Systemic cytokine levels and subsequent risk of gastric cancer in Chinese Women
To whom correspondence should be addressed.
Although control of the host cytokine network is known to influence gastric cancer susceptibility, the specific inflammatory responses in gastric carcinogenesis remain unclear. We prospectively examined the relationships between gastric cancer risk and plasma levels of interleukin (IL)-1β, IL-6, IL-8 and tumor necrosis factor (TNF)-α in a nested case control study within The Shanghai Women’s Health Study. Two controls were matched to each case on the basis of age, menopausal status, and sample collection parameters. The associations between gastric cancer risk and tertiles of cytokine levels were estimated by odds ratios (OR) and 95% confidence intervals (CI) from conditional logistic regression, adjusting for education. During a median follow-up period of 4 years (range 0.1–8 years), 141 women developed gastric cancer and were matched to 282 cancer-free study participants. Elevated levels of plasma IL-6 were associated with an increased risk of gastric cancer (Ptrend = 0.04). Risk increased 70% (OR = 1.7; 95% CI 1.0, 3.0) for women in the highest tertile (>4 pg/mL) of IL-6 compared with those in the lowest tertile (<1.8 pg/mL). The association between gastric cancer risk and IL-6 was stronger after 4 years of follow-up (OR = 2.6 [95% CI 1.0, 6.7] for highest versus lowest tertile) compared with an OR of 1.4 (95% CI 0.7, 2.9) for those diagnosed within 1–4 years of follow-up. No associations were observed with the other pro-inflammatory cytokines examined, namely IL-1β, IL-8, and TNF-α. Systemic plasma IL-6 levels may inform long-term gastric cancer risk. This novel finding awaits confirmation in future studies with sequential plasma collection. (Cancer Sci 2011; 102: 1911–1915)
Gastric cancer has the second highest cancer mortality globally.(1) A strong risk factor for gastric cancer is persistent colonization in the stomach lining by the Class 1 gastric carcinogen Helicobacter pylori.(2)Helicobacter pylori chronically infects half the global human population.(3) Although chronic H. pylori infection induces specific local and systemic inflammatory responses, resulting in gastritis and inflammation-related comorbidities, only a small percentage of infected people develop gastric cancer.(4,5)
Bacterial and human host factors modulate the risk of gastric cancer development. In terms of bacterial factors, the virulence of the H. pylori can be identified by the presence of a 40-kbp region of chromosomal DNA termed “cytotoxin-associated (cag) pathogenicity island” (cag+ versus cag−). People infected with the cag+ strain have a more pronounced inflammatory response, as well as a higher risk of gastric adenocarcinoma, compared with those carrying the cag− strains of H. pylori. However, a large proportion of hosts colonized by the cag+ strain remain asymptomatic.(6) The key players in human host response that underlie H. pylori-associated gastric carcinogenesis remain unclear.
In vitro and in vivo studies have demonstrated the role of cytokines, key regulators of the immune system, in gastric carcinogenesis. We and others have shown that one of the human host factors associated with gastric cancer susceptibility is polymorphisms in genes encoding proinflammatory cytokines(7–9) that may act as proxies for the strength of inflammatory responses to H. pylori infection. Based on in vitro studies of H. pylori persistence, as well as comparison studies of cytokine levels between gastric cancer patients and controls, we hypothesized that higher systemic levels of certain interleukins (IL) may increase gastric cancer risk. We tested this hypothesis in a prospective cohort in which blood samples were collected at baseline prior to cancer diagnosis. To our knowledge, this is the first study into the relationships between gastric cancer risk and prediagnostic levels of IL-1β, IL-6, IL-8, and tumor necrosis factor (TNF)-α.
Materials and Methods
Study population. The study population is nested within the Shanghai Women’s Health Study (SWHS), a population-based prospective cohort, as described previously.(10) Briefly, 81 170 women aged between 40 and 70 years residing in seven urban communities in Shanghai, China, from March 1997 through to May 2000 were invited to participate in the study. Of these, 74 942 women were interviewed, resulting in a response rate of 92.3%. At recruitment, information on lifestyle factors, anthropometric measurements, and biospecimens were collected. Incident gastric cancer cases among cohort members were identified through active biennial follow-ups and record linkage with the population-based Shanghai Cancer Registry, the Shanghai Vital Statistics Unit, and Shanghai Resident Registry. The response rates were 99.8%, 98.7%, and 96.7% for the first (2000–2002), second (2002–2004), and third (2004–2006) in-person follow-up surveys. The Institutional Review Boards at the Shanghai Cancer Institute (China), National Cancer Institute (Bethesda, MD, USA), and Vanderbilt University (Nashville, TN, USA) approved the present study.
Cases. Of the 74 942 women in the cohort, 56 832 provided a blood sample. Exclusion criteria were previous cancer reported at the baseline interview or cancer diagnosis within 1 month of the date of sample collection. As of December 2006, 141 gastric cancer cases (International Classification of Disease version 9 codes 151.0–151.9)(11) were newly diagnosed among cohort members. Of these cases, 13 (9%) were classified as gastric cardiac cancers.
Controls. Two controls per case, resulting in a total of 282 controls, were randomly sampled without replacement from the cohort and matched to the case based on the following variables at sample collection: menopausal status, date (<1 month) and time of sample collection (morning/afternoon), age (<2 years difference), time interval since last meal (<2 h), and availability of plasma samples.
Questionnaire data and non-cytokine serological markers. Fruit and vegetable intake was quantified by trained interviewers using a comprehensive validated food frequency questionnaire representing 90% of foods consumed in Shanghai in 1996.(12) Smokers were classified as subjects who smoked at least one cigarette per day for more than 6 months. Alcohol drinkers were those who drank alcohol more than three times a week for 6 months. Self-reported information regarding aspirin/non-steroidal anti-inflammatory drug (NSAID) use and frequency during the past year was obtained at baseline. Self-reported information regarding medication use during the past 7 days and the past 24 h was obtained at the time of blood collection.
Helicobacter pylori infection and cag seropositivity was examined as the presence of IgG antibodies against whole cell or the CagA H. pylori antigens using ELISA (H. pylori Biohit ELISA kit; Biohit, Helsinki, Finland) and immunoblot assays (Helicoblot 2.0; Geneslab Diagnostics, Singapore), respectively.
Cytokine measurements. Blood samples were collected from study subjects using BD Vacutainer tube serum tubes (Becton-Dickinson, Franklin Lakes, NJ, USA) at the time of the in-person interview. A 10-mL blood sample was drawn into an EDTA tube. Samples were transported in portable insulated bags containing ice packs (at 0–4°C) and processed by centrifugation (standard centrifugation for plasma processing) within 6 h of collection. Plasma was stored at −70°C.(10) Samples were collected between September 14, 1997, and May 18, 2000, with a median collection date of June 17, 1998. Plasma concentrations of IL-1β, IL-6, IL-8, and TNF-α were measured simultaneously in 25-μL plasma samples using he high-sensitivity LINCOplex kit (Luminex xMAP Technology, Millipore, St. Charles, MO, USA) at the Vanderbilt Hormone Assay & Analytical Services Core (Nashville, TN, USA). The sensitivities of the IL-1β, IL-6, IL-8 and TNF-α assays, according to the manufacturer, were 0.1, 1.6, 0.2, and 0.14 pg/mL, respectively.
The absolute concentration of each cytokine (in pg/mL) in each sample was determined by relating the fluorescence intensity to a standard curve fitted by a four-point regression. Cytokine measurements were performed in duplicate. Intra-assay variability was assessed with 19 replicate/duplicates of quality control samples within a single assay/plate of 38 samples. Interassay variability across 12 assays was assessed by using the mean and standard deviation (SD) of two replicate samples each in 12 separate assays. The intra-assay coefficient of variation (CV) ranged from 3% to 17%, whereas the interassay CV ranged from 9% to 20%.
Statistical analysis. To accommodate the matched design, we fit conditional logistic regression (CLR) models to estimate odds ratios (OR) and 95% confidence intervals (CI) for the association of gastric cancer risk with IL-1β, IL-6, IL-8, and TNF-α. Models were further adjusted for education level (continuous). Covariates considered as potential confounders were self-reported antibiotic use within 1 year of blood collection, self-reported NSAID use and frequency during the past year, body mass index, and smoking history. None was included in the final model because they did not substantially alter the parameter estimates for the exposures (plasma cytokine levels).
Plasma cytokine levels were classified into tertiles based on the distribution among controls. Using the lowest tertile as reference, we assessed the OR for the second and third tertiles. Measurements below the detection limits of the assay (0.06, 0.1, 0.11, and 0.05 pg/mL for IL-1β, IL-6, IL-8, and TNF-α, respectively) were included in the lowest tertile. To assess linear trends in risk, ascending tertiles of cytokine levels were treated as ordered categories.
The stratum-specific OR of each cytokine and gastric cancer risk was estimated for early (i.e. case diagnosed 1–4 years from baseline and corresponding controls) or late duration of follow-up (i.e. >4 years). The heterogeneity of OR for different strata was tested using 1 d.f. Wald tests. Sixteen cases diagnosed within 1 year from blood collection were excluded in the stratum-specific analyses; inclusion of cases diagnosed within <1 year of blood collection did not significantly alter the parameter estimates (data not shown).
All P-values are two-sided and statistical analyses were performed using STATA 10.0 (Stata Corporation, College Station, TX, USA).
The median duration between blood sample collection and cancer diagnosis was 3.75 years (range 0.12–8.39 years). Helicobacter pylori prevalence was high (Table 1): specifically, 92% of controls were positive for H. pylori and 89% were seropositive for CagA strains. By virtue of matching, cases and controls had similar distributions of age at the time of interview/recruitment (median: 61 years) and menopausal status (79% postmenopausal). There was no significant difference in education level, alcohol intake, smoking history, fruit and vegetable intake, or H. pylori seropositivity between cases and controls (Table 1). Cases had a lower frequency of self-reported antibiotic use within 1 year of blood collection.
Table 1. Selected characteristics of the gastric cancer cases and controls, Shanghai Women’s Health Study
| College or higher||28 (9.9)||8 (5.7)||0.03|
| High school||59 (20.9)||26 (18.4)|
| Middle school||82 (29.1)||39 (27.7)|
| Elementary/less||113 (40.1)||68 (48.2)|
| Ever||5 (1.8)||2 (1.4)||1.0|
| Never||277 (98.2)||139 (98.6)|
| Ever||13 (4.6)||9 (6.4)||0.49|
| Never||269 (95.4)||132 (93.6)|
|Median no. servings of fruit and vegetables/week (range)||21 (6–35)||21 (7–28)||0.62|
|Antibiotic use within 1 year of blood collection|
| No||256 (90.8)||136 (96.5)||0.05|
| Yes||26 (9.2)||5 (3.5)|
|Helicobacter pylori seropositivity|
| No||22 (7.8)||5 (3.5)||0.14|
| Yes||260 (92.2)||136 (96.5)|
Table 2 summarizes the association of plasma levels of each cytokine and gastric cancer risk. Plasma levels of IL-1β were below the limits of detection in one case and three controls; IL-8 levels were below the limits of detection in one control. Women in the highest tertile (>4 pg/mL) of IL-6 were at increased risk of gastric cancer (OR = 1.7; 95% CI 1.0, 3.0) compared with those in the lowest tertile (<1.8 pg/mL). Similar results were observed when analysis was restricted to women who developed gastric cancer ≥1 year after plasma collection and/or after adjustment for medication use within 1 week prior to plasma collection (data not shown). The exclusion of 13 gastric cardia cancer cases produced similar results (e.g. lowest versus highest tertile of plasma IL-6 levels: OR = 2.0, 95% CI 1.1, 3.8; data not presented in tabular form). Analyses restricted to subjects seropositive for H. pylori infection yield similar results.
Table 2. Odds ratios and 95% confidence intervals for prediagnostic circulating interleukin (IL)-1β, IL-6, IL-8 and tumor necrosis factor-α and gastric cancer risk in Shanghai women
| 0.70–1.55||94/45||0.85 (0.52–1.39)|
| >1.56||92/45||0.87 (0.50–1.56)|
| 1.77–4.05||94/41||1.07 (0.61–1.87)|
| >4.06||93/60||1.73 (1.00–3.00)|
| 2.65–7.16||94/47||1.12 (0.67–1.87)|
| >7.17||94/53||1.41 (0.85–2.36)|
| 4.87–7.16||94/42||0.68 (0.41–1.15)|
| >7.17||94/43||0.74 (0.42–1.30)|
The magnitude of association with IL-6 increased with duration of follow-up. The OR comparing the highest and lowest tertile of IL-6 was 2.6 (95% CI 1.0, 6.7) after >4 years of follow-up compared with an OR of 1.4 (95% CI 0.7, 2.9) for those diagnosed during 1–4 years of follow-up (Table 3). However, the P-value for heterogeneity was not significant.
Table 3. Odds ratios and 95% confidence intervals for prediagnostic circulating interleukin (IL)-1β, IL-6, IL-8 and tumor necrosis factor-α and gastric cancer risk according to follow-up time in Shanghai women
| Medium||41/28||1.41 (0.71, 2.80)||40/15||0.74 (0.33, 1.68)|
| High||37/13||0.62 (0.25, 1.56)||45/28||1.58 (0.64, 3.88)|
| Ptrend|| ||0.55|| ||0.40|
| Medium||35/14||0.85 (0.38, 1.89)||49/20||1.19 (0.48, 2.95)|
| High||42/26||1.37 (0.66, 2.86)||42/29||2.56 (0.98, 6.71)|
| Ptrend|| ||0.41|| ||0.04|
| Medium||43/24||1.21 (0.58, 2.54)||39/21||1.15 (0.52, 2.55)|
| High||36/17||1.07 (0.49, 2.33)||52/27||1.44 (0.65, 3.21)|
| Ptrend|| ||0.88|| ||0.36|
| Medium||37/17||0.74 (0.34, 1.63)||42/24||1.07 (0.48, 2.29)|
| High||36/16||0.70 (0.29, 1.70)||49/21||0.95 (0.39, 2.26)|
| Ptrend|| ||0.41|| ||0.90|
No significant associations were observed with the other proinflammatory cytokines examined, namely IL-1β, IL-8, and TNF-α (Tables 2,3).
Persistent H. pylori infection is associated with a systemic immune response(13) that possibly contributes to the development of gastric adenocarcinoma. The results of the present study suggest that systemic plasma IL-6 levels may increase subsequent gastric cancer risk. Our findings could be suggestive of the concept that systemic IL-6 levels may be a biomarker for low-grade inflammation. Soluble inflammatory markers can stratify populations into groups that may benefit from intervention and these markers themselves are possible intervention targets. To our knowledge, this is the first report demonstrating that elevated IL-6 levels ≥4 years prior to cancer diagnosis are a risk factor for gastric cancer. Furthermore, consistent with in vitro and in vivo data of cytokines and numerous cancers, this is the first report of an association between prediagnostic systemic levels of cytokines with any cancer. The association between plasma IL-6 levels and gastric cancer risk is consistent with the model in which H. pylori infection induces a cascade of inflammatory reactions in the gastric mucosa that leads to gastric atrophy and carcinogenesis.(14) In humans, proinflammatory cytokines are known to accompany acute H. pylori infection. The expression levels of IL-1β, IL-8, and IL-6 were increased in gastric biopsies of 20 human volunteers who were experimentally infected with between 104 and 1010 colony forming units of H. pylori 2 weeks prior to cytokine measurements.(15) Little is known about the key players of chronic inflammation in gastric carcinogenesis. Our findings implicate the pertinent players in the host response as specific members of a class of regulators of the immune system called cytokines, specifically IL-6 in gastric cancer susceptibility. Our findings are consistent with the mouse model of IL-6 superfamily and gastric carcinogenesis.(16) Chronic H. pylori infection induces chronic inflammation, which includes a strong T-helper 1 cell type immune response, and thereby boosts the production of members of the IL-6 family and induces the subsequent downstream effects of disruption of the signal transducer and activator of transcription (STAT) 1/3 and protein tyrosine phosphatase (SHP2)–Ras–ERK signaling pathways that leads to gastric hyperplasia.(16) However, the findings of the present study require careful interpretation, as discussed below.
In case control and cross-sectional studies, IL-6 levels are elevated in the sera and plasma of individuals with gastric cancer,(17–20) as well as patients with other inflammation-related cancers, such as colon(20–22) and prostate(23–25) cancer. However, cancer cells and the surrounding microenvironment secrete inflammatory markers, including cytokines.(26) Our study design minimizes the possibility that the high cytokine levels we observed are a consequence of cancer. To exclude the possibility that our findings may be due to the presence of undetected gastric cancer or early gastric cancer, we assessed the associations by restricting analyses to those patients whose cancer was diagnosed after more than 1 year of follow-up. In addition, we stratified the associations between each cytokine and gastric cancer risk by follow-up time. Risk associated with IL-6 increased further with a longer duration of follow-up, suggesting that this association is unlikely to be due to reverse causation. However, this novel finding awaits confirmation in other prospective studies with sequential plasma collection.
A limitation to our findings is the possibility of a type 1 error due to our small sample size. A potential weakness of our findings is that the increased risk of gastric cancer was observed only with a threshold of higher IL-6 levels. We also explored non-linear associations between cytokine levels and latent factors of the four cytokines via principal components analyses; these alternative analyses did not yield additional insights. Similar significant associations were not observed with the other three proinflammatory cytokines evaluated (IL-1β, IL-8, and TNF-α). In the present study, plasma IL-6 levels in healthy women (controls) were correlated with IL-8, but not with IL-1β and TNF-α. These correlations are consistent with those observed in 60 Japanese gastric cancer patients.(27) Although the in vitro evidence for the role of these four cytokines in gastric carcinogenesis is approximately similar, limited data from cross-sectional studies indicate that preoperative circulating IL-1β(18) and IL-6(17–19) levels are elevated, whereas TNF-α levels are reduced,(18) in gastric cancer patients compared with non-diseased individuals. In the present study, the associations between IL-8 and TNF-α levels in relation to gastric cancer risk were not significant; however, the estimates appear to be consistent with the results of cross-sectional studies. The associations between IL-1β and gastric cancer risk, although not significant, were not consistent with the results of cross-sectional studies. However, the measurement error for plasma IL-1β levels was much higher (CV ∼20%) than that for the other three cytokines (CV <10%). In addition, the volatility of the cytokines levels due to external stimuli may differ; for example, the induction of IL-1β and TNF-α levels is influenced by exercise and the circadian rhythm, whereas IL-6 levels are not affected.(28) Another limitation of the present study is that plasma cytokine levels may not represent robust markers of inflammation in the gastric lining. Although we have accounted for known stimuli (e.g. smoking, obesity and anti-inflammation therapy), in our study design and/or statistical analyses we did not exclude certain, possibly common, non-specific chronic inflammation conditions that drive systemic cytokine levels, such as urinary tract infections(29) or chronic respiratory infections.
The present study has several strengths to address the hypothesis that proinflammatory cytokines predict gastric cancer risk. We capitalized on the prospective nature of our study design to minimize temporal ambiguity in the interpretation of our findings. In our high-risk population of more than 90% infected with H. pylori, we minimize confounding by non-specific H. pylori-related inflammation. In addition, the prevalence of other strong and common non-specific inflammatory stimuli is low in our population: tobacco use (2.4%), obesity (5.1%), and anti-inflammatory medication intake (Table 1). In terms of measurement error, we also used sensitive and quantitative methods to measure the low levels of plasma cytokines.(30,31)
Using gastric cancer as a model of inflammation and cancer, our approach of prospectively examining soluble inflammatory markers in relation to cancer susceptibility has the potential to elucidate the key players underlying the etiology of inflammation-related cancers.
The work was supported by a grant from the National Institutes of Health (R37 CA70867) and an Intramural Research Program contract (N02 CP1101066). The authors thank the Shanghai residents who participated in the study and the research staff of the Shanghai Women’s Health Study. The authors also thank Regina Courtney for plasma sample preparation at the Survey and Biospecimen Shared Resources supported, in part, by the Vanderbilt–Ingram Cancer Center (P30 CA68485).
None of the authors has any financial or personal relationships with other people or organizations that could inappropriately influence this work.