1. Introduction
Globally, pancreatic cancer was the 12th most newly diagnosed cancer in 2020, with 495,773 new cases, and the 7th leading cause of cancer death in 2020 with 466,003 deaths [
1,
2]. Currently, the only cure for pancreatic cancer is surgery [
3], however, only 15–20% of patients have resectable tumours at diagnosis [
3]. Furthermore, up to 80% of patients have tumour recurrence following resection [
4]. There are few clinically available biomarkers that can effectively guide prognosis and treatment in pancreatic ductal adenocarcinoma (PDAC). Measuring serum levels of cancer antigen 19-9 (CA19-9) can be useful to confirm diagnosis as well as monitor disease recurrence. However, it has a relatively low predictive value and is of limited use in selecting patients for resection [
5].
Endoscopic ultrasound guided fine needle aspiration (EUS-FNA) remains the gold standard for the diagnosis of localised PDAC, with a sensitivity of around 85% and specificity of near 100% [
6]. Furthermore, it can be used to provide tissue for cytological and even histological diagnosis and can also be utilised for further applications such as
KRAS detection [
7]. However, it is not currently used to provide prognostic information to guide selection of patients for surgical resection.
In recent years, the molecular landscape of pancreatic cancer has become well defined. The four most commonly mutated genes in pancreatic cancer are
KRAS,
CDKN2A,
TP53, and
SMAD4. Mutations in these genes have been associated with oncogenic cellular processes and with prognosis in pancreatic cancer [
8]. The
KRAS gene encodes a guanosine triphosphatase (GTPase) protein that controls cellular processes by linking membrane growth factor receptors to intracellular signalling pathways and transcription factors. The
KRAS gene is mutated in over 90% of PDAC cases [
9].
The
KRAS mutation is usually a point mutation which frequently occurs on codon 12 of exon 2, affecting the first or second nucleotide, but can also affect other codons and exons such as codon 13 and 61. The normal or wild-type sequence is GGT encoding glycine, and it is present in 8–12% of PDAC [
10]. The most common mutation subtype is G12D, present in 40% of patients, whereby a GAT sequence is coded, producing aspartic acid [
10]. G12V subtype refers to a GTT replacement sequence producing valine, and G12R refers to a CGT replacement sequence producing arginine, and these are present in 33% and 15% of PDAC cases, respectively [
10].
Detection of
KRAS does not currently have a role in screening or diagnosis of pancreatic cancer due to difficulty standardizing sampling, transportation, extraction and detection. However, many studies have demonstrated cytopathology with
KRAS on EUS-FNA materials has a higher sensitivity, specificity and accuracy of PDAC diagnosis compared to cytopathology alone [
10]. Furthermore, liquid biopsies which detect circulating tumour cells and cell-free circulating tumour DNA (ctDNA) have been heavily investigated diagnostically and prognostically, with varying results [
10,
11].
Many studies have investigated the association between
KRAS mutation and PDAC patient survival, predominantly in patients with early-stage, resectable disease. Most studies found that
KRAS mutation was significantly associated with shorter overall survival, although a small number of studies found no statistically significant correlation [
12,
13,
14,
15]. The association between
KRAS mutation and survival in PDAC patients with advanced unresectable disease has also been assessed. EUS-FNA was the main source of tissue for
KRAS detection in this patient cohort [
16,
17,
18]. Ogura et al. analysed 242 patients with unresectable pancreatic cancer and found that those in the
KRAS mutation group had a significantly shorter survival compared to the wild-type group [
16]. Conversely, a recent study by on 219 patients with advanced PDAC found that there was no significant difference in survival between the mutant
KRAS and
KRAS wild-type groups [
18]. However, overall, as demonstrated in a recent meta-analysis, patients with
KRAS mutation had poorer overall survival [
19].
When considering the different mutant
KRAS subtypes, G12D is both the most frequently occurring
KRAS mutation subtype (36.9–67%) and also the subtype that is most commonly reported as being associated with poorer survival [
12,
13,
14,
15,
16,
18,
20,
21]. For example, Qian et al. found that
KRAS G12D subtype patients had an overall median survival of 15.3 months compared to 24.8 months in patients without the
KRAS G12D subtype [
15] (
n = 356). However, the
KRAS G12A and G12R subtypes have also been reported as being associated with a reduction in patient survival [
12,
16], while other investigators have been unable to demonstrate any correlation between a particular
KRAS subtype and survival [
17].
Given the heterogeneity of these results, we sought to assess the correlation between the KRAS G12D mutation subtype and survival in pancreatic cancer patients across all clinical stages using tissue biopsies obtained predominantly via EUS-FNA.
2. Methods
2.1. Ethics Statement
This retrospective cohort study was approved by the Monash Health Human Research Ethics Committee (Monash Health HREC Ref: 17387L). The cohort was comprised of patients who had previously had
KRAS mutations analysis performed on EUS-FNA or resection specimens. The patients had either provided tissue as part of the Victorian Pancreatic Cancer Biobank and various sub studies associated with this (Monash Health HREC Ref: 15450A) or were being screened for enrolment in a prospective cohort study examining the use of panitumumab as second or subsequent line therapy in advanced
KRAS wild-type pancreatic cancer (Monash Health HREC Ref: 16584A) [
22]. All patient information was stored in a confidential and de-identified manner.
2.2. Data and Sample Collection
We included all consenting patients from 2012 to 2020 with
KRAS mutation data. Demographic data, date and method of diagnosis, treatment details and survival were reviewed. Clinical stage was established via review of multi-disciplinary team meeting documentation. In general, clinical staging was determined on anatomical criteria using the NCCN Pancreatic Adenocarcinoma guidelines (2016) [
23,
24]. These guidelines specify criteria such as degree of tumour contact or invasion on the superior mesenteric artery, coeliac artery, or common hepatic artery, to distinguish between resectable, borderline resectable, locally advanced or metastatic pancreatic disease.
KRAS mutation analysis was performed either on EUS FNA specimens or from tissue taken from pancreatic resections (either a snap frozen core biopsy of the tumour taken immediately after resection or from archival formalin fixed paraffin blocks later retrieved from the pathology department). For patients undergoing EUS-FNA an additional biopsy was snap frozen and stored at −80°C or in liquid nitrogen in the Victorian Pancreatic Cancer Biobank (Monash Health HREC Ref: 15450A) or the Monash Surgical Oncology Biobank (Monash Health HREC Ref: 13058A).
DNA was extracted from patient biopsy samples via homogenization in a Buffer RLT plus AllPrep DNA/RNA Universal Kit (Qiagen, Hilden, Germany) as per the manufacturer’s protocol. The isolation of gDNA from FFPE tissue was performed on 5 × 10 micron-thick sections using the ReliaPrep FFPE gDNA Miniprep System (Promega, Madison, WI, USA). The quality and quantity of gDNA were determined on a NanoDrop Spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA) and Qubit Fluorometer (ThermoScientific), and TapeStation (Agilent, St. Clara, CA, USA). gDNA (25–50 ng) was subjected to the KRAS XL StripAssay® (ViennaLab Diagnostics GmbH, Wien, Austria). Mutations were objectively scored using StripAssay Evaluator software. For a small number of patients, KRAS mutation was determined using the TruSight Oncology 500 (TSO-500) gene panel (Illumina, San Diego, CA, USA).
2.3. Statistical Analysis
Survival was measured from date of first tissue diagnosis (either EUS-FNA or surgical resection, whichever came first) until death or censoring date defined as the last known follow-up. Treatment groups were based on if patients received resection with curative intent, chemotherapy and/or radiotherapy without surgery or supportive care only. In patients who received surgery, we further recorded lymph node status, tumour location, resection margin clearance and whether or not adjuvant or neoadjuvant therapy was administered.
A chi-squared test, chi-squared test with Yates’ continuity correction, and Fisher’s exact test were used to assess independent categorical variables. A Mann—Whitney U test was used to compare differences in age. Kaplan-Meier curves were generated for median survival within individual groups, and these were compared using the log-rank test for univariate analysis. KRAS mutant was compared with KRAS wild-type, and the G12D subtype was compared with non-G12D patients, which consists of all other mutant KRAS subtypes and KRAS wild-types combined. Multivariate analysis only included KRAS mutation as a whole, and also the G12D, G12V and G12R subtypes individually, and used covariates associated with survival on univariate analysis (defined by p < 0.1) and was conducted using the Cox proportional hazards regression model. The sample method (FNA vs. resection) was excluded from multivariate analysis as this was not considered clinically relevant to patient survival.
Statistical significance was defined at p < 0.05. All statistical analyses were performed using GraphPad Prism Version 9.1.1 (GraphPad Software, San Diego, CA, USA) and IBM SPSS Statistics 27 (IBM, Armonk, NY, USA).
4. Discussion
In this retrospective cohort study examining the correlation between the presence of a KRAS G12D mutation subtype and prognosis in PDAC across all stages of disease, we found that the presence of a KRAS G12D mutation was associated with reduced survival in those with resectable disease. There was also a suggestion (on univariate analysis only) of reduced survival in patients with a KRAS G12D mutation who underwent resection, however, after adjusting for potential confounders, this was no longer statistically significant (on multivariate analysis). A larger sample size may help clarify this further. There was no association between KRAS G12D mutation and the clinical stage of disease at presentation. Further, we could not demonstrate a statistically significant association between the presence of a KRAS G12D mutation subtype and survival in other clinical stages of disease or with other treatment modalities.
Our study is unique in a number of ways. Firstly, although we predominantly used EUS to assess
KRAS mutation status, we also included analysis of resection specimens (in patients who had not undergone EUS-FNA) allowing us to assess the association between
KRAS mutation and subtype across all stages of disease and in all treatment groups. Furthermore, we found that 87.4% of patients had a
KRAS mutation, which reflects the approximately 90% frequency reported in current literature [
9]. In addition, G12D is typically cited as occurring in 40% of PDAC patients, G12V usually occurs in 33% of patients, and G12R typically occurs in 15% of patients [
10]. Our cohort had 40.3% G12D, 27.7% G12V and 10.4% G12R. As such, our cohort is a representative sample of a PDAC population from a
KRAS mutation frequency perspective. Previous studies which have used EUS-FNA to assess
KRAS mutation status have often reported significantly lower rates of
KRAS mutation detection suggesting significant sampling error [
12,
13,
14]. The fact that we utilized an additional snap frozen biopsy for molecular analysis probably accounts for this difference. If EUS-FNA is to be used to guide treatment decisions based on molecular analyses, it is important that the information provided is indeed reliable and reflective of the tumour.
Secondly, our study represents the first time that the prevalence of
KRAS mutations has been simultaneously analysed across all clinical stages of disease allowing us to assess any potential associations between
KRAS mutations and both stage at presentation and survival. Most previous studies assessed either only resected patients [
25] or advanced unresectable disease [
19]. Furthermore, although some previous studies which focused on resected patients considered TNM as an independent predictor of survival [
19,
25], we chose to use the NCCN staging system [
26,
27]. The rationale for this is that the TNM staging system requires accurate assessment of lymph nodes status and this is only available for the 20% of patients who undergo resection, whereas the NCCN staging system can be applied to the whole population. We hypothesized that given that treatment options such as surgery or chemotherapy remain the most important prognosticators in PDAC, any association between
KRAS mutation and survival may be mediated by differences in the prevalence of
KRAS mutations and/or subtypes in the various NCCN clinical stages at diagnosis. However, as demonstrated above, there was in fact no association between
KRAS mutation and/or subtype and clinical stage at presentation. Moreover, there was no association between G12D mutation and the presence of lymph node metastases within the resected cohort, further suggesting that the deleterious effect of G12D is not mediated through its impact on stage at diagnosis, even in those with resectable disease.
The G12D subtype has previously been shown to be associated with poorer prognosis, however, the literature is somewhat conflicting [
12,
13,
14,
15,
17,
19,
20]. The strongest evidence for G12D being a marker of poor prognosis comes from studies of patients with operable disease, whereas the evidence for advanced disease is less clear cut. Our data support this. Since G12D is present in 40% of PDAC patients [
10], this finding has potentially important clinical implications. For patients who are considered anatomically resectable, but who are marginal surgical candidates due to age and comorbidities,
KRAS mutation assessment could be useful to guide treatment decisions. The fact that
KRAS mutation subtype can be reliably assessed via EUS-FNA enables this information to be available prior to considering surgery.
The reason for exploring
KRAS mutation status as a prognostic biomarker rather than any other potentially useful biomarkers, such as Ki67, is that these patients were being screened for potential inclusion in a clinical study of
EGFR inhibition in
KRAS wild-type pancreatic cancer [
22]. The recent introduction of targeted therapies directed at specific
KRAS subtypes highlights the fact that it will become increasingly important to be able assess prognostic and predictive biomarkers using EUS-FNA biospecimens. For example, targeting G12C has recently demonstrated significant anti-tumour activity in clinical trials, some of which included PDAC patients [
28]. Sotorasib, a G12C inhibitor, has been approved for clinical use in non-small-cell lung cancer, after promising phase I and II trials [
28,
29]. Another approach includes siG12D-LODER™, an siRNA which targets
KRAS loaded into a biodegradable polymetric matrix [
30]. Early phase clinical trials in PDAC have reported it is well tolerated, with promising signs of efficacy [
31]. Engineered inhibitory exosomes have proven to be effective in PDAC preclinical models [
32]. All these approaches rely on accurate assessment of the subtype status prior to surgery. We have convincingly demonstrated that EUS-FNA can be used to assess
KRAS subtype status in patients with PDAC.
There are a number of limitations in our study. These include that it is from a single center and has a relatively small sample size for survival analysis. The latter may have contributed to the failure to demonstrate an association between the presence of a
KRAS mutation and decreased survival (compared to wild-type disease), despite the positive findings of previous papers. In addition, it was difficult to solely compare G12D subtype survival to
KRAS wild-type. The relatively small sample size guided our choice to focus on
KRAS G12D as the most prevalent subtype and to compare this subtype to all other subtypes combined, including G12R, G12V and wild-type. A comparison between each of the various subtypes individually was not possible with this sample size and may have led to the problem of multiple testing. However, comparing G12D to non G12D includes all patients rather than just wild-type patients and is used in existing studies [
11,
19]. Further, we know that adjuvant therapy, lymph node status and surgical margins are associated with survival outcomes [
33,
34,
35], and the fact that none of these factors were associated with survival on multivariate analysis further demonstrates the lack of power in our study. Finally, the exact details of the chemotherapy received were not reported as many patients had been referred from other centers for diagnostic biopsy and returned to the referring center for chemotherapy. Furthermore, the chemotherapy regimens used were too heterogenous and, given the sample size, were unlikely to have provided statistically meaningful results when examined individually. Time to recurrence was also not reported as it is difficult to collect in retrospective studies, especially as many patients returned to their referring centers for ongoing treatment.
A potential perceived limitation of our study was that only 37 out of the 63 resected specimens had KRAS mutation analysis performed on the specimens themselves, with the rest having KRAS analysis conducted on EUS-FNA samples prior to surgery. It could be argued that due to the small volume of material obtained, EUS-FNA evaluation may not be able to reflect tumour heterogeneity as well as resection specimens. However, the presence of more than one KRAS mutation subtype within the same tumour has almost never been reported even in resection specimens, demonstrating that tumour heterogeneity is not a relevant consideration with respect to KRAS mutation subtype. This is probably because an activating KRAS mutation is known to occur early in the dysplasia-carcinoma pathway and is therefore likely to be present in virtually every malignant cell.
Given the conflicting results in the literature, further large cohorts may be required to definitively establish the nature of any subtle correlation between NCCN stage and/or prognosis and KRAS mutation subtype, particularly G12D, in PDAC. A meta-analysis assessing the association between specific KRAS subtypes such as G12D and prognosis in different stages of disease is also warranted given the potential importance of this prognostic information.