Circulating Inflammation Markers and Pancreatic Cancer Risk: A Prospective Case-Cohort Study in Japan

Pancreatic cancer is a highly fatal disease and accounts for more than 200,000 deaths worldwide every year (1). In Japan, pancreatic cancer is the fourth leading cause of cancer deaths with a crude mortality rate that increased from 21.3 in 2009 to 29.4/100,000 in 2019 (2). Age-standardized mortality rates are similar in Eastern and Western countries (3). Chronic pancreatitis, smoking, diabetes, and obesity are established risk factors for pancreatic cancer (1), which may implicate variant inflammatory processes in its etiology (4).

The tissue microenvironment impacts pancreatic tumor development, progression, and metastasis by both intracellular and extracellular components (5–7). Acellular influences include collagen, fibronectin, and multiple soluble factors such as cytokines, chemokines, and growth factors secreted by pancreatic stellate cells residing within the tumor and surrounding stroma (5, 8, 9).

Previous prospective studies of associations between circulating inflammation-related molecules and pancreatic cancer risk have included limited numbers of markers. Two nested case-control studies found that high C-reactive protein (CRP) concentrations were not associated with incident pancreatic cancer (10). Similarly, the EPIC studies did not find any significant associations of CRP and IL6 with pancreatic cancer (4). A pooled study of five US cohorts observed that associations between the risk of pancreatic cancer and pre-diagnostic levels of CRP, IL6, and TNF-R2 were not significant (11). A recent publication of the Swedish AMORIS cohort study reported an increased risk of pancreatic cancer with high prediagnostic serum levels of haptoglobin, CRP, and leukocytes (12). However, to our knowledge, no study has comprehensively evaluated a wide range of circulatory inflammation markers for associations with pancreatic cancer risk.

When a patient is diagnosed with pancreatic cancer, it is often too late for treatment. Pancreatic cancer is generally asymptomatic until it becomes advanced (9), at which point available curative treatment is limited. To probe potential mechanisms of carcinogenesis and evaluate utility for screening, we investigated the association of pancreatic cancer with inflammatory markers of chemokines, cytokines, and growth factors in a population-based prospective cohort in Japan.

The Japan Public Health Center-based Prospective Study (JPHC) Cohort II was initiated in 1993 to 1994. We defined our target popu--lation as Japanese inhabitants of six areas (Niigata, Ibaraki, Osaka, Kochi, Nagasaki, and Okinawa) across Japan aged 40 to 69 years at the baseline survey (n = 78,825). Details of the study design had been described elsewhere (13). This study conformed to the ethical guidelines of the Declaration of Helsinki. Although informed written consent was not obtained from each subject, the Institutional Review Board of the National Cancer Center, Japan (IRB No. 2004–059) and Fukushima Medical University, Japan (IRB No. Generic 2019–325) approved the study protocol.

A total of 23,335 participants answered the study questionnaire and provided 10 mL of blood sample during health checkups. Excluding those individuals with insufficient blood volume, missing information on cancer, fasting status or blood collection time, or those who had diagnosed cancer or moved out before blood collection, we followed 20,900 study participants until December 31, 2010. We identified 111 pancreatic cancer cases in total. We identified incident pancreatic cancer by voluntary reports from the local hospitals and by data linkage with population-based cancer registries, with permission. We used death certificates from the Ministry of Health, Labor and Welfare as supplementary information to identify cancer incidence. Changes in residential and vital status were obtained annually from the population registry of each study area. Cases were coded by using the third edition of International Classification of Diseases for Oncology (C25.0–C25.9; ref. 14).

A subcohort (n = 774) from a total of 20,900 participants was randomly selected and matched to the age and sex distribution of cancer cases. We had a total of 884 participants for analysis in this study including the subcohort participants and 110 additional pancreatic cancer cases from the rest of the cohort (Fig. 1). Diagnosis of pancreatic cancer in 82% (n = 91) was implemented by histologic or cytologic examination or clinical findings (images, tumor markers, etc.).

We collected blood specimens into vacutainer tubes with heparin when participants attended health checkups. These specimens were centrifuged within 12 hours and divided into plasma and buffy coat layers, preserved at −80°C until analysis.

We measured plasma levels of 67 inflammatory and immunity markers using a previously developed and validated assay with Luminex fluorescent bead-based multiplex panel (Millipore Inc.; ref. 15). We estimated the concentrations by interpolation on a four- or five-parameter curve fitted to those of seven calibration standards provided by the manufacturer. Samples of case and subcohort were randomly distributed on test plates. To monitor the reproducibility, each plate included masked samples were tested in triplicate (total n = 56), in two random wells on one plate and the third well on another. We calculated the ICCs using sum of squares from repeated measures analysis of variance as the ratio of between subject-variance over total variance. Two markers (IL3, TNFB) were excluded from the analysis due to levels detectable in less than 10% of samples. Three markers (IL7, CXCL8/IL8, and CCL20/MIP3A) were estimated both by conventional and high-sensitivity assays in different panels. We retained for analysis the respective assays with greater frequency of detection, which were the high-sensitivity result for IL7 and the conventional results for CXCL8/IL8 and for CCL20/MIP3A (16). Of the 62 remaining markers, we log-transformed the marker values and then computed coefficients of variations (CV). The CVs within batch were median 11% (all <30%) and between batch were median 29% (all but three <50%); intraclass correlation coefficients were median 96% (all but nine >80%; ref. 16).

Cox proportional hazards regression with the robust variance was used to estimate the HRs and 95% confidence intervals (CI) for the association between each marker and pancreatic cancer risk. Variance was estimated using the sandwich variance estimate computed using the observed information matrix and the matrix product of the dfbeta residuals; that is, the estimated change in model coefficients due to the deletion of each subject, calculated for all study subjects (17–19). Age was used as a time metric in models. For individuals in the subcohort, follow-up started from the date of blood draw until the diagnosis of pancreatic cancer or censoring, defined as movement out of the study area, death from any cause, or the end of follow-up on December 31, 2010, whichever came first. For cases that were not subcohort members, follow-up time was the 60-day period prior to pancreatic cancer diagnosis, an arbitrarily selected short period to ensure inclusion in only one risk set (17).

Markers were analyzed as quantiles based on the distribution among the subcohort, with the lowest quantile as the reference. Measurements of markers were categorized according to proportions of individuals below the lower limit of detection (LLOD). Markers with less than 25% of individuals below the LLOD were categorized into quartiles; markers with 25% to 50% of individuals below the LLOD were categorized as less than LLOD and tertiles of detectable levels; markers with 50% to 75% of individuals below the LLOD were categorized as less than LLOD and below and above the median; markers with 75% to 90% of individuals below the LLOD were categorized as undetectable and detectable.

HRs were adjusted for the potential confounding variables: age (years), gender, study area, family history of pancreatic cancer (yes or no), history of diabetes mellitus (yes or no), body mass index (<21, 21–<23, 23–<27, and ≥27 kg/m²), smoking habits (never, past, current <20, and current ≥20 cigarettes/day), alcohol drinking (none or occasional, regular <300, and regular ≥300 g ethanol/week), and quartiles of energy expenditure in metabolic equivalents/day. Missing data were treated as a separate category for all covariates. The quantiles were included as ordinal variables in the regression models to examine the linear trend of pancreatic cancer risk per quantile increase of marker levels. We only present results in detail for those markers with P_trend < 0.05. For multiple testing adjustment, we also calculated FDR corrected P values considering <0.05 as statistically significant. We repeated analyses stratified by latency (i.e., time from blood specimen collection to diagnosis) of <5 versus ≥5 years. SAS statistical software, version 9.4 (SAS Institute Inc.), was used for all analyses.

Table 1 shows the baseline characteristics of pancreatic cancer patients and subcohort participants of the JPHC study.

Figure 2 shows associations of pancreatic cancer risk per quantile increase for all the markers examined. Higher quantile of C–C motif ligand-8/monocyte chemoattractant protein-2 (CCL8/MCP2) was marginally associated with increased risk of pancreatic cancer while other markers had no associations. The coefficient of variation was 9.4% and the intraclass correlation coefficient was 0.9 for CCL8/MCP2. In addition, results stratified by gender, age group, or smoking habits were consistent showing no significant associations for each inflammatory marker. A complete list of pancreatic cancer associations for all 62 detectable markers is presented in Supplementary Materials and Methods (Supplementary Table S1).

Per quartile HR and 95% CI is shown for CCL8/MCP2 in Table 2. In the fully adjusted model, the HR in the highest quartile was statistically significantly doubled compared with the lowest quartile. Similar significant associations were observed for cases diagnosed <5 years but not for cases diagnosed ≥5 years after blood draw (Table 2).

We did not find any statistically significant associations between the markers and pancreatic cancer risk after correction for multiple testing.

This is the first comprehensive study to explore associations of a wide range of circulating inflammation markers with pancreatic cancer risk. We observed a suggestive positive association between CCL8/MCP2 levels and risk of pancreatic cancer. Although the CCL8/MCP2 association was not statistically significant after correction of multiple testing, this marker might be a candidate for future research.

Acute inflammatory responses produce a protective microenvironment in tissue to recognize and repair cell damage, but chronic inflammation may boost tumor formation (20). In response to tissue injury, mononuclear phagocytes are guided by chemotactic factors (21). Pro-inflammatory cytokines secreted by immune and inflammatory cells are functionally important components of the tumor microenvironment (22). Chemokines represent the largest family of cytokines and the interacting signaling networks influence tumor cell growth, survival, and migration and also angiogenesis and immune-cell infiltration (23, 24).

In this study, we found a nonsignificant positive association of circulating CCL8/MCP2 level with pancreatic cancer risk. CCL8/MCP2 binds CC receptor 2 and CCR5 in common with other chemokines (23). A recent in vitro study demonstrated that through directly targeting CCL8, miR-345–5p plays a critical role in suppressing cell proliferation and metastasis of pancreatic cancer (25). Thus CCL8/MCP2 might be an appropriate candidate for future studies.

In cross-sectional studies comparing with healthy controls significant higher levels of TNFα (26, 27), IL6 (26–28), and IL8 (26, 28) have been observed in patients with chronic pancreatitis or pancreatic cancer. The CRP level was reported to be an indicator of the aggressiveness of advanced pancreatic cancer (29).

These findings led to epidemiologic studies investigating whether the levels of these inflammation markers among healthy individuals could be a predictor for incidence of pancreatic cancer. In previous prospective studies, pre-diagnostic circulating CRP (4, 10, 11), IL6 (4, 11), and TNF-αR2 (11) were not associated with pancreatic cancer risk, as seen in our study, except for one study reporting an association of CRP (12). The roles of inflammation markers in pancreatic carcinogenesis are still not clear.

One strength in this study is the use of a multiplex analytical method that has been extensively validated (30–32). Also, our blood samples collected long before diagnosis are unlikely to reflect occult pancreatic cancer. Further, the population-based design and multivariate analysis incorporating established risk factors make the study results more robust.

One of the limitations in our study was that we had only a single plasma sample from each study participant, and it is uncertain to what extent an individual's concentrations vary over time (16) during the long lag between blood sample collection and pancreatic cancer incidence. If the stability of the concentration was low, we may have underestimated the association with pancreatic cancer risk. Second, the number of pancreatic cancer cases is limited for this study despite the large number of markers evaluated. Furthermore, due to lack of relevant information, our analysis could not adjust for any nonsteroidal anti-inflammatory drug therapies or other medications that might affect levels of circulating markers and/or pancreatic cancer risk (33, 34).

In conclusion, we observed no association between the concentration of 62 inflammation markers and incident pancreatic cancer in older Japanese adults. Similar studies in other racial/ethnic populations are needed to further explore the roles of local and systematic inflammation in tumor development and evaluate profiles of inflammatory markers, individually or synergistically, in pancreatic cancer risk.

N. Sawada reports grants from National Cancer Center Research and Development Fund during the conduct of the study. T. Yamaji reports grants from Ministry of Health, Labour and Welfare of Japan during the conduct of the study. No disclosures were reported by the other authors.

For information on how to submit an application for gaining access to JPHC data and/or biospecimens, please follow the instructions at https://epi.ncc.go.jp/en/jphc/805/8155.html.

E. Ma: Writing–original draft, writing–review and editing. T. Shimazu: Data curation, formal analysis, writing–original draft, project administration, writing–review and editing. M. Song: Formal analysis, writing–review and editing. H. Charvat: Data curation, writing–review and editing. N. Sawada: Data curation, project administration, writing–review and editing. T. Yamaji: Data curation, writing–review and editing. M. Inoue: Data curation, writing–review and editing. M.C. Camargo: Formal analysis, writing–review and editing. T.J. Kemp: Data curation, writing–review and editing. R.M. Pfeiffer: Formal analysis, writing–review and editing. L.A. Pinto: Project administration, writing–review and editing. C.S. Rabkin: Conceptualization, formal analysis, supervision, methodology, writing–review and editing. S. Tsugane: Conceptualization, data curation, formal analysis, supervision, project administration, writing–review and editing.

We are indebted to the Aomori, Iwate, Ibaraki, Niigata, Osaka, Kochi, Nagasaki, and Okinawa Cancer Registries for providing their incidence data. JPHC members as of May 2021 are listed at the following site: https://epi.ncc.go.jp/en/jphc/781/8510.html. This study was supported by the Japanese National Cancer Center Research and Development Fund [Grant No. 23-A-31 (toku), 26-A-2, and 29-A-4 (since 2011); Principal Investigator, S. Tsugane], and a Grant-in-Aid for Cancer Research from the Ministry of Health, Labour and Welfare of Japan (from 1989 to 2010) (Principal Investigator, S. Tsugane). This study was also supported by the Intramural Research Program, Division of Cancer Epidemiology and Genetics, U.S. National Cancer Institute, National Institutes of Health, Department of Health and Human Services (Grant No. ZIA-CP010212, Principal Investigator, C.S. Rabkin). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.