PLOS ONE

RESEARCH ARTICLE

Clinical validation of a next-generation

sequencing-based multi-cancer early
detection “liquid biopsy” blood test in over
1,000 dogs using an independent testing set:
The CANcer Detection in Dogs (CANDiD) study
Andi Flory1,2,3, Kristina M. Kruglyak1, John A. Tynan1, Lisa M. McLennan1, Jill M. Rafalko ID1*,
Patrick Christian Fiaux ID1, Gilberto E. Hernandez1, Francesco Marass ID1, Prachi Nakashe1,
a1111111111 Carlos A. Ruiz-Perez1, Donna M. Fath1, Thuy Jennings1, Rita Motalli-Pepio1, Kate Wotrang1,
a1111111111 Angela L. McCleary-Wheeler1,4, Susan Lana5, Brenda Phillips2, Brian K. Flesner ID4,6, Nicole
a1111111111 F. Leibman7, Tracy LaDue8, Chelsea D. Tripp9, Brenda L. Coomber ID10, J. Paul Woods11,
a1111111111 Mairin Miller3, Sean W. Aiken2, Amber Wolf-Ringwall12, Antonella Borgatti ID12,
a1111111111 Kathleen Kraska2, Christopher B. Thomson ID3, Alane Kosanovich Cahalane13, Rebecca
L. Murray9, William C. Kisseberth14, Maria A. Camps-Palau7, Franck Floch15,16, Claire Beaudu-
Lange17, Aurélia Klajer-Peres18, Olivier Keravel18, Luc-André Fribourg-Blanc19, Pascale
Chicha Mazetier20, Angelo Marco21, Molly B. McLeod22, Erin Portillo23, Terry S. Clark24,
Scott Judd25, C. Kirk Feinberg21, Marie Benitez21, Candace Runyan26, Lindsey Hackett27,
Scott Lafey28, Danielle Richardson11, Sarah Vineyard29, Mary Tefend Campbell30,
OPEN ACCESS
Nilesh Dharajiya31,33, Taylor J. Jensen32,33, Dirk van den Boom33, Luis A. Diaz, Jr.33,34, Daniel
Citation: Flory A, Kruglyak KM, Tynan JA, S. Grosu ID1, Arthur Polk ID1, Kalle Marsal1, Susan Cho Hicks1, Katherine M. Lytle ID1,
McLennan LM, Rafalko JM, Fiaux PC, et al. (2022) Lauren Holtvoigt1, Jason Chibuk1, Ilya Chorny1, Dana W. Y. Tsui1
Clinical validation of a next-generation sequencing-
based multi-cancer early detection “liquid biopsy” 1 PetDx, La Jolla, California, United States of America, 2 Veterinary Specialty Hospital of San Diego, San
Diego, California, United States of America, 3 Veterinary Specialty Hospital of North County, San Marcos,
blood test in over 1,000 dogs using an independent
California, United States of America, 4 Department of Veterinary Medicine and Surgery, University of Missouri,
testing set: The CANcer Detection in Dogs
Columbia, Missouri, United States of America, 5 Department of Clinical Sciences, Colorado State University,
(CANDiD) study. PLoS ONE 17(4): e0266623. Fort Collins, Colorado, United States of America, 6 Department of Clinical Science and Advanced Medicine,
https://doi.org/10.1371/journal.pone.0266623 University of Pennsylvania, Philadelphia, Pennsylvania, United States of America, 7 The Animal Medical Center,
New York, New York, United States of America, 8 Southeast Veterinary Oncology and Internal Medicine,
Editor: Joseph Alan Bauer, Bauer Research
Orange Park, Florida, United States of America, 9 Bridge Animal Referral Center, Edmonds, Washington,
Foundation, UNITED STATES
United States of America, 10 Department of Biomedical Sciences, Ontario Veterinary College, University of
Received: November 24, 2021 Guelph, Guelph, Ontario, Canada, 11 Institute for Comparative Cancer Investigation at the Mona Campbell
Centre for Animal Cancer, Ontario Veterinary College, University of Guelph, Guelph, Ontario, Canada,
Accepted: March 23, 2022 12 Department of Veterinary Clinical Sciences, University of Minnesota, College of Veterinary Medicine, Saint
Paul, Minnesota, United States of America, 13 Veterinary Specialty Hospital of Hong Kong, Wan Chai, Hong
Published: April 26, 2022 Kong, 14 Department of Veterinary Clinical Sciences, The Ohio State University College of Veterinary Medicine,
Copyright: © 2022 Flory et al. This is an open Columbus, Ohio, United States of America, 15 Oncovet, Villeneuve-D’ascq, France, 16 AniCura TRIOVet,
Rennes, France, 17 Clinique Vétérinaire de la Pierre Bleue, Pipriac, France, 18 Eiffelvet, Paris, France,
access article distributed under the terms of the
19 Clinique Vétérinaire SeineVet, Rouen, France, 20 Clinique Vétérinaire Mazetier, Argenteuil, France,
Creative Commons Attribution License, which
21 Governor Animal Clinic, Inc., San Diego, California, United States of America, 22 City Paws Home Health,
permits unrestricted use, distribution, and Columbus, Ohio, United States of America, 23 VCA Valley Oak Veterinary Center, Chico, California, United
reproduction in any medium, provided the original States of America, 24 VCA Metroplex Animal Hospital, Irving, Texas, United States of America, 25 Prices Creek
author and source are credited. Veterinary Services, Lewisburg, Ohio, United States of America, 26 Carlsbad Animal Hospital, Carlsbad,
California, United States of America, 27 Oceanside Veterinary Hospital, Oceanside, California, United States of
Data Availability Statement: All relevant data are
America, 28 Amici Pet Hospital of Little Italy, San Diego, California, United States of America, 29 Colony
within the paper and its Supporting Information Veterinary Hospital, San Diego, California, United States of America, 30 Carriage Hills Animal Hospital,
files. Montgomery, Alabama, United States of America, 31 Healthbit.ai Inc., San Diego, California, United States of
America, 32 Laboratory Corporation of America, Durham, North Carolina, United States of America, 33 Advisor
Funding: This study received funding from PetDx.
to PetDx, La Jolla, California, United States of America, 34 Division of Solid Tumor Oncology, Memorial Sloan
The funder had the following involvement with the
Kettering Cancer Center, New York, New York, United States of America
study: study design, data collection and analysis,
decision to publish, and preparation of the * [email protected]
manuscript. Study sites were compensated for

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 1 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

submission of samples and for collection and QC

of clinical data.
Abstract
Competing interests: AF, KMK, JAT, LMM, JMR, Cancer is the leading cause of death in dogs, yet there are no established screening para-
PCF, GEH, FM, PN, CAR-P, DMF, TJ, RM-P, KW, digms for early detection. Liquid biopsy methods that interrogate cancer-derived genomic
ALM-W, DSG, AP, KM, SCH, KML, LH, JC, IC, and
alterations in cell-free DNA in blood are being adopted for multi-cancer early detection in
DWYT are employed by or affiliated with PetDx;
and receive compensation from PetDx and/or hold human medicine and are now available for veterinary use. The CANcer Detection in Dogs
vested or unvested equity in PetDx. Drs. Flesner (CANDiD) study is an international, multi-center clinical study designed to validate the per-
and Leibman are members of the PetDx clinical formance of a novel multi-cancer early detection “liquid biopsy” test developed for noninva-
advisory board and receive compensation and
sive detection and characterization of cancer in dogs using next-generation sequencing
equity. Dr. Borgatti has ownership interest
(including patents) in a patent entitled “Reduction (NGS) of blood-derived DNA; study results are reported here. In total, 1,358 cancer-diag-
of EGFR therapeutic toxicity” filed by the University nosed and presumably cancer-free dogs were enrolled in the study, representing the range
of Minnesota Office of Technology of breeds, weights, ages, and cancer types seen in routine clinical practice; 1,100 subjects
Commercialization. Drs. Cahalane, Phillips, and
Aiken are investors in PetDx. Drs. Jensen, van den
met inclusion criteria for analysis and were used in the validation of the test. Overall, the liq-
Boom, Dharajiya, and Diaz are advisors for PetDx uid biopsy test demonstrated a 54.7% (95% CI: 49.3–60.0%) sensitivity and a 98.5% (95%
and hold vested or unvested equity in PetDx. Drs. CI: 97.0–99.3%) specificity. For three of the most aggressive canine cancers (lymphoma,
Chorny, Kruglyak, Grosu, Marass, Ruiz-Perez and
hemangiosarcoma, osteosarcoma), the detection rate was 85.4% (95% CI: 78.4–90.9%);
Tsui hold pending patent applications related to
technology described in this work. This does not and for eight of the most common canine cancers (lymphoma, hemangiosarcoma, osteosar-
alter our adherence to PLOS ONE policies on coma, soft tissue sarcoma, mast cell tumor, mammary gland carcinoma, anal sac adenocar-
sharing data and materials. cinoma, malignant melanoma), the detection rate was 61.9% (95% CI: 55.3–68.1%). The
test detected cancer signal in patients representing 30 distinct cancer types and provided a
Cancer Signal Origin prediction for a subset of patients with hematological malignancies.
Furthermore, the test accurately detected cancer signal in four presumably cancer-free sub-
jects before the onset of clinical signs, further supporting the utility of liquid biopsy as an
early detection test. Taken together, these findings demonstrate that NGS-based liquid
biopsy can offer a novel option for noninvasive multi-cancer detection in dogs.

Introduction
Cancer is by far the leading cause of death in adult dogs [1]. In many cases, canine cancer is
identified only after clinical signs have developed, by which point the disease is often advanced,
the ability to provide long-term control is low, and the prognosis is poor. Just as for human
cancer patients, early detection and treatment are considered essential for achieving the best
possible clinical outcomes for canine cancer patients [2, 3], and studies of cancers detected at
earlier stages (including incidentally detected cases) in dogs have demonstrated improved out-
comes [4–8]. In humans, organ-based cancer screening programs such as mammograms for
women, prostate specific antigen (PSA) testing for men, and colonoscopies are well established
and are covered by most insurance policies, as they have been proven to help detect cancers at
earlier stages, when treatment is more effective, and a cure is more likely to be achieved [9].
Liquid biopsy methods, which detect blood-based analytes such as cancer-derived DNA, were
initially developed as noninvasive alternatives to tissue-based “companion diagnostic” testing
[10] for selection of targeted treatments in human cancer patients [11]; liquid biopsy solutions
are particularly useful in situations where tissue biopsies were difficult or impossible to obtain,
such as in lung cancer [12, 13]. The first FDA approval of a liquid biopsy next-generation
sequencing (NGS) based test was issued in August 2020, for a companion diagnostic indication
[14].

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 2 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

Liquid biopsy methods are now being adopted in human medicine to simultaneously
screen for multiple types of cancer with a simple blood test; these multi-cancer early detection
(MCED) tests represent a paradigm shift in cancer screening and promise to significantly
increase the number of cases that are detected at earlier stages in human patients [15, 16]. The
first MCED test to receive an FDA breakthrough device designation (May 2019) [17], became
commercially available for use in humans in June 2021 [18]. Cell-free DNA (cfDNA) based
methods have been previously explored in the veterinary space for prognosis in canine lym-
phoid neoplasia, using cfDNA concentration measurements [19]; for detection of small geno-
mic alterations in canine histiocytic sarcoma, oral malignant melanoma, and multicentric
lymphoma, using PCR (polymerase chain reaction) techniques [20]; and for detection of mul-
tiple classes of genomic alterations (and demonstration of intra-patient spatial heterogeneity)
in tissue and plasma samples from dogs diagnosed with multiple cancer types, using next-gen-
eration sequencing [21].
Widespread cancer screening programs do not currently exist for dogs. Routine preventive
care, “wellness” physical exams, and commonly available tests (e.g., CBC, chemistry, urinalysis,
and basic imaging) are generally unable to detect canine cancers at a preclinical stage. MCED
liquid biopsy methods may offer clinical utility as a new screening paradigm in canine patients
[21, 22], allowing veterinarians to detect many different cancer types noninvasively before the
emergence of clinical signs by using the latest technologies now available for multi-cancer
screening in humans.
Beyond its use as a screening tool, liquid biopsy may also provide utility in canine patients
as an aid-in-diagnosis for cancer, particularly when traditional diagnostic testing would be
challenging. In some cases, cancer may be suspected, but pursuing a diagnosis through estab-
lished methods (e.g., tissue sampling via surgical biopsy) may be difficult or deemed too risky
due to the anatomical location of the mass or other characteristics of the procedure. In other
cases, cancer may be high on the list of differential diagnoses based on the clinical presenta-
tion, but traditional diagnostic techniques cannot be employed because a specific anatomical
location for the suspected cancer is not clinically evident.
Here, results of the CANcer Detection in Dogs (CANDiD) study are reported. CANDiD is
an international, multi-center clinical study designed to validate the performance of a novel
MCED “liquid biopsy” test using next-generation sequencing of blood-derived DNA, devel-
oped for the noninvasive detection and characterization of cancer in dogs.

Methods
The CANDiD study was based on a prospective sample collection program that enrolled 1,358
client-owned dogs, with and without cancer, at 41 clinical sites across the US, Canada, Brazil,
the Netherlands, France, and Hong Kong between November 2019 and August 2021. Collec-
tion sites included veterinary specialty practices, university/academic veterinary hospitals, and
general practices. All subjects were enrolled under protocols that received Institutional Animal
Care and Use Committee (IACUC) or site-specific ethics approval, according to each site’s
requirements. All subjects were client-owned, and written informed consent was obtained
from all owners.

Criteria for inclusion, exclusion, and characterization of subjects

Blood samples were prospectively collected from an all-comers cohort of dogs with confirmed
or suspected malignancy at the time of enrollment; only subjects in which cancer was defini-
tively diagnosed were ultimately included in the validation. Additionally, blood samples were
collected from an all-comers cohort of dogs who were presumed to be cancer-free due to no

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 3 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

history of cancer and no suspicion of cancer, based on a thorough clinical history and physical
exam by the treating veterinarian at the time of study enrollment; these “presumably cancer-
free" subjects were allowed to enroll if they had known or suspected medical conditions other
than cancer.
Dogs were eligible for study enrollment if: their owner provided written informed consent
for collection and use of their dog’s blood and clinical data; they weighed a minimum of 12
pounds (5.5 kg); they were 1 year of age or older; and they were able, in the professional opin-
ion of the managing veterinarian, to provide the amount of whole blood required for the study
(a minimum of 14 mL collected across two specialized tubes, with a minimum of 7 mL col-
lected in each tube). Dogs of all breeds were eligible to enter the study.
Dogs were excluded from enrollment in the study if they: had experienced physical trauma
(including injury, surgery, or core needle biopsy for any clinical indication) in the 7 days prior
to blood collection (routine blood collection, fine needle aspiration, and cystocentesis were not
excluded); were believed to be pregnant; or if collection of a blood sample provided an unac-
ceptable risk to the site staff and/or the dog, in the professional opinion of the managing veteri-
narian. No subjects were excluded based on known or suspected comorbidities, including
acute and chronic inflammatory, infectious, autoimmune, degenerative, or other conditions.
Dogs were excluded from enrollment in the cancer-diagnosed cohort if they previously under-
went curative-intent surgery for cancer and were considered to be cancer-free at the time of
enrollment; or had undergone targeted or non-targeted chemotherapy, immunotherapy, radi-
ation therapy, or experimental treatment for cancer (other than steroidal or non-steroidal
anti-inflammatory drugs) within 30 days prior to enrollment.
The rationale for exclusion due to recent trauma was based on knowledge from liquid
biopsy testing in humans that tissue disruption (for example as a result of biopsy or surgery)
can lead to the release of large amounts of cell-free DNA and/or circulating tumor DNA
(ctDNA—representing the fraction of cfDNA originating from tumor cells) into the blood-
stream. In cancer-diagnosed subjects, disproportionate disruption of surrounding normal tis-
sue (with release of high amounts of normal cfDNA) could lead to a transient “dilution” of the
ctDNA fraction, potentially making cancer detection by liquid biopsy methods more difficult
[23]. Alternatively, disproportionate disruption of the tumor tissue could trigger a transient
increase in the ctDNA fraction, potentially leading to an artificially inflated rate of detection
by liquid biopsy [24, 25]. In short, tissue trauma could confound the measurement of test per-
formance in unpredictable ways. Trauma was also an exclusion criterion for presumably can-
cer-free subjects, in order to avoid enrollment of subjects where a clinical observation of
“trauma” may represent the first clinical sign of occult cancer. Given that the half-life of
cfDNA in both humans and dogs is estimated to be no longer than a few hours [26, 27], 7 days
was considered to be a reasonable exclusion period for all subjects with recent trauma. The
rationale for exclusion due to pregnancy was likewise due to prior knowledge from human
medicine that fetal or placental chromosomal abnormalities and de novo mutations are detect-
able in cfDNA [28–30]. Although the prevalence and distribution of such events during gesta-
tion (and their rates of detection in the blood of pregnant females by cfDNA methods) are not
well documented in dogs, it was deemed prudent to exclude pregnant subjects given the poten-
tial for pregnancy to confound test results.
All cancer-diagnosed subjects had complete staging, performed by the managing veterinar-
ian according to standard-of-care staging guidelines at the enrolling site for that cancer type.
Metastasis to a local or distant anatomical site was determined by tissue pathology or, in cases
where tissue sampling was not possible due to anatomical location, by imaging diagnosis. “Pos-
sible” but unconfirmed nodules or tissue changes on imaging for which the radiologist pro-
vided both malignant and benign differentials were not included as part of determining the

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 4 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

extent of disease; confirmed nodules or tissue changes determined by the radiologist to most
likely represent malignant disease were included as part of determining the extent of disease.
Subjects with cancers amenable to tissue-based diagnosis, or to imaging-based diagnosis
(such as those with tumors in anatomical locations that precluded tissue diagnosis, for example
heart base masses), were included in the analysis. Subjects that had more than one confirmed
primary cancer type were included, and the subject’s diagnosis was recorded as the union of
diagnoses. Subjects enrolled into the cancer-diagnosed cohort whose tumors were ultimately
determined to be benign by pathology were excluded from analysis.
Cancer size as well as cancer stage are correlated with detection by NGS-based liquid biopsy
methods in humans. This study was designed to allow for evaluation of such correlations in
dogs. The longest diameter of the largest lesion was measured and recorded for the vast major-
ity of cancer-diagnosed subjects. Furthermore, simplified definitions were developed to allow
for classification of extent of disease in cancer-diagnosed subjects, given that the process of
cancer staging is less standardized in dogs than it is in humans, and many canine cancer types
have distinct staging methodologies [31, 32]. Localized/regional was defined as cancer that was
limited to the organ of origin or to nearby lymph nodes, tissues, or organs; or lymphomas lim-
ited to a single lymph node (Stage I) or multiple lymph nodes on one side of the diaphragm
(Stage II). Disseminated/metastatic was defined as cancer that had spread to areas of the body
distant from the primary tumor; or lymphomas that involved two or more lymph nodes on
both sides of the diaphragm and/or one or more extra-nodal sites (Stages III, IV, and V); or
any non-lymphoma hematological malignancy. Undetermined was used in a small number of
cases where it was not possible to accurately determine the extent of disease, despite a complete
cancer staging workup. These definitions allowed for all cancer-diagnosed cases (whether solid
or hematological) to be classified by extent of disease.
Clinical data for each visit were collected on standardized case report forms (CRFs). Full
source-document collection (including but not limited to clinical intake and progress notes,
imaging reports, surgery reports, and histopathology reports) and verification were performed
for each subject. Queries were issued to clinical sites to clarify or correct data items, as appro-
priate, before clinical database lock. Final clinical determinations, such as benign versus malig-
nant tumor and cancer type, stage, and size, were performed by board-certified veterinary
medical oncologists on the central study team and recorded prior to unblinding of liquid
biopsy testing results. This independent clinical adjudication process was informed by all of
the written clinical information available, including the diagnosis provided by each subject’s
managing veterinarian(s); however, primary data derived from imaging tests (e.g., actual
images) and pathology tests (e.g., actual slides or pictures of slides) were not independently
reviewed by the central study team.

Sample collection and laboratory procedures

Whole blood was collected from a peripheral vein (jugular, cephalic, or saphenous) using the
Cell-Free DNA Collection Tube (Roche); these specialized blood collection tubes were
designed to prevent white blood cell (WBC) lysis and cfDNA degradation for up to seven days
in transit at ambient temperature. This allowed for blood samples to be shipped to the central
testing laboratory (PetDx, La Jolla, CA) without any processing or special sample handling
(such as refrigeration, freezing, or centrifugation) at the collection site. Samples were collected
without any restrictions related to the time of day or the time of the dog’s last feeding. Two
tubes were collected from each subject, with a minimum of 7 mL of whole blood per tube. In
one subject, presented below as a case study, tumor tissue was collected from multiple anatom-
ical locations at necropsy using a 2 mm punch biopsy tool, and each biopsy specimen was

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 5 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

stored in a vial containing DNA Shield (Zymo Research) stabilization solution and shipped to
the central testing laboratory at ambient temperature.
Subjects who were unable to provide at least two tubes of blood, or whose tubes did not
meet minimum fill requirements, were excluded from analysis. Samples that were received or
processed more than 7 days from the time of collection were also excluded from analysis.
Upon receipt at the central lab, blood samples were processed with a double-centrifugation
protocol to separate plasma from WBCs [21], and were scored on the extent of hemolysis and
lipemia. Plasma aliquots, WBC pellets, and tissue samples were stored at -80˚C until they were
thawed for testing. Samples were tested between February 2021 and September 2021 and were
randomized for laboratory processing across batches, operators, and reagent lots, for the
avoidance of bias.
Cell-free DNA was extracted from plasma using a proprietary bead-based chemistry opti-
mized to maximize cfDNA yield in canine subjects. Genomic DNA (gDNA) was extracted
from WBCs, and from tissue samples, using QIAamp DNA Mini Blood Kit (Qiagen). Ampli-
fied DNA libraries were generated for each subject from the matched cfDNA and gDNA
extracts. Libraries were prepared by incorporating universal adapters and barcodes into sam-
ple DNA via ligation and universal PCR amplification, and were subjected to next-generation
sequencing on an Illumina NovaSeq 6000 for somatic variant analysis. All sequencing reads
were aligned to the CanFam3.1 reference genome [33] using the BWA-MEM algorithm [34]
with default parameters. Somatic variant calling from cfDNA and gDNA sequence data was
performed using a custom bioinformatics pipeline leveraging Sentieon TNscope [35],
ichorCNA [36], and internally-developed algorithms that are based on the coverage at specific
genomic regions or on fragmentomics profiles. CNV profiles were evaluated using 1 Mb bins.
Small variant (single nucleotide variants and insertion/deletion variants) annotation was per-
formed using Ensembl Variant Effect Predictor (VEP) Release 103 [37]. Liquid biopsy testing
results for each subject underwent dual blinded review, with independent adjudication for dis-
crepant cases prior to final reporting. All analyses were performed on the Google Cloud Plat-
form leveraging the Cloud Life Sciences API.

Analytical repeatability and reproducibility

Prior to testing prospectively collected samples for clinical validation of the test, analytical
repeatability and reproducibility of the test were assessed using contrived samples. Two estab-
lished canine cell lines, MDCK.1 (ATCC CRL-2935) and Cf2Th (ATCC CRL-1430), were
sourced from American Type Culture Collection (ATCC) and were sequenced to confirm
baseline variant profiles. DNA was extracted from these cell lines and from the white blood
cells (WBC) of a single cancer-free canine subject, respectively. Extracted DNA was frag-
mented, and cell line-derived DNA was then serially diluted into WBC-derived DNA at
defined genomic equivalent ratios. Repeatability at each dilution level was assessed by analyz-
ing agreement among multiple within-run replicates, processed by the same operators under
the same conditions. Reproducibility at each dilution level was assessed by analyzing agree-
ment among replicates across multiple runs, operators, days, and reagent lots. The repeatabil-
ity and reproducibility of the test were each >95% at all mixing ratios tested.

Training and testing sets for clinical validation

For clinical validation, cancer-diagnosed and presumably cancer-free subjects were randomly
assigned to training and testing sets in a 1:4 ratio: that is, approximately 20% of subjects were
assigned to training and 80% to testing. Population-level sex, weight, and age matching
between the cancer-diagnosed and presumably cancer-free cohorts was enforced in the

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 6 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

training set to limit the potential for bias due to demographic variables. This resulted in pre-
sumably cancer-free subjects in the training set having a higher median age compared to pre-
sumably cancer-free subjects in the testing set, since cancer subjects tended to be older than
cancer-free subjects overall. Bioinformatics algorithms were optimized and locked based on
analysis of the training set, and the final pipeline was subsequently applied to the separate test-
ing set for independent validation of the test’s clinical performance [38, 39]. Blood samples
from all cancer-diagnosed and presumably cancer-free subjects, across the training and testing
sets, were processed using the same laboratory workflows.

Test performance: Sensitivity and specificity

Test sensitivity was defined as the percentage of all cancer-diagnosed dogs in the testing set
who received a Cancer Signal Detected (positive) result. Test specificity was defined as the per-
centage of all presumably cancer-free dogs in the testing set who received a Cancer Signal Not
Detected (negative) result. Binomial confidence intervals were calculated for all performance
estimates.
Overall test performance was evaluated using the testing set and included subjects with a
single primary cancer as well as subjects with multiple concurrent primary cancers. A sub-
group analysis was performed on cancer-diagnosed subjects from the testing set representing a
predefined list of three of the most aggressive canine cancers: lymphoma, hemangiosarcoma,
and osteosarcoma. Another subgroup analysis focused on cancer-diagnosed subjects from the
testing set representing a predefined list of eight common canine cancers [40] that account for
most cancer deaths in the species: lymphoma, hemangiosarcoma, osteosarcoma, soft tissue sar-
coma, mast cell tumor, mammary gland carcinoma, anal sac adenocarcinoma, and malignant
melanoma. Separately, detection rates by cancer type were determined for a large number of
distinct cancer types using data from all cancer-diagnosed subjects across the training and test-
ing sets. The detection rate for the subgroup of subjects diagnosed with multiple concurrent
primary cancers was also determined in this manner.
Additional analyses were performed to determine the effect of pre-analytical factors such as
the time difference between sample collection and processing (time in-transit), and the levels
of hemolysis and lipemia, on the test’s performance in the cancer-diagnosed as well as the pre-
sumably cancer-free subjects in the testing set. Hemolysis and lipemia were assessed by visual
inspection of each sample based on color and extent of turbidity, respectively, using scales of
0–4 and 0–2, respectively.

Statistical analyses
All metrics were summarized as median and range unless otherwise stated. P-values were cal-
culated using Student’s t-test for continuous variables and the Chi-squared test for categorical
variables. A p-value of <0.05 was considered statistically significant. All confidence intervals
reported are two-sided 95% binomial intervals. Analyses were performed using R package ver-
sion 4.0.5.

Results
Subject disposition
A total of 1,358 dogs were enrolled in the CANDiD study. Of these, 51 subjects were excluded
due to enrollment protocol deviations or clinical criteria, such as: age <1 year at time of enroll-
ment; subject initially presumed to have cancer but a definitive diagnosis could not be con-
firmed, or the mass was determined to be benign by pathology; or subject initially presumed to

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 7 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

Fig 1. Disposition of subjects in the CANDiD study.

https://doi.org/10.1371/journal.pone.0266623.g001

be cancer-free but received a diagnosis of cancer between the time of blood collection and
time of analysis. Additionally, 152 subjects were excluded due to deviations from the standard-
ized laboratory workflow, such as: time from collection to processing >7 days; low blood col-
lection tube fill volume; or only one blood collection tube received. From the remaining 1,155
subjects eligible for clinical validation testing, 55 failed testing for reasons such as: low plasma
volume, sample swap detected, DNA library failure, and sequencing-based QC failure. From
the remaining 1,100 subjects (the “validation set”), 224 were used for algorithm development
(the “training set”), and 876 were used for analysis of test performance (the “testing set”) (Fig
1). Full subject level data are included in S1 Table.

Subject characteristics
The subject characteristics for the training and testing sets are shown in Table 1. There were
no significant differences between the training set and the testing set in the proportions of:
male to female subjects; purebred to mixed-breed subjects; cancer-diagnosed to presumably-
cancer free subjects; subjects with localized/regional disease; and subjects with a tumor diame-
ter measuring �5 cm. Likewise, there was no significant difference between the weights of the
subjects in the training set versus the testing set. There was a statistically significant overall dif-
ference between the ages of subjects in the two sets, with the training set representing a slightly
older cohort (median 7.3 vs 6.6 years, p<0.0001), as expected due to the enforced age matching
in the training set, as described above.

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 8 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

Table 1. Comparison of subject demographics and cancer characteristics for the training and testing sets.
Training set (n = 224) Testing set (n = 876) p-value
Age1 Median (years) 7.3 6.6 <0.0001
Range (years) 1.0–15.6 1.0–15.8
Weight1 Median (kg) 28.4 29.1 0.2111
Range (kg) 5.9–81.7 6.0–106.8
Sex2 Male 107 (48%) 463 (53%) 0.1989
Intact 9 74
Neutered 98 388
Not Provided 0 1
Female 117 (52%) 413 (47%)
Intact 12 42
Spayed 105 371
Breed2 Purebred 113 (50%) (40 distinct breeds) 434 (50%) (78 distinct breeds) 0.8679
Mixed-breed 111 (50%) 442 (50%)
Cancer status2 Cancer-diagnosed 81 (36%) 352 (40%) 0.3064
Presumably cancer-free 143 (64%) 524 (60%)
Extent of disease2 Localized/regional3 (% of cancer-diagnosed subjects) 50/81 (62%) 204/352 (58%) 0.1521
2
Tumor size Tumor diameter �5 cm4 (% of cancer-diagnosed subjects) 47/81 (58%) 185/352 (53%) 0.6727
1
Significance assessed using t-test.
2
Significance assessed using Chi-squared test.
3
Localized/regional was defined as cancer that was limited to the organ of origin or to nearby lymph nodes, tissues, or organs; or lymphomas limited to a single lymph
node (Stage I) or multiple lymph nodes on one side of the diaphragm (Stage II).
4
This measurement was based on the longest diameter of the largest lesion in each subject.

https://doi.org/10.1371/journal.pone.0266623.t001

The 224 subjects in the training set included 113 purebred dogs (as provided by the dog
owners) representing 40 distinct breeds, and 111 mixed-breed dogs; 107 males and 117
females; median age 7.3 years (range: 1.0–15.6 years); and median weight 28.4 kg (range: 5.9–
81.7 kg). The training set included 81 cancer-diagnosed subjects and 143 presumably cancer-
free subjects (Table 1).
The 876 subjects in the testing set included 434 purebred dogs (as provided by the dog own-
ers) representing 78 distinct breeds, and 442 mixed-breed dogs; 463 males and 413 females;
median age 6.6 years (range: 1.0–15.8 years); and median weight 29.1 kg (range: 6.0–106.8 kg).
The testing set included 352 cancer-diagnosed subjects and 524 presumably cancer-free sub-
jects (Table 1).
In the testing set, there were no significant differences between the presumably cancer-free
subjects and the cancer-diagnosed subjects in the proportion of male to female subjects and
purebred to mixed-breed subjects. Likewise, there was no significant difference between the
weights of presumably cancer-free and cancer-diagnosed subjects. There was a statistically sig-
nificant difference between the ages of subjects in the cancer-diagnosed cohort and the pre-
sumably cancer-free cohort, with the cancer-diagnosed subjects representing an older
population (median 9.7 vs. 4.2 years, p<0.0001), as expected due to the preponderance of can-
cer diagnoses in older dogs (Table 2).
The 352 cancer-diagnosed subjects in the testing set comprised 205 (58%) localized/regional
cases, 136 (39%) disseminated/metastatic cases, and 11 (3%) cases where the extent of disease
was undetermined.
Full lists of the breeds and cancer types represented in the CANDiD study (across the train-
ing and testing sets) are provided in Tables 3 and 4, respectively.

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 9 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

Table 2. Demographics of presumably cancer-free vs. cancer-diagnosed subjects in the testing set.
Presumably cancer-free (n = 524) Cancer-diagnosed (n = 352) p-value
Age1 Median (years) 4.2 9.7 <0.0001
Range (years) 1.0–15.0 1.9–15.8
1
Weight Median (kg) 28.5 29.7 0.3844
Range (kg) 6.0–106.8 6.9–67.3
Sex2 Male 273 (52%) 190 (54%) 0.6612
Intact 52 22
Neutered 221 167
Not Provided 0 1
Female 251 (48%) 162 (46%)
Intact 31 11
Spayed 220 151
Breed2 Purebred 260 (50%) 174 (49%) >0.9999
Mixed-breed 264 (50%) 178 (51%)
1
Significance assessed using t-test.
2
Significance assessed using Chi-squared test.

https://doi.org/10.1371/journal.pone.0266623.t002

Overall performance
In the 876 subjects in the testing set, there were 202 Cancer Signal Detected (positive) results
and 670 Cancer Signal Not Detected (negative) results. In four subjects (one cancer-diagnosed
subject and three presumably cancer-free subjects), genomic alterations were detected but
their significance was uncertain; these subjects received Indeterminate results and were
excluded from analysis of test performance.
There were 351 cancer-diagnosed subjects in the testing set that received a positive or a neg-
ative result. In these subjects, the test returned a Cancer Signal Detected (positive) result for
192 subjects, for an overall sensitivity (detection rate) of 54.7% (192/351; 95% CI: 49.3–60.0%)
(Table 5). There was no significant difference in test sensitivity based on age, weight, sex, or
breed status (purebred vs mixed-breed) of the cancer-diagnosed subjects (S3 Table).
There were 521 presumably cancer-free subjects in the testing set that received a positive or
a negative result. In these subjects, the test returned a Cancer Signal Not Detected (negative)
result for 511 dogs, and a Cancer Signal Detected (positive) result for 10 dogs (“putative false
positives”, pFP) (S4 Table). Two of these 10 pFP subjects were diagnosed with cancer after
undergoing a confirmatory cancer evaluation triggered by a Cancer Signal Detected result and
were excluded from analysis of test performance, as they could no longer be presumed cancer-
free. The remaining 519 presumably cancer-free subjects, including 8 subjects that received a
Cancer Signal Detected result, were used for the calculation of specificity. The overall specificity
of the test at the time of manuscript submission was 98.5% (511/519; 95% CI: 97.0–99.3%),
corresponding to a false positive rate (FPR) of 1.5% (8/519; 95% CI: 0.7–3.0%) (Table 5).
One of the two pFP subjects excluded from analysis of test performance (Subject 0374 in S4
Table) was diagnosed with hemangiosarcoma 5 months following collection of the blood sam-
ple and is described in the case study section below. In the second pFP subject excluded from
analysis (Subject 0148 in S4 Table), a Cancer Signal Detected result with “hematological malig-
nancy” signal origin prediction led to fine needle aspiration cytology of a normal-sized popli-
teal lymph node; a clonal B-cell population was suspected on PARR (PCR for Antigen
Receptor Rearrangements) but could not be confirmed due to low cellularity of the sample.
The day following the collection of the lymph node sample, the subject died due to unrelated

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 10 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

Table 3. List of 85 dog breeds represented in the training and testing sets of the CANDiD study.
A D P
Airedale Terrier Dachshund Pembroke Welsh Corgi
Akita Dalmatian Pointer
Alaskan Malamute Doberman Pinscher Pomeranian
American Staffordshire Terrier Dogo Argentino Poodle, Miniature
Anatolian Shepherd Dogue de Bordeaux Poodle, Standard
Australian Cattle Dog E Pug
Australian Shepherd English Bulldog Pyrenean Shepherd
Australian Terrier English Cocker Spaniel R
B English Setter Rat Terrier
Barbet English Springer Spaniel Rhodesian Ridgeback
Basenji F Rottweiler
Basset Hound Flat-Coated Retriever S
Beagle French Bulldog Samoyed
Belgian Malinois G Scottish Terrier
Bernese Mountain Dog German Shepherd Shetland Sheepdog
Bloodhound German Shorthaired Pointer Shih Tzu
Border Collie German Wirehaired Pointer Siberian Husky
Boston Terrier Golden Retriever Silken Windhound
Bouvier des Flandres Gordon Setter Small Munsterlander
Boxer Great Dane Soft Coated Wheaten Terrier
Boykin Spaniel Great Pyrenees Saint Bernard
Brittany Greater Swiss Mountain Dog Stabyhoun
Bull Terrier Greyhound Staffordshire Bull Terrier
Bulldog, American K V
Bullmastiff Kerry Blue Terrier Vizsla
C L W
Cairn Terrier Labrador Retriever Weimaraner
Cane Corso M Welsh Springer Spaniel
Cardigan Welsh Corgi Mastiff West Highland White Terrier
Cavalier King Charles Spaniel McNab Whippet
Chesapeake Bay Retriever Miniature Schnauzer
Chihuahua O
Chinese Shar-Pei Old English Sheepdog
Cocker Spaniel
Collie
https://doi.org/10.1371/journal.pone.0266623.t003

aspiration pneumonia. Necropsy did not reveal evidence of lymphoma; however, PARR per-
formed on spleen and intra-abdominal lymph node samples collected at necropsy revealed a
clonal B-cell population consistent with a diagnosis of lymphoma, 19 months after collection
of the blood sample.
In addition to the 10 pFP subjects described above, 2 subjects that had originally been
enrolled as presumably cancer-free received Cancer Signal Detected results, but were indepen-
dently diagnosed with cancer prior to receiving the liquid biopsy test results: one patient with
a monostotic aggressive bone lesion consistent with a malignant primary bone tumor at 1
month, and one with lymphoma at 6 months, respectively, following collection of the blood
samples. These 2 subjects were excluded from the validation set because they could no longer
be considered as presumably cancer-free.

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 11 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

Table 4. List of all cancer types represented in cancer-diagnosed subjects from the training and testing sets of the
CANDiD study.
A M
Abdominal Cavity Malignant Melanoma
Adrenal Gland� Mammary Gland Carcinoma
Anal Sac Adenocarcinoma Mast Cell Tumor
B Multiple Myeloma�
Bile Duct N
Bone, Fibrosarcoma� Nasal Cavity and Paranasal Sinuses
Bone, Multilobular Osteochondrosarcoma� Nasal Planum�
Bone, Osteosarcoma O
Brain Oral Cavity
C Ovary�
Chondrosarcoma P
E Pancreas, Endocrine�
Ear Canal Peripheral Nerve Sheath
H Pituitary�
Heart Base Prostate�
Hemangiosarcoma S
Histiocytic Sarcoma Salivary Gland
K Skin
Kidney Soft Tissue Sarcoma
L Spinal Cord�
Large Intestine �
Stomach§
Leukemia, Acute Lymphoid (ALL) T
Leukemia, Chronic Lymphoid (CLL) Thymoma�
Liver Thyroid
Lung Transmissible Venereal Tumor
Lymphoma, Indolent U
Lymphoma, Intermediate to Large Cell Urinary Bladder / Urethra
�
Cancer types for which no cancer signal was detected in any cancer-diagnosed subjects in the training and testing
sets in the CANDiD study.
§
Present in one subject with a Cancer Signal Detected result that had one other concurrent primary cancer type.
Cancer-diagnosed subjects in the training and testing sets in the CANDiD study were assigned to one of 42 cancer
types (listed above), based primarily on anatomical location. This simplified classification was adapted from Withrow
and MacEwen’s Small Animal Clinical Oncology (Sixth Edition) and from the American Joint Committee on Cancer
(AJCC) Manual (Eighth Edition).
This simplified list was derived from a more detailed list of 82 cancer subtypes that were additionally defined based
on anatomical sub-location and/or histology, as shown in S2 Table.

https://doi.org/10.1371/journal.pone.0266623.t004

In summary, among 12 presumably cancer-free subjects who received a Cancer Signal

Detected result, two received a cancer diagnosis based upon a confirmatory cancer evaluation
triggered by the liquid biopsy test result, and 2 had been independently diagnosed with cancer
prior to receiving the liquid biopsy test result. Of the remaining 8 pFP subjects, 4 died without
a confirmed diagnosis of cancer, and 4 had cancer evaluations but cancer was not found and
these patients continue to be monitored for evidence of cancer. Because a diagnosis of cancer
could not be confirmed for these 8 pFP subjects as of the time of manuscript submission, they
were considered false positives for the purpose of determining test specificity.

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 12 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

Table 5. Test results and performance in the testing set of the CANDiD study.
Testing Set Test Results Test Performance§
�
Cancer-diagnosed subjects (n = 352) Cancer Signal Detected 192/351 Sensitivity 54.7% (95% CI: 49.3–
60.0%)
Presumably cancer-free subjects Cancer Signal Not Detected 511/ Specificity 98.5% (95% CI: 97.0–
(n = 524) 519�� 99.3%)
�
One of the cancer-diagnosed subjects received an Indeterminate test result and was excluded from analysis of test
performance.
��
Three of the presumably cancer-free subjects received Indeterminate test results and were excluded from analysis of
test performance. Additionally, two of the presumably cancer-free subjects received Cancer Signal Detected results
and were diagnosed with cancer following confirmatory cancer evaluations; these two subjects were also excluded
from analysis of test performance.
§
There was no significant difference in Test Performance between the testing set and the training set. Training set:
detection rate 49.4% (p = 0.4583), true negative rate 97.2% (p = 0.5204).

https://doi.org/10.1371/journal.pone.0266623.t005

Performance in three of the most aggressive canine cancers

Lymphoma, hemangiosarcoma, and osteosarcoma are highly aggressive canine cancers. In the
testing set, there were 137 subjects with a single primary cancer belonging to one of these three
cancer types; all received a positive or negative result. The liquid biopsy test returned a positive
result in 117 of these 137 subjects, resulting in an overall detection rate of 85.4% (95% CI:
78.4–90.9%) across these three aggressive cancer types (Table 6).

Performance in eight of the most common canine cancers

Certain types of cancer are common in dogs and account for most of the cancer deaths in the
species, namely: lymphoma, hemangiosarcoma, osteosarcoma, soft tissue sarcoma, mast cell
tumor, mammary gland carcinoma, anal sac adenocarcinoma, and malignant melanoma [40].
In the testing set, there were 237 subjects with a single primary cancer belonging to one of
these eight cancer types; of these, 236 received a positive or negative result. The liquid biopsy
test returned a positive result in 146 of these 236 subjects, resulting in an overall detection rate
of 61.9% (95% CI: 55.3–68.1%) across these eight common cancer types (Table 7).

Performance by cancer type

Among the 432 cancer-diagnosed subjects across the training and testing sets that received a
positive or a negative result, there were 416 subjects with a confirmed diagnosis of a single pri-
mary cancer, representing 40 cancer types (Table 4); and 16 subjects with confirmed diagnoses
of multiple concurrent primary cancers representing a total of 17 distinct cancer types, 2 of
which were only observed in these 16 subjects (S5 Table). Samples from all cancer-diagnosed
subjects in the study were processed using the same standardized workflow and analyzed
using the same bioinformatics pipeline. The detection rates of the test by cancer type across all
cancer-diagnosed subjects in the training and testing sets are shown in Fig 2.

Table 6. Test results and performance in the testing set for three of the most aggressive canine cancers.
Three of the Most Aggressive Canine Cancers (Lymphoma, Hemangiosarcoma, Osteosarcoma)
Testing Set Test Results Test Performance
Number of subjects tested (n = 137) Cancer Signal Detected 117/137 Detection rate 85.4% (95% CI: 78.4–90.9%)
https://doi.org/10.1371/journal.pone.0266623.t006

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 13 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

Table 7. Test results and performance in the testing set for eight of the most common canine cancers.
Eight of the Most Common Canine Cancers (Lymphoma, Hemangiosarcoma, Osteosarcoma, Soft Tissue
Sarcoma, Mast Cell Tumor, Mammary Gland Carcinoma, Anal Sac Adenocarcinoma, Malignant Melanoma)
Testing Set Test Results Test Performance
Number of subjects tested (n = 237) Cancer Signal Detected 146/236� Detection rate 61.9% (95% CI: 55.3–68.1%)
�
One of the cancer-diagnosed subjects in the testing set (diagnosed with mast cell tumor) received an Indeterminate
test result and was excluded from analysis of test performance.

https://doi.org/10.1371/journal.pone.0266623.t007

Performance by extent of disease and tumor size

The test’s detection rate was calculated as a function of extent of disease (localized/regional or
disseminated/metastatic) and tumor size (�5 cm or >5 cm, for the longest diameter of the
largest lesion). Results are shown in Fig 3. Detection rate increased as a function of stage and
tumor size, with localized/regional disease with tumor size �5 cm having the lowest perfor-
mance at 19.6% (22/112; 95% CI: 12-7-28.2%) and disseminated/metastatic disease with tumor
size >5 cm having the highest performance at 87.5% (58/70; 95% CI: 72.0–90.8%). Subjects for
which extent of disease and/or tumor size was not documented or could not be determined
showed a detection rate of 66.7% (22/33; 95% CI: 48.2–82.0).

Cancer Signal Origin (CSO) prediction

A CSO algorithm for hematological malignancy was developed based on genomic features
associated with hematological malignancies, such as copy number gains in chr13 and/or
chr31, copy number losses in chr14 [41, 42], and/or penetration of somatic genomic alter-
ations into gDNA. The algorithm was developed using subjects diagnosed with hematological
malignancies in the training set; the algorithm was then applied to subjects in the testing set
who received a Cancer Signal Detected result. A total of 96 subjects with a diagnosis of hemato-
logical malignancy in the testing set received a Cancer Signal Detected result. Of these 96 sub-
jects, 40% (n = 38) were also assigned a CSO prediction of “hematological malignancy”, while
the remaining 58 did not receive a CSO prediction. Three subjects who had been diagnosed
with non-hematological malignancies were additionally assigned a CSO prediction of “hema-
tological malignancy”, for an overall CSO prediction accuracy of 92.7% (38/41).

Test performance as a function of pre-analytical variables

Neither the sensitivity nor the specificity of the test had a significant association with any of
the following pre-analytical variables in the testing set: time from collection to processing,
extent of hemolysis, and extent of lipemia (S6 Table).

Case study: Liquid biopsy reveals cancer signal months prior to onset of
clinical signs
The following case study presents an adjudicated putative false positive (pFP) result and illus-
trates the ability of NGS-based liquid biopsy to detect cancer-associated genomic alterations in
blood months before the onset of clinical signs.
The subject was a 7-year-old, 37 kg female spayed mixed-breed dog initially enrolled in the
presumably cancer-free cohort in December 2020. At the time of enrollment, the subject had
no suspicion of cancer based on physical exam findings or clinical signs noted by the owner.
Blood was collected, and plasma and WBCs were separated upon sample receipt and stored at
-80˚C until liquid biopsy testing was performed in April 2021; at that time, a Cancer Signal

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 14 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

Fig 2. Detection rates by cancer type.

https://doi.org/10.1371/journal.pone.0266623.g002

Detected result was issued. Genomic alterations detected in cfDNA from the plasma sample
included copy number variants (CNVs) across the genome, and single nucleotide variants
(SNVs) in two genes that are frequently mutated in human and canine cancers: TP53 and
PIK3CA [43]. Both mutations, TP53 R265Q and PIK3CA N1044K, have human orthologues
(TP53 R213Q and PIK3CA N1044K, respectively) that have been observed in multiple human
cancers as recorded in the COSMIC (Catalogue Of Somatic Mutations In Cancer) database
[44]. In particular, PIK3CA N1044K is a known oncogenic mutation that is targetable in
human breast cancer by an FDA-approved drug [45]. At the time of receipt of the Cancer Sig-
nal Detected result in April 2021, the subject continued to have a normal physical exam and no
clinical signs noted by the owner. Because of the Cancer Signal Detected result, the subject
underwent a confirmatory cancer evaluation in May 2021, five months after the initial blood
sample was collected. Three-view thoracic radiographs identified nodules in the lungs, and

Fig 3. Detection rates by extent of disease and tumor size.

https://doi.org/10.1371/journal.pone.0266623.g003

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 15 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

metastatic neoplasia was suspected. A thoracic CT was subsequently performed, which con-
firmed the presence of the nodules in the lungs and revealed a cavitated mass in the left ventri-
cle of the heart. Ultrasound-guided fine needle aspiration cytology of one lung nodule was
consistent with probable sarcoma (suspect hemangiosarcoma). The subject was managed with
palliative care and remained asymptomatic. Additional blood samples were collected in May
2021 and July 2021 for repeat liquid biopsy testing; both tests confirmed the previously
detected genomic alterations in plasma. Mild clinical signs (e.g., intermittent lethargy) were
first noted in July 2021, seven months after molecular signs of cancer were measurable in the
blood. Clinical monitoring revealed progression of pulmonary nodules, the development of a
cavitated mass in the spleen, and mild pericardial effusion. The subject did not experience clin-
ical signs consistent with acute blood loss or tamponade and continued to be managed pallia-
tively. Elective euthanasia was pursued in August 2021 due to clinical signs of progressive
lethargy, eight months following the collection of the initial blood sample. Necropsy confirmed
a splenic mass, rib mass, diffuse pulmonary nodules, and a multilobular mass infiltrating the
muscle of the entire heart and extending through the pericardium; a histopathologic diagnosis
of hemangiosarcoma was subsequently confirmed in all tissue sites. Further genomic analysis
of the tissue samples confirmed the presence of the genomic alterations previously identified
in plasma. To the authors’ knowledge, this is the first known reported case in which cancer-
associated genomic alterations were shown to be present and detectable in blood several
months prior to the development of clinical signs of cancer in a canine patient; and in which a
positive result from an NGS-based liquid biopsy test triggered a cancer workup in a preclinical
patient that ultimately resulted in a confirmed diagnosis of cancer (Fig 4).

Discussion
A novel, NGS-based MCED liquid biopsy test has demonstrated the ability to identify cancer-
associated genomic alterations in canine patients representing a wide spectrum of cancer
types. To the authors’ knowledge, this is the first NGS-based test in the veterinary space that
leverages a robust training and testing study design for algorithm development and perfor-
mance validation, respectively. A large, representative training set was used to optimize and
lock the test’s algorithms and thresholds. This locked pipeline was separately applied to an
independent testing set for validation of clinical performance. Use of an independent testing
set for validation is critical, as reporting performance from the same dataset used in develop-
ment may result in inflated estimates of test performance due to overfitting. Validation on a
fully independent testing set demonstrates robust generalizability of the test’s clinical perfor-
mance characteristics [38, 39]. Cumulatively, the training and testing sets comprised over
1,000 canine subjects. To the knowledge of the authors, this is the largest clinical validation
study ever published in veterinary cancer diagnostics.

Clinical sensitivity and specificity

In the testing set, the test was able to detect three of the most aggressive canine cancers (lym-
phoma, hemangiosarcoma, osteosarcoma) at an overall detection rate of 85.4%. The test also
demonstrated an overall detection rate of 61.9% in the testing set for a subset of eight com-
monly encountered canine cancers: lymphoma, hemangiosarcoma, osteosarcoma, soft tissue
sarcoma, mast cell tumor, mammary gland carcinoma, anal sac adenocarcinoma, and malig-
nant melanoma. The overall sensitivity of the test across all cancer-diagnosed subjects in the
testing set was 54.7%. The majority (58%) of cancer-diagnosed subjects had localized/regional
disease.

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 16 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

Fig 4. Case study of a presumably cancer-free 7-year-old, 37 kg female spayed mixed-breed dog that received a Cancer Signal Detected result and
was subsequently diagnosed with hemangiosarcoma. (A) Timeline of NGS-based testing of blood and tissue samples, and the corresponding clinical
assessments for the subject. (B) Sequencing data showing no genomic alterations in gDNA derived from white blood cells. (C) Sequencing data showing
genomic alterations present in cfDNA from plasma samples obtained at multiple timepoints. (D) Sequencing data showing genomic alterations present
in various tissue samples collected during necropsy.
https://doi.org/10.1371/journal.pone.0266623.g004

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 17 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

These detection rates were achieved at a specificity of 98.5% in the testing set, correspond-
ing to a false positive rate of 1.5%. This compares favorably to false positive rates in the range
of 3.5% to >10% observed historically in screening programs focused on specific cancer types
in human patients [46–49]. High specificity is important because false positives can impose
significant financial and psychological costs, and can lead to adverse events resulting from
unnecessary diagnostic interventions. For these reasons, two MCED tests recently developed
for annual cancer screening in humans have been optimized for specificity (>99.0% in both
cases) [50, 51].
Importantly, the specificity of the test was not determined in a cohort of “healthy” dogs.
Rather, control subjects were enrolled as “presumably cancer-free” based on having no history
of cancer and no suspicion of cancer on clinical evaluation at the time of study enrollment;
there were no restrictions on the type or number of medical conditions (other than confirmed
or suspected cancer) that could be present in these subjects. The use of “presumably cancer-
free" rather than “healthy” subjects was deliberately chosen in order to document the real-
world specificity of the test, given that many patients in whom such a test might be used (for
screening in older dogs at higher risk of cancer, or for aid-in-diagnosis in dogs suspected of
cancer based on clinical presentation) may have other concurrent medical conditions. Achiev-
ing a very low false positive rate (1.5%) in a large control group of study subjects with a variety
of non-malignant medical conditions was made possible by the fact that NGS-based
approaches rely on detection of genomic alterations that are associated with cancer and are
not typically encountered in other disease states. The test’s high specificity for cancer also
allowed for unambiguous, binary reporting of results (Cancer Signal Detected or Cancer Signal
Not Detected) as opposed to using a continuous risk score or multiple risk categories, which
can create interpretation challenges in the clinic.

Tumor size and extent of disease

The relationship between cancer size and/or spread, and likelihood of detection by liquid
biopsy, has been well established in human cancers [52, 53]. A large clinical validation study of
an MCED test in humans recently demonstrated the following sensitivity performance (at a
specificity of 99.5%): 16.8% for Stage I, 40.4% for Stage II, 77.0% for Stage III, and 90.1% for
Stage IV cases [50]. A similar trend was observed in the current study when comparing test
sensitivity across extent of disease and tumor size categories in dogs (at a specificity of 98.5%):
19.6% and 51.3% for localized/regional cases with �5 cm and >5 cm lesions, respectively;
82.9% and 87.5% for disseminated/metastatic cases with �5 cm and >5 cm lesions, respec-
tively; and 66.7% for cases in which extent of disease and/or tumor size was not documented
or could not be determined (Fig 3). The measurement of tumor size relied on the longest
diameter of the single largest lesion in each cancer-diagnosed patient; this metric was a reliable
indicator of total tumor burden in localized/regional cases, but not in disseminated/ metastatic
cases–where subjects typically had multiple malignant lesions, and the size of the single largest
lesion did not provide an accurate reflection of overall tumor burden.
These findings have implications for the clinical utility of this type of testing in both the
screening and the aid-in-diagnosis scenarios. For example, a smaller (�5 cm) lesion that is sus-
pected to represent localized malignancy and is easily accessible by biopsy or fine needle aspi-
ration (FNA) should likely be pursued with conventional tissue sampling rather than with
liquid biopsy, given the lower sensitivity of the latter for smaller, localized cancers. However,
in a screening situation where there is no prior suspicion of cancer, or in a scenario where a
suspicious lesion is identified on imaging in a difficult-to-access anatomical location, a test
that can detect smaller localized cancers with high specificity through a simple blood draw

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 18 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

could offer significant clinical utility even at lower levels of sensitivity, given the lack of
alternatives.
This utility may extend to certain disseminated/metastatic cases. Dogs may not show clini-
cal signs until cancer has advanced to a stage where it is no longer curable at the time of diag-
nosis. In such situations, preclinical detection of disseminated/metastatic disease may
nevertheless provide significant utility to the clinician and the pet owner. For example, it can
help to shorten the path to diagnosis, or allow the diagnostic workup to take place without the
time constraints that may exist when patients present with acute clinical signs–as, for example,
in many hemangiosarcoma cases wherein cancer is detected after development of potentially
life-threatening hemorrhage; it may also allow for palliative care to be initiated earlier, for
improved quality of life; and it might give the family more time to make important medical
management decisions for their pet. The case study presented above (Fig 4) illustrates these
multiple elements of clinical and personal utility in a canine patient with metastatic disease
who was diagnosed at a preclinical stage following a Cancer Signal Detected liquid biopsy test
result.

Positive and negative predictive values

Though sensitivity and specificity are important test performance metrics, the positive predic-
tive value (PPV) associated with an abnormal result is arguably a more relevant number for cli-
nicians, as it describes the probability that a particular patient has cancer, given the positive
liquid biopsy test result and his/her unique clinical presentation. Similarly, in the case of a neg-
ative result, the negative predictive value (NPV) helps the clinician determine the level of reas-
surance that can be provided regarding the absence of cancer.
Positive predictive values (PPV) and negative predictive values (NPV) can be calculated
based on the estimated prevalence of cancer in the intended use population(s), using the fol-
lowing standard formulas: PPV ¼ ðsensitivity x prevalenceÞ=½sensitivity x prevalence þ ð1
specificityÞ x ð1 prevalenceÞ� and NPV ¼ ½specificity x ð1 prevalenceÞ�=½ð1
sensitivityÞ x prevalence þ specificity x ð1 prevalenceÞ�
The test is currently intended for use in dogs who are at higher risk for cancer: as an annual
screening test for dogs at higher risk of cancer due to age and/or breed; and as an aid-in-diag-
nosis for dogs in which cancer is suspected based on clinical signs or other clinical findings.
Though the prevalence of cancer in these high-risk populations has not been conclusively
established, it is possible to estimate the prior probability of cancer for each population from
existing data sources.
It is important to note that the term “screening” is used here in a strict sense. The American
Cancer Society (ACS) states that “Screening tests are used to find cancer before a person has
any symptoms” (emphasis in the original) [54]; and the National Cancer Institute (NCI) states
that “Checking for cancer. . . in people who have no symptoms is called screening” [55]. Veter-
inary patients do not have symptoms, but they do have clinical signs [56]. In a veterinary con-
text, cancer screening therefore refers to investigations undertaken in patients that are at
higher risk of cancer (e.g., due to age or breed) but do not have a current suspicion of cancer
based on clinical presentation. Tests performed in patients already suspected of having cancer
are not “screening” tests; the term “aid-in-diagnosis" provides a more accurate description for
test use in such clinical scenarios.
In the United States, it is estimated that 4.2 to 6 million dogs receive a new cancer diagnosis
each year [57, 58] out of a total population of approximately 65 to 77 million dogs [58, 59].
Therefore, the estimated annual incidence of canine cancer is approximately 5.5% to 9.2%
across all dogs. Because the liquid biopsy test is intended for annual screening in dogs at higher

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 19 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

Table 8. Positive and negative predictive values based on estimated prior probabilities of cancer in two intended use populations.
Positive test result Negative test result
Clinical use case Prior probability of cancer Intended use population PPV NPV
Screening 8–10% Higher risk of cancer due to age and/or breed 76–80% 95–96%
Aid-in-diagnosis 30–50% Cancer suspected based on clinical presentation 94–97% 68–84%

Estimated ranges for positive predictive value (PPV) and negative predictive value (NPV), calculated using a test sensitivity of 54.7% and specificity of 98.5% (range is
calculated using the lower and higher ends of prior probability).

https://doi.org/10.1371/journal.pone.0266623.t008

risk of cancer based on age or breed, a conservative annual incidence of cancer in this screen-
ing population may be estimated at 8–10%. Though studies of canine cancer incidence are lim-
ited, available data provide support for this estimate in higher risk populations [60].
When used as an aid-in-diagnosis, the test’s PPV is expected to increase due to the higher a
priori cancer prevalence in the population being tested. A survey conducted in June 2020 col-
lected data from over 300 veterinary general practitioners in the United States and found that
approximately 50% of dogs that present with a clinical suspicion of cancer go on to receive a
definitive or presumptive diagnosis of cancer (S1 Fig). Therefore, for the purpose of PPV and
NPV calculations in this study, a cancer prevalence of 30–50% was used for the aid-in-diagno-
sis population.
Using the established overall test performance metrics (54.7% sensitivity, 98.5% specificity),
and the estimated cancer prevalence in each of the intended use populations (8–10% in the
screening population, and 30–50% in the aid-in-diagnosis population), PPVs and NPVs can
be calculated for the liquid biopsy test. As shown in Table 8, when a positive result is issued for
a screening patient, the PPV is estimated to be 76–80%. PPV rises to 94–97% when a positive
result is issued for an aid-in-diagnosis patient, due to the increased prevalence of cancer in this
population. In the context of a negative liquid biopsy result for a screening patient, NPV is
expected to range from 95–96%, providing a high level of reassurance for the absence of cancer
at the time of screening. In the aid-in-diagnosis population, where there is already a clinical
suspicion of cancer, the prior probability is higher than in the screening setting, resulting in an
NPV in the range of 68–84%. This lower NPV range underscores the importance of further
clinical evaluation if the liquid biopsy test is negative as an aid-in-diagnosis, but cancer
remains high on the list of differential diagnoses for that patient.

Cancer screening using an MCED test

Traditional cancer screening programs in humans have focused on one cancer type per screen-
ing test. For example, a patient would undergo mammography, Pap smear, and colonoscopy
to specifically screen for breast, cervical, and colon cancer, respectively [9]; a true positive
result for a given test would detect the targeted cancer type but miss any other malignancies
that might be present in the patient’s body. In contrast, MCED tests are designed to detect as
many cancer types as possible with a single testing method, to maximize the cancer detection
rate across the entire at-risk population (number of cases detected as a fraction of all actual
cancer cases present in the population). Thus, an overall detection rate of 54.7% across 30 can-
cer types in a screening population corresponds to a larger absolute number of cancer cases
detected than the detection rate of 85.4% in a restricted set of “three of the most aggressive can-
cers” or 61.9% in a restricted set of “eight of the most common cancers”. Furthermore, in a
true screening setting (where the patient has no current suspicion of cancer) there is no firm
basis for suspecting a particular cancer type at the outset versus all other possible cancer types
that might be present at a preclinical stage in that patient. As with MCED testing in humans,

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 20 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

the first goal of screening in dogs is to determine if cancer (of any type or stage) is present in a
preclinical patient, before pursuing cancer typing and staging. These screening principles
underscore the fact that an MCED test’s PPV is more clinically relevant for any given patient
than the test’s sensitivity. As previously noted, the overall sensitivity of this test coupled with
its very high specificity for cancer support a PPV estimate above 75% in a canine high-risk
screening population; this is significantly higher than the PPVs of 5–21% that have been
reported for various clinical symptoms, signs, and non-diagnostic tests in human medicine
and are considered highly predictive for cancer in general practice settings [61].
Given these considerations, it is important to note that MCED tests are intended to be used
in a manner that fits with a patient’s overall clinical picture. These tests are not a replacement
for conventional diagnostic and staging tests, and they should not serve as the sole basis for
making important decisions such as treatment selection or euthanasia. Rather, their intent is
to assist in the detection of cancer by noninvasive means. The clinical utility of MCED testing
is the subject of ongoing research in human medicine, including evaluation of the impact on
multiple aspects of patient care and disease management [62–65]. Similar studies will be
required to understand the full clinical implications of MCED testing in veterinary medicine.

Study limitations
A potential limitation of this study is its “case-control” design, wherein all subjects used to
determine test performance had already been classified as either “cancer-diagnosed” or “pre-
sumably cancer-free” based on prior clinical evaluation. Most of the “cancer-diagnosed” sub-
jects had initially presented for veterinary care due to clinical signs of disease, and
subsequently received a definitive diagnosis of malignancy. Many subjects in a real-world
screening setting would be expected to have sub-clinical disease; therefore, the sensitivity of
the test (at a single point in time) for the screening use case is likely to be somewhat lower than
the sensitivity reported here. The currently reported sensitivity is more likely to be reflective of
real-world sensitivity in the aid-in-diagnosis use case, where patients receive the test because
they are currently suspected of cancer based on observable clinical signs. This consideration
regarding the screening use case should be balanced with the possibility that, when screened at
regular intervals (e.g., on an annual basis), dogs at higher risk of cancer might benefit from a
lifetime “cumulative detection rate” significantly higher than the sensitivity documented in
this study at just one point in time; this is because each successive test provides an additional
opportunity for detection, if cancer is indeed present [66]. In general, as cancer increases in
size and becomes more aggressive, more ctDNA will be shed into circulation, increasing the
chances of detection by liquid biopsy. The relationship between cancer size and likelihood of
detection by liquid biopsy is well established in human cancers [52, 53] and was also observed
in this study when comparing test sensitivity across extent of disease and tumor size categories
in dogs.
Of note, each subsequent test also provides an additional opportunity for a false positive to
be reported, increasing the probability that a given patient will receive a false positive screening
result at some point during a multi-year period of repeat testing [67]. This highlights the value
of using a screening test with a high specificity (low false positive rate), as well as the impor-
tance of understanding the actual window of opportunity for preclinical detection of various
cancer types in dogs. For example, one putative false positive subject in this study (Subject
0148 in S4 Table) was diagnosed with lymphoma 19 months following collection of a blood
sample that received a Cancer Signal Detected result; given the long duration of this elapsed
period, it is unclear if the positive test result was in fact related to the eventual cancer diagnosis.
In humans, the latency periods of many cancer types have been extensively researched. A

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 21 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

recent analysis estimated a latency range of 2.2 years to 35.7 years for lymphoproliferative and
hematopoietic cancers, and 6.6 years to 57 years for solid malignancies; 35 of the 44 cancer
types in this analysis were found to progress silently for 10 years or longer prior to detection,
representing 89% of the patients in the analysis [68]. This provides a considerable window of
opportunity for preclinical identification of cancer in human patients by various screening
approaches, including multi-cancer early detection (MCED) testing [15, 63]. The duration of
the preclinical phase is not well documented for any canine cancer type, but it is plausible that
many dogs may have preclinical disease for months to years before onset of clinical signs. Life-
time screening studies are needed to better understand the duration of the preclinical phase in
dogs; the performance of NGS-based liquid biopsy for detecting preclinical cancer in canine
patients; the incremental benefits of cumulative sensitivity with interval screening; and the
ultimate clinical utility of earlier detection (including impact on clinical management deci-
sions, and actual survival benefit after accounting for lead-time bias) for various cancer types.
To explore these important topics, the Cancer Lifetime Assessment Screening Study in
Canines (CLASSiC) was launched in December 2021 (PetDx, La Jolla, CA); the study aims to
prospectively follow over 1,000 initially cancer-free dogs, with semi-annual liquid biopsy test-
ing and comprehensive documentation of cancer-related clinical outcomes, over many years.
Another limitation is the fact that only one sample (collected around the time of initial can-
cer diagnosis) was tested for each cancer-diagnosed subject as part of this validation study; as a
result, the performance of the test for post-diagnosis longitudinal use cases (such as minimal
residual disease detection, recurrence monitoring, and treatment response monitoring) has
not been determined. Over 1,000 serial blood samples were obtained from cancer-diagnosed
subjects as part of the prospective collection that powered this study, along with clinical data
such as the nature and timing of therapeutic interventions, time to recurrence, and extent of
disease at recurrence. Future validation studies using these samples are planned to demon-
strate the performance of the test in disease-monitoring use cases.
Another potential limitation of this study was that clinical status designations (such as
benign vs malignant tumor, and cancer type, stage or size) were collected as provided by the
managing veterinarian(s) according to the standard of care at each collection site. This infor-
mation was independently reviewed and adjudicated by the central study team based on source
documents (including but not limited to clinical intake and progress notes, imaging reports,
surgery reports, and histopathology reports) prior to database lock. However, the original
radiologic images and pathology samples were not independently reviewed by the central
study team. Thus, there exists the possibility that the clinical status designations attributed to
each subject may not have captured the true clinical status of every subject in the study, given
the well-documented variability in interobserver concordance for imaging, pathology, and
other clinical assessment areas [69–74].
Finally, many cancer types were represented by only one or a few subjects in the all-comers
collection used in this study, making it difficult to determine test performance in those cancer
types with a high degree of precision. Further research is planned in order to study the perfor-
mance of the test in larger numbers of cases for specific cancer types, and the list of detectable
cancer types is expected to increase with testing of larger cohorts and with future enhance-
ments to the test. The value of an MCED testing approach lies in its ability to detect cancer-
associated genomic alterations across a broad range of cancer types, and to expand the set of
detectable cancers over time. As described previously, with each addition to the MCED test’s
list of detectable cancer types, both the fraction and the absolute number of cancer cases
detected in the tested population will increase (even if the test’s detection rate is low for the
individual cancer type being added) [50].

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 22 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

Challenges and potential confounders

In humans, some cancers are more challenging to detect by liquid biopsy. For instance, it has
been shown that ctDNA in blood has a lower detection rate in certain cancer types, such as
malignancies of the central nervous system (CNS) or prostate [52, 75]. This observation aligns
with the results of this study, where the detection rate in subjects with CNS tumors and pros-
tate cancer, for example, is also low. Further research is needed to determine if NGS-based liq-
uid biopsy tests can be optimized to improve detection of biologically challenging
malignancies in blood, or whether other bodily fluids offer better conduits for detection of
these cancers [76–78].
The ability to detect a malignant process with very high specificity through a simple blood
test can, by itself, provide significant value by helping the clinician narrow the differential diag-
nosis from among other potential pathological processes in a canine patient. Prediction of a
specific Cancer Signal Origin for patients with a Cancer Signal Detected result would further
enhance the utility of the test, as it would help focus the cancer evaluation process and ideally
lead to a definitive diagnosis earlier and at lower cost for the pet owner. This study demon-
strated the ability of the test to provide CSO prediction for canine hematological malignancies,
the most common of which are lymphomas. As more subjects are tested and adjudicated, and
larger datasets are accumulated for additional cancer types, bioinformatics algorithms may be
developed and optimized to provide accurate CSO predictions for additional cancer types,
including those that are less commonly encountered in dogs. Such capabilities have already
been documented for MCED tests developed for use in humans, with accurate CSO prediction
demonstrated for various cancer types in large numbers of subjects representative of each can-
cer type [50, 51]. The overall sensitivity and specificity of the test can also be expected to
increase over time, with further biochemical and bioinformatics improvements and new learn-
ings gained from testing larger sample sets.
It is important to note that certain biological factors could confound results of NGS-based
liquid biopsy tests. Viable chromosomal aneuploidies, constitutional mosaicism, genomic
alterations present in one or more fetuses of a pregnant female or in a transplanted organ
(including bone marrow), and clonal hematopoiesis of indeterminate potential (CHIP),
among others, have the potential to confound results from blood-derived DNA in humans
[28–30, 79–83]. The robust design of this test provides for parallel testing of matched gDNA
and cfDNA from each sample; the use of an intra-subject gDNA control as part of each test is
intended to mitigate the impact of certain confounders noted above. The very high specificity
observed in the large all-comers cohort of presumably cancer-free dogs in this validation study
suggests that such confounders, if present, would rarely lead to false positive results among
canine patients receiving this test in a routine veterinary practice setting.
Masses designated as “benign” by tissue pathology require special consideration as potential
confounders and present an opportunity for future research. Tumors exist on a continuum
from benign to borderline (premalignant) to malignant, and sometimes these states can coexist
(e.g., in a patient where parts of a benign tumor are undergoing malignant transformation)
[84–86]. The possibility of tumor heterogeneity across spatially separated sites makes it diffi-
cult to ascertain the absence of malignancy in any given subject found to have benign disease
based on tissue sampling from a single anatomical site, especially when interobserver variabil-
ity for benign vs malignant diagnoses (well documented across many tumor types) is addition-
ally considered [22, 87–96]. Cases with benign pathological diagnoses were excluded from
analysis in this study for two reasons. First, in the context of the above considerations, the
absence of malignancy could not be ascertained in such cases, so they could be used for neither
the calculation of sensitivity (which required cancer-diagnosed subjects) nor the calculation of

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 23 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

specificity (which required subjects without a diagnosis or a suspicion of cancer). Second, it

has been established in human oncology that patients with a history of benign masses have an
elevated relative risk of malignancy in multiple organs, such as breast [97, 98], uterus [99],
colon [100], and liver [101, 102]. For perspective, two recent large clinical validation studies
for MCED tests intended for use in humans also excluded potentially confounding cases (such
as those with high-grade dysplasia or with suspected but unconfirmed cancer status at the time
of blood collection) from their analysis, and neither study provided an analysis of subjects with
benign tumors [50, 51]. Although excluded from the current analysis, canine subjects with
benign pathological diagnoses at enrollment in this study will continue to be monitored for
long-term clinical outcomes, which will be evaluated alongside genomic analyses of both tissue
and blood samples.
This study also evaluated the impact of multiple pre-analytical factors that have the poten-
tial to confound the accuracy of test results; these factors have direct relevance for sample col-
lection and handling workflows in the clinic, and for dog owner convenience. The use of
specialized blood collection tubes allowed for whole-blood samples to be shipped to the central
lab without any processing in the clinic, and without any sample handling requirements for
temperature control in the clinic and/or in-transit. Samples were shipped at ambient tempera-
ture from many domestic and international sites located in various climates, and were
accepted for testing if they arrived at the central lab within 7 days of collection. There was no
difference in test performance as a function of days in-transit, within the acceptance period.
Hemolysis and lipemia are also common confounders for blood-based tests. Study subjects
were enrolled without any restrictions related to the timing of the dog’s last feeding, and lipe-
mia scores (which serve as a proxy for the timing of the most recent meal) [103] had no signifi-
cant association with test performance. Likewise, hemolysis scores had no significant
association with test performance. Finally, there was no significant association between test
performance and the following subject-specific characteristics: age, weight, sex (including
spay/neuter status), and purebred vs mixed-breed pedigree. Taken together, these findings
demonstrate the robustness and versatility of an NGS-based liquid biopsy test for multi-cancer
detection in real-world conditions.
Other non-biological factors could also confound test results, such as sample swap (at the
clinic or laboratory), sample contamination, laboratory processing artifacts, and bioinformat-
ics limitations, among others.
In light of the multiple biological and non-biological factors that have the potential to con-
found NGS-based liquid biopsy testing, results that are suspected to be “false positive” or “false
negative” should be comprehensively evaluated: in some cases, by repeat liquid biopsy testing,
and in all cases by incorporating the clinician’s intuition and firsthand knowledge of each
patient’s clinical situation. In particular, positive results from liquid biopsy testing in presum-
ably cancer-free dogs (such as in a screening setting) must be adjudicated by a confirmatory
cancer evaluation; as demonstrated by multiple subjects in the control cohort, including the
case study presented, liquid biopsy has the ability to detect cancer-associated genomic alter-
ations many months prior to the onset of clinical signs. Furthermore, negative liquid biopsy
results in dogs suspected of cancer on clinical grounds should be treated with caution, and a
comprehensive cancer evaluation should be completed to confirm the absence of malignancy.

Conclusion
Unlike the human diagnostics space, the veterinary diagnostics space is unregulated; as a
result, peer-reviewed publication of robust clinical validation studies for new personalized-

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 24 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

medicine solutions is critical for supporting responsible adoption and informing appropriate
clinical decision making [104].
In a large clinical validation study using an independent testing set, a novel NGS-based
MCED liquid biopsy test demonstrated the ability to identify cancer-associated genomic alter-
ations in canine patients–in some cases many months prior to the onset of clinical signs–across
a large and diverse set of cancer types. The test’s detection rate was particularly high in three of
the most aggressive, and in eight of the most common, cancer types in dogs. When added to
the veterinarian’s diagnostic toolkit, this type of noninvasive testing may enable earlier cancer
detection in screening and aid-in-diagnosis clinical scenarios, and lead to a paradigm shift in
the way cancer is diagnosed and managed in dogs.
Though further research is needed to explore the extent to which earlier cancer identifica-
tion and treatment can improve clinical outcomes in canine patients, established screening
programs for various human cancers (breast, colon, prostate, cervix, lung) [9] have historically
provided strong support for improved outcomes associated with high-compliance population-
level screening programs [105, 106]. Major veterinary organizations have also recognized the
value of earlier diagnosis: the American Animal Hospital Association (AAHA) has stated that
"early detection is critical for the best outcome” [3], while the American Veterinary Medical
Association (AVMA) has advised that “neoplasia is frequently treatable and early diagnosis
will aid your veterinarian in delivering the best care possible” [2]. A noninvasive cancer detec-
tion method that requires only a routine blood collection may lead to more early diagnoses in
dogs simply because of the convenience associated with the testing process, which might
encourage adoption as well as compliance with interval screening recommendations.
In addition to its applications in screening and aid-in-diagnosis, NGS-based liquid biopsy
testing is expected to expand in use to assist veterinarians in providing highly personalized
cancer care through targeted treatment selection, minimal residual disease detection (follow-
ing curative-intent treatment), recurrence monitoring, and treatment response monitoring
[22]. Validation studies specific to each of these post-diagnosis use cases will be required to
demonstrate the performance of liquid biopsy methods in these clinical scenarios. Finally,
canine and human cancers share many molecular features [57, 107]; therefore, genomic data
being generated from liquid biopsy testing in canine cancer subjects may provide benefits to
human cancer patients via comparative oncology analyses.

Supporting information
S1 Fig. Questions asked in a survey of over 300 veterinary general practitioners in the
United States.
(PDF)
S1 Table. Full subject level data for subjects enrolled in the CANDiD study.
(PDF)
S2 Table. Cancer types and subtypes evaluated in the CANDiD study.
(PDF)
S3 Table. Analysis of test sensitivity based on demographic characteristics of cancer-diag-
nosed subjects in the testing set.
(PDF)
S4 Table. Overview of putative false positive subjects from the testing set of the CANDiD
study.
(PDF)

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 25 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

S5 Table. Subjects with confirmed diagnoses of multiple concurrent primary cancers in

the training and testing sets (n = 16).
(PDF)
S6 Table. Analysis of test performance based on pre-analytical variables in the testing set.
(PDF)

Acknowledgments
The authors thank all the dogs in the CANDiD study and the humans who love and care for
them. Without their support and contributions, this study would not have been possible. The
authors would also like to thank Jason Loftis and Dominique Lau for assistance with develop-
ing the tables, figures, and graphics in this manuscript, as well as Vu Q. Nguyen, Amanda
Nguyen, the entire PetDx laboratory team for assistance with data generation. Additionally,
the authors would like to thank Oncovet Clinical Research, VCA Clinical Studies, the doctors
and staff at Flint Animal Cancer Center Colorado State University, the College of Veterinary
Medicine at the University of Missouri, Veterinary Specialty Hospital of San Diego, Veterinary
Specialty Hospital of North County, Southeast Veterinary Oncology and Internal Medicine,
Bridge Animal Referral Center, The Animal Medical Center, the team at the University of
Guelph Institute for Comparative Cancer Investigation at the Mona Campbell Centre for Ani-
mal Cancer OVC HSC U of Guelph, Governor Animal Clinic Inc., University of Minnesota
College of Veterinary Medicine, Veterinary Specialty Hospital of Hong Kong, Oncovet, Clini-
que Vétérinaire de la Pierre Bleue, Eiffelvet, Clinique Vétérinaire SeineVet, Clinique Vétéri-
naire Mazetier, Evidensia Animal Hospital Barendrecht, VCA Metroplex Animal Hospital,
VCA Valley Oak Veterinary Center, City Paws Home Health, Prices Creek Veterinary Ser-
vices, Carlsbad Animal Hospital, Oceanside Veterinary Hospital, Amici Pet Hospital of Little
Italy, Colony Veterinary Hospital, Carriage Hills Animal Hospital, and the Integrated Oncol-
ogy Service at The Ohio State University Veterinary Medical Center and the Blue Buffalo Vet-
erinary Clinical Trials Office (BBVCTO). Additionally, the authors would like to give special
thanks to Deirdre Stuart, Victoria Sabine, Aspen Schreiner, Kayla Hufstetler, Olivia Uzan,
Kate Wotrang, Betsy Peet, Sarah Kenney, Sofya Gefter, Daisy Granados Peji, Alma D. Galicia
Ortiz, Natasha Shoemaker, Allison Ball, Dr. Kristina Bowles Miller, Marissa Kroll, Dr. Sarah
Roberts, Dr. Andrea Montaño Hernandez, Dr. Agata Rybicka, Juliette Blondiau, Emmanuel
Bouchaert, Kathleen Stuebner, April Jackson, Sara Pracht, Kelly Bergsrud, Anastasia Glahn,
Dana Jennings, Justin Paul, Angela Wilson, Beth Anderson, and Dana Nielsen.

Author Contributions
Conceptualization: Andi Flory, John A. Tynan, Lisa M. McLennan, Nilesh Dharajiya, Taylor
J. Jensen, Dirk van den Boom, Luis A. Diaz, Jr., Daniel S. Grosu, Lauren Holtvoigt, Ilya
Chorny, Dana W. Y. Tsui.
Data curation: Andi Flory, Kristina M. Kruglyak, John A. Tynan, Lisa M. McLennan, Patrick
Christian Fiaux, Gilberto E. Hernandez, Francesco Marass, Carlos A. Ruiz-Perez, Rita
Motalli-Pepio, Kate Wotrang, Chelsea D. Tripp, Candace Runyan, Daniel S. Grosu, Ilya
Chorny, Dana W. Y. Tsui.
Formal analysis: Andi Flory, Kristina M. Kruglyak, Jill M. Rafalko, Patrick Christian Fiaux,
Francesco Marass, Carlos A. Ruiz-Perez, Nilesh Dharajiya, Taylor J. Jensen, Dirk van den
Boom, Luis A. Diaz, Jr., Ilya Chorny, Dana W. Y. Tsui.

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 26 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

Funding acquisition: Daniel S. Grosu, Arthur Polk, Kalle Marsal, Lauren Holtvoigt, Ilya
Chorny.
Investigation: Andi Flory, John A. Tynan, Gilberto E. Hernandez, Prachi Nakashe, Donna M.
Fath, Kate Wotrang, Susan Lana, Brenda Phillips, Brian K. Flesner, Nicole F. Leibman,
Tracy LaDue, Chelsea D. Tripp, Brenda L. Coomber, J. Paul Woods, Mairin Miller, Sean
W. Aiken, Amber Wolf-Ringwall, Antonella Borgatti, Kathleen Kraska, Christopher B.
Thomson, Alane Kosanovich Cahalane, Rebecca L. Murray, William C. Kisseberth, Maria
A. Camps-Palau, Franck Floch, Claire Beaudu-Lange, Aurélia Klajer-Peres, Olivier Keravel,
Luc-André Fribourg-Blanc, Pascale Chicha Mazetier, Angelo Marco, Molly B. McLeod,
Erin Portillo, Terry S. Clark, Scott Judd, C. Kirk Feinberg, Marie Benitez, Candace Runyan,
Lindsey Hackett, Scott Lafey, Danielle Richardson, Sarah Vineyard, Mary Tefend Campbell,
Dana W. Y. Tsui.
Methodology: Andi Flory, Lisa M. McLennan, Gilberto E. Hernandez, Francesco Marass, Pra-
chi Nakashe, Carlos A. Ruiz-Perez, Nilesh Dharajiya, Taylor J. Jensen, Dirk van den Boom,
Luis A. Diaz, Jr., Daniel S. Grosu, Lauren Holtvoigt, Ilya Chorny, Dana W. Y. Tsui.
Project administration: Andi Flory, Lisa M. McLennan, Jill M. Rafalko, Prachi Nakashe,
Donna M. Fath, Rita Motalli-Pepio, Lauren Holtvoigt, Jason Chibuk, Dana W. Y. Tsui.
Resources: Andi Flory, Lisa M. McLennan, Donna M. Fath, Angela L. McCleary-Wheeler,
Susan Lana, Brenda Phillips, Brian K. Flesner, Nicole F. Leibman, Tracy LaDue, Chelsea D.
Tripp, Brenda L. Coomber, J. Paul Woods, Mairin Miller, Sean W. Aiken, Amber Wolf-
Ringwall, Antonella Borgatti, Kathleen Kraska, Christopher B. Thomson, Alane Kosanovich
Cahalane, Rebecca L. Murray, William C. Kisseberth, Maria A. Camps-Palau, Franck Floch,
Claire Beaudu-Lange, Aurélia Klajer-Peres, Olivier Keravel, Luc-André Fribourg-Blanc,
Pascale Chicha Mazetier, Angelo Marco, Molly B. McLeod, Erin Portillo, Terry S. Clark,
Scott Judd, C. Kirk Feinberg, Marie Benitez, Candace Runyan, Lindsey Hackett, Scott
Lafey, Danielle Richardson, Sarah Vineyard, Mary Tefend Campbell, Dana W. Y. Tsui.
Software: Kristina M. Kruglyak, Patrick Christian Fiaux, Carlos A. Ruiz-Perez, Ilya Chorny.
Supervision: Andi Flory, Kristina M. Kruglyak, John A. Tynan, Lisa M. McLennan, Gilberto
E. Hernandez, Thuy Jennings, Nilesh Dharajiya, Taylor J. Jensen, Dirk van den Boom, Luis
A. Diaz, Jr., Daniel S. Grosu, Arthur Polk, Susan Cho Hicks, Lauren Holtvoigt, Ilya Chorny,
Dana W. Y. Tsui.
Validation: Andi Flory, Kristina M. Kruglyak, John A. Tynan, Patrick Christian Fiaux, Gil-
berto E. Hernandez, Prachi Nakashe, Dana W. Y. Tsui.
Visualization: Andi Flory, Jill M. Rafalko, Dana W. Y. Tsui.
Writing – original draft: Andi Flory, Kristina M. Kruglyak, Jill M. Rafalko, Daniel S. Grosu,
Jason Chibuk, Dana W. Y. Tsui.
Writing – review & editing: Andi Flory, Kristina M. Kruglyak, Jill M. Rafalko, Angela L.
McCleary-Wheeler, Brenda Phillips, Kathleen Kraska, Christopher B. Thomson, Alane
Kosanovich Cahalane, Nilesh Dharajiya, Taylor J. Jensen, Dirk van den Boom, Luis A. Diaz,
Jr., Daniel S. Grosu, Arthur Polk, Kalle Marsal, Susan Cho Hicks, Katherine M. Lytle, Jason
Chibuk, Ilya Chorny, Dana W. Y. Tsui.

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 27 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

References
1. Fleming JM, Creevy KE, Promislow DEL. Mortality in North American Dogs from 1984 to 2004: An
Investigation into Age-, Size-, and Breed-Related Causes of Death. J Vet Intern Med. 2011; 25: 187–
198. https://doi.org/10.1111/j.1939-1676.2011.0695.x PMID: 21352376
2. American Veterinary Medical Association. Cancer in pets. [cited 17 Jan 2022]. Available: https://www.
avma.org/resources/pet-owners/petcare/cancer-pets
3. American Animal Hospital Association. Is my dog at risk for cancer? [cited 17 Jan 2022]. Available:
https://www.aaha.org/your-pet/pet-owner-education/ask-aaha/canine-cancer/
4. Walz JZ, Desai N, Asselt NV, Poirier VJ, Hansen K, Selmic L. Definitive-intent intensity-modulated
radiation therapy for treatment of canine prostatic carcinoma: A multi-institutional retrospective study.
Vet Comp Oncol. 2019. https://doi.org/10.1111/vco.12561 PMID: 31811693
5. Ravicini S, Baines SJ, Taylor A, Amores-Fuster I, Mason SL, Treggiari E. Outcome and prognostic fac-
tors in medically treated canine prostatic carcinomas: A multi-institutional study. Vet Comp Oncol.
2018; 16: 450–458. https://doi.org/10.1111/vco.12400 PMID: 29806232
6. Lee BM, Clarke D, Watson M, Laver T. Retrospective evaluation of a modified human lung cancer
stage classification in dogs with surgically excised primary pulmonary carcinomas. Vet Comp Oncol.
2020. https://doi.org/10.1111/vco.12582 PMID: 32115867
7. Batschinski K, Nobre A, Vargas-Mendez E, Tedardi MV, Cirillo J, Cestari G, et al. Canine visceral
hemangiosarcoma treated with surgery alone or surgery and doxorubicin: 37 cases (2005–2014). Can
Vet J La Revue Veterinaire Can. 2018; 59: 967–972.
8. Nguyen F, Peña L, Ibisch C, Loussouarn D, Gama A, Rieder N, et al. Canine invasive mammary carci-
nomas as models of human breast cancer. Part 1: natural history and prognostic factors. Breast Can-
cer Res Tr. 2018; 167: 635–648. https://doi.org/10.1007/s10549-017-4548-2 PMID: 29086231
9. Smith RA, Andrews KS, Brooks D, Fedewa SA, Manassaram-Baptiste D, Saslow D, et al. Cancer
screening in the United States, 2019: A review of current American Cancer Society guidelines and cur-
rent issues in cancer screening. Ca Cancer J Clin. 2019; 69: 184–210. https://doi.org/10.3322/caac.
21557 PMID: 30875085
10. Papadopoulos N, Kinzler KW, Vogelstein B. The role of companion diagnostics in the development
and use of mutation-targeted cancer therapies. Nat Biotechnol. 2006; 24: 985–995. https://doi.org/10.
1038/nbt1234 PMID: 16900147
11. Lanman RB, Mortimer SA, Zill OA, Sebisanovic D, Lopez R, Blau S, et al. Analytical and Clinical Vali-
dation of a Digital Sequencing Panel for Quantitative, Highly Accurate Evaluation of Cell-Free Circulat-
ing Tumor DNA. PLOS One. 2015; 10: e0140712. https://doi.org/10.1371/journal.pone.0140712
PMID: 26474073
12. Weber B, Meldgaard P, Hager H, Wu L, Wei W, Tsai J, et al. Detection of EGFR mutations in plasma
and biopsies from non-small cell lung cancer patients by allele-specific PCR assays. BMC Cancer.
2014; 14: 294. https://doi.org/10.1186/1471-2407-14-294 PMID: 24773774
13. Odegaard JI, Vincent JJ, Mortimer S, Vowles JV, Ulrich BC, Banks KC, et al. Validation of a plasma-
based comprehensive cancer genotyping assay utilizing orthogonal tissue- and plasma-based meth-
odologies. Clin Cancer Res. 2018; 24: clincanres.3831.2017. https://doi.org/10.1158/1078-0432.
CCR-17-3831 PMID: 29691297
14. U.S. Food & Drug Administration. FDA Approves First Liquid Biopsy Next-Generation Sequencing
Companion Diagnostic Test. 7 Aug 2020 [cited 24 Jan 2022]. Available: https://www.fda.gov/news-
events/press-announcements/fda-approves-first-liquid-biopsy-next-generation-sequencing-
companion-diagnostic-test
15. Hubbell E, Clarke CA, Aravanis AM, Berg CD. Modeled reductions in late-stage cancer with a multi-
cancer early detection test. Cancer Epidemiology Prev Biomarkers. 2020; cebp.1134.2020. https://
doi.org/10.1158/1055-9965.EPI-20-1134 PMID: 33328254
16. Clarke CA, Hubbell E, Ofman JJ. Multi-cancer early detection: a new paradigm for reducing cancer-
specific and all-cause mortality. Cancer Cell. 2021; 39: 447–448. https://doi.org/10.1016/j.ccell.2021.
02.004 PMID: 33606995
17. Businesswire.com. GRAIL Announces Significant Progress with Multi-Cancer Early Detection Test
Including FDA Breakthrough Device Designation. 13 May 2019 [cited 24 Jan 2022]. Available: https://
www.businesswire.com/news/home/20190513005441/en/GRAIL-Announces-Significant-Progress-
Multi-Cancer-Early-Detection/?feedref=JjAwJuNHiystnCoBq_hl-
WFAllVCLJFCqzlmaJ8DKHU4plfPZtlGYRDUHCSmgbii6XkLWuQZD-HgGRnjQvCyg3iCfEFIaJW7-
otp9V1XQiK2eHTJy3ZqOEt7kK
18. GRAIL, Inc. GRAIL Presents Interventional PATHFINDER Study Data at 2021 ASCO Annual Meeting
and Introduces Galleri, a Groundbreaking Multi-Cancer Early Detection Blood Test. 4 Jun 2021 [cited

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 28 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

24 Jan 2022]. Available: https://grail.com/press-releases/grail-presents-interventional-pathfinder-

study-data-at-2021-asco-annual-meeting-and-introduces-galleri-a-groundbreaking-multi-cancer-
early-detection-blood-test/
19. Schaefer DMW, Forman MA, Kisseberth WC, Lehman AM, Kelbick NT, Harper P, et al. Quantification
of plasma DNA as a prognostic indicator in canine lymphoid neoplasia. Vet Comp Oncol. 2007; 5:
145–155. https://doi.org/10.1111/j.1476-5829.2007.00122.x PMID: 19754786
20. Prouteau A, Denis JA, Fornel PD, Cadieu E, Derrien T, Kergal C, et al. Circulating tumor DNA is
detectable in canine histiocytic sarcoma, oral malignant melanoma, and multicentric lymphoma. Sci
Rep-uk. 2021; 11: 877. https://doi.org/10.1038/s41598-020-80332-y PMID: 33441840
21. Kruglyak KM, Chibuk J, McLennan L, Nakashe P, Hernandez GE, Motalli-Pepio R, et al. Blood-Based
Liquid Biopsy for Comprehensive Cancer Genomic Profiling Using Next-Generation Sequencing: An
Emerging Paradigm for Non-invasive Cancer Detection and Management in Dogs. Frontiers Vet Sci.
2021; 8: 704835. https://doi.org/10.3389/fvets.2021.704835 PMID: 34307538
22. Chibuk J, Flory A, Kruglyak KM, Leibman N, Nahama A, Dharajiya N, et al. Horizons in Veterinary Pre-
cision Oncology: Fundamentals of Cancer Genomics and Applications of Liquid Biopsy for the Detec-
tion, Characterization, and Management of Cancer in Dogs. Frontiers Vet Sci. 2021; 8: 664718.
https://doi.org/10.3389/fvets.2021.664718 PMID: 33834049
23. Henriksen TV, Reinert T, Christensen E, Sethi H, Birkenkamp-Demtröder K, Gögenur M, et al. The
effect of surgical trauma on circulating free DNA levels in cancer patients—implications for studies of
circulating tumor DNA. Mol Oncol. 2020; 14: 1670–1679. https://doi.org/10.1002/1878-0261.12729
PMID: 32471011
24. Corbetta M, Chiereghin C, Simone ID, Soldà G, Zuradelli M, Giunta M, et al. Post-Biopsy Cell-Free
DNA From Blood: An Open Window on Primary Prostate Cancer Genetics and Biology. Frontiers
Oncol. 2021; 11: 654140. https://doi.org/10.3389/fonc.2021.654140 PMID: 34109115
25. Baumgartner JM, Riviere P, Lanman RB, Kelly KJ, Veerapong J, Lowy AM, et al. Prognostic Utility of
Pre- and Postoperative Circulating Tumor DNA Liquid Biopsies in Patients with Peritoneal Metastases.
Ann Surg Oncol. 2020; 27: 3259–3267. https://doi.org/10.1245/s10434-020-08331-x PMID: 32767050
26. Kustanovich A, Schwartz R, Peretz T, Grinshpun A. Life and death of circulating cell-free DNA. Cancer
Biol Ther. 2019; 1–11. https://doi.org/10.1080/15384047.2019.1598759 PMID: 30990132
27. Wilson IJ, Burchell RK, Worth AJ, Burton SE, Gedye KR, Clark KJ, et al. Kinetics of Plasma Cell-Free
DNA and Creatine Kinase in a Canine Model of Tissue Injury. J Vet Intern Med. 2018; 32: 157–164.
https://doi.org/10.1111/jvim.14901 PMID: 29230875
28. Dharajiya NG, Grosu DS, Farkas DH, McCullough RM, Almasri E, Sun Y, et al. Incidental Detection of
Maternal Neoplasia in Noninvasive Prenatal Testing. Clin Chem. 2018; 64: 329–335. https://doi.org/
10.1373/clinchem.2017.277517 PMID: 28982650
29. Bianchi DW, Chudova D, Sehnert AJ, Bhatt S, Murray K, Prosen TL, et al. Noninvasive Prenatal Test-
ing and Incidental Detection of Occult Maternal Malignancies. JAMA. 2015; 314: 162–169. https://doi.
org/10.1001/jama.2015.7120 PMID: 26168314
30. Peretz-Machluf R, Rabinowitz T, Shomron N. Genome-Wide Noninvasive Prenatal Diagnosis of De
Novo Mutations. Methods Mol Biology. 2021; 2243: 249–269. https://doi.org/10.1007/978-1-0716-
1103-6_12 PMID: 33606261
31. Vail D, Thamm D, Liptak J, editors. Withrow and MacEwen’s Small Animal Clinical Oncology. 6th Edi-
tion. Saunders; 2019.
32. Amin MB, Gress DM, Vega LRM, Edge SB, Greene FL, Byrd DR, et al. AJCC Cancer Staging Manual.
Eighth Edition. American College of Surgeons; 2018. https://doi.org/10.1007/978-3-319-40618-3
33. Lindblad-Toh K, Wade CM, Mikkelsen TS, Karlsson EK, Jaffe DB, Kamal M, et al. Genome sequence,
comparative analysis and haplotype structure of the domestic dog. Nature. 2005; 438: 803–819.
https://doi.org/10.1038/nature04338 PMID: 16341006
34. Li H. Aligning sequence reads, clone sequences and assembly contigs with BWA-MEM. Arxiv. 2013.
35. Freed D, Pan R, Aldana R. TNscope: Accurate Detection of Somatic Mutations with Haplotype-based
Variant Candidate Detection and Machine Learning Filtering. Biorxiv. 2018; 250647. https://doi.org/10.
1101/250647
36. Adalsteinsson VA, Ha G, Freeman SS, Choudhury AD, Stover DG, Parsons HA, et al. Scalable whole-
exome sequencing of cell-free DNA reveals high concordance with metastatic tumors. Nat Commun.
2017; 8: 1324. https://doi.org/10.1038/s41467-017-00965-y PMID: 29109393
37. Howe KL, Achuthan P, Allen J, Allen J, Alvarez-Jarreta J, Amode MR, et al. Ensembl 2021. Nucleic
Acids Res. 2020; 49: D884–D891. https://doi.org/10.1093/nar/gkaa942 PMID: 33137190
38. Ransohoff DF. Rules of evidence for cancer molecular-marker discovery and validation. Nat Rev Can-
cer. 2004; 4: 309–314. https://doi.org/10.1038/nrc1322 PMID: 15057290

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 29 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

39. Wang Y. Development of cancer diagnostics—from biomarkers to clinical tests. Translational Cancer
Research. 2015; 4. Available: https://tcr.amegroups.com/article/view/4523
40. Biller B, Berg J, Garrett L, Ruslander D, Wearing R, Abbott B, et al. 2016 AAHA Oncology Guidelines
for Dogs and Cats. J Am Anim Hosp Assoc. 2016; 52: 181–204. https://doi.org/10.5326/JAAHA-MS-
6570 PMID: 27259020
41. Richards KL, Suter SE. Man’s best friend: what can pet dogs teach us about non-Hodgkin’s lym-
phoma? Immunol Rev. 2015; 263: 173–191. https://doi.org/10.1111/imr.12238 PMID: 25510277
42. Thomas R, Smith KC, Ostrander EA, Galibert F, Breen M. Chromosome aberrations in canine multi-
centric lymphomas detected with comparative genomic hybridisation and a panel of single locus
probes. Brit J Cancer. 2003; 89: 1530–1537. https://doi.org/10.1038/sj.bjc.6601275 PMID: 14562028
43. Vogelstein B, Papadopoulos N, Velculescu VE, Zhou S, Diaz LA, Kinzler KW. Cancer Genome Land-
scapes. Science. 2013; 339: 1546–1558. https://doi.org/10.1126/science.1235122 PMID: 23539594
44. Tate JG, Bamford S, Jubb HC, Sondka Z, Beare DM, Bindal N, et al. COSMIC: the Catalogue Of
Somatic Mutations In Cancer. Nucleic Acids Res. 2018; 47: gky1015–. https://doi.org/10.1093/nar/
gky1015 PMID: 30371878
45. Chakravarty D, Gao J, Phillips S, Kundra R, Zhang H, Wang J, et al. OncoKB: A Precision Oncology
Knowledge Base. JCO Precis Oncol. 2017; 2017: 1–16. https://doi.org/10.1200/PO.17.00011 PMID:
28890946
46. Kim JJ, Burger EA, Regan C, Sy S. Screening for Cervical Cancer in Primary Care: A Decision Analy-
sis for the US Preventive Services Task Force. Jama. 2018; 320: 706–714. https://doi.org/10.1001/
jama.2017.19872 PMID: 30140882
47. Imperiale TF, Ransohoff DF, Itzkowitz SH, Levin TR, Lavin P, Lidgard GP, et al. Multitarget Stool DNA
Testing for Colorectal-Cancer Screening. New Engl J Medicine. 2014; 370: 1287–1297. https://doi.
org/10.1056/nejmoa1311194 PMID: 24645800
48. Pinsky PF, Gierada DS, Black W, Munden R, Nath H, Aberle D, et al. Performance of Lung-RADS in
the National Lung Screening Trial: A Retrospective Assessment. Ann Intern Med. 2015; 162: 485–
491. https://doi.org/10.7326/M14-2086 PMID: 25664444
49. Lehman CD, Arao RF, Sprague BL, Lee JM, Buist DSM, Kerlikowske K, et al. National Performance
Benchmarks for Modern Screening Digital Mammography: Update from the Breast Cancer Surveil-
lance Consortium. Radiology. 2017; 283: 49–58. https://doi.org/10.1148/radiol.2016161174 PMID:
27918707
50. Klein EA, Richards D, Cohn A, Tummala M, Lapham R, Cosgrove D, et al. Clinical validation of a tar-
geted methylation-based multi-cancer early detection test using an independent validation set. Ann
Oncol. 2021. https://doi.org/10.1016/j.annonc.2021.05.806 PMID: 34176681
51. Cohen JD, Li L, Wang Y, Thoburn C, Afsari B, Danilova L, et al. Detection and localization of surgically
resectable cancers with a multi-analyte blood test. Science. 2018; 359: eaar3247. https://doi.org/10.
1126/science.aar3247 PMID: 29348365
52. Bettegowda C, Sausen M, Leary RJ, Kinde I, Wang Y, Agrawal N, et al. Detection of Circulating Tumor
DNA in Early- and Late-Stage Human Malignancies. Sci Transl Med. 2014; 6: 224ra24–224ra24.
https://doi.org/10.1126/scitranslmed.3007094 PMID: 24553385
53. Avanzini S, Kurtz DM, Chabon JJ, Moding EJ, Hori SS, Gambhir SS, et al. A mathematical model of
ctDNA shedding predicts tumor detection size. Sci Adv. 2020; 6: eabc4308. https://doi.org/10.1126/
sciadv.abc4308 PMID: 33310847
54. American Cancer Society. American Cancer Society Guidelines for the Early Detection of Cancer.
[cited 23 Nov 2021]. Available: https://www.cancer.org/healthy/find-cancer-early/american-cancer-
society-guidelines-for-the-early-detection-of-cancer.html
55. National Cancer Institute. Cancer Screening Overview. [cited 21 Jan 2022]. Available: https://www.
cancer.gov/about-cancer/screening/patient-screening-overview-pdq
56. Donlin JD. Open Letter from the Executive Vice President and Chief Executive Officer of AVMA. 2020.
Available: https://content.naic.org/sites/default/files/inline-files/2020_03_04_NAIC%20Model%
20Act_AVMA%20Comments_Definition%20of%20Preexisting.pdf
57. Schiffman JD, Breen M. Comparative oncology: what dogs and other species can teach us about
humans with cancer. Philosophical Transactions Royal Soc B Biological Sci. 2015; 370: 20140231.
https://doi.org/10.1098/rstb.2014.0231 PMID: 26056372
58. National Cancer Institute. What is Comparative Oncology? [cited 10 Nov 2021]. Available: https://ccr.
cancer.gov/Comparative-Oncology-Program/pet-owners/what-is-comp-onc
59. American Veterinary Medical Association. U.S. pet ownership statistics. [cited 10 Nov 2021]. Avail-
able: https://www.avma.org/resources-tools/reports-statistics/us-pet-ownership-statistics

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 30 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

60. Baioni E, Scanziani E, Vincenti MC, Leschiera M, Bozzetta E, Pezzolato M, et al. Estimating canine
cancer incidence: findings from a population-based tumour registry in northwestern Italy. Bmc Vet
Res. 2017; 13: 203. https://doi.org/10.1186/s12917-017-1126-0 PMID: 28659149
61. Shapley M, Mansell G, Jordan JL, Jordan KP. Positive predictive values of �5% in primary care for
cancer: systematic review. Br J Gen Pract. 2010; 60: e366–e377. https://doi.org/10.3399/
bjgp10X515412 PMID: 20849687
62. Hackshaw A, Cohen SS, Reichert H, Kansal AR, Chung KC, Ofman JJ. Estimating the population
health impact of a multi-cancer early detection genomic blood test to complement existing screening in
the US and UK. Brit J Cancer. 2021; 125: 1432–1442. https://doi.org/10.1038/s41416-021-01498-4
PMID: 34426664
63. Etzioni R, Gulati R, Weiss NS. Multi-Cancer Early Detection: Learning from the past to Meet the
Future. J National Cancer Inst. 2021; djab168–. https://doi.org/10.1093/jnci/djab168 PMID: 34450655
64. Nadauld LD, McDonnell CH, Beer TM, Liu MC, Klein EA, Hudnut A, et al. The PATHFINDER Study:
Assessment of the Implementation of an Investigational Multi-Cancer Early Detection Test into Clinical
Practice. Cancers. 2021; 13: 3501. https://doi.org/10.3390/cancers13143501 PMID: 34298717
65. Lo YMD, Lam WKJ. Towards multi-cancer screening using liquid biopsies. Nat Rev Clin Oncol. 2020;
17: 525–526. https://doi.org/10.1038/s41571-020-0404-0 PMID: 32555396
66. Kooyker AI, Toes-Zoutendijk E, Winden AWJO, Spaander MCW, Buskermolen M, Vuuren HJ, et al.
The second round of the Dutch colorectal cancer screening program: Impact of an increased fecal
immunochemical test cut-off level on yield of screening. Int J Cancer. 2020; 147: 1098–1106. https://
doi.org/10.1002/ijc.32839 PMID: 31853977
67. Haug U, Coupé VMH. The cumulative false-positive rate in colorectal cancer screening: a Markov
analysis. Eur J Gastroen Hepat. 2020; 32: 575–580. https://doi.org/10.1097/MEG.
0000000000001669 PMID: 31990711
68. Nadler DL, Zurbenko IG. Estimating Cancer Latency Times Using a Weibull Model. Adv Epidemiology.
2014; 2014: 1–8. https://doi.org/10.1155/2014/746769
69. Reese DJ, Green EM, Zekas LJ, Flores JE, Hill LN, Winter MD, et al. Intra- and interobserver variability
of board-certified veterinary radiologists and veterinary general practitioners for pulmonary nodule
detection in standard and inverted display mode images of digital thoracic radiographs of dogs. J Am
Vet Med Assoc. 2011; 238: 998–1003. https://doi.org/10.2460/javma.238.8.998 PMID: 21492042
70. Williams JM, Graham JP, Wang C. Interobserver Variability of Radiographic Pulmonary Nodule Diam-
eter Measurements in Dogs and Cats. J Am Anim Hosp Assoc. 2014; 50: 83–88. https://doi.org/10.
5326/JAAHA-MS-5988 PMID: 24446396
71. Ku C -K., Kass PH, Christopher MM. Cytologic–histologic concordance in the diagnosis of neoplasia in
canine and feline lymph nodes: a retrospective study of 367 cases. Vet Comp Oncol. 2017; 15: 1206–
1217. https://doi.org/10.1111/vco.12256 PMID: 27523399
72. Christopher MM, Ku C. Likelihood of Neoplasia for Diagnoses Modified by Probability Terms in Canine
and Feline Lymph Node Cytology: How Probable Is Probable? Frontiers Vet Sci. 2018; 5: 246. https://
doi.org/10.3389/fvets.2018.00246 PMID: 30356759
73. Koehler JW, Miller AD, Miller CR, Porter B, Aldape K, Beck J, et al. A Revised Diagnostic Classification
of Canine Glioma: Towards Validation of the Canine Glioma Patient as a Naturally Occurring Preclini-
cal Model for Human Glioma. J Neuropathology Exp Neurology. 2018; 77: 1039–1054. https://doi.org/
10.1093/jnen/nly085 PMID: 30239918
74. Tecilla M, Gambini M, Forlani A, Caniatti M, Ghisleni G, Roccabianca P. Evaluation of cytological diag-
nostic accuracy for canine splenic neoplasms: An investigation in 78 cases using STARD guidelines.
PLOS One. 2019; 14: e0224945. https://doi.org/10.1371/journal.pone.0224945 PMID: 31697755
75. Hennigan ST, Trostel SY, Terrigino NT, Voznesensky OS, Schaefer RJ, Whitlock NC, et al. Low Abun-
dance of Circulating Tumor DNA in Localized Prostate Cancer. JCO Precis Oncol. 2019; 3:
PO.19.00176. https://doi.org/10.1200/PO.19.00176 PMID: 31528835
76. Mochizuki H, Shapiro SG, Breen M. Detection of BRAF Mutation in Urine DNA as a Molecular Diag-
nostic for Canine Urothelial and Prostatic Carcinoma. PLOS One. 2015; 10: e0144170. https://doi.org/
10.1371/journal.pone.0144170 PMID: 26649430
77. Stark AC, McGrath S, Karn M, Thomson CE. Evaluation of cell-free DNA as a diagnostic marker in
cerebrospinal fluid of dogs. Am J Vet Res. 2020; 81: 416–421. https://doi.org/10.2460/ajvr.81.5.416
PMID: 32343180
78. Miller AM, Shah RH, Pentsova EI, Pourmaleki M, Briggs S, Distefano N, et al. Tracking Tumor Evolu-
tion in Glioma through Liquid Biopsies of Cerebrospinal Fluid. Nature. 2019; 565: 654–658. https://doi.
org/10.1038/s41586-019-0882-3 PMID: 30675060

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 31 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

79. Knight SR, Thorne A, Faro MLL. Donor-specific Cell-free DNA as a Biomarker in Solid Organ Trans-
plantation. A Systematic Review. Transplantation. 2019; 103: 273–283. https://doi.org/10.1097/TP.
0000000000002482 PMID: 30308576
80. Sharon E, Shi H, Kharbanda S, Koh W, Martin LR, Khush KK, et al. Quantification of transplant-
derived circulating cell-free DNA in absence of a donor genotype. PLOS Comput Biol. 2017; 13:
e1005629. https://doi.org/10.1371/journal.pcbi.1005629 PMID: 28771616
81. Razavi P, Li BT, Brown DN, Jung B, Hubbell E, Shen R, et al. High-intensity sequencing reveals the
sources of plasma circulating cell-free DNA variants. Nat Med. 2019; 25: 1928–1937. https://doi.org/
10.1038/s41591-019-0652-7 PMID: 31768066
82. Gajecka M. Unrevealed mosaicism in the next-generation sequencing era. Mol Genet Genomics.
2016; 291: 513–530. https://doi.org/10.1007/s00438-015-1130-7 PMID: 26481646
83. Friedman E, Efrat N, Soussan-Gutman L, Dvir A, Kaplan Y, Ekstein T, et al. Low-level constitutional
mosaicism of a de novoBRCA1 gene mutation. Brit J Cancer. 2015; 112: 765–768. https://doi.org/10.
1038/bjc.2015.14 PMID: 25633036
84. Zamora K, Newton amy, Denham SL, Wei S, Woodard SA. High-Risk Lesions: Review and Manage-
ment Update. J Am Osteopath Coll Radiol. 2021; 1: 6–18. Available: https://www.jaocr.org/articles/
high-risk-lesions-review-and-management-update
85. Aziz H, Brown ZJ, Eskander MF, Aquina CT, Baghdadi A, Kamel IR, et al. A Scoping Review of the
Classification, Diagnosis, and Management of Hepatic Adenomas. J Gastrointest Surg. 2022; 1–14.
https://doi.org/10.1007/s11605-021-05032-y PMID: 34027579
86. Carbonnel M, Layoun L, Poulain M, Tourne M, Murtada R, Grynberg M, et al. Serous Borderline Ovar-
ian Tumor Diagnosis, Management and Fertility Preservation in Young Women. J Clin Medicine.
2021; 10: 4233. https://doi.org/10.3390/jcm10184233 PMID: 34575343
87. Gerlinger M, Rowan AJ, Horswell S, Larkin J, Endesfelder D, Gronroos E, et al. Intratumor Heteroge-
neity and Branched Evolution Revealed by Multiregion Sequencing. New Engl J Medicine. 2012; 366:
883–892. https://doi.org/10.1056/NEJMoa1113205 PMID: 22397650
88. Lurkin A, Ducimetière F, Vince DR, Decouvelaere A-V, Cellier D, Gilly FN, et al. Epidemiological evalu-
ation of concordance between initial diagnosis and central pathology review in a comprehensive and
prospective series of sarcoma patients in the Rhone-Alpes region. BMC Cancer. 2010; 10: 150.
https://doi.org/10.1186/1471-2407-10-150 PMID: 20403160
89. Shen H, Hsu H-C, Tai Y-J, Kuo K-T, Wu C-Y, Lai Y-L, et al. Factors Influencing the Discordancy
Between Intraoperative Frozen Sections and Final Paraffin Pathologies in Ovarian Tumors. Frontiers
Oncol. 2021; 11: 694441. https://doi.org/10.3389/fonc.2021.694441 PMID: 34277439
90. Lindquist KE, Ciornei C, Westbom-Fremer S, Gudinaviciene I, Ehinger A, Mylona N, et al. Difficulties
in diagnostics of lung tumours in biopsies: an interpathologist concordance study evaluating the inter-
national diagnostic guidelines. J Clin Pathol. 2021; jclinpath-2020-207257. https://doi.org/10.1136/
jclinpath-2020-207257 PMID: 33547095
91. Pantanowitz L, Quiroga-Garza GM, Bien L, Heled R, Laifenfeld D, Linhart C, et al. An artificial intelli-
gence algorithm for prostate cancer diagnosis in whole slide images of core needle biopsies: a blinded
clinical validation and deployment study. Lancet Digital Heal. 2020; 2: e407–e416. https://doi.org/10.
1016/S2589-7500(20)30159-X PMID: 33328045
92. Lacey JV, Chia VM. Endometrial hyperplasia and the risk of progression to carcinoma. Maturitas.
2009; 63: 39–44. https://doi.org/10.1016/j.maturitas.2009.02.005 PMID: 19285814
93. Mutter GL, Zaino RJ, Baak JPA, Bentley RC, Robboy SJ. Benign Endometrial Hyperplasia Sequence
and Endometrial Intraepithelial Neoplasia. Int J Gynecol Pathol. 2007; 26: 103–114. https://doi.org/10.
1097/PGP.0b013e31802e4696 PMID: 17413975
94. Ip PP-C. Benign endometrial proliferations mimicking malignancies: a review of problematic entities in
small biopsy specimens. Virchows Arch. 2018; 472: 907–917. https://doi.org/10.1007/s00428-018-
2314-4 PMID: 29445890
95. Rosa M, Agosto-Arroyo E. Core needle biopsy of benign, borderline and in-situ problematic lesions of
the breast: Diagnosis, differential diagnosis and immunohistochemistry. Ann Diagn Pathol. 2019; 43:
151407. https://doi.org/10.1016/j.anndiagpath.2019.151407 PMID: 31634810
96. Patel A. Benign vs Malignant Tumors. JAMA Oncol. 2020; 6: 1488–1488. https://doi.org/10.1001/
jamaoncol.2020.2592 PMID: 32729930
97. London SJ, Connolly JL, Schnitt SJ, Colditz GA. A Prospective Study of Benign Breast Disease and
the Risk of Breast Cancer. JAMA. 1992; 267: 941–944. https://doi.org/10.1001/jama.1992.
03480070057030 PMID: 1734106

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 32 / 33

PLOS ONE Clinical validation of a next-generation sequencing-based “liquid biopsy” test for cancer in over 1,000 dogs

98. Hartmann LC, Sellers TA, Frost MH, Lingle WL, Degnim AC, Ghosh K, et al. Benign Breast Disease
and the Risk of Breast Cancer. New Engl J Medicine. 2005; 353: 229–237. https://doi.org/10.1056/
NEJMoa044383 PMID: 16034008
99. Nees LK, Heublein S, Steinmacher S, Juhasz-Böss I, Brucker S, Tempfer CB, et al. Endometrial
hyperplasia as a risk factor of endometrial cancer. Arch Gynecol Obstet. 2022; 1–15. https://doi.org/
10.1007/s00404-021-06260-y PMID: 34609593
100. CONTEDUCA V, SANSONNO D, RUSSI S, DAMMACCO F. Precancerous colorectal lesions. Int J
Oncol. 2013; 43: 973–984. https://doi.org/10.3892/ijo.2013.2041 PMID: 23900573
101. Desjonqueres E, Campani C, Marra F, Zucman-Rossi J, Nault J. Preneoplastic lesions in the liver:
Molecular insights and relevance for clinical practice. Liver Int. 2022. https://doi.org/10.1111/liv.15152
PMID: 34982503
102. Strasberg SM, Chapman WC. Enucleation of Biliary Cystadenomas: a Review. J Gastrointest Surg.
2021; 25: 2700–2706. https://doi.org/10.1007/s11605-021-05106-x PMID: 34505221
103. Silva NLT, Bonatto NCM, Oliveira PL, Vieira GC, Floriano BP, Barros LD, et al. Post-prandial lipemia
and glycemia in dogs fed with industrialized pet food. Comp Clin Pathology. 2019; 28: 253–258.
https://doi.org/10.1007/s00580-018-2824-0
104. Cawley J, Khanna C. The Challenge of Cognitive Dissonance in the Delivery of Precision Medicine in
Veterinary Oncology. Arch Cancer Biol Ther. 2020; 1: 37–41.
105. Siegel RL, Miller KD, Fuchs HE, Jemal A. Cancer Statistics, 2021. Ca Cancer J Clin. 2021; 71: 7–33.
https://doi.org/10.3322/caac.21654 PMID: 33433946
106. National Cancer Institute. Screening Tests. [cited 1 Feb 2022]. Available: https://www.cancer.gov/
about-cancer/screening/screening-tests
107. Sultan F, Ganaie BA. Comparative oncology: Integrating human and veterinary medicine. Open Vet J.
2018; 8: 25–34. https://doi.org/10.4314/ovj.v8i1.5 PMID: 29445618

PLOS ONE | https://doi.org/10.1371/journal.pone.0266623 April 26, 2022 33 / 33