Guardant360 CDx

P200010 · Guardant Health, Inc. · PQP · Aug 7, 2020 · Pathology

Device Facts

• Record ID: P200010
• Device Name: Guardant360 CDx
• Applicant: Guardant Health, Inc.
• Product Code: PQP · Pathology
• Decision Date: Aug 7, 2020
• Decision: APPR
• Device Class: Class 3

Intended Use

Guardant360® CDx is a qualitative next generation sequencing-based in vitro diagnostic device that uses targeted high throughput hybridization-based capture technology for detection of single nucleotide variants (SNVs), insertions and deletions (indels) in 55 genes, copy number amplifications (CNAs) in two (2) genes, and fusions in four (4) genes. Guardant360 CDx utilizes circulating cell-free DNA (cfDNA) from plasma of peripheral whole blood collected in Streck Cell-Free DNA Blood Collection Tubes (BCTs). The test is intended to be used as a companion diagnostic to identify non-small cell lung cancer (NSCLC) patients who may benefit from treatment with the targeted therapy listed in Table 1 in accordance with the approved therapeutic product labeling. A negative result from a plasma specimen does not assure that the patient’s tumor is negative for genomic findings. NSCLC patients who are negative for the biomarkers listed in Table 1 should be reflexed to tissue biopsy testing for Table 1 biomarkers using an FDA-approved tumor tissue test, if feasible. Additionally, the test is intended to provide tumor mutation profiling to be used by qualified health care professionals in accordance with professional guidelines in oncology for cancer patients with solid malignant neoplasms. The test is for use with patients previously diagnosed with cancer and in conjunction with other laboratory and clinical findings. Genomic findings other than those listed in Table 1 are not prescriptive or conclusive for labeled use of any specific therapeutic product.

Device Story

Guardant360 CDx is a single-site NGS-based IVD for qualitative detection of genomic alterations in cfDNA from plasma. Input: 5-30 ng cfDNA from peripheral whole blood collected in Streck BCTs. Process: Library preparation with blunt-end ligation/inline barcodes; hybrid capture using biotinylated RNA probes; paired-end sequencing by synthesis on Illumina NextSeq 550. Bioinformatics pipeline performs molecular barcode-based reconstruction, alignment to hg19, and alteration calling. Output: Qualitative report of SNVs, indels, CNAs, and fusions. Used in clinical laboratories by trained personnel. Results assist oncologists in identifying NSCLC patients eligible for targeted therapy (TAGRISSO) and provide tumor mutation profiling for solid tumors. Benefits include non-invasive identification of actionable biomarkers, potentially avoiding tissue biopsy risks; negative results trigger reflex to tissue testing.

Clinical Evidence

Clinical evidence derived from two retrospective bridging studies using pre-treatment plasma from FLAURA (N=441) and AURA3 (N=300) trials. Primary endpoint: investigator-assessed PFS. FLAURA: HR 0.41 (95% CI 0.31, 0.54) for TAGRISSO vs SoC in Guardant360-positive patients. AURA3: HR 0.34 (95% CI 0.22, 0.53) for TAGRISSO vs chemotherapy in T790M-positive patients. Analytical validation included concordance with orthogonal NGS (PPA 95-100% for CDx variants), LoD/LoB, precision, and stability studies. No clinical data for T790M plasma-positive/tissue-negative population.

Technological Characteristics

NGS-based assay targeting 74 genes (reporting on 55). Uses hybrid capture with biotinylated RNA probes. Input: 5-30 ng cfDNA. Sequencing: Illumina NextSeq 550 (paired-end). Bioinformatics: Proprietary BIP software module for molecular barcode-based alteration calling. QC: Variant Control (cell line mixture) and NTC. Instruments: Agilent 4200 TapeStation, Hamilton Microlab STAR/STARlet, Veriti 96-Well Thermal Cycler. Single-site assay performed at Guardant Health, Inc.

Indications for Use

Indicated for NSCLC patients to identify eligibility for TAGRISSO (osimertinib) therapy based on EGFR exon 19 deletions, L858R, or T790M mutations. Also indicated for tumor mutation profiling in patients with solid malignant neoplasms previously diagnosed with cancer.

Regulatory Classification

Identification

A next generation sequencing (NGS) oncology panel is a device used for the qualitative detection of germline or somatic variants in one or more cancer-related genes. The device is intended to be used on DNA or RNA isolated from human clinical specimens.

Reference Devices

• cobas® EGFR Mutation Test v2 (Roche Molecular Systems, Inc.)

Submission Summary (Full Text)

{0}

SUMMARY OF SAFETY AND EFFECTIVENESS DATA (SSED)

I. GENERAL INFORMATION

Device Generic Name: Next Generation Sequencing Oncology Panel, Somatic or Germline Variant Detection System

Device Trade Name: Guardant360® CDx

Device Procode: PQP

Applicant's Name and Address: Guardant Health, Inc.
505 Penobscot Drive
Redwood City, CA 94063 USA

Date(s) of Panel Recommendation: None

Premarket Approval Application (PMA) Number: P200010

Date of FDA Notice of Approval:

Breakthrough Device: Granted breakthrough device status (formerly known as the Expedited Access Pathway, or EAP) on January 29, 2018 because the device (1) is intended to provide more effective diagnosis of a life threatening or irreversibly debilitating disease or condition (2) represents a breakthrough technology that provides a clinically meaningful advantage over existing legally marketed technology, and (3) the availability of the device is in the best interest of patients.

A De Novo (DEN200001) for Streck Cell-Free DNA Blood Collection Tubes (BCTs) was also submitted for the use of the Streck Cell-Free BCTs with the Guardant360 CDx. DEN200001 was authorized on August 7, 2020 in conjunction with the approval of P200010.

II. INDICATIONS FOR USE

Guardant360® CDx is a qualitative next generation sequencing-based in vitro diagnostic device that uses targeted high throughput hybridization-based capture technology for detection of single nucleotide variants (SNVs), insertions and deletions (indels) in 55 genes, copy number amplifications (CNAs) in two (2) genes, and fusions in four (4) genes. Guardant360 CDx utilizes circulating cell-free DNA (cfDNA) from plasma of peripheral whole blood collected in Streck Cell-Free DNA Blood Collection Tubes (BCTs). The test is intended to be used as a companion diagnostic to identify non-small cell lung cancer (NSCLC) patients who may benefit from treatment with the targeted therapy listed in Table 1 in accordance with the approved therapeutic product labeling.

PMA P200010: FDA Summary of Safety and Effectiveness Data

{1}

Table 1. Companion Diagnostic Indications

| Indication | Biomarker | Therapy |
| --- | --- | --- |
| Non-small cell lung cancer (NSCLC) | EGFR exon 19 deletions, L858R, and T790M* | TAGRISSO® (osimertinib) |

A negative result from a plasma specimen does not assure that the patient’s tumor is negative for genomic findings. NSCLC patients who are negative for the biomarkers listed in Table 1 should be reflexed to tissue biopsy testing for Table 1 biomarkers using an FDA-approved tumor tissue test, if feasible.

*The efficacy of TAGRISSO® (osimertinib) has not been established in the EGFR T790M plasma-positive, tissue-negative or unknown population and clinical data for T790M plasma-positive patients are limited; therefore, testing using plasma specimens is most appropriate for consideration in patients from whom a tumor biopsy cannot be obtained.

Additionally, the test is intended to provide tumor mutation profiling to be used by qualified health care professionals in accordance with professional guidelines in oncology for cancer patients with solid malignant neoplasms. The test is for use with patients previously diagnosed with cancer and in conjunction with other laboratory and clinical findings.

Genomic findings other than those listed in Table 1 are not prescriptive or conclusive for labeled use of any specific therapeutic product.

Guardant360 CDx is a single-site assay performed at Guardant Health, Inc.

## III. CONTRAINDICATIONS

There are no known contraindications.

## IV. WARNINGS AND PRECAUTIONS

Warnings and precautions are listed below:

- Alterations reported may include somatic (not inherited) or germline (inherited) alterations. The assay filters germline variants from reporting except for pathogenic BRCA1, BRCA2, ATM, and CDK12 alterations. However, if a reported alteration is suspected to be germline, confirmatory testing should be considered in the appropriate clinical context.
- The test is not intended to replace germline testing or to provide information about cancer predisposition.
- Somatic alterations in ATM and CDK12 are not reported by the test as they are excluded from the test's reportable range.

PMA P200010: FDA Summary of Safety and Effectiveness Data

{2}

- Genomic findings from cfDNA may originate from circulating tumor DNA (ctDNA) fragments, germline alterations, or non-tumor somatic alterations, such as clonal hematopoiesis of indeterminate potential (CHIP).
- Allow the tube to fill completely until blood stops flowing into the tube. Underfilling of tubes with less than 5 mL of blood (bottom of the label indicates 5 mL fill when tube is held vertically) may lead to incorrect analytical results or poor product performance. This tube has been designed to fill with 10 mL of blood.

## V. DEVICE DESCRIPTION

Guardant360 CDx is a single-site test performed at Guardant Health, Inc. The test includes reagents, software, and procedures for testing cfDNA from whole blood samples. The test uses 5-30 ng of cfDNA for library construction and next generation sequencing. Sequencing data is processed using a customized bioinformatics pipeline designed to detect several classes of genomic alterations, including nucleotide substitutions, indels, copy number amplifications, and genomic fusions / rearrangements. The device is designed to sequence 74 genes, but only report pre-defined and de novo alterations within the 55 genes outlined in Table 2. The test's reportable range for SNVs and indels covers approximately 46,000 bases.

Table 2. Genes Containing Alterations Detected by the Guardant360 CDx

| Alteration Type | Genes |
| --- | --- |
| Single Nucleotide Variants (SNVs) | AKT1, ALK, APC, AR, ARAF, ATM*, BRAF, BRCA1**, BRCA2**, CCND1, CDH1, CDK4, CDK6, CDK12*, CDKN2A, CTNNB1, EGFR, ERBB2, ESR1, FGFR1, FGFR2, FGFR3, GATA3, GNA11, GNAQ, HRAS, IDH1, IDH2, KIT, KRAS, MAP2K1, MAP2K2, MET, MLH1, MTOR, MYC, NF1, NFE2L2, NRAS, NTRK1, NTRK3, PDGFRA, PIK3CA, PTEN, RAF1, RET, RHEB, ROS1, SMAD4, SMO, STK11, TERT, TSC1, VHL |
| Indels | AKT1, ALK, APC, ATM*, BRAF, BRCA1**, BRCA2**, CDH1, CDK12*, CDKN2A, EGFR, ERBB2, ESR1, FGFR2, GATA3, HNF1A, HRAS, KIT, KRAS, MET, MLH1, NF1, PDGFRA, PIK3CA, PTEN, RET, ROS1, STK11, TSC1, VHL |
| Copy Number Amplifications (CNAs) | ERBB2, MET |
| Fusions | ALK, NTRK1, RET, ROS1 |

*Reporting is enabled for pathogenic germline alterations only. Somatic alterations will not be reported.
** Reporting is enabled for both germline and somatic alterations.

## Test Output

The test report includes variants reported in the following categories; see Table 3:

PMA P200010: FDA Summary of Safety and Effectiveness Data
3 of 75

{3}

Table 3. Category Definitions

| Category | Guardant360 CDx | | | Comments |
| --- | --- | --- | --- | --- |
| | Prescriptive use for a Therapeutic Product | Clinical Performance | Analytical Performance | |
| Category 1: Companion Diagnostic (CDx) | Yes | Yes | Yes | ctDNA biomarkers linked to the safe and effective use of the corresponding therapeutic product, for which Guardant360 CDx has demonstrated clinical performance shown to support therapeutic efficacy and strong analytical performance for the biomarker. |
| Category 2: ctDNA Biomarkers with Strong Evidence of Clinical Significance in ctDNA | No | No | Yes | ctDNA biomarkers with strong evidence of clinical significance presented by other FDA-approved liquid biopsy companion diagnostics for which Guardant360 CDx has demonstrated analytical reliability but not clinical performance. |
| Category 3A: Biomarkers with Evidence of Clinical Significance in tissue supported by: strong analytical validation using ctDNA | No | No | Yes | ctDNA biomarkers with evidence of clinical significance presented by tissue-based FDA-approved companion diagnostics or professional guidelines for which Guardant360 CDx has demonstrated analytical performance including analytical accuracy, and concordance of blood-based testing to tissue-based testing for the biomarker. |
| Category 3B: Biomarkers with Evidence of Clinical Significance in tissue supported by: analytical validation using ctDNA | No | No | Yes | ctDNA biomarkers with evidence of clinical significance presented by tissue-based FDA-approved companion diagnostics or professional guidelines for which Guardant360 CDx has demonstrated minimum analytical performance including analytical accuracy. |
| Category 4: Other Biomarkers with Potential Clinical Significance | No | No | Yes | ctDNA biomarkers with emergent evidence based on peer-reviewed publications for genes/variants in tissue, variant information from well-curated public databases, or in-vitro pre-clinical models, for which Guardant360 CDx has demonstrated minimum analytical performance. |

PMA P200010: FDA Summary of Safety and Effectiveness Data

{4}

PMA P200010: FDA Summary of Safety and Effectiveness Data
5 of 75

# Test Kit Contents

The test includes the Guardant360 CDx Blood Collection Kit (BCK), which is sent to ordering laboratories. Each BCK contains two blood collection tubes. The BCK also contains supporting packaging materials, instructions for use and a return shipping label. The BCK contains the following components:

- Streck blood collection tubes for specimen collection, stabilization, and transport of cfDNA; 2 per kit.
- Cushioning materials to prevent breakage of the blood collection tubes; 2 per kit
- Foam tray for protection of collection tubes during transport
- Absorbent sheet to be used during specimen shipping
- Biohazard specimen bag for protection during specimen transport
- Return shipping label for return of specimen to Guardant Health
- Barcodes for specimen identification and shipping instructions
- Instructions for Use for blood draw
- Patient welcome brochure which contains an overview of the test
- Test requisition form to complete to order Guardant360 CDx for a patient.

The test also includes the Guardant360 CDx Sample Preparation Kit (SPK), which is used in the Guardant Health Clinical Laboratory. The SPK contains reagents for library preparation, library enhancement, and cfDNA quantification/qualification. The kit is assembled into six (6) different boxes (referred to as box 1, 2, 3, 4a, 4b, and 4c) based on the usage of the reagents. The division of reagents amongst the boxes reflects different storage conditions and/or locations (e.g. different laboratory spaces).

# Instruments

Guardant360 CDx is intended to be performed with serial number-controlled instruments as indicated in Table 4. All instruments are qualified by Guardant Health, Inc. under the Guardant Health Quality System.

Table 4. Serial Number Controlled Instruments

| Instrument |
| --- |
| Agilent Technologies 4200 TapeStation Instrument |
| Hamilton Company Microlab STAR |
| Hamilton Company Microlab STARlet |
| Illumina NextSeq 550 Sequencer |
| Veriti 96-Well Thermal Cycler |

{5}

PMA P200010: FDA Summary of Safety and Effectiveness Data
6 of 75

# Test Process

## Whole Blood Collection and Shipping

The Guardant360 CDx Blood Collection Kit is used by ordering laboratories / physicians to collect whole blood specimens and ship them to the Guardant Health Clinical Laboratory. A minimum of 5 mL whole blood must be received in order to achieve optimal performance for the Guardant360 CDx assay. Underfilling of tubes with less than 5 mL of blood may lead to incorrect analytical results or poor product performance.

## Plasma Isolation and cfDNA Extraction

Whole blood specimens are processed in the Guardant Health Clinical Laboratory within 7 days of blood collection. Plasma is isolated from both tubes of whole blood via centrifugation. One tube of plasma is stored, while the second tube is used for cfDNA extraction using the QIAGEN QIAsymphony SP Instrument and reagent system. The resulting cfDNA is quantified using the 4200 TapeStation. Input amounts ranging from 5 to 30 ng of cfDNA are further processed for each sample.

## Library Preparation and Enrichment

Reagents from the Guardant360 CDx Sample Preparation Kit are used during library preparation, enrichment, enrichment wash, and quantitation steps using the Veriti 96-Well Thermal Cycler, Microlab STAR and STARlet, and 4200 TapeStation Instruments. During library preparation, cfDNA fragment ends are repaired and library adapters containing inline barcodes are attached using blunt-end ligation. The resulting DNA is amplified by PCR to create libraries suitable for enrichment.

Amplified libraries are enriched for genes of interest using hybrid target capture with custom biotinylated RNA probes. Each enriched library is amplified by PCR using a unique index primer that also contains a sequencing flow cell attachment sequence. Amplified enriched libraries are pooled in equimolar amounts, denatured, and diluted to appropriate concentration for sequencing.

## DNA Sequencing

Paired-end sequencing by synthesis is performed with the Illumina NextSeq 550 Sequencing system. The amplified cfDNA is analyzed by parallel sequencing of amplified target genes to an average depth of coverage of greater than 2,700 unique molecules.

## Data Analysis and Reporting

The Guardant360 CDx Software uses a custom-developed analysis bioinformatics pipeline (BIP) software module. The BIP software module uses the raw data (output) from the targeted sequencing, partitions the data based on the sample index sequence (barcode) of each read to separate reads originating from individual samples, and executes a proprietary algorithmic reconstruction of the digitized sequencing signals based on molecular barcodes for high-fidelity molecule-based alteration calling downstream. The sequence data then undergoes an alignment process where it is mapped to the human genome (hg19) and an analysis of sequence alteration data is performed.

{6}

Alteration detection is conducted according to alteration calling metrics derived from clinical sample analysis. All alterations must pass alteration calling metrics as described in Table 5.

The SNV and indel cut-offs are defined in terms of mutant allele fraction (MAF) estimate, number and type of molecules supporting the alteration, pseudo-gene assessment, and likelihood ratio (LLR) score. The MAF estimate describes the calculated allelic fraction of an SNV or indel. The number of molecules describes the observed number of molecules meeting requirements for a particular alteration call. The LLR score is a calculated number that reflects how much observed support for the mutation exceeds expectations based on PCR and sequencing induced artifacts.

Table 5. Alteration Analytical Calling Threshold/Cut-Off Metrics

| SNV Calling Property | Metric |
| --- | --- |
| DNA Molecule Support | ≥2 |
| MAF Estimate | ≥0.001% |
| Log likelihood ratio | ≥0 |
| Indel Calling Property | Metric |
| DNA molecule support | ≥2 |
| Log likelihood ratio | ≥10 |
| MAF Estimate | ≥0.01% |
| CNA Calling Property | Metric |
| ERBB2 copy number | ≥2.18 |
| ERBB2 Z-score | ≥10 |
| ERBB2 amplification is not associated with chromosome-arm aneuploidy | TRUE |
| MET copy number | ≥2.16 |
| MET Z-score | ≥10 |
| MET amplification is not associated with chromosome-arm aneuploidy | TRUE |
| Fusion Calling Property | Metric |
| MAPQ score of supporting molecule to fusion sequence | >30 |
| Number of unique fusion molecules | ≥2 |
| Number of unique fusion reads | >2 |

PMA P200010: FDA Summary of Safety and Effectiveness Data

{7}

The laboratory and physician receive a qualitative alteration-level result. A sample will receive an overall “Failed” result when any QC metric is failed. Samples failing any QC metric are automatically held and not released. The laboratory may attempt to rerun a patient sample that has failed a QC metric by using stored plasma or intermediate products.

Results from samples passing all QC metrics are formatted onto an IVD results report with CDx relevant information (Category 1) and all other biomarkers (Categories 2-4) within the LIMS system. The IVD results report will be populated with patient-specific information and may be merged with additional information provided by Guardant Health as a professional service prior to approval and release by the laboratory director or designee.

## Quality Control Measures

The Guardant360 CDx Sample Preparation Kit includes the Variant Control, which is engineered to contain known positive and negative alterations and is treated as a sample. Additionally, a no template negative control (NTC) is run in parallel with patient samples.

The Variant Control consists of a mixture of cfDNA from multiple human cancer cell lines containing all four alteration types, SNVs, indels, CNAs and fusions. The control is treated as a sample and processed starting from 15 ng cfDNA input through sequencing where it is analyzed for the presence and absence of the specific alterations.

Although the Variant Control does not contain all the alterations that the test is capable of detecting, concordant detection of alterations targeted in the Variant Control indicates that assay is performing as expected across the panel.

In addition to assessing Variant Control performance within a batch, the test is assessing multiple per-sample in-process and post-sequencing analytical metrics for each of the patient samples tested. These metrics provide in depth analytical QC information that complements Variant Control performance data and is specific and informative to that sample performance..

The NTC samples are absent of a DNA template, so cfDNA extraction, library preparation, and enrichment steps are expected to result in background level metrics.

## VI. ALTERNATIVE PRACTICES AND PROCEDURES

There is an FDA approved companion diagnostic (CDx) alternative for the detection of genetic alterations using cfDNA, as listed in Table 1 of the Guardant360 CDx intended use statement. The approved CDx test is detailed below: for additional details see FDA List of Cleared or Approved Companion Diagnostic Devices at https://www.fda.gov/MedicalDevices/ProductsandMedicalProcedures/InVitroDiagnostics/ucm301431.htm?source=govdelivery

PMA P200010: FDA Summary of Safety and Effectiveness Data
8 of 75

{8}

- cobas® EGFR Mutation Test v2 (Roche Molecular Systems, Inc.)
- Technology: polymerase chain reaction (PCR)
- Therapy: Tagrisso (osimertinib)
- Indication: Non-small cell lung cancer (NSCLC)

Each alternative has its own advantages and disadvantages. A patient should fully discuss these alternatives with his/her physician to select the method that best meets expectations and lifestyle.

## VII. MARKETING HISTORY

Guardant Health, Inc. initially designed and developed the Guardant360 laboratory developed test (Guardant360 LDT), and the first commercial sample was tested in 2012. Guardant360 LDT has been used to detect the presence of genomic alterations in plasma isolated from whole blood. The Guardant360 LDT is not FDA-cleared or -approved.

Guardant360 CDx has not been marketed in the United States or any foreign country.

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

Failure of the device to perform as expected or failure to correctly interpret test results may lead to incorrect test results, and subsequently, may lead to inappropriate patient management decisions. Patients with false positive results may undergo treatment with the therapy listed in the intended use statement without clinical benefit and may experience adverse reactions associated with the therapy. Patients with false negative results may not be considered for treatment with the indicated therapy. There is also a risk of delayed results, which may lead to delay of treatment with indicated therapy.

For the specific adverse events that occurred in the clinical studies, please see the TAGRISSO® (osimertinib) FDA approved package insert which is available at Drugs@FDA.

## IX. SUMMARY OF NONCLINICAL STUDIES

### A. Laboratory Studies

Guardant360 CDx performance characteristics were established using clinical samples from patients with a wide range of cancer types, including those with NSCLC. The clinical samples consisted of pools of cfDNA from clinical samples from multiple cancer types, pools of cfDNA from clinical samples derived from one cancer type (e.g., samples from patients with NSCLC) or un-pooled clinical samples. Studies included CDx variants, as well as a broad range of representative alteration types (SNVs, indels, CNAs, and fusions) in various genomic contexts across several genes in the assay's reportable range. Due to limitations in clinical sample availability and due to the rarity of the fusions reported by the Guardant360 CDx, contrived samples were utilized for some analytical validation studies. A contrived sample functional characterization

PMA P200010: FDA Summary of Safety and Effectiveness Data
9 of 75

{9}

(CSFC) study was conducted to demonstrate comparable performance of contrived samples made of cell line cfDNA and clinical sample cfDNA so that fusion cell line cfDNA material could be used in some analytical validation studies. Fusion positive clinical samples were used to confirm the estimated limit of detection, analytical accuracy and precision.

## 1. Analytical Accuracy/Concordance with an Orthogonal Method

The detection of alterations by Guardant360 CDx was compared to results of an externally validated NGS assay. Samples from 386 donors with different cancer types were collected for the study. Sixteen (16) samples failed testing with the comparator assay due to instrument failures, while eleven (11) samples failed testing with the Guardant360 CDx assay due to an instrument failure due to a power outage. 359 samples remained comprising three collection sets as follows.

Collection set one consisted of 100 donor samples selected with the comparator assay consecutively without selection for any specific variants. Since the first sample collection was expected to lack many rare variants, in the second collection set, a set of 100 positive samples were selected with the comparator assay. Collection set three consisted of 159 samples selected from the Guardant Health biobank based on Guardant360 LDT results to include additional rare variants including fusions which were not available from collection sets 1 and 2.

Of 359 patients, no samples failed QC on Guardant 360 CDx, and three samples failed QC with the comparator NGS assay. In total, 356 donor samples across 18 cancer types, which all passed every QC metric were used for the concordance analysis. The cancer types represented in this study included lung (178), gastrointestinal (82), colon (25), breast (17), head and neck (13), prostate (12), genitourinary (7), bladder (3), stomach (3), pancreas (3), endocrine (2), liver (2), ovarian (2), kidney (2), gynecologic (1), esophagus (1), skin (1), and other (5). A summary of Positive Percent Agreement (PPA) and Negative Percent Agreement (NPA) with 95% confidence intervals (CI) is provided in Table 6 for CDx alterations in samples from the intended use population, i.e., patients with NSCLC. Agreement rates for each of the CDx variants ranged from 95% to 100% for PPA, and from 98.1% to 99.9% for NPA. The reported PPA and NPA were not adjusted for the distribution of samples from collection set 3 selected using Guardant LDT results.

PMA P200010: FDA Summary of Safety and Effectiveness Data
10 of 75

{10}

Table 6. Summary of Concordance Between Guardant360 CDx and NGS Comparator for CDx variants

| Alteration Type | Guardant360 CDx(+), Comparator(-) | Guardant360 CDx(+), Comparator(-) | Guardant360 CDx(-), Comparator(-) | Guardant360 CDx(-), Comparator(-) | Possible Variants (n) | Patients (n) | PPA (95% CI) | NPA (95% CI) |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| EGFR T790M | 19 | 3 | 1 | 153 | 1 | 176 | 95.0% (75.1%, 99.9%) | 98.1% (94.5%, 99.6%) |
| EGFR L858R | 18 | 1 | 0 | 157 | 1 | 176 | 100.0% (81.5%, 100.0%) | 99.4% (96.5%, 100.0%) |
| EGFR exon 19 deletions | 30 | 1 | 1 | 1024 | 6 | 176 | 96.8% (83.3%, 99.9%) | 99.9% (99.5%, 99.9%) |

A summary of PPA and NPA for other clinically significant variant categories and for panel wide for SNVs and indels over all sample collections is provided in Table 7 and Table 8, respectively. The reported PPA and NPA were not adjusted for the distribution of samples from collection set 3 that were selected using Guardant LDT results.

Positive agreement rates were evaluable for nine (9) patients with clinical Category 2 variants, which consisted of clinically relevant PIK3CA mutations in breast cancer patients that included E545A, E542K, E545K, H1047R, and H1047L variants. Concordance analysis resulted in 100% PPA and 100% NPA for the Category 2 variants.

Positive agreement rates for clinical Categories 3 and 4 variants resulted in 93.5% PPA and 86.1% PPA, respectively. Variants in clinical category 3 and 4 showed 99.8% and 100.0% NPA.

MET amplifications had a PPA of 56%, which is attributed to differences in reporting of copy number amplifications by the Guardant360 CDx and the comparator assay. The Guardant360 CDx reports on only focal amplifications and not chromosome-arm amplifications, while the NGS comparator assay reports all amplifications.

PMA P200010: FDA Summary of Safety and Effectiveness Data
11 of 75

{11}

Table 7. Summary of Concordance Between Guardant360 CDx and NGS Comparator for other clinically significant variant categories.

| Alteration Type | Guardant360 CDx(+), Comparator(-) | Guardant360 CDx(+), Comparator(-) | Guardant360 CDx(-), Comparator(-) | Guardant360 CDx(-), Comparator(-) | Possible Variants (n) | Patients (n) | PPA (95% CI) | NPA (95% CI) |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Category 2 variants | 9 | 0 | 0 | 76 | 5 | 17 | 100.0 % (66.4%, 100.0%) | 100% (95.3%, 100.0%) |
| Category 3 variants | 115 | 11 | 8 | 6191 | 50 | N/A* | 93.5% (87.6%, 97.2%) | 99.8% (99.7%, 99.9%) |
| Category 4 variants | 420 | 58 | 68 | 137582 | 388 | 356 | 86.1% (82.7%, 89.0%) | 100.0% (99.9%, 100.0%) |
| MET CNAs | 13 | 3 | 10 | 330 | 1 | 356 | 56.5% (34.5%, 76.8%) | 99.1% (97.4%, 99.8%) |
| ERBB2 CNAs | 15 | 0 | 2 | 339 | 1 | 356 | 88.2% (63.6%, 98.5%) | 100.0% (98.9%, 100.0%) |
| NTRK1 Fusions | 5 | 0 | 0 | 351 | 1 | 356 | 100.0% (47.8%, 100.0%) | 100.0% (98.9%, 100.0%) |
| RET Fusions | 11 | 2 | 1 | 342 | 1 | 356 | 91.7% (61.5%, 99.8%) | 99.4% (97.9%, 99.9%) |
| ALK Fusions | 10 | 2 | 0 | 344 | 1 | 356 | 100.0% (69.2%, 100.0%) | 99.4% (97.9%, 99.9%) |
| ROS1 Fusions | 11 | 0 | 0 | 345 | 1 | 356 | 100.0% (71.5%, 100.0%) | 100.0% (98.9%, 100.0%) |

* For Category 3, no number is given for the number of patients. This is because Category 3 is a merge of many different variants, each with a specific set of cancer types that qualify the variant to belong in Category 3. This means that a different number of patients was associated with each variant within Category 3. For this level, the concordantly negative population was computed as the sum of the concordantly negative populations if each variant in this category was treated independently.

A summary of PPA and NPA for panel wide SNVs and indels over all sample collections is provided in Table 8. The study demonstrated a PPA of 82.5% for indels, 91.4% for SNVs and >99% NPA for the entire reportable range, demonstrating the analytical accuracy of the device.

PMA P200010: FDA Summary of Safety and Effectiveness Data
12 of 75

{12}

Table 8. Summary of Concordance Between Guardant360 CDx and NGS Comparator panel wide for SNVs and indels.

| Alteration Type | Guardant360 CDx(+), Comparator(-) | Guardant360 CDx(+), Comparator(-) | Guardant360 CDx(-), Comparator(-) | Guardant360 CDx(-), Comparator(-) | Possible Variants (n) | Patients (n) | PPA (95% CI) | NPA (95% CI) |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Panel-wide SNVs | 428 | 48 | 40 | 13,726,844 | 38,560 | 356 | 91.5% (88.5%, 93.8%) | 99.9% (99.9%, 99.9%) |
| Panel-wide Indels | 118 | 19 | 25 | 15,717,238 | 44,150 | 356 | 82.5% (75.3%, 88.4%) | 99.9% (99.9%, 99.9%) |

2. **Contrived Sample Functional Characterization (CSFC) Study**

A CSFC study was performed to demonstrate comparable performance between contrived samples that consisted of fusion cell line cfDNA material and fusion positive clinical sample cfDNA material. The CSFC study was performed using 5 ng DNA input (the lowest cfDNA input for the assay) to compare the performance of the Guardant360 CDx with cfDNA derived from cell lines and cfDNA derived from multiple clinical samples from multiple cancer types with ALK, NTRK1, RET, and ROS1 fusions. The cell line and clinical cfDNA sample pools contained known fusion events that were diluted with pools of wild-type (WT) cfDNA from multiple clinical specimens from multiple cancer types to pre-determined MAF levels (targeted levels were above and below LoD; see Table 9). Cell line cfDNA sample pools were tested across 13-20 replicates, 13 replicates for level 6, 14 replicates for level 2, and 20 replicates for the other levels at 5 ng cfDNA input. Clinical cfDNA sample pools from multiple cancer types were tested with 14 replicates at 5 ng cfDNA input. Both cell line and clinical cfDNA sample pools were tested with the orthogonal method to confirm MAF levels. Detection rates of the 4 fusions, for each titration level, and for each of the two types of pools, are presented in Table 9.

Based on these analyses, the results (Table 9) demonstrate that the performance of the Guardant360 CDx is similar for both fusion positive contrived cfDNA samples and for fusion positive clinical cfDNA samples.

PMA P200010: FDA Summary of Safety and Effectiveness Data

{13}

Table 9. Fusion Detection Rate in the CSFC study

| Fusion | Sample Type | Detection Rate (95% confidence interval) | | | | | |
| --- | --- | --- | --- | --- | --- | --- | --- |
| | | Level 1 Target MAF 0.07% | Level 2 Target MAF 0.175% | Level 3 Target MAF 0.35% | Level 4 Target MAF 0.7% | Level 5 Target MAF 1.4% | Level 6 Target MAF 1.8% |
| EML4-ALK | Cell line | 5.0% (0.1%, 24.9%) | 28.6% (8.4%, 58.1%) | 50.0% (27.2%, 72.8%) | 90.0% (68.3%, 98.8%) | 100.0% (83.2%, 100.0%) | 100.0% (75.3%, 100.0%) |
| EML4-ALK | Clinical | 7.1% (0.2%, 33.9%) | 28.6% (8.4%, 58.1%) | 50.0% (23.0%, 77.0%) | 85.7% (57.2%, 98.2%) | 100.0% (76.8%, 100.0%) | 100.0% (76.8%, 100,0%) |
| CCDC6-RET | Cell line | 15.0% (3.2%, 37.9%) | 35.7% (12.8%, 64.9%) | 80.0% (56.3%, 94.3%) | 95.0% (75.1%, 99.9%) | 100.0% (83.2%, 100%) | 100.0% (75.3%, 100%) |
| TRIM33-RET | Clinical | 7.1% (0.2%, 33.9%) | 14.3% (1.8%, 42.8%) | 64.3% (35.1%, 87.2%) | 85.7% (57.2%, 98.2%) | 100.0% (76.8%, 100%) | 100.0% (76.8%, 100%) |
| ROS1-SLC34A2 | Cell line | 0.0% (0.0%, 16.8%) | 21.4% (4.7%, 50.8%) | 50.0% (27.2%, 72.8%) | 75.0% (50.9%, 91.3%) | 100.0% (83.2%, 100.0%) | 100.0% (75.3%, 100,0%) |
| ROS1-CD74 | Clinical | 7.1% (0.2%, 33.9%) | 42.9% (17.7%, 71.1%) | 85.7% (57.2%, 98.2%) | 100.0% (76.8%, 100%) | 100.0% (83.9%, 100.0%) | ND |
| TPM3-NTRK1 | Cell line | 15.0% (3.2%, 37.9%) | 50.0% (23.0%, 77.0%) | 40.0% (19.1%, 63.9%) | 90.0% (68.3%, 98.8%) | 100.0% (83.2%, 100.0%) | 100.0% (75.3%, 100.0%) |
| PLEKHA 6-NTRK1 | Clinical | 21.4% (4.7%, 50.8%) | 35.7% (12.8%, 64.9%) | 85.7% (57.2%, 98.2%) | 100.0% (76.8%, 100.0%) | ND | 100.0% (76.8%, 100.0%) |

ND: Not Determined

## 3. Analytical Sensitivity

### a. Limit of Blank (LoB)

The LoB was established by evaluating whole blood samples from age-matched healthy donors. Sixty-two (62) donor samples confirmed to be mutation negative based on sequencing with an externally validated orthogonal method were processed using 30 ng of cfDNA input with the Guardant360 CDx (highest DNA input for the assay) across three lots of reagents, operator groups, and instruments. Of the 62 donor samples, 58 donor samples were tested with 4 replicates, while 4 donors were tested with 2 replicates for a total of 240 replicates analyzed to assess the false positive rate of Guardant360 CDx.

PMA P200010: FDA Summary of Safety and Effectiveness Data

{14}

This study demonstrated a near zero false positive rate across the entire assay reportable range, as shown in Table 10. The false positive rate was zero for Category 1 (CDx) and Category 2 variants.

Table 10. LoB Study Summary Results
| Category | Per Position False Positive Rate | Per Sample False Positive Rate |
| --- | --- | --- |
| Category 1: EGFR L858R | 0% | 0 (0/240) |
| Category 1: EGFR T790M | 0% | 0 (0/240) |
| Category 1: EGFR Exon 19 deletions | 0% | 0 (0/240) |
| Category 2 | 0% | 0 (0/240) |
| Panel-wide SNVs (38,560 bp) | <0.00005%
(4/(38,560*240)) | 1.67% (4/240) |
| Panel-wide Indels (44,150 bp) | <0.00002%
(2/(44,150*240)) | 0.83% (2/240) |
| Panel-wide CNAs (2 genes) | 0.2% (1/(2*240)) | 0.42% (1/240) |
| Panel-wide Fusions (4 genes) | 0% | 0 (0/240) |

## b. Limit of Detection (LoD)

The LoD for the Guardant360 CDx variants with CDx claims, representative SNVs and indels, and all reportable CNAs and fusions was established at the lowest and highest claimed cfDNA input amounts (5 and 30ng). LoD established for fusions using cfDNA derived from cell lines was confirmed at 5ng cfDNA input using cfDNA derived from clinical patient samples. LoDs were further confirmed in the clinical pools of relevant cancer types for CDx variants and additional representative variants, including long indels and homopolymers in a combined LoD confirmation and precision study.

For SNVs, indels, including CDx variants and for CNAs, the Guardant360 CDx LoD was established by combining cfDNA from clinical plasma samples from multiple cancers to create pools of material comprising multiple known alterations. The LoD was established with these clinical cfDNA sample pools at 5 ng and 30 ng input, using a combination of probit and empirical approaches. Samples were titrated at 5 different MAF values that included levels above and below the LoD for SNVs, and indels or copy numbers values for CNAs and tested across 20 replicates for 5 ng input and 14 replicates for 30 ng input across two reagent lots.

The LoDs of three (3) CDx alterations representing EGFR T790M, EGFR L858R, and EGFR exon 19 deletions established using pools of cfDNA from clinical plasma samples from multiple cancer types are summarized in Table 11. The LoD was

PMA P200010: FDA Summary of Safety and Effectiveness Data
15 of 75

{15}

confirmed for CDx variants using cfDNA sample pools from patients with NSCLC only; refer to Table 13 below.

Table 11. Summary of Established LoD for Alterations Associated with CDx Claims using pools of cfDNA from Clinical Plasma Samples from Multiple Cancer Types

| Alteration | Alteration Type | LoD, 5ng input (MAF) | LoD, 30 ng input (MAF) |
| --- | --- | --- | --- |
| EGFR T790M | SNV | 1.1 % | 0.2% |
| EGFR L858R | SNV | 1.0 % | 0.2% |
| EGFR exon 19 deletion, p.Glu746_Ala750del | Indel (15 bp) | 1.5% | 0.2% |

The LoD estimates for SNV, indels, and CNA alterations established using pools of cfDNA from clinical plasma samples from multiple cancer types are summarized in Table 12.

For fusions, the Guardant360 CDx LoD was established using cfDNA from cell lines with known fusions titrated into wild-type (WT) cfDNA from clinical plasma samples. Samples were titrated at 5 different MAF values for fusions across 20 replicates for 5 ng cfDNA input and 14 replicates for 30 ng cfDNA input across two reagent lots. The established LoD was then confirmed using fusion positive cfDNA from clinical plasma samples at 5 ng cfDNA input only. Fusion positive cfDNA from clinical samples were titrated across 5 concentrations with 14 replicates across 2 reagent lots. The higher of the LoD values established using cell lines and confirmed using clinical samples were used to claim the LoD performance levels of the test for fusions at 5 ng (Table 12).

Table 12. LoD Establishment Study Summary Results for Representative Variants using Pools of cfDNA Clinical Plasma Samples from Multiple Cancer Types

| Alteration | Alteration Type | LoD, 5 ng (MAF/CN) | LoD, 30 ng (MAF/CN) |
| --- | --- | --- | --- |
| BRAF V600E | SNV | 1.8% | 0.2% |
| KRAS G12V | SNV | 1.5% | 0.5% |
| NRAS Q61R | SNV | 3.0% | 0.8% |
| BRCA1 p.E23fs | Indel (2 bp) | 2.6% | 0.8% |
| BRCA2 p.S1982fs | Indel (1 bp) | 1.3% | 0.4% |
| EGFR exon 20 insertion, p.Ala767_Val769dup | Indel (9 bp) | 0.8% | 0.2% |

PMA P200010: FDA Summary of Safety and Effectiveness Data

{16}

| Alteration | Alteration Type | LoD, 5 ng (MAF/CN) | LoD, 30 ng (MAF/CN) |
| --- | --- | --- | --- |
| ERBB2 exon 20 insertion, p.A775_G776insYVMA | Indel (12 bp) | 1.1% | 0.2% |
| MET | CNA | 2.4 | 2.4 |
| ERBB2 | CNA | 2.3 | 2.3 |
| NTRK1 | Fusion | 0.9% (0.9%) | (0.2%) |
| RET | Fusion | 1.1% (0.7%) | (0.1%) |
| ROS1 | Fusion | 1.9% (1.2%) | (0.2%) |
| ALK | Fusion | 1.4% (1.5%) | (0.2%) |

The numbers in parentheses represent LoD established using cell line derived cfDNA. MAF: Mutant Allele Fraction. CN: copy number.

The established LoD was confirmed for CDx variants by testing clinical patient pools exclusively from NSCLC patients targeting 1-1.5xLoD of the established LoD (refer to Table 11) across at least 20 replicates at 5 ng input using a combined LoD Confirmation and Precision Study. Similarly, the established LoD was confirmed for SNVs and indels in clinical pools made exclusively from the relevant cancer type source material prepared with 5 ng cfDNA input targeting 1-1.5x LoD and run in at least 20 replicates targeting 5 distinct variants. Established LoD targets were used for 5 variants (EGFR L858R, EGFR T790M, EGFR exon 19 deletion, p.E746_A750del, KRAS G12C, and ROS1 fusions), while in silico LoD targets were used for 10 additional variants to target variants to 1-1.5x LoD.

In this combined LoD and Precision study (see Section IX.A.5. below for additional studies demonstrating assay precision starting from cfDNA extraction, and with additional mutation positive and negative samples) samples were tested across three precision combinations that evaluated three operator groups, three instrument combinations, and three SPK reagent lots over at least three different start dates. The higher of the LoD values established using clinical sample pools from cancer patients and confirmed using clinical samples exclusively from the relevant cancer type source material were used to claim LoD performance of the test at 5 ng input as summarized in Table 13.

PMA P200010: FDA Summary of Safety and Effectiveness Data

{17}

Table 13. Combined LoD Confirmation and Precision Study Summary Results for CDx Variants and Representative Variants

| Alteration | MAF | Alteration Type | Cancer Type | Number Positive / Number Expected | PPA |
| --- | --- | --- | --- | --- | --- |
| EGFR L858R | 1.5%* | SNV | NSCLC | 20/20 | 100.0% |
| EGFR T790M | 1.4%* | SNV | NSCLC | 19/20 | 95.0% |
| EGFR exon 19 deletion, p.E746_A750del | 1.5%* | Indel (15bp) | NSCLC | 20/20 | 100.0% |
| EGFR exon 19 deletion, A750_I759delinsPT | 2.3%^ | Indel (29 bp) | NSCLC | 20/20 | 100.0% |
| KIT V654A | 2.5%^ | SNV | Prostate | 20/20 | 100.0% |
| KRAS G12C | 1.8%* | SNV | NSCLC | 19/20 | 95.0% |
| PIK3CA E545K | 2.4%^ | SNV | Breast | 21/21 | 100.0% |
| PIK3CA H1047L | 1.7%^ | SNV | Breast | 21/21 | 100.0% |
| EGFR exon 20 insertion, H773_V774insHPH | 3.5%^ | Indel (9 bp) | NSCLC | 22/22 | 100.0% |
| MET exon 14 skipping 7.116412041.AAGGTATATT TCAGTT>A | 2.7%^ | Indel (15 bp) | NSCLC | 20/20 | 100.0% |
| BRCA2 p.T3033fs | 4.4%^ | Indel (1 bp), homopolymer | NSCLC | 21/21 | 100.0% |
| BRCA2 p.I605fs | 5.0%^ | Indel (1 bp), homopolymer | Prostate | 20/20 | 100.0% |
| BRCA2 p.V1532fs | 4.2%^ | Indel (1 bp), homopolymer | Prostate | 20/20 | 100.0% |
| STK11 p.L282fs | 4.7%^ | Indel (1 bp), homopolymer | NSCLC | 21/21 | 100.0% |
| ROS1 | 1.8%* | Fusion | NSCLC | 21/21 | 100.0% |

* Observed MAF level in LoD Confirmation Study. LoD confirmed with single cancer type clinical pool and ≥95% detection rate is within 1-1.5x LoD MAF level from the original establishment study range.
^ Observed MAF at the level tested with ≥95% detection rate for variants without direct prior LoD establishment data.

PMA P200010: FDA Summary of Safety and Effectiveness Data

{18}

Panel-wide SNV and indels detected by Guardant360 CDx is summarized in Table 14 as median values.

Table 14. Summary of LoD for Alterations Associated with Panel-Wide Claims
| Alteration | Median LoD, 5ng (MAF) | Median LoD, 30ng (MAF) |
| --- | --- | --- |
| Panel-wide SNVs | 1.8% | 0.2% |
| Panel-wide Indels | 2.7% | 0.2% |

## 4. Analytical Specificity

### a. Endogenous and Microbial Interfering Substances

To evaluate the potential impact of endogenous and microbial interfering substances on the performance of Guardant360 CDx, this study evaluated whole blood samples representing 13 cancer types. Whole blood was collected from at least 10 advanced cancer patients per condition. A total of 4 BCTs per patient were collected. Whole blood from a single BCT from each donor was spiked with an interfering substance at a pre-defined concentration.

A total of 146 samples (73 reference-condition pairs, 12 sample pairs in albumin condition, 11 sample pairs in bilirubin (conjugated) condition, 11 sample pairs in bilirubin (unconjugated) condition, 11 sample pairs in hemoglobin condition, 11 sample pairs in Staphylococcus epidermidis condition and 17 sample pairs in triglyceride condition) were tested to analyze the effect of 6 interfering substances on assay performance. After QC checks, a total of 130 samples (65 reference-condition pairs, 10 sample pairs in albumin condition, 11 sample pairs in bilirubin (conjugated) condition, 11 sample pairs in bilirubin (unconjugated) condition, 10 sample pairs in hemoglobin condition, 11 sample pairs in Staphylococcus epidermidis condition and 12 sample pairs in triglyceride condition) were eligible for analysis.

Substances were considered as non-interfering if, when compared to no interferent controls, the sample level molecule recovery, exon-level molecule recovery, and variant call concordance met pre-defined acceptance thresholds.

Sample level molecule recovery was determined by the depth of non-singleton molecule (NSC) coverage across the panel. Median non-singleton molecule coverage across targeted regions was evaluated to demonstrate that microbial or interfering substances do not impact assay performance to sequence unique molecules. Recovery of unique molecules across interfering substance conditions did not show a negative impact of interfering substances (fold change of median NSC in spike condition over reference condition ranged from 0.88 to 1.08).

Relative exon coverage calculated as the ratio of median exon coverage to sample level coverage for each of the 508 exon regions was compared for each condition-reference sample pair. Aggregating across all samples contributing to the analysis, the

PMA P200010: FDA Summary of Safety and Effectiveness Data
19 of 75

{19}

total fraction of all exonic regions within expected level of differences defined as $2^{*}$ $\sigma$ , where $\sigma$ is the pooled standard deviation of the differences observed in historical ( $\sigma = 0.108$ ) were calculated. Under normal distribution assumption, the fraction of such regions is expected to be $95\%$ . The fraction of exons with relative exon level coverage difference between condition and reference within $2\sigma$ ( $2^{*}0.108$ ) was $94.3 - 99.7\%$ , which demonstrate that there was no preferential drop-out of relative exon-level coverage exceeding expected levels due to random variation, and the entire panel was covered consistently between reference and interfering substance conditions. The lower bound of the $95\%$ exact binomial CI for the fraction of genomic targeted exonic regions with relative exon-level non-singleton coverage is shown in Table 15.

The results were aggregated across all variants across all ten whole blood samples, and concordance was assessed within each treatment category across variants. PPAs were calculated for 62 SNVs, 24 indels, and 3 CNAs. The 6 conditions tested showed variant call concordant PPAs ranging from $83.3\% -100.0\%$ . PPA $\geq 1x$ LoD ranged from $90.0\% -100.0\%$ for all 6 interferents, Table 15.

The panel-wide NPAs were also calculated for SNVs and indels within the reportable range. The discordant negative variants were defined as those negative variants that were positive in the non-reference condition. The panel-wide NPA was $99.9\% - 100.0\%$ for all conditions (Table 15).

In conclusion no interference was found in albumin (60 g/L), bilirubin (conjugated) (342 μmol/L), bilirubin (unconjugated) (342 μmol/L), hemoglobin (2 g/L), Staphylococcus epidermidis (106 cfu) and triglycerides (15 g/L).

Table 15. Summary of Interference Results

| Condition | PPA | ≥1x LoD PPA | NPA | Sample Level Molecule Recovery Median of NSC fold (condition / reference) | Relative Exon Level Coverage Lower bound of 95% CI |
| --- | --- | --- | --- | --- | --- |
| Albumin (60 g/L) | 100.0% (13/13) | 100.0% (9/9) | 100.0% (462217/462217) | 0.96 | 99.5% |
| Bilirubin (conjugated) (342 μmol/L) | 90.9% (10/11) | 100.0% (9/9) | 99.9% (508440/508442) | 0.98 | 93.8% |
| Bilirubin (unconjugated) (342 μmol/L) | 84.6% (11/13) | 100.0% (8/8) | 99.9% (508439/5084440) | 0.88 | 96.0% |
| Hemoglobin (2 g/L) | 83.3% (10/12) | 100.0% (7/7) | 100.0% (462218/462218) | 1.08 | 96.1% |

PMA P200010: FDA Summary of Safety and Effectiveness Data

{20}

| Condition | PPA | ≥ 1x LoD PPA | NPA | Sample Level Molecule Recovery Median of NSC fold (condition / reference) | Relative Exon Level Coverage Lower bound of 95% CI |
| --- | --- | --- | --- | --- | --- |
| Staphylococcus epidermidis (106 cfu) | 90.9% (10/11) | 90.0% (9/10) | 100.0% (508442/508442) | 1.03 | 98.9% |
| Triglyceride (15 g/L) | 89.7% (26/29) | 100.0% (25/25) | 99.9% (554644/554647) | 0.96 | 93.7% |

The effect of potential exogenous interfering substances that may carry over from cfDNA extraction on assay performance was not evaluated. A post-market study will be conducted to evaluate the effect of exogenous interfering substances on assay performance. See Section XIII.

## b. In silico Specificity Analysis

An in silico specificity study was conducted to evaluate the potential for Guardant360 CDx test probes to amplify non-specific products from human DNA and to assess the potential for incorrect results due to commensal microorganism contamination. For this in silico study, the probe sequences were mapped against a reference genome that included unplaced contigs and decoy sequences designed to capture known problematic reads in human genome sequence.

In addition, a modified version of the human genome was generated by adding genomic contigs from common bacterial, fungal, and viral sources that can be found in the blood or on the skin. Finally, sequencing primer sequences were checked for specificity using NCBI's primer BLAST tool.

When mapped to the human genome with decoy sequences, unplaced contigs, and representative microbial contaminant genomes, 97.6% of the probes uniquely mapped to the intended targets (MAPQ ≥ 60). None of the probes mapped to the representative microbial contaminant genomes. No off-target amplicons &lt;1000 bp were predicted for the primers using the web-based NCBI Primer-BLAST tool against a database of Ref-Seq representative genomes.

Based on the results, all of the panel coverage results satisfy the desired performance. There is minimal risk of Guardant360 CDx producing a false positive patient result due to detection of microbial sequences or mis-alignment of off-target sequences in the human genome.

## 5. Precision

The purpose of the precision studies was to demonstrate the repeatability and within-site reproducibility of Guardant360 CDx through closeness of agreement between

PMA P200010: FDA Summary of Safety and Effectiveness Data
21 of 75

{21}

measured qualitative output obtained in replicate testing using different combinations of reagent lots, instruments, operators, and days. Additional runs were conducted (1) on mutation negative samples to demonstrate precision of analytically blank samples, and (2) on plasma samples to understand the influence of extraction on test. All studies were conducted exclusively with patient material and no cell line material was used.

## a. Precision from three distinct cfDNA clinical sample pools

Precision was evaluated for alterations associated with CDx claims, as well as representative and specific alterations to support platform-level performance. Repeatability including intra-run performance (run on the same plate under the same conditions) and reproducibility including inter-run performance (run on different plates under different conditions) were assessed and compared across three different precision combinations (PC) of instrument sets, reagent lots, and operators over multiple days. Three distinct precision clinical cfDNA sample pools from multiple cancer types, containing a total of 16 targeted alterations across all pools were prepared targeting 1-1.5 x LoD, and tested at 5ng cfDNA input. Ten (10) replicates per three (3) pools were tested for each of three (3) precision combinations (90 replicate samples total) and comprised of three (3) different reagent lots (G360 SPK, Ampure XP beads, and NextSeq 550 sequencing, reagent lots), three (3) different instrument sets and three (3) different operator groups. Each combination was tested on two (2) batches, sequenced on four (4) flow cells. The QIAsymphony instrument was not paired within each of the three (3) precision combination sets, since the sample pools were generated from previously extracted and stored cfDNA. Precision starting from cfDNA extraction was evaluated in a separate study described in Section IX.A.5.b. In total, 480 alterations were assessed across 90 samples tested. Qualitative results were used to calculate PPA and NPA.

The final levels for the targeted variants tested ranged from 0.7x to 2.6x LoD. Three variants were below 1x LoD (ROS1 fusion at 0.9x LoD, MET amplification at 0.8x LoD, and NRAS Q61R at 0.7x LoD), 8 were within 1-1.5x range, including the CDx variants, and 5 variants were in the 1.7x - 2.6x LoD range.

Across 960 expected negative targeted sites (32 targeted negative variants across 3 sample pools * 30 replicates), the observed NPA was 100.0%.

All CDx alterations demonstrated acceptable precision (PPA 96.7%-100.0%), Table 16.

Table 16. Precision CDx Variant PPA Summary

| Alteration | Number Positive / Number Expected | PPA (95% CI) |
| --- | --- | --- |
| EGFR T790M | 30/30 | 100.0% (88.4%, 100%) |
| EGFR L858R | 30/30 | 100.0% (88.4%, 100%) |

PMA P200010: FDA Summary of Safety and Effectiveness Data
22 of 75

{22}

The variant level PPA for all targeted variants were above 90.0% across all instrument, reagent, and operator combinations, except for MET amplification in pool 1, which may be attributed to the 0.8x LoD range achieved in the titration pool (Table 17). ROS1 fusion detection demonstrated 93.3% PPA, consistent with the achieved 0.9x LoD titration level. BRCA1 E23fs also resulted in a lower variant level PPA (90.0%) than expected. However, the 90.0% detection rate is consistent with the variant being located in a more challenging area of the panel with respect to coverage. Specifically the variant is considered to be in a more challenging area because it is in a region with relatively low GC content and has below average DNA molecule recovery.

In summary, across 480 alterations (150 SNVs, 150 indels, 60 CNAs, and 120 fusions) in a set of 90 cfDNA sample replicates containing 16 unique alterations across 3 cfDNA sample pools made from cfDNA from multiple cancer types, all alterations demonstrated PPA of 86.7%-100.0%. The PPA across all targeted alterations for each condition was evaluated. The PPA across all targeted alterations per precision combination (PC) ranged from 96.3%-99.4%.

Table 17. Variant Level PPA for Four Variant Classes

| Alteration Class | Alteration | Number Positive / Number Expected | PPA (95% CI) |
| --- | --- | --- | --- |
| SNV | KRAS G12V | 30/30 | 100.0% (88.4%, 100%) |
| SNV | NRAS Q61R | 30/30 | 100.0% (88.4%, 100%) |
| SNV | BRAF V600E | 30/30 | 100.0% (88.4%, 100%) |
| SNV | Panel-wide | 150/150 | 100.0% (97.6%, 100%) |
| Indel | ERBB2 A775_G776insYVMA | 30/30 | 100.0% (88.4%, 100%) |
| Indel | EGFR A767_V769dup | 30/30 | 100.0% (88.4%, 100%) |
| Indel | BRCA1 E23fs | 27/30 | 90.0% (73.5%, 97.9%) |
| Indel | BRCA2 S1982fs | 30/30 | 100.0% (88.4%, 100%) |
| Indel | Panel-wide | 146/150 | 97.3% (93.3%, 99.3%) |
| CNA | ERBB2 | 30/30 | 100.0% (88.4%, 100%) |
| CNA | MET | 26/30 | 86.7% (69.3%, 96.2%) |
| Fusion | EML4-ALK | 30/30 | 100.0% (88.4%, 100%) |
| Fusion | TPM3-NTRK1 | 30/30 | 100.0% (88.4%, 100%) |
| Fusion | TRIM33-RET | 30/30 | 100.0% (88.4%, 100%) |
| Fusion | ROS1-CCDC6 | 28/30 | 93.3% (77.9%, 99.2%) |

PMA P200010: FDA Summary of Safety and Effectiveness Data

{23}

Precision from clinical pools with samples from a single clinically relevant cancer type was confirmed in the combined LoD confirmation and precision study described in Section IX.A.3.b above.

## b. Precision from plasma evaluation of extraction precision and precision of downstream steps

The purpose of this study was to show the precision of variant calling for the entire sample workflow (from cfDNA extraction through sequencing) with un-pooled clinical samples.

This study utilized clinical plasma samples from 53 unique patients. Each plasma sample with positive variants (as detected by Guardant360 LDT) and high cfDNA yields was split into six aliquots or six replicates per patient.

The LoD was established for inputs of 5 ng and 30 ng, which are the lower and upper limit of cfDNA mass input for library preparation. Since the purpose of this precision study was to test the full spectrum of sample yields that would be observed in normal use, sample inputs ranged from 5 ng to 30 ng of cfDNA input. The corresponding LoD range was between 1x for the 30 ng LoD MAFs, and 1.5x for the 5 ng LoD MAFs. Variants that were previously observed in this MAF range in the Guardant360 LDT run were selected for this study and evaluated for call agreement.

Eighteen (18) different tumor types were evaluated in this study to support a pan-cancer tumor profiling indication for Guardant360 CDx (see Section IX.A.10. for more details on pan-cancer claim). Each donor specimen was processed in duplicate across three lots for a total of 6 replicates. "Lot" refers to different reagent lots, as well as different combinations of operators, days, and instruments to evaluate precision. The targeted variants evaluated in the study are shown in Table 18.

Table 18. Targeted Variants Amongst the 53 Donor Samples Selected for Study

| Category | Variant | Number of Eligible Based on MAF/CN |
| --- | --- | --- |
| ERBB2 | CNA | 3 |
| MET | CNA | 3 |
| ALK | fusion | 2 |
| RET | fusion | 2 |
| EGFR exon 19 deletion | indel | 6 |
| EGFR exon 20 insertion | indel | 2 |
| Long indel (>30 bp) | indel | 1 |
| MET exon 14 skipping | indel | 1 |
| BRAF V600E | SNV | 3 |
| EGFR L858R | SNV | 6 |
| EGFR T790M | SNV | 4 |
| KRAS G12C | SNV | 3 |
| PIK3CA E542K | SNV | 3 |

PMA P200010: FDA Summary of Safety and Effectiveness Data
24 of 75

{24}

| PIK3CA E545K | SNV | 4 |
| --- | --- | --- |
| PIK3CA H1047L/R | SNV | 2 |
| PIK3CA C420R | SNV | 3 |

A total of 315 replicates passed QC and were analyzed for within-condition and between-condition precision.

For each eligible variant, pairwise comparisons of variant detection were made between the technical replicates in each lot. From the study design with three lots and two replicates within each lot, there were 3 pairs for each variant in calculating within-lot average positive agreement (APA) and 12 pairs for each variant in calculating between-lot APA.

The APA results for eligible SNVs, indels, fusions, CNAs and all three together are shown in Table 19. Workflow or sample QC failures mean there were fewer than 3 lots per variant tested in some cases. The within lot APA for all variant types together was 97.3% as shown in Table 19.

Table 19. Within-Reagent Lot APA Summary

| Variant Type | Variant Lot Comparisons | Concordant (C) | Discordant (D) | APA |
| --- | --- | --- | --- | --- |
| SNV | 150 | 141 | 9 | 96.9% |
| Indel | 35 | 35 | 0 | 100.0% |
| CNA | 15 | 13 | 2 | 92.9% |
| Fusion | 12 | 12 | 0 | 100.0% |
| ALL | 212 | 201 | 11 | 97.3% |

The within-lot average negative agreement (ANA) was 99.9%. This statistic includes all called variant sites panel-wide, not just the eligible variant sites based on LoD in the source samples, so this statistic includes positions with expected stochastic detection due to low mutant molecule count. The number of positions evaluated was 46,217 unique SNV and indel reportable positions, 2 CNAs, and 4 fusions.

The between lot APA for eligible SNVs, indels, fusions, CNAs, and all reportable variants together are shown in Table 20. For each of these variants, there were 12 pairwise comparisons.

Table 20. Between-Lot APA Summary

| Variant Type | Variant Lot Comparisons | Concordant | Discordant | APA |
| --- | --- | --- | --- | --- |
| SNV | 47 | 531 | 26 | 97.6% |
| indel | 11 | 132 | 0 | 100.0% |
| CNA | 8 | 53 | 6 | 94.6% |

PMA P200010: FDA Summary of Safety and Effectiveness Data

{25}

| fusion | 4 | 48 | 0 | 100.0% |
| --- | --- | --- | --- | --- |
| ALL | 70 | 764 | 32 | 98.0% |

The between-lot APA for all variant types together was 98.0%. The between lot ANA was 99.9% across all reportable positions and variants. This statistic includes all called variant sites, not just the eligible variants sites based on LoD in the source samples, so includes positions with expected stochastic detection due to low mutant molecule count. The number of positions evaluated was 46,217 unique SNV and indel reportable positions, 2 CNAs, and 4 fusions.

Notably, for ERBB2 amplifications, within and between-lot APA were observed to be 80.0% and 85.0%, respectively, due to variation in focality determination. Specifically some of the replicates were determined to be focally amplified, and thus reported by the assay, and some were determined to be aneuploid and thus reported negative as the Guardant360 CDx reports CNAs only for focal amplifications and not chromosome-arm amplifications.

In addition to the main study, supplementary samples, starting from plasma, were processed to evaluate precision from extraction. Fusion samples were created by diluting cfDNA extracted from cell lines harboring ROS1 and NTRK1 fusions into plasma of clinical lung cancer samples negative for fusions. These contrived plasma samples were evaluated in lieu of clinical samples for this study due to the rarity of these alterations. Plasma was processed from extraction to sequencing on the same batches as the rest of the study samples. The fusion cfDNA was diluted to &lt; 0.2% MAF for ROS1 and NTRK1 at ~30 ng input. There was 100% detection (6/6) across reagent lots for both fusions when tested at 0.15% MAF at approximately 30 ng of cfDNA.

## c. Precision from mutation negative samples

Samples from healthy donors were pre-screened by an externally validated orthogonal method. Mutation negative samples by the orthogonal method were tested by Guardant360 CDx in three reproducibility conditions (i.e., different reagent lots, operators, instruments, and days). Four replicates from each donor were tested with Guardant360 CDx across the different reproducibility conditions. The study demonstrated a sample-level, within-condition ANA of 97.4% and sample-level between-condition ANA of 97.3%. The within-condition ANA was 99.6% and between-condition ANA was 99.6% for 7 variants that had a positive call in at least one condition. Within-condition and between-condition ANA values were 100.0% for all CDx variants (EGFR L858R, EGFR T790M, and EGFR exon 19 deletions) and category 2 variants.

## 6. Carryover/Cross-Contamination

The carryover/cross-contamination study evaluated the prevalence of cross-contamination when material is transferred between samples in the same batch and

PMA P200010: FDA Summary of Safety and Effectiveness Data

{26}

carry-over when material is transferred between samples across batches processed sequentially on the same instrument using Guardant360 CDx.

A total of 352 plasma samples across 8 batches (44 samples/batch x 8 batches) were run in a consecutive order across instruments within the analytical accuracy study and sequenced on 16 flowcells.

There was no evidence of high positive variants from near-by wells detected in negative samples. In conclusion, no carryover or cross-contamination was observed in 352 samples processed across 8 consecutive batches.

# 7. Reagent Lot Interchangeability

The Guardant360 CDx Sample Preparation Kit (G360 SPK), a single use kit, provides the reagents necessary to process a batch of patient cfDNA samples and prepare libraries for sequencing (Guardant360 workflow steps: library preparation &amp; enrichment). This study evaluated the interchangeability of each G360 SPK box (Box 1, 2, 3, and 4). Box 4 has three components 4a, 4b, 4c, which have to be used together from the same lot.

Reagent lot interchangeability was assessed by testing cfDNA sample pools in five replicates using two different lots of Guardant360 CDx Sample Preparation Kit in seven different lot combinations. Two sample pools contained in total 16 known variants, 9 variants in pool 1 and 7 variants in pool 2. Variant positive cfDNA diluted in WT clinical cfDNA to levels near the LoD were prepared from clinical cfDNA (pool 1) and a mixture of clinical cfDNA and cell line-derived cfDNA (pool 2). Cell-line cfDNA was used to evaluate fusions. For the sample replicates that proceeded to sequencing, all met the performance metrics.

To test the interchangeability of the Guardant360 SPK boxes, 5 replicates of two sample pools harboring 16 targeted variants near LoD were tested in 7 interchangeability run conditions. Kit Lot Interchangeability of G360 SPK boxes was evaluated based on the rate of positive agreement for detection of targeted variants.

Out of 70 samples, 68 passed QC metrics (97% pass rate). The rate of qualitative agreement rate (QDR), i.e., the agreement with the majority call for baseline reagent was calculated. QDR was defined as the number of positively detected targeted variants across eligible samples (D) divided by the total number of targeted variants tested across eligible samples (N), expressed as a percentage (100 * D/N). QDR ranged from 91.6% to 98.7%. There was 100.0% negative agreement among expected negative sites within respective pool replicates.

The panel-wide assessment of NPA was 99.9% calculated from negative variant sites across the Guardant360 CDx reportable range that are not detected in the reference condition represents SPK Lot A for all combinations tested.

PMA P200010: FDA Summary of Safety and Effectiveness Data
27 of 75

{27}

# 8. Stability

## a. Reagent Stability

The objective of this study was to evaluate the closed-container stability and establish the shelf-life of the unused Guardant360 CDx Sample Preparation Kit. Three lots of identical reagents with the same specifications were stored under the specified storage conditions for each box and then tested with two sample pools at defined time points following the same protocol.

Two sample pools contained in total 16 known variants, 9 variants in pool 1 and 7 variants in pool 2. Variant positive cfDNA diluted in WT clinical cfDNA to levels near the LoD were prepared from clinical cfDNA (pool 1) and a mixture of clinical cfDNA and cell line-derived cfDNA (pool 2). Cell-line cfDNA was used to evaluate fusions. These variants are both clinically relevant and are representative of the entire Guardant360 CDx panel based on variation in GC content, sequence context, and coverage.

Three (3) lots of G360 SPK boxes 1-4 were used for this study and were tested as soon as feasible after manufacturing and QC. Time zero (T0) was defined as the first testing timepoint, and reagents were aged and used to test samples at or after 3, 4, 7, 10, 13, and 19 months to support a shelf-life stability claim of 2, 3, 6, 9, 12, or 18 months, respectively. When possible, testing at all time points was conducted using frozen aliquots of the same large sample pool.

The G360 SPK boxes were tested at each timepoint with five (5) replicates per each of the two unique sample pools. At least 4 replicates had to pass sample level sequencing QC metrics for a valid timepoint, and only results from samples passing QC were evaluated against the acceptance criteria. If less than 4 replicates passed sequencing level QC, and after failed samples were re-run, the testing timepoint would be considered failed due to unstable SPK reagents for that lot. Replicates were tested at the minimum (5 ng) cfDNA input, as this represents the most challenging condition.

208 of these sample replicates passed all expected QC metrics (96.2% pass rate) and were included in this study (up to 19 months of testing), representing 2 different pools, processed in 5 replicates each, for each of the 3 SPK lots used, and the 7 time points (0, 3, 4, 7, 10, 13, and 19 months).

Qualitative detection rates (QDR), which is based on the agreement with the majority call at T0 for the number of targeted variants detected, were assessed per lot/per timepoint (Table 21). QDR was defined as the number of positively detected targeted variants that were positively detected in the baseline condition across eligible samples (D) divided by the total number of positively detected targeted variants tested across eligible samples (N), expressed as a percentage (100 * D/N). The study shows no significant difference between time points compared to T0 for all three lots (alpha = 0.05), demonstrating that there was no significant decline in detection rates over the

PMA P200010: FDA Summary of Safety and Effectiveness Data
28 of 75

{28}

course of the study (Table 21). All of the expected negative variants were observed as negative calls across all replicates, indicating 100% negative agreement among all targeted variants expected to be negative across study conditions. The panel-wide assessment of NPA was 99.9% calculated from negative variant sites across the Guardant360 CDx reportable range that are not detected in the reference condition representing time 0 for all time points tested.

Table 21. SPK Concordance Per Timepoint and Lot Summary Results

| Lot | Metric | T0 | T3 | T4 | T7 | T10 | T13 | T19 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Lot 1 | Detected / Expected = QDR | 80/80 = 100.0% | 80/80 = 100.0% | 69/71 = 97.2% | 78/80 = 97.5% | 77/80 = 96.3% | 80/80 = 100.0% | 117/121= 96.7% |
| | QDR LLCI | 95.5% | 95.5% | 90.2% | 91.3% | 89.4% | 95.5% | 91.8% |
| | chi squared statistic | N/A* | 0.637 | 0.506 | 1.359 | N/A* | N/A* | 1.313 |
| | p-value | N/A* | 0.212 | 0.238 | 0.122 | N/A* | N/A* | 0.126 |
| Lot 2 | Detected / Expected = QDR | 71/73 = 97.3% | 69/71 = 97.2% | 80/80 = 100.0% | 77/80 = 96.3% | 80/80 = 100.0% | 71/71 = 100.0% | 70/71= 98.6% |
| | QDR LLCI | 90.5% | 90.2% | 95.5% | 89.4% | 95.5% | 94.9% | 92.4% |
| | chi squared statistic | 0 | 0.605 | 0 | 0.605 | 0.479 | 0 | 0 |
| | p-value | 0.5 | 0.782 | 0.5 | 0.782 | 0.756 | 0.5 | 0.5 |
| Lot 3 | Detected /Expected = QDR | 76/80 = 95.0% | 80/80 = 100.0% | 62/64 = 96.9% | 79/80 = 98.8% | 80/80 = 100.0% | 80/80 = 100.0% | 79/80= 98.8% |
| | QDR LLCI | 87.7% | 95.5% | 89.2% | 93.2% | 95.5% | 95.5% | 93.2% |
| | chi squared statistic | 2.3 | 0.02 | 0.83 | 2.3 | 2.3 | 2.3 | 0.2 |
| | p-value | 0.9 | 0.6 | 0.8 | 0.9 | 0.9 | 0.9 | 0.7 |

*chi-square statistic of proportion test is not defined for the case when the two sample proportions of detection are 100%

Based on the time points tested (up to the 19-month time point) the results support a maximum of 18 months of shelf life for the Guardant360 Sample Preparation Kit at the recommended stored conditions.

## b. Whole Blood Stability

The objective of this study was to demonstrate the stability of whole blood specimens used for Guardant360 CDx collected in the Guardant360 BCK, that is in Streck Cell-

PMA P200010: FDA Summary of Safety and Effectiveness Data
29 of 75

{29}

Free DNA BCTs, across the expected range of sample transport and storage conditions for up to 7 days after blood collection prior to plasma isolation.

A total of four BCTs were drawn from each of 16 cancer patients and subjected to the conditions described in Table 22 below. From each patient, one tube was processed to plasma 1 day after blood draw (storage at room temperature). Plasma was then shipped on dry ice to Guardant Health. This constituted the reference condition. In addition to the reference tube, three more blood tubes per donor were shipped as whole blood to Guardant Health and subjected to Condition 1 (Summer profile), Condition 2 (Winter profile) or Condition 3 (Room temperature). After conditioning, plasma was isolated on the 8th day after blood collection and run on the Guardant360 CDx.

Table 22. Description of Whole Blood Storage Conditions

| Condition | BCT # from Each Patient | Storage Condition / Processing |
| --- | --- | --- |
| Reference | 1 | Reference condition: Plasma processing (day 1 after blood collection). |
| 1 | 2 | Summer Profile Storage: 4h at 22°C, 6h at 37°C and 56h at 22°C, 6h at 37°C) plus remaining time at room temperature. |
| 2 | 3 | Winter Profile Storage: 4h at 18°C, 6h at 0°C, 56h at 10°C, and 6h at 0°C plus remaining time at room temperature. |
| 3 | 4 | Room Temperature Storage: Storage at room temperature 18-25°C. |

All 64 samples passed all QC and were included in analysis. All storage conditions demonstrated acceptable performance (Table 23). All samples in each group demonstrated acceptable sample-level molecule recovery as assessed by depth of NSC coverage across the panel. Fold change of median NSC in test condition over the reference condition or time zero ranged from 0.90 to 0.97.

Exon-level coverage was also acceptable for all conditions evaluated. The fraction of exons with relative exon level coverage difference between condition and reference (Time zero) within 2σ (2 * 0.108) was 95.3-96.3%, which demonstrate that there was no preferential drop-out of relative exon-level coverage exceeding expected levels due to random variation, and the entire panel was covered consistently between reference and interfering substance conditions. The lower bound of the 95% exact binomial CI for the fraction of genomic targeted exonic regions with relative exon-level non-singleton coverage is shown in Table 23.

PPAs were also calculated for the SNVs and indels in the reportable range: 10 SNVs and 6 indels. All conditions showed variant call concordant PPA of 87.5% - 93.8%.

PMA P200010: FDA Summary of Safety and Effectiveness Data
30 of 75

{30}

PPA above LoD was 100% for all conditions (Table 23). The data indicate acceptable sensitivity and specificity when using samples across the storage conditions.

The panel-wide NPAs were also calculated for SNVs and indels within the reportable range within 55 genes, CNAs and fusions. The total set of negative variants was set to the reportable range excluding variants found to be positive in the reference condition. The discordant negative variants were defined as those negative variants that were positive in the non-reference condition. The panel wide NPA was 99.9% for condition 1 (739,550 out of 739,552 variants), 99.9% (739,550 out of 739,552 variants) for condition 2, and 99.9% (739,548 out of 739,552 variants) for condition 3 (Table 23).

Table 23. Whole Blood Stability Summary Results

| Study Endpoint | Metric | Summer profile | Winter profile | Room temperature |
| --- | --- | --- | --- | --- |
| | | Study Result | Study Result | Study Result |
| Sample-level Molecule Recovery Difference (fold change) | Median of NSC fold (condition / reference) | 0.90 | 0.97 | 0.97 |
| Relative Exon-Level Coverage | Lower bound of 95% CI for fraction of exons outside expected coverage | 95.4% | 94.8% | 95.8% |
| Variant Call Concordance | PPA | 87.5% (14/16) | 93.8% (15/16) | 93.8% (15/16) |
| | PPA (≥ 1xLoD) | 100.0% (13/13) | 100.0% (13/13) | 100.0% (13/13) |
| | NPA | 99.9% (739,550/739,552) | 99.9% (739,550/739,552) | 99.9% (739,548/739,552) |

The whole blood stability study described above was supplemented by an additional study with two objectives (1) to demonstrate the concordance between samples processed into plasma on the same day as blood collection and the samples processed into plasma the day after collection; (2) robustness to changes in relative humidity (RH) that tubes may be exposed to during shipping.

PMA P200010: FDA Summary of Safety and Effectiveness Data

{31}

A total of four BCTs were drawn from 19 healthy donors. For each donor, one BCT was processed to plasma within 4 hours after blood collection and shipped to Guardant Health on dry ice on the same day. This served as the reference condition. The other 3 BCTs will be subjected to conditions described below:

- Test condition 1. Intact whole blood in BCTs packed in BCKs was shipped overnight to Guardant Health and plasma isolation was done on the day of receipt (Day 1 after blood collection).
- Test condition 2. Exposure of whole blood in BCT starting on the day of blood collection and for 1 day to low humidity (25% RH, at 23°C) storage profile, followed by storage at Room temperature for 1 day. Plasma isolation occurred on Day 2 after blood collection.
- Test condition 3. Storage of whole blood in BCT starting on the day of blood collection and for 1 day at Room temperature, followed by exposure to high-humidity (90% RH, at 23°C) storage profile for 1 day. Plasma isolation occurred on Day 2 after blood collection.

Out of 76 samples processed, 24 study samples (6 distinct donor samples for all 4 conditions) had cfDNA underloading in some samples and overloading in some other samples due to a Guardant operator error. After QC check, 52 samples from 13 donors passed all sample QC metrics and were included in the analysis. Recovery of unique molecules across the 3 conditions did not show a negative impact of Day 1 processing and exposure of tubes to high (90% RH) and low (25% RH) relative humidity conditions. Fold change of median NSC in storage condition over reference condition ranged from 0.95 to 0.99. For the reportable range of the device, the fraction of exons with relative coverage within 2σ (2 * 0.108) ranged 98.1 – 99.0%.

Based on the evidence from preservation of overall coverage and relative exon coverage the quantity and quality of cfDNA are not impacted by: (1) whole blood collection at vendor site and overnight shipping to Guardant Health at room temperature, followed by standard plasma isolation on day 1 after collection, (2) exposure of whole blood in BCT starting on the day of blood collection and for 1 day to low relative humidity (25% RH, at 23°C) storage profile, followed by storage at Room temperature for 1 day and plasma isolation on Day 2 after blood collection, and (3) Storage of whole blood in BCT starting on the day of blood collection and for 1 day at Room temperature, followed by exposure to high relative humidity (90% RH, at 23°C) storage profile for 1 day and plasma isolation on Day 2 after blood collection.

Based on these study results, whole blood may be stored in Cell-Free DNA BCTs for up to 7 days after blood collection and prior to plasma isolation, and can withstand winter and summer shipping conditions.

c. Plasma Stability

To define the storage conditions and evaluate the stability of plasma isolated from whole blood, stability at defined temperatures and durations was assessed. Four

PMA P200010: FDA Summary of Safety and Effectiveness Data
32 of 75

{32}

BCTs from 12 cancer patients, 48 samples in total, were collected and run on Guardant360 CDx, with plasma stored at the specified storage conditions (Table 24). Plasma from one BCT was processed through cfDNA extraction on the same day as a reference condition (Ref), plasma from a second BCT was stored at 2-8°C for 25 hours before cfDNA extraction (for a 24-hour stability claim at 2-8°C) as condition 1 (C1), plasma from a third BCT was stored at -80°C ± 10°C with two freeze/thaw cycles for 46 days before cfDNA extraction (for a 45-day stability claim at -80°C ± 10°C) as condition 2 (C2), and plasma from a fourth BCT was stored at -80°C ± 10°C for one year before cfDNA extraction to support usage of stored plasma for analytical validation (AV) studies (C3). Extracted cfDNA from each condition was stored at -20°C ± 5°C until further processing.

Table 24. Description of Plasma Storage Conditions

| Condition | BCT # from Each Patient | Storage Condition / Processing |
| --- | --- | --- |
| Reference (Ref) | 1 | Reference condition: cfDNA extracted directly after plasma isolation on the same day. |
| Condition 1 (C1) | 2 | Storage of plasma at 2-8°C for 25 hours before cfDNA extraction (for a 24-hour stability claim at 2-8°C). |
| Condition 2 (C2) | 3 | Storage of plasma at -80°C ± 10°C plus 2 freeze/thaw cycles for 46 days before cfDNA extraction (for a 45-day stability claim at -80°C ± 10°C). |
| Condition 3 (C3) | 4 | Storage of plasma at -80°C ± 10°C plus 2 freeze/thaw cycles for one year to support usage of stored plasma for analytical validation (AV) studies. |

Out of 48 samples processed, 40 study samples (11 samples in reference condition, 8 samples in Condition 1, 10 samples in Condition 2 and 11 samples in Condition 3) passed their respective in-process and post-sequencing QC metrics and had at least one reference-condition sample pair, thus were included in the final analysis. In the three tested storage conditions, samples demonstrated acceptable performance (Table 23). In the three tested storage conditions, samples demonstrated acceptable sample-level molecule recovery, relative exon-level coverage, and variant call concordance.

Sample level molecule recovery showed fold change of 0.93, 1.10 and 0.99, less than 25% change with 95% confidence interval in all three conditions.. Exon-level relative coverage demonstrated 92.8%-97.1% fraction of exons within 2σ of expected relative coverage. The lower bound of the 95% exact binomial CI for the fraction of genomic targeted exonic regions with relative exon-level non-singleton coverage is shown in Table 25.

PMA P200010: FDA Summary of Safety and Effectiveness Data
33 of 75

{33}

PPAs were also calculated for the SNVs and indels in the reportable range within 55 genes that are reportable by test, as well as the reportable CNA and fusion genes: 14 SNVs, 1 indel and 1 CNA. Three conditions showed variant call concordant PPA of $76.9\% - 78.6\%$ . PPA above LoD was $90.9\% - 91.7\%$ for all conditions (a single variant was discordant, see Table 25). NPA across the reportable range was $99.9\% - 99.9\%$ .

Based on these study results, plasma may be stored at $2 - 8^{\circ}\mathrm{C}$ for 24 hours or at $-80^{\circ}\mathrm{C} \pm 10^{\circ}\mathrm{C}$ plus 2 freeze/thaw cycles for 1 year before cfDNA extraction.

Table…