GRAIL Public Comments to FDA Advisory Committee on MCD Tests

December 19, 2023

ELECTRONIC SUBMISSION AT www.regulations.gov

Attn: Jarrod Collier, M.S.
Designated Federal Officer
FDA Advisory Committee
Molecular and Clinical Genetics Panel of the Medical Devices Advisory Committee
Docket FDA–2023–N–4720
US Food and Drug Administration
Silver Spring, MD 20993

Dear Mr. Collier:

GRAIL, LLC (“GRAIL” or “we”) appreciates the opportunity to submit a public comment to docket FDA–2023–N–4720 on the Molecular and Clinical Genetics Panel of the Medical Devices Advisory Committee on the topic of Multi-Cancer Detection (“MCD”) Devices. We have focused our comments on the design of MCD in vitro diagnostic devices (tests) and methodological considerations for evaluation of benefits and risks of MCD tests, together with recommendations for study design and study outcomes of interest. We commend the FDA’s commitment to engage with stakeholders, including test developers, providers, and most importantly patients, to inform the Agency’s regulatory decision-making regarding these important and novel tests.

Background

GRAIL is a healthcare company whose mission is to detect cancer early, when it can be cured. Through our population-scale clinical studies and machine learning and data science, we have developed the Galleri^(R)multi-cancer early detection blood test.

The Galleri test is an analytically- and clinically-validated test for early detection of multiple types of cancer. In clinical studies, Galleri has shown the ability to identify a shared cancer signal across more than 50 types of cancer, often at an early stage, demonstrating a high positive predictive value (“PPV”) and low false positive rate.^[1],^[2] If a cancer signal is detected, Galleri also predicts the tissue type or organ associated with the cancer signal (the cancer signal origin) with high accuracy.^[3] Galleri works by detecting DNA fragments shed into the bloodstream by tumor cells, referred to as cell-free DNA (“cfDNA”).^[4] DNA has specific methylation patterns that can be used to both identify a general cancer signal and to localize that signal to a specific organ or tissue type. Importantly, Galleri is intended as a complement to recommended standard of care single cancer detection tests, not as a replacement.

As part of the FDA breakthrough designation and PMA submission process, GRAIL has consulted numerous experts, and diligently assessed the complex and challenging issues involved in developing novel MCD technology and designing first-of-its-kind study and outcome measures. Our learnings in key areas of discussion identified for consideration by the FDA Advisory Committee are shared below.

Design of Multi-Cancer Detection Tests

Based on discussions and consultations with key opinion leaders, medical practitioners, patient advocates and clinical experts, GRAIL believes the following are essential features for any MCD test to be implemented as a screening test in an average risk, asymptomatic population (e.g., >50 years of age) in addition to currently recommended single cancer screenings, optimizing the benefits of screening overall while mitigating potential risks:^[5]

Ability to identify a broad range of cancer types:^[6] An MCD test should identify a broad spectrum of cancer types to optimize overall benefit for the intended use population and the number of cancers detected in a screened population (e.g., PPV and yield). PPV represents the probability that a positive test result is a true positive, and yield represents the percentage of cancer detected within the screened population.
High PPV and low false positive rate: An MCD test should have a high PPV and low false positive rate to assure reliable and accurate positive test results, and reduce the number of unnecessary workups, and associated harms, in the intended use population.
Ability to limit overdiagnosis of indolent cancers: An MCD test should preferentially detect aggressive cancers that warrant treatment and should not unnecessarily contribute to overdiagnosis of more indolent cancers.
Ability to predict with high accuracy the cancer signal origin and direct diagnostic workup: An MCD test should predict the cancer signal origin with high accuracy to facilitate efficient diagnostic workups.
Application to a diverse population: An MCD test should be supported by a robust clinical evidence program that supports implementation in a broad and diverse average risk intended use population.

The Galleri test detects a cancer signal shared by a broad spectrum of cancers, including the most deadly types of cancers that are not covered by cancer screening guidelines, enabling detection of a broad range of cancer types.^[7]Currently, cancer screening guidelines cover only five cancer types (breast, cervical, colorectal, lung, and prostate). These single-cancer detection (“SCD”) approaches are estimated to cover only 15% of cancer diagnoses and 30% of cancer deaths in the United States, meaning that 70% of cancer deaths are due to cancers without a current screening paradigm.^[8] Of these deaths, many are attributable to rare cancer types. By leveraging the biology of the “shared cancer signal,” Galleri enables the detection of a broad spectrum of cancers, including less common cancer types. We believe that the only way to conduct general population screening for multiple types of cancers, including rare cancers and those that predominantly affect racial/ethnic minority groups, is to aggregate them for screening. By targeting this aggregate incidence, Galleri has the potential to increase the yield of cancers diagnosed through screening from 15% to 49%.^[9]

Additionally, a “shared cancer signal” approach to MCD, which Galleri uses, allows for a single low false positive rate (FPR), which we believe contributes to optimizing PPV.^[10] This shared cancer signal is rarely observed in individuals known not to have cancer. For a condition like cancer that has a low prevalence in the population, PPV is significantly impacted by prevalence and specificity, such that PPV increases with the prevalence of cancer in the population and with the specificity. Asymptomatic screening tests designed to optimize PPV (and likelihood of benefit) appropriately balance the potential risks involved in confirmatory diagnostic workups. We believe that the Galleri test’s “shared cancer signal” approach is clinically preferable to multiple SCD tests intended for multi-cancer screening, whether independently administered or designed by stacking a set of individual screens combined into a single test. SCD tests generally are designed to optimize sensitivity, and thus tolerate a higher FPR. Screening individuals with multiple SCD tests adds to the cumulative FPR. Ten SCD tests, each optimized for one deadly cancer type, used in annual screening of 100,000 persons aged 50-79 in addition to USPSTF-recommended screening, would result in a cumulative FPR of 12%, while the Galleri test FPR remains at 0.4%. A single, low FPR with an MCD test limits unnecessary workups in patients who do not have cancer, and associated harms, which is important in evaluating the overall risks of the test.

The Galleri test has been shown to preferentially detect more aggressive cancer types, potentially limiting overdiagnosis of indolent cancers.^[11],^[12] Data across our clinical studies suggests that although Galleri detects cancer signals for some of the most aggressive cancers, detection of cancer signals for indolent cancer types, which people are less likely to die from, is low. This is because more aggressive cancers shed more DNA into circulation, making them more detectable; conversely, less aggressive, more indolent cancers, like early-stage prostate cancer, shed less, and are less detectable through cfDNA tests. As a result, the design of the Galleri test helps limit potential risks from overdiagnosis and the harms potentially caused by unnecessary procedures and workups.

In addition to a low FPR and limited detection of indolent cancer types, the ability to predict with high accuracy the cancer signal origin and direct diagnostic workup further reduces MCD screening risks.^[13] Accurate cancer signal origin helps limit unnecessary diagnostic tests and procedures. Galleri has a 88% cancer signal origin prediction accuracy for identifying the location of cancer, which supports targeted and more efficient diagnostic resolution through established workup pathways. Our PATHFINDER study showed that 80% of screened subjects achieved diagnostic resolution by cancer signal origin-directed initial evaluation and generally facilitated diagnosis in less than 3 months (median of 79 days) among participants who had a cancer signal detected.^[14]

Further, Galleri has been validated in large clinical studies across diverse populations in behaviors (such as smoking), non-cancer diseases, environmental exposures, age, gender, race, ethnicity, socio-economic status, and other confounding indications and differences.^[15] For example, in published data from our CCGA study, we found no meaningful differences in performance across racial subgroups.^[16] Understanding cancer signals associated with demographic diversity is pivotal to our ability to account for biological noise and underpin the high-specificity of our test while allowing for use in a broad and diverse screening population.

GRAIL has studied and evaluated the most clinically appropriate MCD design features, and believes that any MCD test should optimize early cancer detection, particularly of aggressive cancers, limit risks of unnecessary diagnostic tests and procedures, and have the potential to achieve reduced overall cancer mortality and morbidity, when added to existing guideline-based screening paradigm.

Methodological Considerations for Evaluation of Benefits and Risks of MCD Tests and MCD Study Designs and Study Outcomes of Interest

As part of developing our large clinical evidence program for MCD over the past seven years, GRAIL has diligently assessed and implemented key study designs, methodologies, and outcomes appropriately suited to evaluate the probable benefits and risks of MCD screening tests. This experience, and ongoing consultation with leading experts in the field, has informed our thinking. In this comment, we focus on two methodologic considerations for evaluating the benefits and risks of MCD tests. We believe that (1) the primary study endpoint of absolute incidence of late stage cancer represents the most promising strategy for robust yet accelerated evidence regarding the clinical utility of MCD tests and (2) the performance or clinical utility of MCD should be evaluated by the the aggregate measure of detecting cancer in the intended use population, rather than by individual cancer type, as discussed below.

There have been debates on what constitutes appropriate endpoints for the study of cancer screening trials. Towards that, some have cited the applicability of mortality including all-cause mortality. A primary endpoint of mortality may not capture in a timely manner the substantial morbidity reduction and patient-reported benefits of earlier cancer detection.^[17] Given the high and growing number of cancer deaths still occurring under current screening recommendations, there is substantial opportunity cost in deferring adoption of screening tests with the potential to reduce cancer morbidity and mortality. As such, more rapid and more patient-centric yet rigorous evaluation of their clinical utility is needed. The relevance of trial results reported after a decade or more will be diminished by the rapid evolution of MCD technologies and cancer treatments during that time. In the same way that modern drug trials are being redesigned to fit with modern therapies (e.g. basket and umbrella trials),^[18] cancer screening trials also need a major reconsideration with regards to endpoints, design, and analytical methods to keep up with rapidly advancing technology.

Groups such as the MCED Consortium are working to rethink traditional strategies for clinical utility evidence generation involving trials with nearer-term and more patient-centric endpoints, real-world data collection, and more sophisticated modeling of longer-term endpoints. A range of alternative trial endpoints have been suggested, including reduction in late-stage cancer incidence, candidacy for curative interventions at diagnosis, overall cancer detection rate, reduced treatment morbidity for early-stage cancers, increased treatment response rates, improved quality of life during and after treatment, and rates of metastatic recurrence.

Observations from screening trials for breast, colorectal, and lung cancer provide evidence supporting the demonstration of a reduction in the incidence of late-stage cancer as a relevant surrogate for cancer-specific mortality. For example, while a meta-analysis of nine randomized controlled trials of mammography reported an overall mortality benefit of 22%, the trials that reduced advanced stage disease by >20% showed an even greater (28%) reduction in mortality, corresponding to a 40% reduction in those who actually participated in screening.^[19] Furthermore, reductions in advanced stage disease in these trials accounted for two-thirds of the benefit from screening. As such, at least two ongoing breast and colorectal cancer screening trials have chosen incidence of late stage disease as their measured endpoint and used the mortality reduction modeled from that measure to make decisions that are both timely and less burdensome with respect to sample size/statistical power.

Therefore, we believe that absolute incidence of late stage cancer as a directly measured primary endpoint in clinical studies, followed by sophisticated computational disease modeling to understand potential impacts on cancer-specific mortality, represents the most promising strategy for robust yet accelerated evidence regarding the clinical utility of MCD and other technologies that detect cancer in average-risk populations. Designing and powering clinical utility studies to directly measure cancer-specific or all-cause mortality invoke intractable challenges related to efficiency and complexity under rapid evolution of cancer screening and treatment technologies. While some researchers express interest in the relationship between cancer screening and directly measured all-cause mortality,^[20] we consider that endpoint insensitive for understanding the impact of the cancer-directed intervention,^[21],^[22] and will be difficult to interpret due to competing deaths from non-cancer over long time periods, but point out that it can certainly be modeled and evaluated from studies followed to an endpoint of cancer stage at diagnosis. A recent commentary led by Dr. Ruth Etzioni and National Cancer Institute researchers agree that building a rigorous program of objective analytical and modeling studies is our best bet for accelerating evaluation of novel screening tests in a way that balances scientific rigor with timely answers.^[23]

Another important consideration for the evaluation of MCD tests involves the appropriate endpoint across all cancer types that can be detected by the test under investigation. Studies should primarily aim to determine the aggregate performance or clinical utility of a test in the intended use population, not stratified or powered to subtypes of targeted cancers. It is not possible to know in advance the cancer type to be detected at the time when an MCD test is administered – just as individuals are not able to choose which type of cancer they develop. Although the benefits, risks, and performance characteristics of any MCD test will ultimately vary by cancer type detected, these considerations are secondary to the primary public health goal of increasing the total number of cancers diagnosed in a population at measurable risk of any and all cancer types. Thus, we believe that the appropriate evaluation endpoint for an MCD test is the performance across all cancer types that can be detected by the test.

GRAIL has leveraged these methodological underpinnings in assembling what we believe is the largest clinical development program in genomic medicine, aimed at robustly validating our breakthrough device technology responsibly and expeditiously, to allow for its deployment in improving public health. GRAIL has enrolled over 300,000 subjects in over 8 different studies to develop and demonstrate the clinical validity, utility and value of methylation-based MCD testing that has a shared cancer signal and highly accurate CSO prediction.

These studies include our foundational case-control Circulating Cell-free Genome Atlas (CCGA) and PATHFINDER studies, which included more than 21,000 participants to develop, validate, and launch our Galleri technology, and additional ongoing multiple large-scale observational and interventional studies.^[24],^[25]

The CCGA study is a large multicenter case-control study that enrolled approximately 10,000 participants with newly diagnosed cancer and 5,000 participants without a diagnosis of cancer from medical institutions and health systems and followed them for outcome information.^[26] The CCGA study was divided into three sub-studies to develop, train, and validate classifiers for cancer and non-cancer signal detection, as well as CSO prediction. The CCGA-3 sub-study demonstrated detection of a shared cancer signal across more than 50 cancer types and accurate CSO prediction, with a false positive rate of 0.5%.^[27]

The PATHFINDER study was an interventional multi-center study of 6,662 participants representing the first pilot implementation of the Galleri test into clinical practice.^[28],^[29] The study was conducted under an FDA-approved investigational device exemption (“IDE”) application. Test results, including CSO, were returned to study investigators who determined the appropriate diagnostic evaluation. The study objectives were (a) to assess the extent of cancer diagnostic testing required to achieve diagnostic resolution following a cancer signal detected test result, (b) evaluate clinical performance of the Galleri test, and (c) assess participant-reported outcomes. Critically, performance was consistent with our case-control CCGA study: a pre-specified retrospective re-analysis of samples with the Galleri version of the test showed a PPV of approximately 43%, cancer signal origin prediction accuracy of approximately 88%, and specificity of 99.5%.^[30]

GRAIL is currently conducting two additional population-scale interventional studies, both of which are conducted under FDA-approved IDEs: the PATHFINDER 2 study and the National Health Service (“NHS”)-Galleri trial. These studies have been designed to evaluate the safety and performance of Galleri, including, among other endpoints, a measurable outcome of whether an MCD test program can reduce the proportion of late-stage cancers detected.^[31]

The PATHFINDER 2 Study is an interventional, multi-center study of participants receiving the Galleri test.^[32] The study is enrolling approximately 35,000 participants aged 50 years or older from clinical study sites in North America over an anticipated enrollment period of approximately 36 months. 22% of the enrolled subjects are intended to be from minority and under-served populations.The primary objectives of this study are to evaluate the safety of the Galleri test in terms of cancer diagnostic workup triggered by a test result and performance of the test in individuals eligible for cancer screening, with a three year follow-up period. Test results are being returned to the healthcare providers and information on the ensuing cancer diagnostic workup is recorded.

The NHS-Galleri trial is a blinded, randomized, controlled study of the clinical utility of the Galleri test that is being run through a public-private partnership with The Cancer Research UK and King’s College London Cancer Prevention Trials Unit, in partnership with the NHS and GRAIL.^[33] The NHS-Galleri trial has enrolled 142,966 asymptomatic participants by inviting an estimated 1.5 million people from the general population aged 50-77 years who reside in England. Blood samples will be collected at three annual study visits. Following the baseline visit, participants were randomized 1:1 to either the intervention arm (blood samples tested) or control arm (blood samples not tested but stored for potential future use). Only participants in the intervention arm with a cancer signal detected are receiving the test result and referral for diagnostic investigations by the NHS. The primary objective of the trial is to evaluate whether there is a statistically significant reduction in the absolute numbers of stage III and IV cancers diagnosed in the intervention arm compared with the control arm. This endpoint is expected 3.5 years after randomization. Additional near-term endpoints of interest include cancer-specific mortality modeled from the stage distribution, receipt of curative cancer surgery and treatment, and patient-reported outcomes. Longer-term endpoints of interest include cancer-specific and all-cause mortality. However, serious methodologic challenges with these longer-term endpoints in the context of MCD and/or genomic technologies for early cancer detection are outlined above and provide justification for utilization of nearer-term outcomes to understand the range of benefits and harms potentially associated with MCD and to accelerate adoption.

Conclusion

Despite 50 years of the “war on cancer,” most cancers in the United States still are detected too late, after a person develops symptoms and once the cancer already has spread. This is because the currently recommended cancer screening programs only find at most 15% of the total cancer burden among those eligible for screening.

Recommended screening tests can find cancer early and improve outcomes, but are only available for five types of cancer. We know that cancer found in an early stage is associated with improved survival versus cancer found in later stages, which can be more difficult and costly to treat and may have fewer therapeutic options. In fact, the 5-year survival rate across all cancers is 89% when cancer is found while it still is localized, but drops to 21% once it has spread. Screening for one cancer at a time—and for so few cancers—is simply not going to materially reduce deaths from cancer, as most cancers are not covered by the current screening paradigm.

We believe making validated MCD tests, with features described herein, available as a complement to recommended screening programs will dramatically increase cancer detection from screening in the population, and improve public health. GRAIL feels great urgency to address the burden of cancer and to work in collaboration with the FDA and other key stakeholders to achieve our shared mission to end late-stage cancer as we know it. We greatly appreciate the opportunity to comment and for the FDA’s ongoing work to accelerate the regulatory evaluation of MCD tests.

Respectfully submitted,

Lakshman Ramamurthy
On behalf of GRAIL, LLC
Lakshman Ramamurthy, Ph.D
Vice President, Regulatory Affairs
GRAIL, LLC

^[1]Klein EA, Richards D, Cohn A, et al. Clinical validation of a targeted methylation-based multi-cancer early detection test using an independent validation set. Ann Oncol. 2021;32(9):1167-1177.
^[2]Hubbell E, Venn O, Shanmugam A. Shared cancer signal: evidence from cross-training. USC Computational Biology Symposium; May 19–21, 2022, Los Angeles, CA
^[3]Klein (n 1).
^[4]Jamshidi A, Liu MC, Klein EA, et al. Evaluation of cell-free DNA approaches for multi-cancer early detection. Cancer Cell. 2022;40(12):1537-1549.e12.
^[5]Hackshaw A, Clarke CA, Hartman AR. New genomic technologies for multi-cancer early detection: Rethinking the scope of cancer screening. Cancer Cell. 2022;40(2):109-113.
^[6]Ahlquist DA. Universal cancer screening: revolutionary, rational, and realizable. NPJ Precis Oncol. 2018;2(1):23.
^[7]Klein (n 1).
^[8]Hackshaw A, Cohen SS, Reichert H, et al. Estimating the population health impact of a multi-cancer early detection genomic blood test to complement existing screening in the US and UK. Br J Cancer. 2021;125(10):1432-1442.
^[9]Ibid.
^[10]Hubbell (n 2).
^[11]Chen X, Dong Z, Hubbell E, et al. Prognostic Significance of Blood-Based Multi-cancer Detection in Plasma Cell-Free DNA. Clin Cancer Res. 2021;27(15):4221-4229.
^[12]Bredno J, Lipson J, Venn O, at al. Clinical correlates of circulating cell-free DNA tumor fraction. PLoS One. 2021;16(8):e0256436.
^[13]Klein (n 1).
^[14]Schrag D, Beer TM, McDonnell CH 3rd, et al. Blood-based tests for multicancer early detection (PATHFINDER): a prospective cohort study. Lancet. 2023;402(10409):1251-1260.
^[15]Klein (n 1).
^[16]Tang WHW, Yimer H, Tummala M, et al. Performance of a targeted methylation-based multi-cancer early detection test by race and ethnicity. Prev Med. 2023;167:107384.
^[17]https://www.fda.gov/about-fda/cdrh-patient-science-and-engagement-program/patient-preference-information-ppi-medical-device-decision-making.
^[18]Master Protocols: Efficient Clinical Trial Design Strategies to Expedite Development of Oncology Drugs and Biologics Guidance for Industry. Published March 2022. Accessed November 15, 2023. [PDF Link]
^[19]Tabár L, Yen AM, Wu WY, et al. Insights from the breast cancer screening trials: how screening affects the natural history of breast cancer and implications for evaluating service screening programs. Breast J. 2015;21(1):13-20.
^[20]Carr D, Kent DM, Welch HG. All-cause mortality as the primary endpoint for the GRAIL/National Health Service England multi-cancer screening trial. J Med Screen. 2022;29(1):3-6.
^[21]Duffy SW. All-cause mortality in multi-cancer screening trials. J Med Screen. 2022;29(1):1-2.
^[22]Prasad V, Lenzer J, Newman DH. Why cancer screening has never been shown to “save lives”–and what we can do about it. BMJ. 2016;352:h6080.
^[23]Etzioni R, Gulati R, Patriotis C, et al. Revisiting the Standard Blueprint for Biomarker Development to Address Emerging Cancer Early Detection Technologies [published online ahead of print, 2023 Nov 6]. J Natl Cancer Inst. 2023;djad227.
^[24]Liu MC, Oxnard GR, Klein EA, et al. Sensitive and specific multi-cancer detection and localization using methylation signatures in cell-free DNA. Ann Oncol. 2020;31(6):745-759.
^[25]Klein (n 1).
^[26]Ibid.
^[27]Ibid.
^[28]Nadauld LD, McDonnell CH 3rd, Beer TM, et al. The PATHFINDER Study: Assessment of the Implementation of an Investigational Multi-Cancer Early Detection Test into Clinical Practice. Cancers (Basel). 2021;13(14):3501.
^[29]Schrag (n 14).
^[30]Ibid.
^[31]Neal RD, Johnson P, Clarke CA, et al. Cell-Free DNA-Based Multi-Cancer Early Detection Test in an Asymptomatic Screening Population (NHS-Galleri): Design of a Pragmatic, Prospective Randomised Controlled Trial. Cancers (Basel). 2022;14(19):4818.
^[32]https://classic.clinicaltrials.gov/ct2/show/NCT05155605
^[33]Neal (n 31).