Circulating Resistin Levels and Risk of Colorectal Cancer: A Meta-Analysis

Objectives. Published data on resistin levels in patients with colorectal cancer (CRC) were conflicting and heterogeneous. We conducted a meta-analysis of observational studies to examine the association of circulating resistin levels with carcinogenesis of the CRC. Methods. Potentially eligible studies published up to November 2015 were searched through MEDLINE, EMBASE, Science Citation Index Expanded database, CNKI, and WanFang database. The pooled weighted mean differences (WMDs) with 95% confidence intervals (CIs) calculated by fixed- or random-effect model were used to estimate the effects. Results. A total of 11 studies involving 965 patients were admitted in our meta-analysis. The pooled effects indicated that resistin levels were higher in CRC patients compared to healthy controls (WMD: 1.47 ng/mL; 95% CI: 0.78 to 2.16), with significant heterogeneity across the studies (I 2 = 72%, p < 0.0001). Subgroup analyses and sensitivity analyses revealed that study quality, design, sample type, and resistin assays may account for this heterogeneity. No publication bias was observed. Conclusions. Our meta-analysis suggests that increased circulating resistin levels are associated with greater risk of colorectal cancer. Given the limited number of available studies and significant heterogeneity, larger well-designed randomized studies are warranted.


Introduction
Colorectal cancer is the third most common cancer and the third most common cause of high mortality cancer rates in the united states [1]. Although the incidence of CRC is 18% higher in developed regions in comparison with lowerincome areas [2], there is a rapid increase in developing countries in the past few decades, which makes it a major public health concern worldwide. CRC is recognized as a complex multipathway disease circled with chronic inflammation, metabolic syndrome, insulin resistance (IR), and obesity. The results from epidemiological evidence indicate that obesity is associated with increased incidence of CRC [3,4]. Recent studies suggest that adipocyte-derived secretory factors including adiponectin, leptin, resistin, visfatin, and numerous cytokines including TNF-alpha, IL-6, and IL-10 may be the key to link obesity with CRC, although underlying mechanisms remain unexplained [5][6][7][8].
Resistin, produced by adipose tissue whereas it is mostly secreted from monocytes and macrophages and involved in autocrine and paracrine cell signaling [9], has raised considerable research interest in recent years [10]. It is a cysteine rich 108-amino-acid peptide with a molecular weight of 12.5 kDa, mapping to the p13.3 band of human chromosome 19 [11,12] and circulating in low and high molecular weight forms [13]. As a member of resistin-like molecules (RELMs), which is a small family of secreted proteins associated with the activation of inflammatory [14], resistin can regulate inflammatory response, mediate metabolic disturbances and IR, promote cell proliferation, and enhance cancer progression [15], providing novel insights to obesity and colorectal cancer.
Although there are several systematic reviews evaluating the relationship between several adipocytokines and risk of CRC or CRA, including leptin and adiponectin [16][17][18][19][20], to the best of our knowledge, there is no meta-analysis on the associations of circulating resistin and CRC. In the past few years, 2 BioMed Research International some studies indicated a positive association between resistin levels and CRC [18,[21][22][23][24][25][26][27][28]; others failed to demonstrate such a relationship [29][30][31][32]. Seeing that controversial discrepancy, we summarized the evidence obtained from available data to assess resistin levels in patients with CRC.

Materials and Methods
Since we were evaluating observational studies, we conducted this meta-analysis in accordance with the Observational Studies in Epidemiology [33].

Inclusion and Exclusion Criteria.
Eligible cohort and case-control studies reporting the association between serum or plasma resistin levels and colorectal cancer were included in this meta-analysis. We excluded studies as follows: (1) reviews or nonoriginal studies; (2) studies on animal or cell or tissue or genetic variation; (3) studies without controls or original data; (4) studies related to mechanism or prognosis or survival.

Data Extraction and Quality
Assessment. The following information and data were extracted from each included study: (1) basic information such as first author's name, year of publication, and geographic location; (2) details of each trial, for example, study design and size, type of blood sampling, resistin detection method, and reagent; (3) characteristics of patients and controls including age, gender, BMI estimation, means, and standard deviation (SD) of resistin concentration. Two reviewers independently finished the extraction work and the disagreements were resolved by consensus or the third reviewer. Meanwhile, we systematically assessed the quality of each study with the Newcastle-Ottawa Scale (NOS) [34].

Statistical Analysis.
All statistical analyses were conducted with Stata 11.0 (Stata, College Station, Texas, USA) and Review Manager 5.3 software (Cochrane Collaboration). Weighted mean differences (WMDs) and corresponding 95% confidence intervals (CIs) were calculated based on the sample size, mean, and SD extracted from eligible studies. For five studies that only provided the median and range or interquartile range of outcomes, mean and SD were estimated, as described by Hozo et al. [35] and Liu et al. [36]. The Cochran Q statistic and inconsistency index ( 2 ) were used for examining the heterogeneity among studies, with the statistical significance level 0.05. Considering the presence of significant heterogeneity, we applied a random effects model for pooling effect, otherwise, a fixed-effects model.
In order to explore the sources of heterogeneity, we performed subgroup analysis by geographical area, study design, study size, quality, blood sample, resistin assay method, mean age, and mean BMI. Besides, several sensitivity analyses were conducted by omitting one study at a time, excluding estimated effect sizes and studies not adjusted for BMI. Publication bias was evaluated by funnel plot, complementary with Egger's regression method and Begg's rank correlation method.

Search Results and Studies' Characteristics.
A total of 122 potential records were detected by the primary database search and 78 were left after being autodeduplicated. By reading titles and abstracts for relevance, 54 reports were excluded because they were either review articles, mechanism studies, animal studies, or irrelevant to circulating resistin. Of the 24 records selected for detailed full-text evaluation, only 10 studies fulfilled our inclusion criterion. Three studies did not measure circulating resistin levels in healthy controls. Two trials were excluded because of one provided boxplot without data to describe distribution of serum resistin concentration and the other reported only odd ratio (OR) and 95% CI of the risk of colorectal cancer; we could not get primary data although we tried to contact the authors. Besides, a Chinese study that met the inclusion criterion was added. A flow diagram of selection strategy was presented in Figure 1.
Consequently, 11 studies including 965 patients with colorectal cancer and 1325 healthy controls were identified in this meta-analysis. Among the 11 studies, one was case-cohort study conducted in USA [31], while the other 10 studies were of case-control design conducted either in Asia [18,21,23,25,26,28] or in Europe [24,27,29,32]. Ten of the studies were matched for age and/or gender [18, 21, 23, 24, 26-29, 31, 32] and 6 of them were adjusted for BMI additionally [21,23,24,26,27,31]. The characteristics of each study were summarized in Table 1. The quality assessments of 11 case-control studies were presented in Table 2. The only case-cohort study [31] was finally awarded 7 stars based on Newcastle-Ottawa Scale (NOS) and specific assessment items were not listed.

Subgroups/Sensitivity
Analysis. We subsequently conducted subgroups and sensitivity analysis to trace the potential sources of heterogeneity. In subgroup analysis, the summary WMD of circulating resistin levels between CRC patients and healthy controls did not differ substantially by mean age, sample size, and mean BMI.  Table 3. Three of 11 studies mentioned the relationship between circulating resistin levels and TNM stages of colorectal cancer. The results of those studies were inconsistent (Table 4). Pooled mean differences suggested that circulating resistin levels might not be significantly different in CRC stage II-IV patients (Figure 3).
In sensitivity analysis, there was no significant difference after deleting a single study at a time and calculating pooled WMD of the remainders (Figure 4). Neither separate analysis of studies matching for BMI yielded significant associations of resistin levels and colorectal cancer risk. While removing studies for which the mean and SD were estimated [23,27,29,31,32], a stronger association between resistin and CRC was noted (WMD, 2.10; 95% CI, 1.25-2.95).

Discussion
The results of this meta-analysis suggested that resistin levels were notably higher in patients with CRC than those in healthy controls, indicating that resistin levels may be positively correlated with risk of colorectal cancer, although there was significant heterogeneity among the studies. The effects of estimates were consistent for studies regardless of their sample size, mean age, mean BMI, and whether being adjusted for BMI or not. No publication bias was observed in our meta-analysis. Keeping in mind the role of resistin in linking obesity to colorectal cancer risk, as well as the inconsistent published results regarding the effect of resistin in this setting, this meta-analysis is of special importance.
Although the role of resistin in colorectal cancer is far from being elucidated, several mechanisms may be involved in explaining these outcomes. In vitro studies demonstrate that resistin has proinflammatory effects mediated by TLR4 receptor stimulation and NF-Β pathway [37,38]. Other studies show that resistin regulates production of the matrix metalloproteinases (MMPs) and modulates the secretion of vascular endothelial growth factor (VEGF), which is considered of importance for promoting tumor invasion [39,40]. An animal study, where the mice were dealt with to lack adipocyte-derived mouse resistin but be able to produce human resistin resulting in rapidly accelerated white adipose tissue (WAT) inflammation after the high-fat diet, indicated that resistin contributes to WAT inflammation and insulin resistance [41]. Furthermore, RETN (rs1862513) or -420C>G resistin gene polymorphism is associated with decreased CRC susceptibility [29,42].
Recently, studies demonstrate that obesity is strongly related to CRC, which is involved in several mechanisms, such as obesity-related insulin resistance, oxidative stress [43], adipocytokine production [44], and insulin-like growth factor-1 (IGF-1) [45]. They are all responsible for cancer promoting effects, increasing tumor cell migration, and finally leading to tumor metastasis. The overexpression of resistin results in the accumulation of intracellular lipid and then leads to obesity-mediated insulin resistance and inflammation [46,47]. Therefore, it is widely accepted that high resistin levels are related to inflammatory conditions and malignancies [48,49].
However, we cannot ignore the significant heterogeneity across studies in the interpretation of the findings. To investigate potential sources, we performed subgroup and sensitivity analyses. Our study suggested that BMI may not be important in explaining the heterogeneity, since effect estimates showed no difference in different mean BMI subgroup and studies whether adjusting for BMI or not. However, the stronger and more detailed estimating factor for obesity including waist circumference and waist-to-hip ratio was not available in most involved studies, so effect of    is newly discovered, researches with large sample size are lacked. Third, the estimates of mean and SD from median and range are a cause of bias as described in sensitivity analysis results. At last, we did not rule out residual confounders that may alter resistin levels, including lifestyle parameters such as smoking, alcohol, exercise, nutritional habits, and clinicopathological features such as tumor stage, localization, and type.
In conclusion, our study suggests that circulating resistin levels in patients with CRC are significantly higher than those in healthy controls, increasing the current understanding of the association between resistin levels and risk of colorectal cancer. Given the limitations mentioned above, larger well-designed randomized studies and even experimental researches are needed to confirm the effect of resistin levels in the development of colorectal cancer.

Competing Interests
The authors declare no competing interests.