Introduction
The incidence of hepatocellular carcinoma (HCC) has tripled in the United States over the last 30 years. This is thought to be at least partly due to metabolic syndrome, which has the highest population-attributable risk for HCC in the United States [1,2,3]. Patients with features of metabolic syndrome, including obesity, insulin resistance, and diabetes, also have worsened outcomes from several kinds of cancer, including HCC [4,5].
We and others have found that features of the metabolic syndrome may contribute to more aggressive tumor phenotypes, both in animal models and in patients. For instance, both diet and genetic obesity have been shown to promote HCC development in mouse models [6]. Prior work by our group also suggests an association between increased body mass index (BMI) and vascular invasion in tumors in patients with HCC [7,8].
There has been growing interest in the relationship between adipokines, such as adiponectin, and cancer. Adiponectin is a hormone that is usually inversely correlated with BMI and percent body fat. High levels typically predict lower incidences of several types of solid tumors [9,10] and better prognosis [11].
In this prospective study of patients with HCC, we examined serum adiponectin, leptin, HOMA-IR (homeostasis model assessment-estimated insulin resistance) levels, and clinicopathological features of HCC to evaluate preliminary relationships between adipokines and HCC outcome.
Expression levels of the two adiponectin receptor forms (AR1 and AR2) have also been associated with both physiological and pathological states including obesity and insulin resistance [12]. Reduced AR expression has been implicated in the pathogenesis of nonalcoholic steatohepatitis (NASH) and the development of cirrhosis [13,14]; thus, we also examined a subset of subjects for AR expression in tumor and surrounding tissue.
Methods
Study Population
From 2008 to 2012, 140 adult patients, 18 years and older, with newly diagnosed or recurrent HCC at the Columbia University Medical Center were recruited for this prospective study and followed until December 2012. Subjects had histologically proven HCC or met the American Association for the Study of Liver Diseases criteria for HCC diagnosis [15]. We excluded patients who had any previous malignancy within the last 5 years, those who had undergone liver transplantation, and those who had received any systemic cancer therapy. Those with Child-Pugh (CP) class C disease and/or clinically significant ascites were also excluded. The protocol was approved by the Columbia University Institutional Review Board, and all subjects signed a written informed consent.
Procedures
Subjects completed an epidemiologic questionnaire and underwent a physical exam including weight, height, waist and hip measurements. Fasting morning blood samples were collected at the time of enrollment. Laboratory analyses included a complete blood count, basic metabolic/hepatic panels, hemoglobin A1c, lipid panel, hepatitis screen, α-fetoprotein (AFP), ferritin, transferrin, ceruloplasmin, α1-antitrypsin phenotype, antinuclear antibody, anti-smooth-muscle antibody, and antimitochondrial antibody to define the underlying etiology of HCC.
An additional 30-ml fasting blood sample was collected at the same blood draw for adipokine evaluation. Whole blood, plasma, serum, and buffy coat were divided into aliquots and frozen at -70°C, then run in batches at the Irving Institute's Biomarker Core at Columbia University. Specimens were de-identified with a unique study number and barcode, and results were forwarded to the study coordinator for entry into a secure database. Laboratory personnel were blinded to study hypotheses and outcomes.
Measures
Our main exposures of interest were serum levels of adipokines. We used a radioimmunoassay to measure baseline levels of leptin (RIA Linko, St. Charles, Mo., USA; quantification limit 0.5 ng/ml, interassay precision 4.6%, intra-assay precision 5%), adiponectin (RIA Millipore, Billerica, Mass., USA; quantification limit 1 ng/ml, interassay precision 6.9%, intra-assay precision 6.2%), and insulin (RIA Siemens, Deerfield, Ill., USA; quantification limit 2 μIU/ml, interassay precision 5.3%, intra-assay precision 4.3%). HOMA-IR was used as a measure of insulin resistance and was calculated as fasting insulin (mU/l) × plasma glucose (mmol/l)/22.5 [16]. All samples were run in duplicate, and the median value was used. Our primary outcome of interest was time to death from diagnosis. Mortality data were obtained from medical records and confirmed with the National Death Index.
The following covariates were used in the analysis: age (≥ or < median), gender, race/ethnicity, etiology of liver disease (hepatitis B virus, hepatitis C virus, alcohol, diabetes mellitus), BMI, CP class, metastasis (yes/no), Barcelona Clinic Liver Cancer (BCLC) stage, Milan criteria (within/outside), and serum AFP levels (≥ or < median). Those who self-reported an alcohol use of more than two standard drinks per day (28 oz) consistently for over a year at any point in their lives were classified as having alcohol-related disease. Diabetes was defined as having a fasting glucose ≥126 mg/dl or being on drug therapy for diabetes. NASH was not formally assessed because of difficulties assessing it pathologically in the setting of cirrhosis, and because not all subjects had an available pathology for review. Since the waist to hip ratio has often been shown to better estimate mortality risks due to general and abdominal obesity, we conducted separate analyses assessing the waist to hip ratio (≥ or < median) [17]. Variations in circulating levels of serum adipokines have been shown to play a role in liver fibrogenic processes [18,19]; therefore, in an exploratory analysis in a subset of patients (54%) for whom pathology data were available, we also assessed the association between the Batts-Ludwig stage of liver fibrosis (I-IV) and serum adipokine levels [20].
Immunohistochemistry
Tissue sections from 17 patients were evaluated for AR1 and AR2 expression by immunohistochemistry. Four groups were selected randomly from cases with available formalin-fixed paraffin-embedded tissue based on low and high levels of serum adiponectin (≥ or < median) and the degree of fibrosis assessed by the Batts-Ludwig stage. After antigen retrieval in ULTRA Cell Conditioning Solution (Ventana Medical Systems, Tucson, Ariz., USA), slides from all cases were stained on an autostainer (Ventana Benchmark Ultra; Ventana Medical Systems) for AR1 (dilution 1:100, rabbit monoclonal antibody; Abcam, Cambridge, Mass., USA) and AR2 (374-386, dilution 1:100, rabbit polyclonal antibody; Phoenix Pharmaceuticals, Burlingame, Calif., USA). Slides were examined by a liver pathologist blinded to clinical and laboratory values, who scored the expression levels of AR1 and AR2 in both tumor and surrounding tissue as either positive or negative based on the presence or absence of staining.
Statistical Analysis
Descriptive statistics including mean levels and ranges were used to summarize baseline adipokine levels. Associations between serum adiponectin, leptin, HOMA-IR (all as categorical variables using median values as cutoffs) and covariates including etiology and clinicopathological variables (all as categorical variables) were analyzed using χ2 tests.
Our preliminary data showed a 3-fold increased risk of death for those with higher than median adiponectin levels (>12,000 ng/ml) in univariate and multivariate models (adjusted for stage, AFP, age, and CP class). Our pilot data suggested a median survival of 290 days for the high adiponectin group, and we conservatively assumed a median survival of 600 days for the lower adiponectin group. We would then have had 83% power to identify a survival difference of about 300 days at p = 0.05, with 40 patients in each group, using Kaplan-Meier analyses. With a sample size of 140 patients, we planned to adjust for additional covariates, including BMI, etiology of liver disease, age, race/ethnicity, stage, and tumor features.
Kaplan-Meier plots and Cox proportional hazards models were used to determine the association between variables of interest and time to death. We defined cutoffs for the low and high levels of our main exposure variables (adiponectin, leptin, and HOMA-IR) using quartiles of baseline serum levels. Variables that were clinically relevant and showed statistical significance in univariable analysis were included in the multivariable Cox proportional hazards models. Subjects contributed person-time for survival analyses from the date of enrollment to the date of death/date of censoring. Subjects were censored if they were lost to follow-up or at the end of the study period (December 31, 2012). There were no significant differences in baseline adipokine levels for the 18 patients that were lost to follow-up. Further, excluding these participants did not lead to any significant change in the hazard ratios (HRs) in unadjusted and multivariable adjusted Cox proportional hazards models.
In additional analyses, we examined the relationship between insulin resistance (using HOMA-IR) and adipokine level (as continuous variables) using Pearson correlation coefficients. Since the ratio of adiponectin to leptin has been used as an index to evaluate insulin resistance [21], we assessed the adiponectin-leptin ratio as an independent predictor of survival in patients with HCC. Finally, to further evaluate the predictive value of these markers, we tested for interaction and calculated separate HRs stratified by baseline subject and tumor characteristics using medians as cutoffs to define the high and low levels of adiponectin, leptin, and HOMA-IR. All tests of statistical significance were two-sided, and analyses were done using SAS, version 9.3 (SAS Institute Inc., Cary, N.C., USA).
Results
The median age of the study participants was 62 years. 79% were men, 49% were non-Hispanic White, 59% had hepatitis C, and 36% were diabetic. 64% were CP class A and 44% were within the Milan criteria at enrollment. Mean (median, and range) baseline serum levels were 14.1 ng/ml (7.8, 1.6-89.3) for leptin, 16,540.7 ng/ml (13,050, 850.0-82,400.0) for adiponectin and 5.5 (3.2, 0.35-67.60) for HOMA-IR, respectively. The median follow-up was 8 months, and the median survival time was 18 months. Additional data related to survival and adiponectin levels are shown in the online supplementary table 1 (for all online suppl. material, see www.karger.com/doi/10.1159/000367971). During the follow-up period, 76% of patients received locoregional therapy, 19% underwent resection, 15% received sorafenib, and 20% underwent a liver transplantation.
Subject and tumor characteristics stratified by low (<median) and high (≥median) levels of serum leptin, adiponectin, and HOMA-IR are summarized in table 1. High adiponectin levels (≥13,050 ng/ml) were associated with female gender (p = 0.003), being hepatitis C positive (p = 0.002), and being CP class B (p < 0.001). High leptin (p = 0.003) and HOMA-IR (p = 0.03) levels were significantly associated with a higher degree of liver fibrosis. Interestingly, the adiponectin level was not significantly associated with liver fibrosis (p = 0.29).
Table 1
Subject and tumor characteristics according to low (<median) and high (≥median) levels of serum leptin, adiponectin, and HOMA-IR
In univariable analysis, a serum adiponectin level greater than the median was significantly associated with survival [HR 1.83; 95% confidence interval (CI) 1.05-3.19; p = 0.03], whereas leptin (HR 0.62; 95% CI 0.35-1.08; p = 0.09) and HOMA-IR (HR 0.77; 95% CI 0.44-1.34; p = 0.36) were not (table 2; fig. 1). Adiponectin remained a significant predictor of time to death (HR 1.90; 95% CI 1.05-3.45; p = 0.03) in a multivariable adjusted model that included age, alcohol history, CP class, Milan criteria, and serum AFP level (table 3). However, the HR for adiponectin was attenuated when we replaced the Milan criteria with BCLC stage (HR 1.34; 95% CI 0.73-2.45; p = 0.35) or included surgical treatment (HR 1.71; 95% CI 0.93-3.17; p = 0.09) in the above model. Surprisingly, a history of alcohol intake was a significant predictor of better overall survival in both the unadjusted and multivariable-adjusted model. This may be because a history of alcohol intake was associated with better prognostic variables like lower stage at baseline.
In additional analyses, the relationship between insulin resistance (using HOMA-IR) and the adipokine level (as continuous variables) was examined. A significant positive correlation was found between leptin levels and HOMA-IR (Pearson correlation coefficient r = 0.49; p < 0.001), while correlations between HOMA-IR and adiponectin (r = -0.11; p = 0.19) were not significant. Every standard deviation increase in the adiponectin-leptin ratio was associated with a HR of 1.38 (p = 0.002). This association remained significant (HR 1.38; p = 0.01) after further adjusting for other predictors including age, alcohol history, CP class, BCLC stage, and serum AFP levels.
Expression levels of AR1 or AR2 were assessed for 17 tissue sections from subjects containing both tumor and non-neoplastic liver parenchyma (fig. 2). Tumors of patients with higher adiponectin levels were more likely to stain positive for both AR1 (p = 0.08) and AR2 (p = 0.15). Staining was predominantly cytoplasmic. AR1 and AR2 expression did not significantly correlate with the degree of fibrosis or other clinical or pathological covariates. In general, no significant staining heterogeneity was noted in individual samples. There were also no significant differences in the staining pattern or intensity of the background cirrhotic liver for AR1 or AR2.
Fig. 2
AR1 and AR2 expression in HCC by immunohistochemistry. a, c Cases with negative AR1 and AR2 reactivity. b, d In comparison, strong staining for AR1 and AR2, respectively.
Discussion
In this observational study of 140 HCC patients, we found that the serum adiponectin level was independently associated with worsened overall survival even after adjusting for important clinical covariates. In a subset of patients with available pathology, AR staining in tumors trended with peripheral adiponectin levels, although these results were not statistically significant, likely due to small sample sizes.
Adiponectin is a protein derived from adipocytes which regulates fat and glucose metabolism [22]. Levels are usually lower in those who are obese and diabetic and are typically increased in those with hepatic fibrosis [23]. In most studies, adiponectin levels predict a better prognosis in cancer patients [11,24,25,26,27,28,29]. In contrast, some studies have reported high adiponectin levels to be associated with poor prognosis or more aggressive pathology. For instance, in gastric cancer patients, increased adiponectin levels were associated with higher tumor grade (p = 0.02) and more poorly differentiated tumors [25]. Similarly, in prostate cancer, serum adiponectin levels were found to be higher in locally advanced relative to organ-confined disease [30].
In HCC, relationships between adiponectin levels and clinical features of disease are complex. Cell culture and animal models suggest that adiponectin inhibits leptin-induced proliferation of HCC via a blockade of downstream pathways including STAT-3, AKT, and M-TOR [31]. Adiponectin also leads to suppression of liver tumor growth and metastasis in mice by inhibiting angiogenesis [32]. Similarly, in a study of human HCC, low adiponectin levels were associated with worsened histological grade of HCC [33].
However, in other human studies, adiponectin may reflect an increased risk of HCC [34,35,36]. Elevated adiponectin expression may also correlate with worsened outcome, as noted by a report from Taiwan which suggested that high cytoplasmic staining of HCC tumors with adiponectin was associated with recurrence and worsened survival, possibly via upregulation of AKT. This finding persisted after controlling for BMI and fasting blood sugar. However, this study did not assess peripheral levels of adiponectin [37].
The median values seen for adipokines in our study seemed relatively comparable to other populations in the literature, including those with chronic liver disease and cases with pancreatic cancer compared with controls, although our ranges did seem a bit broader [9,38]. We speculate that this may be due to the significant heterogeneity of our patient population, particularly with respect to tumor stage, nutritional status, and degree of underlying liver dysfunction.
Possible mechanisms that could explain our results include a possible angiogenic role and/or an antiapoptotic effect of adiponectin, particularly in the setting of high levels of glucose [39,40,41]. In a study of 609 patients with type II diabetes, higher adiponectin levels were also associated with worsened overall survival. This and other studies showed a high correlation between total and high-molecular-weight adiponectin, so we chose to present total adiponectin [42].
Another possible explanation for our results is that adiponectin levels may be increased in cirrhosis, and with increasing stages of fibrosis [43,44]. Cirrhosis is thought to lead to a state of adiponectin resistance, with downregulation of ARs in liver tissue and decreased clearance of adiponectin [45,46,47,48]. However, we did not find a statistically significant association between liver fibrosis and peripheral adiponectin levels. Although higher adiponectin levels were seen in those with CP class B disease, adjusting for CP class did not significantly affect the association of adiponectin with overall survival.
Adiponectin and AR expression have been shown to be reduced in patients with NASH [13]. Obesity can lead to decreased adiponectin levels as well as to downregulation of ARs leading to insulin resistance [49,50]. In contrast, cachexia has been associated with increases in adiponectin, even when controlling for BMI [51]. In the elderly, higher adiponectin levels also correlate with mortality, are associated with frailty, and can be modified by exercise [52,53]. High adiponectin levels have been shown to be associated with poor outcomes in patients in the intensive care unit, independent of BMI or markers of inflammation [54]. These results highlight the multiple competing effects regulating adiponectin and AR levels in these patients. Interestingly, in our study, subjects self-reported weights which were generally lower several years prior to diagnosis (data not shown), making cachexia related to illness less likely in our population, since we excluded those with clinically significant ascites.
Our study has several strengths, including its prospective nature and the use of REMARK guidelines as its basis [55]. Limitations include a relatively small sample size, assessment of adiponectin levels at only one time point, and a variety of treatments given before and after enrollment. Although this heterogeneity may be considered a weakness, it also suggests the potential broad applicability of adiponectin as a prognostic biomarker.
Based on our results, we believe that adiponectin deserves further validation as a simple and practical biomarker to assess outcomes in patients with HCC, particularly in those with features of metabolic syndrome. There are multiple variables which could have opposing effects on adiponectin levels, including metabolic syndrome, cirrhosis, cachexia, and frailty; further examination is needed to sort these out in individual patients and to clarify the mechanisms for these associations. Modulation of adiponectin pathways in those with elevated levels of adiponectin as a potential mechanism to alter the disease course (potentially via AKT pathway modulation) also deserves further study.
Acknowledgement
This study was supported by NIH K23 CA149084-01A1 and the Steven J. Levinson Medical Research Foundation (to A.B.S.)
Disclosure Statement
None of the authors has a financial conflict of interest relevant to the manuscript.
Get Permission
PubMed ID
