Role of preoperative circulating tumor DNA in predicting occult metastases in resectable and borderline resectable pancreatic ductal adenocarcinoma

Abstract

BACKGROUND

Some patients with resectable or borderline resectable pancreatic ductal adenocarcinoma (PDAC) may have distant metastases, undetected on preoperative imaging or early recurrence, within 6 months after surgery. Occult metastases (OMs) must be accurately predicted to optimize multidisciplinary treatment.

AIM

To investigate the efficacy of circulating tumor DNA (ctDNA) in predicting OM.

METHODS

Two Japanese institutions prospectively collected preoperative plasma samples from PDAC patients between July 2019 and September 2021 and evaluated ctDNA using a targeted next-generation sequencing panel covering 52 cancer-related genes.

RESULTS

Among 135 PDAC patients, 38 had OM and 35 were positive for ctDNA. The ctDNA positivity rate was significantly higher in patients with OM than in patients without OM. ctDNA-positive patients had significantly shorter median recurrence-free survival than ctDNA-negative patients. Logistic multivariate regression revealed ctDNA positivity as an independent predictor of OM.

CONCLUSION

Preoperative ctDNA in resectable PDAC is an independent predictor of OM and indicates poor prognosis following pancreatectomy and may be a useful biomarker in determining multidisciplinary patient care.

Key Words: Pancreatic ductal adenocarcinoma; Occult metastases; Early recurrence; Multidisciplinary treatment; Circulating tumor DNA

Core Tip: Circulating tumor DNA (ctDNA) levels, when used alongside conventional biomarkers, may offer valuable insights for developing individualized treatment strategies in pancreatic ductal adenocarcinoma. These include determining the optimal regimen and duration of neoadjuvant therapy. In addition, ctDNA may help identify the most appropriate timing for surgery, thereby enabling more precise treatment decisions and potentially contributing to improved clinical outcomes and long-term prognosis.

INTRODUCTION

Pancreatic ductal adenocarcinoma (PDAC) has the poorest prognosis among carcinomas, with only approximately 20% of patients eligible for resection based on imaging findings at the time of diagnosis[1]. Neoadjuvant therapy (NAT) based on resectability status[2,3] improves the outcomes of PDAC patients[4,5]. However, 20%-40% of patients with resectable or borderline resectable PDAC are found to be unresectable because distant metastases, not detected on imaging, are discovered during abdominal examination[6-8]. This underdiagnosis may lead to patients receiving inappropriate treatment regimens or undergoing unnecessary laparotomies. Patients with distant metastases, detected during laparotomy, may have lower induction rates of chemotherapy[9] and delayed chemotherapy initiation[10] than those with distant metastases identified by staging laparoscopy. This has been suggested to have a detrimental impact on prognosis[9,10].

Moreover, approximately 20% of patients who undergo pancreatectomy experience early recurrence within 6 months after surgery[11,12]. These patients may have poorer prognosis than those whose distant metastases are detected at the time of surgery and who receive chemotherapy[13]. Distant metastases within 6 months after surgery were considered manifestations of micrometastases that were undetected intraoperatively[14]. Patients with early recurrence and those whose distant metastases were detected intraoperatively may have received inadequate NAT or inappropriate surgical intervention.

Therefore, distant metastases appearing between the time of surgery and within 6 months postoperatively were considered the same condition[13] and are referred to as occult metastases (OMs) herein. OM detection must be addressed in the preoperative selection of appropriate treatment regimens and surgical decision-making in the multidisciplinary treatment of PDAC. Carbohydrate antigen 19-9 (CA19-9) levels and tumor size are predictors of radiologically undetectable distant metastasis[6,15] or early postoperative recurrence[16] in PDAC. Nevertheless, the prediction accuracy of these factors is limited[6,15,16]; thus, the development of new biomarkers must be prioritized. To address this problem, we focused on circulating tumor DNA (ctDNA) as a biomarker. Recently, ctDNA has attracted considerable research attention as a novel biomarker for cancer treatment[17] and for its potential usefulness in early cancer detection, prognostic stratification, and monitoring of treatment efficacy[18]. ctDNAs may predict the presence of unresectable factors in PDAC[19]. The KRAS ctDNA positivity rate in PDAC patients is as high as approximately 80% for clinical stage (cStage) IV and as low as approximately 30%-40% for cStages I-III[20]. The ctDNA positivity rate is even lower in patients who are truly stage I-III[19] because some cStage I-III patients had distant metastases during surgery. Thus, preoperative KRAS ctDNA positivity may correlate with poor prognosis and early recurrence following pancreatectomy[19,21]. Therefore, preoperative ctDNA positivity in resectable and borderline resectable PDAC may indicate the presence of OM.

A wide variety of genetic mutations have been detected in ctDNA in PDAC[22], although most previous studies have focused only on KRAS[19,20]. In this study, we evaluated ctDNA using targeted next-generation sequencing (NGS) of 52 cancer-related genes to obtain comprehensive information and clarify the association between preoperative ctDNA and OM in resectable and borderline resectable PDAC.

MATERIALS AND METHODS

Patients

This prospective study enrolled patients with resectable and borderline resectable PDAC who underwent planned pancreatectomy at Sapporo Medical University and Hokkaido University from 1 June 2019 to 30 September 2021. All patients were histologically or cytologically diagnosed with PDAC. Patients with in situ carcinoma and those showing a pathologically complete response to NAT were excluded, as no residual tumor capable of shedding ctDNA remained, which could affect the assessment of the association between ctDNA and OM. OM was defined as distant metastases that were not detected during preoperative imaging but became evident within 6 months of surgery as well as intraoperatively.

This study was approved by the Institutional Review Board of Sapporo Medical University (No. 302-240), Hokkaido University (No. 18-053), and the Japanese Foundation for Cancer Research (No. 2018-1016 and No. 2020-1150). The study was conducted according to the principles of the Helsinki Declaration. This study complied with the strengthening the reporting of observational studies in epidemiology guidelines[23] and reporting recommendation for tumor marker prognostic studies criteria[24]. This study was registered in the University Hospital Medical Information Network Clinical Trials Registry (registration No. UMIN 000054546). All eligible patients provided written informed consent before participation.

Neoadjuvant and adjuvant therapy

During the study period, NAT was standardized in Japan, and gemcitabine plus S-1 (tegafur, gimeracil, and oteracil potassium) was an established treatment protocol for resectable PDAC[4]. Prior to this, standard NAT protocol for resectable or borderline resectable PDAC had not been established, and NAT was conducted as part of various clinical studies, leading to a lack of protocol standardization. Adjuvant therapy with S-1 was established for 6 months as the standard[25] and was initiated at physician discretion for patients meeting the institutional criteria.

Perioperative evaluation

Resectability status was assessed at diagnosis using contrast-enhanced multidetector computed tomography (MDCT) according to the National Comprehensive Cancer Network guidelines. Patients who underwent NAT were evaluated using MDCT within 2 weeks of NAT completion. Gadolinium ethoxybenzyl diethylenetriamine pentaacetic acid-enhanced magnetic resonance imaging (EOB-MRI) was performed before and after NAT to explore small liver metastases. Baseline CA19-9 Levels were measured in the absence of elevated bilirubin levels (< 2 mg/dL). Preoperative CA19-9 Levels were measured within 2 weeks of surgery. Pancreatectomy was planned 4-8 weeks after completing NAT in the absence of unresectable imaging factors. Immediately after laparotomy, 100 mL of saline was injected into the pouch of Douglas and lavage fluid was collected for cytology. Liver metastases and peritoneal dissemination were meticulously investigated, and suspicious nodules were collected for pathological evaluation via frozen section analysis. All patients without liver or peritoneal metastases underwent para-aortic lymph node sampling. If liver metastases or peritoneal dissemination were identified in frozen sections, pancreatectomy was not performed. Positive washing cytology and para-aortic lymph node metastases were not treated as absolute contraindications to pancreatectomy.

For postoperative surveillance, CA19-9 Levels were measured every 1-2 months during the first year after surgery and every 3 months thereafter. CT scans were conducted every 3 months during the first year after surgery and every 6 months thereafter. In cases of elevated tumor marker levels, a CT scan was conducted promptly to evaluate recurrence, with additional EOB-MRI or positron emission tomography-CT, if deemed necessary. Recurrence was diagnosed based on radiological findings and confirmed via biopsy whenever feasible.

Blood collection and ctDNA analysis

Peripheral blood (14 mL) was collected from all patients on the day of surgery and placed in ethylenediaminetetraacetic acid disodium salt tubes. In some patients, blood samples were collected on the day NAT was initiated. Blood sample processing and DNA extraction methods have been previously reported[21]. A library for NGS evaluation of ctDNA was constructed according to the manufacturer’s protocol (Thermo Fisher Scientific, Waltham, MA, United States) using the Oncomine Pan Cancer Cell-Free Assay for 52 genes (Supplementary Table 1). The constructed libraries were subjected to template preparation using either the Ion 540 or 550 Chef Kit (Thermo Fisher Scientific) on an Ion Chef (Thermo Fisher Scientific). Subsequently, sequencing was performed using the Ion GeneStudio S5 Prime System (Thermo Fisher Scientific). Detected ctDNA was classified as positive if the mutant allele frequency (MAF) was ≥ 0.065% and the number of mutant alleles copies was two or more. This threshold is based on the manufacturer’s protocol (Thermo Fisher Scientific) for the Oncomine Pan-Cancer Cell-Free Assay and has been used in prior studies[21]. Mutations identified in the buffy coat that matched the plasma cell-free total nucleic acids were not classified as ctDNAs because they were associated with clonal hematopoiesis-related mutations. Details of genetic analysis using liquid biopsy have been reported previously[21].

Statistical analyses

Categorical variables were compared using the χ² test or Fisher’s exact test as appropriate, and continuous variables are reported as medians (interquartile range) and compared using the Mann-Whitney U test. Recurrence-free survival (RFS) and overall survival were defined as the period from the date of surgery to recurrence, death, or last follow-up and to death or last follow-up, respectively. Survival curves were drawn using Kaplan-Meier estimation, and the results were compared using the log-rank test. The cut-off date was July 31, 2023. The optimal cut-off points for preoperative CA19-9 Levels and tumor size to predict OM were identified as the upper left corner of the receiver operating characteristic (ROC) curve. Logistic regression analysis was performed to identify independent factors associated with OM. Multivariate analysis included variables with P values < 0.05 in the univariate analysis. Statistical significance was set at P values < 0.05. All analyses were performed using the JMP Pro statistical software package Pro version 16.0.0 (SAS Institute Inc., Tokyo, Japan).

RESULTS

Patient cohort

During the study period, 141 patients with resectable and borderline resectable PDAC were prospectively enrolled, and their blood samples were collected. The CONSORT diagram is shown in Figure 1. One patient was excluded from the analysis because of pathological carcinoma in situ, two because of a pathologically complete response, and three because of defective blood samples, leaving 135 patients (70 men, 65 women) for analysis. The median patient age was 71 (65-76) years. Median CA19-9 Levels at diagnosis and preoperatively were 76.6 U/mL (21.3-258.2 U/mL) and 30.9 U/mL (11.6-102 U/mL), respectively. The tumors were located in the pancreatic head, body, and tail in 98 (72.6%), 29 (21.5%), and 8 (5.9%) patients, respectively. The median tumor sizes at diagnosis and preoperatively were 22.5 mm (17 mm-30 mm) and 18 mm (14 mm-25 mm), respectively. The resectability status was resectable in 91 (67.4%) patients and borderline resectable in 44 (32.6%) patients. Twenty-six (19.3%) patients underwent upfront surgery. Among the remaining 109 (80.7%) patients, 99 received chemotherapy and 10 received chemoradiotherapy (CRT) as NAT (Table 1). All 44 patients with borderline resectable PDAC received NAT, while 65 of 91 patients with resectable PDAC received NAT. The NAT regimens for patients with resectable PDAC were gemcitabine and nab-paclitaxel (GnP) (31 patients), gemcitabine and S-1 (GS) (25 patients), S-1 (7 patients), and CRT (2 patients). NAT was performed in all 44 patients with borderline resectable PDAC (44 patients). The NAT regimens were GnP (29 patients), CRT (8 patients), modified oxaliplatin, irinotecan, 5-fluorouracil, and leucovorin (FOLFIRINOX) (5 patients) and oxaliplatin, irinotecan, and S-1 (IRIS-OX) (1 patient), and GS (1 patient).

Figure 1 Patient enrollment and sample collection. R: Resectable; BR: Borderline resectable; PDAC: Pancreatic ductal adenocarcinoma; NAT: Neoadjuvant therapy; pCR: Pathologiccaly complete response; pTis: Pathological carcinoma in situ.

Table 1 Baseline characteristics of patients.

Variables	n = 135
Age, years, median (IQR)	71 (65-76)
Sex, male/female	70/65
Baseline CA19-9, U/mL, median (IQR)	76.6 (21.3-258.2)
Preoperative CA19-9, U/mL, median (IQR)	30.9 (11.6-100.2)
1Resectability status, resectable/BR	91/44
Tumor location, Ph/Pb/Pt	98/29/8
Neoadjuvant therapy, present	109
Chemotherapy, present	99
Chemoradiotherapy, present	10
Baseline tumor size, mm, median (IQR)	22.5 (17-30)
Preoperative tumor size, mm, median (IQR)	18 (13.8-25)

¹According to the National Comprehensive Cancer Network criteria.

BR: Borderline resectable; IQR: Interquartile range; Pb: Pancreatic body; Ph: Pancreatic head; Pt: Pancreatic tail; CA19-9: Carbohydrate antigen 19-9.

Intra-abdominal exploration following laparotomy identified 23 patients with unresectable factors (17.0%), i.e., 3 with liver metastases, 7 with peritoneal dissemination, 8 with para-aortic lymph node metastases, and 5 with positive washing cytology. Pancreatectomy was performed in 13 of these patients because positive washing cytology was not considered an unresectable factor according to the Japan Pancreas Society guideline[3] during the study period and para-aortic lymph nodes were sampled for staging in all patients but were not assessed for frozen sections from some patients. After excluding patients with unresectable factors, all 112 patients underwent pancreatectomy and R0 resection was achieved in 101 (90.2%) patients. Adjuvant therapy was administered to 86 patients, 78 of whom received oral S-1. After a median observation period of 37.1 months, 88 patients experienced relapse, with 15 patients experiencing recurrence within 6 months. Thus, 38 of 135 patients with resectable and borderline resectable PDAC had OM (Supplementary Table 2). There were no patients without OM and with an observation period of less than 6 months.

ctDNA analysis

Preoperative peripheral blood samples of 135 patients were analyzed, and after excluding clonal hematopoiesis-related mutations, mutated genes were detected in 49 patients. Fourteen patients had ctDNA positivity levels below the cut-off, resulting in the identification of 59 ctDNA mutations in 35 patients. The most commonly detected ctDNA was TP53 (20 patients), followed by KRAS (9 patients). EGFR, MET, and SMAD4 mutations were each identified in 3 patients, whereas BRAF, GNAS, and PIK3CA mutations were each observed in 2 patients. Nine other mutations were found in 1 patient (Supplementary Table 3). The ctDNA positivity rate in the NAT group (22.9%, 25/109) was lower than that in the upfront surgery group (38.5%, 10/26; P = 0.105).

Table 2 shows a comparison of patient backgrounds between the ctDNA-positive and ctDNA-negative groups. There were no differences between the two groups in clinicopathological factors, including age, sex, CA19-9 Levels, tumor size, and pathological T and N factors. However, the frequency of borderline resectable PDAC was lower in the ctDNA-positive group than in the ctDNA-negative group (P = 0.023). Additionally, OM frequency in the ctDNA-positive group was 51% (18/35), which was significantly higher than the 20% (20/100) frequency of OM in the ctDNA-negative group (P < 0.001). In patients with OM, the ctDNA positivity rate was 47.4%, which was significantly higher than the 18% positivity rate observed in patients without OM (P < 0.001; Supplementary Figure 1A). ctDNA positivity did not differ significantly across pathological stages I-III, ranging from 18.4% to 28.6%, yet showed an increase to 52.2% in stage IV (Supplementary Figure 1B).

Table 2 Comparison of clinicopathological factors between the circulating tumor DNA positive and negative groups.

Variables	ctDNA + (n = 35)	ctDNA - (n = 100)	P value
Age, years, median (IQR)	72 (65-77)	71 (66-76)	0.846
Sex, male/female	23/12	47/53	0.057
Baseline CA19-9, U/mL, median (IQR)	79.9 (18.0-236)	71.2 (21.5-267.6)	0.891
Preoperative CA19-9, U/mL, median (IQR)	30.9 (12.3-125.2)	30.5 (11.3-97.5)	0.572
1Resectability status, resectable/BR	29/6	62/38	0.023
Tumor location, Ph/Pb/Pt	26/9	72/28	0.794
Neoadjuvant therapy, absent/present	10/25	16/84	0.105
Baseline tumor size, mm, median (IQR)	21 (15-26)	23 (17-30)	0.145
Preoperative tumor size, mm, median (IQR)	17 (14-25)	19 (13-25)	0.442
2Pathological T factor, 1/2/3	11/14/5	31/52/11	0.551
2Regional lymph node metastases, 0/1/2	14/11/5	45/35/14	0.973
Post operative therapy, absent/present	5/30	13/87	0.847
Occult metastases, absent/present	18/17	20/80	0.001

¹According to the National Comprehensive Cancer Network criteria.

²According to the Union for International Cancer Control tumor node metastasis classification 8^th edition. Excluded ten patients who had not undergone pancreatectomy.

ctDNA: Circulating tumor DNA; BR: Borderline resectable; IQR: Interquartile range; Pb: Pancreatic body; Ph: Pancreatic head; Pt: Pancreatic tail; CA19-9: Carbohydrate antigen 19-9.

Univariate and multivariate analyses of preoperative predictors for OM

Univariate and multivariate logistic regression analyses were performed to assess whether ctDNA status was associated with OM relative to other clinicopathological factors. The optimal cut-off values of preoperative CA19-9 Levels and tumor size for predicting OM were assessed using ROC curve analysis. The optimal CA19-9 Level and tumor size cut-off values for predicting OM were 150.0 U/mL and 29 mm, respectively. Three factors were identified to be associated with OM: Preoperative CA19-9 Levels, tumor size, and ctDNA. In the multivariate analysis, preoperative CA19-9 Levels 150 U/mL [odds ratio (OR) = 2.89; 95% confidence interval (CI): 1.10-7.61; P = 0.031] and ctDNA positivity (OR = 4.96; 95%CI: 2.07-11.90; P = 0.0003) were independent predictive factors for OM (Table 3). Analysis of the prediction accuracy of OM using ROC curves showed that adding ctDNA status to the prediction of tumor size and CA19-9 Levels significantly increased the area under the curve from 0.617 to 0.724 (P = 0.013) and improved diagnostic accuracy (Supplementary Figure 2).

Table 3 Univariate and multivariate analyses of preoperative factors related to occult metastases.

Factors	Univariate analysis			Multivariate analysis
	OR	95%CI	P value	OR	95%CI	P value
Age, ≥ 70 (years)	0.99	0.46-2.13	0.975
Sex, male	1.21	0.57-2.57	0.620
Tumor location, Pb, Pt	1.11	0.48-2.56	0.802
1Resectability status, BR	0.66	0.29-1.52	0.332
Neoadjuvant therapy, yes	1.82	0.63-5.25	0.265
Preoperative CA19-9, ≥ 150 (U/mL)	3.08	1.28-7.41	0.012	2.89	1.10-7.61	0.031
Preoperative tumor size, ≥ 29 (mm)	3.03	1.10-8.37	0.032	2.82	0.92-8.64	0.069
ctDNA, positive	4.23	1.86-9.66	0.0006	4.96	2.07-11.90	< 0.001

¹According to the National Comprehensive Cancer Network criteria.

OR: Odds ratio; CI: Confidence interval; ctDNA: Circulating tumor DNA; BR: Borderline resectable; Pb: Pancreatic body; Pt: Pancreatic tail; CA19-9: Carbohydrate antigen 19-9.

Association between ctDNA status and prognosis

A Kaplan-Meier curve excluding 10 non-resected patients showed that the median RFS in the ctDNA-positive group was 10.2 months, significantly worse than the 25.8 months in the ctDNA-negative group (P = 0.023; Figure 2A). In the Kaplan-Meier analysis of all 135 patients, including non-resected cases, the median OS was significantly shorter in the ctDNA-positive group (22.9 months) than in the ctDNA-negative group (41.4 months) (P = 0.039; Figure 2B).

Figure 2 Association between circulating tumor DNA status and prognosis. A: Kaplan-Meier curve of postoperative recurrence-free survival (RFS) in circulating tumor DNA (ctDNA)-positive and -negative groups of 125 patients who underwent pancreatectomy, excluding 10 patients who underwent laparotomy. During the observation period of 37.1 months, the median RFS in the ctDNA-positive group was 10.2 months, significantly worse than the 258 months in the ctDNA-negative group (P = 0.023); B: Kaplan-Meier curve of postoperative overall survival in ctDNA-positive and -negative groups of 135 patients. The median overall survival in the ctDNA-positive group was 22.9 months, significantly worse than the 411 months in the ctDNA-negative group (P = 0.039). ctDNA: Circulating tumor DNA.

NAT-induced ctDNA dynamics and prognosis after pancreatectomy

In the subgroup analysis, we assessed the association between ctDNA dynamics during NAT and postoperative recurrence in patients who underwent pancreatectomy. Blood samples were collected before NAT from 32 of the 109 patients who underwent NAT. Excluding 1 patient who did not undergo resection owing to the intraoperative detection of distant metastasis, the analysis of samples collected before NAT in 31 patients revealed that 14 patients tested positive, while 17 were negative for ctDNA. Subsequent analysis of samples collected before pancreatectomy showed that 2 patients remained positive, 2 became positive, 12 became negative, and 15 remained negative for ctDNA (Supplementary Table 4).

The median RFS of 20.2 months for the group whose ctDNA changed from positive to negative after NAT was not significantly different from the median RFS of 18.0 months for the group whose ctDNA remained negative during NAT (P = 0.862; Figure 3). The median RFS in the group that remained or changed to ctDNA-positive after NAT was 6.2 months. Although statistical significance could not be demonstrated owing to the small sample size, our results suggest that the prognosis for this population may be very poor.

Figure 3 Kaplan-Meier curves of circulating tumor DNA dynamics and postoperative recurrence-free survival in 31 patients sampled before and after neoadjuvant therapy. The median recurrence-free survival (RFS) of 20.2 months for the group whose circulating tumor DNA (ctDNA) changed from positive to negative after neoadjuvant therapy (NAT) was not significantly different from the median RFS of 18 months for the group whose ctDNA remained negative during NAT (P = 0.862). ctDNA: Circulating tumor DNA; RFS: Recurrence-free survival.

DISCUSSION

In this study, ctDNA positivity emerged as a strong risk factor for OM in patients with resectable and borderline resectable PDAC. Although there were no discernible differences in the clinicopathological background (e.g., CA19-9 Level, tumor size, T-factor, and N-factor) between the ctDNA-positive and ctDNA-negative groups, significant disparities in OM frequency and prognosis after resection were observed. The frequency of borderline resectable PDAC was lower in the ctDNA-positive group than in the ctDNA-negative group. In an analysis of 31 patients where samples were collected before and after NAT, many patients became ctDNA-negative following NAT. This suggests that the lower incidence of borderline resectable PDAC in the ctDNA-positive group is attributed to the differing rates of NAT administration. Among patients with borderline resectable PDAC, the ctDNA positivity rate may have been lower because surgery was selected for patients who demonstrated a favorable response to NAT and were subsequently deemed eligible for R0 resection.

The addition of ctDNA to CA19-9 and tumor size, which are risk factors for distant metastasis and early recurrence of PDAC[6,15,16], significantly improved the diagnostic accuracy of OM. Consequently, integrating ctDNA positivity into clinical workflows may help identify patients at high risk of OM prior to surgery. By combining ctDNA status with imaging and tumor markers, clinicians may refine risk stratification, avoid non-beneficial surgery, and offer more personalized treatment approaches for patients with resectable or borderline resectable PDAC.

The high incidence of postoperative recurrence in PDAC suggests that despite the appearance of localized disease on imaging, most patients already harbor systemic disease[14]. In particular, recurrence within 6 months after surgery has a dismal prognosis, with an overall survival comparable to that of patients whose distant metastases are identified after laparotomy and are classified to have unresectable disease[13]. Such early recurrences likely reflect micrometastases that were already present but undetectable at the time of surgery[13], supporting our inclusion of metastases occurring within 6 months postoperatively in the definition of OM. Pancreatectomy is a highly invasive procedure that poses challenges in the prompt initiation of adjuvant therapy following surgery. Data from a Japanese multicenter cohort study showed that the median time to the initiation of adjuvant S-1 therapy after pancreatectomy was 50 days[26]. Therefore, controlling recurrence within 6 months of surgery with adjuvant therapy may not be practical, and preoperative determination of this risk and intervention may contribute to improved outcomes.

In PDAC patients, preoperative detection of KRAS ctDNA in the peripheral blood has been reported to be an indicator of poor prognosis[27] and a risk factor for radiologically negative distant metastases[19]. Since peripheral blood from PDAC patients contains KRAS mutations and diverse tumor-derived gene mutations[22], we evaluated ctDNA using a targeted NGS panel covering 52 cancer-related genes to obtain comprehensive information. The positivity rate of ctDNA in PDAC patients without distant metastasis is low because of the biological features of the tumor[28] that prevent ctDNA from entering the blood circulation[29]. Therefore, the positive rate of ctDNA in resectable and borderline resectable PDAC is 19.4%-37.7% with targeted NGS for multiple gene mutations[30,31] and approximately 22.9% using a personalized tumor-informed assay[32]. In our cohort, NAT was administered to most patients, although the ctDNA detection rate was 25.9%, which is similar to the reported rate[30-32]. Furthermore, the sensitivity of ctDNA detection remains limited. False-negative results may occur in patients with low tumor burden or limited ctDNA shedding, and thus a negative result should not exclude the presence of OM.

We previously reported the utility of a tumor-informed approach to increase the detection rate of ctDNA in resected pancreatic cancer[21]. However, in this study, we examined the preoperative predictors of OM and did not use a tumor-informed approach based on resected specimens. Although the ctDNA positivity in this cohort was a strong independent predictor of OM, improvements in the ctDNA detection rate may provide additional information for stratifying patient prognosis. Several studies have demonstrated that targeted NGS is feasible using endoscopic ultrasonography-guided fine-needle aspiration or biopsy specimens collected at the time of diagnosis[33-35]. Therefore, a tumor-informed approach using these specimens may enhance ctDNA detection rates.

Monitoring ctDNA in unresectable PDAC is useful for estimating treatment response during chemotherapy[36,37]. However, the significance of monitoring ctDNA levels during NAT in resectable and borderline resectable PDAC remains unclear. Thus, we investigated the correlation between ctDNA dynamics and prognosis in a limited number of patients from whom blood samples were collected both before and after NAT. In the subgroup of 31 patients whose ctDNA was evaluated before and after NAT, patients who transitioned from ctDNA-positive to ctDNA-negative had RFS comparable to those who were consistently ctDNA-negative. These results suggest that ctDNA dynamics could serve as a key predictor of NAT efficacy. The changes in ctDNA during NAT may provide useful information for determining the optimal duration of NAT and appropriate timing of surgical intervention in a multidisciplinary treatment approach.

The detection rate of ctDNAs in PDAC with distant metastases differs depending on the metastatic site. The detection rate of ctDNA is low in patients with lung metastasis, whereas ctDNA is detected at a high rate in patients with liver metastasis[38]. Most early postoperative recurrences are liver metastases[11], and these may have already involved micrometastases at the time of surgery[14]. Therefore, preoperative ctDNA positivity may indicate the presence of micrometastases. However, because ctDNA positivity varies depending on the site of metastasis or recurrence, solely using ctDNA for the accurate prediction of OM is challenging. Therefore, combining ctDNA with existing markers, including tumor size and CA19-9 Levels, is essential for comprehensive prediction.

This study has several limitations. Although the study was conducted prospectively and across two centers, the number of patients was not sufficiently large. As the study only included patients for whom pancreatectomy was considered, there were no blood samples from patients whose disease progressed during NAT, thereby excluding them from resection. Only a limited number of patients had blood samples available before NAT, which may have introduced selection and analysis bias. In addition, NAT may reduce ctDNA levels, potentially introducing bias in the interpretation of ctDNA results. To improve patient stratification and better assess the prognostic value of ctDNA, future studies should consider performing ctDNA analysis both before and after NAT. Although an association between MAF and tumor progression[20] and prognosis[29,30] has been reported, in this study, MAF values above the cut-off were only treated as ctDNA-positive and were not stratified based on their value.

CONCLUSION

In resectable and borderline resectable PDAC, positive ctDNA was an independent preoperative predictor of the presence of OM and poor prognosis following resection. In the multidisciplinary treatment of PDAC, preoperative ctDNA evaluation may lead to more personalized and effective treatment strategies, thereby improving patient prognosis. A larger prospective validation study is necessary to confirm the relationship between ctDNA dynamics during preoperative treatment and prognosis.