Clinical significance of circulating-tumour DNA analysis by metastatic sites in pancreatic cancer

Abstract

Background

Liquid biopsy is an alternative to tissue specimens for tumour genotyping. However, the frequency of genomic alterations with low circulating-tumour DNA (ctDNA) shedding is shown in pancreatic ductal adenocarcinoma (PDAC). We, therefore, investigated the prevalence of KRAS mutations and ctDNA fraction by the metastatic site in patients with PDAC.

Methods

This study enrolled previously treated PDAC patients from a plasma genomic profiling study; ctDNA analysis was performed using Guardant360 at disease progression before initiating subsequent treatment.

Results

In 512 patients with PDAC, KRAS mutations were detected in 57%. The frequency of KRAS mutation in ctDNA differed depending on the metastatic organ; among patients with single-organ metastasis (n = 296), KRAS mutation detection rate was significantly higher in patients with metastasis to the liver (78%). In addition, the median maximum variant allele frequency (VAF) was higher with metastasis to the liver (1.9%) than with metastasis to the lungs, lymph nodes, peritoneum or with locally advanced disease (0.2%, 0.4%, 0.2% and 0.3%, respectively).

Conclusion

The prevalence of KRAS mutations and maximum VAF were higher in patients with metastasis to the liver than in those with metastasis to other sites. This study indicated the clinical utility of ctDNA analysis, especially in PDAC with liver metastases.

Introduction

Activating KRAS mutations represent an early event during pancreatic tumorigenesis crucial for cancer initiation and progression in about 90% of pancreatic ductal adenocarcinoma (PDAC) [1, 2]. In addition to KRAS, multiple other driver mutations have been identified, some of which are associated with approved targeted therapies. Indeed, in the most recent National Comprehensive Cancer Network (NCCN) Clinical Practice Guidelines for the treatment of PDAC mentions ALK, NRG1, NTRK1-3 and ROS1 fusions; mutations in BRAF, BRCA1-2, ERBB2, KRAS and PALB2; and microsatellite instability (MSI) as molecular driver alterations relevant to targeted therapy [3]. It is particularly important to accurately measure KRAS mutations and eliminate those with KRAS mutations in PDAC since these druggable genes are found in KRAS wild-type PDAC.

Due to the increasing importance of broad genotyping in PDAC and other tumour types, next-generation sequencing (NGS)-based gene panel tests have been widely implemented, most commonly as tissue-based NGS tests, such as FoundationOne^®CDx [4], MSK-IMPACT [5] and the OncoGuide^TMNCC Oncopanel system [6]. However, due to the clinical and technical limitations associated with tissue-based testing, plasma-based liquid biopsy NGS tests, such as Guardant360 CDx and FoundationOne^® Liquid CDx, have been recently approved by the Food and Drug Administration (FDA) and public agencies in other countries as comprehensive genome profiling (CGP) tests capable of identifying circulating-tumour DNA (ctDNA) in blood. Such tests allow for the rapid return of genomic results to patients for whom adequate tissue is not available for more traditional genotyping tests. This problem is particularly acute in advanced PDAC, where many patients are diagnosed by endoscopic ultrasound-guided fine needle aspiration (EUS-FNA), which often yields insufficient tissue for gene panel testing, and thus is an indication where liquid biopsy may be particularly helpful [7].

The GOZILA study was launched in Japan in January 2018 with the aim of studying the effectiveness of liquid biopsy as an alternative to tissue genotyping and has enrolled nearly 6000 patients with gastrointestinal (GI) cancers to date. In a study of the first 1687 patients, ctDNA genotyping significantly improved the proportion of patients enrolled in a matched clinical trial without compromising the treatment efficacy compared with tissue genotyping [8]. However, the prevalence of genomic alterations in ctDNA from patients with PDAC in this cohort was lower than that previously reported using tissue-based testing [9]. Given the importance of accurate measurement of KRAS mutations to find therapeutic targets in PDAC, we investigated the relationship between the prevalence of KRAS mutations in blood and ctDNA fraction, clinicopathologic correlates, including metastatic site, and clinical outcome in PDAC patients enrolled in the GOZILA study.

Methods

GOZILA study design and patient selection

The GOZILA study is a nationwide plasma genomic profiling study conducted in Japan using the SCRUM-Japan GI-SCREEN genome profiling platform. The study, which involves 31 institutions, aims to identify patients with GI cancers who might benefit from targeted therapies [8]. The key inclusion criteria are (i) pathologically confirmed unresectable or metastatic GI cancer, (ii) age ≥20 years, and (iii) a life expectancy of at least 12 weeks. To avoid the suppression of ctDNA shedding because of chemotherapy, patients are included only if they show disease progression during chemotherapy and have not started subsequent therapy at the time of blood sampling. Eligible patients provide written informed consent. For pancreatic cancer (PC) cohort in the GOZILA study, the study was conducted from January 2018 to December 2019. Our present study included only the patients with pathologically diagnosed PDAC from the PC cohort in the GOZILA study.

ctDNA analysis by Guardant360

For patients enrolled in the GOZILA study, the NGS analysis of ctDNA was performed using Guardant360 at Guardant Health, a CLIA-certified, CAP-accredited, New York State Department of Health-approved laboratory. Guardant360 detects single-nucleotide variants (SNVs), indels, fusions, and copy number alterations in 74 genes with a reportable range of ≥0.04%, ≥0.02%, ≥0.04% and ≥2.12 copies, respectively. For the patients enrolled in the GOZILA study, 2 × 10-mL whole-blood samples were collected using Streck Cell-Free DNA blood collection tubes (BCT; Streck, Inc) and sent to Guardant Health. Five to thirty nanograms of cell-free DNA (cfDNA) isolated from each plasma sample was labelled with nonredundant oligonucleotides (“molecular barcoding”), enriched using targeted hybridisation capture, and sequenced on the Illumina NextSeq 550 platform (Illumina, Inc.). Base call files generated by Illumina’s RTA software version 2.12 were demultiplexed using bcl2fastq version 2.19 and processed as previously described [10]. Somatic cfDNA alterations were identified using a proprietary bioinformatics pipeline. To estimate cfDNA clonality for somatic SNVs, indels and fusions, the relative clonality was initially defined as the altered variant allele frequency (VAF)/maximum somatic VAF in the sample. In the GOZILA study, we defined ‘subclonal’ as a clonality of ≤0.3 and set the cut-off value for maximum VAF as 0.6%.

Clinical data and genotyping results

The clinicopathological information and efficacy data regarding the chemotherapy of patients enrolled in the GOZILA study were collected using an electronic data capture system. These clinical data and genotyping results were stored in a clinical-grade database and were used for an integrated clinicogenomic analysis. For the survival analysis according to genomic status of KRAS mutation, anonymized efficacy data from patients treated in clinical practice were used.

Statistical analysis

The difference of maximum VAF in the total ctDNA and maximum VAF of KRAS mutation detected in ctDNA, between clinicopathological factors such as metastatic site, were assessed using the Kruskal–Wallis test. When calculating maximum VAF, we excluded the germline pathogenic variants identified by Guardant360. Overall survival (OS) was calculated using the interval from the day of blood draw for ctDNA analysis during first-line chemotherapy or second-line chemotherapy until the day of death or the last visit. Kaplan–Meier curves were constructed, and the statistical significance was determined using a log-rank test. A P value of <0.05 was considered significant. JMP software (JMP^® Pro 14.2.0; SAS Institute Inc.) was used to perform the statistical analyses.

Results

From January 2018 to December 2019, a total of 540 patients with PC were enrolled in the GOZILA study. Among those enrolled as advanced PC, 26 patients were excluded for the following reasons: duplicate registration (n = 2), ineligible for enrolment (n = 2), sample not submitted (n = 2), acinar cell carcinoma (n = 5), solid pseudopapillary neoplasm (n = 1), neuroendocrine neoplasm (n = 3), pathologically unknown (n = 9) and others (n = 2). For the remaining 514 patients that were pathologically diagnosed with PDAC, blood samples were collected at the time of disease progression during chemotherapy and prior to the start of subsequent therapy. Two patients with insufficient sample quality and/or quantity were excluded. Finally, a total of 512 patients with genomic and clinical data were included in this analysis (Supplementary Fig. 1). Table 1 shows the patient background: the median age was 65 years; 58% were male; 11% had a locally advanced tumour status, 60% had distant metastasis, and 29% had postoperative recurrence; and metastases were present in the liver (54%), lungs (21%), lymph nodes (22%), peritoneum (30%), and others (10%). The first-line chemotherapy was performed with gemcitabine plus nab-paclitaxel (65%), FOLFIRINOX (19%), gemcitabine (6%) and S-1 (3%). In patients with postoperative recurrence, 23% received S-1 as adjuvant chemotherapy.

Table 1.

Patient characteristics.

n = 512, n (%)
Age (median, [range])		65, [24–85]
Sex	Male	296 (58)
Tumour status	Locally advanced	57 (11)
	Metastatic	306 (60)
	Recurrence	149 (29)
Metastatic sites*	Liver	275 (54)
	Lung	110 (21)
	Lymph nodes	114 (22)
	Peritoneum	156 (30)
	Others**	53 (10)
Metastatic sites to a single site	Liver	134 (26)
	Lung	37 (7)
	Lymph nodes	22 (4)
	Peritoneum	57 (11)
The number of treatment line*** (median, [range])		1, [1–4]
The first-line chemotherapy	GEM + nab-PTX	334 (65)
	FOLFIRINOX	96 (19)
	GEM	30 (6)
	S-1	14 (3)
	Others	38 (7)

GEM + nab-PTX gemcitabine plus nab-paclitaxel, FOLFIRINOX 5fluorouracil, irinotecan and oxaliplatin, GEM gemcitabine.

*Overlapping distribution.

**Metastatic sites of others include bone, brain, muscle, ovary, tone, spleen, adrenal gland and local recurrence.

***The available data based on the population of 244 patients.

Figure 1a shows the overall frequency of genomic alterations identified with ctDNA analysis in patients with PDAC. KRAS mutations in ctDNA were detected in 292 patients (57%). Of 43% with KRAS mutations not detected, 145 patients (28%) had other genomic alterations and 78 patients (15%) had no tumour-related alterations detected (Fig. 1b). The frequency of KRAS mutations in ctDNA was analysed according to the number and location of metastatic sites. The detection rate was 42% among patients with no metastatic sites (locally advanced), 53% among patients with a single metastatic site, and 66% among patients with 2 or more metastatic sites (Table 2). In addition, the detection rates among patients with a single or more than one metastatic site were 75% for the liver, 46% for lungs, 60% for lymph nodes and 50% for the peritoneum. The detection rates among patients with two metastatic sites were also shown; the rate with liver metastasis were 71–78% and others without liver metastasis were 33–38% (Table 3).

Fig. 1. Genomic alterations identified with ctDNA analysis in patients with PDAC.

Overall frequency of genomic alterations (n = 512) (a). The distribution of KRAS mutation, RAS wild-type and no alterations (b). ctDNA, circulating DNA, PDAC pancreatic ductal adenocarcinoma, SNV single-nucleotide variant, HRD homologous recombination deficiency.

Table 2.

The overall frequencies of KRAS mutation detected in ctDNA analysis according to the number of metastatic sites.

	The number of metastatic sites
	Locally advanced	1	≥2
n*	57	259	197
KRAS mt detected	24	138	131
(%)	(42)	(53)	(66)

KRAS mt KRAS mutation.

*Each population includes the cases with no tumour-related somatic alterations detected.

Table 3.

The overall frequencies of KRAS mutation detected in ctDNA analysis according to the location of metastatic sites.

	The location of metastatic sites (single or more than one sites).				The location of metastatic sites (two sites)
	Liver	Lung	Lymph nodes	Peritoneum	Liver +lung	Liver+ lymph nodes	Liver+ peritoneum	Lung + lymph nodes	Lung+ peritoneum	Lymph nodes + peritoneum
n*	275	110	114	156	20	31	36	8	6	13
KRAS mt detected	207	51	68	78	15	22	28	3	2	5
(%)	(75)	(46)	(60)	(50)	(75)	(71)	(78)	(38)	(33)	(38)

KRAS mt KRAS mutation.

*Each population includes cases with no tumour-related somatic alterations detected.

Figure 2 shows the ctDNA fraction, as measured by maximum VAF, and KRAS mutation detection rate by metastatic site among patients with single-organ metastasis. The median maximum VAF was higher in patients with metastasis to the liver (1.9%) than in those with metastasis to the lungs, lymph nodes, peritoneum or with locally advanced disease (0.2%, 0.4%, 0.2% and 0.3%, respectively, P < 0.001). Similarly, KRAS mutation detection rate was significantly higher in the liver (78%) than in the lungs, lymph nodes, peritoneum or with locally advanced disease (14%, 38%, 27% and 43%, respectively, P < 0.01). KRAS mutation detection rates by ctDNA fraction category parallel this trend (Supplementary Table 1). In general, the rates of patients with tumour fractions inadequate for genotyping (maximum VAF < 0.6%) [8] were lower with metastasis to the liver (28%) than with metastasis to the lungs, lymph nodes, peritoneum or with locally advanced disease (83%, 59%, 65% and 63%, respectively).

Fig. 2. The maximum VAF (%) and frequencies of the detection of KRAS mutation in ctDNA in patients with metastasis to a single site (n = 296).

Left and right box plots represent the maximum VAF in all patients and patients with KRAS mutation in ctDNA according to each metastatic site, respectively. Max VAF maximum variant allele frequency, KRAS mt KRAS mutation. *Max VAF calculated in population excluded germline mutation. **In GOZILA study, we define subclonal with the clonality as ≤0.3 and set the cut-off value of max VAF as 0.6%. ***Kruskal–Wallis test.

To determine whether the observed differences in ctDNA fraction were associated with clinical outcome, the OS for the patient subgroups defined by ctDNA above were compared. The association between OS and the detection rate of KRAS mutations in ctDNA was analysed with 180 patients available in this analysis. The result shows the prognostic significance of KRAS mutation in ctDNA during first-line chemotherapy (KRAS mutation-not-detected group: 8.9 months [95% C.I.: 6.6–13.2] vs. KRAS mutation-detected group: 7.0 months [95% C.I.: 5.1–8.2], hazard ratio (HR): 0.58, P = 0.0264) and during second-line chemotherapy (KRAS mutation-not-detected group: 8.7 months [95% C.I.: 7.0–12.6] vs. KRAS mutation-detected group: 4.9 months [95% C.I.: 3.6–6.2], HR = 0.27, P < 0.0001). (Supplementary Fig. 2). The OS in patients with liver metastasis stratified with the detection of KRAS mutation during first-line chemotherapy is shown in Supplementary Fig. 3a (KRAS mutation-not-detected group: 8.3 months [95% C.I.: 5.4–10.7] vs. KRAS mutation-detected group: 6.7 months [95% C.I.: 5.1–8.2], HR: 0.75, P = 0.3710) and with other than liver metastasis is also shown in Supplementary Figure 3b (KRAS mutation-not-detected group: 11.4 months [95% C.I.: 7.5–14.5] vs. KRAS mutation-detected group: 7.6 months [95% C.I.: 3.9–9.9], HR: 0.50, P = 0.1024). Supplementary Fig. 4 provides the subgroup analysis for OS during first-line chemotherapy; the alteration of ctDNA favoured KRAS mutation-not-detected group across the majority of prespecified subgroups.

Discussion

The treatment of patients with PDAC is rapidly evolving to encompass biomarker-targeted therapies; however, limitations associated with traditional tissue-based testing approaches prevent some patients from accessing these treatment options. Liquid biopsies hold promise for democratising access; however, ctDNA shedding is inadequate for accurate genotyping in a minority of PDAC patients, and little is known about the clinicopathological determinants for this. To address this, we assessed ctDNA fraction and prevalence of KRAS mutations by the metastatic site in 512 patients with advanced PDAC from the GOZILA study.

As expected, we found that ctDNA fraction indeed varied substantially among PDAC patients, with a significant minority below the threshold for adequate genotyping by current comprehensive liquid biopsy panels. Moreover, we found that metastatic site was the most important determinant of ctDNA levels; among patients with single-organ metastasis, those with liver metastases demonstrated significantly higher ctDNA levels and concomitantly higher KRAS mutation detection than patients with metastases to the lungs, peritoneum, or lymph nodes or with locally advanced disease. Although these findings were consistent with observations previously made in colorectal cancer [11], there have been no reports of clinically investigated ctDNA by the metastatic site in PC. The observation of this same pattern in two biologically and anatomically distinct tumour types suggests that ctDNA levels may be influenced more by the metastatic site rather than the features of tumour itself. Indeed, the liver is unique in that it possesses a discontinuous or ‘fenestrated’ endothelium, which may more readily allow transfer of ctDNA from the hepatic parenchyma to the circulation.

Regardless of the specific reason for varying ctDNA fractions, the implications for plasma-based genotyping are clear: patients with liver metastases are more likely to be adequately genotyped using a plasma-based approach than patients with other disease distributions. This may allow for improved patient selection for liquid vs. tissue testing and/or heightened vigilance in patients with low-shedding disease distributions.

In addition to its implications for genotyping, our data further indicate that KRAS mutation in ctDNA fraction is associated with OS. This finding is congruent with previous reports that have observed this phenomenon [12–14] and likely reflects both tumour burden as well as biological aggressiveness. As such, this finding supports a previous report that ctDNA may serve as a useful surrogate to monitor disease status in a minimally invasive manner with potentially greater dynamic sensitivity than radiographic approaches [15].

Our study has several limitations: first, our study did not assess volumetric tumour burden metrics, such as tumour size and number of metastases per metastatic site, and as such, we are unable to differentiate between tumour volume and metastatic location as contributors to the observed differences in ctDNA fraction. In a related study [11], however, in the patients with lung-only and peritoneum-only metastasis of colorectal cancer, 69.2% and 87.5% of patients with max VAFs of <0.2% exhibited the longest lesion diameter of <20 mm [11]. Second, the GOZILA study enrolled only patients progressing on therapy, while genotyping is most beneficial when conducted prior to the initiation of therapy. As such, the data do not directly inform the most relevant PDAC population, though we have no reason to suspect that the observations here could not be generalised to first-line patients. Third, we did not evaluate the genetic alteration of tumour in these patients. Therefore, the correlation with ctDNA analysis is unknown. Fourth, we do not have enough sample size to run subgroup analysis for OS stratified by KRAS status.

In conclusion, a liquid biopsy is an essential screening test to confirm the status of KRAS mutations or genomic alterations that could be therapeutic targets in patients with PDAC, in whom the collection of tumour tissue can be difficult. Although our study indicated that the frequency of KRAS mutation in ctDNA was basically low and the maximum VAF differed depending on the metastatic organ in patients with PDAC, the prevalence of KRAS mutations and maximum VAF were higher in patients with metastasis to the liver than in those with metastasis to other sites. Our findings provide important insight that PDAC patients with metastasis to the liver may benefit most from liquid biopsy, while tissue-based NGS tests may be taken priority over liquid biopsy in patients with metastasis other than liver including lungs, lymph nodes and peritoneum. Further studies are needed to clarify whether different clinical management is warranted for these populations.

Author contributions

KU and YS contributed to the planning and conducting of studies, recruiting patients, acquisition of data, analysis and interpretation of the data and writing of the manuscript. MF, N Mizuno, KS, YK, TK, KO, NO, N Matsuhashi, SI, T Matsumoto, SS, T Otsuru, HH, H Okuyama, H Ohama, T Moriwaki and T Ohta contributed to the recruitment of patients and acquisition of data. JIO, YN, HB and TY contributed to the planning, conducting of studies and interpretation of data. MU, MI and CM contributed to the recruitment of patients, acquisition of data, planning and interpretation of data. All authors agree to be accountable for all aspects of the work and will ensure that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Funding

This work was supported by SCRUM-Japan Funds (http://www.scrum-japan.ncc.go.jp/index.html).

Data availability

The authors declare that all variant data used in the conduct of the analyses are available within the article and its Supplementary information. To protect the privacy and confidentiality of patients in this study, clinical data are not made publicly available in a repository or the supplementary material of the article but will be made available following reasonable request to the corresponding author.

Competing interests

MF, KO, N Matsuhashi, SI, T Matsumoto, T Otsuru, HH, H Okuyama and H Ohama have nothing to disclose. KU reports honoraria from Chugai Pharmaceutical, Taiho Pharmaceutical and Yakult Honsha. YS reports honoraria from Takeda Pharmaceutical, Eli Lilly Japan, Chugai Pharmaceutical, Taiho Pharmaceutical, Bristol-Myers Squibb, Ono Pharmaceutical, Bayer, Daiichi Sankyo and Merck Biopharma; and grants from IQVIA and Parexel; and contributions or endowed chair from Takeda Pharmaceutical, Taiho Pharmaceutical, Chugai Pharmaceutical, Eli Lilly Japan, Sanofi and EN Otsuka Pharma. MU reports grants and personal fees from Taiho Pharmaceutical, AstraZeneca, Merck Biopharma, MSD, Ono Pharmaceutical, Incyte Corporation and Chugai Pharmaceutical; and personal fees from Nihon Servier; and grants from Astellas Pharma, Eisai and DFP. N Mizuno reports grants and personal fees from AstraZeneca, Novartis, Yakult Honsha, Ono Pharmaceutical, Taiho Pharmaceutical; and grants from MSD, Dainippon Sumitomo Pharma, ASLAN Pharmaceuticals, Incyte Corporation. and Seagen; and personal fees from Teijin Pharma, FUJIFILM Toyama Chemical, outside the submitted work. KS reports honoraria from Ono Pharmaceutical and Yakult Honsha; and grants (for the institution) from Bristol-Myers Squibb/Ono Pharmaceutical, Eisai and Incyte corporation. YK reports honoraria from Taiho Pharmaceutical, Incyte, Merck Biopharma, Yakult Honsha, and Eli Lilly; and grants from Takeda Pharmaceutical. TK reports honoraria from Taiho Pharmaceutical, Chugai Pharmaceutical, Ono Pharmaceutical and Eli Lilly. NO reports honoraria from Taiho Pharmaceutical, Eli Lilly, Eisai, Bayer Yakuhin, Chugai Pharmaceutical, Ono Pharmaceutical and Takeda Pharmaceutical; and advisory board from GlaxoSmithKline. SS reports grants from AstraZeneca, Incyte Corporation and Delta-Fly Pharma. T Moriwaki reports honoraria from Taiho Pharmaceutical, Eli Lilly Japan, Takeda Pharmaceutical, Chugai Pharmaceutical, Sanofi, Bayer Yakuhin, Merck Biopharma, Ono Pharmaceutical and Yakult Honsha; and grants from Taiho Pharmaceutical, MSD, Takeda Pharmaceutical, and Yakult Honsha. T Ohta reports honoraria from Chugai Pharmaceutical, Teijin Pharma, Takeda Pharmaceutical, Eisai, Yakult-Honsha, Daiichi Sankyo, Merck Biopharma, Ono Pharmaceutical, Taiho Pharmaceutical and Bristol-Myers Squibb; and grants from Takeda Pharmaceutical. JIO reports from, and stock interests in, Guardant Health. YN reports grants from Taiho Pharmaceutical, Chugai Pharmaceutical, Guardant Health, Genomedia, Daiichi Sankyo, Seagen, Roche Diagnostics. HB reports honoraria from Taiho Pharmaceutical, Eli Lilly Japan, Ono Pharmaceutical; and grants from Ono Pharmaceutical outside the submitted work. TY reports honoraria from Taiho Pharmaceutical, Chugai Pharmaceutical, Eli Lilly, Merck Biopharma, Bayer Yakuhin, Ono Pharmaceutical and MSD; and grants from Ono Pharmaceutical, Sanofi, Daiichi Sankyo, Parexel International, Pfizer Japan, Taiho Pharmaceutical, MSD, Amgen, Genomedia, Sysmex, Chugai Pharmaceutical and Nippon Boehringer Ingelheim. MI reports honoraria from Bayer, Bristol-Myers Squibb, Eli Lilly, Eisai, Sumitomo Dainippon Pharma, EA Pharma, Teijin Pharma, Yakult Honsha, Taiho Pharmaceutical, Otsuka, MSD, Mylan, Nihon Servier, Chugai Pharmaceutical, AstraZeneca, AbbVie, Abbott, Takeda, Novartis and Astellas Pharma; and advisory roles with Bayer, Eli Lilly, Eisai, Chugai Pharmaceutical, AstraZeneca, Takeda Pharmaceutical, Ono Pharmaceutical, GlaxoSmithKline and Nihon Servier; and grants from Bayer, Bristol-Myers Squibb, Eli Lilly, Eisai, Takeda Pharmaceutical, AstraZeneca, Chugai Pharmaceutical, Merck Biopharma, ASLAN Pharmaceuticals, Novartis, Yakult Honsha, Taiho Pharmaceutical, Ono Pharmaceutical, MSD, Merus N.V., Nihon Servier, Pfizer, Chiome Bioscience, Delta-Fly Pharma, and J-Pharma. CM reports honoraria from Nihon Servier, Novartis, Yakult Honsha, Teijin Pharma, Taiho Pharmaceutical, Eisai and MSD; advisory roles with Yakult Honsha, Novartis, Servier, Taiho Pharmaceutical and Abbvie; and grants from Eisai, Yakult Honsha, Ono Pharmaceutical, Taiho Pharmaceutical, J-Pharma, Daiichi Sankyo, HITACHI, AstraZeneca and Merck Biopharma.

Ethics approval and consent to participate

The study was conducted in accordance with the Declaration of Helsinki and the Japanese Ethical Guidelines for Medical and Health Research Involving Human Subjects. The protocol of the GOZILA study was approved by the institutional review board of each participating institution and registered at the University Hospital Medical Information Network (UMIN) Clinical Trials Registry (protocol no. UMIN000029315 for GOZILA.

Consent to publish

This manuscript does not contain any individual person’s data such as individual details, images or videos.

Supplementary information

The online version contains supplementary material available at 10.1038/s41416-023-02189-y.