SEARCH

SEARCH BY CITATION

Keywords:

  • hepatocellular carcinoma;
  • testosterone;
  • sex hormone-binding globulin;
  • insulin-like growth factor I;
  • prospective study;
  • EPIC

Abstract

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

Elevated prediagnostic testosterone and insulin-like growth factor I (IGF-I) concentrations have been proposed to increase risk of hepatocellular carcinoma (HCC). However, the metabolism of these hormones is altered as a consequence of liver damage and they may have clinical utility as HCC risk markers. A case–control study was nested within the European Prospective Investigation into Cancer and Nutrition cohort and included 125 incident HCC cases and 247 individually matched controls. Testosterone, sex hormone-binding globulin (SHBG) and IGF-I were analyzed by immunoassays. Odds ratios (ORs) and 95% confidence intervals (CIs) were estimated by conditional logistic regression. The area under the receiver operating curves (AUC) was calculated to assess HCC predictive ability of the tested models. After adjustments for epidemiological variables (body mass index, smoking, ethanol intake, hepatitis and diabetes) and liver damage (a score based on albumin, bilirubin, aspartate aminotransaminase, alanine aminotransaminase, gamma-glutamyltransferase and alkaline phosphatase concentrations), only SHBG remained significantly associated with risk [OR for top versus bottom tertile of 3.86 (1.32–11.3), ptrend = 0.009]. As a single factor SHBG had an AUC of 0.81 (0.75–0.86). A small, but significant increase in AUC was observed when SHBG was added to a model including the liver damage score and epidemiological variables (from 0.89 to 0.91, p = 0.02) and a net reclassification of 0.47% (0.45–0.48). The observed associations of HCC with prediagnostic SHBG, free testosterone and IGF-I concentrations are in directions opposite to that expected under the etiological hypotheses. SHBG has a potential to be tested as prediagnostic risk marker for HCC. © 2013 UICC

Abbreviations
ALT

alanine aminotransaminase

AP

alkaline phosphatase

AST

aspartate aminotransaminase

AUC

area under the receiver operator curve

BMI

body mass index

CI

confidence interval

CV

coefficient of variation

EPIC

the European Prospective Investigation into Cancer and Nutrition

GGT

gamma-glutamyltransferase

HBV

hepatitis B virus

HCC

hepatocellular carcinoma

HCV

hepatitis C virus

IGF-I

insulin-like growth factor I

NR

normal range

NRI(>0)

continuous net reclassification improvement

OR

odds ratio

SHBG

sex hormone-binding globulin

Elevated circulating concentrations of testosterone and insulin-like growth factor I (IGF-I) have been implicated in the development of several cancers,[1, 2] including hepatocellular carcinoma (HCC).[3, 4]

Impetus for research on testosterone as an etiologic factor for HCC comes from the twofold to fourfold higher HCC incidence in men than in women.[5] However, results from the five prospective studies reported so far have been inconsistent. Positive associations of prediagnostic testosterone with risk of HCC were observed in three studies,[6-8] but no association was reported by another two.[9, 10] Despite the strengths of each of the individual studies, most of them were small and three included 35 cases of HCC or less.[7-9] With one exception,[9] the previous studies were conducted in Asian populations, in which the prevalence of hepatitis infections is high, and some were specifically limited to HBsAg carriers.[6] It has been demonstrated that androgens and its receptor can enhance hepatitis B replication[11] and a cross-talk between HCV and the androgen receptor has also been reported.[12] Interestingly, in a study from China,[10] a positive association of testosterone with risk of HCC was observed only among HBsAg-positive cases and controls, but not among HBsAg-negative participants and no association of testosterone with risk was observed in the only study from Europe.[9] Furthermore, valid characterization of the association of circulating testosterone with HCC requires a concurrent investigation of sex hormone-binding globulin (SHBG), testosterone's major protein carrier in the circulation.[13] Determination of SHBG is essential because it is exclusively of liver origin, its synthesis may increase with liver damage and lead to reduced clearance of the SHBG-bound hormone.[13-15] Free testosterone is of interest, because it represents the bioavailable fraction of the hormone that can readily reach target tissues[1] and its positive association with risk will substantiate an etiological role of testosterone in HCC. Only two of the five prospective studies on HCC and testosterone reported on the associations with free testosterone and SHBG, including 20 and 29 cases of HCC, but the results were conflicting.[7, 9]

Most of circulating IGF-I is synthesized in the liver,[16] the blood concentrations decrease proportionally with the degree of liver damage[17-19] and HCC patients may have lower IGF-I concentrations than patients with a similar degree of liver damage, but without HCC.[20-22] Recently, low IGF-I concentrations at HCC diagnosis were correlated with advanced clinicopathologic parameters and poor survival and were proposed for use in the prognostic stratification of HCC patients.[23] In contrast to a suggested etiological role of elevated IGF-I in HCC,[4] one prospective study reported that low IGF-I concentrations measured at least 5 years before cancer diagnosis are associated with increased risk of liver cancer.[24]

We nested a case–control study within the European Prospective Investigation into Cancer and nutrition (EPIC) cohort, a population with low prevalence of hepatitis infections, to investigate whether associations of HCC risk with prediagnostic concentrations of total and bioavailable testosterone, SHBG and IGF-I provide evidence for an etiologic role for these hormones or, inversely, whether these associations reflect preexisting liver pathology. The potential use of the investigated hormones for HCC risk prediction was also examined.

Material and Methods

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

Study population

The EPIC cohort was designed to identify nutritional, lifestyle, metabolic and genetic risk factors for cancer occurrence.[25] In brief, between 1992 and 2000 about 520,000 apparently healthy men and women from ten European countries (Denmark, France, Germany, Greece, Italy, The Netherlands, Norway, Spain, Sweden and the United Kingdom), of ages mostly between 35 and 75 years, were enrolled into the study. Dietary, life style and medical history data were collected through standardized questionnaires and anthropometric measures were recorded. Information on race was not recorded, but the vast majority of the recruited subjects were Caucasian, reflecting the composition of the European populations. Blood samples were collected from 385,747 participants. The protocol of the EPIC study was approved by the ethical committees of the International Agency for Research on Cancer and of the participating centers. All participants provided informed consent, and procedures were in line with the Helsinki Declaration for human rights.

Follow-up for cancer incidence and vital status

Incident cancers were identified through linkages with regional cancer registries except in France, Germany, Greece and Naples (Italy) where follow-up was conducted by review of health insurance records, contacts with cancer and pathology registries and/or direct contact with cohort members. Vital status was ascertained through linkages with regional and national mortality registries and active follow-up (Germany and Greece). For our study, the latest dates of complete follow-up for cancer incidence and vital status in the EPIC centers ranged from 2002 to 2006.

Selection of case and control subjects

We identified 156 cases with liver cancer (ICD-O-2 code C22.0) who had provided a blood sample and had no prior cancer diagnosis. Cases with non-HCC histology (e.g., cholangiocarcinoma, adenocarcinoma and large cell lymphoma) were excluded (n = 31). Of the 125 eligible primary liver cancers, 105 had histologically verified HCC (ICD-O code 8170). No HCC case was excluded on the basis of possible etiology.

Two controls per case were selected at random from risk sets that included all cohort members who had donated a blood sample, were alive and free of cancer at the time of HCC diagnosis of the index case, were not selected for other EPIC case–control studies and matched the case on recruitment center, sex, age (±12 months), date (±2 months), fasting status (<3, 3–6 and >6 hr) and time of the day (±3 hr) at blood collection. Women were additionally matched for menopausal status (premenopausal, peri/unknown and postmenopausal) and exogenous hormone use (yes, no and missing) at blood donation. For three cases, only one eligible control was available.

Laboratory analyses

For each hormone, plasma samples from cases and their matched controls were analyzed in the same analytical batch and laboratory personnel were unaware of the case–control status of the samples. Testosterone and SHBG were quantified by a radioimmunoassay (IM1119, Beckman Coulter) and immunoradiometric assay (2098507, CisBio), respectively. IGF-I was measured using an enzyme-linked immunosorbent assay (ELISA, DSL-10-2800, Beckman Coulter) with an acid-ethanol precipitation step before quantification to eliminate IGF-binding proteins. Mean intra-assay and interassay coefficients of variation were 7.4 and 17% for a testosterone concentration of 0.25 ng/mL, 4.6 and 9.7% for an SHBG concentration of 20 nmol/L and 6.4 and 9.5% for an IGF-I concentration of 73 ng/mL, respectively. Free testosterone concentrations were calculated from measured total testosterone, SHBG and albumin using mass action equations.[26]

Concentrations of albumin [normal range (NR): 35–50 g/L], total bilirubin (NR: 3.4–20.5 µmol/L), alanine aminotransaminase (ALT, NR: <55 U/L), aspartate aminotransaminase (AST, NR: 5–34 U/L), gamma-glutamyltransferase [GGT, NR: 12–64 U/L (men); 9–36 U/L (women)] and alkaline phosphatase (AP, NR: 40–150 U/L) were quantified on an ARCHITECT analyzer. Hepatitis B surface antigen and antibodies to hepatitis C virus were measured using ARCHITECT chemiluminescent microparticle immunoassays (Abbott Diagnostics, Lyon, France).[27]

Statistical analysis

Spearman's correlation coefficients were calculated between hormones and continuous variables. General linear models were applied to compare mean hormone differences across categories of risk factors in cases and controls separately with adjustments for EPIC center, sex, age and fasting status at blood donation.

The associations between hormones and risk of HCC were assessed by conditional logistic regression. Sex-specific tertiles were based on the distributions of the analytes among the controls. Likelihood ratio tests were applied to assess linear trends in odds ratios with assigned quantitative scores 1, 2 and 3 for the tertiles. Multivariable models were adjusted for factors known to be associated with both liver cancer and hormone concentrations: body mass index (BMI, in categories <25.0, 25.0–29.9 and ≥30 kg/m2), smoking (never, former and current), habitual alcohol consumption at recruitment (sex-specific categories of low, moderate and high consumption), diabetes at recruitment (yes or no) and chronic HBV or HCV infection (yes or no). The few missing data on major risk factors were coded in a separate “missing” category for each variable. Models were further adjusted for liver damage. A composite quantitative liver damage score was constructed (range 0–6), with each of the six hepatic markers contributing 1 point when outside of the normal range. A similar pattern of the associations between HCC risk and hormone concentrations was observed after adjustments for liver damage by including either the six liver markers on continuous scale simultaneously or categories of the liver damage score. The few missing liver marker values (n = 13) were replaced by the case–control sex-specific median concentrations.

Hormone-risk associations were also examined by median lag time to diagnosis and by sex and among those with histologically confirmed HCC, with no indication of hepatitis infection or with all liver markers within the normal range or with up to one marker outside the normal range. Risk estimates for these additional analyses were calculated on the continuous scale of the log2-transformed hormone variables, a unit increase of which corresponds to a doubling of concentrations. Statistical heterogeneity was tested by Cochran's Q-statistics.

The risk prediction ability of logistic regression models including the major confounders and the liver damage score (basic model) with addition of SHBG or IGF-I or both was evaluated by calculating the area under the receiver operator curve (AUC). Improvement of the predictive ability of the basic model after addition of SHBG, IGF-I or both was assessed by a nonparametrical test[28] and by computing the continuous net reclassification improvement [NRI(>0)], which is independent from the strength of the baseline model.[29]

Two-sided p-values below 0.05 were considered as statistically significant. SAS was used for all statistical analyses (SAS statistical software, version 9, SAS Institute, Cary, NC).

Results

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

Baseline characteristics of the study participants are summarized in Table 1. Median ages at blood donation and cancer diagnosis were 60.6 and 65.3 years, respectively. Median lag time between recruitment and HCC diagnosis was 4.7 years. Case subjects were more frequently obese, reported more often smoking, high ethanol intake and diabetes at recruitment and more frequently tested positive for HBV or HCV infections (14 and 22%, respectively, versus only 2–3% in the controls). Women were predominantly postmenopausal (95%).

Table 1. Selected characteristics of HCC cases and their matched controls at enrolment in the EPIC study [median (range) or number (percentage)]
CharacteristicCases (n = 125)aControls (n = 247)
  1. a

    Distribution of cancer cases (all/histologically confirmed) across EPIC countries was: 31/25 (Denmark), 22/21 (Germany), 12/9 (Greece), 20/15 (Italy), 4/4 (the Netherlands), 7/4 (Spain), 21/19 (Sweden) and 8/8 (the United Kingdom).

  2. b

    Low intake: men (0–<10 g/day), women (0–<5 g/day); moderate: men (10–<40 g/day), women (5–<20 g/day); high: men (≥40 g/day), women (≥20 g/day).

  3. c

    HBsAg positive when ≥0.05 IU/mL; HCV positive when the ratio of sample relative light units to cutoff relative light units ≥1 in two measurements.

  4. d

    Geometric mean (10th–90th percentile).

  5. e

    A quantitative liver damage score was constructed from concentrations of albumin, total bilirubin, AST, ALT, GGT and AP, each contributing 1 point when outside of the normal range.

Sex  
Men85 (68%)168 (68%)
Women40 (32%)79 (32%)
Age at blood collection (years)60.5 (46.1–77.1)60.6 (45.7–77.0)
Age at diagnosis (years)65.3 (47.6–84.6) 
Lag time (years)4.7 (0.01–13.3) 
BMI (kg/m2)  
<2539 (31%)73 (30%)
25–3049 (39%)124 (50%)
≥3037 (30%)50 (20%)
Smoking  
Never34 (27%)104 (42%)
Former41 (33%)95 (38%)
Current48 (38%)47 (19%)
Ethanol intake at baseline (g/day)b 
None to low70 (56%)123 (50%)
Moderate29 (23%)96 (39%)
High26 (21%)28 (11%)
Chronic HBV/HCV infectionc 
No82 (66%)231 (94%)
Yes40 (32%)12 (5%)
Diabetes  
No105 (84%)222 (90%)
Yes16 (13%)16 (6%)
Testosterone (nmol/L)d  
Men8.33 (3.82–15.3)7.29 (3.82–12.8)
Women0.69 (0.35–1.74)0.69 (0.35–1.39)
Free testosterone (nmol/L)d  
Men0.39 (0.20–0.77)0.50 (0.29–0.86)
Women0.03 (0.01–0.08)0.04 (0.02–0.09)
SHBG (nmol/L)d  
Men52 (25–122)28 (15–49)
Women61 (25–145)34 (17–73)
IGF-I (nmol/L)d  
Men8.12 (3.80–18.0)13.23 (8.52–21.1)
Women8.78 (5.24–16.24)12.18 (7.47–20.3)
Liver damage scoree  
034 (29%)194 (84%)
117 (14%)30 (13%)
219 (16%)7 (3%)
335 (29%)1 (0%)
410 (8%)0 (0%)
52 (2%)0 (0%)
62 (2%)0 (0%)

Among controls, there were no significant differences in hormone concentrations by HBsAg and anti-HCV positivity, diabetes, smoking or alcohol use at recruitment with the exception of increased testosterone and free testosterone concentrations in the heavy alcohol drinkers. In HCC cases, significantly increased SHBG and testosterone were observed in hepatitis-positive versus -negative subjects. Spearman's correlation coefficients between hormones and liver markers in cases and controls separately are presented in Table 2. Testosterone and free testosterone were not correlated with the measured liver markers in either the case or control groups. By contrast, distinct correlations of SHBG with bilirubin, AST, ALT, GGT and AP and of IGF-I with albumin, bilirubin, GGT and AP were observed in the case and control groups. SHBG was inversely correlated with IGF-I, but the correlation was stronger in the cases than in controls (male cases r = −0.56, male controls r = −0.10, female cases r = −0.34, female controls r = −0.24). A moderate positive correlation of SHBG with testosterone was observed in men in both case and control groups (r = 0.59 and r = 0.52, respectively), but not in women. SHBG was inversely correlated with free testosterone in the male cases (r = −0.21) and more strongly so in the female cases (r = −0.46) and controls (r = −0.48).

Table 2. Spearman's correlation coefficients between endogenous hormones and liver markers in case and control subjects (correlations in the controls are shown within parentheses)a
VariableAlbuminBilirubinASTALTGGTAP
  1. a

    Analyses based on 119 cases and 235 controls, respectively, adjusted for EPIC center, sex, age at blood donation and fasting status.

  2. Significant correlation coefficients are given in bold. *p < 0.001, †p < 0.01, ‡p < 0.05.

Testosterone0.08 [−0.03]0.13 [0.01]0.12 [−0.01]0.14 [−0.05]0.02 [0.04]0.03 [−0.09]
Free testosterone0.08 [−0.10]−0.15 [0.04]−0.13 [0.00]−0.05 [0.02]−0.07 [0.09]−0.17 [−0.13]
SHBG−0.09 [−0.01]0.31† [−0.07]0.34† [−0.02]0.22‡ [−0.16‡]0.06 [−0.15‡]0.19 [−0.08]
IGF-I0.30† [−0.14‡]0.46* [0.07]0.33† [−0.24†]−0.20 [−0.15‡]0.23‡ [−0.09]0.27† [0.12]

Both male and female cases had significantly higher SHBG, but lower IGF-I and free testosterone in comparison with respective sex-matched controls (Table 1). Total testosterone concentrations were marginally higher in the male cases (p = 0.06), but not in women. Levels of albumin were lower and all other liver markers were higher in cases than in controls (data not shown). Conditional logistic regression analysis showed positive associations of HCC with testosterone and SHBG and inverse association with free testosterone and IGF-I. After adjustment for major HCC risk factors and the liver damage score, all associations were attenuated and only SHBG and IGF-I (for doubling of concentrations) remained significantly associated with risk (Table 3).

Table 3. OR (95% confidence interval) from conditional regression models for HCC across tertiles and for doubling of circulating concentrations of testosterone, free testosterone, SHBG and IGF-I
 TertilesDoubling of concentrations
 Tertile 1 [cases/controls]Tertile 2 [cases/controls]Tertile 3 [cases/controls]ptrendOR [cases/controls]pvalue
  1. a

    Adjusted for BMI (<25.0, 25.0–29.9 and ≥30), smoking (never, former, current and missing), ethanol at baseline (low, medium and high), HBsAg/anti-HCV (negative, positive and missing) and diabetes (no, yes and missing).

  2. b

    Adjusted as above plus a score based on albumin, bilirubin, AST, ALT, GGT and AP concentrations outside normal range (0–6).

Testosterone[36/83][25/82][64/82] [125/247] 
Cruderef.0.82 (0.44–1.51)2.50 (1.30–4.79)0.0041.45 (1.03–2.05)0.03
Adjustedaref.0.48 (0.21–1.11)1.85 (0.76–4.46)0.111.35 (0.87–2.09)0.18
Adjusted bref.0.34 (0.11–1.01)1.22 (0.42–3.58)0.441.09 (0.64–1.86)0.75
Free testosterone[60/79][33/79][27/78] [121/236] 
Cruderef.0.44 (0.24–0.80)0.34 (0.18–0.66)0.0010.47 (0.32–0.68)<0.001
Adjustedaref.0.54 (0.25–1.18)0.46 (0.19–1.11)0.080.57 (0.36–0.90)0.02
Adjustedbref.0.61 (0.23–1.62)0.68 (0.24–1.93)0.460.60 (0.35–1.03)0.07
SHBG[12/83][20/82][93/81] [125/246] 
Cruderef.2.09 (0.90–4.84)9.88 (4.49–21.7)<0.0015.32 (3.42–8.28)<0.001
Adjustedaref.1.83 (0.73–4.59)6.64 (2.58–17.1)<0.0014.38 (2.59–7.41)<0.001
Adjustedbref.1.52 (0.52–4.39)3.86 (1.32–11.3)0.0092.88 (1.62–5.12)0.001
IGF-I[88/83][19/82][18/82] [125/247] 
Cruderef.0.18 (0.10–0.35)0.12 (0.06–0.26)<0.0010.19 (0.12–0.31)<0.001
Adjustedaref.0.33 (0.16–0.69)0.21 (0.09–0.50)<0.0010.24 (0.13–0.42)<0.001
Adjustedbref.0.86 (0.35–2.14)0.41 (0.14–1.18)0.100.45 (0.24–0.85)0.01

In analyses limited to cases diagnosed after the median lag time between blood donation and cancer diagnosis (≥4.7 years), risk estimates for HCC–hormone associations were somewhat stronger in comparison with the overall study findings (Tables 3 and 4). Formal tests for homogeneity were rejected for testosterone, but ORs for free testosterone were very similar for cases diagnosed before and after the median lag time. Risk estimates for SHBG were several times higher for cases diagnosed more than 4.7 years before blood donation in comparison to those diagnosed within 4.7 of blood donation, but the homogeneity test reached only borderline significance (phomogeneity = 0.06)

Table 4. OR (95% confidence interval) from conditional logistic regression models for HCC for doubling of testosterone, free testosterone, IGF-I and SHBG by median lag time between blood donation and cancer diagnosis
 Lag < 4.7 years [cases/controls]Lag ≥ 4.7 years [cases/controls]phomogeneity
  1. a

    Adjusted for BMI (<25.0, 25.0–29.9 and ≥30), smoking (never, former, current and missing), ethanol at baseline (low, medium and high), HBsAg/anti-HCV (negative, positive and missing) and diabetes (no, yes and missing).

  2. b

    Adjusted as above plus a score based on albumin, bilirubin, AST, ALT, GGT and AP concentrations outside normal range (0–6).

Testosterone[61/121] [64/126]  
Crude1.11 (0.73–1.70)0.632.19 (1.23–3.92)0.0080.06
Adjusteda0.99 (0.55–1.78)0.972.38 (1.06–5.37)0.040.09
Adjustedb0.59 (0.24–1.42)0.243.00 (0.91–9.88)0.070.03
Free testosterone[57/112] [64/124]  
Crude0.57 (0.37–0.89)0.010.33 (0.17–0.64)0.0010.18
Adjusteda0.57 (0.31–1.04)0.070.45 (0.19–1.05)0.070.67
Adjustedb0.43 (0.16–1.15)0.090.54 (0.20–1.48)0.230.75
SHBG[61/121] [64/125]  
Crude3.48 (2.09–5.78)<0.00110.92 (4.50–26.52)<0.0010.03
Adjusteda3.55 (1.68–7.48)<0.00114.66 (4.10–52.47)<0.0010.06
Adjustedb2.22 (0.91–5.45)0.0811.12 (2.61–47.37)0.0010.06
IGF-1[61/121] [64/126]  
Crude0.30 (0.17–0.51)<0.0010.09 (0.04–0.23)<0.0010.03
Adjusteda0.33 (0.15–0.70)0.0040.12 (0.04–0.37)<0.0010.15
Adjustedb0.50 (0.19–1.28)0.150.21 (0.06–0.83)0.030.32

Subgroup analyses by sex or restricted to histologically confirmed HCC (105 case–control sets) or to subjects with no indication for hepatitis B or C infection (82 case–control sets) were largely similar to those overall (data not shown). The same pattern of associations was observed in sensitivity analyses restricted to subjects with liver markers within their normal ranges (32 case–control sets) or subjects with up to one marker altered (51 case–control sets). These analyses, however, were based on very limited number of subjects and risk of HCC was significantly associated only with SHBG in subjects with up to one liver marker exceeding the normal range (Table 5).

Table 5. OR (95% confidence interval) from conditional regression models for HCC for doubling of circulating concentrations of testosterone, free testosterone, SHBG and IGF-I limited to subjects with either all studied liver markers (albumin, total bilirubin, AST, ALT, GGT and AP) within the normal range or with alterations in up to one marker
 All markers within the NR [cases/controls]Up to one marker outside the NR [cases/controls]
  1. a

    Adjusted for BMI (<25.0, 25.0–29.9 and ≥30), smoking (never, former, current and missing), ethanol at baseline (low, medium and high), HBsAg/anti-HCV (negative, positive and missing) and diabetes (no, yes and missing).

Testosterone[31/52] [51/98] 
Crude1.09 (0.54–2.22)0.801.09 (0.66–1.79)0.74
Adjusteda1.65 (0.62–4.39)0.311.36 (0.72–2.57)0.34
Free testosterone[31/52] [51/98] 
Crude0.69 (0.40–1.19)0.190.68 (0.44–1.03)0.07
Adjusteda0.94 (0.46–1.93)0.870.75 (0.44–1.27)0.28
SHBG[31/52] [51/98] 
Crude2.33 (1.23–4.42)0.0092.45 (1.48–4.05)0.0005
Adjusteda1.82 (0.84–3.94)0.132.48 (1.34–4.59)0.004
IGF-I[31/52] [51/98] 
Crude0.49 (0.21–1.13)0.090.49 (0.27–0.90)0.02
Adjusteda0.76 (0.26–2.25)0.620.55 (0.25–1.20)0.13

The receiver operator characteristic analyses showed an AUC of 0.81 (0.76–0.86) for SHBG and 0.76 (0.71–0.82) for IGF-I as single predictors (Fig. 1). When compared with the basic prediction model (including epidemiological variables and the liver damage score) only the addition of SHBG resulted in significant (p = 0.02), albeit small increase in the AUC from 0.89 (0.85–0.93) to 0.91 (0.88–0.95) (Fig. 1). The NRI(>0) indicated excellent improvement in net reclassification after addition of SHBG [0.47 (0.45–0.48), p < 0.001] or IGF-I [0.25 (0.24–0.27), p = 0.01] to the basic model. In subjects whose liver markers were all within the normal range the addition of both SHBG and IGF-I increased the AUC of the basic model from 0.84 (0.76–0.93) to 0.88 (0.81–0.95), p = 0.09 and NRI(>0) was 0.61 (0.56–0.66), p = 0.003.

image

Figure 1. AUC for HCC for the basic model (including epidemiologic variables and a liver damage score), SHBG, IGF-I and for the basic model with SHBG (121 cases and 237 controls). Epidemiologic variables include BMI, smoking, alcohol intake, HBsAg/anti-HCV and diabetes.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

In our prospective study, we observed a strong positive association between risk of HCC and prediagnostic concentrations of SHBG. Risk estimates for SHBG concentrations were attenuated, but remained statistically significant after adjustment for major confounders (BMI, smoking, alcohol consumption at recruitment, diabetes and hepatitis infection) and a liver damage score. HCC was inversely associated with prediagnostic IGF-I, but after adjustment for liver damage, the association remained significant only for doubling of hormone concentrations. Testosterone and free testosterone were not related to HCC risk in the fully adjusted models. In comparison with the overall results, risk estimates were somewhat stronger in analyses limited to subjects with follow-up time above the median (4.7 years). Results from analyses stratified by sex or limited to histologically verified HCC or to subjects testing negative for hepatitis infection were similar to those from the overall study.

The lack of association of HCC with testosterone after adjustments in our data, as also observed in some of the previous prospective investigations,[8, 9] together with the lack of indication for a positive association with the concentrations of the free, bioavailable fraction of the hormone argue against a strong etiological role of testosterone in HCC in the 5–10 years preceding disease diagnosis in a population with low prevalence of hepatitis infection.

The directions of the association of HCC with IGF-I and SHBG in our data are in contrast to what could be postulated from an etiological view point. Elevated IGF-I would be expected to increase risk because of its mitogenic and antiapoptotic properties, which are believed to underlie its association with a number of cancers.[2, 4] Similarly, elevation in SHBG would be expected to be inversely related to risk as it reduces the testosterone's bioavailable fraction. Instead, the observed associations of IGF-I and SHBG with HCC risk and with markers of liver damage in future HCC cases suggest that alterations in their concentrations reflect a developing liver pathology that could ultimately culminate in HCC. Such an interpretation is supported by data from two longitudinal and one recent prospective study, but all including a small number of HCC cases.[9, 21, 24]

For IGF-I, in a group of 114 patients with mild cirrhosis (Grade A) in which 20 HCC were diagnosed during a mean follow-up of 56 months, a significant decrease in IGF-I concentrations (∼50%) was observed about 9 months before a subsequent HCC diagnosis, independent of liver function impairment.[21] The recent prospective study by Major et al. suggests that the decrease in IGF-I concentrations may be evident more than 5 years before HCC diagnosis.[24] The lower circulating IGF-I in patients with chronic liver disease is believed to be a consequence of direct damage to the liver parenchyma, the decrease in the expression of liver growth hormone receptors (growth hormone resistance) and organ inflammation.[30] Furthermore, animal experiments[31-34] and preliminary human data[35] suggest that reduced liver synthesis of IGF-I may, in turn, aggravate an existing liver damage.

For SHBG, in 101 male patients with cirrhosis in which 29 HCC were diagnosed during a mean follow-up of 38 months, elevations in SHBG had an independent predictive value for subsequent diagnosis of HCC after controlling for a number of liver damage markers, cirrhosis etiology and stage.[9] The mechanisms underlying an increase in SHBG in some chronic liver conditions remain to be identified,[9, 14, 15] but it is likely that a decreased liver IGF-I production plays a role. In healthy men and women IGF-I downregulates SHBG production by the liver, although the strength of the inverse association between circulating IGF-I and SHBG concentrations is generally modest.[2] In patients with chronic liver damage and hence a significantly diminished hepatic IGF-I production, the escape of SHBG from IGF-I downregulation may be particularly pronounced. In support, in our data, the inverse correlation between IGF-I and SHBG was stronger in the cases than in the controls.

In our data, both SHBG and IGF-I were strong individual predictors of HCC risk (AUC of 0.81 and 0.76, respectively). Although only addition of SHBG increased slightly (but significantly) the AUC of the basic model (including epidemiological and a liver damage score), the NRI(>0), which is independent of the predictive ability of the baseline model, indicated excellent improvement in net reclassification after addition of either SHBG or IGF-I. The association of SHBG with HCC risk was stronger than that for IGF-I and its predictive ability [as estimated by AUC and NRI(>0)] was also superior to that of IGF-I. However, as discussed above, elevations in SHBG and its performance as a predictive marker may result from decreased liver IGF-I production. Additionally, circulating SHBG is exclusively of liver origin, whereas IGF-I is produced in many tissues of the body.[36] It has been shown that the extrahepatic contribution to circulating IGF-I increases in a background of decreased liver synthesis of the hormone.[36]

The strengths of our study are its prospective design in a population with low prevalence of hepatitis infection, the availability of data on major risk factors for HCC and a set of markers of liver damage. Our study has several limitations. Despite being the largest prospective study on endogenous hormones and HCC, the statistical power for detailed subgroups analyses and homogeneity tests was limited. Liver damage markers were not analyzed at recruitment (not feasible in a large prospective study), but in stored specimens, precluding the analyses of additional established markers (e.g., platelet counts). Despite recruitment of apparently healthy men and women, most cases (but not controls) had a number of liver damage markers outside the normal range, indicating that the presence of some degree of liver pathology and adjustments for liver markers substantially attenuated risk estimates. No information about cirrhosis or use of androgen-lowering medication (e.g., for treatment of benign prostate hyperplasia) at baseline or during follow-up is available in EPIC. Nevertheless, the association of HCC with prediagnostic concentrations of SHBG was strong and remained significant in most fully adjusted subgroup analyses, despite that they were based on small number of subjects. Finally, it would have been of interest to have information on additional hormones, such as estrogens or other androgens. It will take future clinical studies to determine if prediagnostic concentrations of SHBG and IGF-I could contribute to the building of risk prediction models independently of other widely used markers of liver damage and function also in patients with chronic liver damage, in whom the associations with risk are expected to be less strong. Nevertheless, given recent encouraging findings on the potential use of IGF-I in the prognostic stratification of HCC patients,[23] it would be of interest to explore utility of SHBG also in such setting.

In summary, our study provides evidence that prediagnostic increased SHBG concentrations are associated with greater risk of HCC and reveal the potential of this binding protein and of IGF-I to be included for testing as risk markers in patients at elevated risk of HCC. After accounting for major risk factors and liver damage, our data do not support a strong etiological role of testosterone or IGF-I in HCC in the decade preceding diagnosis in a population with low prevalence of hepatitis infection.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

The authors are grateful for the helpful discussions with Profs. Drs. P. Schirmacher and M. Müller and Dr. Jutta Kneisel. They thank Mrs. Bettina Ehret, Ms. Britta Lederer and Ms. Sigrid Henke for their meticulous technical work during the performance of the analyte measurements. Reagents for the hepatitis infection determinations were kindly provided by Abbott, Lyon, France.

References

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References