Serum Undercarboxylated Osteocalcin as Biomarker of Vitamin K Intake and Risk of Prostate Cancer: A Nested Case-Control Study in the Heidelberg Cohort of the European Prospective Investigation into Cancer and Nutrition

Vitamin K is a fat-soluble vitamin that physiologically acts as a cofactor during the posttranslational γ-carboxylation of vitamin K–dependent proteins, including blood-clotting factors and the bone protein osteocalcin (1). Vitamin K from food includes phylloquinone (vitamin K₁), which is abundant in green leafy vegetables, and the group of menaquinones (vitamin K₂), which mainly occur in fermented dairy products, such as cheese, as well as in meat (2). The aliphatic side chain of dietary relevant menaquinones contains between 4 (menaquinone-4, MK-4) and 14 (MK-14) isoprenoid units. The anticancer activities of vitamin K have been observed in a variety of cancer cell lines, including prostate cancer cells (3). In principle, phylloquinone and menaquinones share growth-inhibitory effects on cancer cells by the modulation of proto-oncogenes that foster cell cycle arrest and apoptosis (4-6). However, the growth-inhibitory potential of menaquinones has been observed to be 5-fold higher than that of phylloquinone (7). We previously investigated the association between the dietary intake of phylloquinone and menaquinones and the risk of prostate cancer in the Heidelberg cohort of the European Prospective Investigation into Cancer and Nutrition (EPIC; ref. 8). We found an inverse association between dietary intake of menaquinones and prostate cancer, especially advanced prostate cancer. Phylloquinone, however, was not associated with prostate cancer risk. This was the first hint from an epidemiologic study that habitual dietary intake of vitamin K may affect the risk of prostate cancer in humans. The advantage of using biomarkers in nutritional epidemiologic studies is that the measurement errors are uncorrelated with measurement errors originating from questionnaire-based assessment of habitual dietary intake (9). Serum undercarboxylated osteocalcin (ucOC) has been proposed as a sensitive functional biomarker of vitamin K nutritional status that is inversely associated with dietary vitamin K intake (10). The enzyme vitamin K epoxide reductase is responsible for the recycling of the active form of vitamin K (11). The vitamin K antagonist warfarin interferes with this enzyme, resulting in reduced blood coagulation. The +2255 polymorphism of the vitamin K epoxide reductase gene (VKORC1) has been reported to be significantly associated with the warfarin dose requirement (12). Carriers of the minor A allele of this polymorphism require the highest warfarin dose for inhibition of blood coagulation and were observed to have lower concentrations of the undercarboxylated forms of osteocalcin and prothrombin (13). These observations suggest that this allele contributes to a higher functional efficiency of the enzyme vitamin K epoxide reductase.

In this nested case-control study, we aimed to further elucidate the association between vitamin K and prostate cancer using serum ucOC as a biomarker of vitamin K status and to assess potential effect modification by the +2255 polymorphism in the VKORC1 gene.

The prospective cohort study EPIC-Heidelberg comprises 25,540 participants, including 11,928 men; at recruitment in 1994 to 1998, men were between 40 and 64 y old. At baseline, information on lifestyle factors, health characteristics, and diet was assessed by self-administered questionnaires and a personal computer–guided interview. Blood samples were provided at baseline by 95.8% of the study participants. The collected blood was drawn into serum or plasma tubes and separated by centrifugation into serum, plasma, buffy coat, and erythrocytes. The aliquots of 0.5 mL are being stored in liquid nitrogen at −196°C.

Follow-up questionnaires were mailed to the participants every 2 to 3 y. All self-reported prostate cancer cases were verified on the basis of medical records and/or death certificates. Information on the tumor-node-metastasis stage and Gleason grade of prostate cancer was extracted from pathology reports originating from the procedures or tests conducted during the initial diagnosis. Advanced-stage prostate cancer was defined as prostate cancer with a tumor-node-metastasis staging score of T₃/T₄, N₁ to N₃, or M₁. Prostate cancer cases with a Gleason sum score ≥7 were considered as high-grade prostate cancer cases. Aggressive prostate cancer was defined as cases in advanced stage or with high Gleason grade, or with prostate cancer as underlying cause of death.

The selection of cases and controls for this nested case-control study was based on all male EPIC-Heidelberg participants with available blood samples who were free of prevalent cancer (except nonmelanoma skin cancer) at baseline. All incident cases of prostate cancer that had occurred by the end of February 2007 were included (n = 250). For each case, two controls were selected, matched by age (5-y age groups) and time of recruitment (6-mo intervals) after an incidence density sampling protocol. The present nested case-control study includes 250 case subjects and 494 control subjects.

Habitual dietary intake was assessed at baseline by a semiquantitative food frequency questionnaire (14). Each food item was specified by the participants with regard to typical portion size and consumption frequency ranging from once or less per month to 5 or more times per day. Dietary intakes of phylloquinone and menaquinones (MK-4 to MK-14) were calculated from the food frequency questionnaires using previously published food composition data based on high-performance liquid chromatography analyses (15-17). Further details on vitamin K intake calculation are given elsewhere (8).

All laboratory analyses were carried out with the laboratory personnel blinded to the case-control status. Commercially available ELISA based on monoclonal antibodies were used to quantify serum concentrations of ucOC (GLU-OC EIA Kit, Takara Biomedical Group) and total intact osteocalcin (iOC, Metra Osteocalcin EIA Kit, Quidel Corporation), which corresponds to total osteocalcin independent of carboxylation status. The intra-assay coefficients of variation of the ucOC and iOC ELISA were 8.1% and 6.0%, respectively. Genomic DNA was extracted from buffy coat and the +2255 polymorphism of the VKORC1 gene (rs2359612) was genotyped by means of RFLP as described previously (12, 13).

The baseline characteristics of the 250 prostate cancer cases and 492 matched controls were compared through paired tests. For this purpose, generalized estimating equation models (logistic or linear regression) accounting for the matched design were used, whereby the characteristic of interest was the dependent variable and the case-control status was the independent variable. Dietary and serum osteocalcin variables were log transformed.

ucOC was expressed relative to iOC (i.e., as ucOC/iOC ratio). The association between serum ucOC/iOC ratio as well as dietary intake of menaquinones and risk of total, advanced-stage, localized-stage, high-grade, low-grade, aggressive, and nonaggressive prostate cancer was analyzed by means of conditional logistic regression calculating odds ratios (OR) and corresponding 95% confidence intervals (95% CI). Analyses were conducted both for categorized variables (quartiles) and continuous variables. Quartile cut points were defined by the distribution of serum ucOC/iOC ratio or menaquinone intake, respectively, among controls. Tests for trend across quartile categories were done by modeling the median values of the quartiles as continuous variables in the matched and multivariate models.

For the analysis of the effect of dietary intake of menaquinones on the risk of prostate cancer, multivariate models were adjusted for phylloquinone intake (μg/d, quartiles), total energy intake (kcal/d, quartiles), calcium intake (mg/d, quartiles), vitamin D intake (μg/d, quartiles), dietary intake of tomatoes/tomato products (g/d, quartiles), smoking status (never, former, current), education (none/primary school, technical school, secondary school, university degree), vigorous physical activity (none, <2 h/wk, ≥2 h/wk), and family history of prostate cancer in first-degree relatives. In the analysis of the association between serum ucOC/iOC ratio and risk of prostate cancer, multivariate models were adjusted for season of blood collection (spring, summer, fall, winter), smoking status, education, vigorous physical activity, and family history of prostate cancer in first-degree relatives.

For analysis of vitamin K status indicated jointly by estimated dietary menaquinone intake and serum ucOC/iOC ratio, we created a vitamin K supply score (18). This score was based on the deciles of menaquinone intake and serum ucOC/iOC ratio. Because of the inverse association between dietary vitamin K and serum ucOC/iOC ratio, the deciles of serum ucOC/iOC were coded from 1 to 10, with 1 indicating the highest ucOC/iOC ratio and 10 indicating the lowest ucOC/iOC ratio. The combined score was calculated as the sum of the menaquinone intake deciles and the reversely coded deciles of ucOC/iOC ratio, resulting in a score ranging from 2 to 20. Thus, the higher the vitamin K supply score, the better the vitamin K status. The association between quartiles of the vitamin K supply score (based on the distribution among control subjects) was analyzed in conditional logistic regression models. Multivariate models were adjusted for season of blood collection, phylloquinone intake, total energy intake, calcium intake, vitamin D intake, dietary intake of tomatoes/tomato products, smoking status, education, vigorous physical activity, and family history of prostate cancer in first-degree relatives. Tests for trend were done by modeling the vitamin K supply score as a continuous variable.

The main effects of the VKORC1 +2255 polymorphism on the risk of total and advanced prostate cancer were examined with the GG genotype being the reference category. To evaluate whether the association between dietary intake of menaquinones or the serum ucOC/iOC ratio, respectively, and prostate cancer was modified by the VKORC1 +2255 genotype, logistic regression models were calculated stratified by genotype. We used unconditional logistic regression models for the stratified analyses, adjusting for the original matching variables age and time of recruitment. Because potential confounders, such as education, smoking status, vigorous physical activity, or dietary intake of calcium, did not significantly differ by VKORC1 +2255 genotype, stratified analyses were carried out without adjustment for these factors. To test for multiplicative interaction, product terms of the VKORC1 +2255 polymorphism (coded as 0, 1, and 2 according to the number of A alleles) and the continuous variables of menaquinone intake or the ucOC/iOC ratio, respectively, were created. P values for interaction were calculated by analyzing conditional logistic regression models with and without the interaction term and comparing them according to the log-likelihood statistic. Because of the low number of cases in the strata, analyses were conducted only for total, aggressive, and nonaggressive prostate cancer.

The baseline characteristics by case-control status are presented in Table 1. There were no significant differences between cases and controls with respect to body mass index and vigorous physical activity. Case subjects had less often a university degree and were more often never smokers than the control subjects although the differences were not statistically significant. Phylloquinone intake did not differ significantly by case-control status whereas median intakes of menaquinones were appreciably lower in cases than in controls (P = 0.006). Apart from dietary calcium intake, which was lower in cases than in controls (P = 0.10), the other dietary intake variables did not differ distinctly by case-control status. The median absolute serum ucOC concentration was 1.36 ng/mL in cases and 1.42 ng/mL in controls, and the ucOC/iOC ratios were 0.16 and 0.17 in cases and controls, respectively. No significant differences in absolute concentration of ucOC (ng/mL) or ucOC/iOC ratio were observed in comparing cases with control subjects. The genotype frequencies of the +2255 polymorphism of the VKORC1 gene were similar to those reported for other Caucasian populations and did not deviate from the Hardy-Weinberg equilibrium (χ² test, P = 0.86 in control subjects). No significant differences in the distribution of the genotypes were observed between case and control subjects.

As reported from the analysis of the full male EPIC-Heidelberg cohort (8), dietary intake of menaquinones was inversely associated with prostate cancer incidence also in this nested case-control study. The multivariate adjusted ORs of total prostate cancer comparing the 2nd, 3rd, and 4th quartiles with the 1st quartile were 0.60 (95% CI, 0.38-0.95), 0.63 (95% CI, 0.36-1.08), and 0.36 (95% CI, 0.19-0.68), respectively. Each 10 μg/d increment in dietary intake of menaquinones was associated with a significant 11% reduction in risk of total prostate cancer (multivariate adjusted OR, 0.89; 95% CI, 0.76-1.00). The OR of aggressive prostate cancer was significantly reduced comparing the 4th with the 1st quartile of menaquinone intake (multivariate adjusted OR, 0.35; 95% CI, 0.13-0.95 and OR per 10 μg/d increment, 0.87; 95% CI, 0.72-1.06). Dietary intake of phylloquinone was not associated with prostate cancer.

As shown in Table 2, serum ucOC/iOC ratio was not associated with total prostate cancer. The multivariate adjusted continuous OR per 0.1 increment in ucOC/iOC ratio was 1.01 (95% CI, 0.91-1.13). With regard to advanced-stage prostate cancer, each 0.1 increment in ucOC/iOC ratio was associated with a significant 38% increase in risk of advanced-stage prostate cancer (OR, 1.38; 95% CI, 1.03-1.86). Positive associations were also observed for high-grade prostate cancer as well as aggressive prostate cancer; multivariate adjusted ORs were 1.21 (95% CI, 0.995-1.46) and 1.18 (95% CI, 0.98-1.41), respectively. In the quartile analyses, we observed increased ORs of advanced-stage, high-grade, and aggressive prostate cancer in the 3rd and 4th quartiles of ucOC/iOC ratio compared with the 1st quartile, which, however, were not statistically significant.

The vitamin K supply score was nonsignificantly inversely associated with total prostate cancer (Table 3). However, subjects in the 4th quartile of vitamin K supply score had a significantly decreased risk of being diagnosed with advanced-stage prostate cancer (OR, 0.25; 95% CI, 0.07-0.96) compared with subjects in the 1st quartile (P for linear trend = 0.01).

We observed no main effects of VKORC1 +2255 genotype on the risk of prostate cancer (data not shown). Analyses stratified by VKORC1 +2255 genotype revealed stronger inverse associations between dietary intake of menaquinones and risk of prostate cancer in homozygous and heterozygous carriers of the A allele compared with GG carriers (Table 4; P_interaction = 0.14). Whereas the OR per 10 μg/d increment was close to one in GG carriers, it was significantly decreased in AG carriers (OR, 0.88; 95% CI, 0.79-0.99) and even lower, although not statistically significant, in carriers of the AA genotype (OR, 0.79; 95% CI, 0.61-1.02). Similar patterns of nonsignificant effect modification were observed for aggressive prostate cancer (P_interaction = 0.14) and nonaggressive prostate cancer (P_interaction = 0.44). The numbers of subjects in the subgroups of the stratified analyses were small, and no interaction effect was statistically significant. With regard to serum ucOC/iOC ratio, a slightly stronger effect in AA carriers was observed with aggressive prostate cancer (P_interaction = 0.37) whereas ORs were homogeneous across VKORC1 +2255 genotypes for total and nonaggressive prostate cancer.

Considering the ucOC/iOC ratio as a biomarker of vitamin K status, the findings of this nested case-control study strengthen the previously hypothesized inverse association between dietary intake of menaquinones and risk of prostate cancer. Serum ucOC/iOC ratio was significantly positively associated with advanced-stage prostate cancer whereas no association was observed with total prostate cancer. The vitamin K supply score combining information on serum ucOC/iOC ratio with dietary menaquinones intake data was significantly inversely associated with advanced-stage prostate cancer. The +2255 polymorphism of the VKORC1 gene was not related to the risk of prostate cancer, but there was some suggestion that the association between dietary intake of menaquinones and prostate cancer may be modified by this polymorphism.

As expected, the results on dietary intake of vitamin K and risk of prostate cancer were in agreement with those obtained from the full cohort analysis (8). Dietary intake of menaquinones was inversely associated with the risk of prostate cancer whereas no association was observed for phylloquinone intake. This observation is consistent with experimental studies showing substantially higher growth-inhibitory effects for menaquinones than for phylloquinone (4-7) and seems plausible in the context of higher bioavailability and longer half-life in plasma of menaquinones compared with phylloquinone (2, 16, 21). Further evidence for the anticancer effects of menaquinones stems from the only intervention study linking vitamin K with cancer incidence in humans. In that study, megadoses of menaquinones reduced the risk of hepatocellular carcinoma among female subjects with viral cirrhosis of the liver by 80% in the intervention group compared with the control group (22).

The primary aim of the present nested case-control study was to confirm the findings on estimated menaquinone intake by means of a biomarker reflective of vitamin K status. For this purpose, we quantified serum ucOC/iOC ratio and evaluated its association with the risk of prostate cancer. Serum ucOC/iOC ratio has been suggested as a sensitive biomarker of extra-hepatic vitamin K status (23). Because a higher ucOC/iOC ratio indicates a poorer vitamin K status, a positive association between ucOC/iOC ratio and risk of prostate cancer was expected. We observed a significant positive association between serum ucOC/iOC ratio and advanced-stage prostate cancer and a borderline significant positive association with high-grade prostate cancer. These results strengthen the hypothesis that vitamin K status plays a role in the etiology of prostate cancer although no association was observed with total prostate cancer. Bone metastases, which occur frequently in advanced prostate cancer cases (24), are associated with increased osteocalcin synthesis (25, 26) and, thus, may influence serum ucOC concentration and ucOC/iOC ratio. Serum samples in the present study were prediagnostic. Hence, the observed results for advanced-stage prostate cancer are unlikely to arise from reverse causation due to increased osteocalcin synthesis among advanced prostate cancer patients. Besides the hypothesis that vitamin K itself is associated with prostate cancer by the beforementioned mechanisms, alternatively it could be hypothesized that there is a link between the carboxylation of osteocalcin and the formation of bone metastasis. This may possibly pose as an alternative explanation of our observed positive association between ucOC/iOC ratio and advanced prostate cancer. However, to our knowledge, no studies have been conducted specifically addressing this question.

Serum ucOC/iOC ratio is a marker of vitamin K status that reflects supply with both phylloquinone and menaquinones (27, 28); because only menaquinones seem to exert relevant cancer-preventive effects (8), it could be expected that the effect of serum ucOC/iOC ratio on prostate cancer will be less prominent than the previously observed effect of estimated dietary menaquinones. In fact, we observed no significant association between serum ucOC/iOC ratio and total prostate cancer but could show a statistically significant association with advanced-stage prostate cancer. Dietary intake estimations are prone to error due to imprecise assessment of dietary intake by the use of food frequency questionnaires and inaccuracies of food composition tables. The use of biomarkers of intake should overcome some of these limitations. However, the measurement of serum ucOC/iOC ratio may also imperfectly reflect vitamin K status. In the present study, serum ucOC/iOC ratio was considered the preferable measure of vitamin K status because it is a functional biomarker that is not as much influenced by recent dietary intake as was found for plasma vitamin K concentrations (10, 23). However, there is a discrepancy in terms of the reflected time period between ucOC/iOC ratio determined from serum samples collected at one point in time and dietary intake data assessed with the use of a food frequency questionnaire covering habitual dietary intake during the previous 12 months. Possibly, ucOC/iOC ratio measured in multiple blood samples collected throughout one year would have provided a better estimate of habitual vitamin K status. Measurement of plasma menaquinones may pose an alternative biomarker of (recent) menaquinone intake although there is evidence of tissue-specific conversion of phylloquinone to MK-4, which subsequently may be released to plasma (29-31). The implementation of this measure in epidemiologic studies is complex because very sensitive high-performance liquid chromatography methods are required and, to date, only certain menaquinones can be detected in human plasma (e.g., MK-4, MK-7; ref. 32).

The measurement errors arising from the quantification of serum ucOC/iOC ratio are expected to be uncorrelated with the measurement errors of dietary menaquinone intake estimation. Both menaquinone intake estimation and serum ucOC/iOC measurement therefore provide independent reports of vitamin K status. It has been shown that the use of two independent exposure reports can provide a less biased risk estimate even if one exposure report is less accurate than the other (33). Therefore, it can be assumed that the vitamin K supply score presented in our study, which is determined equally by estimated dietary menaquinone intake and serum ucOC/iOC ratio, gives a more reliable risk estimate of the association between vitamin K status and prostate cancer than using menaquinone intake or serum ucOC/iOC ratio as individual exposure variables. Thus, the observation that the vitamin K supply score was inversely associated with advanced-stage prostate cancer can be considered as a further support of our study hypothesis.

We also hypothesized that the association between vitamin K status and prostate cancer may be modified by the VKORC1 +2255 genotype, which affects the activity of the vitamin K cycle (12, 34). Carriers of the A allele of this polymorphism are characterized by higher expression of vitamin K epoxide reductase and, thus, by a more active vitamin K cycle (12, 34). Stratified analyses showed stronger inverse associations between dietary intake of menaquinones and risk prostate cancer in heterozygous and homozygous carriers of the A allele although no statistically significant interaction was observed. The observed stronger effects in carriers of the A allele suggest that the vitamin K cycle may play a role in the anticancer activities of menaquinones. Experimental studies report controversial opinions about the involvement of γ-carboxylation in the antitumor activities of vitamin K. Although one study found hints for the growth inhibition induced by vitamin K to be independent of γ-carboxylation (5), another study hypothesized that vitamin K exerts its inhibitory effect through its role in γ-carboxylation (4). Considering the anticancer effects of vitamin K as a process of modulation of transcription factors (3), an involvement of γ-carboxylation is rather unlikely. However, it could be speculated that a high menaquinone turnover due to a high activity of the vitamin K cycle, as is the case in carriers of the A allele, may indirectly enhance the modulating effects of menaquinones on proto-oncogenes. The association between ucOC/iOC ratio and aggressive prostate cancer was slightly more prominent in subjects carrying the AA genotype compared with carriers of the G allele.

The strengths of the present study include the prospective design with a mean follow-up time of 8 years and little loss to follow-up (response rates >92%), the detailed information on the subjects' characteristics that allows for comprehensive control of confounding, and the careful handling of blood samples according to standardized methods.

A major limitation of the present nested case-control study is the limited power, especially for the analysis of prostate cancer subgroups and subanalyses stratified by VKORC1 +2255 genotype. Considering this, we cannot rule out that our findings may be due to chance. Larger studies are required for an estimation of the effect of uOC/iOC ratio on the risk of prostate cancer with sufficient statistical power. Although we were able to control for a large number of potential confounders, we cannot exclude that our findings may be influenced by residual confounding due to measurement errors in confounders accounted for or due to unaccounted factors. For example, because in EPIC-Heidelberg the available data does not allow for quantification of calcium intakes from supplements, the multivariate analyses were adjusted only for calcium intake from food. Thus, although food is the main source of calcium in Germany, we cannot rule out residual confounding by unmeasured calcium intake from supplements.

In conclusion, the increased risks of advanced-stage and high-grade prostate cancer with increasing serum ucOC/iOC ratio strengthen the hypothesis that vitamin K status may play a role in the etiology and progression of prostate cancer. The findings of this nested case-control study may pose as motivation for the initiation of further and larger studies on the association between serum ucOC/iOC ratio or other biomarkers of vitamin K status, such as plasma menaquinones and prostate cancer.

No potential conflicts of interest were disclosed.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.