Phylloquinone intake, insulin sensitivity, and glycemic status in adult men and women

Abstract

Background

Limited evidence suggests that vitamin K may have a beneficial role in glucose homeostasis. No observational data exist on the associations between vitamin K intake and insulin sensitivity.

Objective

The aim was to examine associations between vitamin K intake and measures of insulin sensitivity and glycemic status in men and women, aged 26-81 y.

Design

We assessed the cross-sectional associations between self-reported phylloquinone (vitamin K₁) intake and insulin sensitivity, and with glycemic status in the Framingham Offspring Cohort. Dietary and supplemental phylloquinone intakes were assessed by food-frequency questionnaire. Insulin sensitivity was measured by fasting and 2-h post oral glucose tolerance test (OGTT) insulin, the homeostasis model assessment of insulin resistance (HOMA-IR) and the insulin sensitivity index (ISI_0,120). Glycemic status was assessed by fasting and 2-h post OGTT glucose and hemoglobin A_1c (HbA_1c).

Results

Higher phylloquinone intake was associated with greater insulin sensitivity and glycemic status, as measured by 2-h post OGTT insulin and glucose, and ISI_0,120, after adjustment for age, sex, waist circumference, lifestyle characteristics and diet quality (2-h post OGTT insulin: lowest vs. highest quintile, 81.0 vs. 72.7 μU/mL, P for trend = 0.003; ISI_0,120: 26.3 vs. 27.3 mg·L²/mmol·mU·min, P for trend = 0.01; 2-h post OGTT glucose: 106.3 vs. 101.9 mg/dL, P for trend = 0.01). Phylloquinone intake was not associated with fasting insulin and glucose concentrations, HOMA-IR and HbA_1c.

Conclusions

Our findings support a potential beneficial role for phylloquinone in glucose homeostasis in adult men and women.

INTRODUCTION

Recent evidence suggests a potential beneficial role of vitamin K in glucose homeostasis (1-3). The major form of vitamin K in the North American diet is phylloquinone (vitamin K₁), which is concentrated in green vegetables and certain plant oils (4). Rats fed a low phylloquinone diet had higher glucose and delayed acute insulin response to intravenous glucose infusion compared with those fed a high phylloquinone diet (1). In clinical studies of young Japanese healthy men, short-term supplementation of vitamin K improved acute insulin response and glucose disposal in peripheral tissues among those with low baseline vitamin K status (2, 3). Although these findings are suggestive, available data on the role of vitamin K in glucose metabolism are limited, and the potential mechanism behind the association between vitamin K and glucose homeostasis is uncertain. Vitamin K is a cofactor specific to the formation of γ-carboxyglutamyl residues in certain proteins including the coagulation factors and the bone formation protein, osteocalcin. Based on data from the osteocalcin knockout mouse model, osteocalcin may be involved in insulin sensitivity and insulin secretion (5). The role of the vitamin K-dependent carboxylation of osteocalcin in glucose homeostasis is not known.

Currently, there are no published data on the associations between phylloquinone intake and measures of insulin sensitivity and glycemic status in a community-based sample. We examined the cross-sectional associations between phylloquinone intake and both insulin sensitivity and glycemic status, as measured by fasting and 2-h post oral glucose tolerance test (OGTT) insulin and glucose concentrations, hemoglobin A_1c (HbA_1c), the homeostasis model assessment of insulin resistance (HOMA-IR) and the insulin sensitivity index (ISI_0,120) in men and women who participated in the Framingham Offspring Study. We hypothesized that higher phylloquinone intake is associated with greater insulin sensitivity and improved glycemic status in adult men and women.

SUBJECTS AND METHODS

Subject

The Framingham Offspring Study is a longitudinal, community-based study of cardiovascular disease in the children of the participants in the original Framingham Heart Study cohort and their spouses (6). In 1971, 5135 participants were enrolled into the study (7). During examination cycle 5 (1991-1995), 3799 participants underwent an extensive examination, including comprehensive questionnaires, anthropometric measures, blood chemistries, and a physical examination with assessment of cardiovascular and other risk factors by trained clinical personnel. Of 3799 participants, 1040 individuals were excluded from the current analyses for following reasons: invalid dietary information (n = 381), missing data on fasting insulin or glucose measures (n = 143), presence of diabetes (n = 324), use of anticoagulant including the vitamin K antagonist, warfarin (n = 21), missing information on major covariates (n = 221). Our final sample size was 2719 (1247 men and 1472 women). The Institutional Review Boards for Human Research at Boston University and the Tufts-New England Medical Center approved this study.

Phylloquinone intake assessment

Usual dietary intakes for previous 12 months were assessed by a semiquantitative food-frequency questionnaire (FFQ), as described elsewhere (8). This FFQ was validated for various nutrients and foods (8, 9), including phylloquinone (10, 11). A significant correlation was observed between phylloquinone intake calculated from the FFQ and that from three 4-d diet record (r = 0.53) (10). A biomarker-based validation study has shown plasma phylloquinone levels increased approximately two-fold in a linear fashion across FFQ assessed phylloquinone intakes between 50 and 250 μg/d (11). The questionnaire was mailed to participants before examination, and the participants were asked to bring the completed questionnaire with them to their appointment. The FFQ consisted of a list of foods with a standard serving size, and a selection of nine frequency categories ranging from never or < 1 serving/mo to > 6 servings/d. Phylloquinone intake was calculated by multiplying the frequency of consumption of each unit of food from the FFQ by the phylloquinone contents of the specific portion. Separate questions about use of vitamin and mineral supplements were also included in the FFQ. Phylloquinone intakes reported here included intakes from both diets and supplements. Data from the FFQ was judged as reliable if reported total energy intakes were ≥ 600 kcal/d (2.51 MJ/d) for men and women, but < 4200 kcal/d (17.54 MJ/d) for men or < 4000 kcal/d (16.74 MJ/d) for women, and if fewer than 13 items were left blank.

Insulin sensitivity measures

Fasting blood samples (≥ 8 hr) were collected at each examination cycle. Plasma and serum samples were stored at −70 °C. Plasma insulin concentrations were determined using the Coat-A-Count ¹²⁵I-radioimmunoassay (Diagnostic Products, Los Angeles, CA). This assay has cross-reactivity with proinsulin at the midcurve of 40%, intra- and inter-assay CVs of 5 to 10%, and a lower limit of sensitivity was 1.1 μU/mL (7.9 pmol/L). Plasma glucose concentrations were measured with a hexokinase regent kit (A-gent glucose test, Abbott Laboratories, Inc. South Pasadena, CA). Glucose assays were performed in duplicate and the CVs for this assay were < 3%. The 75-g OGTT was administered and 2-h post OGTT plasma insulin and glucose concentrations were measured. HbA_1c was measured by HPLC after an overnight dialysis against normal saline to remove the labile fraction. The inter- and intra-assay CVs were < 2.5%. The assay was standardized against the glycosylated hemoglobin assay used in the Diabetes Control and Complication Trial (12).

We calculated two indices of insulin sensitivity, HOMA-IR (13) and ISI_0,120 (14). HOMA-IR is a surrogate measure of insulin sensitivity at basal state, and tends to represent hepatic insulin sensitivity whereas ISI_0,120 reflects peripheral insulin resistance, glucose disposal and β-cell response to an energy load (15). The HOMA-IR was calculated with the following formula (13):

$$\text{HOMA-IR}=\left[{\text{fasting plasma insulin}}^{\ast }\text{fasting plasma glucose}\right]∕22.5.$$

The ISI_0,120 was calculated using the following formula (14):

$${\mathrm{ISI}}_{0,120}=(\mathrm{m}∕\mathrm{MPG})∕\log \mathrm{MSI}$$

where

$$\begin{array}{cc}\hfill \mathrm{m}=& [75,000\mathrm{mg}+(\text{fasting plasma glucose}-2\text{-h post OGTT plasma glucose})\times 0.19\times \text{body weight}\left(\mathrm{kg}\right)]∕120\min \hfill \\ \hfill \mathrm{MPG}=& [\text{fasting plasma glucose}(\mathrm{mg}∕\mathrm{dL})+2\text{-h post OGTT plasma glucose}(\mathrm{mg}∕\mathrm{dL})]∕2\hfill \\ \hfill \mathrm{MSI}=& [\text{fasting serum insulin}(\mu \mathrm{U}∕\mathrm{L})+2\text{-h post OGTT insulin}(\mu \mathrm{U}∕\mathrm{L})]∕2.\hfill \end{array}$$

Covariate information

Covariates used to examine associations between phylloquinone intake and insulin sensitivity included: age; sex; waist circumference (cm); alcohol intake (g/d); cigarette smoking status (smoked regularly in the past year: yes or no); multivitamin supplementation use (current use at the time of the examination: yes or no); estrogen use (reported current use of oral conjugated estrogen: yes or no); and physical activity. Physical activity was assessed by computing physical activity score based on a validated questionnaire of self-reported 24-hr history of activity (16). Total energy intake (kcal/d) was assessed by the semiquantitative FFQ (8). Overall diet quality was measured by using the 2005 Dietary Guidelines for Americans Adherence Index (DGAI), as described elsewhere (17). The DGAI score was calculated based on age- and sex-specific energy requirement with a range between 0 and 20. Higher DGAI score indicates greater compliance with the 2005 Dietary Guidelines for Americans.

Statistical analysis

SAS statistical software (version 9; SAS institute, Cary, NC) was used for all statistical analyses. Statistical significance was defined as a P-value < 0.05. Normality of insulin sensitivity and glycemic status measures was tested. Since insulin concentrations and HOMA-IR were skewed to the right, we analyzed these variables with the natural logarithm transformation. Phylloquinone intake was categorized based on quintiles of participants’ intake levels. HbA_1c was categorized into two groups (HbA_1c < 6.5% and HbA_1c ≥ 6.5%) to capture individuals with long-term hyperglycemia.

To describe subject characteristics across phylloquinone intake, analysis of covariance (ANCOVA) was performed. Age and sex adjusted means or percentages (95% CI) were presented.

To assess the association between phylloquinone intake and insulin sensitivity and glycemic status measures, we applied ANCOVA and logistic regression for continuous and dichotomous markers of insulin sensitivity and glycemic status, respectively. Phylloquinone intake is a potential surrogate marker for healthy lifestyle and dietary pattern (18), which may relate to greater insulin sensitivity and improved glycemic status. As a result, lifestyle and diet quality potentially confound the association. Therefore, lifestyle characteristics were adjusted in model 1, and both lifestyle and dietary factors were controlled in model 2. Model 1 included age, sex, waist circumference, physical activity, smoking, alcohol consumption, estrogen use, and multivitamin supplementation use. Model 2 adjusted for total energy intake and diet quality measured by the DGAI in addition to covariates used in model 1. In all models, tests for trend by quintile category of phylloquinone intake were performed by assigning median values of phylloquinone intake for each quintile category and treating them as continuous variables. We presented least squares means (95% CIs) and odds ratios (95% CIs) for results from ANCOVA and logistic regression analyses, respectively. We tested each association for interaction with age and sex; however, none of them were significant.

RESULTS

The final sample included 2719 participants (1247 men and 1472 women) with a mean age of 54.0 ± 9.7 (mean ± SD) (range 26-81 y). Phylloquinone intake ranged from 10 to 1975 μg/d (Table 1). Participants in the highest phylloquinone intake quintile category were more likely to be women than men, to use multivitamin supplements and were less likely to be current smokers. Phylloquinone intake was positively associated with physical activity, alcohol consumption, total energy intake, the DGAI score, and among women, estrogen use. Participants in the highest quintile category had lower waist circumference. However, there was no association between phylloquinone intake and BMI.

Table 1.

Characteristics of men and women in the Framingham Offspring Cohort across quintiles of phylloquinone intake (N = 2719)

	Phylloquinone intake quintile category
	Q1	Q2	Q3	Q4	Q5	P for trend1
N	543	544	544	544	544
Phylloquinone intake median (μg/d)	63	103	139	185	282
Phylloquinone intake range (μg/d)	10 - 83	84 -120	120 -160	160 - 218	218-1975
Characteristics2
Age (y)	53.3 (52.5, 54.2)	54.2 (53.3, 55.0)	53.8 (53.0, 54.6)	54.6 (53.8, 55.4)	54 (53.2, 54.8)	0.32
Female (%)	42 (38, 47)	49 (45, 53)	52 (48, 56)	61 (57, 66)	66 (62, 70)	< 0.001
Waist circumference (cm)	92.7 (91.6, 93.7)	92.4 (91.3, 93.4)	92.2 (91.2, 93.2)	92.4 (91.4, 93.5)	90.9 (89.8, 91.9)	0.02
BMI (kg/m2)	27.3 (26.9, 27.6)	27.1 (26.8, 27.5)	27.1 (26.7, 27.5)	27.2 (26.8, 27.6)	26.9 (26.5, 27.3)	0.19
Physical activity score (MET)	34.6 (34.1, 35.1)	34.5 (34.0, 35.0)	35.0 (34.5, 35.5)	35.2 (34.7, 35.7)	35.2 (34.7, 35.7)	0.04
Current smokers (%)	27 (24, 31)	21 (18, 24)	17 (13, 20)	14 (11, 18)	16 (13, 20)	< 0.001
Alcohol consumption (g/d)	10.5 (9.1, 11.8)	10.2 (8.8, 11.5)	11.1 (9.8, 12.4)	12.3 (10.9, 13.6)	12.1 (10.7, 13.4)	0.02
Estrogen use (% in females)	15 (10, 20)	13 (9, 18)	15 (10, 19)	20 (16, 24)	20 (16, 24)	0.01
Multivitamin use (%)	22 (18, 25)	22 (18, 26)	27 (23, 31)	32 (29, 36)	34 (30, 38)	< 0.001
Total energy (kcal/d)	1498 (1451, 1545)	1750 (1703, 1797)	1891 (1845, 1938)	2029 (1982, 2076)	2205 (2157, 2252)	< 0.001
Dietary guideline adherence index score	7.2 (7.0, 7.4)	8.1 (7.9, 8.3)	9.2 (9.0, 9.3)	10 (9.8, 10.2)	11.2 (11.0, 11.4)	< 0.001

¹Analysis of covariance was used to examine the participant characteristics across phylloquinone intake quintile categories. P for trend by quintile category of phylloquinone intakes were performed by assigning median values of phylloquinone intake for each quintile category and treating them as continuous variables.

²Values are least squares mean or percentage (95% CI). Least squares means and percentages are adjusted for sex and age. Age is adjusted for sex only, and sex and estrogen use are adjusted for age only. Since dietary guideline adherence index score was calculated based on age and sex specific criteria, the P for trend is unadjusted for age and sex.

A higher phylloquinone intake was associated with indices of insulin sensitivity after adjustment for age, sex, waist circumference and lifestyle characteristics (Table 2). The associations between phylloquinone intake and fasting insulin concentrations, and with HOMA-IR became non-significant after adjustment for total energy and diet quality, as assessed by the DGAI. However, these associations remained significant when diet quality was adjusted by a healthy-choice subscore, a sub-component of DGAI which does not include dark green vegetable intake component (fasting insulin: P for trend = 0.01; HOMA-IR: P for trend = 0.01). In contrast, the associations between phylloquinone intake and 2-h post OGTT insulin, and with ISI_0,120 remained significant with additional adjustment for total energy intake and DGAI. A higher phylloquinone intake was associated with lower 2-h post OGTT glucose in the fully adjusted model. Further adjustment for other dietary factors, which include total fiber, saturated fatty acid, n-3 fatty acids (EPA and DHA), or potassium, did not affect associations between phylloquinone and 2-h post OGTT insulin and glucose, or ISI_0,120 (data are not shown). There was no association between phylloquinone intake and fasting glucose concentrations or HbA_1c. Energy-adjusted phylloquinone intakes based on regression residuals provided same results.

Table 2.

Insulin sensitivity and glycemic status measures across phylloquinone intake quintiles (N = 2719)

	Phylloquinone intake quintile category
	Q1	Q2	Q3	Q4	Q5	P for trend1
Median (μg/d)	63	103	139	185	282
Range (μg/d)	10 - 83	84 -120	120 -160	160 - 218	218 -1975
N	543	544	544	544	544
Insulin sensitivity measures2
Fasting insulin (μU/mL)
Model 13	29.2 (28.5, 29.8)	29.1 (28.4, 29.7)	28.3 (27.7, 28.9)	28.4 (27.8, 29.0)	27.9 (27.3, 28.5)	0.003
Model 24	29.0 (28.2, 29.7)	28.9 (28.3, 29.6)	28.3 (27.7, 28.9)	28.4 (27.8, 29.1)	28.1 (27.4, 28.8)	0.14
2-h OGTT insulin (μU/mL)5
Model 1	80.4 (76.8, 84.2)	81.1 (77.4, 84.8)	80.4 (76.8, 84.1)	78.6 (75.0, 82.3)	73.1 (69.8, 76.6)	0.001
Model 2	81.0 (76.8, 85.3)	81.0 (77.3, 84.9)	80.3 (76.7, 84.0)	78.4 (74.8, 82.2)	72.7 (68.9, 76.6)	0.003
HOMA-IR
Model 1	6.51 (6.35, 6.68)	6.53 (6.37, 6.70)	6.33 (6.17, 6.49)	6.37 (6.21, 6.53)	6.23 (6.07, 6.39)	0.007
Model 2	6.46 (6.27, 6.64)	6.48 (6.31, 6.65)	6.33 (6.17, 6.49)	6.38 (6.22, 6.55)	6.29 (6.11, 6.47)	0.22
ISI0,1205 (mg·L2/mmol·mU·min)
Model 1	26.5 (26.0, 27.0)	26.1 (25.6, 26.6)	26.1 (25.6, 26.7)	26.5 (26.0, 27.1)	27.1 (26.6, 27.7)	0.03
Model 2	26.3 (25.7, 26.9)	26.1 (25.5, 26.6)	26.1 (25.6, 26.7)	26.6 (26.0, 27.2)	27.3 (26.7, 27.9)	0.009
Glycemic status measures2
Fasting glucose (mg/dL)
Model 1	94.3 (93.6, 95.1)	95.0 (94.2, 95.7)	94.7 (94.0, 95.4)	94.8 (94.1, 95.6)	94.4 (93.6, 95.1)	0.77
Model 2	94.2 (93.4, 95.0)	94.8 (94.0, 95.5)	94.7 (94.0, 95.4)	94.9 (94.2, 95.6)	94.6 (93.8, 95.4)	0.76
2-h OGTT glucose (mg/dL) 5
Model 1	105.3 (103.0, 107.5)	106.5 (104.3, 108.7)	105.9 (103.7, 108.1)	105.0 (102.7, 107.3)	102.9 (100.6, 105.2)	0.06
Model 2	106.3 (103.8, 108.9)	106.9 (104.5, 109.2)	105.8 (103.6, 108.1)	104.6 (102.3, 106.9)	101.9 (99.3, 104.5)	0.009
HbA1c (≥ 6.5 % of total hemoglobin) 6
Model 1	1.00	0.66 (0.30, 1.44)	0.81 (0.38, 1.75)	0.91 (0.43, 1.90)	0.39 (0.15, 1.00)	0.11
Model 2	1.00	0.71 (0.32, 1.58)	0.91 (0.40, 2.09)	1.07 (0.45, 2.53)	0.50 (0.15, 1.55)	0.37

¹Analysis of covariance was performed to examine the association between phylloquinone intake quintile categories and measures of insulin sensitivity and glycemic status. P value is based on linear regression β coefficient for phylloquinone intake quintile as a continuous variable. For HbA_1c, logistic regression was performed to assess the association between phylloquinone intake quintiles and HbA_1c (≥ 6.5 % HbA_1c). P value is based on the logistic regression maximum likelihood estimate for phylloquinone intake quintile as a continuous variable.

²Values of insulin and HOMA-IR are adjusted geometric means (95% CI). Values of glucose and ISI_0,120 are least squares means (95% CI). Values of HbA_1c are odds ratio (95% CI) for prevalence of HbA_1c ≥ 6.5% of total hemoglobin.

³Model 1 adjusted for age, sex, waist circumference, physical activity score, smoking, alcohol consumption, multivitamin use, and estrogen use.

⁴Model 2 adjusted for total energy intake, the DGAI score and all covariates in model 1.

⁵Due to data availability 2655 and 1992 subjects were included in the analyses for the association between phylloquinone intake and 2-h post OGTT insulin and glucose and HbA_1c, respectively. For 2-h post OGTT insulin and glucose, and ISI_0,120, the number of participants in each quintile is 534, 535, 532, 528, and 526 for Q1, Q2, Q3, Q4, and Q5, respectively. For HbA_1c, sample size for each quintile is 391, 389, 403, 391, and 418, for Q1, Q2, Q3, Q4, and Q5, respectively.

DISCUSSION

The major finding of the present study was that higher phylloquinone intake was associated with greater insulin sensitivity, as measured by 2-h post OGTT insulin and ISI_0,120, and better glycemic status by 2-h post OGTT glucose concentrations in a community-based sample of men and women. These observations are consistent with a previous small metabolic study, in which young men with lower vitamin K status, as assessed by biochemical markers, had higher 2-h post OGTT insulin concentrations compared with those with greater vitamin K status (3). Accordingly, it has been proposed that vitamin K may have a potential biological role in glucose homeostasis.

Although phylloquinone intake was significantly associated with insulin sensitivity and glycemic status, as assessed from 2-h OGTT measurements in the present study, we did not find significant associations between phylloquinone intake and insulin sensitivity and glycemic status measures assessed in the fasting state. Currently we do not have an explanation for observed disparity in baseline and 2-h measures. However, our results are consistent with findings from a previous report from a small metabolic study of young men (2). Men with lower reported phylloquinone intake had lower insulin and higher glucose concentrations at 30 min after oral glucose loading, compared with those with higher phylloquinone intake, but there was no association between phylloquinone intake and either fasting glucose or insulin concentrations (2). Potentially, a delayed early response of insulin release by the β-cells to oral glucose loading may explain observed elevation of 2-h post OGTT insulin and glucose concentrations in individuals with lower reported phylloquinone intake in our study. However, our interpretation of data is limited since we did not assess the effect of phylloquinone intake on the acute insulin and glucose responses to oral glucose loading in the present study.

The non-significant association between phylloquinone intake and fasting insulin and HOMA-IR is potentially due to over-adjustment in statistical models. Consumption of dark green vegetables, which are major sources of phylloquinone in this population, is part of the DGAI. When a sub-score of the DGAI that did not include the dark green vegetable intake component was used instead of the overall DGAI in our statistical models, higher phylloquinone intake was associated with fasting insulin and HOMA-IR.

We also observed that higher phylloquinone intake was associated with higher ISI_0,120, which indicates greater insulin sensitivity. ISI_0,120 may capture more of the complexity of insulin resistance and glucose homeostasis than either HOMA-IR or the individual measures of insulin and glucose concentrations. The ISI_0,120 incorporates body weight, and fasting and 2-h post OGTT insulin and glucose concentrations. It is a complex assessment of insulin sensitivity that accounts for β-cell response to glucose loading, peripheral and hepatic insulin sensitivity, and glucose disposal (15). A previous study has shown a high correlation between ISI_0,120 and insulin sensitivity, as measured by the hyperinsulinemic-euglycemic clamp technique (14). In the absence of appropriate measures, the current study cannot determine if phylloquinone intake is associated with β-cell response, insulin sensitivity, glucose disposal, or all. However, all these components are involved in glucose homeostasis and their dysfunction contributes to diabetes (19, 20).

The potential biological mechanisms relating phylloquinone to insulin resistance and glucose homeostasis are not understood. Two forms of vitamin K, phylloquinone and menaquinone-4, are found in the pancreas (21). However, vitamin K-dependent proteins specific to the pancreas have not been identified. A recent study proposed osteocalcin, one of the vitamin K dependent proteins in the bone, may improve insulin sensitivity and increase β-cell functions partially through the enhancement of adiponectin expression (5). Alternatively, it has been suggested that vitamin K has potential physiological functions in addition to its classic role as a cofactor for γ-carboxylation (22). Recent in vivo, in vitro and observational studies have shown that vitamin K decreases inflammation induced cytokines (23-25), so it is plausible that phylloquinone may improve insulin sensitivity and glycemic status by the suppression of inflammation.

There are several limitations in the present study. First, higher phylloquinone intake is a potential surrogate marker for a healthy dietary pattern, as characterized by higher intakes of fruit, vegetables, fish, dietary fiber, and lower intakes of saturated fat (18). Furthermore, since green leafy vegetables are also rich in other components (e.g. dietary fiber and potassium) which have been reported to improve insulin sensitivity, phylloquinone intake may be tracking components in green leafy vegetables that may be beneficial to insulin sensitivity and glycemic status. Although we cannot rule out the presence of residual confounding by overall lifestyle characteristics and diet quality, which may lead to the overestimation of the associations between phylloquinone intake and measures of insulin sensitivity and glycemic status, our finding does not support the hypothesis that these associations are due solely to diet quality. Additional adjustment for diet quality did not alter the significant associations between phylloquinone intake and measures of insulin sensitivity and glycemic status, as measured by 2-h post OGTT insulin and glucose, and ISI_0,120. Therefore, observed associations between phylloquinone intake and insulin sensitivity may indicate a biological role of phylloquinone in glucose homeostasis. A second limitation of this study is its cross-sectional nature, which limits any causal inference from our observations. Finally, most participants in the Framingham Offspring Study have a Northern European ancestry. However, our findings are consistent with those in Japanese young adults (2, 3). Thus, generalizability to other populations may not present a major limitation.

In summary, our findings suggest phylloquinone intake may have beneficial effect on glucose homeostasis, or may serve as a surrogate marker of other dietary or lifestyle factors that were not controlled in our analysis. Our findings may lead to further areas of research that may help to elucidate potential novel functions of vitamin K. Future studies should focus on prospective relationships between phylloquinone intakes and surrogate markers for insulin sensitivity, glycemic status, and type 2 diabetes, as well as the putative biological mechanism to explain associations between phylloquinone and glucose homeostasis.