In this prospective study of 101,695 American adults, we found that dietary intakes of phylloquinone and dihydrophylloquinone were inversely associated with the risk of pancreatic cancer in a nonlinear dose–response pattern. Moreover, these observed associations were not modified by predefined stratification variables and remained in sensitivity analyses. No significant association was found for dietary menaquinone intake and the risk of pancreatic cancer.
Vitamin K acts first and foremost as a cofactor for γ-glutamyl-carboxylase, which catalyzes the posttranslational carboxylation of glutamate residues in coagulation factors. Therefore, promoting blood coagulation is a primary function of vitamin K initially recognized by scientific communities. During the past decade, other health impacts of vitamin K have been successively revealed. Extensive epidemiologic and clinical studies have investigated the roles of vitamin K in cardiovascular health. However, evidence of the association of dietary vitamin K intake with the risk of cancer is limited and inconclusive. Specifically, researchers in 2 prospective studies in European populations 40–65 years of age found that dietary intake of menaquinones, but not phylloquinone, was inversely associated with risks of prostate and lung cancers,[18,19] whereas our group recently found that a higher dietary intake of menaquinones, but not phylloquinone, conferred higher incidence of death from breast cancer in a cohort of 51,662 American women aged 55–74 years. On the basis of prospective data of the PLCO Cancer Screening Trial, we observed for the first time that subjects with higher intake of phylloquinone or dihydrophylloquinone, but not menaquinones, were at a lower risk of pancreatic cancer. In fact, investigators in an experimental study found that sorafenib in combination with phylloquinone resulted in stronger growth inhibition on human pancreatic cancer cells than di sorafenib alone, indicating the potential antitumor ability of phylloquinone, which supports our results to some degree. On one hand, our observation suggests that increasing dietary intake of phylloquinone or dihydrophylloquinone could be an attractive approach for reducing the risk of pancreatic cancer, which is especially relevant in the US population, in which up to 57.0% of men and 37.5% of women do not meet the adequate intake for phylloquinone (120 μg/day and 90 μg/day for men and women, respectively). On the other hand, it reminds us that phylloquinone and dihydrophylloquinone possibly play a role in the etiology of pancreatic cancer.
Dietary supplement use is common in the US adult population, with 51% consuming multivitamin/minerals that comprise ≥9 micronutrients from 2007 to 2010. An observational study of 30,899 US adults aged ≥20 years found that the mean vitamin K intake from supplements was 29.6 (SD, 0.8) μg/day, which accounted for 21.8% of total vitamin K intake; moreover, a large randomized controlled trial of 14,641 US men aged ≥50 years showed that daily multivitamin use significantly decreased the risk of overall cancer (HR = 0.92, 95% CI: 0.86, 1.00). Hence, supplemental vitamin K is a significant source of total vitamin K intake and can impact the ranking of an individual's vitamin K intake and the estimation of risks or benefits involved. Thus, ideally, supplemental vitamin K intake should be considered in the assessment of its health impacts. Unfortunately, in the PLCO Cancer Screening Trial, data on supplemental vitamin K are not available; therefore, we cannot directly assess its potential role in the association of interest. Nevertheless, our subgroup analysis found that the observed association of dietary vitamin K intake with the risk of pancreatic cancer could not be modified by single or multivitamin supplement use, indirectly reminding us that supplemental vitamin K seems to have minimal influence on the observed association. In fact, supplemental vitamin K intake has been found to be not associated with overall cancer mortality in the US adults, which provides some supports for its above-mentioned supposed minimal impacts on the association of interest, at least in part.
The observed association of dietary vitamin K intake with the risk of pancreatic cancer may be explained by several potential mechanisms. First, phylloquinone intake has been related to reduced levels of proinflammatory cytokines, such as interleukin-6 and tumor necrosis factor-α, which suggests that phylloquinone has inhibitory impact on inflammation. It is now well known that chronic inflammation is an important driver of cancer initiation and progression. Second, vitamin K can function as a cofactor in the activation process of vitamin K-dependent proteins, including protein C and protein S. Notably, endogenous activated protein C has the ability to limit cancer cell extravasation; protein S overexpression can inhibit the proliferation of human pancreatic cancer cells. These facts indicate that the observed association between dietary vitamin K intake and the risk of pancreatic cancer may be dependent on the activation of vitamin K–dependent proteins. Third, a randomized controlled trial showed that compared with placebo, phylloquinone supplementation resulted in a significant increase in the serum level of adiponectin, which has been found to be inversely related to the risk of pancreatic cancer and to suppress the proliferation of human pancreatic cancer cells via inactivating β-catenin signalling. In addition, vitamin K has been suggested to act as a potent antioxidant; our recent work showed that higher total antioxidant capacity from diet conferred a lower risk of pancreatic cancer. Hence, the increased serum level of adiponectin and the attenuated oxidative stress induced by vitamin K may account for its association with the risk of pancreatic cancer. Fourth, phylloquinone intake has been found to be related to a reduced risk of type 2 diabetes and improved insulin resistance. Thus, it is possible that the inverse association of phylloquinone intake with the risk of pancreatic cancer is mediated by the above 2 conditions, given their roles in the etiology of pancreatic cancer.[51,52] Finally, the inverse association of dietary vitamin K intake with the risk of pancreatic cancer may also be mediated by diet quality. Nevertheless, such an explanation seems to be not supported by the fact that this inverse association was not modified by Healthy Eating Index-2015 score in subgroup analysis and remained after further adjustment for Healthy Eating Index-2015 score in sensitivity analyses. The exact reasons for the differential association of phylloquinone (or dihydrophylloquinone) and menaquinone with the risk of pancreatic cancer are unclear and may be related to their differences in half-life, tissue distribution, and bioavailability in the human body.
Our study has several limitations. First, dietary vitamin K intake was evaluated using a food frequency questionnaire and thus might be susceptible to measurement errors resulting from potential recall bias. Nevertheless, this bias is nondifferential, considering that it is not expected to be related to the future risk of pancreatic cancer, and thus tends to weaken the association of interest (i.e., the true association is expected to be stronger). Second, food consumption used for the estimation of dietary vitamin K intake was evaluated once at baseline in the PLCO Cancer Screening Trial. Given that dietary habits may alter over time, the evaluation of food consumption at one time point may lead to nondifferential bias. Nevertheless, an adult's dietary habits are not expected to alter dramatically during several years; additionally, it has been indicated that the methods using baseline diet in general result in a weaker association than do those using the cumulative averages. Third, we fully adjusted for the potential confounders, but our results may still be biased by residual confounding because of the observational design of our study. For example, warfarin is a frequently used vitamin K antagonist in clinical practice; its use has been found to be associated with a reduced risk of overall cancer (incidence rate ratio = 0.84, 95% CI: 0.82, 0.86). However, data on warfarin use were not available in this study, resulting in our inability to adjust for this potential confounder. Fourth, in our subgroup analyses, we found that the association between dietary vitamin K intake and the risk of pancreatic cancer could not be modified by prespecified stratification variables. However, it should be noted that these subgroup analyses only included a small number of pancreatic cancer cases. Thus, a likelihood ratio test may have insufficient power to detect the potential interactions. Fifth, although the direction and the magnitude of the initial associations did not change substantially in both 2-year and 4-year lagged analyses, the possibility of reverse causation cannot be ruled out in the present study, considering that the relatively long-term latency of pancreatic cancer may require additional year of exclusion to evaluate the possibility of this bias. Finally, in the PLCO Cancer Screening Trial, all participants were American adults aged 55–74 years; 85.6% of them were non-Hispanic white, more than half were current or past smokers, and approximately one-third were college students. Thus, our results may not be generalizable to other populations and settings.
Am J Epidemiol. 2021;190(10):2029-2041. © 2021 Oxford University Press