Low adiponectin levels are associated with renal cell carcinoma: A case-control study
Article first published online: 4 JAN 2007
Copyright © 2006 Wiley-Liss, Inc.
International Journal of Cancer
Volume 120, Issue 7, pages 1573–1578, 1 April 2007
How to Cite
Spyridopoulos, T. N., Petridou, E. Th., Skalkidou, A., Dessypris, N., Chrousos, G. P., Mantzoros, C. S. and the Obesity and Cancer Oncology Group (2007), Low adiponectin levels are associated with renal cell carcinoma: A case-control study. Int. J. Cancer, 120: 1573–1578. doi: 10.1002/ijc.22526
- Issue published online: 30 JAN 2007
- Article first published online: 4 JAN 2007
- Manuscript Accepted: 20 NOV 2006
- Manuscript Received: 10 JUL 2006
- Athens University Medical School, Greece
- Harvard Medical School, Boston
- Humboldt Foundation
- renal cell carcinoma;
- intra-abdominal obesity
Adiponectin is a novel endogenous insulin sensitizer, secreted by mature adipocytes. Circulating levels of adiponectin are inversely associated with obesity and insulin resistance. Because obesity is a risk factor for renal cell carcinoma (RCC), we hypothesized that low adiponectin levels are associated with RCC. To evaluate this hypothesis, we conducted a case- control study of 70 patients with histologically confirmed RCC and 280 healthy controls matched by gender, age and county of residence. Study subjects were interviewed and blood samples were collected during a 32-month period in Athens, Greece. Serum adiponectin levels were statistically, significantly and inversely associated with RCC when compared with controls (OR = 0.76, p = 0.05) and this association remained practically unchanged after controlling for BMI; the introduction of waist to hip ratio along with adiponectin in the multiple logistic regression analysis model rendered the association between adiponectin and RCC risk insignificant, indicating that altered levels of adiponectin may mediate the effect of central or intra-abdominal obesity on RCC. Prospective studies as well as studies exploring underlying mechanisms are needed to fully explore the role of adiponectin in predicting future risk of RCC in humans. © 2006 Wiley-Liss, Inc.
Obesity, and more importantly central obesity, is closely associated with insulin resistance.1, 2, 3 Adiponectin, a 247 amino acid protein with properties of an endogenous insulin sensitizer, was discovered in the mid-1990s4 and has been widely studied in relation to many disease entities associated with obesity and insulin resistance.5, 6
Adiponectin is produced and secreted exclusively by differentiated adipocytes. It is one of the most abundant serum proteins and circulates in tightly associated trimers and higher order oligomers.4, 7, 8 Adiponectin decreases with increasing overall and central adiposity and increases with prolonged weight reduction4, 9, 10; thus, it is currently thought that adiponectin represents a biochemical link between obesity, insulin resistance, and various other metabolic abnormalities associated with obesity.7, 11 Consistently, interventional studies have shown that adiponectin improves insulin sensitivity12, 13 and lipid profile,14 and suppresses inflammatory mechanisms involved in atherosclerosis.15 Finally, there is emerging mechanistic and epidemiological evidence supporting a protective role for adiponectin in obesity-associated malignancies.16, 17, 18, 19
Obesity is a known risk factor for renal cell carcinoma (RCC), but the molecular pathways underlying this association remain to be elucidated.5, 20, 21, 22, 23, 24, 25 We have hypothesized that decreased adiponectin levels may mediate the effect of central or intra-abdominal visceral obesity on RCC. To evaluate this hypothesis we have conducted a case-control study of 70 patients with RCC and 280 healthy controls.
Material and methods
During a 32-month period (October 2002 to May 2005), 70 cases of histologically confirmed RCC were diagnosed, newly presented with signs and/or symptoms in 4 hospitals in Athens and were recruited in the study by the study investigators. Seven cases of transitional epithelium cancer and three with other renal malignancies that were diagnosed during the same period were not included in the study, because these disease entities may have different etiology and pathophysiology than RCC. Thirty cases (42.9%) were Stage I, 17 (24.3%) Stage II, 15 (21.4%) Stage III and 8 (11.4%) Stage IV disease, whereas none of the cases included in the study had ever been previously diagnosed with any other type of cancer.26 In addition to the overall tumor staging, we used the TNM (Tumor-Node-Metastasis) system (updated version 1997) for a more detailed pathologic staging, the Fuhrman scale for tumor grading and we also classified tumors according to their histological subtype.27, 28, 29
For each one of these patients, 4 eligible controls matched for gender and of approximately the same age (±5 years) and place of residence were also enrolled consecutively in the study. Controls had a negative self reported medical history for cancer, hepatic disease, major hormonal or hematological disorders, asthma, autoimmune disease, HIV infection, advanced heart failure, chronic kidney failure, recent myocardial infarction, stroke, acute pancreatitis or bone fracture. Five potential controls refused or were unable to collaborate and were properly substituted for a total of 280 matched control subjects.
The study protocol was approved by the University of Athens Medical School Ethics Committee and the study design and performance were in accordance with the Helsinki Declaration of 1975. All participants provided written informed consent.
Each case and the 4 matched controls were interviewed by one of 4 trained interviewers. The interview lasted about 30 min and information pertaining to socio-demographic, anthropometric, lifestyle and medical history variables was obtained. Anthropometric variables were measured by specially trained health professionals (height, with subjects wearing no shoes; waist and hip circumference, with subjects wearing no clothes) as previously described.16, 18, 30 In addition, blood samples were taken from all cases and controls (no later than 9 a.m.) for hormonal measurements and for the determination of fasting glucose levels.
All coded samples, blinded as to case control status, were centrifuged and stored at −70°C, prior to being air shipped with dry ice to the Beth Israel Deaconess Medical Center, in Boston USA. Serum adiponectin levels were measured in 1 batch by trained technicians utilizing a radioimmunoassay procedure with a sensitivity of 2 ng/ml and intra-assay coefficient of variation of <10%, as previously described.16, 17, 18 Average preservation time was similar for cases and controls, although adiponectin levels do not systematically change with storage time.31
To investigate how adiponectin levels and the somatometric variables correlate with tumor characteristics, we performed analysis of mean values of adiponectin, weight, Body Mass Index (BMI) and Waist to Hip Ratio (WHR) by tumor stage, grade and histological subtype. For the case control statistical analyses, representative values (mean, SD, percentiles) of the anthropometric variables and serum adiponectin levels were calculated. We, subsequently, evaluated the correlation of serum adiponectin with the somatometric variables to identify possible predictors of adiponectin levels among healthy individuals. To study the association of adiponectin with RCC, we modeled the data through multiple conditional logistic regression analysis using case control as the outcome variable and adiponectin (in increments of 1 standard deviation of the hormone among controls), as well as a series of possible confounders as predictor variables. Potential confounders considered in the study were BMI at the time of diagnosis and 2 months before the time of diagnosis (in 2 kg/m2 increments), WHR (in 5% increments), education, smoking, coffee alcohol consumption, diabetes mellitus status (self reported history of diabetes mellitus or fasting glucose level higher than 125 mg/dl) and weight change during the last 2 months.
To further explore the possible effect of obesity we run an additional analysis restricted to obese (BMI ≥ 30 kg/m2) patients and controls (n = 25 and 87, respectively). Using unconditional logistic regression the risk (OR) for RCC in relation to adiponectin was calculated with and without adjustment for the pre-defined potential cofounders, as well as for age and gender. The SAS statistical package was used in all analyses and the level of statistical significance was set at 0.05.32
In Table I data concerning somatometric variables and adiponectin levels of the 70 RCC patients are presented tabulated by tumor stage, grade and histological subtype. No statistically significant association of adiponectin with any of the studied variables was observed. Nevertheless, it was noticed that mean serum levels of adiponectin were decreasing with increasing tumor grade.
|Variable||N||Adiponectin||BMI||Waist to hip ratio||Weight|
In Table II we present data on demographic and lifestyle variables as well as presence of diabetes for the 70 patients with RCC and the 280 gender-age- and residence matched controls. These data serve mostly descriptive purposes and are not directly interpretable because of mutual confounding. However, they reveal most of the established risk characteristics of RCC. Thus, patients with RCC were statistically significantly less educated (p = 0.02), they smoked and drank coffee more frequently, they had more frequently positive history of diabetes mellitus and had more pronounced weight loss during the 2 months prior to presentation when compared with their respective controls. Controls, on the other hand, appeared to consume alcohol on a more regular basis. Both total abstinence, but also over-consumption of alcohol, was positively associated with RCC in comparison to drinking 1 alcoholic drink per day.
|Yes or ex-smoker||43||61.4||160||57.1|
Table III shows representative values of somatometric variables and serum adiponectin levels by case-control status. Mean values of height, weight, BMI and hip circumference were not significantly different between cases and controls (p-values 0.30, 0.72, 0.69 and 0.55, respectively). Cases had more pronounced weight loss than controls (mean values −1.0 and 0.1 kg, respectively). Mean values of waist circumference and WHR were higher among cases than among controls (p-values 0.003 and 0.0001 respectively), indicating more pronounced intra-abdominal obesity among cases. Conversely, mean serum levels of adiponectin were higher among controls, when compared with those of patients (p = 0.06). Finally, no differences of mean values of adiponectin serum levels were observed between localized (9.62 μg/ml) and metastatic cases (9.26 μg/ml).
|Variable||N||Min value||25%||Median||75%||Max value||Standard deviation||Mean||p-value t-test|
|Weight change(kg): last 2 months|
|Waist circumference (cm)|
|Hip circumference (cm)|
|Waist/Hip ratio (%)|
We then examined associations between serum adiponectin and somatometric variables as well as age among healthy controls and the Pearson correlation coefficients are presented in Table IV. Adiponectin was negatively correlated with obesity, especially central obesity. Again, these results may not be directly interpretable because of the possibility of mutual confounding.
|Variable||Height||Weight||BMI||Waist circum.||Hip circum.||Waist to hip ratio||Adipo-nectin|
|Age||−0.24 (0.0001)||−0.04 (0.47)||0.15 (0.01)||0.23 (0.0001)||0.07 (0.22)||0.24 (0.0001)||−0.01 (0.80)|
|Height||0.55 (0.0001)||−0.10 (0.11)||0.23 (0.0001)||0.01 (0.95)||0.27 (0.0001)||−0.28 (0.0001)|
|Weight||0.78 (0.0001)||0.71 (0.0001)||0.54 (0.0001)||0.38 (0.0001)||−0.30 (0.0001)|
|BMI||0.73 (0.0001)||0.68 (0.0001)||0.26 (0.0001)||−0.15 (0.008)|
|Waist circumference||0.62 (0.0001)||0.65 (0.0001)||−0.24 (0.0001)|
|Hip circumference||−0.18 (0.0001)||0.00 (0.99)|
|Waist to hip ratio||−0.30 (0.0001)|
Subsequently, multiple logistic regression analyses were performed and the odds ratio for RCC for each increment of adiponectin, corresponding to 1 standard deviation among controls were determined (Table V). The odds ratios were derived from 5 different models. The crude odds ratio for adiponectin (Model 1) was 0.76 (p-value = 0.05). The statistically significant inverse association between serum adiponectin and RCC remained essentially unchanged (OR = 0.75, p-value = 0.05) after controlling for BMI at the time of diagnosis (Model 2) and for BMI 2 months before diagnosis (data not shown), namely in both instances they were not found to be significantly associated with the risk of RCC. Further adjustment for the potential confounders from Table II, including BMI, education, coffee consumption, smoking, alcohol and weight change during the last 2 months (Model 4) did not essentially alter the inverse association with adiponectin (OR = 0.77, p-value = 0.08); the associations of the other variables with RCC in this model were in agreement with those initially presented in Table II. More specifically, the risk for RCC is reduced as education level is increasing. The absence of alcohol consumption or the increased consumption was positive related with the RCC in relation to drinking 1 alcoholic drink per day. Recent weight loss was significantly related with RCC risk whereas coffee consumption, smoking and BMI were not found to be associated with RCC risk. When WHR was included along with adiponectin (Model 3), however, the association of adiponectin with RCC became statistically non-significant (OR = 0.84, p-value = 0.26) and was further attenuated when all studied potential confounding variables were mutually adjusted in Model 5 (OR = 0.89, p-value = 0.46). In Model 5, WHR index was associated with increased RCC risk and the impact of the other variables, that were also included in Model 4, remained practically the same, and whereas diabetes was not statistical significantly associated with RCC risk. Introducing adiponectin as a categorical variable in these models yielded similar results, whereas the exclusion of coffee consumption or the introduction of smoking as a categorical variable in 3 levels (never smoking, ex-smoker, current smoker) or in 4 levels based on the number of cigarettes (0, 1–19, 20–39, 40+) in Models 4 and 5 did not alter the initial results (data not shown). Concerning the possible effect of obesity, the results from the stratified analysis (restricted to obese patients and controls) were practically the same with those of the total study group. The OR for adiponectin in the univariate model (Model 1) was 0.65 (CI = 0.39–1.07) and for the corresponding Models 4 and 5 were 0.71 (CI = 0.38–1.35) and 0.66 (CI = 0.33–1.36), respectively. The introduction of WHR in both Models 4 and 5 decreased the effect of adiponectin in relation to RCC, indicating that altered adiponectin levels may mediate the effect of central or intra-abdominal obesity on RCC.
|Variable||Category or increment||ORs1 (95% CIs) p-value||ORs2 (95% CIs) p-value||ORs3 (95% CIs) p-value||ORs4 (95% CIs) p-value||ORs5 (95% CIs) p-value|
|Adiponectin||1 SD among controls||0.76 (0.57–1.00)||0.75 (0.56–0.99)||0.84 (0.62–1.13)||0.77 (0.57–1.03)||0.89 (0.64–1.22)|
|BMI||2 kg/m2||0.96 (0.78–1.17)||0.95 (0.76–1.19)||0.84 (0.66–1.07)|
|Waist/Hip ratio||5% more||1.89 (1.39–2.55)||2.09 (1.45–3.00)|
|Education||1 level more||0.71 (0.54–0.92)||0.72 (0.54–0.96)|
|Coffee consumption||yes vs. rare or not||2.26 (0.78–6.56)||2.37 (0.77–7.29)|
|Smoking||yes or ex-smoker vs. no||1.19 (0.64–2.22)||1.17 (0.60–2.31)|
|Alcohol||0 vs. 1–31 glasses/month||2.71 (1.30–5.65)||3.20 (1.46–7.01)|
|32+ vs. 1–31 glasses/month||2.26 (1.13–4.52)||2.52 (1.20–5.31)|
|Diabetes mellitus||no vs. yes||1.56 (0.78–3.12)|
|Weight change: last 2 months||1 kg more||0.86 (0.77–0.96)||0.88 (0.78–1.00)|
This case-control study shows, for the first time, that serum adiponectin levels are significantly and inversely associated with the risk of RCC. The association between adiponectin levels and RCC risk persists after adjustment for BMI, whereas further adjustment for other lifestyle characteristics, including educational level, smoking, coffee, alcohol consumption, diabetes mellitus status and recent weight change does not materially alter the observed association, suggesting little confounding effect from these factors. Conversely, the association between adiponectin levels and RCC risk becomes statistically non-significant after adjusting for indicators of central obesity including WHR. This somatometic index is significantly positively associated with RCC risk and negatively with adiponectin levels. Since the introduction of WHR in multivariate models attenuated the significance of the association of adiponectin with RCC risk, we propose that decreased circulating adiponectin levels may mediate the effect of central or intra-abdominal visceral obesity on RCC. Finally, the analysis of mean values of adiponectin levels and weight, BMI, WHR by tumor stage, grade and histological subtype did not reveal any statistically significant difference; however, a trend of decreasing serum adiponectin levels with increasing tumor grade was observed in our study. Further studies of a larger size should be conducted in the future to compare these variables between metastatic and non-metastatic RCC patients.
Overweight and obese subjects have a 1.5- to 3-fold higher risk for RCC when compared with subjects of normal weight.21 Most studies have reported a dose-response relationship of RCC with increasing weight or BMI, which is more pronounced among females.22 Adiposity is also related with hypertension and diabetes, disorders clearly associated with insulin resistance, which are also considered to be risk factors for RCC.23, 24, 25 The molecular pathways linking obesity, especially central obesity, with RCC remain to be fully elucidated, although several mechanisms have been proposed to date. Obesity leads to increased levels of several hormones including insulin as well as activation of the insulin-like growth factor 1 pathway, which may promote carcinogenesis.23 Obesity may also increase the risk for RCC by increasing the circulating levels of estrogens.23, 33 Importantly, visceral obesity is related to elevated insulin levels and insulin resistance, more closely than overall obesity.34
Adiponectin, a novel endogenous insulin sensitizer, is inversely associated with overall and visceral obesity, as well as insulin resistance.13, 34, 35, 36 Of note, central obesity expressed by WHR, is much more closely associated with decreased production of adiponectin as well as development of insulin resistance than BMI.37 More specifically, central fat distribution is an important, independent negative predictor of adiponectin and accumulating data suggests that adiponectin may represent an important link between central obesity and the development of insulin resistance or diabetes later in life.37, 38, 39 The negative association of adiponectin with HOMA-IR (Homeostatic Model Approach-Insulin Resistance) index, fasting insulin and glucose levels remains significant after adjusting for fat mass, but becomes non-significant after adjusting for markers of central adiposity such as WHR or Waist Circumference, suggesting that the relation between central obesity and insulin resistance may be mediated, at least in part, by adiponectin.37, 40
Adiponectin may also play an important role in several malignancies associated with obesity and insulin resistance, independently of other established risk factors for cancer.16, 17, 18, 19 Specifically, adiponectin levels are inversely associated with risk for endometrial,17, 18 breast,16 prostate,19 gastric cancer41 as well as leukemia.30 In addition to these prior case- control studies,16, 17, 18, 19 we have found, utilizing a prospective cohort study design that men in the highest quintile of adiponectin levels have an ∼60% reduced risk for colorectal cancer compared with those in the lowest quintile.31 Data from the present case-control study are consistent with those previous studies and extend their findings by demonstrating that low adiponectin levels are also associated with another obesity related malignancy, i.e. RCC.
There is accumulating experimental evidence supporting the protective role of adiponectin in the development of several types of cancer, as recently reviewed elsewhere.42 In brief, increasing levels of adiponectin improve insulin resistance and thus decrease circulating levels of insulin and insulin like growth factors in vivo.37, 43In vitro treatment of acute myelomonocytic leukemia cell lines with adiponectin suppresses the growth of myelomonocyte cells, induces apoptosis and down-regulates Bcl-2 gene expression.44 Adiponectin is also a direct inhibitor of angiogenesis through activation of caspases in endothelial cells in vivo and in vitro.45 Finally, adiponectin may act on tumor cells directly through activation of c-jun NH2-terminal kinase, a member of mammalian MAPK family involved in the regulation of cell proliferation and apoptosis,46, 47 and also through suppression of the transcription factor STAT3,48 which may also regulate cell proliferation, differentiation, survival and apoptosis.49
Our study has several strengths, but also limitations. The size of the study sample is fairly large given that RCC is a relatively rare type of cancer. The exclusion of cases with cancer of the renal pelvis, a histological type of kidney cancer distinct from RCC, prevents bias from potential inclusion of renal cancer cases with different underlying pathophysiology. We utilized state-of-the-art methodology and assays were performed blindly in 1 batch minimizing variability. Although random misclassification remains a possibility, this could have merely attenuated the statistical significance of our findings. To eliminate any selection bias the main confounding factors, such as the educational level and alternatively the profession (data not shown) have been controlled for in the analysis. The educational level of the control group was similar with that of the general population of the study area indicating that no major bias is likely to have been introduced in the selection of the controls. One may suggest that the results of this study may be biased due to the retrospective nature of case-control studies. Given however, the low incidence of RCC and the time needed to collect a sizeable group of eligible cases, we applied this study design, which is appropriate for rare diseases. Inherent with the study design issue is that case- control studies lack the time sequence criterion for causality; since prior prospective studies have also demonstrated an inverse association between adiponectin and colon cancer, the data on RCC need to be further confirmed by cohort studies, which can prove causality i.e. clinical prospective trials. Despite its limitations, this is the first study that investigated the association of adiponectin with RCC.
Our findings support the hypothesis that low adiponectin levels are significantly associated with RCC and may mediate the effect of central or visceral adiposity on the pathogenesis of RCC. Further prospective studies that incorporate the time criterion for causality, as well as studies exploring underlying mechanisms, should be undertaken to confirm and expand the findings of this study. Further studies for the elucidation of the molecular mechanisms underlying the effects of adiponectin on the pathogenesis of RCC are also warranted.
We thank Drs. T. Kakkavas, K. Livadas, S. Papadoukakis, A. Zarogiannos, I. Liakatas, G. Papadopoulos and A. Paschalis for their long-standing commitment to the study as well as S. Gialamas, T. Katsifotis and A. Pagonis, students of the Athens University Medical School for their valuable contribution to the study.
This work was supported by a pilot feasibility grant to C.S.M. from Harvard Medical School, Boston. C.S.M. is a Bessel Award recipient of the Humboldt Foundation.
- 26Epithelial tumors of the kidney. In: Belldegrun A, Ritchie A, Figlin R, Oliver RTD, Vaughan EDJr, eds. Renal and adrenal tumors. New York: Oxford University Press, 2003. 40–54., .
- 27AJCC cancer staging manual, 5th edn. Philadelphia: Lippincot-Raven, 1997. 231–2., , , , , , , .
- 32SAS Institute Inc. SAS/STAT User's Guide, Version 6, 4th edn., SAS Institute Inc, Cary, NC, 1989.
- 37Serum adiponectin levels are inversely associated with overall and central fat distribution but are not directly regulated by acute fasting of leptin administration in humans: cross-sectional and interventional studies. J Clin Endocrinol Metab 2003; 88: 4823–31., , , , , , .