Letter to the editor: Serum S100A6 as complementary biomarker for pancreatic cancer-an advance needing clearer delineation

Abstract

The study by Bae et al was the first prospective study to evaluate serum S100A6 as a tumor marker for early detection of pancreatic cancer (PC). The study was commendably well designed, included 414 patients, underwent bootstrap validation, and demonstrated that adding S100A6 to carbohydrate antigen 19-9 (CA19-9) and carcinoembryonic antigen enhanced detection of early PC compared with chronic pancreatitis (area under the curve = 0.821, P = 0.017). However, some issues need more exploration. The “early stage” group is composed of 140 stage II tumors and 9 stage I cases; thus, the results apply mainly to resectable PC, not stage I disease. Second, Lewis antigen status for second malignancy cases is not available, and it is unclear whether Lewis-negative cases are present in CA19-9–negative/S100A6–positive cases (41% of CA19-9 false negatives) or constitute a distinct subgroup. Third, the study was a single-center using samples from 2008 to 2015; so multicenter validation is needed. Fourth, the mechanism by which S100A6 enters the circulation needs detailed study. Fifth, clinical net benefit needs assessment using decision curve analysis. Collectively, there is evidence for serum S100A6 as a complementary biomarker for PC, but outstanding questions must be resolved before the marker can be used clinically.

Key Words: S100A6; Pancreatic cancer; Early detection; Carbohydrate antigen 19-9; Complementary biomarker; Chronic pancreatitis; Lewis antigen; External validation; Decision curve analysis

Core Tip: This letter commends the innovative study by Bae et al as the first large prospective study evaluating serum S100A6 for early detection of pancreatic cancer, and outlines five points requiring attention. The “early-stage” group consists predominantly of stage II; absence of Lewis antigen status information limits mechanistic understanding of S100A6’s complementary role; single center design and use of historical samples limit generalizability; absence of decision curve analysis means net benefit is unassessed; and how S100A6 enters the circulation is unidentified. Overall, serum S100A6 is promising, but these limitations must be addressed before clinical implementation.

TO THE EDITOR

We read with interest the article by Bae et al[1] published in the World Journal of Gastroenterology which was the first prospective, large-scale assessment of serum S100A6 for early detection of pancreatic cancer (PC). This study has several features of interest. It includes a well described cohort [414 subjects: 301 PC, 52 chronic pancreatitis (CP), and 61 healthy controls], uses pre-treatment samples, reports clear tumor, node, metastasis staging, performs 2000 bootstrap resamples, and demonstrates transparency regarding the limitations. The data reveal that the levels of serum S100A6 are higher in early-stage (I-II) and later-stage (III–IV) PC when compared to CP and healthy controls. The addition of S100A6 to carbohydrate antigen 19-9 (CA19-9) and carcinoembryonic antigen (CEA) further improves the ability to distinguish early-stage PC from CP [area under the curve (AUC) = 0.821, P = 0.017 vs CA19-9 alone][1]. As a whole, these results support the use of S100A6 as a valuable marker for complementary biomarker for PC. However, there are some things that need to be explored.

First, several considerations apply to the definition of “early-stage” in this study

As shown in Table 1, there are only 9 cases of stage I (6%) and 140 cases of stage II (94%) in the early-stage group (n = 149). The conclusions primarily apply to resectable (stage II) and not necessarily to truly early, asymptomatic disease. This is not a criticism, as stage I PC is uncommon in clinical practice. The goal of biomarker-based early detection is to identify disease at a curable stage, preferably before symptoms arise. However, as Ballehaninna and Chamberlain[2] noted, CA19-9 has no established role in screening asymptomatic populations and is unreliable for detecting small, surgically resectable cancers. Vitale et al[3] recently summarized that CA19-9 has significant limitations, low specificity, and is inefficient as a screening tool. Future studies should evaluate the S100A6/CA19-9/CEA panel specifically in stage I cohorts, requiring multi-center collaboration. Thus, the conclusions apply primarily to resectable (stage II) disease, and the performance of this panel in true stage I, asymptomatic PC remains to be established.

Table 1 Comparison of the original study’s “early-stage” cohort with an ideal screening population (asymptomatic stage I pancreatic cancer).

Characteristic	Original study “early” group (n = 149)	Ideal screening population (asymptomatic stage I)	Implication
TNM stage I	9 (6%)	100%	Results largely reflect stage II, not very early disease
TNM stage II	140 (94%)	0%	-
Symptom status	Mostly symptomatic (tertiary referral)	Asymptomatic	Unable to assess screening performance
Tumor size	Not reported, but stage II typically > 4 cm1	Usually ≤ 4 cm	Sensitivity for small tumors unknown

¹Based on general tumor, node, metastasis staging guidelines (stage II: T3–T4, tumor > 4 cm or extending beyond pancreas).

Data for the original study are from Bae et al[1]. The ideal screening population is defined as asymptomatic individuals with stage I pancreatic cancer detected through population-based screening programs; a scenario not represented in the original study. TNM: Tumor, node, metastasis.

Second, the complementary role of S100A6 in relation to CA19-9’s Lewis antigen dependency remains incompletely elucidated

Among the 96 CA19-9-false-negative PC patients (CA19-9 < 37 U/mL), 41% (39/96) were S100A6-positive[1]. This finding suggests that S100A6 could help fill the gap in the blind spot of the CA19-9 test. Table 2 presents a possible stratification according to Lewis antigen status. However, without Lewis genotyping, it is not possible to differentiate these individuals from Lewis-negative (who cannot produce CA19-9 at all) or to determine whether they represent a genuine distinct biological subgroup. Luo et al[4] have previously pointed out that Lewis antigen status should be considered when using CA19-9 as a biomarker. This is important to understand in order to fully appreciate the mechanism of complementary biomarker action. We hope that future validation studies include Lewis genotyping to determine whether S100A6 adds clinical value by identifying Lewis-negative patients or by a different biological mechanism.

Table 2 Hypothetical stratification of carbohydrate antigen 19-9-false-negative pancreatic cancer patients by Lewis antigen status to interpret S100A6 complementarity.

Subgroup of CA19-9 false-negative PC patients	Expected proportion	Hypothetical S100A6 positivity	Interpretation of complementarity
Lewis antigen-negative (cannot produce CA19-9)	5%-10% of general population1	Possibly high	S100A6 acts as a substitute, not a biological complement
Lewis antigen-positive (can produce CA19-9)	Remaining patients	Could be independent	True biological complementarity (different pathway)
Unknown (original study)	100%	41% (39/96)2	Cannot distinguish between the above two mechanisms

¹Prevalence of Lewis-negative (le/Le) genotype in most populations[2].

²Based on Bae et al[1], S100A6 positivity defined by Youden-index-derived cutoff (≥ 60.2 ng/mL).

This table presents hypothetical scenarios because Lewis antigen status was not assessed in the original study. The interpretation of complementarity depends on whether S100A6 positivity occurs predominantly in Lewis-negative individuals (substitution) or in Lewis-positive individuals (true biological complementarity). PC: Pancreatic cancer; CA19-9: Carbohydrate antigen 19-9.

Third, although the authors validated their model internally via bootstrap resampling, external validation in an independent multicenter cohort is the necessary next step

In the present study, samples were acquired from a single tertiary hospital in South Korea from 2008 to 2015. The extent to which the results can be generalized to Western and other ethnic populations is not established. In a multi-laboratory study by Haab et al[5], the authors demonstrated that, for reproducible biomarker performance, it is critical to have coordinated validation across multiple sites and the combination of CA19-9 with other biomarkers improved sensitivity from 0.44 to 0.71 at 0.98 specificity. Likewise, the IMMray PanCan-d test was performed at 11 sites in the United States and Europe, and showed 98% specificity and 85% sensitivity across the stages of pancreatic ductal adenocarcinoma I-II[6]. These precedents demonstrate that the promising results reported by Bae et al[1] now need to be replicated by independent multi-center studies.

Fourth, the molecular rationale underlying S100A6’s elevation in PC deserves further exploration in the serum context

Single-cell RNA-seq has revealed two distinct subpopulations of ductal cells in PC, including S100A6+ and FXYD2+ cells, and has demonstrated that S100A6+ ductal cells are enriched in high-risk groups of patients with poor prognosis[7]. These observations are at the tissue level and provide a compelling biological rationale. S100A6 could be actively secreted, leak out from dying tumor cells, or be carried by extracellular vesicles-or some combination thereof. Knowing these release mechanisms would aid in sample handling, assay standardization, etc. In addition to its role as a biomarker, S100A6 has been identified as a regulator of epithelial–mesenchymal transition and a promoter of migration and invasion via a β-catenin-dependent mechanism[8], which could explain the relationship between circulating levels and tumor biology. An intriguing issue is whether serum S100A6 is predictive of the percentage of S100A6-positive cells in the tumor microenvironment, which is likely to be influenced by the dense desmoplastic stroma and fibrosis characteristic of PC. This relationship would help determine if serum S100A6 is an indicator of overall tumor burden or a reflection of the biology of the tumor.

Fifth, the clinical utility of adding S100A6 should be assessed using decision curve analysis

A statistically significant improvement in AUC does not guarantee clinical net benefit. Decision curve analysis (DCA) quantifies net benefit across clinically reasonable risk thresholds and has become a standard tool for evaluating novel biomarkers[9,10]. In distinguishing early PC from CP, false positives could lead to unnecessary invasive procedures (e.g., endoscopic ultrasound with biopsy), while false negatives could delay curative resection. We encourage the authors to perform DCA on their bootstrap-validated models to determine whether adding S100A6 truly helps clinicians make better decisions.

In summary, Bae and colleagues present the first large-scale prospective evidence that adding serum S100A6 to CA19-9 and CEA enhances the detection of resectable PC, particularly in the challenging clinical setting of CP. The elevation of S100A6 in 41% of CA19-9-negative PC cases is a notably compelling observation that merits further investigation at the mechanistic level.

These results point to several future directions. External validation of the S100A6/CA19-9/CEA panel in multi-center cohorts that include diverse populations is a clear priority[5,6]. Other studies should also include genotyping of the Lewis antigen to further define the complementarity[4] and the performance of the panel in the context of stage I PC should be evaluated separately. Net clinical benefit needs to be quantified using DCA[9,10]. Lastly, the mechanism of S100A6 entering the circulation requires further study[7,8] and longitudinal studies are required to determine if the marker can predict or monitor disease progression. Future studies should also examine whether serum S100A6 levels differ according to tumor location (head vs body/tail). Direct comparison of serum S100A6 levels between resectable and unresectable patients would also be necessary. Longitudinal studies with pre- and post-resection samples are needed to assess changes in S100A6 levels after treatment.

Taken as a whole, the effort of Bae and cohorts is an important step in the research on PC biomarkers and serves as a solid foundation for future studies that can be validated. The study is well designed. We look forward to seeing whether serum S100A6 will eventually function as a meaningful complement to CA19-9 in clinical practice.

ACKNOWLEDGEMENTS

We are grateful for the original authors’ contribution to the field of pancreatic cancer diagnostics.