Meta-analysis: Circulating adiponectin levels and risk of colorectal cancer and adenoma

Authors


Zhi Hua RAN, Department of Gastroenterology, Renji Hospital, Shanghai Jiaotong University School of Medicine Shanghai Institute of Digestive Disease, 145 Shandong Zhong Road, Shanghai 200001, China. Email: ranzhihua62@yahoo.com

Abstract

OBJECTIVE:  To provide a systematic review with a meta-analysis for addressing the association between circulating adiponectin levels and the risk of colorectal cancer and adenoma.

METHODS:  Multiple electronic sources including MEDLINE, EMBASE and the Science Citation Index Expanded databases were searched to identify relevant studies for this systematic review. All existing observational studies that examined the relationship between circulating adiponectin and colorectal cancer or adenoma were included. Weighted mean difference and 95% confidence intervals (CI) were estimated and pooled using meta-analysis methods.

RESULTS:  Overall 13 case control or nested case control studies met the inclusion criteria. A total of 6175 participants and 3015 cases of colorectal cancer and adenoma were included in this meta-analysis. The weighted mean difference (95% CI) were −1.084 µg/mL (−1.836, −0.331), P = 0.005 in colorectal cancer and −1.43 µg/mL (−2.231, −0.628), P = 0.000 in adenoma. In men, a 2% decreased risk of colorectal neoplasm for a 1 µg/mL increment in adiponectin levels was observed (OR = 0.98, 95% CI 0.96–0.99) whereas among women there is no evidence of such a trend (OR = 0.99, 95% CI 0.97–1.01).

CONCLUSIONS:  Patients with colorectal cancer and adenoma demonstrated markedly lower adiponectin values than controls, yet there was significant heterogeneity among studies. A negative dose response relationship between levels of adiponectin and the risk of colorectal neoplasm was observed in men.

INTRODUCTION

Colorectal cancer is a common cancer worldwide. In the past several decades, much has been learned about the interaction of genetic and environmental factors for this malignancy. In recent years there has been emerging evidence indicating that obesity is associated with an increased risk of colorectal cancer compared with normal weight individuals but the mechanisms underlying this relationship remain to be fully elucidated.1,2

Experimental and epidemiological evidence has indicated that obesity leads to altered levels of several adipocytokines,3 which may further contribute to an increased risk of colorectal cancer. The adipose tissue produces various growth factors, hormones and cytokines, known as adipocytokines. The adipocytokines include leptin, resistin, visfatin, adiponectin and numerous cytokines including tumor necrosis factor (TNF)-α, interleukin (IL)-6, IL-8, IL-10 and IL-1 receptor agonist.4 Adiponectin (ACRP30, ADIPOQ, APM-1 or GBP28) is one of the most interesting adipocytokines associated with obesity. Recent research has found that adiponectin has antidiabetic,5 antiangiogenic,6 anti-inflammatory and antiatherogenic properties.7–9 The circulating adiponectin modulates several key physiological processes and has been recently reported to have anti-carcinogenic properties in various malignancies, including cancers of the breast,10 endometrium,11 stomach12 and prostate,13 and leukemia.14 Several studies have examined an association of circulating adiponectin levels with risk of colorectal cancer and adenoma.15–27 However, the results of numerous epidemiological studies have been inconsistent. The reasons underlying these heterogeneous findings need to be investigated.

To our knowledge, there has been no systematic review conducted that evaluates the associations of circulating adiponectin levels with colorectal cancer or adenoma. Hence, the purpose of this current review is to summarize all of the available data to provide a quantitative assessment of the association and to summarize the results in a meta-analysis.

MATERIALS AND METHODS

Search strategy

To identify all observational studies that examined the relationship between circulating adiponectin levels and colorectal cancer or adenoma, we conducted a literature search of the MEDLINE, EMBASE and the Science Citation Index Expanded, covering all articles published up to November 2010 without language restriction. The following terms were used for the search: ‘(adiponectin or ACDC or ADPN or APM1 or APM-1 or GBP28 or ACRP30 or ADIPOQ) and (colorectal or colon or rectal) and (cancer or carcinoma or neoplasia or tumor or adenoma or neoplasm)’. In addition, the reference lists of identified articles were also reviewed for additional pertinent studies. Abstracts of research presented at related conferences were also searched, including the World Congress of Gastroenterology and the Digestive Disease Week of the American Gastroenterological Association. This systematic review was presented in accordance to the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines.28

Inclusion and exclusion criteria

The studies considered for the meta-analysis were case control and cohort studies, which reported the relationship between circulating adiponectin levels and colorectal cancer or adenoma. Conference abstracts were also included if they had sufficient information to extract effect estimates. We excluded literature reviews, studies on animals or cell lines, studies of genetic variation in adiponectin-related genes and studies with no healthy control group. For duplicate publications, we included the study with the largest number of patients. We contacted the authors to obtain the relevant information when the data were insufficient or there was doubt about the publications.

Data extraction

The following information was extracted from each study: the first author's name, year of publication, country of population studied, participants' age and gender, study design, study size and number of cases, adiponectin measurement method, total adiponectin means and standard deviation (SD). The quality of each study was assessed with the nine-star Newcastle–Ottawa scale.29 Data extraction was conducted by the first author (XT XU) and the second author (Q XU) independently in a standardized manner and any disagreement was resolved by consensus. For four studies19,21,24,26 that provided outcomes in medians and interquartile ranges, the medians and interquartile ranges were converted to means and standard deviations, as described by Hozo et al.30

Statistical analysis

The primary outcome of adiponectin and colorectal cancer or adenoma risk was indicated as weighted mean difference (WMD) and the corresponding 95% confidence interval (CI). A fixed or random-effects model was used where appropriate. Two studies that reported only gender-specific results were entered separately for men and women analysis.19,26

To evaluate the dose response relationship between circulating adiponectin levels and colorectal neoplasm, we performed a dose response meta-analysis based on studies considering at least three categories of adiponectin levels. The method, developed by Greenland and Longnecker31,32 based on generalized least-squares trend estimation (GLST) analysis was used to back calculate and pool the risk estimates.

Statistical heterogeneity among studies was evaluated by using the Q and I2 statistics.29 Sensitivity analysis was performed to evaluate whether the results could have been affected significantly by repeating the meta-analysis but excluding one study at a time. We also conducted subgroup analysis by study size, study quality, geographical area, and participants' body mass index (BMI), gender and laboratory assay. An estimate of potential publication bias was assessed by visual inspection of funnel plots and further evaluated with the use of the Begg's adjusted rank correlation test and Egger's regression asymmetry test.33,34 We performed a nonparametric ‘trim and fill’ procedure to further assess the possible effect of publication bias in our meta-analysis. All statistical analyses were performed with Stata 11.0 (Stata, College Station, Texas, USA). P < 0.05 was considered statistically significant.

RESULTS

Description of the studies

A total of 390 records were identified during our initial search. Titles and abstracts were manually evaluated for relevance. In all 294 studies were excluded because they were either duplicated in different databases or irrelevant to the current analysis. Of the 96 reports selected for detailed evaluation, only 22 studies investigated the relationship between circulating adiponectin levels and colorectal cancer or adenoma. Five studies were excluded because circulating adiponectin was not measured in the healthy control group, four studies were excluded because they were published as abstracts only, with insufficient information or the identified abstracts were also reported in a full article. Finally, 13 studies, including a total of 6175 individuals, fulfilled our inclusion criteria. Among the included studies, 10 provided data on colorectal cancer and seven on adenoma, respectively. A flow diagram of the search process is given in Figure 1. Of the articles included, eight were case control, three nested case control and two cross-sectional in design. Among the 13 studies, seven were conducted in Europe, five in Asia and one in the USA. Table 1 summarizes the characteristics of the 13 studies included in the review. The quality assessment of all the published studies that evaluated the association between circulating adiponectin levels and colorectal cancer or adenoma is shown in Table 2.

Figure 1.

Selection of studies for inclusion in meta-analysis.

Table 1.  Characteristics of the selected studies in the systematic review
StudyCountryParticipants' age (years)GenderNo. of cases (tumor type)Participants (N)Adiponectin assayStudy design
  1. CC, colon cancer; CRA, colorectal adenoma; CRC, colorectal cancer; ELISA, enzyme-linked immunosorbent assay; F, female; M, male; RIA, radioimmunoassay.

Wei et al. 200515USA40–75M179 (CRC)535RIANested case control
Otake et al. 200516Japan≥40F and M51 (CRA)103ELISACase control
Lukanova et al. 200617Norwegian44 (median)M381 (CRC)762RIANested case control
Stocks et al. 200819Sweden50–60F and M95 (CRC) (F)277 (F)ELISANested case control
125 (CRC) (M)370 (M)
Fukumoto et al. 200818Japan47–58M656 (CRA)1304ELISACross sectional
Guadagni et al. 200921Italy36–80F and M90 (CRC)120ELISACase control
Erarslan et al. 200920Turkey≥40F and M23 (CRC)104ELISACase control
31 (CRA)
Kumor et al. 200922Poland50–65F and M36 (CRC)98ELISACase control
37 (CRA)
Yamaji et al. 201026Japan40–79F/M255 (CRA) (F)510 (F)ELISACase control
523 (CRA) (M)1003 (M)
Otake et al. 201025Japan≥31M51 (CRC)124ELISACross-sectional
47 (CRA)
Nakajima et al. 201024Japan63.6 (mean)F and M115 (CRC)374ELISACase control
72 (CRA)
Gonullu et al. 201023Turkey28–77F and M36 (CRC)73ELISACase control
Kemik et al. 201027Turkey37–74F and M126 (CC)164RIACase control
Table 2.  Assessment of study quality
StudiesQuality indicators from Newcastle–Ottawa scale
12345A5B678Score
  1. For nonrandomized studies, 1 indicates cases independently validated (if yes, one score); 2, cases are representative of population (if yes, one score); 3, community controls (if drawn from the same community as the reference group, one score; no score if drawn from a different source or selection of group was not described); 4, controls have no history of colorectal cancer, adenoma, diabetes mellitus, inflammatory bowel disease (if yes, one score); 5A, study controls for age, gender, and date of blood draw (if yes, one score); 5B, study controls for additional factor (if yes, one score); 6 ascertainment of exposure by blinded interview or record (if yes, one score); 7, same method of ascertainment used for cases and controls (if yes, one score); 8, non-response rate the same for cases and controls (if yes, one score).

Wei et al.200515YesYesYesNoYesYesYesYesYes8
Otake et al. 200516YesYesNoNoYesYesYesYesNo6
Lukanova et al. 200617YesYesYesNoYesNoYesYesNo6
Stocks et al. 200819YesYesYesNoYesYesYesYesNo7
Fukumoto et al. 200818YesYesYesNoYesYesYesYesYes8
Guadagni et al. 200921YesYesNoNoNoYesYesYesYes6
Erarslan et al. 200920YesYesNoYesYesYesYesYesNo7
Kumor et al. 200922YesYesNoNoYesYesYesYesNo6
Yamaji et al. 201026YesYesYesNoYesYesYesYesYes8
Otake et al. 201025YesYesNoNoYesYesYesYesNo6
Nakajima et al. 201024YesYesNoNoYesYesYesYesNo6
Gonullu et al. 201023YesYesNoYesYesYesYesYesNo7
Kemik et al. 201027YesYesNoNoNoYesYesYesNo5

Main analysis

The results of meta-analysis on the association between circulating adiponectin levels and colorectal cancer or adenoma are shown in Figure 2. As regards colorectal cancer, 11 data points were included in the meta-analysis and a remarkable association was observed in pooled analyses using a random-effects model (WMD −1.084 µg/mL; 95% CI −1.836, −0.331; P = 0.005), yet with significant heterogeneity among studies (I2 = 82.3%). Neither Begg's test (P = 0.06) nor Egger's test (P = 0.19) for publication bias provided evidence of significant effect. Eight data points were included in the meta-analysis on the association of adiponectin with colorectal adenoma. In these adiponectin in patients with colorectal adenoma was significantly different than in corresponding controls under random effects models (WMD, −1.43 µg/mL; 95% CI −2.231, −0.628; P = 0.000). However, statistical heterogeneity was present (I2 = 87%). Neither Begg's test (P = 0.11) nor Egger's test (P = 0.05) suggested a publication bias.

Figure 2.

Weighted mean adiponectin (WMA) values in colorectal neoplasm and matched controls and pooled estimate (random effects).

Dose response analysis

The GLST dose response analysis based on five studies15,17–19,26 included five data points for men and two data points for women. Overall, there is no evidence of a significant trend (OR = 0.98, 95% CI 0.97–1.00). According to studies reporting that circulating adiponectin levels of colorectal neoplasm suggested a possible modification of effect by gender, we fitted separate dose responses for men and women. In men, a 2% decreased risk of colorectal neoplasm for a 1 µg/mL increment in adiponectin levels was observed (OR = 0.98, 95% CI 0.96–0.99) whereas among women there was no evidence of such a trend (OR = 0.99, 95% CI 0.97–1.01). The model fit statistics revealed that these models fitted the data well and there was no statistically significant heterogeneity among the results of individual studies for both genders combined (Q-test = 18.45, P = 0.492), for men (Q-test = 14.23, P = 0.432) and for women (Q-test = 3.20, P = 0.524).

Subgroup/sensitivity analysis

We evaluated potential sources of heterogeneity between studies included: study size, study quality, geographical area, BMI, gender and laboratory assay. Low levels of circulating adiponectin in colorectal cancer and adenoma were consistently observed and there were no significant differences between colorectal cancer and adenoma. Associations tended to be stronger in smaller sized and lower quality studies than in larger sized and higher quality studies, respectively, indicating that the levels of the relative difference in study size and study quality between colorectal cancer and adenoma and the control group contributed significantly to the heterogeneity across studies. Associations did not differ substantially by geographical area, the participants' BMI, gender and laboratory assay. WMD in adiponectin across categories did not seem to be explained by these variables. The results are summarized in Table 3 and 4. A single study involved in the meta-analysis was deleted at a time and the pooled WMD for the remainder of the studies was calculated. This sensitivity analysis showed that no results were materially altered (Figs 3,4).

Table 3.  Stratified meta-analysis of circulating adiponectin levels and colorectal cancer
CharacteristicData points (N)Colorectal cancer (N)Random effects WMD (95% CI)P value effectI2 (%)
YesNo
  1. BMI, body mass index; CI, confidence interval; ELISA, enzyme-linked immunosorbent assay; I2, inconsistency index; RIA, radioimmunoassay; WMD, weighted mean difference.

All studies1113431653−1.084 (−1.836, −0.331)0.00582.3
Study size      
 <1005236168−2.893 (−4.601, −1.185)0.00154.8
 ≥100611071485−0.469 (−1.241, 0.302)0.23385.8
Study quality      
 Low (<7 score)6799615−2.079 (−3.607, −0.55)0.00888.9
 High (≥7 score)55441038−0.371 (−0.974, 0.231)0.22735.3
Geographical area      
 Asia2166141−1.775 (−5.398, 1.848)0.01284.2
 Europe89981156−1.233 (−2.327, −0.139)0.00085
 USA1179356−0.4 (−0.766, −0.034)
Adjusted for BMI      
 Yes5493772−0.6 (−1.902, 0.702)0.36765.8
 No3253432−1.822 (−3.780, 0.136)0.06881.8
 Unknown3597449−1.535 (−3.515, 0.441)0.12893.4
Gender      
 Male88901088−1.81 (−3.008, −0.613)0.00395.1
 Female5302423−1.59 (−2.854, −0.319)0.01465.0
 Combined2151140−2.70 (−8.373, 2.968)0.35088.4
Laboratory assay      
 ELISA8657878−1.475 (−2.78, −0.17)0.02772.6
 RIA3686775−0.875 (−1.948, 0.198)0.1193.5
Table 4.  Stratified meta-analysis of circulating adiponectin levels and adenoma
CharacteristicData points (N)Adenoma (N)Random effects WMD (95% CI)P value effectI2 (%)
YesNo
  1. BMI, body mass index; CI, confidence interval; I2, inconsistency index; WMD, weighted mean difference.

All studies816721608−1.43 (−2.231, −0.628)0.00087
Study size      
 <1005238225−2.66 (−3.785, −1.536)0.00046.7
 ≥100314341383−0.312 (−0.536, −0.089)0.0060
Study quality      
 Low(<7 score)4207175−2.901 (−4.215, −1.586)0.00046.6
 High(≥7 score)414651433−0.381 (−0.681, −0.081)0.01329.7
Geographic area      
 Asia616041533−1.35 (−2.22, −0.481)0.00290.2
 Europe26875−1.895 (−3.457, −0.334)0.0170
Adjusted for BMI      
 Yes4415404−1.918 (−3.801, −0.035)0.04687.4
 No412571204−0.723 (−1.397, −0.048)0.03675.9
Gender      
 Men412431181−0.68 (−1.35, −0.01)0.04774.5
 Women2269278−0.59 (−1.27, 0.09)0.0870
 Combined3160149−2.58 (−4.33, −0.84)0.00461.1
Figure 3.

Influence analysis in studies of circulating adiponectin and colorectal cancer.

Figure 4.

Influence analysis in studies of circulating adiponectin and colorectal adenoma.

Publication bias

We found the funnel plots were slightly asymmetrical in distribution, which raises the possibility of publication bias (Figs 5,6), although Egger's regression test suggested no significant asymmetry of the funnel plot (colorectal cancer: P = 0.19 and adenoma: P = 0.05). We further undertook analysis using the trim and fill method.35 The results from the trim-and-fill analysis did not change the summary estimate of effect.

Figure 5.

Begg's funnel plot with pseudo 95% confidence limits showing weighted mean difference (WMD) of adiponectin in colorectal cancer by standard error of WMD.

Figure 6.

Begg's funnel plot with pseudo 95% confidence limits showing weighted mean difference (WMD) of adiponectin in colorectal adenoma by standard error of WMD.

DISCUSSION

Some studies have shown that circulating adiponectin levels can be regulated by physiological, pharmacological and nutritional interventions.36,37 Keeping in mind the clinical importance of low levels of adiponectin for colorectal cancer and adenoma and that patients are likely to benefit most from strategies directed at increasing adiponectin production, as well as the inconsistent published results regarding the effect of adiponectin in this setting, this meta-analysis finding is of special importance.

In this meta-analysis of 13 observational studies, patients with colorectal cancer and adenoma demonstrated significantly lower adiponectin values, although there was obvious heterogeneity across the studies. As regards colorectal cancer, 11 data points, which comprised a total of 2996 participants including 1343 patients, were included in the meta-analysis, a significant association was observed in pooled analysis. Eight data points, which comprised a total of 3280 participants including 1672 patients, were included in the meta-analysis regarding the association of adiponectin with colorectal adenoma. Adiponectin in patients with colorectal adenoma was also significantly different compared to corresponding controls. A 2% decreased risk of colorectal neoplasm for a 1 µg/mL increment in adiponectin levels was observed in men, whereas among women there is no evidence of such a trend.

To explore the possible variables that explain why the results varied from study to study, we performed subgroup and sensitivity analysis to investigate the underlying causes. We observed that effect estimates were consistently stronger in small sized and low quality studies than those in large sized and high quality studies of colorectal adenoma, suggesting that the true effect of adiponectin may be weaker than indicated by the small sized and low quality studies. There was a weak correlation in large sized and high quality studies of adiponectin with colorectal cancer risk and showed no statistical significance as compared with those in the control group. Recently, Otake et al.25 showed that a low adiponectin level was strongly associated with an increased risk of colorectal adenoma and early cancer but not with advanced cancer. This may partly explain the result observed, and thus further studies involving patients with different cancer stage are needed.

Circulating adiponectin concentration is known to decrease with obesity. We conducted subgroup analyses to investigate obesity between-study variation in effect estimates. However, the association between BMI, a measure of general obesity, and risk of colorectal neoplasm seems not to be important in explaining the heterogeneity observed. Some studies reported that adiponectin may be a good biomarker of colorectal neoplasm independently from BMI;16,20 however, other studies suggest that visceral fat accumulation rather than BMI is associated with colorectal neoplasm.38,39 The influence of visceral fat could not be checked, as the visceral fat area or waist-to-hip ratio was not provided in most studies. More work is needed to clarify the association of visceral fat and circulating adiponectin with the development of colorectal neoplasm.

According to studies of colorectal neoplasm reporting circulating adiponectin levels by gender, the difference in circulating adiponectin levels between men and women may be substantial. It has been shown that the levels of adiponectin in men are significantly lower than in women.40,41 A recent study by Yamaji et al.26 reported that the inverse association between adiponectin levels and colorectal adenoma risk was confined to men. Because most studies reported the mean concentrations of circulating adiponectin for men and women combined, we asked the corresponding authors of studies to provide the mean concentrations for men and women separately. We conducted a subgroup analysis by gender and found that low adiponectin was not associated with colorectal adenoma in women, but a significant association was nevertheless observed in men and women with colorectal cancer. Given the very small numbers of studies available to date, more studies are needed to assess potential differences in the association between adiponectin with male and female risk in more detail. Insulin resistance has been shown to be associated with low adiponectin status.42,43 We could not pool these results, however, because each study adjusted for different markers.

Our study was the first to perform a meta-analysis that adds to the current understanding of the association between circulating adiponectin and colorectal cancer/adenoma. This analysis included a large number of cases and controls pooled from different studies, which significantly increased its statistical power. However, there are several potential limitations in our analysis. First, our meta-analysis is based on observational studies and is particularly vulnerable to the potential biases and confounding inherent in the original studies. The cohort study conducted by Lukanova et al.17 revealed negative results for cancer, but their observing term was very long. The cross-sectional study performed by Fukumoto et al.18 was also negative for adenoma. The participants of the study were mainly troops, which might have a different life style from regular people. Second, converting non-normally distributed statistics (median and range) to normally distributed statistics (mean and SD) may be less accurate than if individual data had been available. Furthermore, although this pooled study included 1343 colorectal cancer and 1672 adenoma patients, the sample size for some subgroup analyses was still small. The interpretation and conclusions made from the results of this meta-analysis should be regarded cautiously.

In summary, findings from this meta-analysis indicate that low circulating adiponectin levels are associated with a risk of colorectal cancer and adenoma although there was a significant heterogeneity across the studies. High adiponectin levels could decrease risk for colorectal neoplasm among men in a dose response manner. Most studies in this meta-analysis adjusted for colorectal cancer and adenoma risk factors. However, alcohol, food, exercise or medications may also alter adiponectin levels and residual confounding may still be present. Studies of the association between the adiponectin forms (trimers, hexamers and high molecular weight oligomers), the participation of the different receptors (AdipoR1 and AdipoR2), and adiponectin gene polymorphisms may be used in the future to overcome residual confounding.

ACKNOWLEDGEMENTS

We are very grateful to Dr Elife ERARSLAN (Fatih University Medical School, Ankara, Turkey), Dr Fiorella GUADAGNI (IRCCS San Raffaele Pisana, Rome, Italy), Dr Guzin GONULLU (Ondokuz Mayis University Medical School, Samsun, Turkey), Dr Hiroaki TAKEDA (Yamagata University, Yamagata, Japan), Dr Ozgur KEMIK (Yuzuncu Yil University Medical Faculty, Van, Turkey), Dr Par STATTIN (Umea University Hospital, Umea, Sweden), Dr Suminori KONO (Kyushu University, Fukuoka, Japan), Dr Taiki YAMAJI (National Cancer Center, Tokyo, Japan) and Dr Yasuhide YAMADA (National Cancer Center Hospital, Tokyo, Japan), who all provided the additional data requested on their publications. This research was supported by the outstanding academic leader's project, Science and Technology Commission of Shanghai (No.10XD1402600) and the National Natural Science Foundation of China (No. 81000929).

Ancillary