Pro-Inflammatory Adipokines as Predictors of Incident Cancers in a Chinese Cohort of Low Obesity Prevalence in Hong Kong

Background Cytokines released from adipose tissues induce chronic low-grade inflammation, which may enhance cancer development. We investigated whether indices of obesity and circulating adipokine levels could predict incident cancer risk. Materials and Methods This longitudinal community-based study included subjects from the Hong Kong Cardiovascular Risk Factors Prevalence Study (CRISPS) study commenced in 1995-1996 (CRISP-1) with baseline assessments including indices of obesity. Subjects were reassessed in 2000-2004 (CRISPS-2) with measurement of serum levels of adipokines including interleukin-6 (IL-6), soluble tumor necrosis factor receptor 2 (sTNFR2; as a surrogate marker of tumor necrosis factor-α activity), leptin, lipocalin 2, adiponectin and adipocyte-fatty acid binding protein (A-FABP). Incident cancer cases were identified up to 31 December 2011. Results 205 of 2893 subjects recruited at CRISPS-1 had developed incident cancers. More of the subjects who developed cancers were obese (22.1 vs 16.1%) or had central obesity (36.6 vs 24.5%) according to Asian cut-offs. Waist circumference (adjusted HR 1.02 [1.00-1.03] per cm; p=0.013), but not body mass index (adjusted HR 1.04 [1.00-1.08] per kg/m2; p=0.063), was a significant independent predictor for incident cancers after adjustment for age, sex and smoking status. 99 of 1899 subjects reassessed at CRISPS-2 had developed cancers. Subjects who developed cancers had significantly higher level of hsCRP, IL-6, sTNFR2 and lipocalin 2. After adjustment for conventional risk factors, only IL-6 (HR 1.51, 95% CI 1.18-1.95) and sTNFR2 (HR 3.27, 95%CI 1.65-6.47) predicted cancer development. Conclusions Our data supported the increased risk of malignancy by chronic low grade inflammation related to central obesity.


Introduction
The prevalence of obesity has increased dramatically in the last three decades globally [1]. In a World Health Organization report, over 400 million people were estimated to be obese in the association of obesity, in particular central obesity, with cancers has been increasingly recognized [3]. Obesity is associated with adipose tissue inflammation [4] and an increased production of the pro-inflammatory adiopkines such as leptin, interleukin-6(IL-6) and tumor necrosis factor-alpha (TNF-α), lipocalin 2 and adipocyte fatty acid binding protein (A-FABP) which have been shown to enhance tumorigenesis in different tumor cell line models and animal based studies [5,6,7,8] . On the other hand, adiponectin, an antiinflammatory adipokines with reduced expression in obesity, may be protective against the development of cancers such as breast, endometrial and colorectal cancers [9] . In humans, prospective epidemiological studies have also reported the association of high levels of interleukin-6(IL-6), tumor necrosis factor-alpha (TNF-α) and A-FABP [10,11,12,13], and low level of adiponectin [9], with cancer development. Whether the proinflammatory adipokines are associated with cancer development in the Chinese population, which has a lower prevalence of obesity and different body fat distribution when compared with Caucasians [14,15], remains unclear. In this study, the role of body adiposity and circulating levels of proinflammatory adipokines as predictors of incident cancers was addressed in a population-based cohort of Hong Kong Chinese.

Patients
The Hong Kong Cardiovascular Risk Factor Prevalence Study (CRISP) Cohort. The Hong Kong Cardiovascular Risk Factor Prevalence Study Cohort (CRISP) is a population-based prospective study on Chinese, mostly originated from Southern China. The study commenced as a cross-sectional survey in 1995-1996 (CRISP1) [16] with subjects recruited from the general population through random selections of telephone numbers. After baseline assessment, subjects were invited for prospective follow-up to assess the development of major cardiovascular diseases and cancers with subsequent assessments carried out in 2000-2004 (CRISPS2) [17]. Subjects attended all assessments at the Queen Mary Hospital after an overnight fast. At each assessment, demographic data, medical and drug histories, smoking and family histories of cardiovascular diseases and diabetes were obtained using a standardized questionnaire. Subjects were characterized for anthropometric parameters as previously described [16,17]. At CRISPS-2, plasma and serum samples were taken for the measurement of biomarkers including adipokines. Diagnoses of cancers, based on ICD-9 (140-165,170-176,179-209,230-239) and date of diagnosis were verified from the Hong Kong Hospital Authority database, with cut-off date of 31 December 2011. The prospective relationship between incident cancers over 16 years and obesity was assessed using baseline anthropometric parameters measured in 1995-6 (CRISP1). On the other hand, adipokine levels measured in 2000-4 (CRISP2) were used as baseline data in examining the relationship between 9.5-year cancer incidence and obesity-related adipokines. This study was approved by the Institutional Review Board of the University of Hong Kong / Hospital Authority Hong Kong West Cluster, and all participants gave written informed consent.

Statistical Methods
All statistical analyses were performed using Statistical Package for the Social Sciences (SPSS Version 16.0; SPSS Inc, Chicago). Baseline variables were compared using chisquare, partial correlation or analysis of variance as appropriate. Variables were tested for normality using Kolmogorox-Smirnov test, and skewed variables were naturallogarithmically transformed before analysis. Biologically relevant variables were entered into Cox proportional hazard regression analyses by forced entry to identify the independent predictors for incident cancers.  (Table S1 in File S1) with an incidence rate of 4.62 per 1000 person-years. The commonest type of cancer was carcinoma of the lung (44 subjects), followed by colorectal cancer (36), breast cancer (20), prostate cancer (18) and cancers of the female reproductive tract (14). The median time for cancer development from baseline was 8.7 years (interquartile range 4.9 to 13.2 years). Table 1 shows that, compared to subjects with no incident cancer on follow up, those who developed cancers were more likely to be male (p=0.004), older (p<0.001), current or ex-smoker (p=0.001), have higher body mass index (p=0.002), higher waist circumference (p<0.001, sex-adjusted), and hypertension (systolic blood pressure ≥140 mmHg and/or diastolic blood pressure ≥90 mmHg or on treatment [21]; p<0.001), type 2 diabetes ( p<0.001), and dyslipidemia (p=0.004) at baseline (1995-6; CRISP1). There was no significant difference in HOMA-IR (p=0.119), fasting insulin level (p=0.683) and alcohol use (p=0.678). More of the subjects who developed cancers were obese (p<0.001) or had central obesity (p<0.001) according to Asian cut-offs (BMI≥27.5 kg/m 2 for obesity and WC≥90cm in male or 80cm in female for central obesity [14,22]). If the general cut-off for obesity (BMI≥30 kg/m 2 ) was used, the prevalence of obesity was 8.3% and 6.1% (p=0.197) for cancer and non-cancer group respectively. After adjustment for potential confounding factors, including age, sex, smoking status, WC (adjusted HR 1.02 [1.00-1.03] per cm increase in WC; p=0.013), but not BMI (adjusted HR 1.04 [1.00-1.08] per kg/m 2 increase in BMI ; p=0.063), was found to be a significant independent predictor for incident cancers (
After adjustment for age, sex, and smoking status (Model 1 of Table 6), only baseline IL-6 and sTNFR2 remained to be independent predictors for incident cancers with hazard ratio (HR) of 1.51 (95% CI 1.18-1.95, p=0.001) per natural log unit increase in IL-6, and 3.27 (95% CI 1.65-6.47, p=0.001) per natural log unit increase in sTNFR2, respectively. Similar conclusion was reached when WC and/or diabetes status were also included as covariates (Model 2, 3 and 4 of Table 6). Since tumors are possible sources for inflammatory cytokines like IL-6 and TNF-α, analyses were repeated after excluding 8 patients who developed cancers within one year from baseline, to reduce the effect of a preclinical malignancy. Similar results were obtained with baseline IL-6 and sTNFR2 remaining as independent predictors of incident cancers, with HR of 1.53 (1.18-2.00, p=0.002) per natural log unit increase in IL-6, and 3.15 (1.53-6.47, p=0.002) per natural log unit increase in sTNFR2, respectively.

Discussion
There is now increasing evidence that obesity contributes to the development of many cancers, including cancers of colon, breast, liver, gall bladder, endometrium, kidney, prostate and pancreas [24,25]. A 5 kg/m 2 increase in body mass index (BMI) above normal was found to increase the risk of developing many types of cancers by 24 to 59% in both sexes [3]. Study in Asia-Pacific region also showed a 5-unit increase in BMI above 18.5 kg/m 2 raised all cancer-mortality by 1.09 times [26]. Although the prevalence of obesity has been shown to be lower in Chinese populations when compared with Western populations even when Asian-Pacific cut-offs for obesity are adopted [15], a rising trend, along with the incidence and mortality of many cancers, is observed and is believed to be related to the westernization of lifestyle [27]. Moreover, body fat composition in Chinese is different from their Caucasian counterparts and Chinese tend to have more central obesity, and thus visceral fat, for a given BMI after adjustment for age and sex [14]. Since the circulating levels of the proinflammatory adipokines are higher in the portal vein than in the peripheral arteries in obese subjects [28], visceral fat may play a more important role in contributing to systemic inflammation than the subcutaneous fat. In this Chinese community cohort with a relative low prevalence of obesity (16.5% and 6.2% having BMI ≥27.5 kg/m 2 and BMI≥30 kg/m 2 respectively), we demonstrated that waist circumference, an indicator of central adiposity, but not BMI which represents general obesity, was a significant independent predictor of incident cancer development, even after adjusting for the effect of age, the strongest risk factor in this cohort. This differential effect of central and general adiposity on cancer risk has also been demonstrated in other studies in the Asian-Pacific region.
In a case-control study of relatively lean Chinese population with mean BMI 21.9 kg/m 2 , an increase in waist-hip ratio increased the risk of prostate cancer by 3 folds while the risk incurred by BMI was not significant [29]. Again, in a metaanalysis of 30 cohort studies in Asian-Pacific region, WC but not BMI, was associated with increased mortality from pancreatic cancer [30].   Table 6. Cox proportional hazard regression models showing baseline IL-6 and sTNFR2 independently predicted incident cancers. Pro-inflammatory adipokines have been shown in vitro to play significant pathophysiological roles in different stages of tumorigenesis including cancer cell proliferation, angiogenesis and metastasis [6,31] . Apart from their direct effects, adipokines may also interact with sex hormones, sex-hormone binding globulin, insulin-like growth factor or the binding proteins to mediate the development of obesity-related cancers [32]. In this study, we demonstrated the significant correlations between indices of obesity and the baseline levels of pro-  inflammatory markers, among which IL-6 and sTNFR2 independently predicted incident cancer development. IL-6 and TNF-α are key cytokines involved in inflammation and immunity [33,34]. Both of them had been shown in vitro to be the links between chronic inflammatory states and cancers, by promoting cell proliferation, enhancing angiogenesis and cancer cell metastasis [35,36]. Epidemiological data also supported the role of IL-6 and TNF-α in cancer promotion. In a nested case-control study of 1298 postmenopausal women in the United States, Ho GY et al. demonstrated that subjects with IL-6 in highest quartile had a relative risk of 1.41 of developing colorectal cancer though the association was insignificant after adjustment for baseline insulin level [10]. Heikkila K et al. also showed IL-6 was associated with increased lung and breast cancer risks in a meta-analysis of two prospective cohorts, the British Women's Heart and Health study and the Caerphilly Cohort [11]. In addition, in a sub-group analysis of the Nurses' Health Study, Chan et al. showed that the relative risk of development colorectal cancer was 1.67(95% CI 1.05-268, p=0.03) in participants with highest quartile of sTNFR-2 level, when compared with those in the lowest quartile [12]. While our study was the first prospective study examining the effects of the adipokines on cancer development in Chinese population, our findings on IL-6 and sTNFR-2 were in line with those of western cohorts.

Model 1 Model 2 Model 3 Model 4 (Base Model b ) (Base Model + WC) (Base Model + Diabetes) (Base Model + WC +Diabetes)
Smoking, a well known carcinogenic risk factor, has been associated with level of both IL6 and TNF-α [37,38]. In the current study, smoking was more prevalent at baseline among those who subsequently developed cancers. In the multivariate analysis, however, smoking was not a significant independent risk factor of cancer development, possibly suggesting that the effect of smoking on cancer development is, at least in part, mediated by these proinflammatory cytokines.
The negative findings on adiponectin in this study was unexpected as this anti-inflammatory adipokine inhibited the growth of several breast cancer cell lines in vitro [39]. In epidemiological studies, hypoadiponectinemia had been shown to predict the development of breast, endometrial, colonic, renal cell carcinoma and many hematological malignancies [9]. The association between adiponectin and prostate or lung cancers, on the other hand, was less consistent [40,41]. This might explain why adiponectin did not emerge as a protective factor in our cohort as prostate and lung cancers constituted a significant proportion of our cancer cases.
This study has several limitations. Firstly, in view of the relatively small numbers of incident cancer cases, the findings of this study should be validated in studies involving larger sample size. Secondly, since baseline assessments were performed more than a decade ago and the list of adipokines is growing, not all adipokines were measured in this study. Thirdly, estradiol and its metabolites might be other important confounding factors which were not measured. Numerous studies have demonstrated the effects of estrogens on the development and progression of different cancers [23] and estrogen production in obesity state will be increased by peripheral aromatization in the adipose tissues. However, our subgroup analyses on obesity parameters and estrogen dependent cancers were limited by small numbers of events, resulting in insignificant power to make a definitive conclusion. Nonetheless, levels of CRP, IL-6, sTNFR2 were found to be higher in subjects with low estrogen levels (post-menopausal females, defined as women aged older than 55, and male subjects) when compared to female subjects aged less than 55 with presumably higher estrogen levels (Table S2 in File S1). Previous studies have also reported a suppressive effect of estrogen on both adipokines in high levels, and at low levels, no effect on IL-6 but stimulatory effect on TNF-α secretion [42]. In obesity, the increased estrogen production by the adipose tissue would be tend to lower the secretion of these adipokines and hence their effect on cancer development. Fourthly, other factors which may impact on cancer development such as family history of malignancy, lifestyle factors like dietary factors and physical activity, and personal history of chronic inflammatory diseases, were not included in the analyses. Furthermore, detailed data on concomitant medications which may impact on inflammation or adipokines levels, such as aspirin, statins, blockers of the renin-angiotensin system and thiazolidinediones, were not recorded. Finally, due to the limitation of an observational study design, our findings could only imply association but not causal relationship, though multiple molecular mechanisms have been identified to link cancer development with elevated cytokines in chronic inflammatory states.

Conclusions
In conclusion, we have demonstrated in a population-based study that high waist circumference, and raised serum levels of the pro-inflammatory adipokines, IL-6 and TNFR2, can predict the development of cancers in a Chinese cohort with relatively low obesity prevalence. These associations are supportive of the role of chronic low grade inflammation, related to central obesity, in the development of cancers even though a causal relationship cannot be concluded. When more epidemiological data become available and with a better understanding of the roles of these adipokines in tumourigenesis, risk models can be constructed to identify high risk patients for the development of malignancy so that appropriate measures may be employed to reduce their risks. On the other hand, in patients with established malignancy, the inflammatory cytokines level may act as novel prognostic factors. Furthermore, the inhibition of the inflammatory cytokines provides potential new strategies for the treatment of cancers or for sensitizing tumors to chemotherapeutic agents. While IL-6 and TNF-alpha receptor antagonists are now used clinically for the treatment of a wide range of rheumatological or hematological disorders, it will be of interest to follow-up on patients treated with such agents for any potential beneficial effects on cancer development.