Lung Cancer

AN INTERNATIONAL JOURNAL FOR LUNG CANCER
AND OTHER THORACIC MALIGNANCIES

Affiliated with the International Lung Cancer Consortium, the European Thoracic Oncology
Platform and the British Thoracic Oncology Group.

Aims and Scope
Lung Cancer is an international publication covering the clinical, translational and basic science of malignancies of the lung and chest region.
Original research articles, early reports, review articles, editorials and correspondence covering the prevention, epidemiology and etiology, basic
biology, pathology, clinical assessment, surgery, chemotherapy, radiotherapy, combined treatment modalities, other treatment modalities and
outcomes of lung cancer are welcome.

Editor-in-Chief:
R. Stahel (Switzerland)

Deputy Editors:
P. Baas (Netherlands)
M. Edelman (USA)
S. Popat (UK)
Y.-L. Wu (China)

Associate Editors:
Epidemiology and Prevention:
M. Janssen-Heijnen (Netherlands)

Pathology:
E. Brambilla (France)
S. Finn (Ireland)

Systemic Treatments:
T. Berghmans (France)
B.C. Cho (Korea)
N. Girard (France)
S. Lu (China)
I. Okomoto (Japan)
C. Zhou (China)

Radiation Oncology:
M. Guckenberger (Switzerland)
S. Senan (Netherlands)
P. van Houtte (Belgium)

Surgery:
F. Yang (China)

Pneumonology:
D. Baldwin (UK)

Radiology and Nuclear Medicine:
I. Burger (Switzerland)
C.A. Ridge (UK)

Preclinical and Translational Research:
D. Costa (USA)
R. Soo (Singapore)

Statistics Editor:
U. Dafni (Greece)

Social Media Editor:
A. Passaro (Italy)

Editorial Board: V. Ninane (Belgium) C. Rolfo (Belgium) G. Veronesi (Italy)
K.J. O’Byrne (Ireland) J.-P. Sculier (Belgium) H. Wakelee (USA)
L. Bubendorf (Switzerland) S. Peters (Switzerland) T. Stinchcombe (USA) Y.-C. Wang (Taiwan)
C. Faivre–Finn (UK ) R. Pirker (Austria) P. VanderLaan (USA) S. Yom (USA)
E. Felip (Spain) M. Reck (Germany) J.P. Van Meerbeeck (Belgium)
S. Lam (Canada) G. Rocco (USA) A. Vergnenègre (France)
C. Mascaux (France)
T. Mitsudomi (Japan)

Editorial Office:

Hannah Searle, Lung Cancer Editorial Office, Elsevier UK Ltd., Stover Court, Bampfylde Street, Exeter, EX1 2AH, UK. +44 (0)1392 285881.
e-mail: [email protected]

Processed at Thomson Digital, Gangtok (India)

doi:10.1016/S0169-5002(21)00650-4

Vol. 163,  January 2022 www.journals.elsevier.com/lung-cancer
CONTENTS

Cited in: EMBASE/Excertpa Medica, Oncology Information Service, Elsevier BIOBASE/Current Awareness in Biological Sciences;
Current Contents/Clinical Medicine; SciSearch, Index Medicus/MEDLINE. Also covered in the abstract and citation database SCOPUS®.

Full text available on ScienceDirect®.

Reviews 1

The highlights of the 15th international conference of the international mesothelioma interest group – Do molecular
concepts challenge the traditional approach to pathological mesothelioma diagnosis?

S. Klebe (Australia), F. Galateau Salle (France), R. Bruno (Italy), L. Brcic (Austria), H. I. Chen‑Yost (United States),
M.‑C. Jaurand (France)

Antibody drug conjugates in non-small cell lung cancer: An emerging therapeutic approach 59
S. Marks (Republic of Ireland), J. Naidoo (Republic of Ireland, USA)

Antibody-drug conjugates: A promising novel therapeutic approach in lung cancer 96
A. Desai (USA), P. Abdayem (France), A.A. Adjei (USA), D. Planchard (France)

Original Research Articles

Genetic landscape of patients with ALK-rearranged non–small-cell lung cancer (NSCLC) and response to ceritinib in ASCEND-1 study
D.S.‑W. Tan (Singapore), M. Thomas (Germany), D.‑W. Kim (Korea), S. Szpakowski (Cambridge), P. Urban (Switzerland),

R. Mehra (Philadelphia), L.Q.M. Chow (Seattle), S. Sharma (Salt Lake City), B.J. Solomon (Australia), E. Felip (Spain),

D.R. Camidge (Aurora), J. Vansteenkiste (Belgium), L. Petruzzelli (Cambridge), S. Pantano (Switzerland), A.T. Shaw

(Boston) 7

A phase 1b study evaluating the safety and preliminary efficacy of berzosertib in combination with gemcitabine in patients 19
with advanced non-small cell lung cancer

R. Plummer, E. Dean, H.‑T. Arkenau (United Kingdom), C. Redfern, A.I. Spira, J.M. Melear, K.Y. Chung (United States),
J. Ferrer‑Playan, T. Goddemeier, G. Locatelli (Germany), J. Dong, P. Fleuranceau‑Morel (United States), I. Diaz‑Padilla
(Germany), G.I. Shapiro (United States)

A risk prediction model for selecting high-risk population for computed tomography lung cancer screening in China 27
L.‑W. Guo, Z.‑Y. Lyu, Q.‑C. Meng, L.‑Y. Zheng, Q. Chen, Y. Liu, H.‑F. Xu, R.‑H. Kang, L.‑Y. Zhang, X.‑Q. Cao, S.‑Z. Liu,
X.‑B. Sun, J.‑G. Zhang, S.‑K. Zhang (China)

Prognostic and predictive value of neutrophil-to-lymphocyte ratio with adjuvant immunotherapy in stage III non-small-cell 35
lung cancer

A.K. Bryant, K. Sankar, G.W. Strohbehn, L. Zhao, D. Elliott, A. Qin, S. Yentz, N. Ramnath, M.D. Green (USA)

Modifiable factors associated with health-related quality of life among lung cancer survivors following curative intent therapy
D.M. Ha, A.V. Prochazka, D.B. Bekelman, J.E. Stevens‑Lapsley, J.L. Studts, R.L. Keith (United States) 42

Surgical or medical strategy for locally-advanced, stage IIIA/B-N2 non-small cell lung cancer: Reproducibility of decision- 51
making at a multidisciplinary tumor board

J. Mainguene, C. Basse, P. Girard, S. Beaucaire‑Danel, K. Cao, E. Brian, M. Grigoroiu, D. Gossot, M. Luporsi, L. Perrot,
T. Vieira, R. Caliandro, C. Daniel, A. Seguin‑Givelet, N. Girard (France)

(Contents continued overleaf)

08036

doi:10.1016/S0169-5002(21)00652-8

vi Contents continued

Impacts of lung cancer multidisciplinary meeting presentation: Drivers and outcomes from a population registry 69
retrospective cohort study 77
87
T. Lin, J. Pham, E. Paul, M. Conron, G. Wright, D. Ball, P. Mitchell, N. Atkin, M. Brand, J. Zalcberg, R.G. Stirling (Australia)
14
Brain penetration and efficacy of tepotinib in orthotopic patient-derived xenograft models of MET-driven non-small cell 107
lung cancer brain metastases

M. Friese‑Hamim (Germany), A. Clark (USA), D. Perrin (Germany), L. Crowley (USA), C. Reusch, O. Bogatyrova (Germany),
H. Zhang, T. Crandall, J. Lin, J. Ma, D. Bachner (USA), J. Schmidt, M. Schaefer, C. Stroh (Germany)

Preoperative CT-based peritumoral and tumoral radiomic features prediction for tumor spread through air spaces in clinical
stage I lung adenocarcinoma

G. Liao (China), L. Huang (China, Germany), S. Wu, P. Zhang, D. Xie, L. Yao, Z. Zhang, S. Yao, L. Shanshan, S. Wang, G. Wang,
L. Wing‑Chi Chan, H. Zhou (China)

Short Communications
Incorporating circulating tumor DNA detection to radiographic assessment for treatment response in advanced EGFR-mutant
lung cancer

P.‑S. Kok (Australia), K. Lee (Hong Kong), S. Lord, T. John, I. Marschner (Australia), Y.‑L. Wu (China), T.S.K. Mok (Hong Kong),
C.K. Lee (Australia)

Acknowledgement to reviewers

Lung Cancer 163 (2022) 19–26
Contents lists available at ScienceDirect

Lung Cancer

journal homepage: www.elsevier.com/locate/lungcan

A phase 1b study evaluating the safety and preliminary efficacy of
berzosertib in combination with gemcitabine in patients with advanced
non-small cell lung cancer

Ruth Plummer a,*,3, Emma Dean b,1, Hendrik-Tobias Arkenau c, Charles Redfern d, Alexander
I. Spira e, Jason M. Melear f, Ki Y. Chung g, Jordi Ferrer-Playan h, Thomas Goddemeier i,
Giuseppe Locatelli i, Jennifer Dong j, Patricia Fleuranceau-Morel j, Ivan Diaz-Padilla h,2, Geoffrey
I. Shapiro k

a Newcastle University and Northern Centre for Cancer Care, Newcastle Hospitals NHS Trust, Newcastle Upon Tyne, United Kingdom
b The University of Manchester and The Christie NHS Foundation Trust, Manchester, United Kingdom
c Sarah Cannon Research Institute, London, United Kingdom
d Sharp Healthcare, San Diego, CA, United States
e Virginia Cancer Specialists Research Institute and US Oncology Research, Fairfax, VA, United States
f Texas Oncology, Austin, TX, United States
g Prisma Health, Greenville, SC, United States
h Ares Trading SA, Eysins, Switzerland, an affiliate of Merck KGaA, Darmstadt, Germany
i Merck Healthcare KGaA, Darmstadt, Germany
j EMD Serono Research & Development Institute, Inc., Billerica, MA, United States
k Dana-Farber Cancer Institute and Harvard Medical School, Boston, MA, United States

ARTICLE INFO ABSTRACT

Keywords: Objectives: Berzosertib (formerly M6620, VX-970) is an intravenous, highly potent and selective, first-in-class
Berzosertib ataxia telangiectasia and Rad3-related (ATR) protein kinase inhibitor. We assessed the safety, tolerability, pre­
Gemcitabine liminary efficacy, and pharmacokinetics (PK) of berzosertib plus gemcitabine in an expansion cohort of patients
Non-small cell lung cancer with advanced non-small cell lung cancer (NSCLC). The association of efficacy with TP53 status and other tumor
ATR inhibitor markers was also explored.
DNA-damage response Materials and methods: Adult patients with advanced histologically confirmed NSCLC received berzosertib 210
mg/m2 (days 2 and 9) and gemcitabine 1000 mg/m2 (days 1 and 8) at the recommended phase 2 dose established
in the dose escalation part of the study.
Results: Thirty-eight patients received at least one dose of study treatment. The most common treatment-
emergent adverse events were fatigue (55.3%), anemia (52.6%), and nausea (39.5%). Gemcitabine had no
apparent effect on the PK of berzosertib. The objective response rate (ORR) was 10.5% (4/38, 90% confidence
interval [CI]: 3.7–22.5%). In the exploratory analysis, the ORR was 30.0% (3/10, 90% CI: 9.0–61.0%) in patients
with high loss of heterozygosity (LOH) and 11.0% (1/9, 90% CI: 1.0–43.0%) in patients with low LOH. The ORR
was 33.0% (2/6, 90% CI: 6.0–73.0%) in patients with high tumor mutational burden (TMB), 12.5% (2/16, 90%
CI: 2.0–34.0%) in patients with intermediate TMB, and 0% (0/3, 90% CI: 0.0–53.6%) in patients with low TMB.

* Corresponding author.
E-mail addresses: [email protected] (R. Plummer), [email protected] (E. Dean), [email protected] (H.-T. Arkenau),

[email protected] (J.M. Melear), [email protected] (K.Y. Chung), [email protected] (G.I. Shapiro).
1 Emma Dean was employed by The University of Manchester and The Christie NHS Foundation Trust, Manchester at the time of study. This author is no longer an

employee of The University of Manchester and The Christie NHS Foundation Trust, Manchester. Current address: Oncology R&D, AstraZeneca, Cambridge, United
Kingdom.

2 Ivan Diaz-Padilla was employed by Ares Trading SA, Eysins, Switzerland, an affiliate of Merck KGaA, Darmstadt, Germany, at the time of manuscript preparation.
This author is no longer an employee of Ares Trading SA, Eysins, Switzerland, an affiliate of Merck KGaA, Darmstadt, Germany. Current address: GlaxoSmithKline,
Zug, Switzerland.

3 ORCID ID: 0000-0003-0107-1444.

https://doi.org/10.1016/j.lungcan.2021.11.011
Received 21 October 2021; Received in revised form 16 November 2021; Accepted 22 November 2021
Available online 1 December 2021
0169-5002/© 2021 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

R. Plummer et al. Lung Cancer 163 (2022) 19–26

Conclusions: Berzosertib plus gemcitabine was well tolerated in patients with advanced, pre-treated NSCLC. Based
on the observed clinical efficacy, future clinical trials should involve genomically selected patients.

1. Introduction evaluate the safety, tolerability, pharmacokinetics (PK), and preliminary
efficacy of berzosertib combined with gemcitabine in patients with
For patients with advanced non-small cell lung cancer (NSCLC) and advanced NSCLC, with or without TP53 mutations. An exploratory
no targetable mutations, cytotoxic chemotherapies, including DNA- analysis of potential response biomarkers was also conducted (Clin­
damaging or anti-mitotic agents, achieve response rates of 7–21%, as icalTrials.gov, identifier: NCT02157792).
per Response Evaluation Criteria in Solid Tumors (RECIST), when used
as single agents in the second- and/or third-line treatment setting [1–7]. 2. Materials and methods

Ataxia-telangiectasia-mutated (ATM) and Rad3-related protein ki­ 2.1. Study design
nases (ATR) play critical roles in the DNA-damage response (DDR) by
regulating the cell cycle checkpoint control and repairing damaged DNA This trial was part of a multicenter, open-label, non-randomized,
by homologous recombination [8]. In response to DNA replication phase 1 study separated into six parts (A, B, B2, C1, C2, and C3). The
stress, induced or exacerbated by chemotherapies such as gemcitabine, initial dose escalation phase of the study (parts A and B) established the
ATR is recruited to regions of exposed single-stranded DNA to mediate recommended phase 2 dose (RP2D) of berzosertib when combined with
replication fork stabilization, whereas ATM responds to DNA double- chemotherapeutic agents, including gemcitabine and cisplatin [22,23].
strand breaks [9]. These doses were further evaluated in the expansion phase of the study
in patients with NSCLC (part C1), triple-negative breast cancer (part C2),
Berzosertib (formerly M6620, VX-970) is an intravenous (IV), highly and small-cell lung cancer (part C3). The focus of this manuscript is part
potent, and selective first-in-class inhibitor of ATR [10]. In preclinical C1; the other parts have been or will be reported separately.
studies, berzosertib sensitizes lung cancer cells to DNA-damage-
inducing chemotherapeutics such as gemcitabine [10]. Previous clin­ Part C1 was a single-arm expansion cohort study of berzosertib
ical studies have shown that berzosertib in combination with chemo­ combined with gemcitabine in patients with advanced NSCLC, with or
therapy is well tolerated with preliminary efficacy signals in several without TP53 mutations. This part of the study was conducted across
solid tumors [11,12]. Furthermore, a recent proof-of-concept phase 2 three sites in the UK and eight sites in the USA between December 2015
study evaluating the berzosertib–topotecan combination reported an and March 2020. The study was conducted in accordance with the
objective response rate of 36% (9/25), with a median duration of ethical principles of the International Council for Harmonization
response of 6.4 months, in patients with SCLC, including those with Guidelines for Good Clinical Practice and the Declaration of Helsinki, as
platinum-resistant disease [13]. well as with applicable local regulations.

Berzosertib efficacy can be enriched in the presence of specific tumor 2.2. Patients
genetic alterations. Tumor protein p53 (TP53) mutational status has
been shown preclinically to correlate with response to DNA-damaging The plan was to enroll approximately 40 patients, including at least
agents combined with ATR inhibition [14]. This is explained by the 20 patients with TP53 mutations prospectively determined from
dependence of tumor cells on a functional TP53 to maintain genomic archival tumor biopsies.
stability when ATR is inhibited [15], as well as the importance of the
ATR-CHK1 axis for G2/M checkpoint control in response to DNA dam­ Eligible patients were adults ≥ 18 years of age with advanced
age when TP53 is mutated, thus harnessing the synthetic lethal rela­ (metastatic or locally advanced unresectable tumors and not eligible for
tionship between ATR and TP53 in TP53-mutant tumors. Since definitive treatment), histologically confirmed NSCLC, with available
mutations of the TP53 gene are present in approximately 50% of NSCLC archival tumor biopsies, who were intolerant to standard approved
[16], ATR inhibition represents a potential therapeutic combination targeted therapies, and measurable disease defined by RECIST v1.1
strategy for DNA-damaging chemotherapy in pretreated NSCLC. In [24]. Patients who had received more than two lines of cytotoxic
addition, recent studies have suggested that molecular alterations in chemotherapy in the advanced setting were excluded. Patients who had
other genes, such as ATM, switch/sucrose non-fermentable related, received treatment with gemcitabine within 6 months were also
matrix associated, actin dependent regulator of chromatin, subfamily A, excluded. Full inclusion and exclusion criteria are shown in the Sup­
member 4 (SMARCA4) and AT-rich interaction domain (ARID1A), could plementary Information.
be potential predictive biomarkers of ATR inhibition by exploiting
mechanisms of synthetic lethality related to the DDR or to replication 2.3. Treatments
fork stability [17–19]. However, when leveraging the synthetic lethal
relationship between ATR and ATM, it may be important to consider the Following screening and baseline evaluations, patients received
emerging role of ATM in promoting tumor cell ferroptosis [20]. berzosertib IV (210 mg/m2; days 2 and 9) approximately 24 h after
SMARCA4 is frequently mutated in NSCLC and is involved in the acti­ gemcitabine (1000 mg/m2; days 1 and 8) in 21-day cycles, which was
vation of replication stress responses, while ARID1A mutations increase the RP2D established in part A of this study [22]. The timing of berzo­
tumor cells reliance on ATR-mediated checkpoint activity. Furthermore, sertib relative to gemcitabine was based on the synergy demonstrated
ARID1A-mutant tumor cells may be more susceptible to oxidative stress when berzosertib was administered 12–24 h after gemcitabine in pre­
due to low levels of antioxidant factors such as glutathione [21]. clinical models [25]. Patients received treatment until progressive dis­
ease (PD) or unacceptable toxicity.
This phase 1 study was separated into six parts (A, B, B2, C1, C2, and
C3). In the dose escalation part of this study with berzosertib and 2.4. Objectives
gemcitabine (part A), the most common treatment–emergent adverse
events (TEAE) of any grade included fatigue, nausea, anemia, and in­ The primary objectives of this study were to evaluate the safety,
creases in alanine aminotransferase (ALT), and the most common grade tolerability, and the objective response rate (ORR) of berzosertib when
≥ 3 TEAEs were neutropenia, increases in ALT and fatigue. These TEAEs combined with gemcitabine in patients with advanced NSCLC, with and
were consistent with patient populations treated with gemcitabine [22].

The main purpose of this expansion cohort study (part C1) was to

20

R. Plummer et al. Lung Cancer 163 (2022) 19–26

without TP53 mutations. The secondary objectives were to evaluate the calculated using standard non-compartmental methods and the actual
preliminary efficacy and PK of berzosertib combined with gemcitabine. administered dose. Computation of PK parameters was performed using
The evaluation of potential biomarkers associated with the efficacy of Phoenix® WinNonlin® Version 8.0 (Certara, L.P., Princeton, New Jer­
berzosertib in combination with gemcitabine was an exploratory sey, USA).
objective.
3. Results
2.5. Assessments and endpoints
3.1. Patient demographics and disposition
The safety profile was continuously monitored clinically and with
standard laboratory parameters. TEAEs were coded according to the Baseline and disease history characteristics of all 38 patients enrolled
Medical Dictionary for Regulatory Activities (MedDRA) v21.0 [26] and are presented in Table 1. For those patients whose baseline genotype
graded according to the National Cancer Institute (NCI) Common Ter­ was determined, TP53 mutations were found in 60.5% of tumors. All
minology Criteria for Adverse Events (CTCAE) v4.0 [27]. patients except for one (37, 97.4%) received at least one dose of ber­
zosertib and 38 (100.0%) patients received at least one dose of
To evaluate the efficacy of berzosertib in combination with gemci­ gemcitabine.
tabine, tumor assessments were performed every two cycles for the first
12 cycles, then every two to three cycles, and 5 (±1) weeks after 3.2. Safety
completion of therapy. Tumor response assessments were classified ac­
cording to RECIST v.1.1 [24]. The ORR (primary efficacy endpoint) was All 38 patients experienced TEAEs, 36 (94.7%) of whom experienced
defined as the proportion of participants who achieved a best overall berzosertib or gemcitabine-related TEAEs (Table 2). The most common
response (BOR) of partial response (PR) or complete response (CR) TEAEs (of any grade) were fatigue (55.3%), anemia (52.6%), and nausea
(summarized as objective response [OR]), where both CR and PR were (39.5%). The most common berzosertib-related TEAEs were fatigue
confirmed by repeat assessments performed no < 4 weeks after the (44.7%), anemia (39.5%), and thrombocytopenia (28.9%). Further­
criteria for response were first met. The ORR was calculated with the more, 22 (57.9%) patients experienced berzosertib-related grade ≥ 3
two-sided 90% confidence interval (CI) using the Clopper–Pearson TEAEs and 14 (36.8%) patients experienced berzosertib-related serious
method [28]. TEAEs.

The efficacy of berzosertib in combination with gemcitabine was Five patients experienced berzosertib-related TEAEs (neutropenia,
further explored through the assessment of progression-free survival thrombocytopenia, fatigue, aspartate aminotransferase, and ALT in­
(PFS), duration of response (DOR), overall survival (OS), and disease creases) leading to a dose reduction in berzosertib. Eleven (28.9%) pa­
control. tients discontinued treatment with berzosertib, including seven (18.4%)
patients who discontinued primarily due to TEAEs; three (7.9%) of
Blood samples for PK analysis of berzosertib in plasma were collected whom due to berzosertib-related TEAEs (anemia, thrombocytopenia,
pre-dose, mid-infusion, and 0, 0.5, 1, 2, 3, and optionally 7 h after the
end of the 1-hour berzosertib infusion on cycle 1 day 2. Bioanalysis in Table 1
plasma was performed using validated liquid chromatography-tandem Patient demographics and baseline characteristics.
mass spectrometry (LC-MS/MS) methods [29].
Characteristic Total
Archival tumor biopsies were collected to assess TP53 status and N = 38
other genetic alterations by DNA next–generation sequencing (NGS)
with FoundationOne® CDx NGS assay (Foundation Medicine, Cam­ Sex, n (%) 20 (52.6)
bridge, Massachusetts, US) [30]. A post-hoc exploratory analysis was Male 18 (47.4)
conducted to investigate the potential association between specific ge­ Female
netic tumor alterations and treatment outcomes. The loss of heterozy­ 34 (89.5)
gosity (LOH), tumor mutational burden (TMB) and microsatellite Race, n (%) 1 (2.6)
instability (MSI) results were discretized according to the criteria White 1 (2.6)
established by Foundation Medicine [31]. LOH levels were reported as a Asian 2 (5.3)
percentage of the affected genome and were discretized to either high Other 62.5 (36–76)
LOH (LOH score ≥ 16) or low LOH (LOH score < 16). TMB levels were Unknown
classified as high, if somatic mutations per megabase (MB) were ≥ 20; 11 (28.9)
intermediate, if somatic mutations per MB were ≥ 6 and < 20; and low, Median (range) age, years 27 (71.1)
if somatic mutations per MB were < 6. Baseline ECOG PS,a n (%)
38 (100.0)
2.6. Statistical analysis 0 8 (21.1)
1 6 (15.8)
Based on historical response rates of 10% for single-agent gemcita­ Prior anticancer therapy, n (%) 5 (13.2)
bine in second-line NSCLC [5], and an enrollment of 30 patients, six Chemotherapyb
responders would result in an exact one-sided 90% CI for ORR of Investigational therapy 3 (7.9)
9.1–35.7%. The probability to observe at least six responders was Immunotherapy 7 (18.4)
calculated under the assumption of different response rates. In case of a Other 36 (94.7)
true response rate of 30%, the likelihood of at least six responders was Number of previous anticancer chemotherapy regimens, n (%) 22 (57.9)
84%. Neoadjuvant 16 (42.1)
Adjuvant
The modified full analysis set included all patients who had a base­ 1st line, metastatic disease 6 (15.8)
line tumor assessment with a measurable target lesion and at least one 2nd line, metastatic disease 23 (60.5)
dose of the study drug. The safety analysis set included all patients who >2nd line, metastatic disease 9 (23.7)
received at least one dose of study drug. The PK analysis set included all TP53,c n (%)
patients who received at least one dose of study drug and who provided Wild-type
at least one measurable post-dose concentration. Mutant
Unknown
Summary statistics were provided for berzosertib concentrations by
group and time and for berzosertib PK parameters. PK parameters were aData reported for modified Full Analysis Set.
bSeven patients received prior treatment with gemcitabine.
cOnly patients with biomarker status determined by FoundationOne® CDx next-
generation sequencing were reported.
Abbreviations: ECOG PS, Eastern Cooperative Oncology Group performance
status; TP53, tumor protein p53.

21

R. Plummer et al. Lung Cancer 163 (2022) 19–26

Table 2 Table 3
Overview of TEAEs for berzosertib + gemcitabine occurring in > 20% of patients Best overall response (modified full analysis set; N = 38).
by preferred term (N = 38; safety analysis set).
Efficacy outcome Patients, n (%)

Patients, n (%) Berzosertib + Best overall response
gemcitabine CR
N = 38 PR 0
SD 4 (10.5)
Any Grade ≥ PD 22 (57.9)
grade 3 Not evaluable 7 (18.4)
5 (13.2)
All TEAEs 38 33 ORR, n (%), [90% CI] 4 (10.5) [3.7–22.5]
(100.0) (86.8) DCR,a n (%), [90% CI] 26 (68.4) [53.9–80.7]
Berzosertib-related TEAE 34 (89.5) 22 Median PFS (months), [90% CI] 4.0 [3.2–5.0]
(57.9) Median OS (months), [90% CI] 7.4 [5.4–8.5]
Berzosertib- or gemcitabine-related TEAE 36 (94.7) 24 Median DOR (months), [90% CI] 6.0 [3.6–nd]
(63.2)
TEAEs occurring in ≥ 20% of patients Any Grade aDisease control was defined as a BOR of CR, PR, or SD and the DCR was
grade 3–4 calculated with a two-sided 90% CI using the Clopper–Pearson method.
Fatigue 21 (55.3) 9 (23.7) Abbreviations: BOR, best overall response; CI, confidence interval; CR, complete
Anemia 20 (52.6) 8 (21.1) response; DCR, disease control rate; DOR, duration of response; nd, not defined;
Nausea 15 (39.5) 0 ORR, objective response rate; OS, overall survival; PD, progressive disease; PFS,
Dyspnea 14 (36.8) 5 (13.2) progression-free survival; PR, partial response; SD, stable disease.
Pyrexia 13 (34.2) 1 (2.6)
Thrombocytopenia 13 (34.2) 7 (18.4) erlotinib).
Vomiting 12 (31.6) 2 (5.3) A 75-year-old male patient with anaplastic large-cell lymphoma ki­
AST increased 11 (28.9) 2 (5.3)
Decreased appetite 11 (28.9) 1 (2.6) nase wildtype, EGFR wild-type, programmed death-ligand 1 negative
Neutropenia 11 (28.9) 5 (13.2) NSCLC (adenocarcinoma), with evidence of lymph node metastases,
ALT increased 10 (26.3) 2 (5.3) with progression through two aggressive regimens (carboplatin +
Lower RTI 9 (23.7) 5 (13.2) pemetrexed, docetaxel + nintedanib followed by nintedanib mainte­
Headache 8 (21.1) 0 nance), achieved a confirmed, durable PR lasting 6.0 months.
WBC count decreased 8 (21.1) 1 (2.6)
Serious TEAEs 24 (63.2) 22 3.4. Exploratory biomarker analyses
(57.9)
Berzosertib-related serious TEAE 14 (36.8) 1 (2.6) Archival tumor biopsies from the 38 patients were analyzed by DNA
Berzosertib- or gemcitabine-related serious TEAE 14 (36.8) 10 NGS, of which nine samples did not pass the laboratory quality check
(26.3) process. Overall, 29 samples were included in the data analysis. The
Serious TEAE leading to dose interruption 9 (23.7) NR subgroup analysis did not demonstrate a clear association between
Serious TEAE leading to dose interruption in berzosertib 8 (21.1) NR clinical outcome (ORR and PFS) and any alterations in 324 genes
Serious TEAE leading to dose interruption in gemcitabine 8 (21.1) NR explored, including TP53, and other potential predictive biomarkers of
8 (21.1) NR sensitivity to ATR inhibition including ATM, ARID1A (Supplementary
TEAE leading to dose reduction in ≥ 1 study drug 6 (15.8) NR Table 1) and SMARCA4 (Table 4).
TEAE leading to dose reduction in berzosertib 7 (18.4) NR
TEAE leading to dose reduction in gemcitabine 8 (21.1) NR Regarding genomic signatures, LOH was measurable in 19 patients.
The ORR was 30.0% (3/10, 90% CI: 9.0–61.0%) in patients with high
TEAE leading to permanent discontinuation of ≥ 1 study 3 (7.9) NR LOH, and 11.0% (1/9, 90% CI: 1.0–43.0%) in patients with low LOH.
drug TMB was measurable in 25 patients. The ORR was 33.0% (2/6, 90% CI:
Berzosertib-related TEAE leading to permanent 4 (10.5) NR 6.0–73.0%) in patients with high TMB, 12.5% (2/16, 90% CI:
discontinuation of berzosertib 2.0–34.0%) in patients with intermediate TMB, and 0% (0/3, 90% CI:
Gemcitabine-related TEAE leading to permanent 4 (10.5) 0.0–53.6%) in patients with low TMB.
discontinuation of gemcitabine 1 (2.6)
3.5. Pharmacokinetics
TEAE leading to death
Berzosertib-related TEAE leading to death The berzosertib 210 mg/m2 (IV) dose administered in this study was
within the dose range previously shown to exhibit dose-dependent
ALT, alanine aminotransferase; AST, aspartate aminotransferase; NR, not re­ berzosertib PK as monotherapy, or in combination with either carbo­
ported; RTI, respiratory tract infection; TEAE, treatment-emergent adverse platin or gemcitabine [11,22]. The observed berzosertib concentration
event; WBC, white blood cell. data in this expansion cohort were consistent with those reported pre­
viously at the same dose level [22]. Gemcitabine demonstrated no
and fatigue). Four (10.5%) patients experienced a TEAE leading to apparent effect on berzosertib pharmacokinetics (Fig. 2).
death. One of the four deaths was related to study treatment (hemop­
tysis, hypovolemic shock, and cardiac arrest); the death occurred after The geometric mean (percentage coefficient of variation [%CV])
the patient had experienced a grade 3 lower respiratory tract infection maximum observed concentration (Cmax) of berzosertib was 882 ng/mL
and shortness of breath (both unrelated to treatment) for > 1 month. (55.2%), which was similar to the previously reported Cmax of 899 ng/
mL in part A of this study [22]. A population PK model was developed
3.3. Efficacy based on pooled data from two phase 1 studies, including this expansion
cohort [32]. The model confirmed that gemcitabine had no apparent
The median treatment duration for berzosertib in combination with effect on berzosertib PK, and that berzosertib PK in patients with NSCLC
gemcitabine was 14.0 weeks (2.0–63.0 weeks). There were four was comparable to patients with other advanced solid tumors.
confirmed partial responders (10.5%), two of which had a particularly
notable response (Table 3, Fig. 1).

Amongst the responders, it is worth noting that a 67-year-old female
with epidermal growth factor receptor (EGFR) wild-type NSCLC
(adenocarcinoma), with evidence of lymph node and lung metastases,
achieved a confirmed PR lasting 13.2 months (57.6% maximum tumor
shrinkage). This patient was heavily pretreated with several different
anticancer regimens (carboplatin + pemetrexed + bevacizumab, fol­
lowed by pemetrexed + bevacizumab as maintenance, nivolumab, and

22

R. Plummer et al. Lung Cancer 163 (2022) 19–26

Fig. 1. Best percentage change in tumor size from baseline (modified full analysis set) with genetic profiles. Only patients with a baseline scan, at least one post-
baseline assessment, and at least one response assessment are included in Fig. 1 (n = 34). Only patients with biomarker status determined by FoundationOne® CDx
next-generation sequencing were reported. The dashed line at 20% represents PD whereas the dashed line at –30% represents PR. ARID1A, AT-rich interaction
domain 1A; ATM, ataxia telangiectasia mutated; LOH, loss of heterozygosity; NE, not evaluable; PD, progressive disease; PR, partial response; SMARCA4, switch/
sucrose non-fermentable related, matrix associated, actin dependent regulator of chromatin, subfamily A, member 4; SD, stable disease; TMB, tumor mutational
burden; TP53, tumor protein p53.

Table 4 responder among the few cases who did not have TP53 alterations (the
Objective response rate for selected biomarker subgroups (modified full analysis study was designed to enrich for tumors with TP53 mutations). This
set). clinical finding suggests that TP53 mutations alone are insufficient to
enhance the efficacy of the berzosertib–gemcitabine combination in
Subgroups Mutation statusa Responders ORR [90% CI] patients with advanced NSCLC, despite the fact that preclinical experi­
ments have highlighted TP53 mutations as an efficacy surrogate for
Overall Overall 4/38 10.5 [3.7–22.5] treatment with ATR inhibitors [14].
TP53 Wild-type 0/6 0.0 [0.0–39.3]
ARID1A Mutant 4/23 17.4 [6.2–35.5] The exploratory biomarker subgroup analysis also demonstrated no
ATM Wild-type 4/27 14.8 [5.2–30.8] clear association between treatment outcome (ORR and PFS) and gene
SMARCA4 Mutant 0/2 0.0 [0.0–77.6] alterations, including ATM, ARID1A and SMARCA4, which have previ­
LOH Wild-type 4/23 17.4 [6.2–35.5] ously been associated with sensitivity to ATR inhibition [17–19]. There
TMB Mutant 0/6 0.0 [0.0–39.3] was also no association observed between treatment response and al­
Wild-type 4/24 16.7 [5.9–34.2] terations in other cell cycle genes such as CCNE1, MYC, or RB, whose
Mutant 0/5 0.0 [0.0–45.1] dysregulation is associated with DNA replication stress [33,34].
High 3/10 30.0 [8.7–60.7]
Low 1/9 11.1 [0.6–42.9] However, since only single digit cases carrying genetic alterations in
High 2/6 33.3 [6.3–72.9] each of these genes were identified, the relationship between treatment
Intermediate 2/16 12.5 [2.3–34.4] response and these individual genomic alterations could not be deter­
Low 0/3 0 [0.0–63.2] mined. Additionally, there were limitations to our analyses, including
the lack of genomic data that would have enabled the evaluation of
aOnly patients with high impact or predicted high impact mutations reported. biomarker zygosity status of tumor suppressor genes on treatment out­
Only patients with biomarker status determined by FoundationOne® CDx next- comes. Another limitation was the lack of confirmation of the observed
generation sequencing were reported. The ORR was calculated with the two- ATM, SMARCA4, ARID1A or other tumor suppressor loss at the protein
sided 90% CI using the Clopper–Pearson method. level. Although ATM immunohistochemistry was planned, the available
Abbreviations: ATM, ataxia telangiectasia mutated; ARID1A, AT-rich interaction tumor samples were largely exhausted after DNA NGS. As the clinical
domain 1A; CI, confidence interval; LOH, loss of heterozygosity; ORR, objective development of berzosertib continues, further investigations are
response rate; TMB, tumor mutational burden; TP53, tumor protein p53; required to identify genomic alterations that confer susceptibility to ATR
SMARCA4, switch/sucrose non-fermentable related, matrix associated, actin inhibition. Identifying such alterations, with confirmation at the protein
dependent regulator of chromatin, subfamily A, member 4 level, may ultimately help define patient populations most likely to
benefit from the addition of berzosertib to DNA damage-inducing che­
4. Discussion motherapies or other anticancer therapies.

In this phase 1b expansion cohort study, the combination of the ATR We observed a trend towards increased response rate in patients with
inhibitor berzosertib with gemcitabine, according to the dosing regimen high TMB (33.0%) and LOH scores (30.0%), versus those patients with
previously determined in the dose escalation portion of the trial (ber­ low TMB and LOH scores, respectively. TMB and LOH are markers of
zosertib 210 mg/m2 [days 2 and 9] and gemcitabine 1000 mg/m2 [days genetic instability and homologous recombination deficiency [35,36].
1 and 8] every 3 weeks) [22], was well tolerated in patients with pre- High TMB is also indicative of DNA DSB repair deficiency [37]. In fact,
treated advanced NSCLC. The safety profile was consistent with that both TMB and LOH are emerging as predictive biomarkers to poly (ADP-
of the individual agents [11,23]. However, the observed clinical efficacy ribose) polymerase and immune checkpoint inhibitors [36,38–40],
(ORR of 10.5% and median treatment duration of 14.0 weeks) suggests which could well synergize with increased DNA damage resulting from
limited benefit of combining an ATR inhibitor with gemcitabine in this the combination of berzosertib and gemcitabine. However, the associ­
unselected population of patients with advanced NSCLC, with and ations were not statistically significant, likely due to the small sample
without TP53 mutations, when compared with historical controls of size. Nevertheless, given the biologic rationale for an association
gemcitabine monotherapy [5]. The ORR was only marginally higher in
the subgroup of patients with TP53 mutations, because there was no

23

R. Plummer et al. Lung Cancer 163 (2022) 19–26

Fig. 2. Plasma berzosertib concentration–time profile after the first intravenous infusion of berzosertib monotherapy and of berzosertib in combination with
gemcitabine. an = 3 for the 8-hour timepoint. bn = 1 for the 3-hour timepoint. StD, standard deviation.

between TMB and LOH and a higher sensitivity to DDR inhibitors, Writing – review & editing. Alexander I. Spira: Investigation, Re­
further investigations in this direction are warranted. sources, Data curation, Writing – review & editing, Project administra­
tion, Funding acquisition. Jason M. Melear: Investigation, Resources,
5. Conclusions Writing – review & editing. Ki Y. Chung: Investigation, Resources, Data
curation, Writing – review & editing. Jordi Ferrer-Playan: Writing –
The combination of berzosertib and gemcitabine in patients with review & editing, Supervision, Project administration. Thomas God­
advanced, pre-treated NSCLC was well tolerated, but given the observed demeier: Formal analysis, Writing – review & editing, Visualization.
clinical efficacy, future clinical trials may best be undertaken in a Giuseppe Locatelli: Conceptualization, Formal analysis, Investigation,
genomically selected patient population. In other malignancies, such as Resources, Writing – review & editing, Visualization, Supervision.
platinum-resistant ovarian cancer, the combination of berzosertib and Jennifer Dong: Formal analysis, Writing – review & editing. Patricia
gemcitabine has shown an encouraging efficacy signal, serving as a Fleuranceau-Morel: Formal analysis, Writing – review & editing. Ivan
reminder of the molecular heterogeneity and notable clinical differences Diaz-Padilla: Writing – review & editing, Supervision. Geoffrey I.
between disease entities [12]. Shapiro: Conceptualization, Investigation, Resources, Writing – review
& editing, Supervision.
Role of the funding source
Declaration of competing interest
The trial was sponsored by Merck Healthcare KGaA, Darmstadt,
Germany (CrossRef Funder ID: https://doi.org/10.13039/100009945) Ruth Plummer: received honoraria for serving on advisory boards
and Vertex Pharmaceuticals Incorporated., Boston, MA, USA. Both on behalf of Vertex and Merck Healthcare KGaA, Darmstadt, Germany
sponsors were involved in the development of the study design and the relating to this compound; reimbursement for her institution for clinical
collection of data. Merck Healthcare KGaA, Darmstadt, Germany, pro­ trials costs; consultancy services to Pierre Faber, Bayer, Octimet, Clovis
vided financial support for the preparation of this article, and provided Oncology, Novartis, Karus Therapeutics, Biosceptre, BMS, Cybrexa, El­
technical assistance for the analysis and interpretation of data. The de­ lipses, CV6 Therapeutics, Astex Therapeutics, and Sanofi Aventis;
cision to submit the article for publication was made by the authors research funding from AstraZeneca; and medical writing support for
together with Merck Healthcare KGaA, Darmstadt, Germany. developing this manuscript from Merck Healthcare KGaA, Darmstadt
Germany.
CRediT authorship contribution statement
Emma Dean: is an employee and stakeholder of AstraZeneca; and
Ruth Plummer: Conceptualization, Formal analysis, Investigation, received medical writing support for developing this manuscript from
Resources, Writing – review & editing, Supervision. Emma Dean: Merck Healthcare KGaA, Darmstadt Germany.
Investigation, Resources, Data curation, Writing – review & editing,
Supervision. Hendrik-Tobias Arkenau: Investigation, Resources, Hendrik-Tobias Arkenau: reports honoraria from Bicycle Thera­
Writing – review & editing. Charles Redfern: Investigation, Resources, peutics, Biontech, Bayer, Beigene, Servier, Roche and Guardant Health;
served as an advisor/consultant for Bicycle Therapeutics, Biontech,
Bayer, Beigene, Servier, Roche and Guardant Health; received support

24

R. Plummer et al. Lung Cancer 163 (2022) 19–26

for attending meetings from Amgen; participated on a Data Safety Healthcare KGaA, Darmstadt, Germany, provided technical and opera­
Monitoring Board or Advisory Board for Beigene; and received medical tional support of biomarker sample analysis. Annick Seithel-Keuth, PhD,
writing support for developing this manuscript from Merck Healthcare an employee of Merck Healthcare KGaA, Darmstadt, Germany,
KGaA, Darmstadt Germany. contributed to the analysis and interpretation of PK data. Bart Hendriks,
PhD, a former employee of Merck Healthcare KGaA, Darmstadt, Ger­
Charles Redfern: received medical writing support for developing many, contributed to the analysis and interpretation of PK data. Medical
this manuscript from Merck Healthcare KGaA, Darmstadt Germany. writing assistance was provided by David Lester of Bioscript Stirling Ltd,
Macclesfield, UK and funded by Merck Healthcare KGaA, Darmstadt,
Alexander I Spira: reports grants or contracts from BMS, Mirati, Germany (CrossRef Funder ID: 10.13039/100009945).
Amgen, Novartis, Merck KGaA, Darmstadt, Germany, AstraZeneca,
Sanofi, Abbvie, Cytomx, Macrogenics, Pfizer; and consulting fees from Appendix A. Supplementary data
Mirati, Amgen, Sanofi, AstraZeneca, Merck KGaA, Darmstadt, Germany,
BMS; and received medical writing support for developing this manu­ Supplementary data to this article can be found online at https://doi.
script from Merck Healthcare KGaA, Darmstadt Germany. org/10.1016/j.lungcan.2021.11.011.

Jason M Melear: reports honoraria from Janssen Pharmaceuticals, References
TG Therapeutics and A2 Biotherapeutics; and received medical writing
support for developing this manuscript from Merck Healthcare KGaA, [1] C.A. Rabik, M.E. Dolan, Molecular mechanisms of resistance and toxicity associated
Darmstadt Germany. with platinating agents, Cancer Treat. Rev. 33 (1) (2007) 9–23, https://doi.org/
10.1016/j.ctrv.2006.09.006.
Ki Y Chung: received medical writing support for developing this
manuscript from Merck Healthcare KGaA, Darmstadt Germany. [2] N.A. Howlader, M. Krapcho, J. Garshell, N. Neyman, S.F. Altekruse, C.L. Kosary, M.
Yu, J. Ruhl, Z. Tatalovich, H. Cho, A. Mariotto, D.R. Lewis, H.S. Chen, E.J. Feuer, K.
Jordi Ferrer-Playan: is an employee of Ares Trading SA, Eysins, A. Cronin, SEER Cancer Statistics Review, 1975-2010, 2013. http://seer.cancer.go
Switzerland, an affiliate of Merck KGaA, Darmstadt, Germany; and v/csr/1975_2010/sections.html.
received medical writing support for developing this manuscript from
Merck Healthcare KGaA, Darmstadt Germany. [3] B. Besse, A. Adjei, P. Baas, P. Meldgaard, M. Nicolson, L. Paz-Ares, M. Reck, E.
F. Smit, K. Syrigos, R. Stahel, E. Felip, S. Peters, M. Panel Esmo, 2nd ESMO
Thomas Goddemeier: is an employee of Merck Healthcare KGaA, Consensus Conference on Lung Cancer: non-small-cell lung cancer first-line/second
Darmstadt, Germany; and received medical writing support for devel­ and further lines of treatment in advanced disease, Ann Oncol 25 (8) (2014)
oping this manuscript from Merck Healthcare KGaA, Darmstadt 1475–1484, https://doi.org/10.1093/annonc/mdu123.
Germany.
[4] A. Kumar, H. Wakelee, Second- and third-line treatments in non-small cell lung
Giuseppe Locatelli: is an employee of Merck Healthcare KGaA, cancer, Curr. Treat. Options Oncol. 7 (1) (2006) 37–49, https://doi.org/10.1007/
Darmstadt, Germany; and received medical writing support for devel­ s11864-006-0030-9.
oping this manuscript from Merck Healthcare KGaA, Darmstadt
Germany. [5] C. Manegold, G. Koschel, D. Hruska, K. Scott-von-Romer, J. Mezger, L.R. Pilz,
Open, randomized, phase II study of single-agent gemcitabine and docetaxel as
Jennifer Dong: is an employee of EMD Serono Research & Devel­ first- and second-line treatment in patients with advanced non-small-cell lung
opment Institute, Inc., Billerica, MA, USA, an affiliate of Merck KGaA; cancer, Clin Lung Cancer 8 (4) (2007) 245–251, https://doi.org/10.3816/clc.2007.
and received medical writing support for developing this manuscript n.001.
from Merck Healthcare KGaA, Darmstadt Germany.
[6] Y.H. Ko, M.A. Lee, Y.S. Hong, K.S. Lee, H.J. Park, R. Yoo Ie, Y.S. Kim, Y.K. Kim, K.
Patricia Fleuranceau-Morel: is an employee of EMD Serono H. Jo, Y.P. Wang, K.Y. Lee, J.H. Kang, Docetaxel monotherapy as second-line
Research & Development Institute, Inc., Billerica, MA, USA, an affiliate treatment for pretreated advanced non-small cell lung cancer patients, Korean J.
of Merck KGaA; and received medical writing support for developing Intern. Med. 22(3) (2007) 178-85 10.3904/kjim.2007.22.3.178.
this manuscript from Merck Healthcare KGaA, Darmstadt Germany.
[7] D.H. Kang, J.O. Kim, S.S. Jung, H.S. Park, C. Chung, D. Park, J.E. Lee, Efficacy of
Ivan Diaz-Padilla: is a former employee of Ares Trading SA, Eysins, vinorelbine monotherapy as third- or further-line therapy in patients with
Switzerland, an affiliate of Merck KGaA, Darmstadt, Germany; is a advanced non-small-cell lung cancer, Oncology 97 (6) (2019) 356–364, https://
current employee of GlaxoSmithKline, Zug, Switzerland; and received doi.org/10.1159/000502343.
medical writing support for developing this manuscript from Merck
Healthcare KGaA, Darmstadt Germany. [8] A.N. Blackford, S.P. Jackson, ATM, ATR, and DNA-PK: The Trinity at the Heart of
the DNA Damage Response, Mol. Cell 66(6) (2017) 801-817 10.1016/j.
Geoffrey I Shapiro: reports grants/contracts from Eli Lilly, Merck & molcel.2017.05.015.
Co., Inc., Sierra Oncology and Pfizer; holds a patent entitled, ‘Dosage
regimen for sapacitabine and seliciclib’, also issued to Cyclacel Phar­ [9] M.K. Zeman, K.A. Cimprich, Causes and consequences of replication stress, Nat.
maceuticals, and a pending patent, entitled, ‘Compositions and Methods Cell Biol. 16 (1) (2014) 2–9, https://doi.org/10.1038/ncb2897.
for Predicting Response and Resistance to CDK4/6 Inhibition’; received
medical writing support from Pfizer; and received medical writing [10] A.B. Hall, D. Newsome, Y. Wang, D.M. Boucher, B. Eustace, Y. Gu, B. Hare, M.A.
support for developing this manuscript from Merck Healthcare KGaA, Johnson, S. Milton, C.E. Murphy, D. Takemoto, C. Tolman, M. Wood, P. Charlton,
Darmstadt Germany. J.D. Charrier, B. Furey, J. Golec, P.M. Reaper, J.R. Pollard, Potentiation of tumor
responses to DNA damaging therapy by the selective ATR inhibitor VX-970,
Acknowledgements Oncotarget. 5(14) (2014) 5674-85 10.18632/oncotarget.2158.

The authors would like to thank patients, investigators, co- [11] T.A. Yap, B. O’Carrigan, M.S. Penney, J.S. Lim, J.S. Brown, M.J. de Miguel Luken,
investigators, and the study teams at each of the participating centers N. Tunariu, R. Perez-Lopez, D.N. Rodrigues, R. Riisnaes, I. Figueiredo, S. Carreira,
and at Merck Healthcare KGaA, Darmstadt, Germany. The authors thank B. Hare, K. McDermott, S. Khalique, C.T. Williamson, R. Natrajan, S.J. Pettitt, C.J.
Vertex Pharmaceuticals for their involvement in the development of Lord, U. Banerji, J. Pollard, J. Lopez, J.S. de Bono, Phase I Trial of First-in-Class
berzosertib (formerly M6620, VX-970). The UK study sites received ATR Inhibitor M6620 (VX-970) as Monotherapy or in Combination With
support as part of the Experimental Cancer Medicine Centres. Thomas Carboplatin in Patients With Advanced Solid Tumors, J. Clin. Oncol. 38(27) (2020)
Grombacher, PhD, an employee of Merck Healthcare KGaA, Darmstadt, 3195-3204 10.1200/JCO.19.02404.
Germany, contributed to the analysis and interpretation of the
biomarker results. Danyi Wang, MD, PhD, an employee of EMD Serono [12] P.A. Konstantinopoulos, S.C. Cheng, A.E. Wahner Hendrickson, R.T. Penson, S.T.
Research & Development Institute, Inc., Billerica, MA, USA, an affiliate Schumer, L.A. Doyle, E.K. Lee, E.C. Kohn, L.R. Duska, M.A. Crispens, A.B.
of Merck KGaA, provided technical and operational support of Olawaiye, I.S. Winer, L.M. Barroilhet, S. Fu, M.T. McHale, R.J. Schilder, A.
biomarker sample analysis. Jens Derbinski, PhD, an employee of Merck Farkkila, D. Chowdhury, J. Curtis, R.S. Quinn, B. Bowes, A.D. D’Andrea, G.I.
Shapiro, U.A. Matulonis, Berzosertib plus gemcitabine versus gemcitabine alone in
platinum-resistant high-grade serous ovarian cancer: a multicentre, open-label,
randomised, phase 2 trial, Lancet Oncol. 21(7) (2020) 957-968 10.1016/S1470-
2045(20)30180-7.

[13] A. Thomas, N. Takahashi, V.N. Rajapakse, X. Zhang, Y. Sun, M. Ceribelli, K.M.
Wilson, Y. Zhang, E. Beck, L. Sciuto, S. Nichols, B. Elenbaas, J. Puc, H. Dahmen, A.
Zimmermann, J. Varonin, C.W. Schultz, S. Kim, H. Shimellis, P. Desai, C. Klumpp-
Thomas, L. Chen, J. Travers, C. McKnight, S. Michael, Z. Itkin, S. Lee, A. Yuno, M.-
J. Lee, C.E. Redon, J.D. Kindrick, C.J. Peer, J.S. Wei, M.I. Aladjem, W.D. Figg, S.M.
Steinberg, J.B. Trepel, F.T. Zenke, Y. Pommier, J. Khan, C.J. Thomas, Therapeutic
targeting of ATR yields durable regressions in small cell lung cancers with high
replication stress, Cancer Cell. 39(4) (2021) 566-579.e7 10.1016/j.
ccell.2021.02.014.

[14] P.M. Reaper, M.R. Griffiths, J.M. Long, J.D. Charrier, S. Maccormick, P.A. Charlton,
J.M. Golec, J.R. Pollard, Selective killing of ATM- or p53-deficient cancer cells

25

R. Plummer et al. Lung Cancer 163 (2022) 19–26

through inhibition of ATR, Nat. Chem. Biol. 7 (7) (2011) 428–430, https://doi.org/ [26] MedDRA, Medical Dictionary for Regulatory Activities Version 21.0, 2018. https://
10.1038/nchembio.573. admin.new.meddra.org/sites/default/files/guidance/file/dist_file_format_21_0_en
[15] K.T. Bieging, S.S. Mello, L.D. Attardi, Unravelling mechanisms of p53-mediated glish.pdf.
tumour suppression, Nat. Rev. Cancer 14 (5) (2014) 359–370, https://doi.org/
10.1038/nrc3711. [27] NCI, Common Terminology Criteria for Adverse Events (CTCAE) Version 4.0, 2009.
[16] A. Mogi, H. Kuwano, TP53 mutations in nonsmall cell lung cancer, J. Biomed. https://evs.nci.nih.gov/ftp1/CTCAE/CTCAE_4.03/Archive/CTCAE_4.0_2009-05
Biotechnol. 2011 (2011), 583929, https://doi.org/10.1155/2011/583929. -29_QuickReference_8.5x11.pdf.
[17] T.A. Yap, D.S.P. Tan, A. Terbuch, R. Caldwell, C. Guo, B.C. Goh, V. Heong, N.R.
M. Haris, S. Bashir, Y. Drew, D.S. Hong, F. Meric-Bernstam, G. Wilkinson, J. Hreiki, [28] C.J. Clopper, E.S. Pearson, The Use of Confidence or Fiducial Limits Illustrated in
A.M. Wengner, F. Bladt, A. Schlicker, M. Ludwig, Y. Zhou, L. Liu, S. Bordia, the Case of the Binomial, Biometrika 26 (4) (1934) 404–413, https://doi.org/
R. Plummer, E. Lagkadinou, J.S. de Bono, First-in-Human Trial of the Oral Ataxia 10.1093/biomet/26.4.404.
Telangiectasia and RAD3-Related (ATR) Inhibitor BAY 1895344 in Patients with
Advanced Solid Tumors, Cancer Discov (2020), https://doi.org/10.1158/2159- [29] C. Medpace Bioanalytical Laboratories, OH, USA., Validation of an LC/MS/MS
8290.CD-20-0868. Bioanalytical Method for the Determination of VX-970 Concentration in Human
[18] M. Gupta, C.P. Concepcion, C.G. Fahey, H. Keshishian, A. Bhutkar, C.F. Brainson, F. K2EDTA Plasma, MBL Study No. MBL13206, Report No. RPT13206 and Report
J. Sanchez-Rivera, P. Pessina, J.Y. Kim, A. Simoneau, M. Paschini, M.C. Beytagh, C. Amendment No’s RPT13206-1, 1A and RPT13206-2, 2A.
R. Stanclift, M. Schenone, D.R. Mani, C. Li, A. Oh, F. Li, H. Hu, A. Karatza, R.
T. Bronson, A.T. Shaw, A.N. Hata, K.K. Wong, L. Zou, S.A. Carr, T. Jacks, C.F. Kim, [30] Foundation Medicine, FoundationOne®CDx Technical Information. https://www.
BRG1 Loss Predisposes Lung Cancers to Replicative Stress and ATR Dependency, accessdata.fda.gov/cdrh_docs/pdf17/P170019S006C.pdf. (Accessed 27 January
Cancer Res. 80 (18) (2020) 3841–3854, https://doi.org/10.1158/0008-5472.CAN- 2021).
20-1744.
[19] C.T. Williamson, R. Miller, H.N. Pemberton, S.E. Jones, J. Campbell, A. Konde, [31] FDA, FoundationFocus CDxBRCA LOH - P160018/S001, 2018. https://www.fda.
N. Badham, R. Rafiq, R. Brough, A. Gulati, C.J. Ryan, J. Francis, P.B. Vermulen, A. gov/medical-devices/recently-approved-devices/foundationfocus-cdxbrca-loh-p1
R. Reynolds, P.M. Reaper, J.R. Pollard, A. Ashworth, C.J. Lord, ATR inhibitors as a 60018s001. (Accessed 27 January 2021).
synthetic lethal therapy for tumours deficient in ARID1A, Nat. Commun. 7 (2016)
13837, https://doi.org/10.1038/ncomms13837. [32] N. Terranova, M. Jansen, M. Falk, B.S. Hendriks, Population pharmacokinetics of
[20] X. Lang, M.D. Green, W. Wang, J. Yu, J.E. Choi, L. Jiang, P. Liao, J. Zhou, Q. Zhang, ATR inhibitor berzosertib in phase I studies for different cancer types, Cancer
A. Dow, A.L. Saripalli, I. Kryczek, S. Wei, W. Szeliga, L. Vatan, E.M. Stone, Chemother. Pharmacol. 87 (2) (2021) 185–196, https://doi.org/10.1007/s00280-
G. Georgiou, M. Cieslik, D.R. Wahl, M.A. Morgan, A.M. Chinnaiyan, T.S. Lawrence, 020-04184-z.
W. Zou, Radiotherapy and Immunotherapy Promote Tumoral Lipid Oxidation and
Ferroptosis via Synergistic Repression of SLC7A11, Cancer Discov. 9 (12) (2019) [33] L.M.F. Primo, L.K. Teixeira, DNA replication stress: oncogenes in the spotlight,
1673–1685, https://doi.org/10.1158/2159-8290.Cd-19-0338. Genet Mol Biol 43(1 suppl 1) (2019) e20190138 10.1590/1678-4685GMB-2019-
[21] H. Ogiwara, K. Takahashi, M. Sasaki, T. Kuroda, H. Yoshida, R. Watanabe, A. 0138.
Maruyama, H. Makinoshima, F. Chiwaki, H. Sasaki, T. Kato, A. Okamoto, T. Kohno,
Targeting the Vulnerability of Glutathione Metabolism in ARID1A-Deficient [34] S.A. Hills, J.F. Diffley, DNA replication and oncogene-induced replicative stress,
Cancers, Cancer Cell. 35(2) (2019) 177-190.e8 10.1016/j.ccell.2018.12.009. Curr. Biol. 24 (10) (2014) R435–R444, https://doi.org/10.1016/j.
[22] M.R. Middleton, E. Dean, T.R.J. Evans, G.I. Shapiro, J. Pollard, B.S. Hendriks, M. cub.2014.04.012.
Falk, I. Diaz-Padilla, R. Plummer, Phase 1 study of the ATR inhibitor berzosertib
(formerly M6620, VX-970) combined with gemcitabine ± cisplatin in patients with [35] V. Abkevich, K.M. Timms, B.T. Hennessy, J. Potter, M.S. Carey, L.A. Meyer,
advanced solid tumours, British Journal of Cancer (2021) 10.1038/s41416-021- K. Smith-McCune, R. Broaddus, K.H. Lu, J. Chen, T.V. Tran, D. Williams, D. Iliev,
01405-x. S. Jammulapati, L.M. FitzGerald, T. Krivak, J.A. DeLoia, A. Gutin, G.B. Mills, J.
[23] G.I. Shapiro, R. Wesolowski, C. Devoe, S. Lord, J. Pollard, B.S. Hendriks, M. Falk, S. Lanchbury, Patterns of genomic loss of heterozygosity predict homologous
I. Diaz-Padilla, R. Plummer, T.A. Yap, Phase 1 study of the ATR inhibitor recombination repair defects in epithelial ovarian cancer, Br. J. Cancer 107 (10)
berzosertib in combination with cisplatin in patients with advanced solid tumours, (2012) 1776–1782, https://doi.org/10.1038/bjc.2012.451.
Br. J. Cancer 125 (4) (2021) 520–527, https://doi.org/10.1038/s41416-021-
01406-w. [36] C.E. Steuer, S.S. Ramalingam, Tumor mutation burden: leading immunotherapy to
[24] E.A. Eisenhauer, P. Therasse, J. Bogaerts, L.H. Schwartz, D. Sargent, R. Ford, J. the era of precision medicine? J. Clin. Oncol. 36 (7) (2018) 631–632, https://doi.
Dancey, S. Arbuck, S. Gwyther, M. Mooney, L. Rubinstein, L. Shankar, L. Dodd, R. org/10.1200/JCO.2017.76.8770.
Kaplan, D. Lacombe, J. Verweij, New response evaluation criteria in solid tumours:
revised RECIST guideline (version 1.1), Eur J Cancer 45(2) (2009) 228-47 [37] J. Deng, A. Thennavan, I. Dolgalev, T. Chen, J. Li, A. Marzio, J.T. Poirier, D.
10.1016/j.ejca.2008.10.026. H. Peng, M. Bulatovic, S. Mukhopadhyay, H. Silver, E. Papadopoulos, V. Pyon,
[25] J. Pollard P. Reaper A. Peek S. Hughes S. Gladwell J. Jones P. Chiu M. Wood C. C. Thakurdin, H. Han, F. Li, S. Li, H. Ding, H. Hu, Y. Pan, V. Weerasekara, B. Jiang,
Tolman M. Johnson P. Littlewood M. Penney K. McDermott B. Hare S.Z. Fields M. E.S. Wang, I. Ahearn, M. Philips, T. Papagiannakopoulos, A. Tsirigos,
Asmal B. O’Carrigan T.A. Yap Abstract 3717: Defining optimal dose schedules for E. Rothenberg, J. Gainor, G.J. Freeman, C.M. Rudin, N.S. Gray, P.S. Hammerman,
ATR inhibitors in combination with DNA damaging drugs: Informing clinical M. Pagano, J.V. Heymach, C.M. Perou, N. Bardeesy, K.-K. Wong, ULK1 inhibition
studies of VX-970, the first-in-class ATR inhibitor Cancer Res 76 14 Supplement overcomes compromised antigen presentation and restores antitumor immunity in
2016 3717 3717 10.1158/1538-7445.Am2016-3717. LKB1-mutant lung cancer, Nature Cancer 2 (5) (2021) 503–514, https://doi.org/
10.1038/s43018-021-00208-6.

[38] J.Y. Kim, A. Kronbichler, M. Eisenhut, S.H. Hong, H.J. van der Vliet, J. Kang, J.
I. Shin, G. Gamerith, Tumor mutational burden and efficacy of immune checkpoint
inhibitors: a systematic review and meta-analysis, Cancers (Basel) 11 (11) (2019).

[39] J. Hu, S. Zhang, K. You, L. Chen, P. Zhang, J. Shi, M. Yao, M. Wang, K. Wang,
Abstract 3548: Loss of heterozygosity (LOH) as a candidate biomarker of PARP
inhibitor sensitivity in Chinese solid tumor patients, Cancer Res. 80 (16
Supplement) (2020).

[40] P. Vikas, N. Borcherding, A. Chennamadhavuni, R. Garje, Therapeutic Potential of
Combining PARP Inhibitor and Immunotherapy in Solid Tumors, Front. Oncol. 10
(2020) 570, https://doi.org/10.3389/fonc.2020.00570.

26

Lung Cancer 163 (2022) 27–34
Contents lists available at ScienceDirect

Lung Cancer

journal homepage: www.elsevier.com/locate/lungcan

A risk prediction model for selecting high-risk population for computed
tomography lung cancer screening in China

Lan-Wei Guo a,1, Zhang-Yan Lyu b,1, Qing-Cheng Meng c, Li-Yang Zheng a, Qiong Chen a,
Yin Liu a, Hui-Fang Xu a, Rui-Hua Kang a, Lu-Yao Zhang a, Xiao-Qin Cao a, Shu-Zheng Liu a,
Xi-Bin Sun a, Jian-Gong Zhang a, Shao-Kai Zhang a,*

a Department of Cancer Epidemiology and Prevention, Henan Engineering Research Center of Cancer Prevention and Control, Henan International Joint Laboratory of
Cancer Prevention, The Affiliated Cancer Hospital of Zhengzhou University, Henan Cancer Hospital, Zhengzhou 450008, China
b Department of Cancer Epidemiology and Biostatistics, Tianjin Medical University Cancer Institute and Hospital, National Clinical Research Center for Cancer, Key
Laboratory of Cancer Prevention and Therapy of Tianjin, Tianjin’s Clinical Research Center for Cancer, Key Laboratory of Molecular Cancer Epidemiology of Tianjin,
Key Laboratory of Breast Cancer Prevention and Therapy of the Ministry of Education, Tianjin, China
c Department of Radiology, The Affiliated Cancer Hospital of Zhengzhou University, Henan Cancer Hospital, Zhengzhou 450008, China

ARTICLE INFO ABSTRACT

Keywords: Objective: Two large randomized controlled trials (RCTs) have demonstrated that low dose computed tomography
Lung cancer (LDCT) screening reduces lung cancer mortality. Risk-prediction models have been proved to select individuals
Prospective cohort for lung cancer screening effectively. With the focus on established risk factors for lung cancer routinely available
Risk assessment in general cancer screening settings, we aimed to develop and internally validated a risk prediction model for
lung cancer.
Materials and methods: Using data from the Cancer Screening Program in Urban China (CanSPUC) in Henan
province, China between 2013 and 2019, we conducted a prospective cohort study consisting of 282,254 par­
ticipants including 126,445 males and 155,809 females. Detailed questionnaire, physical assessment and follow-
up were completed for all participants. Using Cox proportional risk regression analysis, we developed the Henan
Lung Cancer Risk Models based on simplified questionnaire. Model discrimination was evaluated by concordance
statistics (C-statistics), and model calibration was evaluated by the bootstrap sampling, respectively.
Results: By 2020, a total of 589 lung cancer cases occurred in the follow-up yielding an incident density of 64.91/
100,000 person-years (pyrs). Age, gender, smoking, history of tuberculosis and history of emphysema were
included into the model. The C-index of the model for 1-year lung cancer risk was 0.766 and 0.741 in the training
set and validation set, respectively. In stratified analysis, the model showed better predictive power in males,
younger participants, and former or current smoking participants. The model calibrated well across the deciles of
predicted risk in both the overall population and all subgroups.
Conclusions: We developed and internally validated a simple risk prediction model for lung cancer, which may be
useful to identify high-risk individuals for more intensive screening for cancer prevention.

Abbreviations: LDCT, low dose computed tomography; USPSTF, the United States Preventive Services Task Force; CanSPUC, the Cancer Screening Program in
Urban China; STROBE, the Strengthening the Reporting of Observational Studies in Epidemiology; ICD-O-3, the International Classification of Diseases for Oncology,
3rd edition; ICD-10, the International Statistical Classification of Diseases and Related Health Problems, 10th edition; BMI, body mass index; HR, hazard ratio; CI,
confidence interval; ROC, receiver-operating characteristic; C-statistics, concordance statistics; AUC, rea under curvea; pyrs, person-years; COPD, chronic obstructive
pulmonary disease; LLP, Liverpool Lung Project; PLCO, Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial; SNPs, single nucleotide polymorphisms; IQR,
interquartile range.

* Corresponding author at: Department of Cancer Epidemiology and Prevention, Henan Engineering Research Center of Cancer Prevention and Control, Henan
International Joint Laboratory of Cancer Prevention, The Affiliated Cancer Hospital of Zhengzhou University, Henan Cancer Hospital, Dongming Road No. 127, PO
Box 0061, Zhengzhou 450008, China.

E-mail address: [email protected] (S.-K. Zhang).
1 Lan-Wei Guo and Zhang-Yan Lyu contributed equally to the article.

https://doi.org/10.1016/j.lungcan.2021.11.015

Received 10 August 2021; Received in revised form 18 November 2021; Accepted 22 November 2021

Available online 1 December 2021 Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
0169-5002/© 2021 The Author(s).

(http://creativecommons.org/licenses/by-nc-nd/4.0/).

L.-W. Guo et al. Lung Cancer 163 (2022) 27–34

1. Introduction recommended to undergo a one-round LDCT examination free of charge
at a tertiary-level hospital designated by the program. Participants were
Lung cancer remains the leading cause of cancer-related mortality followed up from the date of baseline attendance and ended on the date
among both men and women in China and worldwide [1]. Due to the of the first diagnosis of cancer, death, loss to follow-up, or administra­
asymptomatic nature of lung cancer at an early stage, most newly tive censoring (March 2020), whichever came first. During the study
diagnosed lung cancer cases are with advanced-stage disease. Patients period, incident lung cancer cases were collected by follow-up through
with advanced-stage lung cancer have a poor prognosis with 5-year telephone and personal encounters, as well as linking with the provincial
relative survival estimated at less than 10% [2]. In addition, the sur­ cancer registry database for all participants.
vival rate of lung cancer cases in China was even poor (16.1%) [3].
Therefore, detection of lung cancer at an early stage (before the onset of For the present paper, data from the CanSPUC conducted between
lethal metastasis), could have a dramatic effect on the reduction of lung October 2013 and October 2019 in Henan Province, which covered a
cancer mortality. total of 8 cities (Zhengzhou, Zhumadian, Anyang, Luoyang, Nanyang,
Jiaozuo, Puyang, and Xinxiang) was included in the analysis.
It has been established that screening for lung cancer using low dose
computed tomography (LDCT) is an effective screening modality that The study was approved by the Ethics Committee of both the Na­
can reduce the mortality of lung cancer [4,5]. But considering the po­ tional Cancer Center/Cancer Hospital, Chinese Academy of Medical
tential harms from high false positives, over-diagnoses, economic Sciences and Peking Union Medical College, and Henan Cancer Hospital.
burden, and repeated radiation, LDCT screening should be pointed to the All participants have signed written informed consent forms. This study
high-risk population rather than all [6,7]. Consequently, in 2021, the followed the Strengthening the Reporting of Observational Studies in
United States Preventive Services Task Force (USPSTF) annual screening Epidemiology (STROBE) reporting guideline.
for lung cancer with LDCT in adults aged 50 to 80 years who have a 20
pack-year smoking history and currently smoke or have quit within the 2.2. Data collection and quality control
past 15 years [8]. Despite simple and practical, this risk stratification of
smokers for lung cancer screening is ideal because it recommends that We used paper or computer-based standardized document forms
individuals at very low risk are less likely to benefit from high-risk (epidemiological questionnaires, LDCT reports, follow-up information,
screening and excludes individuals who are more likely to benefit pathology reports, etc.) for information collection. A unique identifier
from screening for lung cancer [9,10]. Therefore, accurate identification for each participant is used to track and link all relevant documentation
of high-risk subpopulations to be screened is critical to maximizing the forms for the individual.
efficacy of lung cancer screening.
Lung cancer cases were diagnosed pathologically, radiologically, or
To address this limitation of existing guidelines, several lung cancer clinically by specialists from the tertiary-level hospitals. All lung cancer
risk prediction models have been developed in recent years, primarily incidence cases were confirmed by medical record review by senior
based on established risk factors such as smoking, occupational expo­ thoracic surgeons. For difficult cases, a panel of thoracic surgeons, ra­
sures, family history of lung cancer, and respiratory diseases [11–40]. diologists, and pathologists from the tertiary-level hospitals were con­
However, there is no risk prediction model for lung cancer among the sulted. Lung cancer was coded as C33 and C34 according to the
Chinese mainland population-based on prospective lung cancer International Classification of Diseases for Oncology, 3rd edition (ICD-
screening programs. Considering the huge difference in the prevalence O-3) and the International Statistical Classification of Diseases and
of risk factors for lung cancer such as smoking by sex [41,42], passive Related Health Problems, 10th edition (ICD-10).
smoking [42,43], obesity [44,45], household air pollution [46] between
the populations studied for the prior risk prediction models versus pa­ 2.3. Statistical methods
tients in mainland China, a new model developed and validated among
Chinese cancer screening population is needed. Therefore, in the present Descriptive analyses were performed on the characteristics of the
study, with the focus on established risk factors for lung cancer routinely study population, including recruitment age, sex (binary), education
available in general cancer screening settings, we aimed to develop and (low: elementary school or less, intermediate: elementary to high school,
internally validated a risk prediction model for lung cancer. high school: high school or more), body mass index (BMI, continuous),
family history of lung cancer in first-degree relatives (binary), and his­
2. Methods tory of chronic respiratory disease (binary, mainly including tubercu­
losis, chronic bronchitis, emphysema, asthmatic bronchiectasis, silicosis
2.1. Data source or pneumoconiosis). Ever smokers, including former and current
smokers, were defined as having smoked or currently smoking > 1 time/
Data were obtained from a multi-center population-based cohort day for at least 6 months. Pack-years of smoking (a pack-year was
study in the context of the Cancer Screening Program in Urban China defined as twenty cigarettes per day for a year) and quit years were
(CanSPUC) which is an ongoing national cancer screening program counted for ever smokers only. Categorical variables were described by
initiated in October 2012 and funded by the Ministry of Finance and the percentages and the Chi-square test was used to compare the difference
National Health Commission of China. The target population of Can­ between different groups. Continuous variables were described by
SPUC was 40–74 years old residents (40–69 years old between 2012 and means (standard deviation).
2015), and the targeted types of cancer were the five most prevalent
types of cancer in urban China including lung cancer, female breast From 2013 to 2019, 282,254 participants were included in the
cancer, liver cancer, upper digestive tract cancer. analysis. The entire population was randomly divided by geographic
distribution into a training set (N = 141,127) and a validation set (N =
The details of the CanSPUC could be found in previously published 141,127).
articles [47]. Briefly, after obtaining signed written informed consent,
all the participants completed a questionnaire that included de­ For each risk factor, the association with lung cancer risk was first
mographic characteristics, lifestyle factors, items assessing general assessed adjusting for the age group by Cox proportional risk regression
health and medical history, and family history of common disease. analysis. Stepwise multivariable-adjusted COX regressions (P entry =
Participants also underwent a simple physical assessment including 0.15, P stay = 0.10) were conducted to choose the variables included in
height and weight, and clinical examination for one or more cancer the prediction model. Hazard ratios (HR) and 95% confidence intervals
screening if necessary, according to the CanSPUC protocol. Only par­ (CI) were calculated as estimates of relative risk.
ticipants who were assessed to be at high risk of lung cancer were
The models with the independent risk factors were selected to plot
the nomogram. The nomogram was plotted with RStudio (version
1.4.1103, with packages “Hmisc”, “lattice”, “Formula”, “ggplot2”, and

28

L.-W. Guo et al. Lung Cancer 163 (2022) 27–34

“rms”). Model discrimination was evaluated by receiver-operating Table 1
characteristic (ROC) curves and concordance statistics (C-statistics) Baseline characteristics of the study population.
which measures the ability to distinguish individuals who experience a
lung cancer event from those who do not. Confidence intervals for area Total no. No lung Lung χ2 P-
under curve (AUC) were estimated based on 2000 stratified bootstrap (%) cancer, n cancer, n value
replicates. ROC curves were generated using the “roc” function in the (%) (%)
pROC R package. In addition, the internal validation of model
discrimination was evaluated by ten-fold cross-validation. All participants 282,254 281,665 589 268.63 <0.001
Person-years, 3.21 ± 3.22 ± 1.84 ±
The bootstrap sampling approach was used to evaluate the calibra­ 1.65 1.65 1.36
tion of the present model by comparing the observed and predicted mean ± SD 55.26 ± 55.24 ± 61.14 ±
probabilities. Correction for deviation of estimates from observations Age, mean ± SD, 8.68 8.67 7.28
(overfitting correction) estimates was based on predictions for a subset
of the interval. The median absolute error is also used to evaluate the years 34,871 34,858 13 (2.21)
calibration performance. The calibration function is implemented in the Age (years) (12.35) (12.38)
R package “rms”. 52,290 52,257 33 (5.60)
40–44 (18.53) (18.55)
All analyses were conducted using the SAS or R software. All statis­ 52,869 52,804 65 84.98 <0.001
tical tests were two-sided, and the significance level was set as P value < 45–49 (18.73) (18.75) (11.04)
0.05. 43,189 43,079 110 2.91 0.088
50–54 (15.30) (15.29) (18.68)
3. Results 47,626 47,478 148
55–59 (16.87) (16.86) (25.13)
3.1. Basic characteristics of the study population 37,560 37,403 157
60–64 (13.31) (13.28) (26.66)
A total of 282,254 participants (male: 126,445; female: 155,809) 13,849 13,786 63
with a mean age of 55.26 years were recruited in this study. The rate of 65–69 (4.91) (4.89) (10.70)
smoking, passive smoking, and alcohol drinking was 23.91%, 37.09%,
and 24.46%, respectively. The proportion of participants with a history 70–74 126,445 126,070 375
of chronic respiratory disease, tuberculosis, chronic bronchitis, (44.80) (44.76) (63.67)
emphysema, and asthma bronchiectasis was 17.40%, 1.52%, 13.63%, Gender 155,809 155,595 214
1.43%, and 3.79%, respectively. (Table 1). Male (55.20) (55.24) (36.33)

By March 2020, among 282,254 eligible participants, 589 lung Female 277,212 276,639 573
cancer cases occurred in the follow-up yielding an incident density of (98.21) (98.22) (97.28)
64.91/100,000 person-years (pyrs) (65.41/100,000 pyrs and 64.41/ Race 5042 5026 16 (2.72)
100,000 pyrs in the training and validation tests, respectively). Han nationality (1.79) (1.78)
Compared with participants without lung cancer, lung cancer cases were 21.46 <0.001
more likely to smoke, drink alcohol, have passive smoking exposure, Others 49,246 49,112
have a family history of lung cancer, and have a history of chronic (17.45) (17.44) 134
respiratory disease, tuberculosis, chronic bronchitis, emphysema, and Education a 190,609 190,210 (22.75)
asthma bronchiectasis (all P vales < 0.05) (Table 1). Low (67.53) (67.53) 399
42,399 42,343 (67.74)
In addition, an overview of the characteristics of the training and Medium (15.02) (15.03) 56 (9.51)
validation cohorts was shown in Supplementary Table 1 and Supple­
mentary Table 2. High 3535 3521 9.63 0.022
(1.25) (1.25)
3.2. Predictors included in the model BMI (kg/m2) 121,156 120,884 14 (2.38)
<18.5 (42.92) (42.92)
Table 2 presents the HRs (95% CI) for each predictor. In the training 125,789 125,544 272
set, we found a positive association with high-risk lung cancers for age 18.5–24.0 (44.57) (44.57) (46.18)
(≥50 years: 2.40, 1.15–5.00; ≥55 years: 4.87, 2.41–9.84; ≥60 years: 31,774 31,716 245
5.51, 2.75–11.04; ≥65 years: 7.57, 3.79–15.16; ≥70 years: 11.94, 24.0–28.0 (11.26) (11.26) (41.60)
5.71–24.96), gender (male: 1.72, 1.28–2.31), smoking packyears 58 (9.85)
(30–49: 1.64, 1.10–2.45; ≥50: 1.95, 1.22–3.13), history of tuberculosis ≥28.0 214,764 214,420
(1.88, 1.03–3.45), and history of emphysema (2.09, 1.21–3.61). Thus, (76.09) (76.13) 101.54 <0.001
we used these variables to build the model. The results in the validation Smoking status 53,528 53,335
set were consistent with that in the training set in terms of the direction Never (18.96) (18.94) 344
and degree of associations. We plotted a 1-year, 3-year, and 5-year lung 13,962 13,910 (58.40)
cancer risk prediction nomogram (Fig. 1). Current (4.95) (4.94) 193
(32.77)
3.3. Predictive performance of the model Former 214,764 214,420 52 (8.83)
(76.09) (76.13)
When we stratified participants into three groups according to the Smoking 44,139 44,010 134.44 <0.001
predicted risk of lung cancer (low risk, medium risk, and high risk) by packyears (15.64) (15.62)
our model, we can see the fraction of event free was statistically sig­ 0 15,890 15,816 344
nificant lower for low-risk groups compared with medium-risk and high- (5.63) (5.62) (58.40)
risk groups (Fig. 2). 1–29 7461 7419 129
(2.64) (2.63) (21.90)
Using this model, the C-index was 0.766, 0.739, and 0.745 for 1-year, 30–49 6 (2–12) 6 (2–12) 74
3-year, and 5-year lung cancer risk in the training set, respectively. (12.56)
≥50 177,558 177,206 42 (7.13)
(62.91) (62.91)
Smoking cessation 104,696 104,459 4.5
years, median (37.09) (37.09) (2–10.5)
(IQR)
2.50 0.114
Passive smoking
No 352
(59.76)
Yes 237
(40.24)
Alcohol Drinking
25.33 <0.001

(continued on next page)

29

L.-W. Guo et al. Lung Cancer 163 (2022) 27–34

Table 1 (continued ) year, 3-year, and 5-year lung cancer risk in the validation set, respec­
tively (Figure S2).
Total no. No lung Lung χ2 P-
(%) cancer, n cancer, n value The model showed good calibration across deciles of predicted risk in
(%) (%) training set, validation set, and all sub-populations (Fig. 4, Figure S3).

Never 213,209 212,814 395 12.23 <0.001 4. Discussions
(75.54) (75.56) (67.06) 6.83 0.009
Current 58,597 58,439 158 In this study, using data from a large prospective lung cancer
(20.76) (20.75) (26.83) screening cohort study, we developed and internally validated a simple
Former 10,448 10,412 36 (6.11) risk prediction model for lung cancer. Our results showed that the model
(3.70) (3.70) has moderate discriminatory accuracy and goodness-of-fit for both men
Physical activity 272 and women, smokers and never-smokers. Because all the indicators
<3 times/week 150,611 150,339 (46.18) included in this model can be acquired easily from general screening
(53.36) (53.38) 317 sets, it may be useful to identify high-risk individuals for more intensive
≥3 times/week 131,643 131,326 (53.82) screening to better prevent lung cancer.
(46.64) (46.62)
Family history of 523 It is well accepted that an effective lung cancer screening program
lung cancer 258,978 258,455 (88.79) necessitates risk stratification based on well-established risk factors
(first-degree (91.75) (91.76) 66 [39,48]. Accurate selection of high-risk individuals is essential to
relative) 23,276 23,210 (11.21) minimize harms associated with screening, including false-positive
No (8.25) (8.24) findings and unnecessary invasive diagnostic procedures. The evi­
27.85 <0.001 dence base of the included predictor variables is one of the important
Yes 233,139 232,701 criteria to measure the validity of risk prediction models. Several risk
(82.60) (82.62) 438 16.38 <0.001 factors have been associated with and used as predictors of lung cancer
History of chronic 49,115 48,964 (74.36) 12.79 <0.001 risk. The most dominant risk factors for lung cancer are smoking and
respiratory (17.40) (17.38) 151 19.04 <0.001 age. Sex, race/ethnicity, family history of lung cancer, chronic
disease (25.64) obstructive pulmonary disease (COPD), emphysema, and exposure to
No 277,951 277,383 6.37 0.012 asbestos and radon are some additional risk factors associated with lung
(98.48) (98.48) 568 cancer. To be specific, it has been proven that smoking is causally
Yes 4303 4282 (96.43) associated with the risk of lung cancer since the 1950s [49]. Addition­
(1.52) (1.52) 21 (3.57) ally, history of tuberculosis and history of emphysema was shown to be
History of related to elevated lung cancer risk [50,51].
tuberculosis 243,793 243,314 479
No (86.37) (86.38) (81.32) In addition to reliable predictor variables, risk prediction models
38,461 38,351 110 should meet predictive efficacy criteria, which are defined primarily as
Yes (13.63) (13.62) (18.68) the ability to distinguish lung cancer cases from healthy populations and
as the ability to calibrate consistency between the observed and pre­
History of chronic 278,213 277,645 568 dicted risk of lung cancer development. There have been several lung
bronchitis (98.57) (98.57) (96.43) cancer prediction models for the general population developed in
No 4041 4020 21 (3.57) different populations [52]. For study design, multiple case-control
(1.43) (1.43) studies (e.g., Liverpool Lung Project [LLP] model), and cohorts or ran­
Yes 555 domized trials (e.g., Prostate, Lung, Colorectal, and Ovarian Cancer
271,561 271,006 (94.23) Screening Trial [PLCO] m2014 model) [53] were used for the develop­
History of (96.21) (96.22) 34 (5.77) ment of lung cancer risk prediction model. In terms of the study popu­
emphysema 10,693 10,659 lation, never-smokers (e.g., EPIC model) [22], or overall population (e.
No (3.79) (3.78) g., LLPi model) [31] were included for developing risk models. To our
knowledge, this study is the first study to develop a lung cancer risk
Yes prediction model in a screening set in China. It is difficult to directly
compare the discriminatory performance of the models constructed in
History of asthma this study with previous models because each model was developed in a
bronchiectasis different type of population with different baseline risks or length of
No follow-up. However, the discriminatory power of each model was rela­
tively similar, with C-statistics ranging from 0.72 to 0.86. In comparison
Yes with previous studies, the models constructed in this study showed
comparable predictive performance.
a. Low, primary school or below; Medium, junior or senior high school; High,
undergraduate or over. Our study had a few limitations. First, all the relevant data were self-
Abbreviations: BMI, body mass index; IQR, Interquartile range. reported and subject to measurement error. However, given the pro­
spective design, any error in exposure assessment would have likely
Stratified analysis by gender showed that the C-statistic of the model was been attenuated. Second, detailed information on lung function,
higher among men (1-year: 0.758, 3-year: 0.734, and 5-year: 0.759) asbestos exposure, history of pneumonia, history of COPD, cooking
than women (1-year: 0.701, 3-year: 0.655, and 5-year: 0.624). Stratified index threshold and lack of ventilator while cooking, and biomarkers
analysis by age showed that the C-statistic of the model was higher including single nucleotide polymorphisms (SNPs) was not collected.
among younger participants (<60 years) (1-year: 0.763, 3-year: 0.696, Third, the performance of our risk prediction model was not validated
and 5-year: 0.703) than elder participants (≥60 years) (1-year: 0.657, 3- on an external dataset. However, the results of the internal validation
year: 0.666, and 5-year: 0.675). When examined by smoking status, the suggest promisingly that this model will obtain well performance when
model yielded higher C-statistic for current smoking participants (1- applied to other populations. Furthermore, the competing risks for death
year: 0.791, 3-year: 0.735, and 5-year: 0.775) or never smoking par­ were not corrected in the present model, which may lead to potential
ticipants (1-year: 0.740, 3-year: 0.714, and 5-year: 0.697) than former bias in terms of the predictive accuracy of the models.
smoking participants (1-year: 0.591, 3-year: 0.668, and 5-year: 0.645).
We additionally evaluated the performance of model for ever smokers
(current smokers and former smokers), and the C-index was 0.750,
0.721, and 0.751 for 1-year, 3-year, and 5-year lung cancer risk in the
training set, respectively (Fig. 3). As for the performance of our model in
the validation dataset, the C-index was 0.741, 0.734, and 0.735 for 1-

30

L.-W. Guo et al. Lung Cancer 163 (2022) 27–34

Table 2
Multivariable Cox-regression prediction model for lung cancer risk in the training set and validation set.

Variables Training set Validation set

β HR (95% CI) P β HR (95% CI) P

Age (years) − 0.02 1.00 0.961 1.12 1.00 0.043
40–44 0.88 0.98 (0.42–2.26) 0.019 1.55 3.05 (1.04–8.97) 0.004
45–49 1.58 2.40 (1.15–5.00) <0.001 2.25 4.73 (1.67–13.43) <0.001
50–54 1.71 4.87 (2.41–9.84) <0.001 2.57 9.53 (3.44–26.38) <0.001
55–59 2.02 5.51 (2.75–11.04) <0.001 2.97 13.07 (4.78–35.75) <0.001
60–64 2.48 7.57 (3.79–15.16) <0.001 3.13 19.53 (7.16–53.28) <0.001
65–69 11.94 (5.71–24.96) 22.83 (8.02–65.00) <0.001
70–74 0.54 <0.001 0.54
1.72 (1.28–2.31) 1.71 (1.28–2.29) 0.221
Gender 0.32 1.00 0.055 0.21 1.00 0.004
Male 0.49 0.016 0.58 0.055
Female 0.67 1.00 0.005 0.51 1.00
1.38 (0.99–1.91) 1.23 (0.88–1.72) 0.174
Smoking packyears 0.63 1.64 (1.10–2.45) 0.039 0.47 1.78 (1.20–2.63)
0 1.95 (1.22–3.13) 1.67 (0.99–2.81) 0.665
1–29 0.74 0.008 − 0.18
30–49 1.00 1.00
≥50 1.88 (1.03–3.41) 1.60 (0.81–3.16)

History of tuberculosis 1.00 1.00
No 2.09 (1.21–3.61) 0.83 (0.36–1.91)
Yes
History of emphysema

No
Yes

Fig. 1. Nomogram to calculate the personal 1-, 3- and 5-year risk of lung cancer.

Despite the limitations, our study suggests the great potential of easily assessed in a screening setting and thus can be easily adapted and
identifying high-risk individuals for better detection of lung cancer using automated in the electronic health record system. In addition, our
a simple risk prediction tool. It has been reported that most primary care findings have significant health service implications, given the growing
providers are willing to use such a risk prediction model, if it adds demand and limited capacity of LDCT in China and other countries.
minimal burden to routine clinical workflow, is easy to use, and requires Thus, tailoring healthcare resources to high-risk individuals may
minimal time to complete. The simple model developed in the current significantly improve the cost-effectiveness of population-based
study meets all these requirements by utilizing risk factors that can be screening.

31

L.-W. Guo et al. Lung Cancer 163 (2022) 27–34
Fig. 2. The lung cancer risk across different cancer risk categories.
5. Conclusions

In summary, we developed and internally validated a simple risk
prediction model for lung cancer in a prospective cohort study of the
Chinese population. The model consists of predictors that are readily
available or easily accessible in a general massive screening setting and
has shown adequate discriminatory accuracy. Thus, the model has po­
tential utility for shared decision-making and individualized risk
assessment for tailored lung cancer screening. Meanwhile, external
validation was not conducted in this study, so more studies focus on
evaluating the performance of the model in other populations should be
the future direction.

6. Funding/Support

The study was supported by the Chinese National Key Research and
Development Project (No. 2018YFC1315600), the Natural Science
Foundation of Henan Province (No. 212300410261) and the Henan
Province Medical Science and Technology Tackling Program (No.
SBGJ202001004).

Fig. 3. The receiver operating characteristic curves of prediction models in the training set among the (a) whole population; (b) male; (c) female; (d) age < 60 years;
(e) age ≥ 60 years; (f) never-smokers; (g) current-smokers; (h) former-smokers.

32

L.-W. Guo et al. Lung Cancer 163 (2022) 27–34

9. Data availability statement

The datasets for this manuscript are not publicly available because
all our data are under the regulation of both the National Cancer Center
of China and Henan Cancer Hospital. Requests to access the datasets
should be directed to Shao-Kai Zhang ([email protected]).

CRediT authorship contribution statement

Lan-Wei Guo: Conceptualization, Methodology, Software, Formal
analysis, Investigation, Writing – original draft, Funding acquisition.
Zhang-Yan Lyu: Methodology, Software, Validation, Writing – original
draft. Qing-Cheng Meng: Investigation, Writing – review & editing. Li-
Yang Zheng: Investigation, Resources. Qiong Chen: Software, Investi­
gation. Yin Liu: Investigation. Hui-Fang Xu: Investigation. Rui-Hua
Kang: Investigation. Lu-Yao Zhang: Investigation. Xiao-Qin Cao:
Investigation. Shu-Zheng Liu: Methodology, Supervision. Xi-Bin Sun:
Methodology, Supervision. Jian-Gong Zhang: Supervision, Project
administration, Funding acquisition. Shao-Kai Zhang: Resources, Data
curation, Writing – review & editing, Supervision.

Declaration of Competing Interest

The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influence
the work reported in this paper.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.
org/10.1016/j.lungcan.2021.11.015.

Fig. 4. Calibration curves of the prediction model for (A) 1-year, (B) 3-year and References
(C) 5-year lung cancer risk in the training set.
[1] J. Ferlay, M. Colombet, I. Soerjomataram, C. Mathers, D.M. Parkin, M. Pin˜eros,
7. Role of the Funder/Sponsor A. Znaor, F. Bray, Estimating the global cancer incidence and mortality in 2018:
GLOBOCAN sources and methods, Int. J. Cancer 144 (8) (2019) 1941–1953.
The funders had no role in the design and conduct of the study;
collection, management, analysis, and interpretation of the data; prep­ [2] F.C. Detterbeck, D.J. Boffa, A.W. Kim, L.T. Tanoue, The eighth edition lung cancer
aration, review, or approval of the manuscript; and decision to submit stage classification, Chest 151 (1) (2017) 193–203.
the manuscript for publication.
[3] H. Zeng, R. Zheng, Y. Guo, S. Zhang, X. Zou, N. Wang, L. Zhang, J. Tang, J. Chen,
8. Ethics approval and consent to participate K. Wei, S. Huang, J. Wang, L. Yu, D. Zhao, G. Song, J. Chen, Y. Shen, X. Yang,
X. Gu, F. Jin, Q. Li, Y. Li, H. Ge, F. Zhu, J. Dong, G. Guo, M. Wu, L. Du, X. Sun,
The study was conducted following the guidelines of the Helsinki Y. He, M.P. Coleman, P. Baade, W. Chen, X.Q. Yu, Cancer survival in China,
Declaration and was approved by the Medical Ethics Committee of the 2003–2005: a population-based study, Int. J. Cancer 136 (8) (2015) 1921–1930.
Henan Cancer Hospital. Written informed consent forms were obtained
from all participants. [4] D.R. Aberle, A.M. Adams, C.D. Berg, W.C. Black, J.D. Clapp, R.M. Fagerstrom, I.
F. Gareen, C. Gatsonis, P.M. Marcus, J.D. Sicks, Reduced lung-cancer mortality
with low-dose computed tomographic screening, N. Engl. J. Med. 365 (5) (2011)
395–409.

[5] H.J. de Koning, C.M. van der Aalst, P.A. de Jong, E.T. Scholten, K. Nackaerts, M.
A. Heuvelmans, J.-W. Lammers, C. Weenink, U. Yousaf-Khan, N. Horeweg, S. van ’t
Westeinde, M. Prokop, W.P. Mali, F.A.A. Mohamed Hoesein, P.M.A. van Ooijen, J.
G.J.V. Aerts, M.A. den Bakker, E. Thunnissen, J. Verschakelen, R. Vliegenthart, J.
E. Walter, K. ten Haaf, H.J.M. Groen, M. Oudkerk, Reduced Lung-Cancer Mortality
with Volume CT Screening in a Randomized Trial, N. Engl. J. Med. 382 (6) (2020)
503–513.

[6] D.R. Aberle, S. DeMello, C.D. Berg, W.C. Black, B. Brewer, T.R. Church, K.
L. Clingan, F. Duan, R.M. Fagerstrom, I.F. Gareen, C.A. Gatsonis, D.S. Gierada,
A. Jain, G.C. Jones, I. Mahon, P.M. Marcus, J.M. Rathmell, J. Sicks, Results of the
two incidence screenings in the National Lung Screening Trial, N. Engl. J. Med. 369
(10) (2013) 920–931.

[7] P.F. Pinsky, C.D. Berg, Applying the National Lung Screening Trial eligibility
criteria to the US population: what percent of the population and of incident lung
cancers would be covered? J. Med. Screen. 19 (3) (2012) 154–156.

[8] A.H. Krist, K.W. Davidson, C.M. Mangione, M.J. Barry, M. Cabana, A.B. Caughey,
E.M. Davis, K.E. Donahue, C.A. Doubeni, M. Kubik, C.S. Landefeld, L. Li,
G. Ogedegbe, D.K. Owens, L. Pbert, M. Silverstein, J. Stevermer, C.-W. Tseng, J.
B. Wong, Screening for Lung Cancer: US Preventive Services Task Force
Recommendation Statement, JAMA 325 (10) (2021) 962, https://doi.org/
10.1001/jama.2021.1117.

[9] S.S. Han, E. Chow, K. Ten Haaf, I. Toumazis, P. Cao, M. Bastani, M. Tammemagi,
J. Jeon, E.J. Feuer, R. Meza, S.K. Plevritis, Disparities of national lung cancer
screening guidelines in the US population, J. Natl. Cancer Inst. 112 (11) (2020)
1136–1142.

[10] M.C. Tammema¨gi, Selecting lung cancer screenees using risk prediction models-
where do we go from here, Transl. Lung Cancer Res. 7 (3) (2018) 243–253.

33

L.-W. Guo et al. Lung Cancer 163 (2022) 27–34

[11] P.B. Bach, M.W. Kattan, M.D. Thornquist, M.G. Kris, R.C. Tate, M.J. Barnett, L. [32] X. Wang, K. Ma, J. Cui, X. Chen, L. Jin, W. Li, An individual risk prediction model
J. Hsieh, C.B. Begg, Variations in lung cancer risk among smokers, J. Natl. Cancer for lung cancer based on a study in a Chinese population, Tumori. 101 (1) (2015)
Inst. 95 (2003) 470–478. 16–23.

[12] M.R. Spitz, W.K. Hong, C.I. Amos, X. Wu, M.B. Schabath, Q. Dong, S. Shete, C. [33] M.W. Marcus, O.Y. Raji, S.W. Duffy, R.P. Young, R.J. Hopkins, J.K. Field,
J. Etzel, A risk model for prediction of lung cancer, J. Natl Cancer Inst. 99 (9) Incorporating epistasis interaction of genetic susceptibility single nucleotide
(2007) 715–726. polymorphisms in a lung cancer risk prediction model, Int. J. Oncol. 49 (1) (2016)
361–370.
[13] A. Cassidy, J.P. Myles, M. van Tongeren, R.D. Page, T. Liloglou, S.W. Duffy, J.
K. Field, The LLP risk model: an individual risk prediction model for lung cancer, [34] X. Wang, K. Ma, L. Chi, J. Cui, L. Jin, J.-F. Hu, W. Li, Combining telomerase reverse
Br. J. Cancer 98 (2) (2008) 270–276. transcriptase genetic variant rs2736100 with epidemiologic factors in the
prediction of lung cancer susceptibility, J. Cancer 7 (7) (2016) 846–853.
[14] C.J. Etzel, S. Kachroo, M. Liu, A. D’Amelio, Q. Dong, M.L. Cote, A.S. Wenzlaff, W.
K. Hong, A.J. Greisinger, A.G. Schwartz, M.R. Spitz, Development and validation of [35] X. Wu, C.P. Wen, Y. Ye, M. Tsai, C. Wen, J.A. Roth, X. Pu, W.H. Chow, C. Huff,
a lung cancer risk prediction model for African-Americans, Cancer Prev. Res. S. Cunningham, M. Huang, S. Wu, C.K. Tsao, J. Gu, S.M. Lippman, Personalized
(Phila) 1 (4) (2008) 255–265. Risk Assessment in Never, Light, and Heavy Smokers in a prospective cohort in
Taiwan, Sci. Rep. 6 (2016) 36482.
[15] M.R. Spitz, C.J. Etzel, Q. Dong, C.I. Amos, Q. Wei, X. Wu, W.K. Hong, An expanded
risk prediction model for lung cancer, Cancer Prev. Res. (Phila) 1 (4) (2008) [36] D.C. Muller, M. Johansson, P. Brennan, Lung cancer risk prediction model
250–254. incorporating Lung function: development and validation in the UK biobank
prospective cohort study, J. Clin. Oncol. 35 (8) (2017) 861–869.
[16] R.P. Young, R.J. Hopkins, B.A. Hay, M.J. Epton, G.D. Mills, P.N. Black, H.
D. Gardner, R. Sullivan, G.D. Gamble, I. Schrijver, Lung cancer susceptibility model [37] M. Weber, S. Yap, D. Goldsbury, D. Manners, M. Tammemagi, H. Marshall,
based on age, family history and genetic variants, PLoS ONE 4 (4) (2009) e5302. F. Brims, A. McWilliams, K. Fong, Y.J. Kang, M. Caruana, E. Banks, K. Canfell,
Identifying high risk individuals for targeted lung cancer screening: Independent
[17] A.M. D’Amelio, A. Cassidy, K. Asomaning, O.Y. Raji, S.W. Duffy, J.K. Field, M. validation of the PLCOm2012 risk prediction tool, Int. J. Cancer 141 (2) (2017)
R. Spitz, D. Christiani, C.J. Etzel, Comparison of discriminatory power and 242–253.
accuracy of three lung cancer risk models, Br. J. Cancer 103 (3) (2010) 423–429.
[38] H. Charvat, S. Sasazuki, T. Shimazu, S. Budhathoki, M. Inoue, M. Iwasaki,
[18] O.Y. Raji, O.F. Agbaje, S.W. Duffy, A. Cassidy, J.K. Field, Incorporation of a genetic N. Sawada, T. Yamaji, S. Tsugane, Development of a risk prediction model for lung
factor into an epidemiologic model for prediction of individual risk of lung cancer: cancer: The Japan Public Health Center-based Prospective Study, Cancer Sci. 109
the Liverpool Lung Project, Cancer Prev. Res. (Phila) 3 (5) (2010) 664–669. (3) (2018) 854–862.

[19] P. Maisonneuve, V. Bagnardi, M. Bellomi, L. Spaggiari, G. Pelosi, C. Rampinelli, [39] H.A. Katki, S.A. Kovalchik, L.C. Petito, L.C. Cheung, E. Jacobs, A. Jemal, C.D. Berg,
R. Bertolotti, N. Rotmensz, J.K. Field, A. DeCensi, G. Veronesi, Lung cancer risk A.K. Chaturvedi, Implications of nine risk prediction models for selecting ever-
prediction to select smokers for screening CT–a model based on the Italian smokers for computed tomography lung cancer screening, Ann. Intern. Med. 169
COSMOS trial, Cancer Prev. Res. (Phila) 4 (11) (2011) 1778–1789. (1) (2018) 10, https://doi.org/10.7326/M17-2701.

[20] C.M. Tammemagi, P.F. Pinsky, N.E. Caporaso, P.A. Kvale, W.G. Hocking, T. [40] M. Markaki, I. Tsamardinos, A. Langhammer, V. Lagani, K. Hveem, O.D. Røe,
R. Church, T.L. Riley, J. Commins, M.M. Oken, C.D. Berg, P.C. Prorok, Lung cancer A validated clinical risk prediction model for lung cancer in smokers of all ages and
risk prediction: Prostate, Lung, Colorectal And Ovarian Cancer Screening Trial exposure types: a HUNT Study, EBioMedicine 31 (2018) 36–46.
models and validation, J. Natl. Cancer Inst. 103 (13) (2011) 1058–1068.
[41] S. Li, L. Meng, A. Chiolero, C. Ma, B. Xi, Trends in smoking prevalence and
[21] M.C. Tammemagi, S.C. Lam, A.M. McWilliams, D.D. Sin, Incremental value of attributable mortality in China, 1991–2011, Prev. Med. 93 (2016) 82–87.
pulmonary function and sputum DNA image cytometry in lung cancer risk
prediction, Cancer Prev. Res. (Phila) 4 (4) (2011) 552–561. [42] Centers for Disease Control and Prevention, State Tobacco Activities Tracking &
Evaluation (STATE) System. https://www.cdc.gov/tobacco/data_statistics/fact_sh
[22] C. Hoggart, P. Brennan, A. Tjonneland, U. Vogel, K. Overvad, J.N. Østergaard, eets/adult_data/cig_smoking/, 2018 (accessed 17 November 2021).
R. Kaaks, F. Canzian, H. Boeing, A. Steffen, A. Trichopoulou, C. Bamia,
D. Trichopoulos, M. Johansson, D. Palli, V. Krogh, R. Tumino, C. Sacerdote, [43] J. Zeng, S. Yang, L. Wu, J. Wang, Y. Wang, M. Liu, D. Zhang, B. Jiang, Y. He,
S. Panico, H. Boshuizen, H.B. Bueno-de-Mesquita, P.H.M. Peeters, E. Lund, I. Prevalence of passive smoking in the community population aged 15 years and
T. Gram, T. Braaten, L. Rodríguez, A. Agudo, E. Sa´nchez-Cantalejo, L. Arriola, M.- older in China: a systematic review and meta-analysis, BMJ Open 6 (4) (2016),
D. Chirlaque, A. Barricarte, T. Rasmuson, K.-T. Khaw, N. Wareham, N.E. Allen, e009847.
E. Riboli, P. Vineis, A risk model for lung cancer incidence, Cancer Prev. Res.
(Phila) 5 (6) (2012) 834–846. [44] Y. Guo, X. Yin, H. Wu, X. Chai, X. Yang, Trends in Overweight and Obesity Among
Children and Adolescents in China from 1991 to 2015: A Meta-Analysis, Int. J.
[23] H. Li, L. Yang, X. Zhao, J. Wang, J.i. Qian, H. Chen, W. Fan, H. Liu, L.i. Jin, Environ. Res. Public Health 16 (23) (2019) 4656.
W. Wang, D. Lu, Prediction of lung cancer risk in a Chinese population using a
multifactorial genetic model, BMC Med. Genet. 13 (1) (2012), https://doi.org/ [45] Centers for Disease Control and Prevention, The Behavioral Risk Factor
10.1186/1471-2350-13-118. Surveillance System (BRFSS). https://www.cdc.gov/niosh/topics/surveillance
/brfss/default.html, 2021 (accessed 17 November 2021).
[24] O.Y. Raji, S.W. Duffy, O.F. Agbaje, S.G. Baker, D.C. Christiani, A. Cassidy, J.
K. Field, Predictive accuracy of the Liverpool Lung Project risk model for [46] World Health Organization (WHO), Global Health Observatory (GHO). Indicator
stratifying patients for computed tomography screening for lung cancer: a case- 7.1.2 Proportion of population with primary reliance on clean fuels and
control and cohort validation study, Ann. Intern. Med. 157 (4) (2012) 242, https:// technologies, https://www.who.int/data/gho/data/themes/air-pollution/househ
doi.org/10.7326/0003-4819-157-4-201208210-00004. old-air-pollution, 2021 (accessed 17 November 2021).

[25] S. Park, B.-H. Nam, H.-R. Yang, J.A. Lee, H. Lim, J.T. Han, I.S. Park, H.-R. Shin, J. [47] L.-W. Guo, Q. Chen, Y.-C. Shen, Q.-C. Meng, L.-Y. Zheng, Y. Wu, X.-Q. Cao, H.-
S. Lee, O.Y. Gorlova, Individualized risk prediction model for lung cancer in Korean F. Xu, S.-Z. Liu, X.-B. Sun, Y.-L. Qiao, S.-K. Zhang, Evaluation of a Low-Dose
men, PLoS ONE 8 (2) (2013) e54823, https://doi.org/10.1371/journal. Computed Tomography Lung Cancer Screening Program in Henan, China. JAMA
pone.0054823. Netw Open 3 (11) (2020) e2019039, https://doi.org/10.1001/
jamanetworkopen.2020.19039.
[26] M.R. Spitz, C.I. Amos, S. Land, X. Wu, Q. Dong, A.S. Wenzlaff, A.G. Schwartz, Role
of selected genetic variants in lung cancer risk in African Americans, J. Thorac. [48] M. Oudkerk, A. Devaraj, R. Vliegenthart, T. Henzler, H. Prosch, C.P. Heussel,
Oncol. 8 (4) (2013) 391–397. G. Bastarrika, N. Sverzellati, M. Mascalchi, S. Delorme, D.R. Baldwin, M.
E. Callister, N. Becker, M.A. Heuvelmans, W. Rzyman, M.V. Infante, U. Pastorino, J.
[27] M.C. Tammemagi, H.A. Katki, W.G. Hocking, T.R. Church, N. Caporaso, P.A. Kvale, H. Pedersen, E. Paci, S.W. Duffy, H. de Koning, J.K. Field, European position
A.K. Chaturvedi, G.A. Silvestri, T.L. Riley, J. Commins, C.D. Berg, Selection criteria statement on lung cancer screening, Lancet Oncol. 18 (12) (2017) e754–e766.
for lung-cancer screening, N. Engl. J. Med. 368 (8) (2013) 728–736.
[49] IARC Working Group on the Evaluation of Carcinogenic Risks to Humans, Tobacco
[28] G. Veronesi, P. Maisonneuve, C. Rampinelli, R. Bertolotti, F. Petrella, L. Spaggiari, smoke and involuntary smoking, IARC Monogr. Eval. Carcinog. Risks Hum. 83
M. Bellomi, Computed tomography screening for lung cancer: results of ten years of (2004) 1–1438.
annual screening and validation of cosmos prediction model, Lung Cancer 82 (3)
(2013) 426–430. [50] H.-Y. Liang, X.-L. Li, X.-S. Yu, P. Guan, Z.-H. Yin, Q.-C. He, B.-S. Zhou, Facts and
fiction of the relationship between preexisting tuberculosis and lung cancer risk: a
[29] R.A. El-Zein, M.S. Lopez, A.M. D’Amelio, M. Liu, R.F. Munden, D. Christiani, L. Su, systematic review, Int. J. Cancer 125 (12) (2009) 2936–2944.
P. Tejera-Alveraz, R. Zhai, M.R. Spitz, C.J. Etzel, The cytokinesis-blocked
micronucleus assay as a strong predictor of lung cancer: extension of a lung cancer [51] Ramón.A. Tubió -Pérez, M. Torres-Durán, M. Pérez-Rió s, A. Fernández-Villar,
risk prediction model, Cancer Epidemiol Biomarkers Prev 23 (11) (2014) A. Ruano-Raviña, Lung emphysema and lung cancer: what do we know about it?
2462–2470. Ann. Transl. Med. 8 (21) (2020), 1471.

[30] K. Li, A. Hüsing, D. Sookthai, M. Bergmann, H. Boeing, N. Becker, R. Kaaks, [52] E.P. Gray, M.D. Teare, J. Stevens, R. Archer, Risk Prediction Models for Lung
Selecting high-risk individuals for lung cancer screening: a prospective evaluation Cancer: A Systematic Review, Clin Lung Cancer 17 (2) (2016) 95–106.
of existing risk models and eligibility criteria in the german EPIC cohort, Cancer
Prev. Res. 8 (9) (2015) 777–785. [53] M.C. Tammema¨gi, T.R. Church, W.G. Hocking, G.A. Silvestri, P.A. Kvale, T.L. Riley,
J. Commins, C.D. Berg, M. Massad, Evaluation of the lung cancer risks at which to
[31] M.W. Marcus, Y. Chen, O.Y. Raji, S.W. Duffy, J.K. Field, LLPi: Liverpool Lung screen ever- and never-smokers: screening rules applied to the PLCO and NLST
Project Risk Prediction Model for Lung Cancer Incidence, Cancer Prev. Res. (Phila) cohorts, PLoS Med. 11 (12) (2014) e1001764.
8 (6) (2015) 570–575.

34

Lung Cancer 163 (2022) 77–86
Contents lists available at ScienceDirect

Lung Cancer

journal homepage: www.elsevier.com/locate/lungcan

Brain penetration and efficacy of tepotinib in orthotopic patient-derived
xenograft models of MET-driven non-small cell lung cancer
brain metastases

Manja Friese-Hamim a,1, Anderson Clark b,1, Dominique Perrin c, Lindsey Crowley b,
Christof Reusch a, Olga Bogatyrova a, Hong Zhang b, Timothy Crandall b, Jing Lin b, Jianguo Ma b,
David Bachner b, Jürgen Schmidt a, Martin Schaefer a, Christopher Stroh a,*

a Translational Innovation Platform, Oncology & Immuno-Oncology, Merck Healthcare KGaA, Darmstadt, Germany
b Translational Innovation Platform, Oncology & Immuno-Oncology, EMD Serono Research & Development Institute, Inc., Billerica, MA, USA, an affiliate of Merck
KGaA
c Discovery & Development Technologies, Merck Healthcare KGaA, Darmstadt, Germany

ARTICLE INFO ABSTRACT

Keywords: Central nervous system-penetrant therapies with intracranial efficacy against non-small cell lung cancer (NSCLC)
Brain metastasis brain metastases are urgently needed. We report preclinical studies investigating brain penetration and intra­
MET amplification cranial activity of the MET inhibitor tepotinib. After intravenous infusion of tepotinib in Wistar rats (n = 3),
Non-small cell lung cancer mean (±standard deviation) total tepotinib concentration was 2.87-fold higher in brain (505 ± 22 ng/g) than
Orthotopic implantation plasma (177 ± 20 ng/mL). In equilibrium dialysis experiments performed in triplicate, mean tepotinib unbound
Patient-derived xenograft fraction was 0.35% at 0.3 and 3.0 µM tepotinib in rat brain tissue, and 4.0% at 0.3 and 1.0 µM tepotinib in rat
Tepotinib plasma. The calculated unbound brain-to-plasma ratio was 0.25, indicating brain penetration sufficient for
intracranial target inhibition. Of 20 screened subcutaneous patient-derived xenograft (PDX) models from lung
cancer brain metastases (n = 1), two NSCLC brain metastases models (LU5349 and LU5406) were sensitive to the
suboptimal dose of tepotinib of 30 mg/kg/qd (tumor volume change [%TV]: –12% and –88%, respectively).
Molecular profiling (nCounter®; NanoString) revealed high-level MET amplification in both tumors (mean MET
gene copy number: 11.2 and 24.2, respectively). Tepotinib sensitivity was confirmed for both subcutaneous
models at a clinically relevant dose (125 mg/kg/qd; n = 5). LU5349 and LU5406 were orthotopically implanted
into brains of mice and monitored by magnetic resonance imaging (MRI). Tepotinib 125 mg/kg/qd induced
pronounced tumor regression, including complete or near-complete regressions, compared with vehicle in both
orthotopic models (n = 10; median %TV: LU5349, –84%; LU5406, –63%). Intracranial antitumor activity of
tepotinib did not appear to correlate with blood–brain barrier leakiness assessed in T1-weighted gadolinium
contrast-enhanced MRI.

Abbreviations: %TV, percent tumor volume change; ALK, anaplastic lymphoma kinase; BEH, ethylene bridged hybrid; BBB, blood–brain barrier; Cbrain, total
concentration of tepotinib in brain; Cplasma, total concentration of tepotinib in plasma; CNS, central nervous system; DPBS, Dulbecco’s PBS; EGFR, epidermal growth
factor receptor; fu, brain, unbound fraction of tepotinib in plasma; fu, plasma, unbound fraction of tepotinib in plasma; GCN, gene copy number; Kp, total brain-to-
plasma ratio; Kp, uu, unbound brain-to-plasma ratio; MET, mesenchymal-epithelial transition factor; METex14, MET exon 14; MRI, magnetic resonance imaging;
NSCLC, non-small cell lung cancer; PDX, patient-derived xenograft; po, orally; qd, once daily; ROI, region of interest; SD, standard deviation; SEM, standard error of
the mean; T1CE, T1-weighted gadolinium contrast-enhanced; T2w, T2-weighted; UPLC-MS/MS, ultra-performance liquid chromatography–tandem mass
spectrometry.

* Corresponding author at: Merck Healthcare KGaA, Frankfurter Str. 250, F128/103, 64293 Darmstadt, Germany.
E-mail address: [email protected] (C. Stroh).

1 Manja Friese-Hamim and Anderson Clark contributed equally to this work.

https://doi.org/10.1016/j.lungcan.2021.11.020

Received 20 October 2021; Received in revised form 25 November 2021; Accepted 28 November 2021

Available online 11 December 2021
0169-5002/© 2021 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license

(http://creativecommons.org/licenses/by-nc-nd/4.0/).

M. Friese-Hamim et al. Lung Cancer 163 (2022) 77–86

1. Introduction reproduce the anatomic situation of the original tumor [30]. Here, we
report preclinical experiments conducted in rodents to evaluate tepoti­
The treatment landscape for non-small cell lung cancer (NSCLC) is nib brain penetration and activity against subcutaneous or orthotopic
increasingly based on the identification of molecular driver alterations PDXs derived from MET-driven NSCLC brain metastases.
that can be targeted therapeutically [1,2]. The spectrum of targetable
alterations now includes MET exon 14 (METex14) skipping mutations 2. Materials and methods
[1,2], which occur in 3–4% of patients [3]. In these tumors, dysregu­
lation of the proto-oncogene MET promotes invasive growth and 2.1. Ethics
metastasis [4], but also confers sensitivity to MET inhibitors [5]. A
further 1–5% of NSCLCs harbor MET amplification, which is an Animal experiments in the US were conducted in accordance with
emerging biomarker of response to MET inhibition [6]. Institutional Animal Care and Use Committee guidelines, while those
performed in Germany complied with Directive 2010/63/EU of the
The potent and highly selective MET inhibitor tepotinib has recently European Parliament and of the Council, as well as with the German
been approved for treatment of metastatic METex14 skipping NSCLC Animal Welfare Act (Tierschutzgesetz) and Laboratory Animal Welfare
[7,8]. In the phase 2 VISION trial, once daily oral tepotinib 500 mg (450 Regulation (Tierschutz-Versuchstierverordnung). All animal experi­
mg active moiety) provided durable clinical responses in patients with ments received institutional review board approval (protocol numbers:
advanced NSCLC harboring METex14 skipping alterations [9,10]. The rat brain penetration study, 55.2-1-54-2532.2-1-06; subcutaneous PDX
study additionally showed antitumor activity of tepotinib in a separate screen, 20-004; confirmatory subcutaneous and orthotopic PDX studies,
cohort of patients with NSCLC with MET amplification as a primary EB17-030).
driver [11]. Tepotinib is also active in combination with gefitinib in
patients with epidermal growth factor receptor (EGFR)-mutant 2.2. Tepotinib brain penetration
advanced NSCLC and MET-driven resistance to anti-EGFR therapy, as
demonstrated in the phase 1b/2 INSIGHT trial [12]. Brain penetration was evaluated at Merck Serono Institute of Drug
Metabolism and Pharmacokinetics (Merck Healthcare KGaA, Grafing,
Up to 40% of patients with NSCLC, overall, develop brain metastases Germany) in male Wistar rats (Charles River Laboratories, Inc., Wil­
during the course of the disease [13], and limited available data suggest mington, MA, USA). Rats were housed in polycarbonate cages type M III
a similarly high incidence in patients with METex14 skipping [14,15]. with a wire netting level and had free access to commercial pellet diet
MET dysregulation has been implicated in the biology of metastasis to (Standard Diet SDS RM1 (E); Special Diets Services, Witham, UK) and
the central nervous system (CNS) [16], and MET amplification and/or water from the public supply. Tepotinib was administered to three an­
overexpression are frequently observed in NSCLC brain lesions [17]. imals by intravenous infusion at a rate of 3.66 mg/kg/hour. After 24 h,
Patients with brain metastases have an especially poor prognosis, with which corresponds to > 5 times the half-life of tepotinib in rats (3.2 h
short duration of overall survival [18,19]. Furthermore, brain metasta­ [31]) and therefore steady-state, rats were bled under anesthesia and the
ses cause substantial morbidity and are often accompanied by disabling brains were removed. The total concentration of tepotinib in plasma
neurocognitive symptoms that can profoundly diminish quality of life (Cplasma) and brain (Cbrain) was measured by ultra-performance liquid
[18,19]. Brain metastases have traditionally been treated with local chromatography–tandem mass spectrometry (UPLC-MS/MS). To pre­
therapies, including stereotactic radiosurgery, whole-brain radio­ pare samples, tepotinib was extracted from plasma using tert-butyl
therapy, and/or surgery, which have neurotoxic effects [20]. Chemo­ methylether and from brain using ethanol/water (80:20 [v/v]), and a
therapy is not commonly used due to the low blood–brain barrier (BBB) deuterated internal standard (2H-tepotinib) was added. The UPLC-MS/
penetration of most conventional cytotoxic agents [18]. Poor CNS MS system comprised an Acquity binary pump and autosampler (Wa­
permeability could be less problematic for immunotherapies, which may ters, Milford, MA, USA) connected to an ethylene bridged hybrid (BEH)
act by modulating immune cells in the periphery [21], but evidence for C18 column (2.1 × 50 mm; 1.7 μm; Waters). Separation was achieved by
these agents in patients with NSCLC brain metastases is currently limited gradient elution (mobile phase A: acetonitrile; mobile phase B: ammo­
[22]. nium bicarbonate 10 mM at pH 10) with a flow rate of 0.75–1.00 mL/
min over 3.5 min. Detection utilized a triple quadrupole API 4000 de­
For patients with tumors harboring oncogenic driver alterations, the tector (ThermoFisher Scientific, Waltham, MA, USA) with multiple re­
development of CNS-penetrant targeted therapies with high intracranial action monitoring (m/z: 493.3–112 for tepotinib; 496.3–115.1 for 2H-
activity has enabled a new, radiotherapy-sparing approach for the tepotinib). Total brain-to-plasma ratio (Kp) was calculated as Cbrain/
management of brain lesions [21]. However, the brain penetration of Cplasma.
different targeted agents is highly variable, with some having poor brain
exposure due to low intrinsic permeability of the drug to the BBB, and/ 2.3. Tepotinib plasma protein and brain binding
or active transport out of the CNS by BBB efflux transporters, such as P-
glycoprotein [23]. As patients with brain metastases are frequently Tepotinib was added to aliquots of rat plasma (strain: HsdCpb:WU;
excluded from clinical trials, information regarding the activity of tar­ Harlan Winkelmann GmbH, Borchen, Germany) at final concentrations
geted agents in the CNS is often lacking [24], and there is a high unmet of 0.3 and 1.0 μM (in triplicate). Samples were dialyzed against phos­
need for CNS-penetrant agents with demonstrated activity against brain phate buffer (70 mM; pH 7.4) in an equilibrium dialysis device con­
lesions [18]. taining a 10-kDa molecular weight cut-off Diachema membrane
(Dianorm GmbH, Munich, Germany). After 1 h, 2H-tepotinib was added
Clinical evidence supporting the intracranial activity of tepotinib in to 20 μL aliquots of the plasma and buffer chambers, and samples were
patients with METex14 skipping NSCLC has recently been provided by extracted in tert-butyl methylether. Samples were separated on an
several case reports [25–28], and an ad hoc analysis of intracranial re­ Acquity UPLC instrument equipped with a BEH phenyl column (2.1 ×
sponses in VISION patients who had brain metastases at baseline [10]. 50 mm; 1.7 μm; Waters) by gradient elution (mobile phase A: acetoni­
To complement such clinical data, preclinical experiments can provide trile; mobile phase B: 0.1% formic acid) at a 0.75–1.00 mL/min flow rate
important information regarding the potential for activity against brain over 4 min. Detection was as previously described for the brain pene­
lesions [24]. For example, the ability of an agent to cross the intact BBB tration study. The unbound fraction of tepotinib in rat plasma (fu, plasma)
in humans can be predicted based on rodent biodistribution and brain was calculated by dividing the concentration of tepotinib in the buffer
binding data, with the unbound brain-to-plasma ratio (Kp,uu) providing chamber by that in the plasma chamber and correcting for the increase
the most useful measure of CNS penetration [23,29]. Furthermore,
antitumor activity against brain metastases can be assessed in patient-
derived xenograft (PDX) models [24], which may be especially rele­
vant when implanted orthotopically into the brain to more closely

78

M. Friese-Hamim et al. Lung Cancer 163 (2022) 77–86

in plasma chamber volume due to the Donnan effect. nSolver v4.0 software according to the manufacturer’s instructions
Rat brain tissue obtained from BioIVT (Westbury, NY, USA) was (NanoString Technologies, Inc., Seattle, WA, USA).

homogenized in Dulbecco’s PBS (DPBS; 1 g brain tissue per 2 mL DPBS). Gene expression was analyzed in 50 ng input RNA using the Pan­
Tepotinib was spiked into the homogenate at final concentrations of 0.3 Cancer Pathways Panel (NanoString), which contains 770 genes from 13
and 3.0 μM (in triplicate) and equilibrated at 37 ◦C for 10 min. A total of canonical cancer-associated pathways. Raw counts were corrected by
300 μL of the equilibrated homogenate and 500 μL DPBS were added to subtracting the mean + 2 standard deviations (SDs) from negative
the donor and receiver chambers, respectively, of a rapid equilibrium controls and then normalizing against the geometric means of both
dialysis device containing a membrane with an 8 kDa molecular weight positive controls and reference genes. Expression levels were summa­
cut-off (ThermoFisher Scientific). After incubating the device at 37 ◦C rized as z-scores, which indicate the number of SDs by which the
for 6 h with gentle rocking, 50 μL aliquots of the donor and receiver expression level differs from the mean. Gene copy number (GCN) was
chambers were extracted with acetonitrile containing 2H-tepotinib. evaluated in 280 ng input DNA using the Cancer Copy Number Assay v2
Samples were separated on an Acquity UPLC instrument equipped with (NanoString), which contains 87 genes commonly amplified or deleted
a BEH phenyl column (2.1 × 30 mm; 1.7 μm) by gradient elution (mobile in cancer. Raw counts were normalized against 10 invariant reference
phase A: 0.1% formic acid in water; mobile phase B: 0.1% formic acid in probes, positive and negative controls in each hybridization reaction,
acetonitrile) at a 900 μL/min flow rate over 1 min. Detection utilized a and a DNA reference sample to generate GCN calls. Average GCN esti­
PE Sciex API 4000 detector (PerkinElmer, Inc., Waltham, MA, USA) with mate value was calculated per gene based on all probes for that gene
multiple reaction monitoring (m/z: 493.3–112.3 for tepotinib; relative to the GCN estimate in the normal DNA sample. High-level
496.4–115.2 for 2H-tepotinib). The concentrations of tepotinib in the amplification was defined as GCN > 10. Mutations were analyzed in
donor and receiver chambers were used to calculate the unbound frac­ 5 ng input DNA using the Vantage 3D Single Nucleotide Variant Solid
tion of tepotinib in brain (fu, brain) after accounting for dilution, ac­ Tumor Panel (NanoString) to analyze 104 actionable mutations across
cording to published methods [32]. Kp,uu was calculated as (Cbrain × fu, 25 genes relevant to cancer.
brain)/(Cplasma × fu, plasma).
2.6. Confirmation of tepotinib sensitivity in subcutaneous PDX models
2.4. Screening for tepotinib sensitivity in subcutaneous PDX models
Subcutaneous PDX models that were sensitive to tepotinib in the
PDX models were established at Crown Bioscience, Inc. (San Diego, screen were further evaluated to confirm the antitumor activity of
CA, USA) using samples from all human lung cancer brain metastases tepotinib in these models when administered at a clinically relevant dose
available from the supplier at the time of the experiment (n = 20). Tu­ in a larger sample size. These experiments were conducted at Crown
mors were not selected for MET alterations or other molecular features. Bioscience, Inc. (San Diego, CA) in 7–8.5-week-old female NOD-SCID
Tumor cell slurries were subcutaneously injected into the right flank of mice (Jackson Laboratory, Bar Harbor, ME, USA). Animals were
6–8-week-old female NOD-SCID mice (NOD.CB17-Prkdcscid/NCrHsd; housed individually in ventilated cages at 68–74◦F and 30–70% hu­
Envigo, Indianapolis, IN, USA). midity, with a 12-hour dark/12-hour light photoperiod, and free access
to irradiated soy rodent feed and sterile acidified water (pH 2.5–3.0).
Implanted mice were shipped to EMD Serono Research and Devel­
opment Institute (Billerica, MA, USA) 5–14 days after implantation. For each model, animals with established subcutaneous tumors were
Mice were housed and maintained in individually ventilated cages under assigned using the multi-task method to either tepotinib 125 mg/kg qd
specific pathogen-free conditions, and received a standard diet of irra­ po or vehicle (20% Solutol/80% 100 mM sodium acetate buffer, pH 5.5)
diated feed with free access to sterile water and, during treatment, a qd po (n = 5). Based on pharmacokinetic/pharmacodynamic modeling
nutritional supplement (ClearH2O, Westbrook, ME, USA). Tumor width [33], the 125 mg/kg qd dose in mice is predicted to achieve comparable
(w) and length (l) were regularly measured with digital calipers and minimal exposure to free tepotinib and similar levels of phospho-MET
used to estimate tumor volume using the formula l × w2 / 2. inhibition as the approved clinical dose of 500 mg (450 mg active
moiety) qd in humans. Tumors were measured three times per week and
Mice with established subcutaneous PDXs (tumor volume: 150–250 treatment continued until mean tumor volume in the control group
mm3) were treated with either vehicle (20% Solutol/80% 100 mM so­ reached 1,200 mm3.
dium acetate buffer, pH 5.5 or 0.5% Methocel, 0.25% Tween-20 in so­
dium citrate buffer; n = 1 per model) or tepotinib (n = 1 per model). In 2.7. Evaluation of tepotinib sensitivity in orthotopic PDX models
this initial screen, tepotinib was administered at a suboptimal dose (30
mg/kg qd po) to facilitate combination with other anticancer agents at PDX models with confirmed sensitivity to tepotinib after subcu­
the same dose level in separate experiments (data not shown). Duration taneous engraftment and verified MET alterations were investigated
of tepotinib treatment was determined by the time taken for the tumor of further through orthotopic implantation in the brains of 6–8-week old
the vehicle-treated mouse of the same model to reach 1,000 mm3 (±3 female NOD-SCID mice (Jackson Laboratory) at Crown Bioscience, Inc.
days). Tumors were measured 2–3 times per week. Mice were eutha­ Mice were housed in disposable, group caging under the same condi­
nized when tumor volume reached 1,000–1,500 mm3. Other humane tions as described above for the subcutaneous models.
endpoints included loss of > 20% body weight, wet ulceration of the
tumor, moribund condition, signs of distress and swollen abdomen. At Prior to tumor implantation, mice were anesthetized by intramus­
the time of euthanasia, tumor tissue was harvested from vehicle-treated cular injection of ketamine and xylazine, and the surgical site was
mice and snap frozen in liquid nitrogen for subsequent molecular shaved and disinfected with povidone-iodine followed by 75% ethanol.
profiling. A 2–3 mm incision was made just to the right of the midline and anterior
to the interaural line. Approximately 200,000 tumor cells were injected
2.5. Molecular profiling of screened PDX models intracranially in 2 µL of PBS at the right frontal lobe, 2 mm lateral and
0.5 mm anterior from the bregma, at a depth of 3.5 mm. The hole in the
For molecular characterization, RNA and DNA were extracted from skull was sealed with bone wax and the incision closed with a #6 suture,
approximately 10 mg of frozen tumor tissue using the RNAqueous-4PCR before disinfection with povidone-iodine. Mice were kept warm until
Kit or PureLink® Genomic DNA Mini Kit according to the manufac­ recovery from anesthesia and left undisturbed for 3 days thereafter.
turer’s instructions (Thermo Fisher Scientific). Purified DNA and RNA
were quantified using a Qubit® 3.0 Fluorometer (Thermo Fisher Sci­ For each model, 20 animals with established intracranial tumors
entific). Gene expression, amplification and mutation were evaluated were assigned based on tumor volume using the matched distribution
using the multiplex digital nCounter® platform, and analyzed with method to receive either tepotinib (125 mg/kg qd po; n = 10 per model)
or vehicle control (20% Solutol/80% 100 mM sodium acetate buffer, pH

79

M. Friese-Hamim et al. Lung Cancer 163 (2022) 77–86

5.5; qd po; n = 10 per model). Treatment was initiated on day 0. fraction of tepotinib remaining unbound to rat brain tissue (fu, brain) was
Duration of treatment (28 days for LU5349 and 16 days for LU5406) and 0.35% (±0.06) at 0.3 µM tepotinib and 0.35% (±0.10) at 3 µM. The
timing of tumor volume assessments were based on observations from a mean fraction (±SD) remaining unbound to rat plasma (fu, plasma) was
pilot study (data not shown). Mice were euthanized prior to treatment 4.0% (±0.4) at both 0.3 and 1.0 µM tepotinib. The unbound brain-to-
completion if humane endpoints were reached. plasma ratio (Kp,uu) was calculated at 0.25, indicating brain penetra­
tion with a fourfold lower concentration of free tepotinib in the brain
Tumor response and permeability were assessed using an 11.7 T relative to plasma.
Avance III BioSpec preclinical MRI system (Bruker, Billerica, MA, USA)
at Invicro Molecular Imaging Center (La Jolla, San Diego, CA, USA). 3.2. Efficacy of tepotinib in subcutaneous PDX models of lung cancer
Mice were anesthetized with isoflurane prior to imaging procedures. brain metastases
After tuning, shimming and pulse calibrations, a scout scan was per­
formed and used to design a contiguous slice packet covering the whole To investigate the preclinical activity of tepotinib against lung cancer
brain for subsequent scans. Intracranial tumor volume was assessed in brain metastases, we screened 20 subcutaneous PDX models for sensi­
T2-weighted (T2w) scans using a rapid acquisition with relaxation tivity to tepotinib at the suboptimal dose of 30 mg/kg qd. Tumors of two
enhancement protocol with eight echoes, 17.5 ms echo time, 2.5 s models (LU5349 and LU5406) regressed (Fig. 1A), both of which were
repetition time, and contiguous 0.5 mm slices over a 25 × 25 mm field of derived from patients with NSCLC that had metastasized to brain
view and 128 × 128 pixel matrix. T2-weighted scans were acquired on (Supplementary Table 1). At the end of treatment, %TV was –12% for
days 0, 4, 7, 11, 19 and 28 for LU5349, and on days 0, 2, 6, 11 and 16 for LU5349 and –88% for LU5406. Molecular profiling revealed that these
LU5406. For imaging of vascular permeability (i.e. BBB ‘leakiness’), T1- two PDXs were the only models in the tested set with high-level MET
weighted gadolinium contrast-enhanced (T1CE) magnetic resonance amplification (MET GCN > 10) (Fig. 1B). Mean MET GCN was 11.2 in
imaging (MRI) scans were acquired using a spin-echo protocol with 4.9 LU5349 and 24.2 in LU5406. MET amplification was accompanied by
ms echo time, 1 s repetition time, and contiguous 0.5 mm slices over a overexpression of MET mRNA in both models, and a concomitant KRAS
25 × 25 mm field of view and 128 × 128 pixel matrix. T1CE scans were mutation (G12C) was detected in LU5349 (Fig. 1C, D).
acquired in a subset of animals (n = 4 per group) on day 4 for LU5349
and on day 2 for LU5406. To confirm the sensitivity of subcutaneously implanted LU5349 and
LU5406 to tepotinib when administered at a clinically relevant dose, we
Regions of interest (ROIs) were manually segmented in each MRI treated mice with established tumors with tepotinib 125 mg/kg qd or
transverse slice using VivoQuant™ software (Invicro). Hyperintensities vehicle (n = 5) (see Supplementary Table 2 for baseline characteristics).
consistent with apparent ‘satellite tumors’ (i.e. secondary tumors sepa­ Tepotinib induced pronounced tumor regression in both models
rated from the primary tumor at the implant site) were excluded from (Fig. 2A, B). At day 21, mean (±standard error) tumor volume was 13.5
tumor region of interest segmentations, whereas fluid-filled spaces mm3 (±0.0) in the tepotinib-treated group, and 1,280.3 mm3 (±114.8)
within the primary tumor were included. Tumor volume (from T2w in the vehicle-treated group for LU5349. Corresponding values for
images) and leaky volume (from T1CE images) were calculated by LU5406 at day 16 were 13.5 mm3 (±0.0) and 1,452.6 mm3 (±150.7),
multiplying the area of each segmented slice by the slice thickness. The respectively.
leakiness of the tumor is reported in terms of ‘sensitivity’ and ‘precision’,
which were assessed via specific comparisons of ROIs of T1CE with T2w 3.3. Efficacy of tepotinib in orthotopic PDX models of NSCLC brain
voxels (i.e. three-dimensional pixels in the images). Sensitivity was metastases
defined as the percentage of the total tumor volume that was leaky (i.e.
T1CE volume that overlaps with T2w volume/total T2w volume × 100). We next evaluated intracranial efficacy by treating mice bearing
In other words, sensitivity represents the percentage of voxels in the orthotopically implanted tumors of both models with either tepotinib or
T2w tumor volume that are also included in the T1CE volume. Precision vehicle (n = 10) (see Supplementary Table 3 for baseline characteris­
was defined as the percentage of voxels within the total T1CE leaky tics). Treatment was discontinued early in some animals implanted with
volume ROI that do not extend outside the limits of the T2w tumor LU5349 due to tolerability issues relating to isoflurane anesthesia during
volume ROI (i.e. T1CE volume that overlaps with T2w volume/total imaging (tepotinib, n = 3; vehicle, n = 1).
T1CE volume × 100).
Median tumor volume increased >300% at termination in vehicle-
2.8. Activity endpoints treated mice for LU5349 and LU5406. Treatment with tepotinib resul­
ted in pronounced tumor regression in both orthotopic models: median
Percent tumor volume change (%TV) was calculated as the differ­ %TV at study end was –84% in LU5349 and –63% in LU5406 (Fig. 2C,
ence in tumor volume between the end and start of treatment as a per­ D). As shown in the waterfall plots in Fig. 3A, B, complete or near-
centage of starting tumor volume. For the orthotopic study, statistical complete regressions were observed with tepotinib in several animals.
analyses were conducted in Excel 2016 (Microsoft, Redmond, WA, USA)
and GraphPad Prism V8 (GraphPad Software, San Diego, CA, USA). 3.4. Relationship between antitumor activity and BBB leakiness in
orthotopic PDX models
3. Results
To investigate BBB integrity in the orthotopic models, BBB leakiness
3.1. Tepotinib brain penetration and binding was visualized as areas of contrast enhancement in T1CE MRI images
(Supplementary Fig. 1). All implanted tumors had regions with intact
To investigate the biodistribution of tepotinib in Wistar rats, tepo­ and disrupted BBB. BBB leakiness was relatively variable among models,
tinib was administered to three animals at a rate of 3.66 mg/kg/hour by with sensitivity (i.e. percentage of the tumor that is leaky) ranging be­
intravenous infusion. After 24 h (i.e. at steady-state), mean total tepo­ tween 32% and 77% at day 4 for LU5349, and between 20% and 56% at
tinib concentration was 505 ng/g (SD: 22; range: 480–519) in the brain day 2 for LU5406 (Fig. 3C, D). There were no apparent associations
and 177 ng/mL (SD: 20; range: 162–200) in plasma. The mean (±SD) between sensitivity at day 4 (LU5349) or day 2 (LU5406) and tumor
total brain-to-plasma (Kp) ratio was 2.87 (±0.24), indicating almost volume change at the end of the study.
threefold higher total tissue concentration of tepotinib in the brain
relative to plasma. 4. Discussion

We next evaluated in vitro tepotinib binding to rat brain and plasma There is an urgent need for targeted therapies that can effectively
in equilibrium dialysis experiments. After equilibration, the mean (±SD)

80

M. Friese-Hamim et al. Lung Cancer 163 (2022) 77–86

Fig. 1. Tepotinib induced tumor regression in two of 20 screened subcutaneous PDX models of lung cancer brain metastasis, both of which were found to harbor
high-level MET amplification. Results of screening for tepotinib sensitivity in subcutaneous lung cancer brain metastasis PDX models (A) and subsequent molecular
profiling via nCounter® of copy number variation (B), mutation status (C), and mRNA expression level (D) of selected genes. Amplification was defined as GCN > 4
and ≤ 10, high-level amplification as GCN > 10, deletion as GCN < 1.2, and wild-type as GCN ≥ 1.2 and ≤ 4. %TV = percent tumor volume; GCN = gene copy
number; PDX = patient-derived xenograft.

treat brain metastases of NSCLC [18]. Since patients with brain lesions evidence of systemic and intracranial responses to tepotinib in patients
are often underrepresented in clinical trials, preclinical experiments can
provide evidence regarding the potential for intracranial activity in the with METex14 skipping NSCLC and brain metastases [9,10,25–28],
clinical setting, for example, by evaluating drug CNS permeability and these data support the further prospective evaluation of the intracranial
activity in xenograft models of brain metastases [24]. In the present
study, we demonstrate that tepotinib penetrates the intact BBB in rats, efficacy of tepotinib in patients with NSCLC with MET alterations.
has relatively high binding to rat brain tissue, and is active against MET- The principal measure of tepotinib brain penetration in the present
driven NSCLC brain metastases in two different murine PDX models.
Antitumor activity was shown in both subcutaneously and orthotopi­ study was Kp,uu, which is considered a more relevant metric of brain
cally implanted models, the latter of which mimics the anatomic situa­ exposure than total brain-to-plasma ratio (Kp) [23]. Total tepotinib
tion of the parent tumor. In the orthotopic experiments, tepotinib concentrations were 2.87-fold greater in the brain than in plasma (Kp =
induced pronounced tumor regression, including complete or near- 2.87), whereas, due to the relatively high binding of tepotinib in the
complete regressions, in both models tested. Together with clinical
brain (unbound fraction < 0.4%), free tepotinib concentrations in the
brain were 25% of that in plasma (Kp,uu = 0.25). The Kp,uu of 0.25
suggests penetration of the BBB by tepotinib and is similar to that re­
ported for the CNS-penetrant EGFR inhibitor osimertinib (0.21) [34].

81

M. Friese-Hamim et al. Lung Cancer 163 (2022) 77–86

Fig. 2. Tepotinib had pronounced antitumor efficacy at a clinically relevant dose in subcutaneous and orthotopic PDX models of NSCLC brain metastases with high-
level MET amplification. Mean tumor volume ± SEM of mice bearing SC tumors (n = 5) of LU5349 (A) or LU5406 (B), or intracranial tumors (n = 10) of LU5349 (C)
or LU5406 (D), which were treated with tepotinib 125 mg/kg qd po or vehicle. NSCLC = non-small cell lung cancer; PDX = patient-derived xenograft; po = orally; qd
= once daily; SC = subcutaneous; SEM = standard error of the mean.

Tepotinib brain penetration may be favored by its lipophilicity (logP of METex14 skipping NSCLC brain metastases [9,10,25–28]. For example,
3), moderate-to-high passive permeability of 14.7 × 10-6 cm/s, and in VISION Cohort A, the efficacy of tepotinib in patients with asymp­
plasma protein binding (97–98% in humans) [31,35]. The Kp,uu of < 1 tomatic and/or neurologically stable brain metastases at baseline was
suggests active export out of the brain [29], consistent with data comparable to that in the overall METex14 skipping population [9,10].
showing tepotinib is a substrate for P-glycoprotein [8]. Furthermore, intracranial activity of tepotinib was demonstrated in an
ad hoc analysis of brain lesion responses per Response Assessment in
In the clinical setting, brain permeability of tepotinib can be inferred Neuro-Oncology Brain Metastases criteria [10,42]. Marked intracranial
from intracranial responses observed in patients with METex14 skipping responses to tepotinib have also been documented in case reports based
NSCLC both in VISION [10,25] and case reports from clinical practice on four patients in clinical practice with NSCLC with MET alterations
[26–28] (discussed further below). Direct evidence of CNS penetration and symptomatic CNS metastases [26–28,43]. Recently, complete
was provided in two of these cases, which documented cerebrospinal response to adjuvant tepotinib has been described in a patient with
fluid concentrations of tepotinib of 62 nM and 49.7–59.3 nM [27,28], disseminated glioblastoma with MET amplification [44]. Collectively,
which exceed the half-maximal inhibitory concentration of tepotinib for these data indicate promising intracranial activity of tepotinib in pa­
MET (1.7 nM) [7]. To our knowledge, Kp,uu has not been reported for tients with MET-driven brain lesions.
other MET inhibitors. However, it is noteworthy that the Kp for tepotinib
(2.87) is considerably greater than that reported for capmatinib (0.09) Clinical activity against METex14 skipping NSCLC brain metastases
in rats, and for cabozantinib (0.20) in mice [36,37]. Furthermore, has been observed with other MET inhibitors in phase 2 trials of cap­
limited BBB penetration has been linked to CNS failure of crizotinib in a matinib [45] and savolitinib [46]. Case reports or retrospective studies
patient with ALK-positive NSCLC with brain metastases [38]. have described responses of brain metastases to crizotinib or cabo­
zantinib in patients with METex14 skipping NSCLC, and to capmatinib
The preclinical activity of tepotinib in PDXs from NSCLC brain me­ in combination with EGFR inhibitors in patients with MET amplification
tastases with MET amplification adds to the body of data demonstrating and EGFR mutation [15,47–49]. These results support the concept of
tumor growth inhibition and regression with tepotinib in subcutaneous targeted therapy with MET inhibitors in patients with NSCLC harboring
xenograft models of MET-driven NSCLC, hepatocellular carcinoma and MET alterations that has metastasized to the CNS. Recent clinical prac­
gastric cancer, as well as glioblastoma [39–41]. In line with preclinical tice guidelines recommend CNS-penetrant tyrosine kinase inhibitors for
data, there is mounting evidence for the clinical activity of tepotinib in

82

M. Friese-Hamim et al. Lung Cancer 163 (2022) 77–86

Fig. 3. Tepotinib induced tumor regression in multiple animals bearing orthotopic PDX models of NSCLC brain metastases with high-level MET amplification, with
no apparent correlation between efficacy and BBB leakiness. Waterfall plots of %TV for each animal treated with tepotinib (125 mg/kg qd po) or vehicle in models
LU5349 (A) and LU5406 (B). %TV at termination and corresponding sensitivity (percentage of tumor leakiness inside the tumor) per animal at day 4 for LU5349 (C)
and day 2 for LU5406 (D). BBB leakiness was evaluated in a subset of animals in the orthotopic study (n = 4). *Sensitivity was not evaluable due to pronounced tumor
regression. %TV = percent tumor volume; BBB = blood–brain barrier; NSCLC = non-small cell lung cancer; PDX = patient-derived xenograft; po = orally; qd =
once daily.

83

M. Friese-Hamim et al. Lung Cancer 163 (2022) 77–86

treatment of NSCLC with driver mutations, such as METex14 skipping, Acknowledgements
and brain metastases [50].
Medical writing assistance was provided by Mark Dyson, DPhil
In the orthotopic PDX models, contrast-weighted MRI indicated (Berlin, Germany) on behalf of Syneos Health (London, UK), and funded
parallel presence of regions with intact and disrupted BBB in all by Merck Healthcare KGaA, Darmstadt, Germany.
implanted tumors, which is thought to mirror the clinical situation [51].
While variability in vascular (i.e. BBB) leakiness among individual tu­ Appendix A. Supplementary data
mors was high, there was no apparent association between the degree of
leakiness and response to tepotinib. This suggests that ability to cross the Supplementary data to this article can be found online at https://doi.
intact BBB may be more important for intracranial efficacy against org/10.1016/j.lungcan.2021.11.020.
existing brain metastases than the extent of BBB disruption, but further
data are required for a firm conclusion. Agents that cross the intact BBB References
may also be more effective in preventing the occurrence of new brain
lesions [24]. Interestingly, the rate of new brain lesions in VISION was [1] National Comprehensive Cancer Network, Non-Small Cell Lung Cancer Version
low (3.9%) [10], and a case report from the study showed no re- 4.2021, (2021). https://www.nccn.org/professionals/physician_gls/pdf/nscl.pdf
emergence of brain lesions during tepotinib treatment in a patient (accessed March 18, 2021).
with intracranial response to prior therapy [25].
[2] N.H. Hanna, A.G. Robinson, S. Temin, S. Baker, J.R. Brahmer, P.M. Ellis, L.
Limitations of the present work include the potential impact of BBB E. Gaspar, R.Y. Haddad, P.J. Hesketh, D. Jain, I. Jaiyesimi, D.H. Johnson, N.
disruption during orthotopic implantation and the use of immunodefi­ B. Leighl, P.R. Moffitt, T. Phillips, G.J. Riely, R. Rosell, J.H. Schiller, B.J. Schneider,
cient mice, in which antitumor immune responses are attenuated [24]. N. Singh, D.R. Spigel, J. Tashbar, G. Masters, Therapy for Stage IV Non–Small-Cell
Study strengths include the use of PDX models to avoid artifacts asso­ Lung Cancer With Driver Alterations: ASCO and OH (CCO) Joint Guideline Update,
ciated with ex vivo passaging of cell lines, and intracranial PDX im­ JCO 39 (9) (2021) 1040–1091.
plantation to reproduce the localization of the original tumor [30].
[3] L. Hong, J. Zhang, J.V. Heymach, X. Le, Current and future treatment options for
In summary, tepotinib penetrated the intact rat BBB to an extent that MET exon 14 skipping alterations in non-small cell lung cancer, Ther. Adv. Med.
was sufficient to achieve intracranial target inhibition, and induced Oncol. 13 (2021), https://doi.org/10.1177/1758835921992976.
pronounced tumor regression following orthotopic implantation of two
murine PDXs from MET-driven NSCLC brain metastases. Alongside the [4] Y.L. Wu, R.A. Soo, G. Locatelli, U. Stammberger, G. Scagliotti, K. Park, Does c-Met
clinical evidence to date [9,10,25–28], these data provide support for remain a rational target for therapy in patients with EGFR TKI-resistant non-small
the intracranial activity of tepotinib, which could provide a cell lung cancer? Cancer Treat. Rev. 61 (2017) 70–81, https://doi.org/10.1016/j.
radiotherapy-sparing treatment approach for patients with METex14 ctrv.2017.10.003.
skipping NSCLC and brain metastases.
[5] G.M. Frampton, S.M. Ali, M. Rosenzweig, J. Chmielecki, X. Lu, T.M. Bauer,
5. Funding statement M. Akimov, J.A. Bufill, C. Lee, D. Jentz, R. Hoover, S.-H. Ou, R. Salgia, T. Brennan,
Z.R. Chalmers, S. Jaeger, A. Huang, J.A. Elvin, R. Erlich, A. Fichtenholtz, K.
This work was funded by Merck Healthcare KGaA, Darmstadt, A. Gowen, J. Greenbowe, A. Johnson, D. Khaira, C. McMahon, E.M. Sanford,
Germany. S. Roels, J. White, J. Greshock, R. Schlegel, D. Lipson, R. Yelensky, D. Morosini, J.
S. Ross, E. Collisson, M. Peters, P.J. Stephens, V.A. Miller, Activation of MET via
6. Disclosures diverse exon 14 splicing alterations occurs in multiple tumor types and confers
clinical sensitivity to MET inhibitors, Cancer Discov. 5 (8) (2015) 850–859.
Manja Friese-Hamim, Christof Reusch, Olga Bogatyrova, Dominique
Perrin, Jürgen Schmidt, Martin Schaefer, and Christopher Stroh are [6] A. Drilon, F. Cappuzzo, S.-H. Ou, D.R. Camidge, Targeting MET in lung cancer: Will
employees of Merck Healthcare KGaA, Darmstadt, Germany. Anderson expectations finally be MET? J. Thorac. Oncol. 12 (1) (2017) 15–26, https://doi.
Clark, Lindsey Crowley, Hong Zhang, Timothy Crandall, Jing Lin, org/10.1016/j.jtho.2016.10.014.
Jianguo Ma, and David Bachner are employees of EMD Serono, Billerica,
MA, USA, an affiliate of Merck KGaA. [7] G.S. Falchook, R. Kurzrock, H.M. Amin, W. Xiong, S. Fu, S.A. Piha-Paul, F. Janku,
G. Eskandari, D.V. Catenacci, M. Klevesath, R. Bruns, U.z. Stammberger, A. Johne,
CRediT authorship contribution statement F. Bladt, M. Friese-Hamim, P. Girard, S. El Bawab, D.S. Hong, First-in-man Phase I
trial of the selective MET inhibitor tepotinib in patients with advanced solid
Manja Friese-Hamim: Conceptualization, Data curation, Formal tumors, Clin. Cancer Res. 26 (6) (2020) 1237–1246, https://doi.org/10.1158/
analysis, Methodology, Supervision, Visualization, Writing – original 1078-0432.CCR-19-2860.
draft, Writing – review & editing. Anderson Clark: Conceptualization,
Formal analysis, Supervision, Writing – original draft, Writing – review [8] FDA, TEPMETKO (tepotinib) Prescribing Information, (2021). https://www.
& editing. Dominique Perrin: Formal analysis, Writing – review & accessdata.fda.gov/drugsatfda_docs/label/2021/214096s000lbl.pdf (accessed
editing. Lindsey Crowley: Data curation, Formal analysis, Methodol­ March 3, 2021).
ogy, Writing – original draft, Writing – review & editing. Christof
Reusch: Data curation, Formal analysis, Supervision, Writing – review [9] P.K. Paik, E. Felip, R. Veillon, H. Sakai, A.B. Cortot, M.C. Garassino, J. Mazieres,
& editing. Olga Bogatyrova: Formal analysis, Visualization, Writing – S. Viteri, H. Senellart, J. Van Meerbeeck, J.o. Raskin, N. Reinmuth, P. Conte,
review & editing. Hong Zhang: Data curation, Formal analysis, Meth­ D. Kowalski, B.C. Cho, J.D. Patel, L. Horn, F. Griesinger, J.-Y. Han, Y.-C. Kim, G.-
odology, Supervision, Writing – review & editing. Timothy Crandall: C. Chang, C.-L. Tsai, J.-H. Yang, Y.-M. Chen, E.F. Smit, A.J. van der Wekken,
Data curation, Methodology, Writing – review & editing. Jing Lin: Data T. Kato, D. Juraeva, C. Stroh, R. Bruns, J. Straub, A. Johne, J. Scheele, J.
curation, Methodology, Writing – review & editing. Jianguo Ma: Data V. Heymach, X. Le, Tepotinib in non-small-cell lung cancer with MET exon 14
curation, Methodology, Supervision, Writing – review & editing. David skipping mutations, N. Engl. J. Med. 383 (10) (2020) 931–943, https://doi.org/
Bachner: Methodology, Writing – review & editing. Jürgen Schmidt: 10.1056/NEJMoa2004407.
Data curation, Formal analysis, Writing – review & editing. Martin
Schaefer: Data curation, Formal analysis, Writing – review & editing. [10] X. Le, H. Sakai, E. Felip, R. Veillon, M.C. Garassino, J. Raskin, A.B. Cortot, S. Viteri,
Christopher Stroh: Conceptualization, Formal analysis, Supervision, J. Mazieres, E.F. Smit, M. Thomas, W.T. Iams, B.C. Cho, H.R. Kim, J. Chih-Hsin
Writing – original draft, Writing – review & editing. Yang, Y.-M. Chen, J. Patel, C.M. Bestvina, K. Park, F. Griesinger, M. Johnson,
M. Gottfried, C. Britschgi, J. Heymach, E. Sikoglu, K. Berghoff, K.-M. Schumacher,
R. Bruns, G. Otto, P.K. Paik, Tepotinib efficacy and safety in patients with MET
exon 14 skipping NSCLC: outcomes in patient subgroups from VISION relevant for
clinical practice, Clin. Cancer Res. (2021), https://doi.org/10.1158/1078-0432.
CCR-21-2733.

[11] X. Le, L.G. Paz-Ares, J. Van Meerbeeck, S. Viteri, C. Cabrera Galvez, D. Vicente Baz,
Y.-C. Kim, J.-H. Kang, K.-M. Schumacher, N. Karachaliou, S. Adrian, R. Bruns, P.
K. Paik, Tepotinib in patients (pts) with advanced non-small cell lung cancer
(NSCLC) with MET amplification (METamp), 9021–9021, J. Clin. Oncol. 39 (15_
suppl) (2021), https://doi.org/10.1200/JCO.2021.39.15_suppl.9021.

[12] Y.-L. Wu, Y. Cheng, J. Zhou, S. Lu, Y. Zhang, J. Zhao, D.-W. Kim, R.A. Soo, S.-
W. Kim, H. Pan, Y.-M. Chen, C.-F. Chian, X. Liu, D.S.W. Tan, R. Bruns, J. Straub,
A. Johne, J. Scheele, K. Park, J.-H. Yang, Y.-L. Wu, X. Liu, Z. Liu, S. Lu, X.i. Chen,
H. Pan, M. Wang, S. Yu, H. Zhang, Y. Zhang, J. Fang, W. Li, J. Zhou, J. Zhao,
Y. Cheng, C.-H. Yang, G.-C. Chang, Y.-M. Chen, T.-C. Hsia, C.-F. Chian, C.-T. Yang,
C.-C. Wang, S.-W. Kim, K. Park, D.-W. Kim, B.C. Cho, K.H. Lee, Y.-C. Kim, H.J. An,
I.S. Woo, J.Y. Cho, S.W. Shin, J.-S. Lee, J.-H. Kim, S.S. Yoo, T. Kato, N. Shinagawa,
R.A. Soo, S.W.D. Tan, L.-M. Ngo, K. Ratnavelu, A.R. Ahmad, C.K. Liam, F. de
Marinis, P. Tassone, A.I. Molla, A. Calles Blanco, M.E. Lazaro Quintela, E. Felip
Font, A.-M. Dingemans, L. Bui, Tepotinib plus gefitinib in patients with EGFR-
mutant non-small-cell lung cancer with MET overexpression or MET amplification
and acquired resistance to previous EGFR inhibitor (INSIGHT study): an open-

84

M. Friese-Hamim et al. Lung Cancer 163 (2022) 77–86

label, phase 1b/2, multicentre, randomised trial, Lancet Respir. Med. 8 (11) (2020) [32] J. Cory Kalvass, T.S. Maurer, Influence of nonspecific brain and plasma binding on
1132–1143, https://doi.org/10.1016/S2213-2600(20)30154-5. CNS exposure: Implications for rational drug discovery, Biopharm. Drug Dispos. 23
[13] P.W. Sperduto, T.J. Yang, K. Beal, H. Pan, P.D. Brown, A. Bangdiwala, R. Shanley, (8) (2002) 327–338, https://doi.org/10.1002/bdd.325.
N. Yeh, L.E. Gaspar, S. Braunstein, P. Sneed, J. Boyle, J.P. Kirkpatrick, K.S. Mak, H.
A. Shih, A. Engelman, D. Roberge, N.D. Arvold, B. Alexander, M.M. Awad, [33] W. Xiong, M. Friese-Hamim, A. Johne, C. Stroh, M. Klevesath, G.S. Falchook, D.
J. Contessa, V. Chiang, J. Hardie, D. Ma, E. Lou, W. Sperduto, M.P. Mehta, The S. Hong, P. Girard, S. El Bawab, Translational pharmacokinetic-pharmacodynamic
effect of gene alterations and tyrosine kinase inhibition on survival and cause of modeling of preclinical and clinical data of the oral MET inhibitor tepotinib to
death in patients with adenocarcinoma of the lung and brain metastases, Int. J. determine the recommended phase II dose, CPT Pharmacometrics Syst. Pharmacol.
Radiat. Oncol. Biol. Phys. 96 (2) (2016) 406–413, https://doi.org/10.1016/j. 10 (5) (2021) 428–440, https://doi.org/10.1002/psp4.12602.
ijrobp.2016.06.006.
[14] S.R. Digumarthy, D.P. Mendoza, E.W. Zhang, J.K. Lennerz, R.S. Heist, [34] N. Colclough, K. Chen, P. Johnstro¨m, N. Strittmatter, Y. Yan, G.L. Wrigley,
Clinicopathologic and imaging features of non-small-cell lung cancer with MET M. Schou, R. Goodwin, K. Varn¨as, S.J. Adua, M. Zhao, D.X. Nguyen, G. Maglennon,
exon 14 skipping mutations, Cancers (Basel) 11 (2019) 1–11, https://doi.org/ P. Barton, J. Atkinson, L. Zhang, A. Janefeldt, J. Wilson, A. Smith, A. Takano,
10.3390/cancers11122033. R. Arakawa, M. Kondrashov, J. Malmquist, E. Revunov, A. Vazquez-Romero, M.
[15] M. Offin, J. Luo, R. Guo, J.K. Lyo, C. Falcon, J. Dienstag, O. Wilkins, J. Chang, C. M. Moein, A.D. Windhorst, N.A. Karp, M.R.V. Finlay, R.A. Ward, J.W.T. Yates, P.
M. Rudin, G. Riely, N. Rekhtman, M.E. Arcila, G. Heller, M. Ladanyi, B.T. Li, M. D. Smith, L. Farde, Z. Cheng, D.A.E. Cross, Preclinical comparison of the blood-
G. Kris, P. Paik, A. Drilon, CNS metastases in patients with MET exon 14–altered brain barrier permeability of osimertinib with other EGFR TKIs, Clin. Cancer Res.
lung cancers and outcomes with crizotinib, JCO Precis. Oncol. 4 (2020) 871–876, 27 (1) (2021) 189–201, https://doi.org/10.1158/1078-0432.CCR-19-1871.
https://doi.org/10.1200/po.20.00098.
[16] G. Stella, A. Corino, G. Berzero, S. Kolling, A. Filippi, S. Benvenuti, Brain [35] C. Fink, M. Lecomte, L. Badolo, K. Wagner, K. M¨ader, S.-A. Peters, Identification of
metastases from lung cancer: Is MET an actionable target? Cancers (Basel) 11 (3) solubility-limited absorption of oral anticancer drugs using PBPK modeling based
(2019) 271, https://doi.org/10.3390/cancers11030271. on rat PK and its relevance to human, Eur. J. Pharm. Sci. 152 (2020) 105431,
[17] M. Preusser, B. Streubel, A.S. Berghoff, J.A. Hainfellner, A. von Deimling, https://doi.org/10.1016/j.ejps.2020.105431.
G. Widhalm, K. Dieckmann, A. Wo¨hrer, M. Hackl, C. Zielinski, P. Birner,
Amplification and overexpression of CMET is a common event in brain metastases [36] FDA, TABRECTA (capmatinib) Prescribing Information, (2020). https://www.
of non-small cell lung cancer, Histopathology 65 (5) (2014) 684–692. accessdata.fda.gov/drugsatfda_docs/label/2020/213591s000lbl.pdf (accessed
[18] M.C. Chamberlain, C.S. Baik, V.K. Gadi, S. Bhatia, L.Q.M. Chow, Systemic therapy March 23, 2021).
of brain metastases: Non-small cell lung cancer, breast cancer, and melanoma,
Neuro Oncol. 19 (1) (2017) i1–i24, https://doi.org/10.1093/neuonc/now197. [37] A. Abdelaziz, U. Vaishampayan, Cabozantinib for the treatment of kidney cancer,
[19] S. Peters, C. Bexelius, V. Munk, N. Leighl, The impact of brain metastasis on quality Expert Rev. Anticancer Ther. 17 (7) (2017) 577–584, https://doi.org/10.1080/
of life, resource utilization and survival in patients with non-small-cell lung cancer, 14737140.2017.1344553.
Cancer Treat. Rev. 45 (2016) 139–162, https://doi.org/10.1016/j.
ctrv.2016.03.009. [38] D.B. Costa, S. Kobayashi, S.S. Pandya, W.-L. Yeo, Z. Shen, W. Tan, K.D. Wilner, CSF
[20] S. Amin, M.J. Baine, J.L. Meza, C. Lin, Association of immunotherapy with survival concentration of the anaplastic lymphoma kinase inhibitor crizotinib, J. Clin.
among patients with brain metastases whose cancer was managed with definitive Oncol. 29 (15) (2011) e443–e445, https://doi.org/10.1200/JCO.2010.34.1313.
surgery of the primary tumor, JAMA Netw. Open 3 (9) (2020) e2015444, https://
doi.org/10.1001/jamanetworkopen.2020.15444. [39] F. Bladt, B. Faden, M. Friese-Hamim, C. Knuehl, C. Wilm, C. Fittschen, U. Gra¨dler,
[21] I. Eguren-Santamaria, M.F. Sanmamed, S.B. Goldberg, H.M. Kluger, M.A. Idoate, B. M. Meyring, D. Dorsch, F. Jaehrling, U. Pehl, F. Stieber, O. Schadt, A. Blaukat, EMD
Y. Lu, J. Corral, K.A. Schalper, R.S. Herbst, I. Gil-Bazo, PD-1/PD-L1 blockers in 1214063 and EMD 1204831 constitute a new class of potent and highly selective c-
NSCLC brain metastases: Challenging paradigms and clinical practice, Clin. Cancer Met inhibitors, Clin. Cancer Res. 19 (11) (2013) 2941–2951.
Res. 26 (16) (2020) 4186–4197, https://doi.org/10.1158/1078-0432.CCR-20-
0798. [40] F. Bladt, M. Friese-Hamim, C. Ihling, C. Wilm, A. Blaukat, The c-Met inhibitor
[22] N. Vilarin˜o, J. Bruna, J. Bosch-Barrera, M. Valiente, E. Nadal, Immunotherapy in MSC2156119J effectively inhibits tumor growth in liver cancer models, Cancers
NSCLC patients with brain metastases. Understanding brain tumor (Basel) 6 (2014) 1736–1752, https://doi.org/10.3390/cancers6031736.
microenvironment and dissecting outcomes from immune checkpoint blockade in
the clinic, Cancer Treat. Rev. 89 (2020) 102067, https://doi.org/10.1016/j. [41] M. Friese-Hamim, F. Bladt, G. Locatelli, U. Stammberger, A. Blaukat, The selective
ctrv.2020.102067. c-Met inhibitor tepotinib can overcome epidermal growth factor receptor inhibitor
[23] S. Varadharajan, S. Winiwarter, L. Carlsson, O. Engkvist, A. Anantha, T. Kogej, resistance mediated by aberrant c-Met activation in NSCLC models, Am. J. Cancer
M. Frid´en, J. Stålring, H. Chen, Exploring in silico prediction of the unbound brain- Res. 7 (2017) 962–972. www.ajcr.us/ (accessed February 25, 2021).
to-plasma drug concentration ratio: Model validation, renewal, and interpretation,
J. Pharm. Sci. 104 (3) (2015) 1197–1206, https://doi.org/10.1002/jps.24301. [42] N.U. Lin, E.Q. Lee, H. Aoyama, I.J. Barani, D.P. Barboriak, B.G. Baumert,
[24] D.R. Camidge, E.Q. Lee, N.U. Lin, K. Margolin, M.S. Ahluwalia, M. Bendszus, S. M. Bendszus, P.D. Brown, D.R. Camidge, S.M. Chang, J. Dancey, E.G.E. de Vries, L.
M. Chang, J. Dancey, E.G.E. de Vries, G.J. Harris, F.S. Hodi, A.B. Lassman, D. E. Gaspar, G.J. Harris, F.S. Hodi, S.N. Kalkanis, M.E. Linskey, D.R. Macdonald,
R. Macdonald, D.M. Peereboom, D. Schiff, R. Soffietti, M.J. van den Bent, J. K. Margolin, M.P. Mehta, D. Schiff, R. Soffietti, J.H. Suh, M.J. van den Bent, M.
S. Wefel, P.Y. Wen, Clinical trial design for systemic agents in patients with brain A. Vogelbaum, P.Y. Wen, Response assessment criteria for brain metastases:
metastases from solid tumours: a guideline by the Response Assessment in Neuro- Proposal from the RANO group, Lancet Oncol. 16 (6) (2015) e270–e278, https://
Oncology Brain Metastases working group, Lancet Oncol. 19 (1) (2018) e20–e32, doi.org/10.1016/S1470-2045(15)70057-4.
https://doi.org/10.1016/S1470-2045(17)30693-9.
[25] K.G. Roth, I. Mambetsariev, R. Salgia, Prolonged survival and response to tepotinib [43] F. Blanc-Durand, R. Alameddine, A.J. Iafrate, D. Tran-Thanh, Y. Lo, N. Blais,
in a non-small-cell lung cancer patient with brain metastases harboring MET exon B. Routy, M. Tehfe´, C. Leduc, P. Romeo, P. Stephenson, M. Florescu, Tepotinib
14 mutation: a research report, Cold Spring Harb. Mol. Case Stud. 6 (6) (2020) efficacy in a patient with non-small cell lung cancer with brain metastasis
a005785, https://doi.org/10.1101/mcs.a005785. harboring an HLA-DRB1-MET gene fusion, Oncologist 25 (2020) 916–920, https://
[26] S. Takamori, T. Matsubara, T. Fujishita, K. Ito, R. Toyozawa, T. Seto, doi.org/10.1634/theoncologist.2020-0502.
M. Yamaguchi, T. Okamoto, Dramatic intracranial response to tepotinib in a
patient with lung adenocarcinoma harboring MET exon 14 skipping mutation, [44] L. Pham, C. Gann, K.M. Schumacher, S. Vlassak, T. Swanson, K. Highsmith, A. Ou,
Thorac. Cancer 12 (2021) 978–980, https://doi.org/10.1111/1759-7714.13871. N. Clarke, A. Aaroe, L. Robell, B. O’Brien, S. Nash, J. DeGroot, N. Majd, INNV-21.
[27] T. Ninomaru, H. Okada, M. Fujishima, K. PhD, S. Irie, A.H. Fukushima, Lazarus Complete response to adjuvant tepotinib in a patient with newly diagnosed
response to tepotinib for leptomeningeal metastases in a MET exon 14 skipping disseminated glioblastoma (GBM) harboring MET amplification, Neuro. Oncol. 23
mutation-positive lung adenocarcinoma patient: Case report, JTO Clin. Res. Rep. (2021) vi109–vi109. https://doi.org/10.1093/NEUONC/NOAB196.432.
(2021) 100145, https://doi.org/10.1016/j.jtocrr.2021.100145.
[28] H. Tanaka, K. Taima, T. Makiguchi, J. Nakagawa, T. Niioka, S. Tasaka, Activity and [45] J. Wolf, T. Seto, J.-Y. Han, N. Reguart, E.B. Garon, H.J.M. Groen, D.S.W. Tan,
bioavailability of tepotinib for leptomeningeal metastasis of NSCLC with MET exon T. Hida, M. de Jonge, S.V. Orlov, E.F. Smit, P.-J. Souquet, J. Vansteenkiste,
14 skipping mutation, Cancer Commun. 41 (1) (2021) 83–87, https://doi.org/ M. Hochmair, E. Felip, M. Nishio, M. Thomas, K. Ohashi, R. Toyozawa, T.
10.1002/cac2.12124. R. Overbeck, F. de Marinis, T.-M. Kim, E. Laack, A. Robeva, S. Le Mouhaer,
[29] A. Reichel, Addressing central nervous system (CNS) penetration in drug discovery: M. Waldron-Lynch, B. Sankaran, O.A. Balbin, X. Cui, M. Giovannini, M. Akimov, R.
Basics and implications of the evolving new concept, Chem. Biodivers. 6 (11) S. Heist, Capmatinib in MET exon 14–mutated or MET-amplified non–small-cell
(2009) 2030–2049, https://doi.org/10.1002/cbdv.200900103. lung cancer, N. Engl. J. Med. 383 (10) (2020) 944–957.
[30] M. Patrizii, M. Bartucci, S.R. Pine, H.E. Sabaawy, Utility of glioblastoma patient-
derived orthotopic xenografts in drug discovery and personalized therapy, Front. [46] S. Lu, J. Fang, X. Li, J. Zhou, L. Cao, Y. Cheng, L. Jiang, Q. Guo, Z. Liang, Y. Chen,
Oncol. 8 (2018), https://doi.org/10.3389/fonc.2018.00023. H. Zhang, N. Yang, H. Xu, X. Zhang, B. Wu, J. Shi, Z. Han, J. Huang, Z. Yang,
[31] D. Dorsch, O. Schadt, F. Stieber, M. Meyring, U. Gra¨dler, F. Bladt, M. Friese- X. Zhang, G. Chen, Y. Hu, J. Wu, S. Zeng, S. Sun, L. Zhang, R. Ma, X. Dong,
Hamim, C. Knühl, U. Pehl, A. Blaukat, Identification and optimization of D. Zhang, J. Li, L. Wang, Y. Ren, W. Su, Abstract 5707: Phase II Study of Savolitinib
pyridazinones as potent and selective c-Met kinase inhibitors, Bioorgan. Med. in Patients with NSCLC Harboring MET Exon 14 Skipping Mutations: Preliminary
Chem. Lett. 25 (7) (2015) 1597–1602, https://doi.org/10.1016/j. Efficacy and Safety Results, CSCO (2019).
bmcl.2015.02.002.
[47] S.J. Klempner, A. Borghei, B. Hakimian, S.M. Ali, S.-H. Ou, Intracranial activity of
cabozantinib in MET exon 14–positive NSCLC with brain metastases, J. Thorac.
Oncol. 12 (1) (2017) 152–156, https://doi.org/10.1016/j.jtho.2016.09.127.

[48] J.F. Gainor, S.E. Stevens, H. Willers, H.A. Shih, R.S. Heist, Intracranial activity of
gefitinib and capmatinib in a patient with previously treated non–small cell lung
cancer harboring a concurrent EGFR mutation and MET amplification, J. Thorac.
Oncol. 15 (2020) e8–e10, https://doi.org/10.1016/j.jtho.2019.07.024.

85

M. Friese-Hamim et al. Lung Cancer 163 (2022) 77–86

[49] O. Gautschi, R. Menon, M. Bertrand, C. Murer, J. Diebold, Capmatinib and P. Wesseling, M. Weller, M. Preusser, EANO-ESMO Clinical Practice Guidelines for
osimertinib combination therapy for EGFR-mutant lung adenocarcinoma, diagnosis, treatment and follow-up of patients with brain metastasis from solid
J. Thorac. Oncol. 15 (2020) e13–e15, https://doi.org/10.1016/j.jtho.2019.07.027. tumours, Ann. Oncol. 32 (11) (2021) 1332–1347.
[51] C.D. Arvanitis, G.B. Ferraro, R.K. Jain, The blood–brain barrier and blood–tumour
[50] E. Le Rhun, M. Guckenberger, M. Smits, R. Dummer, T. Bachelot, F. Sahm, barrier in brain tumours and metastases, Nat. Rev. Cancer 20 (1) (2020) 26–41,
N. Galldiks, E. de Azambuja, A.S. Berghoff, P. Metellus, S. Peters, Y.-K. Hong, https://doi.org/10.1038/s41568-019-0205-x.
F. Winkler, D. Schadendorf, M. van den Bent, J. Seoane, R. Stahel, G. Minniti,

86

Lung Cancer 163 (2022) 87–95
Contents lists available at ScienceDirect

Lung Cancer

journal homepage: www.elsevier.com/locate/lungcan

Preoperative CT-based peritumoral and tumoral radiomic features
prediction for tumor spread through air spaces in clinical stage I
lung adenocarcinoma

Guoqing Liao a,d,1, Luyu Huang a,e,1, Shaowei Wu a,1, Peirong Zhang f, Daipeng Xie a,
Lintong Yao a, Zhengjie Zhang a, Su Yao g, Lyu Shanshan g, Siyun Wang h, Guangyi Wang i,
Lawrence Wing-Chi Chan c,*, Haiyu Zhou a,b,*

a Department of Thoracic Surgery, Guangdong Provincial People’s Hospital, Guangdong Academy of Medical Sciences, Guangzhou, China
b Department of Thoracic Surgery, Jiangxi Lung Cancer Institute, Jiangxi Cancer Hospital, Nanchang, China
c Department of Health Technology and Informatics, The Hong Kong Polytechnic University, Hong Kong, China
d Department of Thoracic Surgery, Cancer Hospital of Shantou University Medical College, Shantou, China
e Department of Surgery, Competence Center of Thoracic Surgery, Charit´e University Hospital Berlin, Berlin, Germany
f Department of Thoracic Surgery, Maoming People’s Hospital, Maoming, China
g Department of Pathology, Guangdong Provincial People’s Hospital, Guangdong Academy of Medical Sciences, Guangzhou, China
h Department of PET Center, Guangdong Provincial People’s Hospital, Guangdong Academy of Medical Sciences, Guangzhou, China
i Department of Radiology, Guangdong Provincial People’s Hospital, Guangdong Academy of Medical Sciences, Guangzhou, China

ARTICLE INFO ABSTRACT

Keywords: Objectives: This study aims to develop and evaluate preoperative CT-based peritumoral and tumoral radiomic
Nomograms features to predict tumor spread through air space (STAS) status in clinical stage I lung adenocarcinoma (LUAD).
Adenocarcinoma of Lung Materials and methods: From June 2018 to December 2019, a retrospective diagnostic investigation was done.
Algorithms Patients with pathologically confirmed STAS status (N = 256) were eventually enrolled. The development cohort
Tomography consisted of 191 patients (74.6%) chosen randomly in a 7:3 ratio, whereas the validation group consisted of 65
X-Ray Computed patients (25.4%). The performance of models was assessed using receiver operating characteristic analysis, ac­
curacy, sensitivity, specificity, negative predictive values, and positive predictive values.
Results: The STAS positive status was found in 85 (33.2%) of the 256 patients (female: 53.2%; median [IQR] age:
62.0, [53.0–79.0] years), while the STAS negative status was found in 171 patients (66.8%) (female:50.6%;
median [IQR] age: 62.0, [53.0–87.0] years). The combined TRS and PRS-15 mm model had an AUC of 0.854
(95% CI, 0.799–0.909) in the development cohort and 0.870 (95% CI, 0.781–0.958) in the validation cohort,
indicating that the tumor radiomic signature (TRS) model and different peritumoral radiomic signature (PRS)
models were used to build the optimal gross radiomic signature (GRS) model. The radiomic nomogram achieves
superior discriminatory performance than GRS and clinical and radiological signatures (CRS), with an AUC of
0.871 (95% CI, 0.820–0.922) in the development cohort and AUC of 0.869 (95% CI, 0.776–0.961) in the vali­
dation cohort. Based on the Akaike information criterion (AIC) and decision curve analysis (DCA), the radiomic
nomogram provided greater clinical predictive capacity than clinical or any radiomic signatures alone.
Conclusion: In conclusion, we discovered that peritumoral characteristics were substantially related to STAS
status. This study revealed the unit of radiomic signature and clinical signatures may have a better performance
in STAS status.

Abbreviations: AIC, Akaike information criterion; AUC, Area under the receiver operating characteristics curve; CI, Confidence interval; DCA, Decision curve
analysis; STAS, Spread through air spaces; GRS, Optimal Gross Radiomic Signature; TRS, Tumor Radiomic Signature; PRS, Peritumor Radiomic Signature; CRS,
Clinical and Radiomic Signature.

* Corresponding authors at: Department of Thoracic Surgery, Guangdong Provincial People’s Hospital, Guangdong Academy of Medical Sciences, Guangzhou,
China (H. Zhou).

E-mail addresses: [email protected] (L. Wing-Chi Chan), [email protected] (H. Zhou).
1 The authors contributed equally and should be considered co-first authors).

https://doi.org/10.1016/j.lungcan.2021.11.017

Received 20 June 2021; Received in revised form 30 October 2021; Accepted 25 November 2021

Available online 6 December 2021
0169-5002/© 2021 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license

(http://creativecommons.org/licenses/by-nc-nd/4.0/).

G. Liao et al. Lung Cancer 163 (2022) 87–95

1. Introduction tumoral and the peritumoral radiomic signatures and compare their
performance of prediction for STAS.
The World Health Organization (WHO) first described tumor spread
through air spaces (STAS) as a new pathological phenomenon in 2015, The purpose of this study was to develop a preoperative CT-based
characterized by micropapillary clusters, solid nests, or single cells radiomics utilizing a clinical noninvasive parameter model to predict
spreading within air spaces in the surrounding lung parenchyma, STAS in clinical stage I lung adenocarcinomas.
beyond the primary tumor’s edge [1–4].
2. Materials and methods
STAS has been discovered in 14.8 percent − 56.4 percent of patients
with lung adenocarcinoma (LUAD), and STAS-positive patients have 2.1. Study population
been shown to have lower recurrence-free survival (RFS) and overall
survival (OS) [5–7]. High-grade histological patterns, lymphovascular This retrospective investigation was approved by the academic ethics
invasion, and lymph node metastases have all been linked to LUAD with committees of Guangdong Provincial People Hospital (GDPH), and
STAS positive status [8–10]. informed patient consent was waived. TNM staging was assessed clini­
cally and pathologically using the 2017 8th TNM staging system (Union
The sublobar resection for clinical stage T1 lung peripheral adeno­ for International Cancer Control, 8th Edition) [19].
carcinomas is becoming more common around the world as a minimally
invasive procedure [11,12]. The STAS-positive in recent studies in­ From June 2018 to December 2019, 536 patients with pathologically
dicates that sublobar resection plus nodal sampling were not an confirmed invasive LUAD who accepted surgical resection at Guangdong
appropriate surgical procedure and may result in a poorer prognosis Provincial People’s Hospital were included in this retrospective diag­
[5,13,14]. Thus, preoperatively identifying STAS status in LUAD could nostic research. Following the exclusion criteria, 256 patients who met
facilitate optimum surgical resection choosing. the requirements were enrolled and randomly divided at a 7:3 propor­
tion. The development cohort consisted of 191 patients (74.6%),
However, most researchers have ignored the peritumoral microen­ whereas the validation group consisted of 65 patients (25.4%). Fig. 1
vironment and only assessed the primary tumor’s radiomic feature [15]. depicts the patient exclusion criteria and recruitment method.
Several studies have already demonstrated that peritumoral pulmonary
parenchyma can provide significant value for the clinical assessment of The inclusive criteria were as follow: (a) the maximum diameter of
tumor metastasis [16–18]. But there remains unclear that peritumoral tumor in preoperative CT images ≤ 3 cm; (b) CT performed before the
radiomic features are valuable tools for predicting STAS status. surgery less than one month; (c) postoperative pathologically confirmed
invasive LUAD; (d) No lymph nodule metastasis reported on preopera­
As a tumor invasion biological behavior, STAS may manifest itself in tive PET/CT or mediastinoscopy; Patients with multiple lesions were
the peritumoral image, and our study was attempted to evaluate the

Fig. 1. The patient recruitment process in the GDPH center. Abbreviation, GDPH center, Guangdong Provincial people’s Hospital; LUAD, Lung adenocarcinoma;
STAS, Spread through air space.

88

G. Liao et al. Lung Cancer 163 (2022) 87–95

excluded from our analysis. The patient receiving Neoadjuvant therapy and Communication System (PACS), and the Digital Imaging and
were also excluded. Communications in Medicine (DICOM) images were transferred to an
open-source software (3D Slicer, https://www. slicer. org/, version 4.
Baseline clinical information, including age, sex, smoking status, 10. 2) [21]. The standard CT protocol can be seen in eTable 1 in the
BMI, feature descriptors for tumor, pleural involvement, consolidation Supplementary materials.
tumor ratio (CTR), T stage, surgery type, histological grade, predomi­
nant type for adenocarcinoma, were derived from medical records. Two experienced radiologists S.Y.W(with 13 years of experience in
thoracic oncology), G.Y.W(with 15 years of experience in thoracic
2.2. Histological evaluation oncological), and a physician G.Q.L(with three years of cardiothoracic
research experience), manually segmented the tumors on each layer
The criteria of the International Association for the Study of Lung after using a semi-automatic segmentation based on the “level tracing”
Cancer/American Thoracic Society/European Respiratory Society and “smoothing” functions to delineate the tumors in three orthogonally
(IASLC/ATS/ERS) were used for LUAD pathological classification [20]. orientated planes. All tumor segmentation was reviewed in consensus by
a radiologist (G.Y.W.) and thoracic surgeon (H.Y.Z.), and any discrep­
Two expert pathologists (Y.S. and L.S.S.) blinded to the patients’ ancies were resolved by discussion among them.
clinical outcomes assessed all tumor slides at the same time. The WHO
[1] classified tumor STAS as tumor cells within air gaps in para carci­ The maximum distances between the lung nodule edge and STAS
noma normal alveolar spaces beyond the margin of the primary tumor, were reported to be 1.35 cm and 0.87 cm in a recent study [6]; in our
which are defined as alveolar spaces not filled with tumor mucin in study, the peritumoral region was captured by dilating the tumor mask
invasive mucinous adenocarcinoma (IMA). If controversies or discor­ 20 mm in three dimensions to obtain a lung parenchyma ring around the
dances occurred, a discussion was held before a decision was reached. tumor and dividing it into 5-mm rings [22]. The 3D-slicer software
According to the criterion, 85 patients were finally identified as STAS- automatically reconstructed the three-dimensional volumes of interest
positive, dividing the study population into two groups. (3D-VOIs) of the tumor (VOI-T), peritumor-5 mm, peritumor-10 mm,
peritumor-15 mm, and peritumor-20 mm (VOI-P5/10/15/20). The flow
2.3. Image review and feature extraction chart depicted the CT acquisition and tumor segmentation processes
(Fig. 2).
Preoperative CT images were downloaded from Picture Archiving
We used Radiomics, an open-source tool for standardizing the

Fig. 2. Overall radiomic workflow and pipeline in this study. (Step 1) CT image (transverse section) in a 64-year-old male patient with a 2.0 cm pulmonary nodule in
the right upper lung (dotted box) on contrast-enhanced CT and biopsy confirmed as lung adenocarcinoma. (Step 2) Five regions of interest (ROIs) were constructed
into volumes of interests (VOIs), and radiomic features were extracted from five VOIs. (Step 3) Radiomic features were selected by the LASSO algorithm and
constructed into a radiomic signature. (Step 4) Model construction, calibration, and discrimination. Abbreviation, PRS5, Peritumoral radiomic signature selected
from the 5 mm peritumoral area; PRS10, Peritumoral radiomic signature selected from the 10 mm peritumoral area; PRS15, Peritumoral radiomic signature selected
from the 15 mm peritumoral area; PRS20, Peritumoral radiomic signature selected from the 20 mm peritumoral area.

89

G. Liao et al. Lung Cancer 163 (2022) 87–95

extraction of radiomics data (https://github.com/Radiomics/pyra Table 1
diomics), to extract the radiomic feature from VOI-T/ P5/10/15/20. Characteristics baseline of patients in the total cohort.
Shape, first-order statistics, neighborhood gray-tone difference matrix
(NGTDM), gray-level size zone matrix (GLSZM), gray-level dependence Variables STAS (− ) STAS (+) p value
matrix (GLDM), neighborhood gray-tone difference matrix (NGTDM), Status Status 0.792
and gray-level co-occurrence matrix (GLCM) were among the 5034 0.422
quantitative 3D-VOI radiomic variables collected. To obtain a zero mean (N = 171) (N = 85)
and unit variance, all feature values were normalized. 0.07
Gender, No. (%) 80 (46.8) 42 (49.4) 0.92
2.4. Radiomics feature selection and model building and validation Male 91 (53.2) 43 (50.6) ＜0.001**
Female 0.001**
To assess STAS status, a radiomic signature was created using the 0.279
least absolute shrinkage and selection operator algorithm (LASSO) Age, (years) 60.7(10.7) 59.5(10.9) 0.568
classifier. The best parameter configuration was determined using Mean ± SD 62 [53,79] 62 [53,87] 0.002**
tenfold cross-validation. The radiomic signatures of the VOI-T/ P5/10/ Median (IQR) 0.001**
15/20 were then extracted. 0.814
Smoking status, No. (%) 136 (79.5) 62 (72.9) <0.001**
Both the tumor radiomic signature (TRS) and the peritumor radiomic Non-smokers 35 (20.5) 22 (27.1) <0.001**
signature (PRS) models were created. To integrate the tumor feature and Smoker
construct a gross image, the peritumoral region feature with the largest 0.07
area under the receiver operating characteristics curve (AUC) was BMI(Kg/m^2), No. (%) 10 (5.8) 4 (4.7) 0.416
chosen. ＜18.5 99 (57.9) 49 (57.6)
18.5–23.9 62 (36.3) 32 (37.6) 0.011*
2.5. Statistical analysis ＞24 0.007*

The significance of associations with the outcome of STAS status CEA, No. (%) 152 (89.4) 59 (69.4) 0.07
from the clinical parameters was determined using univariate logistic Normal 18 (10.6) 26 (30.6)
regression analysis. Statistical significance was determined by a two- Abnormal
sided p < 0.05. To create a clinical signature, variables having a p <
0.05 in univariable analysis were added into stepwise logistical regres­ Tumor size, (mm) 17.58(6.68) 20.51(6.06)
sion studies. Mean ± SD

The multivariable logistic regression analysis was utilized to discover Air bronchogram, No. (%) 110 (64.3) 48 (56.5)
independent prognostic factors of radiomic signature and clinical No 61 (35.7) 37 (43.5)
signature, and a radiomic nomogram was created. The validation cohort Yes
is used to test the radiomic signatures and nomogram.
Vacuous, No. (%) 109 (63.7) 58 (68.2)
The Hosmer–Lemeshow test was performed to evaluate the nomo­ No 62 (36.3) 27 (31.8)
gram’s calibration. To assess the discrimination performance of the Yes
models, the AUCs were calculated. In addition, the DeLong test was used 38 (22.2) 5 (5.9)
to compare multiple ROC curves statistically. For selecting the best Spicule, No. (%) 133 (77.8) 80 (94.1)
combinations of radiomic and clinical signatures, the Akaike informa­ No
tion criterion (AIC) was applied. We also employed decision curve Yes
analysis (DCA) to assess the nomogram’s clinical utility. The R platform
was used to run the glmnet packages and statistical analysis (R Phyllodes, No. (%) 65 (38.0) 14 (16.5)
version3.6.2). No 106 (62.0) 71 (83.5)
Yes
3. Result
Pleural Indentation, No. (%) 42 (24.6) 19 (22.4)
3.1. Patient baseline characteristics No 129 (75.4) 66 (77.6)
Yes
The imaging cases of 256 preoperative patients with clinical stage I
lung adenocarcinoma was obtained from China’s International Lung Boundary, No. (%) 105 (61.4) 30 (35.3)
Cancer Research Institute (Guangdong Provincial People’s Hospital), No 66 (38.6) 55 (64.7)
and the baseline characteristics of the patients in this investigation are Yes
listed in Table 1. The STAS positive status was found in 85 (33.2%) of the
256 patients (female: 53.2%; median [IQR] age: 62.0, [53.0–79.0] Density, No. (%) 10 (5.8) 1 (1.2)
years), while the STAS negative status was found in 171 patients pGGN 76 (44.4) 4 (4.7)
(66.8%) (female:50.6%; median [IQR] age: 62.0, [53.0–87.0] years). PSN 85 (49.7) 80 (94.1)
Solid
3.2. Development and validation of the clinical signature
CTR of the PSNs, No. (%) 45 (59.2) 0 (0.0)
The clinical-radiological characteristics show that there are signifi­ ＜0.5 31 (40.8) 4 (100.0)
cant differences in CEA status, tumor size, spicule, phyllodes, and ≥0.5
boundary (P = <0.001, 0.001, 0.002, 0.001, <0.001, respectively,
eTable 2 in Supplementary materials) between different STAS status and Tumor Location, No. (%) 42 (24.6) 24 (28.2)
other CT characteristics remained, including air bronchogram, vacuole, LUL 21 (12.3) 16(18.8)
pleural indentation, are not statistically significant in univariate LLL 50 (29.2) 22 (25.9)
analysis. RUL 19 (11.1) 5 (5.9)
RML 39 (22.8) 18 (21.2)
RLL

Surgery type, No. (%) 123 (71.9) 74 (87.1)
Lobectomy 48 (28.1) 11 (12.9)
Sublobar resection

Histological grade, No. (%) 22 (12.9) 2 (2.4)
Grade 1 134 (78.4) 47 (55.3)
Grade 2 15 (8.8) 36 (42.4)
Grade 3

Predominant type, No. (%) 24(14.0) 1(1.2)
Lepidic 134(78.4) 70(82.4)
Acinar 5(2.9) 6(7.1)
Solid 7(4.1) 7(8.2)
Papillary 1(0.6) 1(1.2)
MPP

(continued on next page)

90

G. Liao et al. Lung Cancer 163 (2022) 87–95

Table 1 (continued ) 3.3. Feature selection and acquisition of radiomic signatures

Variables STAS (− ) STAS (+) p value The LASSO algorithm was used to select the radiomic features for
Status Status five VOIs from the tumor and four peritumoral areas. Each group’s
selected features, including First-order, Shape, GLCM, GLSZM, GLRLM,
(N = 171) (N = 85) GLDM, and NGTDM, are presented in eTable 3.

T Stage, No. (%) 0.001** The LASSO algorithm picked 4105 radiomic features from five VOIs
T1a (TRS: 4 features for VOI-T, PRS-5 mm: 3 features for VOI-P5, PRS-10
T1b 31 (18.1) 3 (3.5) mm: 7 features for VOI-P10, PRS-15 mm: 16 features for VOI-P15, PRS-
T1c 79 (46.2) 37 (43.5) 20 mm: 5 features for VOI-P20) (Supplementary eFig. 2).
61 (35.7) 45 (52.9)
As shown in Supplementary eFig. 1, the TRS model’s AUC in the
Lymphatic Metastasis, No. ＜0.001** development cohort was 0.726 (95% CI, 0.649–0.803), whereas it was
(%) 0.796 in the validation cohort (95% CI, 0.686–0.906). In the develop­
N0 132 (88.9) 47 (57.7) ment set, each peritumoral model (PRS-5 mm, PRS-10 mm, PRS-15 mm,
N1 6 (3.5) 12 (14.1) PRS-20 mm) performed well, with AUC values of 0.703, 0.741, 0.845,
N2 13 (7.6) 24 (28.2) 0.736, respectively, and in the validation set, with AUC values of 0.709,
0.740, 0.831, 0.784, respectively (Table 2 and Supplementary eFig. 1).
Pleural Invasion, No. (%) ＜0.001**
No The most predictive signature was found to be the peritumoral region
Yes 156 (91.2) 56 (65.9) of 15 mm around the tumor (PRS-15 mm), which with the highest values
15 (8.8) 29 (34.1) of AUC (AUC = 0.845, 0.831) in the development and validation co­
horts. The radiomics signature of the TRS model was combined with the
ALK Status, No. (%) ＜0.001** radiomics signatures of several PRS models to create the GRS model
Wild type (Combined TRS and PRS-15 mm), and the rad-score was subsequently
Mutant 160 (93.6) 65 (76.5) calculated according to the formula (supplementary eTable 3). ROC
Missing 6 (3.6) 15 (17.6) analysis is used to assess each model’s performance, and the GRS has the
5 (3.5) 5 (5.9) greatest AUC in all cohorts (AUC = 0.854, 0.870). The radiomic score for
each patient was significantly different based on STAS status in two
EGFR Status, No. (%) ＜0.001** cohorts (P < 0.001 for Supplementary eFig. 3A/ Supplementary eFig.
Wild type 4A; P < 0.001 for Supplementary eFig. 3B/ Supplementary Fig. 4B).
Mutant 34 (19.9) 26 (30.6)
Missing 92 (53.8) 19 (22.4) Patients with STAS (+) status have considerably greater mean
45 (26.3) 40 (47.1) radiomic scores in both the development cohort and validation cohort
(0.658 and 1.21) than patients with STAS (–) status (− 2.296 and − 2.19,
SD, Standard deviation; IQR, Interquartile range; STAS, Spread through air respectively, eTable 4 in supplementary materials).
space; pGGN, Pure ground-glass nodule; PSN, Part solid nodule; CTR, Consoli­
dation tumor ratio; LLL, left lower lobe; LUL, Left upper lobe; RLL, Right lower 3.4. Development and validation of the nomogram
lobe; RML, Right middle lobe; RUL, Right upper lobe; BMI, Body mass index;
CEA, Carcinoembryonic antigen; Grade 1, lepidic predominant with no or <20% We created a radiomic nomogram that took CRS and GRS into ac­
of high-grade patterns; Grade 2, acinar or papillary predominant with no or count to establish a clinically usable predictive technique that may
<20% of high-grade patterns; Grade 3, any tumor with 20% or more of high- identify pathologically STAS status (Supplementary eFig. 5). After
grade patterns; ALK, Anaplastic lymphoma kinase; EGFR, Epidermal growth multivariate logistic regression analysis, the CRS (odds ratio (OR) =
factor receptor; *, Significant at p＜0.05; **, Significant at p＜0.005. 1.60; 95% CI, 1.03–2.59; P = 0.04) and GRS (odds ratio (OR) = 2.43;

After doing a univariate study, five of them (age, CEA status,
vacuous, spicule, and boundary) were chosen to develop the clinical and
radiomic signature (CRS) model using a stepwise logistic regression
model (Supplementary eTable 3).

According to the ROC curve analysis, the clinical signature showed
AUCs of 0.729 (95% CI, 0.653–0.805) and 0.665 (95% CI, 0.522–0.808)
(Supplementary eFig. 1A/1B, Table 2) in two cohorts, respectively.

Table 2
Performance evaluation of the models in the development cohort and validation cohort.

Cohort Signature Signature Performance

Sensitivity Specificity Accuracy PPV NPV AUC (95%CI)

Development Cohort CRS 0.65 0.70 0.68 0.51 0.80 0.73 (0.65–0.81)
TRS 0.77 0.65 0.69 0.52 0.86 0.73 (0.65–0.80)
PRS5 0.74 0.58 0.63 0.46 0.82 0.70 (0.63–0.78)
PRS10 0.58 0.81 0.73 0.59 0.80 0.74 (0.67–0.82)
PRS15 0.77 0.80 0.79 0.65 0.88 0.85 (0.79–0.90)
PRS20 0.69 0.70 0.70 0.52 0.83 0.74 (0.66–0.81)
GRS 0.76 0.81 0.80 0.66 0.88 0.85 (0.80–0.91)
Nomogram 0.82 0.80 0.81 0.66 0.90 0.87 (0.82–0.92)

Validation Cohort CRS 0.48 0.83 0.71 0.61 0.75 0.67 (0.52–0.81)
TRS 0.87 0.64 0.72 0.57 0.90 0.80 (0.69–0.91)
PRS5 0.83 0.52 0.63 0.49 0.85 0.71 (0.58–0.84)
PRS10 0.87 0.67 0.74 0.59 0.90 0.74 (0.62–0.86)
PRS15 0.83 0.71 0.75 0.61 0.88 0.83 (0.73–0.93)
PRS20 0.83 0.71 0.75 0.61 0.88 0.78 (0.67–0.89)
GRS 0.87 0.81 0.83 0.71 0.92 0.87 (0.78–0.96)
Nomogram 0.74 0.91 0.85 0.81 0.86 0.87 (0.78–0.96)

PPV, Positive predictive values; NPV, Negative predictive values; AUC, Area under the receiver operating characteristics curve; CI, Confidence interval; CRS, Clinical
and radiomic signature model; TRS, Tumoral radiomic signature model; PRS5, Peritumoral radiomic signature selected from the 5 mm peritumoral area; PRS10,
Peritumoral radiomic signature selected from the 10 mm peritumoral area; PRS15, Peritumoral radiomic signature selected from the 15 mm peritumoral area; PRS20,
Peritumoral radiomic signature selected from the 20 mm peritumoral area; GRS, Gross tumoral and peritumoral radiomic signature selected from the 15 mm peri­
tumoral area; Nomogram, Combine clinical and radiomic signature model.

91

G. Liao et al. Lung Cancer 163 (2022) 87–95

95% CI, 1.76–3.66; P < 0.001) were independent predictors for STAS Hosmer–Lemeshow test found that the nomogram had an excellent
status prediction match with a non-significant difference (P > 0.05) (Supplementary eFig.
6).
The radiomic nomogram achieves superior discriminatory perfor­
mance than GRS and CRS, with an AUC of 0.871 (95% CI, 0.820–0.922) The clinical utility of the five models was assessed using DCAs based
in the development cohort and AUC of 0.869 (95% CI, 0.776 to 0.961) in on the CRS, TRS, PRS-15 mm, GRS, and nomogram (Fig. 3). The pre­
the validation cohort (See Table 2). diction model may have a clinical net benefit, and the nomogram line
with the largest benefit throughout the whole range of threshold prob­
Among all prediction models in eTable 5, the nomogram model had abilities produces the best clinical results.
the lowest AIC value of 193.09, which was comparable to GRS (AIC =
199.61) and PRS-15 mm (AIC = 205.5). In comparison to the GRS and 4. Discussion
PRS-15 mm, there were no notable increases in AIC in the nomogram.
The discrepancies between the nomogram and TRS, CRS, PRS-5 mm, This study investigated whether preoperative CT-based radiomics
PRS-10 mm, and PRS-20 mm were statistically significant (P＜0.001, features extracted from tumor and peritumor (tissue) could predict STAS
respectively, see Supplementary eTable 5). status in clinical stage T1 stage LUAD. Our preliminary results showed
that the PRS-15 mm have the best predictive features. The unit combi­
3.5. Calibration and discrimination of the models nation of radiomic signature and clinical signature features with tumor
and peritumor radiomics signatures had a better performance in pre­
The nomogram calibration curve revealed strong correlations be­ dicting STAS.
tween nomogram prediction and actual observation in two cohorts. The

Fig. 3. Decision curve analysis for four models in the development and validation cohort. Decision curve analysis for the nomogram and signatures in the devel­
opment (A) and validation cohort(B).

92

G. Liao et al. Lung Cancer 163 (2022) 87–95

With the widespread usage of low-dose chest CT (LDCT) screening, Particularly features, which predicted STAS with statistical signifi­
early-stage lung cancers are increasingly being detected. With an cance in this study, include the original shape features: Least Axis
attempt to preserve the pulmonary function and reduce complication Length, voxel volume, and surface area. The inter-observer delineation
rates, sublobar resection (segmentectomy or wedge resection) was variability of these features has been appraised, and intraclass correla­
usually considered in early-stage LUAD less invasively than lobectomy tion (ICC) greater than 0.8 was found in different tumor sites, including
[23]. However, multiple studies found that patients with STAS-positive non-small cell lung cancer (NSCLC) [29]. However, the top five STAS
T1 stage LUAD who got restricted resection had a lower RFS and OS than status predictive features identified by the random forest (RF) model in
those who received lobectomy [13]. It should be noted that STAS in another study had one from shape, maximum 3D diameter [30].
stage 1A lung cancer patients treated with lobectomy was no longer a Although maximum 3D diameter was not found in this study, it is closely
significant risk factor for recurrence and overall survival [14]. Hence, related to the three shape features in describing the tumor morphology.
prediction of STAS is crucial to guide surgical strategies for early-stage The relevant but not the same features were identified from different
LUAD. STAS status was associated with pathologic and genetic charac­ radiomics studies because only a subset of redundant features could be
teristics. We found STAS-positive patients have higher pathological selected in the model building process, such as Lasso and random forest.
staging, which indicated STAS-positive patients are most micropapillary
predominant and solid predominant. Moreover, our study showed that The first-order entropy was found to be strongly linked with the
STAS was associated with wide-type EGFR. These pathologic and genetic overall survival of NSCLC in another investigation about the reproduc­
characteristics were identified by previous studies as well [24]. ibility of segmentation algorithms [31]. Even though the features were
extracted from PET images, the first-order entropy in computed to­
To date, many studies are devoted to preoperative STAS prediction. mography has been demonstrated to have predictive value in an imaging
A stratified and stepwise flowchart was developed based on frozen biomarker analysis of antiangiogenic chemotherapy [32].
section and preoperative PET-CT data, which predicted STAS-positive
patients in early-stage LUAD with an unsatisfactory sensitivity and At present, there are two radiomic models reported to predict the
specificity of 79.3% and 68.5%, respectively [25]. Walts et al. pitched STAS status in LUAD. With an AUC of 0.754 (Shenzhen People’s Hos­
into STAS evaluation in the frozen section, showing unacceptably low pital) [30] and 0.69 (Shanghai Pulmonary Hospital) [33], they both
sensitivity and negative predictive value, possibly due to difficulty integrate tumor radiomic characteristics and clinical factors. The AUC of
inappropriate normal lung tissue selection for diagnosis [26]. Therefore, our radiomic nomogram is greater than that of these two articles, which
intraoperative detection of STAS by frozen section is not recommended could be attributable to the peritumoral features that were investigated
from recent studies. in our study. Zhou et al. establish a radiomics nomogram of tumors and
peritumoral regions to predict STAS [34], however, the Hosmer-
Clinical parameters and CT characteristics of early-stage LUAD have Lemeshow test of their peritumoral model reveals that the calibration
been investigated for the prediction of STAS status. Previous research curves are not well fit. Our peritumoral model had a stronger prediction
has looked at how clinical factors and CT characteristics of lung cancer effect than the results of the previous studies, which could be due to the
can be used to predict STAS status. Jae Lee et al. explored correlated fact that our study’s 3D-VOI could provide more information for STAS
clinicopathologic characteristics in resected in LUAD, reporting that prediction.
STAS was reported to have significantly associated with tumor size,
smoking history, the presence of notches and percentage of solid com­ This research has several limitations. Firstly, our study is a retro­
ponents, vascular, spiculation [27]. In this study, we discovered that spective single-institutional study with potential selection bias. In the
tumor STAS was significantly linked with clinical parameters such as future, our GRS model still needs to be further verified in prospective,
age, CEA, vascular, spicule, and border. high-quality, and multi-omics research. Secondly, we used manual
segmentation of CT images in our study, which is susceptible to sub­
Furthermore, our study found that STAS-positive patients signifi­ jective influences, resulting in inconsistency in our findings. Fully con­
cantly correlated with larger tumor size and pleural invasion. Raj G. volutional networks will advocate for the automatic segmentation
Vaghjiani et al. suggested that STAS was a significant independent method in the future study. Furthermore, since our study included pa­
predictor of occult lymph node metastasis, which was in line with our tients from June 2018 to December 2019, long-term follow-ups are
result that lymph node metastasis was high with the rate of 42.3% (36/ required to evaluate the more accurate effect of STAS on prognosis.
85) [28]. Despite the limitations listed above, our study used radiomic analysis to
investigate the relationship between the peritumor region and STAS
In our study, the features of the tumoral region and peritumoral re­ status, resulting in a novel STAS predictive model with improved pre­
gion of 2 cm were used to develop the TRS model and PRS model. The dictive performance.
AUC of the TRS model was 0.73 in the development set and 0.80 in the
validation set. Referring to the predictive performance of the peritu­ 5. Conclusion
moral model, the PRS-15 mm model has the highest values of AUC,
which could be supported by evidence that the maximum distances In conclusion, we discovered that peritumoral characteristics were
between the lung nodule edge and STAS were 1.35 cm in the study substantially related to STAS status. This study revealed the unit of
cohort and 0.87 cm in the validation cohort [6]. radiomic signature and clinical signatures may have a better perfor­
mance in STAS status.
Our study indicates that the discriminative ability of peritumoral
signatures was better than the tumoral signatures. After combining the Funding
TRS model and PRS-15 model, the GRS model has achieved the AUC
0.85 in the development set and 0.87 in the validation set, which indi­ This study was funded by the Guangdong Province Medical Scientific
cated the peritumoral signature could predict the STAS, and the com­ Research Foundation [grant number B2018148], Science and Technol­
bination of tumor and peritumoral signature could achieve higher ogy Program of Guangzhou [grant number 201903010028,
predictive performance. However, no statistically significant differences 2017B030314026], Guangdong Provincial People’s Hospital Intermural
were found between the PRS-15 and GRS models. Program [grant number KJ012019447], National Natural Science
Foundation of China in 2020 [grant number KY012020523], Mandatory
The radiomic nomogram combining the CRS model and GRS model project of Guangdong Medical Science and Technology Research Fund
has achieved the AUC of 0.87. A total of 18 signature was admitted, [grant number C2020107], Health and Medical Research Funds [grant
which consisted of 2 clinical parameters and 16 radiomics signatures. number HMRF 02131026, HMRF 16172561].
The radiomics nomogram model of STAS outperformed the clinical,
tumor, and all peritumor models in terms of prediction efficacy. The GRS
and radiomics nomogram models, on the other hand, showed no sig­
nificant changes in both groups.

93

G. Liao et al. Lung Cancer 163 (2022) 87–95

CRediT authorship contribution statement [10] M.A. Lee, J. Kang, H.Y. Lee, W. Kim, I. Shon, N.Y. Hwang, et al., Spread through air
spaces (STAS) in invasive mucinous adenocarcinoma of the lung: Incidence,
Guoqing Liao: Conceptualization, Data curation, Formal analysis, prognostic impact, and prediction based on clinicoradiologic factors, Thorac
Investigation, Writing – review & editing. Luyu Huang: Conceptuali­ Cancer. 11 (11) (2020) 3145–3154, https://doi.org/10.1111/1759-7714.13632.
zation, Methodology, Project administration, Writing – original draft,
Writing – review & editing. Shaowei Wu: Data curation, Software, [11] M. Sagawa, H. Oizumi, H. Suzuki, H. Uramoto, K. Usuda, A. Sakurada, et al.,
Investigation, Writing – original draft. Peirong Zhang: Data curation, A prospective 5-year follow-up study after limited resection for lung cancer with
Funding acquisition. Daipeng Xie: Data curation, Formal analysis, ground-glass opacity, Eur. J. Cardiothorac. Surg. 53 (4) (2018) 849–856, https://
Methodology, Software. Lintong Yao: Software, Investigation, Writing doi.org/10.1093/ejcts/ezx418.
– original draft. Zhengjie Zhang: Investigation. Su Yao: Methodology,
Software. Lyu Shanshan: Methodology, Software. Siyun Wang: [12] C. Cao, D. Chandrakumar, S. Gupta, T.D. Yan, D.H. Tian, Could less be more?-A
Methodology, Software. Guangyi Wang: Methodology, Software. systematic review and meta-analysis of sublobar resections versus lobectomy for
Lawrence Wing-Chi Chan: Funding acquisition, Methodology, Vali­ non-small cell lung cancer according to patient selection, Lung Cancer. 89 (2)
dation, Visualization. Haiyu Zhou: Conceptualization, Funding acqui­ (2015) 121–132, https://doi.org/10.1016/j.lungcan.2015.05.010.
sition, Project administration, Resources, Supervision.
[13] T. Eguchi, K. Kameda, S. Lu, M.J. Bott, K.S. Tan, J. Montecalvo, et al., Lobectomy Is
Declarations of Competing Interest Associated with Better Outcomes than Sublobar Resection in Spread through Air
Spaces (STAS)-Positive T1 Lung Adenocarcinoma: A Propensity Score-Matched
The authors of this manuscript declare no relationships with any Analysis, J. Thoracic Oncol. 14 (1) (2019) 87–98, https://doi.org/10.1016/j.
companies, whose products or services may be related to the subject jtho.2018.09.005.
matter of the article. The funding organizations had no role in the design
and conduct of the study; collection, management, analysis, and inter­ [14] Y. Ren, H. Xie, C. Dai, Y. She, H. Su, D. Xie, H. Zheng, L. Zhang, G. Jiang, C. Wu,
pretation of data; preparation, review, or approval of the manuscript; C. Chen, Prognostic Impact of Tumor Spread Through Air Spaces in Sublobar
and decision to submit the manuscript for publication. Resection for 1A Lung Adenocarcinoma Patients, Ann. Surg. Oncol. 26 (6) (2019)
1901–1908, https://doi.org/10.1245/s10434-019-07296-w.
Acknowledgments
[15] Y. Huang, Z. Liu, L. He, X. Chen, D. Pan, Z. Ma, C. Liang, J. Tian, C. Liang,
We thank all the authors for their contributions to this manuscript. Radiomics Signature: A Potential Biomarker for the Prediction of Disease-Free
Survival in Early-Stage (I or II) Non-Small Cell Lung Cancer, Radiology 281 (3)
Appendix A. Supplementary data (2016) 947–957, https://doi.org/10.1148/radiol.2016152234.

Supplementary data to this article can be found online at https://doi. [16] X. Wang, X. Zhao, Q. Li, W. Xia, Z. Peng, R. Zhang, Q. Li, J. Jian, W. Wang, Y. Tang,
org/10.1016/j.lungcan.2021.11.017. S. Liu, X. Gao, Can peritumoral radiomics increase the efficiency of the prediction
for lymph node metastasis in clinical stage T1 lung adenocarcinoma on CT? Eur.
References Radiol. 29 (11) (2019) 6049–6058, https://doi.org/10.1007/s00330-019-06084-0.

[1] W.D. Travis, E. Brambilla, A.G. Nicholson, Y. Yatabe, J.H.M. Austin, M.B. Beasley, [17] N. Beig, M. Khorrami, M. Alilou, P. Prasanna, N. Braman, M. Orooji, S. Rakshit,
L.R. Chirieac, S. Dacic, E. Duhig, D.B. Flieder, K. Geisinger, F.R. Hirsch, K. Bera, P. Rajiah, J. Ginsberg, C. Donatelli, R. Thawani, M. Yang, F. Jacono,
Y. Ishikawa, K.M. Kerr, M. Noguchi, G. Pelosi, C.A. Powell, M.S. Tsao, I. Wistuba, P. Tiwari, V. Velcheti, R. Gilkeson, P. Linden, A. Madabhushi, Perinodular and
The 2015 World Health Organization Classification of Lung Tumors: Impact of intranodular radiomic features on lung CT images distinguish adenocarcinomas
Genetic, Clinical and Radiologic Advances Since the 2004 Classification, from granulomas, Radiology 290 (3) (2019) 783–792, https://doi.org/10.1148/
J. Thoracic Oncol. 10 (9) (2015) 1243–1260, https://doi.org/10.1097/ radiol.2018180910.
JTO.0000000000000630.
[18] P. Prasanna, J. Patel, S. Partovi, A. Madabhushi, P. Tiwari, Radiomic features from
[2] A. Warth, M.B. Beasley, M. Mino-Kenudson, Breaking New Ground: The Evolving the peritumoral brain parenchyma on treatment-naive multi-parametric MR
Concept of Spread through Air Spaces (STAS), J. Thoracic Oncol. 12 (2) (2017) imaging predict long versus short-term survival in glioblastoma multiforme:
176–178, https://doi.org/10.1016/j.jtho.2016.10.020. preliminary findings, Eur. Radiol. 27 (10) (2017) 4188–4197, https://doi.org/
10.1007/s00330-016-4637-3.
[3] D. Chen, Y. Mao, J. Wen, Y. She, E. Zhu, F. Zhu, Y. Zhang, M. Fan, C. Chen, Y. Chen,
Tumor spread through air spaces in non-small cell lung cancer: a systematic review [19] F.C. Detterbeck, D.J. Boffa, A.W. Kim, L.T. Tanoue, The eighth edition lung cancer
and meta-analysis, Ann. Thoracic Surg. 108 (3) (2019) 945–954, https://doi.org/ stage classification, Chest 151 (1) (2017) 193–203, https://doi.org/10.1016/j.
10.1016/j.athoracsur.2019.02.045. chest.2016.10.010.

[4] A.R. Shih, M. Mino-Kenudson, Updates on spread through air spaces (STAS) in lung [20] W.D. Travis, E. Brambilla, M. Noguchi, A.G. Nicholson, K.R. Geisinger, Y. Yatabe,
cancer, Histopathology 77 (2) (2020) 173–180, https://doi.org/10.1111/ D.G. Beer, C.A. Powell, G.J. Riely, P.E. Van Schil, K. Garg, J.H.M. Austin,
his.14062. H. Asamura, V.W. Rusch, F.R. Hirsch, G. Scagliotti, T. Mitsudomi, R.M. Huber,
Y. Ishikawa, J. Jett, M. Sanchez-Cespedes, J.-P. Sculier, T. Takahashi, M. Tsuboi,
[5] K. Kadota, J.-I. Nitadori, C.S. Sima, H. Ujiie, N.P. Rizk, D.R. Jones, P.S. Adusumilli, J. Vansteenkiste, I. Wistuba, P.-C. Yang, D. Aberle, C. Brambilla, D. Flieder,
W.D. Travis, Tumor Spread through Air Spaces is an Important Pattern of Invasion W. Franklin, A. Gazdar, M. Gould, P. Hasleton, D. Henderson, B. Johnson,
and Impacts the Frequency and Location of Recurrences after Limited Resection for D. Johnson, K. Kerr, K. Kuriyama, J.S. Lee, V.A. Miller, I. Petersen, V. Roggli,
Small Stage I Lung Adenocarcinomas, J. Thoracic Oncol. 10 (5) (2015) 806–814, R. Rosell, N. Saijo, E. Thunnissen, M. Tsao, D. Yankelewitz, International
https://doi.org/10.1097/JTO.0000000000000486. association for the study of lung cancer/american thoracic society/european
respiratory society international multidisciplinary classification of lung
[6] C. Dai, H. Xie, H. Su, Y. She, E. Zhu, Z. Fan, et al., tumorTumor Spread through Air adenocarcinoma, J. Thoracic Oncol. 6 (2) (2011) 244–285, https://doi.org/
Spaces Affects the Recurrence and Overall Survival in Patients with Lung 10.1097/JTO.0b013e318206a221.
Adenocarcinoma >2 to 3 cm, J. Thoracic Oncol. 12 (7) (2017) 1052–1060, https://
doi.org/10.1016/j.jtho.2017.03.020. [21] J.J.M. van Griethuysen, A. Fedorov, C. Parmar, A. Hosny, N. Aucoin, V. Narayan,
R.G.H. Beets-Tan, J.-C. Fillion-Robin, S. Pieper, H.J.W.L. Aerts, Computational
[7] G. Qu, Y. Shi, Progress on the Study of Tumor Spread Through Air Spaces in the radiomics system to decode the radiographic phenotype, Cancer Res. 77 (21)
Clinicopathological Characteristics of Lung Adenocarcinoma and Its Influence on (2017) e104–e107, https://doi.org/10.1158/0008-5472.CAN-17-0339.
the Surgical Treatment and Prognosis of Lung Cancer, Zhongguo Fei Ai Za Zhi. 22
(6) (2019) 363–368, https://doi.org/10.3779/j.issn.1009-3419.2019.06.06. [22] M. Khorrami, P. Prasanna, A. Gupta, P. Patil, P.D. Velu, R. Thawani, G. Corredor,
M. Alilou, K. Bera, P. Fu, M. Feldman, V. Velcheti, A. Madabhushi, Changes in CT
[8] Y. Terada, T. Takahashi, S. Morita, K. Kashiwabara, K. Nagayama, J.-I. Nitadori, M. radiomic features associated with lymphocyte distribution predict overall survival
Anraku, M. Sato, A. Shinozaki-Ushiku, J. Nakajima, Spread through air spaces is an and response to immunotherapy in non-small cell lung cancer, Cancer Immunol.
independent predictor of recurrence in stage III (N2) lung adenocarcinoma 29(3), Res. 8 (1) (2020) 108–119, https://doi.org/10.1158/2326-6066.CIR-19-0476.
2019 442–448 doi:10.1093/icvts/ivz116.
[23] A.D. Sihoe, P. Van Schil, Non-small cell lung cancer: when to offer sublobar
[9] Y. Liu, D. Chen, X. Qiu, S. Duan, Y. Zhang, F. Li, C. Chen, Y. Zhou, Y. Chen, resection, Lung Cancer. 86 (2) (2014) 115–120, https://doi.org/10.1016/j.
Relationship between MTA1 and spread through air space and their joint influence lungcan.2014.09.004.
on prognosis of patients with stage I-III lung adenocarcinoma, Lung Cancer. 124
(2018) 211–218, https://doi.org/10.1016/j.lungcan.2018.07.040. [24] C. Deng, Y. Zhang, Z. Ma, F. Fu, L. Deng, Y. Li, H. Chen, Prognostic value of
epidermal growth factor receptor gene mutation in resected lung adenocarcinoma,
J. Thorac. Cardiovasc. Surg. 162 (3) (2021) 664–674.e7, https://doi.org/10.1016/
j.jtcvs.2020.05.099.

[25] J.W. Suh, Y.H. Jeong, A. Cho, D.J. Kim, K.Y. Chung, H.S. Shim, C.Y. Lee, Stepwise
flowchart for decision making on sublobar resection through the estimation of
spread through air space in early stage lung cancer1, Lung Cancer 142 (2020)
28–33, https://doi.org/10.1016/j.lungcan.2020.02.001.

[26] A.E. Walts, A.M. Marchevsky, Current evidence does not warrant frozen section
evaluation for the presence of tumor spread through alveolar spaces, Arch. Pathol.
Lab. Med. 142 (1) (2018) 59–63, https://doi.org/10.5858/arpa.2016-0635-oa.

[27] J.S. Lee, E.K. Kim, M. Kim, H.S. Shim, Genetic and clinicopathologic characteristics
of lung adenocarcinoma with tumortumor spread through air spaces, Lung Cancer.
123 (2018) 121–126, https://doi.org/10.1016/j.lungcan.2018.07.020.

[28] R.G. Vaghjiani, Y. Takahashi, T. Eguchi, S. Lu, K. Kameda, Z. Tano, J. Dozier, K.
S. Tan, D.R. Jones, W.D. Travis, P.S. Adusumilli, Tumor spread through air spaces
is a predictor of occult lymph node metastasis in clinical stage IA lung
adenocarcinoma, J. Thoracic Oncol. 15 (5) (2020) 792–802, https://doi.org/
10.1016/j.jtho.2020.01.008.

94

G. Liao et al. Lung Cancer 163 (2022) 87–95

[29] M. Pavic, M. Bogowicz, X. Würms, S. Glatz, T. Finazzi, O. Riesterer, J. Roesch, [32] K. Hayano, N.M. Kulkarni, D.G. Duda, R.S. Heist, D.V. Sahani, Exploration of
L. Rudofsky, M. Friess, P. Veit-Haibach, M. Huellner, I. Opitz, W. Weder, imaging biomarkers for predicting survival of patients with advanced non-small
T. Frauenfelder, M. Guckenberger, S. Tanadini-Lang, Influence of inter-observer cell lung cancer treated with antiangiogenic chemotherapy, AJR Am. J.
delineation variability on radiomics stability in different tumor sites, Acta Oncol. Roentgenol. 206 (5) (2016) 987–993, https://doi.org/10.2214/ajr.15.15528.
57 (8) (2018) 1070–1074, https://doi.org/10.1080/0284186X.2018.1445283.
[33] D. Chen, Y. She, T. Wang, H. Xie, J. Li, G. Jiang, et al., Radiomics-based prediction
[30] C. Jiang, Y. Luo, J. Yuan, S. You, Z. Chen, M. Wu, G. Wang, J. Gong, CT-based for tumour spread through air spaces in stage I lung adenocarcinoma using
radiomics and machine learning to predict spread through air space in lung machine learning, Eur. J. Cardiothorac. Surg. 58 (1) (2020) 51–58, https://doi.
adenocarcinoma, Eur. Radiol. 30 (7) (2020) 4050–4057, https://doi.org/10.1007/ org/10.1093/ejcts/ezaa011.
s00330-020-06694-z.
[34] Y. Zhuo, M. Feng, S. Yang, L. Zhou, D.i. Ge, S. Lu, L. Liu, F. Shan, Z. Zhang,
[31] U. Bashir, G. Azad, M.M. Siddique, S. Dhillon, N. Patel, P. Bassett, D. Landau, Radiomics nomograms of tumors and peritumoral regions for the preoperative
V. Goh, G. Cook, The effects of segmentation algorithms on the measurement of prediction of spread through air spaces in lung adenocarcinoma, Transl. Oncol. 13
(18)F-FDG PET texture parameters in non-small cell lung cancer, EJNMMI Res. 7 (10) (2020) 100820, https://doi.org/10.1016/j.tranon.2020.100820.
(1) (2017), https://doi.org/10.1186/s13550-017-0310-3.

95