cancer Detection	Proprietary bioinformatics algorithms using genomic and epigenomic features	Sensitivity for CRC > 65% (lower bound of 95% CI)	Sensitivity 83.1% (95% CI 72.2%, 90.3%)	1,470 CRC cases and 2,340 cancer-free controls (training); 1,050 CRC cases and 710 colonoscopy-confirmed controls (testing)	—	ECLIPSE study: 7,861 subjects with valid colonoscopy and Shield results	—
Advanced Adenoma Detection	Proprietary bioinformatics algorithms using genomic and epigenomic features	—	Sensitivity 13.2% (95% CI 11.3%, 15.3%)	1,470 CRC cases and 2,340 cancer-free controls (training); 1,050 CRC cases and 710 colonoscopy-confirmed controls (testing)	—	ECLIPSE study: 7,861 subjects with valid colonoscopy and Shield results	—

Indications for Use

The Shield test is a qualitative, in vitro diagnostic test intended to detect colorectal cancer derived alterations in cell-free DNA from blood collected in the Guardant Shield Blood Collection Kit. Shield is intended for colorectal cancer screening in individuals at average risk of the disease, age 45 years or older. Patients with a positive result should be followed by colonoscopy. Shield is not a replacement for diagnostic colonoscopy or for surveillance colonoscopy in high-risk individuals. This test is performed at Guardant Health, Inc.

Device Story

Shield is a blood-based, in vitro diagnostic test for colorectal cancer (CRC) screening. It analyzes cell-free DNA (cfDNA) from patient blood samples collected in specialized tubes. The process involves plasma isolation, cfDNA extraction, methylation partitioning, library preparation, and next-generation sequencing (NGS) on an Illumina NovaSeq system. Proprietary bioinformatics algorithms analyze genomic (somatic mutations) and epigenomic (methylation and fragmentation patterns) signals to generate an Integrated Score and an MR Score. These scores are compared against predefined cutoffs to produce a qualitative 'Positive' or 'Negative' result. The test is performed exclusively at the Guardant Health clinical laboratory. Physicians use the result to determine if a patient requires a follow-up diagnostic colonoscopy. By providing a non-invasive screening option, the device aims to increase screening compliance among average-risk individuals who might otherwise forgo recommended procedures.

Clinical Evidence

Pivotal prospective, multi-site, observational study (ECLIPSE) of 7,861 average-risk subjects (45-84 years). Primary endpoints: CRC sensitivity and advanced neoplasia (AN) specificity. Results: CRC sensitivity 83.1% (95% CI: 72.2%-90.3%); AN specificity 89.6% (95% CI: 88.8%-90.3%). Advanced adenoma (AA) sensitivity was 13.2% (95% CI: 11.3%-15.3%). Stage I CRC sensitivity was 54.5%. No unanticipated adverse device effects reported.

Technological Characteristics

Next-generation sequencing (NGS) based assay. Analyzes cfDNA for somatic mutations, methylation patterns, and fragmentation patterns. Requires cfDNA extraction, methylation partitioning, library preparation, and Illumina NovaSeq sequencing. Components include Guardant cfDNA blood collection tubes, Hamilton Microlab STAR, Tecan SPARK reader, and QIAGEN QIASymphony SP. Software-based analysis using proprietary bioinformatics algorithms. Standardized laboratory workflow with exogenous DNA controls and no-template controls.

Indications for Use

Indicated for colorectal cancer screening in average-risk individuals aged 45 years or older. Contraindicated for patients with a personal history of colorectal cancer (CRC), adenomas, or related cancers; family history of CRC in a first-degree relative; positive result on another CRC screening method within the last 6-36 months (depending on method); or personal history of high-risk conditions including IBD (CUC/Crohn's), FAP, or hereditary cancer syndromes (e.g., Lynch, Peutz-Jeghers, MUTYH Polyposis, Gardner’s, Turcot’s, Cowden’s, Juvenile Polyposis, Cronkhite-Canada, Neurofibromatosis, Familial Hyperplastic Polyposis).

Regulatory Classification

Identification

The colorectal neoplasia DNA methylation and hemoglobin detection system is a qualitative, non-invasive device used to screen for the detection of methylation of specific DNA markers associated with colorectal cancer and for the presence of occult hemoglobin in human specimens, including stool, blood, urine, and buccal swabs. It is intended as an aid in the detection of colorectal cancer or advanced adenomas in conjunction with diagnostic colonoscopy. It is indicated for use in adults 50 years or older, who are typical average-risk candidates for colorectal cancer (CRC) screening.

Reference Devices

• Cologuard
• Epi proColon

Submission Summary (Full Text)

{0}

# SUMMARY OF SAFETY AND EFFECTIVENESS DATA (SSED)

## I. GENERAL INFORMATION

Device Generic Name: Blood-based Qualitative Colorectal Cancer Screening Test

Device Trade Name: Shield

Device Procode: PHP

Applicant’s Name and Address: Guardant Health, Inc
505 Penobscot Dr.
Redwood City, CA 94063

Date(s) of Panel Recommendation: May 23, 2024

Premarket Approval Application (PMA) Number: 230009

Date of FDA Notice of Approval: July 26, 2024

## II. INDICATIONS FOR USE

**Intended Use:**

The Shield test is a qualitative, in vitro diagnostic test intended to detect colorectal cancer derived alterations in cell-free DNA from blood collected in the Guardant Shield Blood Collection Kit.

Shield is intended for colorectal cancer screening in individuals at average risk of the disease, age 45 years or older. Patients with a positive result should be followed by colonoscopy. Shield is not a replacement for diagnostic colonoscopy or for surveillance colonoscopy in high-risk individuals.

This test is performed at Guardant Health, Inc.

## III. PRECAUTION

Based on data from clinical studies, Shield has limited detection (55%-65%) of Stage I colorectal cancer and does not detect 87% of precancerous lesions. One out of 10 patients with a negative Shield result may have a precancer that would have been detected by a screening colonoscopy. Shield demonstrated high detection of Stages II, III, and IV colorectal cancer.

{1}

IV. CONTRAINDICATIONS

The Shield test is NOT indicated for use for patients that have the following:

- Personal history of colorectal cancer (CRC), adenomas, or other related cancers
- Family history of CRC, defined as having one or more first-degree relative (parent, sibling, or child) diagnosed with CRC at any age
- Positive result on another colorectal cancer screening method within the last six months, or:
- 12 months for fecal occult blood test (FOBT) or fecal immunochemical test (FIT)
- 36 months for FIT-DNA test
- Personal history of any of the following high-risk conditions for colorectal cancer:
- Inflammatory Bowel Disease (IBD), including chronic ulcerative colitis (CUC) and Crohn’s disease
- Familial adenomatous polyposis (FAP)
- Other hereditary cancer syndromes including but not limited to:
- Hereditary non-polyposis colorectal cancer syndrome (HNPCC) or “Lynch Syndrome”, Peutz- Jeghers Syndrome, MUTYH Polyposis (MAP), Gardner’s Syndrome, Turcot’s (or Crail’s) Syndrome, Cowden’s Syndrome, Juvenile Polyposis, Cronkhite-Canada Syndrome, Neurofibromatosis and Familial Hyperplastic Polyposis

V. LIMITATIONS

- Providers should discuss the most appropriate screening test to use with patients depending on their medical history and individual circumstances. The Shield test is not intended as a screening test for individuals who are at high risk for CRC.
- Shield has limited ability to prevent the development of colorectal cancer from advanced precancerous lesions and lower detection rates for Stage I CRC, given the current data available.
- Shield has lower performance of stage I CRC [54.5% (12/22); 95%CI (34.7%, 73.1%)]. The majority [6/10] of missed Stage I cancers were less than 10mm. Shield did not detect CRC lesions smaller than 10mm [0% (0/6); 95% CI (0.0% - 39.0%)].
- Shield may fail to detect as many as 88.7% of patients with advanced precancerous lesions which can later become neoplastic because of its limited ability for the detection of advanced adenomas [(13.2% (147/1116); 95%CI (11.3, 15.3)].
- Shield has a false negative rate of 17% for colorectal cancer, meaning 17 of 100

{2}

people who have colorectal cancer will incorrectly have a Shield negative result.

- Shield has a false positive rate of 10%, meaning one of 10 people who do not have Advanced Neoplasia (colorectal cancer or advanced adenoma) will have a false positive test result.

- CRC screening guideline recommendations vary for persons over the age of 75. The decision to screen patients over the age of 75 should be made on an individualized basis in consultation with a healthcare provider.

- A positive Shield test result suggests patients may have colorectal cancer or advanced adenoma. Patients with a positive result should be followed by colonoscopy.

- A negative Shield test result does not guarantee absence of colorectal cancer or advanced adenoma. Patients with a negative result should continue participating in colorectal cancer screening programs, at the appropriate guideline recommended intervals.

- One out of 10 patients testing negative will be falsely reassured that they are negative for advanced adenoma, given the negative predictive value for advanced adenoma of 90%.

- One out of 1000 patients testing negative will be falsely reassured that they are negative for CRC, given the negative predictive value of 99.9%.

- A false positive result may occur when the Shield test generates a positive result while a colonoscopy will not find colorectal cancer or advanced adenoma. A false negative result may occur when the Shield test does not detect a colorectal tumor signal while a colonoscopy identifies a colorectal cancer.

- The performance of Shield has been established in a prospectively designed, cross-sectional study. The benefits and risks of programmatic colorectal screening (i.e., repeated testing over an established period of time) with Shield has not been studied.

- Non-inferiority or superiority of Shield sensitivity as compared to other recommended screening methods for colorectal cancer or advanced adenoma has not been established.

- Cross-reactivity was observed in analytical studies using samples from subjects with non-CRC cancers, including gastric, pancreatic, liver, bladder, breast, lung, prostate, ovarian, melanoma and kidney cancers.

- Consult the Guardant Shield Blood Collection Kit (BCK) instructions for use (LBL-000324), for precautions and limitations specific to the collection and shipping of blood samples.

VI. DEVICE DESCRIPTION

Shield is a next generation sequencing based qualitative test to detect genomic (somatic mutations) and epigenomic alterations (methylation and fragmentation patterns) associated

{3}

with colorectal cancer from whole blood samples collected from individuals at average risk for CRC. These samples are shipped to Guardant Health, where cfDNA is extracted from the plasma component of whole blood and prepared for analysis through the DNA sequencing workflow. The resulting cfDNA data are then analyzed using proprietary bioinformatics algorithms trained to detect the presence of colorectal cancer associated signals. Following analysis, a test report is generated for the sample. This test yields a final qualitative test result of "Positive" or "Negative". Patients with a positive result may have colorectal cancer or advanced adenomas and should be followed by colonoscopy.

Reagents, materials, and equipment needed to perform the test are used exclusively in the Guardant Health Clinical Laboratory. The reagents distributed outside of Guardant Health are contained within the blood collection tubes (BCTs) that are part of the Shield Blood Collection Kit (BCK).

The following components are required for the use of Shield:

**Shield Blood Collection Kit:**
- Guardant cfDNA blood collection tubes
- BCT label with 2D Barcode
- Biohazard specimen bag
- Foam Tray
- Absorbent sheet
- BCK barcode sheet
- BCK Instructions for Use

The Guardant Shield Blood Collection Kit (BCK) comprises all components used in the collection, stabilization, packaging, and transportation of whole blood samples and is the only test component intended for external distribution (i.e., the Shield test itself is performed in Guardant's clinical laboratory). The kit contains four Guardant-labeled blood collection tubes and packaging material with instructions for kit storage, sample collection, and shipping after samples are collected.

**Shield Assay:**
The following reagents are required to perform the Shield assay and qualified by Guardant Health, Inc. under the Guardant Health Quality System:
- cfDNA extraction beads and buffers
- cfDNA quantitation dye
- DNA purification magnetic beads
- Methylated DNA detection reagents
- Library preparation enzyme mix
- Library enrichment probes
- DNA control - exogenous
- Next generation sequencing kit

{4}

Instrumentation:

The following instruments are required to perform the Shield test and qualified by Guardant Health, Inc. under the Guardant Health Quality System:

- Hamilton Microlab STAR
- Tecan SPARK Microplate Multimode Reader
- QIAGEN QIASymphony SP Instrument
- Illumina NovaSeq Sequencing System

Software:

The Shield Test includes software used for sample processing, data analysis, and report generation. The Shield software is comprised of two software subsystems:

- Guardant Assay Platform: The Guardant Assay Platform (GAP) is a software package that supports the execution of assay specific workflows. It provides common infrastructure and functionalities that are assay agnostic. This infrastructure facilitates the execution of the assay by providing the user a graphic interface for managing samples, executing the test, viewing status, receiving notifications, and viewing results.
- Guardant Screening Software: The Guardant Screening Software (GSS) is a software package that implements the Shield screening assay specific workflows and analytics. The GSS runs and is supported by the infrastructure GAP provides. GSS contains all the data analytics, algorithms, assay logic, and business rules for the Shield test.

Additional information about the Shield test is listed in the instructions for use.

Test Procedure:

The Shield test begins with the collection of whole blood in up to 4 Guardant cfDNA blood collection tubes that are provided as part of the Guardant BCK. The patient specimen is then shipped to Guardant Health. Plasma is isolated from whole blood in each tube and then pooled. The cfDNA is extracted from a minimum of 2 mL of plasma for processing through the DNA sequencing workflow. In the cfDNA workflow, cfDNA is prepared in a manner that allows for simultaneous analysis of genomic and epigenomic changes. Epigenomic modifications may be detected as either altered methylation patterns in cfDNA, or as changes in the cfDNA fragment positions. A library is prepared and enriched for informative genomic regions, followed by sequencing of the enriched library. The resulting cfDNA data are analyzed using proprietary bioinformatics algorithms designed to detect the presence of colorectal neoplasm-associated signals.

Whole Blood Collection and Shipping

Peripheral whole blood is collected in Guardant cfDNA BCTs provided with the kit and is then shipped to Guardant Health at ambient temperature. Prior to blood draw, the Guardant BCK may be stored in conditions consistent with its labeling until the expiration date printed on the BCK label. Complete instructions for sample collection and shipping can be found in the BCK Instructions for Use (LBL-000324).

{5}

# Plasma Isolation and cfDNA Extraction

Upon receipt, whole blood specimens are processed in the Guardant Health Clinical Laboratory within 7 days of blood collection. Plasma is isolated from the tubes of whole blood via centrifugation. Plasma is divided into primary and retain aliquots with a minimum volume of 2 mL and a maximum volume of 8 mL in each aliquot. cfDNA (cell-free DNA) is extracted from plasma aliquot using the QIAGEN QIAsymphony SP Instrument qualified by Guardant.

# Methylation Partitioning

After extraction, cfDNA is separated into methylated and unmethylated partitions based on the overall methylation state of each molecule. The cfDNA is partitioned based on the differential binding affinity of the methylated nucleic acid molecules to a binding agent (i.e., a binding agent that binds to methylated nucleotides).

# Library Preparation and Enrichment

The DNA in each partition is tagged with a distinct set of barcodes, which uniquely identifies the partition associated with every molecule. All partitions are then PCR amplified and enriched via hybridization to oligonucleotides representing genomic regions of interest targeting approximately 1 Mb of the human genome.

# Pooling and Sequencing

Enriched partitions are pooled and tagged with an index uniquely identifying each sample prior to pooling multiple enriched samples into sequencing pools. Sequencing pools are sequenced on the Illumina NovaSeq Sequencing System qualified by Guardant.

# Sequencing Data Analysis

Following sequencing, reads are demultiplexed. The methylation partitions associated with every molecule are identified by unique partition labels added during library preparation to enable differentiation of methylated and unmethylated partitions in the analysis step. Only unique molecules which align to genomic regions within the enrichment panels are leveraged in the downstream algorithms.

# Calling Algorithm and Result Reporting

The classification of a clinical sample relies upon the multiple biomarkers derived from cfDNA and known to be distinct between normal and cancer-derived tissues. The Shield test interrogates thousands of individual features that characterize three types of cfDNA signals or patterns: epigenetic changes resulting in the aberrant methylation state, epigenetic changes resulting in the aberrant cfDNA molecule fragmentation patterns, and genomic changes resulting in somatic mutations.

These features are used to generate four individual classification scores: Fragmentomics Mixture Score, Methylation LR Score, Methylation MR score and Somatic mutation detection. Individual scores are combined into the "Integrated Score", and the MR Score is also evaluated independently. The cfDNA MR Score and cfDNA Integrated Scores, are compared to predefined cutoffs to generate positive vs negative results for each cfDNA MR Call and cfDNA Integrated Call. If either of these calls is positive, then the result is positive. A negative call only occurs when both the cfDNA Integrated Call and the cfDNA MR Call are negative.

{6}

These results classify samples as either "Positive" or "Negative".

## Quality Control Measures

Shield includes an exogenous DNA control which is designed to contain known features which would result in a positive classification as well as known negative control features that should not be detected on the Shield test. Additionally, a no template negative control (NTC) is run in parallel with patient samples.

The controls are treated as individual samples with processing starting from methylation partition through cfDNA sequencing workflow. Positive classification and absence of negative control regions for the exogenous DNA control and absence of molecules for the NTC are both required quality control measures for reporting valid patient test results.

In addition to assessing the control performance within a batch, the test utilizes multiple per-sample in-process and post-sequencing analytical metrics from clinical sample data that are specific and informative to sample performance.

## VII. ALTERNATIVE PRACTICES AND PROCEDURES

Recommended screening for colorectal cancer includes both invasive and non-invasive options. Invasive options include colonoscopy, flexible sigmoidoscopy, flexible sigmoidoscopy with FIT, and CT colonography.

Non-invasive screening options include stool DNA-based colorectal cancer screening test, guaiac-based fecal occult blood testing (gFOBT), fecal immunochemical test (FIT) and blood-based plasma DNA testing.

Colonoscopy is considered to be the most accurate screening tool available, which can involve the removal of precancerous lesions to prevent cancer.

Patients who have a positive or abnormal test by an invasive or non-invasive screening method, except for colonoscopy, warrant further investigation through conventional colonoscopy. A patient should discuss these alternatives with their Healthcare Provider to select the method that best meets the patient's needs, expectations and lifestyle.

## VIII. MARKETING HISTORY

The Shield test has not been marketed as an IVD in the United States or any foreign country.

## IX. POTENTIAL ADVERSE EFFECTS OF THE DEVICE ON HEALTH

Due to the nature of the noninvasive blood collection process, potential adverse events (AEs) caused by or related to testing with Shield are unlikely. The primary risk associated with the Shield test is a false assay result (i.e., a false positive or a false

{7}

negative result). All positive test results should lead to a colonoscopy.

In the instance of a false negative result on Shield, there is a possibility that a case of colorectal cancer or advanced adenoma could go undetected.

# X. SUMMARY OF NONCLINICAL STUDIES

Nonclinical studies were conducted at Guardant Health to evaluate the analytical performance of Shield. The studies are described below.

# A. Algorithm Development and Cut-Off Determination

The Shield classification models, including assessment of methylation levels, cfDNA fragmentation patterns and cfDNA genomic alterations, were trained on 1,470 known CRC cases representative of all cancer stages and 2,340 cancer-free controls. A second set of samples including 1,050 known CRC cases across all cancer stages and 710 colonoscopy-confirmed controls without AN was used to test the device.

# B. Analytical Sensitivity

# Limit of Detection (LoD)

The LoD for Shield was expressed in terms of tumor fraction (estimated by somatic mutation frequency, max-MAF), which is the analyte detected by the assay. The LoD was first established as $0.05\%$ max-MAF using cfDNA from 3 clinical CRC samples diluted with pooled normal samples to create independent pools targeting 6 tumor fractions (MAF) levels from $0.01\%$ to $0.5\%$ . All samples were targeted at an input of 600 Non-Singleton Coverage (NSC) which is a challenging input level that minimally exceeds input QC thresholds based on measured NSC variability in cfDNA. Overall LoD was established based on the Shield device positive call rate, and additional analysis was performed to assess positive call rate of the two underlying callers (Integrated and MR positive call rate) at each of the dilution levels. The integrated positive call rate was $100\%$ for dilution levels of $0.10\%$ and above, while the MR positive call rate was $100\%$ for dilution levels of $0.05\%$ and above, resulting in $100\%$ Shield overall positive call rate for dilution levels of $0.05\%$ and above (Table 1).

Table 1. Positive Call Rate of Shield overall call, Integrated call and MR call in the LoD establishing study

| Target Max-MAF (%) | Shield Positive Call Rate (%) | Integrated Positive Call Rate (%) | MR Positive Call Rate (%) |
| --- | --- | --- | --- |
| 0.01 | 49% (53/108) | 4% (4/108) | 49% (53/108) |
| 0.02 | 86% (95/111) | 19% (21/111) | 86% (95/111) |
| 0.05 | 100% (117/117) | 62% (72/117) | 100% (117/117) |
| 0.10 | 100% (119/119) | 95% (113/119) | 100% (119/119) |
| 0.30 | 100% (119/119) | 100% (119/119) | 100% (119/119) |

{8}

| Target Max-MAF (%) | Shield Positive Call Rate (%) | Integrated Positive Call Rate (%) | MR Positive Call Rate (%) |
| --- | --- | --- | --- |
| 0.50 | 100% (115/115) | 100% (115/115) | 100% (115/115) |

The 0.05% max-MAF LoD was confirmed by testing 3 CRC positive sample pools at 3 dilutions levels (0.05%, 0.07%, 1% MAF). A 100% overall positive call rate was observed for dilution levels of 0.05% and above, verifying the detection capability at 0.05% max-MAF.

A supplemental LoD confirmation study was performed using 2 clinical stage I CRC, 1 stage III CRC, a clinical stage unknown CRC and pooled clinical AA samples diluted with pooled self-declared normal cfDNA to the target dilution levels from 0.01% to 0.07%. 100% positive call rate was observed for all samples tested at targeted tumor fraction of 0.05% and above.

## Limit of Blank (LoB)

LoB was evaluated with two healthy donor sample pools in 10 different batches using a unique combination of two critical reagent lots. Each sample was tested in greater than 60 replicates each. A total of 198 sample replicates were assessed. The false positive rate (FPR) for the two sample pools was 0%, and 3%, indicating that the FPR is below 5%. Considering both sample pools, the FPR is 1.52%, with a Clopper Pearson Confidence Interval of [0.39%, 4.72%].

## C. Precision

Precision was evaluated with a total of 24 clinical samples, including 20 individual clinical samples and 4 clinical sample pools created by spiking CRC patient plasma into plasma from healthy donors. These 24 samples included 3 positive, 5 low positive, 2 high negative, 9 negative, and 5 borderline (close to the test decision boundary) samples. Precision across runs (batches), operators, instruments, reagent lots, and test days was assessed, in addition to concordance between sequencing instruments and batch pooling tolerance for variation in sequencing read depth. Six batches of samples were tested starting on 6 different days using various combinations of reagent lots (3 groups), instrument lines (2 groups), and operators (3 groups) and each sample was tested in 4-36 replicates. The primary analysis was based on the primary sequencing of replicates from all 6 batches. Following primary sequencing, the same libraries were sequenced in the secondary sequencer to assess instrument to instrument precision. Tolerance to pooling of multiple batches within a sequencing flow cell was also assessed. Each of the 6 batches was sequenced on an individual flow cell to compare to pooling with three batches per flow cell. The PPA and NPA for each sample for each level (e.g., primary analysis, instrument to instrument, and pooling tolerance) is shown in Table 2 below.

Of the 24 samples, 16 samples had overall agreement for final Shield call over 92% based on data from the primary sequencing runs. Eight (8) samples had agreement ranging from 58.33% to 91.67% with integrated score and/or MR score close to their predefined cutoffs.

{9}

Table 2. Positive call rates for overall Shield call

| Sample ID (A) | Sample Disease Status (B) | Sample Category (C) | Level (D) | Number of Replicates (E) | % Called Positive Shield Test Result (F) | % Called Negative Shield Test Result (G) |
| --- | --- | --- | --- | --- | --- | --- |
| CRC1 | CRC | Positive | Primary analysis | 23 | 100.0 | 0.0 |
| | | | Instrument to instrument | 23 | 100.0 | 0.0 |
| | | | Pooling tolerance | 23 | 100.0 | 0.0 |
| CRC2 | CRC | Positive | Primary analysis | 18 | 100.0 | 0.0 |
| | | | Instrument to instrument | 18 | 100.0 | 0.0 |
| | | | Pooling tolerance | 18 | 100.0 | 0.0 |
| CRC3 | pool_CRC_and_self_declared_cancer_free | Positive | Primary analysis | 23 | 100.0 | 0.0 |
| | | | Instrument to instrument | 23 | 100.0 | 0.0 |
| | | | Pooling tolerance | 23 | 100.0 | 0.0 |
| CRC4 | CRC | Low Positive | Primary analysis | 12 | 91.67 | 8.33 |
| | | | Instrument to instrument | 12 | 91.67 | 8.33 |
| | | | Pooling tolerance | 12 | 91.67 | 8.33 |
| CRC5 | CRC | Low Positive | Primary analysis | 12 | 100.0 | 0.0 |
| | | | Instrument to instrument | 12 | 100.0 | 0.0 |
| | | | Pooling tolerance | 12 | 100.0 | 0.0 |
| CRC6 | CRC | Low Positive | Primary analysis | 9 | 100.0 | 0.0 |
| | | | Instrument to instrument | 10 | 100.0 | 0.0 |
| | | | Pooling tolerance | 10 | 100.0 | 0.0 |
| CRC7 | pool_CRC_and_self_declared_cancer_free | Low Positive | Primary analysis | 28 | 100.0 | 0.0 |
| | | | Instrument to instrument | 28 | 100.0 | 0.0 |
| | | | Pooling tolerance | 28 | 100.0 | 0.0 |
| CRC8 | pool_CRC_and_self_declared_cancer_free | Low Positive | Primary analysis | 29 | 100.0 | 0.0 |
| | | | Instrument to instrument | 29 | 100.0 | 0.0 |
| | | | Pooling tolerance | 29 | 100.0 | 0.0 |
| CRC9 | CRC | Borderline | Primary analysis | 4 | 100.0 | 0.0 |
| | | | Instrument to instrument | 4 | 100.0 | 0.0 |
| | | | Pooling tolerance | 4 | 100.0 | 0.0 |
| CRC10 | CRC | Borderline | Primary analysis | 12 | 58.33 | 41.67 |
| | | | Instrument to instrument | 12 | 50.00 | 50.00 |
| | | | Pooling tolerance | 12 | 66.67 | 33.33 |

{10}

| Sample ID (A) | Sample Disease Status (B) | Sample Category (C) | Level (D) | Number of Replicates (E) | % Called Positive Shield Test Result (F) | % Called Negative Shield Test Result (G) |
| --- | --- | --- | --- | --- | --- | --- |
| CRC11 | CRC | Borderline | Primary analysis | 17 | 52.94 | 47.06 |
| | | | Instrument to instrument | 18 | 55.56 | 44.44 |
| | | | Pooling tolerance | 17 | 76.47 | 23.53 |
| Healthy 1 | self_declared_cancer_free | Borderline | Primary analysis | 17 | 11.76 | 88.24 |
| | | | Instrument to instrument | 17 | 11.76 | 88.24 |
| | | | Pooling tolerance | 17 | 23.53 | 76.47 |
| CRC12 | pool_CRC_and_self_declared_cancer_free | Borderline | Primary analysis | 12 | 91.67 | 8.33 |
| | | | Instrument to instrument | 12 | 91.67 | 8.33 |
| | | | Pooling tolerance | 12 | 91.67 | 8.33 |
| Healthy 2 | self_declared_cancer_free | High Negative | Primary analysis | 30 | 6.67 | 93.33 |
| | | | Instrument to instrument | 30 | 3.33 | 96.67 |
| | | | Pooling tolerance | 30 | 3.33 | 96.67 |
| Healthy 3 | self_declared_cancer_free | High Negative | Primary analysis | 30 | 16.67 | 83.33 |
| | | | Instrument to instrument | 30 | 13.33 | 86.67 |
| | | | Pooling tolerance | 30 | 13.33 | 86.67 |
| Healthy 4 | self_declared_cancer_free | Negative | Primary analysis | 30 | 3.33 | 96.67 |
| | | | Instrument to instrument | 30 | 0.0 | 100.0 |
| | | | Pooling tolerance | 30 | 3.33 | 96.67 |
| Healthy 5 | self_declared_cancer_free | Negative | Primary analysis | 30 | 6.67 | 93.33 |
| | | | Instrument to instrument | 30 | 0.0 | 100.0 |
| | | | Pooling tolerance | 30 | 6.67 | 93.33 |
| Healthy 6 | self_declared_cancer_free | Negative | Primary analysis | 36 | 2.78 | 97.22 |
| | | | Instrument to instrument | 36 | 0.0 | 100.0 |
| | | | Pooling tolerance | 36 | 13.89 | 86.11 |
| Healthy 7 | self_declared_cancer_free | Negative | Primary analysis | 29 | 13.79 | 86.21 |
| | | | Instrument to instrument | 29 | 10.34 | 89.66 |
| | | | Pooling tolerance | 29 | 6.90 | 93.10 |
| Healthy 8 | self_declared_cancer_free | Negative | Primary analysis | 29 | 17.24 | 82.76 |
| | | | Instrument to instrument | 29 | 17.24 | 82.76 |
| | | | Pooling tolerance | 29 | 13.79 | 86.21 |
| Healthy 9 | self_declared_cancer_free | Negative | Primary analysis | 34 | 0.0 | 100.0 |
| | | | Instrument to instrument | 34 | 0.0 | 100.0 |
| Healthy 10 | self_declared_cancer_free | Negative | Primary analysis | 34 | 0.0 | 100.0 |
| | | | Instrument to instrument | 34 | 0.0 | 100.0 |
| | | | Pooling tolerance | 34 | 0.0 | 100.0 |

{11}

Additional analysis was performed based on data from the primary sequencing runs to assess precision with respect to the underlying Integrated and MR Scores and the corresponding predefined cutoffs and calls. With regards to their scores, samples are classified as positive for integrated call ( $> -1.73$ ) or negative ( $<= -1.73$ ); samples are classified as positive for MR call ( $> -11.06$ ) or negative ( $<= -11.06$ ). Of the 24 samples, 16 samples had agreement for both integrated call and MR call ranging from $89.66\%$ to $100\%$ . For 5 samples with integrated scores close to the -1.73 cut-off, ranging from -2.0488 to -1.3562, agreement for reproducibility ranged from $52.94\%$ to $58.33\%$ . For 4 samples with MR scores close to the -11.06 cut-off, ranging from -11.5505 to -10.8642, agreement for reproducibility ranged from $50.00\%$ to $88.24\%$ . The observed integrated score and/or MR scores of these 7 samples were close to their respected cut-off, thus explaining the disagreement of Shield overall results near the threshold. These results are presented in Table 3 below.

Table 3. Positive call rates for Integrated call and MR call

| Sample ID | Sample Disease Status | Sample Category | Level | Number of Replicates | % Called Positive Int Caller | % Called Negative Int Caller | Mean Int Score | % Called Positive MR Caller | % Called Negative MR Caller | Mean MR Score |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| CRC 1 | CRC | Positive | Primary analysis | 23 | 100.0 | 0.0 | 3.2229 | 100.0 | 0.0 | -7.7724 |
| | | | Instrument to instrument | 23 | 100.0 | 0.0 | 3.1107 | 100.0 | 0.0 | -7.8003 |
| | | | Pooling tolerance | 23 | 100.0 | 0.0 | 3.0405 | 100.0 | 0.0 | -7.7967 |
| CRC 2 | CRC | Positive | Primary analysis | 18 | 100.0 | 0.0 | 1.7009 | 100.0 | 0.0 | -8.1913 |
| | | | Instrument to instrument | 18 | 100.0 | 0.0 | 1.7691 | 100.0 | 0.0 | -8.1899 |
| | | | Pooling tolerance | 18 | 94.44 | 5.56 | 1.7554 | 100.0 | 0.0 | -8.1412 |

{12}

| Sample ID | Sample Disease Status | Sample Category | Level | Number of Replicates | % Called Positive Int Caller | % Called Negative Int Caller | Mean Int Score | % Called Positive MR Caller | % Called Negative MR Caller | Mean MR Score |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| CRC 3 | pool_CRC_and_self_declared_cancer_free | Positive | Primary analysis | 23 | 91.3 | 8.7 | 0.387 | 100.0 | 0.0 | -8.7927 |
| | | | Instrument to instrument | 23 | 95.65 | 4.35 | 0.3208 | 100.0 | 0.0 | -8.7912 |
| | | | Pooling tolerance | 23 | 95.65 | 4.35 | 0.3348 | 100.0 | 0.0 | -8.7934 |
| CRC 4 | CRC | Low Positive | Primary analysis | 12 | 91.67 | 8.33 | 0.9091 | 50.0 | 50.0 | -10.8642 |
| | | | Instrument to instrument | 12 | 91.67 | 8.33 | 0.9321 | 66.67 | 33.33 | -11.0013 |
| | | | Pooling tolerance | 12 | 91.67 | 8.33 | 0.9798 | 58.33 | 41.67 | -10.884 |
| CRC 5 | CRC | Low Positive | Primary analysis | 12 | 100.0 | 0.0 | 0.5933 | 100.0 | 0.0 | -10.0333 |
| | | | Instrument to instrument | 12 | 91.67 | 8.33 | 0.4539 | 100.0 | 0.0 | -10.058 |
| | | | Pooling tolerance | 12 | 91.67 | 8.33 | 0.4207 | 100.0 | 0.0 | -10.0666 |
| CRC 6 | CRC | Low Positive | Primary analysis | 9 | 100.0 | 0.0 | 0.9889 | 100.0 | 0.0 | -9.9319 |
| | | | Instrument to instrument | 10 | 100.0 | 0.0 | 1.1593 | 100.0 | 0.0 | -9.9039 |
| | | | Pooling tolerance | 10 | 100.0 | 0.0 | 1.3446 | 100.0 | 0.0 | -9.8339 |
| CRC 7 | pool_CRC_and_self_declared_cancer_free | Low Positive | Primary analysis | 28 | 42.86 | 57.14 | -1.9915 | 100.0 | 0.0 | -10.2519 |
| | | | Instrument to instrument | 28 | 42.86 | 57.14 | -1.899 | 100.0 | 0.0 | -10.2046 |
| | | | Pooling tolerance | 28 | 42.86 | 57.14 | -1.9634 | 100.0 | 0.0 | -10.2226 |
| CRC 8 | pool_CRC_and_self_declared_cancer_free | Low Positive | Primary analysis | 29 | 44.83 | 55.17 | -1.6539 | 100.0 | 0.0 | -9.8775 |
| | | | Instrument to instrument | 29 | 51.72 | 48.28 | -1.6367 | 100.0 | 0.0 | -9.8776 |
| | | | Pooling tolerance | 29 | 48.28 | 51.72 | -1.6773 | 100.0 | 0.0 | -9.8846 |
| CRC 9 | CRC | Borderline | Primary analysis | 4 | 100.0 | 0.0 | 4.0064 | 100.0 | 0.0 | -7.5687 |
| | | | Instrument to instrument | 4 | 100.0 | 0.0 | 4.2094 | 100.0 | 0.0 | -7.6007 |
| | | | Pooling tolerance | 4 | 100.0 | 0.0 | 4.1896 | 100.0 | 0.0 | -7.3983 |
| CRC 10 | CRC | Borderline | Primary analysis | 12 | 58.33 | 41.67 | -1.3562 | 16.67 | 83.33 | -11.5284 |
| | | | Instrument to instrument | 12 | 50.0 | 50.0 | -1.469 | 8.33 | 91.67 | -11.6253 |
| | | | Pooling tolerance | 12 | 66.67 | 33.33 | -1.2061 | 16.67 | 83.33 | -11.3715 |
| CRC 11 | CRC | Borderline | Primary analysis | 17 | 52.94 | 47.06 | -1.6064 | 11.76 | 88.24 | -11.5505 |
| | | | Instrument to instrument | 18 | 55.56 | 44.44 | -1.6252 | 11.11 | 88.89 | -11.5562 |
| | | | Pooling tolerance | 17 | 70.59 | 29.41 | -1.1275 | 29.41 | 70.59 | -11.2605 |
| Healthy 1 | self_declar ed_cancer free | Borderline | Primary analysis | 17 | 5.88 | 94.12 | -3.9711 | 5.88 | 94.12 | -11.536 |
| | | | Instrument to instrument | 17 | 5.88 | 94.12 | -3.9241 | 5.88 | 94.12 | -11.5041 |
| | | | Pooling tolerance | 17 | 5.88 | 94.12 | -3.6675 | 17.65 | 82.35 | -11.3342 |

{13}

| Sample ID | Sample Disease Status | Sample Category | Level | Number of Replicates | % Called Positive Int Caller | % Called Negative Int Caller | Mean Int Score | % Called Positive MR Caller | % Called Negative MR Caller | Mean MR Score |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| CRC 12 | pool_CRC_and_self_declared_cancer_free | Borderline | Primary analysis | 12 | 41.67 | 58.33 | -2.0488 | 91.67 | 8.33 | -10.384 |
| | | | Instrument to instrument | 12 | 50.0 | 50.0 | -2.0673 | 91.67 | 8.33 | -10.3794 |
| | | | Pooling tolerance | 12 | 41.67 | 58.33 | -2.1271 | 91.67 | 8.33 | -10.3785 |
| Healthy 2 | self_declared_cancer_free | High Negative | Primary analysis | 30 | 3.33 | 96.67 | -3.8289 | 6.67 | 93.33 | -11.5008 |
| | | | Instrument to instrument | 30 | 3.33 | 96.67 | -3.7074 | 3.33 | 96.67 | -11.4957 |
| | | | Pooling tolerance | 30 | 3.33 | 96.67 | -3.8131 | 3.33 | 96.67 | -11.503 |
| Healthy 3 | self_declared_cancer_free | High Negative | Primary analysis | 30 | 13.33 | 86.67 | -3.1822 | 3.33 | 96.67 | -11.5721 |
| | | | Instrument to instrument | 30 | 13.33 | 86.67 | -3.2277 | 0.0 | 100.0 | -11.6256 |
| | | | Pooling tolerance | 30 | 13.33 | 86.67 | -3.2514 | 0.0 | 100.0 | -11.6321 |
| Healthy 4 | self_declared_cancer_free | Negative | Primary analysis | 30 | 0.0 | 100.0 | -5.0185 | 3.33 | 96.67 | -11.6048 |
| | | | Instrument to instrument | 30 | 0.0 | 100.0 | -5.1673 | 0.0 | 100.0 | -11.6603 |
| | | | Pooling tolerance | 30 | 0.0 | 100.0 | -5.0608 | 3.33 | 96.67 | -11.6075 |
| Healthy 5 | self_declared_cancer_free | Negative | Primary analysis | 30 | 0.0 | 100.0 | -4.5761 | 6.67 | 93.33 | -11.598 |
| | | | Instrument to instrument | 30 | 0.0 | 100.0 | -4.6926 | 0.0 | 100.0 | -11.6777 |
| | | | Pooling tolerance | 30 | 3.33 | 96.67 | -4.5928 | 6.67 | 93.33 | -11.547 |
| Healthy 6 | self_declared_cancer_free | Negative | Primary analysis | 36 | 0.0 | 100.0 | -5.6064 | 2.78 | 97.22 | -11.6008 |
| | | | Instrument to instrument | 36 | 0.0 | 100.0 | -5.6882 | 0.0 | 100.0 | -11.6244 |
| | | | Pooling tolerance | 36 | 2.78 | 97.22 | -5.5178 | 13.89 | 86.11 | -11.4241 |
| Healthy 7 | self_declared_cancer_free | Negative | Primary analysis | 29 | 10.34 | 89.66 | -3.8179 | 6.9 | 93.1 | -11.4639 |
| | | | Instrument to instrument | 29 | 10.34 | 89.66 | -3.9848 | 3.45 | 96.55 | -11.5322 |
| | | | Pooling tolerance | 29 | 6.9 | 93.1 | -3.9993 | 0.0 | 100.0 | -11.538 |
| Healthy 8 | self_declared_cancer_free | Negative | Primary analysis | 29 | 3.45 | 96.55 | -4.0325 | 13.79 | 86.21 | -11.5209 |
| | | | Instrument to instrument | 29 | 6.9 | 93.1 | -4.0957 | 10.34 | 89.66 | -11.568 |
| | | | Pooling tolerance | 29 | 3.45 | 96.55 | -4.0558 | 10.34 | 89.66 | -11.5503 |
| Healthy 9 | self_declared_cancer_free | Negative | Primary analysis | 34 | 0.0 | 100.0 | -4.5917 | 0.0 | 100.0 | -11.6993 |
| | | | Instrument to instrument | 34 | 0.0 | 100.0 | -4.6661 | 0.0 | 100.0 | -11.706 |
| | | | Pooling tolerance | 34 | 0.0 | 100.0 | -4.7296 | 0.0 | 100.0 | -11.7092 |
| Healthy 10 | self_declared_cancer_free | Negative | Primary analysis | 27 | 0.0 | 100.0 | -4.1272 | 0.0 | 100.0 | -11.6632 |
| | | | Instrument to instrument | 27 | 0.0 | 100.0 | -4.1657 | 0.0 | 100.0 | -11.6713 |
| | | | Pooling tolerance | 25 | 0.0 | 100.0 | -4.1356 | 0.0 | 100.0 | -11.6778 |

{14}

| Sample ID | Sample Disease Status | Sample Category | Level | Number of Replicates | % Called Positive Int Caller | % Called Negative Int Caller | Mean Int Score | % Called Positive MR Caller | % Called Negative MR Caller | Mean MR Score |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Healthy 11 | self_declared_cancer_free | Negative | Primary analysis | 29 | 3.45 | 96.55 | -4.3717 | 0.0 | 100.0 | -11.685 |
| | | | Instrument to instrument | 29 | 3.45 | 96.55 | -4.3397 | 0.0 | 100.0 | -11.6784 |
| | | | Pooling tolerance | 29 | 0.0 | 100.0 | -4.3482 | 0.0 | 100.0 | -11.656 |
| Healthy 12 | self_declared_cancer_free | Negative | Primary analysis | 16 | 0.0 | 100.0 | -4.7814 | 0.0 | 100.0 | -11.7901 |
| | | | Instrument to instrument | 16 | 0.0 | 100.0 | -4.6508 | 0.0 | 100.0 | -11.7691 |
| | | | Pooling tolerance | 16 | 0.0 | 100.0 | -4.6448 | 0.0 | 100.0 | -11.7984 |

A summary of the variance component analysis for integrated score and MR score evaluated in the precision study is presented in Table 4 and Table 5, respectively. The results are based on data from both the primary sequencing runs and the runs on the second sequencing instrument. The mean, standard deviation (SD), coefficient of variation (%CV), number of replicates (N) were calculated for each test sample. The overall variability (SD) for integrated score ranged from 0.8685 to 2.3246 (Table 4) and the overall variability (SD) for MR score ranged from 0.2135 to 0.6775 (Table 5).

Table 4. Variance component estimates for the Shield test Integrated score for each clinical sample

| Clinical Sample ID | Integrated Score | | SD | Variance Component Estimates for the Shield Test Integrated Score | | | | | | | N** |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | Mean | %CV | Total | Total | Between-Run | | | | | Repeatability Within-Batch | |
| | | | | | Between-Lot | | Between-Instrument | | Between-Operator | | |
| | | | | | Sample Prep Reagent | Sequencing Reagent | Sample Prep Instrument | Sequencing | | | |
| CRC 9 | 4.1079 | 160.35% | 6.5872* | 43.3911 (100.00%) | 35.9180 (82.78%) | 0.0000 (0.00%) | 6.0286 (13.89%) | 1.4444 (3.33%) | 0.0000 (0.00%) | 0.0000 (0.00%) | 8 |
| CRC 1 | 3.1668 | 51.64% | 1.6352 | 2.6740 (100.00%) | 0.2886 (10.79%) | 0.0000 (0.00%) | 0.0000 (0.00%) | 0.0483 (1.81%) | 0.0000 (0.00%) | 2.3371 (87.40%) | 46 |
| CRC 2 | 1.735 | 106.29% | 1.844 | 3.4005 (100.00%) | 0.7251 (21.32%) | 0.0000 (0.00%) | 0.9224 (27.13%) | 0.0569 (1.67%) | 0.0000 (0.00%) | 1.6961 (49.88%) | 36 |
| CRC 6 | 1.0786 | 132.44% | 1.4285 | 2.0405 (100.00%) | 0.5413 (26.53%) | 0.4365 (21.39%) | 0.0000 (0.00%) | 0.0446 (2.18%) | 0.0000 (0.00%) | 1.0181 (49.89%) | 19 |
| CRC 4 | 0.9206 | 234.81% | 2.1616 | 4.6724 (100.00%) | 0.0000 (0.00%) | 0.7598 (16.26%) | 0.0000 (0.00%) | 0.1272 (2.72%) | 0.0000 (0.00%) | 3.7854 (81.02%) | 24 |
| CRC 5 | 0.5236 | 393.07% | 2.0581 | 4.2360 (100.00%) | 0.2437 (5.75%) | 0.0000 (0.00%) | 1.1056 (26.10%) | 0.0884 (2.09%) | 0.0000 (0.00%) | 2.7983 (66.06%) | 24 |
| CRC 3 | 0.3539 | 632.05% | 2.2365 | 5.0021 (100.00%) | 1.2529 (25.05%) | 0.0000 (0.00%) | 2.3806 (47.59%) | 0.0794 (1.59%) | 0.3383 (6.76%) | 0.9508 (19.01%) | 46 |
| CRC 10 | -1.4126 | -131.49% | 1.8575 | 3.4503 (100.00%) | 0.0000 (0.00%) | 0.4635 (13.43%) | 0.0000 (0.00%) | 0.2301 (6.67%) | 0.0000 (0.00%) | 2.7567 (79.90%) | 24 |

{15}

| Clinical Sample ID | Integrated Score | | SD | Variance Component Estimates for the Shield Test Integrated Score | | | | | | | N** |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | Mean | %CV | Total | Total | Between-Run | | | | | Repeatability Within-Batch | |
| | | | | | Between-Lot | | Between-Instrument | | Between-Operator | | |
| | | | | | Sample Prep Reagent | Sequencing Reagent | Sample Prep Instrument | Sequencing | | | |
| CRC 11 | -1.6161 | -123.59% | 1.9973 | 3.9890(100.00%) | 0.0000(0.00%) | 2.0345(51.00%) | 0.0000(0.00%) | 0.0839(2.10%) | 0.0000(0.00%) | 1.8706(46.89%) | 35 |
| CRC 8 | -1.6453 | -102.13% | 1.6803 | 2.8235(100.00%) | 0.0000(0.00%) | 1.0327(36.58%) | 0.0000(0.00%) | 0.0602(2.13%) | 0.0000(0.00%) | 1.7306(61.29%) | 58 |
| CRC 7 | -1.9452 | -67.76% | 1.3182 | 1.7376(100.00%) | 0.0000(0.00%) | 0.1458(8.39%) | 0.0000(0.00%) | 0.0443(2.55%) | 0.0000(0.00%) | 1.5475(89.06%) | 56 |
| CRC 12 | -2.058 | -112.95% | 2.3246 | 5.4039(100.00%) | 0.0000(0.00%) | 2.8168(52.13%) | 0.0000(0.00%) | 0.0755(1.40%) | 0.0000(0.00%) | 2.5116(46.48%) | 24 |
| Healthy 3 | -3.2049 | -58.74% | 1.8825 | 3.5440(100.00%) | 0.6015(16.97%) | 0.0000(0.00%) | 1.2402(34.99%) | 0.0734(2.07%) | 0.3988(11.25%) | 1.2301(34.71%) | 60 |
| Healthy 2 | -3.7681 | -36.37% | 1.3706 | 1.8785(100.00%) | 0.0000(0.00%) | 0.3691(19.65%) | 0.0000(0.00%) | 0.0639(3.40%) | 0.2089(11.12%) | 1.2366(65.83%) | 60 |
| Healthy 7 | -3.9014 | -43.64% | 1.7028 | 2.8994(100.00%) | 0.1243(4.29%) | 0.0000(0.00%) | 0.1356(4.68%) | 0.2877(9.92%) | 0.1494(5.15%) | 2.2023(75.96%) | 58 |
| Healthy 1 | -3.9476 | -38.9% | 1.5341 | 2.3535(100.00%) | 0.0000(0.00%) | 0.0000(0.00%) | 0.0000(0.00%) | 0.0443(1.88%) | 0.4678(19.88%) | 1.8414(78.24%) | 34 |
| Healthy 8 | -4.0641 | -32.6% | 1.326 | 1.7582(100.00%) | 0.0403(2.29%) | 0.1859(10.57%) | 0.0000(0.00%) | 0.0991(5.63%) | 0.0000(0.00%) | 1.4330(81.50%) | 58 |
| Healthy 10 | -4.1465 | -28.9% | 1.1978 | 1.4347(100.00%) | 0.0742(5.17%) | 0.0000(0.00%) | 0.0000(0.00%) | 0.0622(4.34%) | 0.0000(0.00%) | 1.2984(90.49%) | 54 |
| Healthy 11 | -4.3557 | -31.35% | 1.3656 | 1.8650(100.00%) | 0.1918(10.28%) | 0.0000(0.00%) | 0.0000(0.00%) | 0.0258(1.38%) | 0.2349(12.60%) | 1.4125(75.74%) | 58 |
| Healthy 9 | -4.6289 | -27.66% | 1.2801 | 1.6387(100.00%) | 0.0415(2.53%) | 0.0214(1.31%) | 0.0000(0.00%) | 0.0428(2.61%) | 0.0000(0.00%) | 1.5330(93.55%) | 68 |
| Healthy 5 | -4.6344 | -34.74% | 1.6102 | 2.5927(100.00%) | 0.3098(11.95%) | 0.0000(0.00%) | 0.5583(21.53%) | 0.1439(5.55%) | 0.1590(6.13%) | 1.4219(54.84%) | 60 |
| Healthy 12 | -4.7161 | -18.42% | 0.8685 | 0.7544(100.00%) | 0.0000(0.00%) | 0.1361(18.04%) | 0.0000(0.00%) | 0.0385(5.10%) | 0.3098(41.06%) | 0.2700(35.79%) | 32 |
| Healthy 4 | -5.0929 | -20.92% | 1.0653 | 1.1348(100.00%) | 0.0000(0.00%) | 0.1472(12.97%) | 0.0000(0.00%) | 0.1414(12.46%) | 0.0000(0.00%) | 0.8462(74.56%) | 60 |
| Healthy 6 | -5.6473 | -17.10% | 0.9656 | 0.9323(100.00%) | 0.0017(0.18%) | 0.0000(0.00%) | 0.0000(0.00%) | 0.0646(6.93%) | 0.0329(3.52%) | 0.8331(89.36%) | 72 |
| * Results for clinical sample CRC 9 were determined to be numerically unstable and are not recommended for use** N = Total number of replicates | | | | | | | | | | | |

{16}

Table 5. Variance component estimates for the Shield test MR score for each clinical sample

| Clinical Sample ID | MR Score | | SD | Variance Component Estimates for the Shield Test MR Score | | | | | | | N** |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | Mean | %CV | Total | Total | Between-Run | | | | | Repeatability Within-Batch | |
| | | | | | Between-Lot | | Between-Instrument | | Between-Operator | | |
| | | | | | Sample Prep Reagent | Sequencing Reagent | Sample Prep Instrument | Sequencing | | | |
| CRC 9 | -7.5847 | -10.94% | 0.8298 | 0.6886 (100.00%) | 0.5012 (72.80%) | 0.0000 (0.00%) | 0.1285 (18.67%) | 0.0588 (8.54%) | 0.0000 (0.00%) | 0.0000 (0.00%) | 8 |
| CRC 1 | -7.7863 | -5.80% | 0.4515 | 0.2038 (100.00%) | 0.0415 (20.36%) | 0.0002 (0.10%) | 0.0214 (10.51%) | 0.0053 (2.61%) | 0.0000 (0.00%) | 0.1354 (66.43%) | 46 |
| CRC 2 | -8.1906 | -8.27% | 0.6775 | 0.4590 (100.00%) | 0.1467 (31.97%) | 0.0000 (0.00%) | 0.1107 (24.12%) | 0.0099 (2.16%) | 0.0000 (0.00%) | 0.1916 (41.75%) | 36 |
| CRC 3 | -8.7919 | -4.25% | 0.3735 | 0.1395 (100.00%) | 0.0120 (8.59%) | 0.0000 (0.00%) | 0.0187 (13.39%) | 0.0025 (1.76%) | 0.0000 (0.00%) | 0.1064 (76.26%) | 46 |
| CRC 8 | -9.8776 | -5.28% | 0.5216 | 0.2721 (100.00%) | 0.0950 (34.91%) | 0.0000 (0.00%) | 0.0264 (9.69%) | 0.0021 (0.77%) | 0.0131 (4.81%) | 0.1355 (49.82%) | 58 |
| CRC 6 | -9.9171 | -4.05% | 0.402 | 0.1616 (100.00%) | 0.0678 (41.94%) | 0.0554 (34.26%) | 0.0164 (10.13%) | 0.0051 (3.18%) | 0.0000 (0.00%) | 0.0169 (10.49%) | 19 |
| CRC 5 | -10.0456 | -5.13% | 0.5151 | 0.2653 (100.00%) | 0.0000 (0.00%) | 0.1279 (48.20%) | 0.0000 (0.00%) | 0.0045 (1.68%) | 0.0000 (0.00%) | 0.1329 (50.11%) | 24 |
| CRC 7 | -10.2283 | -3.01% | 0.3083 | 0.0950 (100.00%) | 0.0078 (8.22%) | 0.0000 (0.00%) | 0.0000 (0.00%) | 0.0070 (7.37%) | 0.0110 (11.59%) | 0.0692 (72.82%) | 56 |
| CRC 12 | -10.3817 | -5.92% | 0.6149 | 0.3781 (100.00%) | 0.0000 (0.00%) | 0.2385 (63.09%) | 0.0000 (0.00%) | 0.0018 (0.49%) | 0.0000 (0.00%) | 0.1377 (36.43%) | 24 |
| CRC 4 | -10.9327 | -3.71% | 0.406 | 0.1648 (100.00%) | 0.0000 (0.00%) | 0.0952 (57.77%) | 0.0082 (4.99%) | 0.0551 (33.41%) | 0.0000 (0.00%) | 0.0063 (3.84%) | 24 |
| Healthy 7 | -11.498 | -3.67% | 0.4221 | 0.1782 (100.00%) | 0.0000 (0.00%) | 0.0694 (38.94%) | 0.0000 (0.00%) | 0.0452 (25.35%) | 0.0000 (0.00%) | 0.0636 (35.71%) | 58 |
| Healthy 2 | -11.4982 | -2.33% | 0.2674 | 0.0715 (100.00%) | 0.0019 (2.59%) | 0.0000 (0.00%) | 0.0000 (0.00%) | 0.0016 (2.20%) | 0.0205 (28.72%) | 0.0475 (66.49%) | 60 |
| Healthy 1 | -11.52 | -3.03% | 0.3486 | 0.1215 (100.00%) | 0.0000 (0.00%) | 0.0169 (13.93%) | 0.0000 (0.00%) | 0.0042 (3.44%) | 0.0000 (0.00%) | 0.1004 (82.63%) | 34 |
| Healthy 8 | -11.5445 | -3.23% | 0.3729 | 0.1390 (100.00%) | 0.0125 (8.99%) | 0.0000 (0.00%) | 0.0000 (0.00%) | 0.0256 (18.40%) | 0.0202 (14.55%) | 0.0807 (58.07%) | 58 |
| CRC 11 | -11.5534 | -3.97% | 0.4591 | 0.2108 (100.00%) | 0.0000 (0.00%) | 0.0624 (29.59%) | 0.0147 (6.99%) | 0.0153 (7.26%) | 0.0279 (13.22%) | 0.0905 (42.93%) | 35 |
| CRC 10 | -11.5769 | -5.50% | 0.6367 | 0.4054 (100.00%) | 0.0000 (0.00%) | 0.0967 (23.87%) | 0.0000 (0.00%) | 0.0676 (16.67%) | 0.0000 (0.00%) | 0.2410 (59.46%) | 24 |
| Healthy 3 | -11.5988 | -2.56% | 0.2967 | 0.0880 (100.00%) | 0.0059 (6.72%) | 0.0003 (0.33%) | 0.0203 (23.10%) | 0.0142 (16.10%) | 0.0000 (0.00%) | 0.0473 (53.74%) | 60 |
| Healthy 6 | -11.6126 | -1.8% | 0.2135 | 0.0456 (100.00%) | 0.0000 (0.00%) | 0.0032 (7.10%) | 0.0000 (0.00%) | 0.0093 (20.37%) | 0.0000 (0.00%) | 0.0331 (72.53%) | 72 |
| Healthy 4 | -11.6325 | -2.1% | 0.2426 | 0.0589 (100.00%) | 0.0014 (2.42%) | 0.0000 (0.00%) | 0.0000 (0.00%) | 0.0327 (55.51%) | 0.0033 (5.68%) | 0.0214 (36.40%) | 60 |

{17}

| Clinical Sample ID | MR Score | | SD | Variance Component Estimates for the Shield Test MR Score | | | | | | | N** |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | Mean | %CV | Total | Total | Between-Run | | | | | Repeatability Within-Batch | |
| | | | | | Between-Lot | | Between-Instrument | | Between-Operator | | |
| | | | | | Sample Prep Reagent | Sequencing Reagent | Sample Prep Instrument | Sequencing | | | |
| Healthy 5 | -11.6379 | -2.3% | 0.2728 | 0.0744 (100.00%) | 0.0067 (9.02%) | 0.0000 (0.00%) | 0.0067 (9.01%) | 0.0300 (40.24%) | 0.0000 (0.00%) | 0.0311 (41.72%) | 60 |
| Healthy 10 | -11.6672 | -2.46% | 0.2866 | 0.0822 (100.00%) | 0.0240 (29.25%) | 0.0000 (0.00%) | 0.0039 (4.73%) | 0.0027 (3.33%) | 0.0099 (12.04%) | 0.0416 (50.65%) | 54 |
| Healthy 11 | -11.6817 | -2.92% | 0.3407 | 0.1161 (100.00%) | 0.0000 (0.00%) | 0.0512 (44.13%) | 0.0000 (0.00%) | 0.0011 (0.95%) | 0.0014 (1.20%) | 0.0624 (53.73%) | 58 |
| Healthy 9 | -11.7026 | -1.97% | 0.2309 | 0.0533 (100.00%) | 0.0024 (4.54%) | 0.0000 (0.00%) | 0.0054 (10.21%) | 0.0015 (2.85%) | 0.0000 (0.00%) | 0.0439 (82.40%) | 68 |
| Healthy 12 | -11.7796 | -2.80% | 0.3294 | 0.1085 (100.00%) | 0.0107 (9.82%) | 0.0000 (0.00%) | 0.0201 (18.52%) | 0.0022 (2.04%) | 0.0000 (0.00%) | 0.0755 (69.63%) | 32 |
| * Results for clinical sample CRC 9 were determined to be numerically unstable and are not recommended for use ** N = Total number of replicates | | | | | | | | | | | |

Based on data from precision and other analytical studies where multiple replicates of the same sample were tested, samples with integrated scores ranging from -2.3590 to -0.1243 yielded less than $95\%$ concordant integrated calls across replicates; samples with MR scores ranging from -11.1935 to -10.6569 yielded less than $95\%$ concordant MR calls across replicates. Based on the clinical study, $\sim 15\%$ samples generated an integrated score within the range that will yield less than $95\%$ concordant call (Table 6), and $\sim 15\%$ samples generated a MR score within the range that will yield less than $95\%$ concordant call (Table 7).

{18}

Table 6. Prevalence, agreement rates, and misclassification rates for samples with Shield test Integrated scores close to the decision threshold

| Shield Test Integrated Score | | | | | | | | | |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Range of Shield Test Integrated Score | Source of Intermediate Precision Estimate | | Distance to Cutoff (SD) | Number of CV Samples Within this Range / Total Samples | Percentage of CV Samples Within this Range | NPA [95% CI] | False Positive Calls / Total Expected Negative Replicates in AV | PPA [95% CI] | False Negative Calls / Total Expected Positive Replicates in AV |
| | VCA Category | Intermediate Precision Estimate | | | | | | | |
| [-4.2353, -2.9844] | High Negative | 1.2509 | -2 to -1 | 3448 / 7861 | 43.86% | 95.62% [93.99, 96.91] | 36 / 822 | Not Available | Not Available |
| [-2.9844, -2.3590] | | | -1 to -0.5 | 1217 / 7861 | 15.48% | 90.62% [82.95, 95.62] | 9 / 96 | 22.45% [11.77, 36.62] | 38 / 49 |
| [-2.3590, -2.0462] | | | -0.5 to -0.25 | 370 / 7861 | 4.71% | 68.75% [41.34, 88.98] | 5 / 16 | 40.48% [29.90, 51.75] | 50 / 84 |
| [-2.0462, -1.7335] | | | -0.25 to Threshold | 287 / 7861 | 3.65% | 62.50% [35.43, 84.80] | 6 / 16 | 43.08% [34.43, 52.05] | 74 / 130 |
| [-1.7335, -1.3312] | Low Positive | 1.6092 | Threshold to 0.25 | 229 / 7861 | 2.91% | Not Available | Not Available | 52.63% [46.20, 59.00] | 117 / 247 |
| [-1.3312, -0.9289] | | | 0.25 to 0.5 | 127 / 7861 | 1.60% | Not Available | Not Available | 68.46% [60.35, 75.82] | 47 / 149 |
| [-0.9289, -0.1243] | | | 0.5 to 1 | 103 / 7861 | 1.31% | 0.00% [0.00, 45.93] | 6 / 6 | 80.00% [70.54, 87.51] | 19 / 95 |
| [-0.1243, 1.4849] | | | 1 to 2 | 59 / 7861 | 0.75% | Not Available | Not Available | 94.58% [91.75, 96.66] | 20 / 369 |

{19}

Table 7. Prevalence, agreement rates, and misclassification rates for samples with Shield test MR Scores close to the decision threshold

| Shield Test MR Score | | | | | | | | | |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Range of Shield Test MR Score | Source of Intermediate Precision Estimate | | Distance to Cutoff (SD) | Number of CV Samples Within this Range / Total Samples | Percentage of CV Samples Within this Range | NPA [95% CI] | False Positive Calls / Total Expected Negative Replicates AV | PPA [95% CI] | False Negative Calls / Total Expected Positive Replicates in AV |
| | VCA Category | Intermediate Precision Estimate | | | | | | | |
| [-11.5896, -11.3255] | High Negative | 0.2641 | -2 to -1 | 3567 / 7861 | 45.38% | 95.24% [93.50, 96.63] | 37 / 777 | 13.43% [6.33, 23.97] | 58 / 67 |
| [-11.3255, -11.1935] | | | -1 to -0.5 | 938 / 7861 | 11.93% | 100.00% [90.26, 100.00] | 0 / 36 | 33.33% [17.29, 52.81] | 20 / 30 |
| [-11.1935, -11.1274] | | | -0.5 to -0.25 | 281 / 7861 | 3.57% | 66.67% [34.89, 90.08] | 4 / 12 | 50.00% [18.71, 81.29] | 5 / 10 |
| [-11.1274, -11.0614] | | | -0.25 to Threshold | 209 / 7861 | 2.66% | 66.67% [22.28, 95.67] | 2 / 6 | Not Available | Not Available |
| [-11.0614, -10.9603] | CRC Donor* | 0.4045 | Threshold to 0.25 | 183 / 7861 | 2.32% | 0.00% [0.00, 45.93] | 6 / 6 | 52.17% [30.59, 73.18] | 11 / 23 |
| [-10.9603, -10.8592] | | | 0.25 to 0.5 | 107 / 7861 | 1.36% | 0.00% [0.00, 45.93] | 6 / 6 | 61.90% [45.64, 76.43] | 16 / 42 |
| [-10.8592, -10.6569] | | | 0.5 to 1 | 99 / 7861 | 1.26% | Not Available | Not Available | 80.00% [51.91, 95.67] | 3 / 15 |
| [-10.6569, -0.2524] | | | 1 to 2 | 84 / 7861 | 1.07% | 0.00% [0.00, 45.93] | 6 / 6 | 91.67% [84.24, 96.33] | 8 / 96 |
| *CRC Donor category used in place of Low Positive for Shield Test MR Scores because the Low Positive category only had 1 sample (with 12 reps in the primary analysis). By way of contrast, the Low Positive category for Shield Test Integrated Scores contained data for 65 replicates in the primary analysis derived from 4 unique clinical CRC samples. | | | | | | | | | |

# D. Analytical Specificity

# 1) In-silico Analysis of Primers and Probes

The specificity of the primers and probes used in the Shield test was evaluated by testing the potential of the probes and primers to amplify the non-specific products from human DNA and assessing the potential for incorrect results due to commensal microorganism using publicly available in-silico analysis.

The probe specificity was determined by mapping probe sequences to the human genome and performing in silico analyses to detect possible contaminant non-human sequences. Study results found no potential for off-target amplification or binding from human, enterobacterial or viral non-specific targets (common bacterial, fungal, and viral sources that are commonly found in the human blood or on the skin).

{20}

# 2) Cross-Reactivity with Non-Colorectal Cancers and Diseases

# Non-CRC Cancers

The potential for cross-reactivity with non-colorectal cancers was evaluated two ways:

1. The incidence of cancer in subjects enrolled in the clinical study were evaluated for a diagnosis of cancer (ECLIPSE) within one year of enrollment (as of March 2024). The rate of non-CRC cancers was compared between participants who had false positive results for advanced neoplasia (AN), and true negative for AN. No statistically significant difference in the incidence rate was observed (0.8% [5/640] versus 0.9% [51/5,502] respectively, adjusted p-value=0.4584).
2. Evaluation of Shield positivity on 218 specimens from subjects with a known diagnosis of a non-CRC cancer. The Shield positivity rates range from $50.0\%$ to $92.9\%$ across 9 cancer types. The samples in the cross-reactivity study were collected from individuals with a known diagnosis of cancer, which is not representative of an asymptomatic intended use population. The result is an overestimation of the non-CRC cancer detection. Table 8 estimates the worst-case scenario for false positives based on the estimated incidence rate and false positive findings. The test is not intended for detection of other cancers.

Table 8. Cross reactivity in the context of Cancer Incident Rates

| Cancer Type | Incidence in US in 20224 | Incident Rate per 10,000 individuals | Number of positive calls in 10,000 subjects |
| --- | --- | --- | --- |
| Bladder | 81,180 | 2.44 | 2.07 |
| Breast | 290,560 | 8.74 | 6.99 |
| Gastric* | 47,020 | 1.41 | 1.30 |
| Kidney | 79,000 | 2.34 | 1.17 |
| Liver | 41,260 | 1.24 | 1.09 |
| Lung | 236,740 | 7.12 | 5.44 |
| Melanoma | 99,780 | 3.00 | 1.71 |
| Ovarian | 19,880 | 0.60 | 0.37 |
| Pancreatic | 62,210 | 1.87 | 1.73 |
| Prostate | 268,490 | 8.07 | 5.10 |
| Total | | | 26.97 |
| *Gastric cancer incidences were estimated by totaling Esophagus and Stomach cancer incidences | | | |

{21}

# Non-Cancer Diseases

The potential for non-cancer disease cross-reactivity was evaluated in 2440 subjects from the interim analysis dataset in the ECLIPSE study that did not have CRC or AA detected by colonoscopy. False positive results were evaluated in combination with the disease prevalence which was observed in interim analysis dataset in ECLIPSE to assess the impact of disease on Shield specificity. The results are shown in Table 9 below.

Table 9. Positivity rates for observed co-morbidities within ECLIPSE interim analysis data set

| Disease | Number of samples | Prevalence Observed in Interim Analysis Set | Positivity Rate | Projected number of positive Shield calls in 10,000 subjects |
| --- | --- | --- | --- | --- |
| Hypertension | 569 | 23.3% | 10.5% | 245.58 |
| Dyslipidemia | 515 | 21.1% | 12.4% | 262.27 |
| Diabetes Mellitus | 266 | 10.9% | 13.9% | 151.62 |
| Gastroesophageal Reflux Disease | 212 | 8.7% | 8.5% | 73.81 |
| Allergy | 142 | 5.8% | 3.5% | 20.49 |
| Hypothyroidism | 140 | 5.7% | 12.9% | 73.84 |
| Anxiety | 86 | 3.5% | 5.8% | 20.48 |
| Depression | 85 | 3.5% | 11.8% | 41.00 |
| Osteoporosis | 76 | 3.1% | 17.1% | 53.29 |
| Arthritis | 73 | 3.0% | 8.2% | 24.59 |
| Asthma | 72 | 3.0% | 5.6% | 16.69 |
| Constipation | 66 | 2.7% | 24.2% | 65.57 |
| Benign Prostatic Hypertrophy | 64 | 2.6% | 15.6% | 40.97 |
| Indigestion | 59 | 2.4% | 13.6% | 33.02 |
| Drug Allergy | 57 | 2.3% | 12.3% | 28.80 |

## 3) Interfering Substances

To evaluate the impact of potential endogenous interfering substances (unconjugated bilirubin, conjugated bilirubin, triglycerides, gDNA, albumin, and hemoglobin) on the performance of Shield, four clinical specimens (2 minimally manipulated positive and 2

{22}

negative samples) were tested in 8 replicates in the presence and absence of the above-mentioned substances at levels shown in Table 10 below. No interference with the Shield test was observed for any of the substances at the concentrations tested.

Table 10. Endogenous Interfering Substances Tested
| Interferent | Amount added [g/L or as specified] |
| --- | --- |
| Albumin | 60 |
| Conjugated Bilirubin | 0.4 |
| Unconjugated Bilirubin | 0.4 |
| Hemoglobin | 10 |
| Triglycerides, Total | 15 |
| gDNA | 100 ng per replicate |

E. Cross-Contamination / Carry Over

Cross-contamination and carry over was evaluated using a checkerboard design alternating pre-characterized high positive male donor samples and low negative female donor samples in two plates processed consecutively using the same line of instruments. A total of 47 samples of each type were tested in 2 batches for a total of 188 replicates. The PPA and NPA for first plate is 100% and the PPA and NPA for the second plate is 100.00% and 97.83%, respectively. One false positive Shield test result was observed across the 94 Low Negative sample replicates. The false positive was from a single replicate among 31 replicates tested of a Low Negative sample, due to a false positive Integrated Call. The Integrated Score for that replicate was 2.414 x SD (per Precision Study) from the mean of the 31 replicates for the discordant Low Negative sample. The probability of observing a maximum deviation from the mean 2.5 x SD or greater across a set of 31 normally distributed scores is approximately 17.6%, supporting the hypothesis that the outlying value is not necessarily attributed to a mechanism outside of expected variability in Shield Test Integrated Scores.

F. Robustness

Assay Workflow Guardbanding

The objective of the robustness study was to evaluate the tolerance of the Shield Test to variation in critical assay workflow parameters in the categories listed below:

- Reagent Volumes relative to instrument tolerances for key reagents
- Incubation Times including library enrichment and cfDNA hybridization times
- Hold Point durations for extracted cfDNA and samples during processing

{23}

- Temperature variations during the enrichment hybridization process.

Six (6) positive and seven (7) negative clinical specimens including samples that consisted of extracted cfDNA from pooled self-declared healthy individuals spiked with cfDNA from clinical CRC positive samples to varying levels of the Shield component test scores were tested in 8-15 replicates. A minimum of 6 positive and 6 negative samples were tested per guardbanded condition and compared to baseline reference samples. PPA for the overall Shield call observed at all conditions tested is 100%, and NPA is 100% except three conditions. The variation of scores between the control and testing conditions is consistent with the assay measurement variability observed in the precision study, indicating that the Shield performance is robust with respect to critical assay workflow parameters.

## Input Guardbanding

An input guardbanding study was conducted to evaluate the robustness of the Shield assay at different input levels. cfDNA workflow for Shield Test uses post-sequencing metrics to evaluate if a sufficient amount of input material was sequenced successfully and one of post-sequencing metrics reflective of DNA molecule recovery is Non-Singleton Coverage (NSC), which is highly correlated with the input amount into the assay.

The robustness of Shield to the cfDNA input level was assessed with 5 cfDNA positive and 4 cfDNA negative samples including samples with scores near the clinical decision cutoff that were selected at NSC > 1000 and diluted to NSC ranging from 400 (below the QC threshold of 500) to 1000. Each of the 5 positive and 4 negative samples was tested in 5 replicates at NSC level of 400, 600, 800 and 1000, while 1 replicate was tested for 3 positive and 3 negative samples at high cfDNA input (2000-5000 NSC), yielding a total of 186 replicates. The study results demonstrated at least 95% agreement for positive and negative samples, indicating the robustness of the assay at low cfDNA input levels.

## G. Stability

### Whole Blood Stability

Whole blood stability was assessed using blood samples collected from 70 self-declared cancer-free donors and 60 independent pooled clinical positive samples at different time points (day 1, day 6, day 8/9 and day 10) to support stability of whole blood specimens collected in BCTs for up to 9 days across the expected range of sample transport and storage conditions. An additional stability study was also performed to demonstrate that the device performance for whole blood processed at Day 1 can represent the device performance for samples processed on the date of collection. Plasma was isolated from two BCTs per donor at each of the isolation time points and resulting plasma was processed through the downstream Shield workflow steps. A replicate for each sample was processed for this study, and classification calls were compared between reference and storage conditions.

PPA is 100% for both winter and summer shipping conditions, while NPA for summer shipping condition are 96.15%, 96.97% and 96.30% at day 6, day 8, and day 10, respectively and NPA for winter shipping condition at day 6 is 93.10%. No significant

{24}

difference in Shield integrated score and MR score was observed at different timepoint for each sample transport and storage condition tested.

## Plasma Stability: Short-term

A short-term plasma stability study was conducted to evaluate the stability of plasma collected in BCT tubes for short-term storage conditions in the Shield workflow. A total of 29 healthy and 42 CRC samples were tested at 3 time-points (day 8, day 16 and day 31 with 12-15 CRC samples per timepoint) and 2-3 freeze-thaw cycles. Each healthy donor plasma sample was tested in 2-3 replicates at each timepoint (3 replicates at baseline and 2 replicates at later timepoints) and 1 replicate was tested per CRC donor sample per timepoint. PPAs for overall Shield call of CRC donor samples were 100%, while 7 samples from self-declared cancer free donors yielded discordant calls. Based on analysis of the integrated score and MR score, the difference between time points and freeze-thaw cycles is within the variation of scores observed in the precision study and discordant calls are due to samples with integrated score and MR score close to the cutoff. The results support the short-term stability of plasma isolated from BCTs for 30 days with up to 3 freeze-thaw cycles when stored at -80°C.

## Plasma Stability: Long-term

A long-term plasma stability study was designed to evaluate the impact of long-term plasma storage and demonstrate equivalence to plasma processed within 30 days of isolation from BCT tubes. A total of 30 healthy and 70 CRC samples are to be tested across 4 time-points. This study is ongoing and results were obtained for the T0 and T1 (7 month) time-points. No statistically significant degradation in stability was observed at 7 months. Currently, this study has demonstrated stability for at least 6 months.

## cfDNA stability

cfDNA stability was evaluated with 4 clinical samples and 2 minimally manipulated cfDNA samples constructed by mixing cfDNA extracted from normal plasma pool and cfDNA extracted from CRC plasma to varying levels of the Shield component test scores. Samples were tested in 20 replicates at each timepoint (e.g., baseline, 4 months, 7 months, 10 months and 13 months). No significant differences in either integrated score and MR score were observed between baseline and later timepoint, demonstrating cfDNA stability for 12 months when stored at -20°C.

## Reagent Shelf-Life Stability

Shelf-life stability testing for the Shield reagents was conducted by evaluating the end-to-end functional performance of 3 lots of reagents over a period of 13 months to support a 12 month shelf-life claim. Each reagent lot comprised unique lots of the individual reagents, which were tested at various time points over 13 months. Reagents stored at ≤ -20°C were freeze-thawed 3 times prior to testing. A set of cfDNA samples consisting of extracted cfDNA from pooled self-declared healthy individuals spiked with cfDNA from clinical CRC positive samples to varying levels of the Shield component test scores were evaluated in 20 replicated at each time point (e.g., baseline, 4 months, 7 months, 10 months and 13 months). The results of the reagent shelf-life stability study showed that there was no change in assay performance throughout the 13 months of testing for all reagents

{25}

demonstrating that Shield assay reagents are stable for 12 months.

## On-Board Stability

The purpose of the on-board stability studies was to determine reagent stability after using and holding reagents under different process steps. All reagents required for the Shield workflow were tested. Three clinical and minimally manipulated positive samples and three clinical negative samples were tested in 16-24 replicates per condition to evaluate the tolerance of the Shield assay to the final onboard hold conditions tested. No significant difference in assay performance or Shield scores between test condition and reference condition were observed. Samples were processed with the Shield workflow with the following hold points and were shown to be stable for the stated hold times:

- Shield reagents are stable for at least 30 minutes at room temperature on the Hamilton Microlab STAR deck.
- Samples are stable at 2-8°C in PCR master mix for at least 24 hours.

## In-Use Stability

An in-use stability study was conducted to evaluate the stability of 3 hold-points in the Shield workflow. The in-use stability study utilized 1,116 sample replicates representative of the positive and negative Shield component scores to verify the following hold conditions:

- Samples stored after Library Prep Clean-up at -15 to -25°C for 20 days with one freeze-thaw cycle
- Samples stored after Enrichment Transfer at -15 to -25°C for 14 days with one freeze-thaw cycle
- Samples stored after Sequencing Normalization at -15 to -25°C for 13 days with one freeze-thaw cycle

Test condition samples were tested for the listed conditions and compared against the reference condition samples that were not held during processing. There was no significant difference between the test or reference samples for the 3 tested hold conditions, supporting the use of the hold points as part of the assay process.

## Animal Studies

N/A

## Additional Studies

N/A

## XI. SUMMARY OF PRIMARY CLINICAL STUDY

The pivotal study ECLIPSE (“Evaluation of the ctDNA LUNAR Test in an Average Patient Screening Episode”) was conducted to generate data to support the safety and effectiveness of Shield as a blood-based screening test for the detection of alterations

{26}

associated with the presence of colorectal cancer (CRC) from whole blood samples. To evaluate the performance of Shield, the test result (negative or positive) was compared with the histopathological result from colonoscopy examination and histopathological diagnosis of all lesions discovered during the colonoscopy. Based on this comparison, Shield sensitivity (true positive fraction) was $83.1\%$ (54/65) for subjects with a histopathological diagnosis of CRC and $13.2\%$ (147/1116) for subjects with a diagnosis of advanced adenoma (Category 2a - 2e, Table 6). For subjects without a diagnosis of CRC or AA, Shield specificity (true negative fraction) was $89.6\%$ (5982/6680).

# A. Study Design

The ECLIPSE study was a multi-site, prospective, non-randomized, observational study designed to evaluate the clinical performance of Shield in patients 45 – 84 years of age who were of average risk for CRC. Patients eligible for CRC screening and intending to undergo colonoscopy were enrolled in the study. The study enrolled a total of 24,876 subjects from 265 sites across the US between October 8, 2019, and the data cutoff on September 30, 2022. Blood samples were collected from all patients who consented to enroll in the study and met eligibility criteria. Blood collection was performed prospectively using Guardant Blood Collection Kits from all enrolled subjects prior to the patient undergoing standard of care colonoscopy and were processed and analyzed at Guardant Health. Performance of the Shield test was compared against colonoscopy result. Central pathology reviews were conducted for lesion classification. The lesion of greatest clinical significance was used to classify each subject into one of the histopathology categories according to the pre-specified standards outlined in Table 11. These categories were used to designate the reference result for the purpose of determining test sensitivity and specificity, positive predictive value and negative predictive value.

Table 11. Colonoscopy/Histopathology Diagnosis Category Descriptions

| Category | Colonoscopy Findings | Class for Reference Result |
| --- | --- | --- |
| 1 | Colorectal cancer, any stage | CRC |
| 2 | Advanced adenoma | AA |
| 2a | Carcinoma in situ, any size | |
| 2b | High-grade dysplasia, any size | |
| 2c | Villous growth % (>25%), any size | |
| 2d | Tubular adenoma, ≥10 mm | |
| 2e | Serrated lesion, ≥10 mm (includes sessile serrated adenoma/polyp) | |
| 3 | Non-advanced adenoma, >3 adenomas, <10 mm | Non-AN |
| 4 | Non-advanced adenoma, 1 or 2 adenomas, >5 mm, <10 mm | |
| 5 | Non-advanced adenoma, 1 or 2 adenomas, ≤5 mm | |
| 6 | Negative, or other findings | |
| 7 | Not evaluable | |

{27}

B. Clinical Inclusion and Exclusion Criteria

Enrollment in the ECLIPSE study was limited to subjects who met the following inclusion criteria:

- Aged 45 to 84 years at time of consent.
- Intended to undergo screening colonoscopy.
- Considered by a physician or healthcare provider as being of average risk for CRC.
- Willing to consent to blood draw pre-bowel preparation administration prior to undergoing colonoscopy within 60 days (amended to 6 months) of the date of the investigational blood draw.
- Willing to consent to follow-up for two years as per protocol.

Subjects were not permitted to enroll in the ECLIPSE study if any of the following exclusion criteria was met:

- Undergoing colonoscopy for investigation of symptoms.
- Has undergone colonoscopy within preceding 9 years.
- Positive FIT/fecal occult blood test result within the previous 6 months.
- Has completed Cologuard or Epi proColon testing within the previous 3 years.
- Personal history of CRC.
- Personal history of any malignancy (patients who have undergone surgical removal of skin squamous cell cancer may be enrolled provided the procedure was completed at least 12 months prior to the date of provision of informed consent for the study).
- Known diagnosis of inflammatory bowel disease.
- Currently taking any anti-neoplastic or disease-modifying anti-rheumatic drugs.
- Family history of CRC, defined as having one or more first-degree relatives (parent, sibling, or child) with CRC at any age.
- Known hereditary/germline risk of CRC (for example, Lynch syndrome or hereditary nonpolyposis CRC, or familial adenomatous polyposis).
- Any major physical trauma (e.g., disruption of tissue, surgery, organ transplant, blood product transfusion) within the 30 days leading up to the provision of informed consent.
- Known medical condition which, in the opinion of the Investigator, should preclude enrollment into the study.
- Participation in a clinical research study in which an experimental medication has been administered or may be administered within the 30 days leading up to providing informed consent or may be administered through the time of colonoscopy.

C. Follow-up Schedule

All patients were contacted at 1 and 2 years after blood sample collection to confirm diagnoses of interval malignancies.

D. Clinical Performance Measures

Primary objectives

The primary objective of this study was to establish the performance characteristics of

{28}

the Shield test sensitivity for CRC (category 1, Table 11) and specificity of non-advanced neoplasia (categories 3, 4, 5, and 6), Table 11) in average-risk patients against the clinical results defined by colonoscopy/histopathology diagnosis. The primary performance measures were analyzed based on the Guardant's predefined acceptance criteria for sensitivity of CRC and specificity for AN:

- With regards to the Shield sensitivity for subjects with CRC, the lower bound of the two-sided 95% confidence interval must be greater than 65%.
- With regards to the Shield specificity for subjects with non-AN, the lower bound of the two-sided 95% confidence interval must be greater than 85%.

## Secondary objectives

The secondary objective was to establish the sensitivity of the Shield test in the detection of AA in average-risk patients. No performance goal was predefined for the secondary performance measure of AA sensitivity.

## E. Accountability of PMA Clinical Validation Dataset

Samples were collected from a total of 24,876 subjects at 265 sites for the Shield test. The disposition of the specimens and colonoscopy results from patients enrolled into the clinical study is as follows:

- Of the total 24,876 subjects, 1,999 subjects from a prespecified enrollment time window were used toward the device development.
- Of the remaining 22,877 subjects, 10,179 subjects were randomly selected not to be screened with the Shield test. The remaining 12,698 subjects included all CRC subjects and a proportion of non-CRC subjects selected through random down-sampling to match US Census age distribution.
- Of the 12,698 subjects, 10,297 subjects met study inclusion / exclusion criteria and have valid colonoscopy within 183 days and have valid Shield results. This population included 65 subjects with CRC.
- Of the 10,297 subjects, 2,436 were randomly selected for interim specificity analysis and cut-offs selection, therefore, 7,861 subjects were included in pivotal clinical validation dataset.
- The total number of patients in the final clinical validation evaluable dataset consisted of 7,861 subjects with valid colonoscopy and valid Shield test results that were analyzed in the primary analysis dataset.

A sample flowchart for the whole validation dataset is shown in Figure 1 below.

Figure 1. Patient accountability diagram showing breakdown of patient samples and colonoscopy results with sequential application of exclusions between the total available for primary analysis and those not included in the primary analysis.

{29}


The percent of Shield Invalid results was $3.1\%$ (345/11,211) with $95\% \mathrm{CI}$ : $(2.8\%, 3.4\%)$ . Invalid results were excluded from the data analysis.

Guardant initiated enrollment into the ECLIPSE study on October 8, 2019, with database cutoff on September 30, 2022. The device underwent modifications twice during the course of the study (addition of a protein testing workflow in July 2021 and removal of the same workflow October 2022). The final cfDNA-only device (the Shield test) remained unchanged i.e., the same cfDNA assay workflow, calling algorithms, and classification scores. The final Shield device was locked prior to unblinding of the colonoscopy/biopsy results. The performance reported by Guardant using the final Shield cfDNA cutoff was comparable to a published methodology that enables the use of a pre-specified fixed target for specificity in the primary analysis dataset (Kondratovich et al, 2005). Analyses comparing performance of the primary dataset excluding and including the interim analysis population demonstrated that there were no significant differences in clinical performance observed between subgroups. FDA concluded that the sensitivity and specificity data presented are representative of the device performance.

{30}

# F. Study Population Demographics and Baseline Parameters

The demographic and baseline characteristics for subjects in the primary analysis dataset considered by Guardant (7,861 subjects constituting final clinical validation evaluable dataset) are presented in Table 12. There was generally a balance of male and female study participants, and the average age was 60 years. $79\%$ of the subjects were White, $12\%$ were Black or African American, and $13\%$ were of Hispanic or Latino. The majority of subjects $(70.2\%)$ never smoked.

Table 12. Demographics and Baseline Characteristics of Subjects by Procedural and Lesion Findings

| Characteristic | Primary dataset (N=7,861) | CRC (Category 1) (N = 65) | AA (Category 2) (N = 1116) | Non-AN (Category 3-6) (N = 6680) | Non-CRC (Category 2-6) (N = 7796) |
| --- | --- | --- | --- | --- | --- |
| Age (years) | | | | | |
| Mean (SD) | 60.3 (9.14) | 63.2 (8.26) | 61.6 (8.67) | 60.0 (9.20) | 60.3 (9.14) |
| Median | 60 | 63 | 62 | 60 | 60 |
| Min, Max | 45, 84 | 45, 82 | 45, 82 | 45, 84 | 45, 84 |
| Age Group, n (%) | | | | | |
| 45-49 | 640 (8.1) | 4 (6.2) | 56 (5.0) | 580 (8.7) | 636 (8.2) |
| 50-59 | 3055 (38.9) | 13 (20.0) | 385 (34.5) | 2657 (39.8) | 3042 (39.0) |
| 60-69 | 2440 (31.0) | 34 (52.3) | 417 (37.4) | 1989 (29.8) | 2406 (30.9) |
| 70-79 | 1670 (21.2) | 13 (20.0) | 252 (22.6) | 1405 (21.0) | 1657 (21.3) |
| 80+ | 56 (0.7) | 1 (1.5) | 6 (0.5) | 49 (0.7) | 55 (0.7) |
| Gender, n (%) | | | | | |
| Female | 4218 (53.7) | 30 (46.2) | 511 (45.8) | 3677 (55.0) | 4188 (53.7) |
| Male | 3643 (46.3) | 35 (53.8) | 605 (54.2) | 3003 (45.0) | 3608 (46.3) |
| Race, n (%) | | | | | |
| American Indian or Alaska Native | 14 (0.2) | 0 | 2 (0.2) | 12 (0.2) | 14 (0.2) |
| Asian | 560 (7.1) | 4 (6.2) | 56 (5.0) | 500 (7.5) | 556 (7.1) |
| Black or African American | 931 (11.8) | 10 (15.4) | 121 (10.8) | 800 (12.0) | 921 (11.8) |
| Native Hawaiian or Other Pacific Islander | 19 (0.2) | 0 | 2 (0.2) | 17 (0.3) | 19 (0.2) |
| White | 6167 (78.5) | 49 (75.4) | 917 (82.2) | 5201 (77.9) | 6118 (78.5) |
| Other | 137 (1.7) | 1 (1.5) | 16 (1.4…