Soluble receptor for advanced glycation end products and risk of liver cancer


  • Kristin A. Moy,

    Corresponding author
    1. Nutritional Epidemiology Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute, National Institutes of Health, Bethesda, MD
    • Nutritional Epidemiology Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute, 6120 Executive Boulevard, Suite 320, Rockville, MD 20852, Fax: (301) 496-6829
    Search for more papers by this author
  • Li Jiao,

    1. Section of Gastroenterology and Hepatology, Department of Medicine, Baylor College of Medicine, Houston, TX
    Search for more papers by this author
  • Neal D. Freedman,

    1. Nutritional Epidemiology Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute, National Institutes of Health, Bethesda, MD
    Search for more papers by this author
  • Stephanie J. Weinstein,

    1. Nutritional Epidemiology Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute, National Institutes of Health, Bethesda, MD
    Search for more papers by this author
  • Rashmi Sinha,

    1. Nutritional Epidemiology Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute, National Institutes of Health, Bethesda, MD
    Search for more papers by this author
  • Jarmo Virtamo,

    1. Department of Chronic Disease Prevention, National Institute for Health and Welfare, Helsinki, Finland
    Search for more papers by this author
  • Demetrius Albanes,

    1. Nutritional Epidemiology Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute, National Institutes of Health, Bethesda, MD
    Search for more papers by this author
  • Rachael Z. Stolzenberg-Solomon

    1. Nutritional Epidemiology Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute, National Institutes of Health, Bethesda, MD
    Search for more papers by this author

  • Potential conflict of interest: Nothing to report.

  • Supported by the Intramural Research Program of the National Institutes of Health, Division of Cancer Epidemiology and Genetics, National Cancer Institute, National Institutes of Health, Department of Health and Human Services, Bethesda, MD.


Binding of advanced glycation end products (AGEs) to their receptor (RAGE) increases oxidative stress and inflammation and may be involved in liver injury and subsequent carcinogenesis. Soluble RAGE (sRAGE) may neutralize the effects mediated by the AGE/RAGE complex. Epidemiologic studies examining sRAGE or AGEs in association with liver cancer are lacking. We examined the associations between prediagnostic serum concentrations of sRAGE or Nϵ-(carboxymethyl)-lysine (CML)-AGE and hepatocellular carcinoma in a case-cohort study within a cohort of 29,133 Finnish male smokers who completed questionnaires and provided a fasting blood sample between 1985 and 1988. During follow-up beginning 5 years after enrollment through April 2006, 145 liver cancers occurred. Serum concentrations of sRAGE, CML-AGE, glucose, and insulin were measured in case subjects and 485 randomly sampled cohort participants. Chronic hepatitis B virus (HBV) and hepatitis C virus (HCV) were available in most cases and in a subset of the study population. Weighted Cox proportional hazards regression was used to calculate relative risks (RR) and 95% confidence intervals (CI) adjusted for age, years of smoking, and body mass index. sRAGE and CML-AGE concentrations were inversely associated with liver cancer. Further adjustment for glucose and insulin or exclusion of case subjects with chronic HBV or HCV did not change the associations. Conclusion: Our results support the hypothesis that sRAGE is inversely associated with liver cancer. The findings need confirmation, particularly in populations that include women and nonsmokers. (HEPATOLOGY 2013 )

Worldwide, primary liver cancer is the sixth most commonly occurring cancer and the third most common cause of cancer-related deaths.1, 2 Established risk factors for hepatocellular carcinoma or HCC, the most common type, include aflatoxin B exposure, chronic infection with hepatitis B virus (HBV) or hepatitis C virus (HCV), excessive alcohol consumption, and obesity and diabetes, which increase the risk of nonalcoholic steatohepatitis (NASH).3, 4 These risk factors are typified by their ability to cause chronic inflammation in the liver, which is associated with subsequent carcinogenesis.5 Advanced glycation end products (AGE) and their receptor (RAGE) are implicated in both inflammation and cancer (reviewed by Riehl et al.,6 Singh et al.,7 and Sparvero et al.8). However, the potential role of the AGE-RAGE axis in the development of HCC is unknown.

AGEs are heterogeneous irreversible adducts formed by the nonenzymatic glycation of proteins, lipids, and nucleic acids.7 The two major sources of AGEs are endogenous AGEs that form during normal metabolism and exogenous AGEs derived from tobacco smoke or food.7, 9 Dietary AGEs are formed when food is processed at high temperatures using methods such as deep frying, broiling, and grilling.10 AGEs accumulate in tissues, and the rate of accumulation increases with aging and under hyperglycemic conditions.7 Of the approximately 20 different AGEs identified, Nϵ-(carboxymethyl)-lysine (CML)-AGE is the best characterized.11, 12

AGEs up-regulate inflammation by binding their full-length membrane-bound receptor RAGE.7, 13 RAGE is a multiligand receptor that belongs to the immunoglobulin superfamily. Binding ligand triggers the activation of cell signaling pathways such as p38 and p44/42 mitogen-activated protein kinase, as well as nuclear factor kappa B, generating reactive oxygen species and the production of proinflammatory cytokines (reviewed by Riehl et al.,6 and Bierhaus and Nawroth14). Because the liver is important for the clearance and catabolism of circulating AGEs (e.g., removing >90% of intravenously injected AGEs via endocytosis has been shown in rats13, 15), the AGE-RAGE axis may be particularly important for liver carcinogenesis and chronic liver diseases, including NASH and liver cirrhosis.16-18

In addition to the full-length receptor for AGEs, RAGE has truncated soluble isoforms (sRAGE) containing only the RAGE extracellular domain,13, 15 including a splice variant of the full-length receptor, endogenous secretory RAGE, and an isoform formed by proteolytic cleavage.19, 20 Both forms can be detected in human serum, are capable of binding ligands, and are thought to bind free AGEs, exerting a cytoprotective effect by preventing ligands from binding to cell surface RAGE.8, 15 In vitro and experimental studies suggest a protective role of sRAGE in hepatocellular injury.21-27

Although several hospital-based studies have found sRAGE levels to be lower in lung, breast, and pancreatic cancer case subjects compared with healthy controls,28-30 only three prospective epidemiologic studies have examined the potential association of AGEs with cancer, finding no significant associations between CML-AGE and pancreatic31, 32 or colorectal cancers.33 In the current study, we examined the associations of serum levels of sRAGE and CML-AGE with liver cancer risk in the Alpha-Tocopherol, Beta-Carotene Cancer Prevention (ATBC) Study.34 We hypothesized that lower levels of sRAGE or higher levels of CML-AGE are associated with increased risk.


AGE, advanced glycation end product; anti-HBc, antibody to hepatitis B core antigen; anti-HBV, antibody to hepatitis B virus; ATBC, Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study; BMI, body mass index; CI, confidence interval; CML, Nϵ-(carboxymethyl)-lysine; HBsAg, hepatitis B surface antigen; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; RAGE, receptor for advanced glycation end product; RR, relative risks; sRAGE, soluble receptor for advanced glycation end product; NASH, nonalcoholic steatohepatitis.

Subjects and Methods

Study Population.

The design of the ATBC Study has been described in detail elsewhere.34 Briefly, the ATBC Study was a randomized, double-blind, placebo-controlled, primary prevention trial conducted in southwest Finland to determine the effects of supplementation with α-tochopherol and β-carotene on cancer incidence among male smokers. Between 1985 and 1988, a total of 29,133 Caucasian men aged 50 to 69 years who smoked at least five cigarettes per day were randomized to receive an active supplement or a placebo. Potential participants were excluded from the trial if they reported malignancy other than nonmelanoma skin cancer or carcinoma in situ, severe angina on exertion, chronic renal insufficiency, liver cirrhosis, chronic alcoholism, receiving anticoagulant therapy, other medical problems that might limit participation for 6 years, or current use of supplements containing vitamin E, vitamin A, or β-carotene in excess of predefined doses.34 The trial ended in April 1993, but participants continued to be followed for health outcomes through national registries. The ATBC Study was approved by the institutional review boards of both the US National Cancer Institute and the National Public Health Institute of Finland (now the National Institute for Health and Welfare, Helsinki, Finland). All participants provided written informed consent before randomization.

At enrollment, participants completed a self-administered questionnaire that assessed demographics, including medical, smoking, and occupational histories. Dietary intake during the previous year was also assessed using a validated self-administered dietary history questionnaire. Height and weight were measured by trained nurses. Body mass index (BMI) was calculated as weight in kilograms per height in meters squared.34 Study participants also provided venous blood samples after an overnight fast at baseline, and serum was aliquoted and stored at −70°C.

Cases and Subcohort.

The present study is a case-cohort study within the parent study. To reduce to the potential effect of subclinical disease on serum levels of CML-AGE or sRAGE in samples collected at baseline prior to randomization, subjects in this study were drawn from cohort members who were alive and cancer-free as of the sixth year of follow-up (N = 24,708).34 Thus, follow-up began 5 years after blood collection at randomization and ended at date of liver cancer diagnosis, death, or on April 30, 2006, whichever occurred first. A total of 146 incident liver cancer cases (defined based on the International Classification of Diseases 9; codes 155.0, 155.1, and 155.2) were identified from the Finnish Cancer Registry.35 From the remaining eligible cohort members, 500 subjects were randomly selected as the reference group. After excluding 1 case and 15 subcohort participants with missing data on one or more of the serological biomarkers, the present analysis included 145 liver cancer cases and 485 subcohort participants.

Laboratory Methods.

Serum sRAGE and CML-AGE were measured in duplicate by Microcoat Biotechnologie Company using the human sRAGE Quantikine ELISA kit (R&D Systems, Inc.) and the AGE-CML-ELISA kit (Microcoat Biotechnologie Company), respectively. The AGE-CML-ELISA kit uses a CML-specific monocolonal antibody (mouse monocolonal 4G9; Alteon Inc).36 The sRAGE Quantikine ELISA kit detects a heterogeneous group of total sRAGE proteins, including cleavage forms of membrane-bound full-length RAGE,19 endogenous secretory RAGE,20 and other splice variant forms of RAGE. Case and subcohort samples were randomly ordered in each batch along with 10% blinded quality control samples from a single pooled serum sample. The intrabatch coefficients of variation for sRAGE and CML-AGE were 3% and 7%, respectively, and the corresponding interbatch coefficients of variation were 6% and 14%. Serum concentrations of glucose and insulin were measured previously for 51 cases and 406 subcohort participants.37 In the present study, serum glucose and insulin were determined on an additional 95 cases that occurred after 2001 and on an additional 94 subcohort participants using the same method in the same laboratory as the earlier study.

Statistical Analysis.

The Wilcoxon rank sum test for continuous variables and the χ2 test for categorical variables were used to compare the distributions of selected demographics, including dietary and other characteristics between the cases and the subcohort. Dietary variables were adjusted for total energy intake using the residuals method.38 Selected demographic and dietary variables were examined across tertiles of serum CML-AGE and sRAGE using the analysis of covariance method.

Standard statistical methods for case-cohort studies were used in the present study. Weighted Cox proportional hazards regression was used to calculate relative risks (RR) and their corresponding 95% confidence intervals (CIs) and P values.39 Study subjects were grouped by tertiles of sRAGE and CML-AGE defined according to the distribution in the subcohort. Continuous RRs were calculated to the average size of the central tertile to represent the change in risk of liver cancer per tertile, and P values for linear trends across tertiles were calculated using a score variable based on the median value of each tertile. Follow-up time was used as the underlying time metric. We used forward and backward stepwise regression to evaluate potential confounding. Factors evaluated as potential confounders of the sRAGE or CML-AGE and liver cancer associations include the variables listed in Table 1. A variable was defined as a confounding factor if it was significantly associated with liver cancer risk and sRAGE or CML-AGE, and inclusion of the variable in the model changed the risk estimate for sRAGE or CML-AGE by more than 10%. Although none of the variables (Table 1) met the latter criterion, final multivariate models included age at randomization, years of smoking, and BMI. Further adjustment for number of cigarettes per day, trial treatment arm, or other variables associated with sRAGE or CML-AGE (Table 2) did not change risk estimates more than 10% and were not included in the final models. To assess whether the associations between sRAGE or CML-AGE and liver cancer were independent of other serum analytes, insulin and glucose concentration were added into the models individually, but because their inclusion did not change the risk estimates more than 10%, they were not retained in the final models.

Table 1. Baseline Characteristics of Liver Cancer Case Subjects and Subcohort Participants, ATBC Study (1985-2006)
CharacteristicCase Subjects (n = 145)Subcohort Participants (n = 485)P*
  • Abbreviations: AT, α-tocopherol; BC, β-carotene; IQR, interquartile range.

  • *

    Two-sided P values are based on Wilcoxon rank sum tests for continuous variables and χ2 tests for categorical variables.

  • Food and nutrient variables were adjusted for total energy intake. Information was available for 133 liver cancer cases and 465 subcohort participants.

Age, years, mean (SD)58.1 (4.8)56.4 (5.0)0.0001
BMI, kg/m2, mean (SD)27.6 (4.7)26.7 (4.0)0.03
Years of smoking, mean (SD)37.2 (8.3)35.2 (8.2)0.0002
No. of cigarettes per day, mean (SD)21.5 (9.3)20.8 (8.5)0.50
Clinical diabetes, n (%)   
 Glucose <126 mg/dL125 (86.2)461 (95.0)0.0002
 Glucose ≥126 mg/dL20 (13.8)24 (5.0) 
Dietary or nutrient intake/day, mean (SD)   
 Total meat, g197 (73.3)194 (68.8)0.65
  Red meat73.1 (30.0)68.2 (28.2)0.06
  Processed meat69.4 (46.4)74.3 (52.4)0.59
 Alcohol, g27.7 (33.1)18.9 (21.7)0.01
 Coffee, g502 (337)595 (354)0.002
 Total fat, g119 (22.0)122 (18.0)0.28
  Saturated fat, g48.9 (15.0)52.8 (13.9)0.01
 Protein, g92.3 (13.2)94.4 (12.7)0.14
 Carbohydrates, g257 (38.5)264 (39.1)0.03
 Glucose, g8.69 (4.37)9.30 (5.76)0.69
 Sucrose, g48.9 (26.1)51.5 (24.1)0.15
 Iron, mg17.6 (3.87)17.9 (3.43)0.29
 Mean (SD)561 (262)615 (307)0.09
 Median (IQR)520 (377-706)571 (413-738) 
 Mean (SD)484 (131)565 (168)<0.0001
 Median (IQR)480 (404-680)561 (471-668) 
Treatment arm, n (%)   
 Placebo35 (24.1)110 (22.7)0.48
 BC only40 (27.6)116 (23.9) 
 AT only33 (22.8)131 (27.0) 
 AT + BC37 (25.5)128 (26.4) 
Table 2. Selected Age-Adjusted Baseline Characteristics by Tertiles of sRAGE and CML-AGE in Subcohort Participants, ATBC Study (1985-2006)
CharacteristicTertile 1Tertile 2Tertile 3P trend
  • Data are expressed as means.

  • *


  • †, ‡

    Analysis of covariance, adjusted for age at randomization.

  • Dietary variables were adjusted for total energy intake (n = 465).

 Age at randomization*56.556.256.50.97
 BMI, kg/m227.126.726.40.12
 Years of smoking35.235.634.70.57
 No. of cigarettes/day21.520.820.00.09
 Total meat, g/day,196.7193.9190.50.42
  Red meat, g/day,67.768.068.90.71
  Processed meat, g/day77.475.170.10.22
 Alcohol, g/day19.920.116.60.30
 Coffee, g/day,590.0567.3629.80.33
 Total fat, g/day,123.4122.3120.40.15
  Saturated fat, g/day,53.152.552.80.86
 Carbohydrates, g/day,258.7262.1271.00.006
 Glucose intake, g/day,
 Sucrose intake, g/day,48.451.155.30.01
 Iron, mg/day,
 Serum insulin, μU/mL,
 Serum glucose, mg/dL,102.797.598.00.10
 Age at randomization*56.855.856.60.74
 BMI, kg/m227.326.926.00.004
 Years of smoking35.335.534.80.50
 No. of cigarettes/day21.120.410.70.70
 Total meat, g/day,193.5189.8197.90.57
  Red meat, g/day,69.666.668.30.70
  Processed meat, g/day,75.673.174.20.82
 Alcohol, g/day,18.518.719.50.67
 Coffee, g/day,566.7636.3582.70.69
 Total fat, g/day,126.1122.0118.0<0.0001
  Saturated fat, g/day,55.553.449.50.0001
 Carbohydrates, g/day,255.2263.2273.1<0.0001
 Glucose intake, g/day,8.38.611.0<0.0001
 Sucrose intake, g/day,48.751.754.20.04
 Iron, mg/day,17.618.018.40.04
 Serum insulin, μU/mL,
 Serum glucose, mg/dL,99.697.3101.40.50

Effect modification of sRAGE and CML-AGE by BMI (continuous and <25 kg/m2 versus ≥25 kg/m2), number of years of smoking (continuous), number of cigarettes per day (continuous), serum glucose (continuous), clinical diabetes (yes or no), energy-adjusted red meat intake (continuous), coffee intake (continuous), and trial intervention (placebo, β-carotene only, α-tocopherol only, β-carotene plus α-tocopherol) was tested by including cross-product terms in multivariate models and testing for statistical significance using the Wald test. Additionally, the joint effects of sRAGE and CML-AGE (both dichotomous) on the risk of liver cancer were examined. To examine potential differential effects of sRAGE and CML-AGE at different stages of liver carcinogenesis, we examined the associations between sRAGE or CML-AGE and liver cancer stratified by follow-up time (subjects with 5-9 years of follow-up and subjects with ≥10 years of follow-up).

Exposure to HBV and HCV infection serologic status was available for 135 cases and 26 subcohort subjects in the present study. Sensitivity analyses excluding participants who were positive for antibody to hepatitis B core antigen (anti-HBc), hepatitis B surface antigen (HBsAg), or antibody to HCV (anti-HCV) were conducted. Statistical analysis was performed using SAS software version 9.1 (SAS Institute, Cary, NC) and SUDAAN software (RTI, Research Triangle Park, NC). All statistical tests were two-sided, and P < 0.05 was considered statistically significant.


In the present study, the average time (± SD) between serum collection and liver cancer diagnosis was 12.2 (± 4.2) years, ranging from 5 to 21 years. The mean age (± SD) at liver cancer diagnosis was 70.3 (± 5.4) years.

Compared with subcohort participants, liver cancer case subjects were slightly older at time of randomization, had a higher BMI, and reported smoking more cigarettes per day for a longer period (Table 1). Case subjects were also more likely to report having been diagnosed with diabetes and consuming more red meat and alcohol. Case subjects also consumed less coffee, saturated fat, and carbohydrates than their subcohort counterparts. Case subjects had significantly lower serum levels of CML-AGE and borderline significantly lower serum levels of sRAGE compared with the subcohort population (Table 1).

Table 2 shows the age-adjusted means of selected baseline characteristics associated with liver cancer or AGEs of the subcohort participants according to tertiles of serum sRAGE and CML-AGE. There was higher carbohydrate and sucrose intake across tertiles of sRAGE (P ≤ 0.01). With increasing tertiles of CML-AGE, mean BMI and total and saturated fat intake were significantly lower (P ≤ 0.004), whereas intake of carbohydrate, glucose, sucrose, and iron all were significantly higher (P ≤ 0.04).

Serum sRAGE concentrations were associated with a statistically significant reduction in risk of liver cancer when examined as a continuous variable (RR, 0.86; 95% CI, 0.75-0.99; P = 0.02) and nonsignificant reduced risk comparing the highest and lowest tertile (RR, 0.77; 95% CI, 0.48-1.24; P for trend = 0.28) (Table 3). Higher CML-AGE concentrations were significantly associated with reduced risk of liver cancer when examined as a continuous variable standardized to the central tertile (RR, 0.74; 95% CI, 0.65-0.84; P < 0.0001) and in a dose-dependent manner when examined categorically (T3 versus T1, RR, 0.19; 95% CI, 0.10-0.35; P for trend < 0.0001). Further adjustment for serum glucose and insulin or trial intervention did not materially change these associations (data not shown).

Table 3. Serum CML-AGE and sRAGE in Relation to Risk of Liver Cancer, ATBC Study (1985-2006)
  • Abbreviation: HR, hazard ratio.

  • *

    Trend test based on median values of each tertile.

  • HRs for continuous sRAGE and CML-AGE were standardized to the average size of the central tertile (sRAGE, 205.5 pg/mL; CML-AGE, 111.4 ng/mL).

  • Adjusted for age.

  • §

    Adjusted for age, BMI, and years of smoking.

 Range, pg/mL<465.4465.4 to <670.9≥670.9   
 Case/subcohort, n54/16247/16244/161   
 Model 1, HR (95% CI)1.000.89 (0.55-1.43)0.74 (0.46-1.18)0.210.84 (0.73-0.97)0.02
 Model 2, HR (95% CI)§1.000.91 (0.56-1.47)0.77 (0.48-1.24)0.280.86 (0.75-0.99)0.04
 Range, ng/mL<505.4505.4 to <616.9≥616.9   
 Case/subcohort, n85/16245/16215/162   
 Model 1, HR (95% CI)1.000.50 (0.32-0.79)0.17 (0.09-0.32)<0.00010.73 (0.64-0.82)<0.0001
 Model 2, HR (95% CI)§1.000.52 (0.33-0.81)0.19 (0.10-0.35)<0.00010.74 (0.65-0.84)<0.0001

There was no evidence of effect modification of the sRAGE or CML-AGE and liver cancer associations by BMI, smoking duration and intensity, serum glucose, diabetes; intake of energy-adjusted red meat, alcohol, or coffee; or trial intervention arm (P for interaction > 0.10), nor the joint effect of serum sRAGE and CML-AGE (P for interaction = 0.34). We stratified the associations by follow-up time (subjects with 5-9 years of follow-up and subjects with ≥10 years of follow-up, based on 52 and 93 cases, respectively) and observed no evidence that the associations differed by time from blood collection to diagnosis (P for interaction > 0.14, data not shown).

As part of another nested case-control study on liver cancer within the ATBC cohort, information on hepatitis B and C titers were measured. Within that study, the prevalence of HBsAg was 1.2% among liver cancer cases (2/167) and 0.7% among controls (6/817), whereas the prevalence of anti-HBc was 15% among cases (25/167) and 7% among controls (57/817) and the prevalence of hepatitis C exposure (anti-HCV) was 5% among cases (8/167) and 0.6% (5/817) among controls (Neal Freedman, personal communication). In the present study, 135 cases and 26 subcohort members had hepatitis B and C information. Only 1 subcohort member and 18 case subjects tested positive for anti-HBc, and one case subject and no subcohort members were positive for HBsAg. Similarly, four case subjects and two subcohort members were positive for anti-HCV. Because hepatitis infection is an important risk factor for liver cancer, though clearly not in our study population, we repeated the analyses after excluding the 19 case subjects and two subcohort members who tested positive for HBsAg, anti-HBc, or anti-HCV, and the associations did not change.


In our prospective study of male Finnish smokers with a low prevalence of chronic HBV and HCV infections, serum levels of sRAGE were associated with a modest reduction in risk of liver cancer. A borderline statistically significant inverse association between continuous sRAGE and liver cancer risk suggests that the soluble receptor of AGEs may protect against the inflammatory effects caused by RAGE activation. We also observed an unexpected, highly statistically significant, inverse association between serum CML-AGE and liver cancer.

Experimental studies suggest the critical role of RAGE activation in liver injury and furthermore indicate that blocking RAGE may mitigate liver injury. For example, in a mouse model of total hepatic ischemia/reperfusion, blockade of RAGE with sRAGE administered via intraperitoneal injections improved survival, protected against hepatocellular necrosis, and enhanced expression of proregenerative cytokine tumor necrosis factor-α.27 In another study, blockade of RAGE with sRAGE attenuated liver injury caused by toxic doses of acetaminophen, and a similar increase in the proregenerative cytokines tumor necrosis factor-α and interleukin-6 was observed.24 These and other rodent models (reviewed by Basta et al.15) support the role of RAGE in liver fibrosis and that blocking RAGE activation may prevent the progression of liver fibrosis, indirectly preventing liver carcinogenesis.

Several human studies have investigated the potential roles of RAGE and sRAGE in inflammation. In the only study that includes HCC, expression of RAGE messenger RNA is lower in normal liver than in chronic hepatitis and is the highest in HCC, suggesting that RAGE activation may be involved in HCC etiology16; this study, however, was cross-sectional, and temporality cannot be determined. In a second study, patients with NASH had significantly lower circulating levels of sRAGE compared with healthy controls with normal liver function tests and liver sonograms.26 Obese Caucasian prepubertal children with nonalcoholic fatty liver disease also had significantly lower sRAGE levels compared with obese but otherwise healthy children.40 Because NASH may precede liver cancer, our borderline statistically significant, inverse association between serum sRAGE and liver cancer is in line with these previous studies. Together, these results suggest that sRAGE may act as a decoy receptor, binding free AGEs and other RAGE ligands, and perhaps mitigating the effects of RAGE activation in the liver.

Because CML-AGE is one of the most abundant AGEs and binds readily with full-length RAGE, we hypothesized that higher serum levels of CML-AGE would be associated with increased risk of liver cancer. Instead, we observed an inverse association. Explanations for these results are unclear. Serum CML-AGE may not be the optimal marker of AGEs or RAGE activation in the liver, because CML-AGE is only one of many RAGE ligands. For example, high mobility group box-1 is another ligand of RAGE that activates the proinflammatory cell signaling cascade and plays a critical role in the mechanisms leading to liver injury.6, 8, 15, 41 One possible explanation for our inverse association is confounding, if levels of CML-AGE were associated with another liver cancer risk factor. For example, we and others42 have shown CML-AGE to be inversely associated with body fat in older adults and preferentially deposited in fat tissues. It is possible that the inverse association between CML-AGE and liver cancer might be explained by body fat. Adjustments for BMI did not affect our results; however, BMI is an imperfect proxy for body fat deposition, particularly in the liver.

As with all observational studies, the present investigation has some limitations. First, the study findings in Finnish male smokers may not be generalizable to other populations that include women and nonsmokers. Nevertheless, the modest inverse association between serum sRAGE and liver cancer is both biologically plausible and in line with laboratory and limited clinical data. Although the ATBC study population included Finnish male smokers, previously observed associations in this population have been replicated in populations that include women and nonsmokers.43, 44 The present study also did not examine other AGEs beside CML-AGE, which may limit our ability to tease out the true role of total AGEs in RAGE activation and subsequent liver cancer etiology. Other RAGE ligands that were not measured in this study, such as high mobility group box-1, may be associated with liver cancer risk. The use of a single measurement of biomarkers at only one point in time may not characterize long-term concentrations, whereas repeated measurements within subjects over time may provide a more accurate assessment of exposure. However, it is rarely feasible to assess multiple time points due to the high cost and logistical complexity of collecting biospecimens from a large number of participants in a cohort study. Although the single, prediagnostic measurements of sRAGE and CML-AGE limits our ability to draw conclusions of the role of the AGE-RAGE axis in the etiology of liver cancer, our study is the first to examine sRAGE and CML-AGE in association with incident liver cancer.

The strengths of our study include the prospective nature and the availability of fasting blood samples collected at least 5 years before the diagnosis of liver cancer, as well as the long follow-up, which reduces the possibility of reverse causality. However, we note that chronic liver disease progression and carcinogenesis can occur over a very long period, and we lacked information on preexisting liver disease at baseline. The low prevalence of HBV and HCV in our population (14% among cases) diminishes the possibility that HBV/HCV status, the primary risk factor of liver cancer in other populations, may confound our results. Moreover, when we performed sensitivity analyses excluding the few subjects who were positive for chronic infection of HBV and HCV, the associations between serum CML-AGE or sRAGE and liver cancer did not change. Finland is known to have among the lowest prevalence of HBV and HCV in the world. In the general population of Finland, the prevalence of HBV and HCV are 0.2% and <2%, respectively,45 which is similar to that of the ATBC study.

In conclusion, the findings from our prospective study among Finnish male smokers support the hypothesis that sRAGE may be protective against liver cancer. Our results, particularly the unexpected inverse association observed with serum levels of CML-AGE, warrant examination in other populations.