- Email: [email protected]
Vitamin k and maintenance of skeletal integrity in adults
Vitamin K and Maintenance of Skeletal Integrity in Adults HAROLD N. ROSEN, M.D., I-AURI A. MAITLAND, B.A., Boston, Massachusetts, JOHN W. SUITIE, Ph.D., Madison, Wisconsin, WARREN J. MANNING, M.D., ROBERT J. GLYNN, Ph.D., SC.D., SUSAN L. GREENSPAN, M.D., Boston, Massachusetts
PURPOSE To determine the role of vitamin K status in the maintenance of skeletal integrity in adults. PATIENTS AND MFXHODSz 1. Bone mineral density (BMD) was measured by quantitative digital radiography (QDR) in 50 patients taking a vitamin K antagonist (warfarin) who were recruited from a large urban cardiology practice, and 50 age-, sex- and race-matched controls recruited from the community. 2. The relationship of BMD versus indices of vitamin K status (determined by measuring levels of vitamin K and descarboxyprothrombin in the plasma) in 113 nonanticoagulated adults was asses~L BESULTS: Measurements of BMD in the hip and spine were similar in anticoagulated subjects and matched controls. Multivariate analysis revealed that use of warfarin was not sssociated with a lower BMD. Ninety-five percent confidence intervals excluded a 0.06 g/cm2 reduction in BMD associated with the use of warfarm. Indices of vitamin K status did not correlate with BMD in normal subjects. CONCLUSIONS: Patients receiving long-term maintenance therapy with a vitamin K antagonist have normal bone density. BMD is unrelated to vitamin K status in nonanticoagulated subjects. These data suggest that vitamin K does not have a major role in maintenance of skeletal integrity in adults.
From the Charles A. Dana Research Institute and the Harvard-Thorndike Laboratory of the Beth Israel Hospital. Department of Medicine (Divisions of Gerontology [HNR, LAM, WJM, SLG]. Endocrinology [HNR. SLG]. and Cardiology [WJM]). Beth Israel Hospital. Boston, Massachusetts, the Department of Medicine (RJG), Brigham and Women’s Hospital, Boston, Massachusetts, and the Department of Biochemistry (JWS), University of Wisconsin-Madison,Madison, Wisconsin. This work was supported by a grant from the National Institute on Aging (No. AG08812). by the Public Health Service’s Bureau of Health Professions Faculty Training Project in Geriatric Medicine (Grant No. D31 PE91000). by Beth Israel Hospital Clinical Research Center Grant No. MO1RRO1032, and by a grant from the Eisai Corporation, Tokyo. Japan. Portions of the data presented here were presented at the 73rd Annual Meeting of the Endocrine Society, Washington, D.C.. June 19-22. 1991. Requests for reprints should be addressed to Harold N. Rosen, M.D., Room SL-435. Divisionof Gerontology, Beth Israel Hospital, 330 Brookline Avenue, Boston, Massachusetts 02215. Manuscript submitted February 25, 1992, and accepted in revised form August 11.1992.
62
January 1993
The American Journal of Medicine
ormal bone growth and development depend on adequate vitamin K effect [l-18]. Vitamin K is necessary for posttranslational gamma-carboxylation of selected glutamic acid residues of osteocalcin, the major noncollagenous bone protein [ 1,2]. While properly carboxylated osteocalcin is able to bind hydroxyapatite, des-carboxyosteocalcin is unable to [1,2,5,6,10]. There are several observations suggesting that lack of vitamin K impairs skeletal integrity. Fetuses with in utero exposure to warfarin, a vitamin K antagonist, are born with skeletal malformations [ll-131. Fractures have been reported to heal more slowly in animals treated with warfarin than in untreated animals [14,15]. Studies in humans show that supplemental vitamin K in postmenopausal women decreases bone turnover [16]. Finally, vitamin K levels in patients with fractures have been reported to be lower than those of matched controls [17,18]. We hypothesized that if vitamin K is indeed necessary for maintenance of skeletal integrity in adults, bone mineral density (BMD) should be lower in patients treated with vitamin K antagonists than in matched controls. Furthermore, BMD should correlate with indices of vitamin K status in normal subjects. We examined the role of vitamin K in the adult skeleton by testing these hypotheses.
N
METHODS FOR PART 1 Subjects Fifty patients (age 41 to 85 years, 20 men and 30 women) who had received maintenance warfarin therapy for more than 1 year were recruited. Eighteen patients took warfarin because they had a prosthetic heart valve, 18 for chronic atrial fibrillation, 8 for valvular heart disease without valve replacement, 5 for a history of embolic stroke, and 1 for recurrent venous thromboembolic disease. For comparison, we recruited 50 controls matched by age, sex, and race. Subjects who had any condition that affects bone density (such as renal or hepatic failure, renal tubular acidosis, malabsorption, Cushing’s syndrome, thyrotoxicosis, or hyperparathyroidism) or who took any medication that adversely affects bone density (such as phenytoin, glucocorticoids, cyclosporine, or heparin) were excluded. The Committee on Clinical Investigations of the Beth Israel Hospital approved our pro-
Volume 94
VITAMIN
tocol. Informed individual. M~rement
consent
was obtained
from each
Variables
We determined the level of physical activity, calcium intake, history of smoking, and use of estrogens, diuretics, and thyroid hormone in each subject. A calcium score was determined for each individual as follows: subjects with an estimated calcium intake under 400 mg/d were given a 0, subjects with a dietary calcium intake over 400 mg/d received a 1, and subjects receiving calcium supplements received a 2. Activity acore was 0 for those reporting no exercise beyond that of activities of daily living, or 1 for those reporting exercise beyond activities of daily living. Thiazide and furosemide scores were 0 if a subject had never been treated with the respective diuretic, or 1 if the subject had ever been treated. Estrogen use was graded as 0 if a woman had never been treated with estrogen replacement therapy, or 1 if she had ever been treated. Thyroid hormone use was graded as 0 if a subject had never been treated with thyroid hormone, or 1 if the subject had ever been treated. Tobacco use was graded as 0 if the subject had never smoked, or 1 if the subject had ever smoked. BMD in g/cm2 was assessed by quantitative digital radiography (QDR) using a Hologic QDR-1000 (Hologic Inc., Waltham, MA). Spine BMD was measured at Ll-L4, and a weighted total spine BMD was determined. Compressed vertebrae, as determined by QDR, were excluded from analysis. Hip BMD was measured at the neck, greater trochanter, and intertrochanteric areas, and a weighted total hip BMD was determined. An automatic internal reference system corrects for any transient variations in the radiograph tube voltage. Published values for the coefficient of variation of replicate measurements of the same subject over the short and long term are less than 1% [19,20] and are similar to those obtained on our machine. The coefficient of variation of replicate measurements of the same subject over the short and long term is constant for subjects of varying bone density and adiposity [19,20]. All measurements were performed by the same technician, on the same machine, and in the same room. Daily quality control was done by measuring the BMD of a spine phantom of known hydroxyapatite. composition; the measurement never varied from the actual density by more than 1%. Body mass index (BMI) was calculated based on the ratio (weight in kg)/(height in meters)2. Analysis Mean, standard deviation (SD), and standard error of the mean (SEM) were calculated for each
K AND SKELETAL INTEGRITV
/ RG6EN El AL
variable for patients and controls. The statistical significance of differences between the patients and control groups were assessed by x2 analysis if the variable was dichotomous, and by paired difference t-test if not. The two groups were pooled for linear regression analysis to determine if use of warfarin, duration of use of warfarin, or any other unmatched variables (BMI, calcium intake score, activity score, smoking score, thyroid hormone score, diuretic score, estrogen score) independently predicted BMD. Linear regression analysis was performed using the PROC REG procedure in the Statistical Analysis System (SAS) statistics program.
METHOOS FOf? PART 2 Subjects One hundred thirteen healthy elderly community-dwelling subjects over the age of 65 (35 men and 78 women) were recruited. Patients with any condition that affects bone density or who had taken any medications that adversely affect bone density were excluded (see Methods for Part l), as were subjects who had taken oral warfarin for more than 1 month. The Committee on Clinical Investigations of the Beth Israel Hospital approved our protocol. Informed consent was obtained from each individual. Measurement Variables BMI, BMD, and scores for the level of physical activity, calcium intake, history of smoking, and use of estrogens, diuretics, and thyroid hormone were calculated as described in Methods Part 1. Serum and plasma were collected from each patient for measurement of vitamin K status at the time of their visits for BMD measurement. Vitamin K status was evaluated by measuring plasma vitamin K (phylloquinone) [21-271 and plasma PIVKA-II [28-351 (Protein Induced by Vitamin K Absence) levels. Plasma vitamin K levels reflect relatively recent vitamin K consumption [21-241, while long-term vitamin K status is reflected by the degree of gamma-carboxylation of proteins that normally are gamma-carboxylated, such as prothrombin. High levels of under-carboxylated prothrombin reflect long-term vitamin K deficiency [22,28-311. An antibody specific to des-carboxyprothrombin is used in an enzyme-linked immunosorbent assay to measure levels of descarboxyprothrombin, also known as PIVKA-II. Plasma vitamin K levels were measured by highpressure liquid chromatography, as previously described [27]. Serum levels of des-carboxyprothrombin (PIVKA-II) were measured by sandwich enzyme immunoassay using a kit obtained from Eisai, Tokyo, Japan [28]. PIVKA-II results were expressed in AU/mL, or arbitrary units/ml; 1 AU/mL
January
1993
The American
Journal
ol Medicine
Volume
94
63
VITAMIN K AND SKELETAL INTEGRITY / ROSEN ET AL
TABLEI
B E
Characteristicsof PatientsReceivingWarfarinVersusControls
cotltrels
Patients
Variable
Mean(SEMI
z z
66.7E.6,;p5i.;; 2;.; Nl:w;
!$i (y) BMI(kg/m*)
Yearsreceivingwarfarin Furosemidescore 0.46(0.07)' Thiazidescore 0.18(0.05) Smokingscore 0.58(0.07) Estrogen score 0.06fO.03) Tqscore 0.12(0.05)' Activity score 0.34(0.07)+ Caintakescore 1.00(0.03)' BMDtspine 0.97(0.03) BMDrneck 0.73(0.02) BMDttrcchanter 0.62(0.02) BMDtintertrochanter 1.01 (0.03) BMDttotalhip 0.85(0.02)
0.12PO05) 0.44(0:07) 0.08(0.041 0.02(0.02) 0.70(0.07) 1.15(0.05) 0.96(0.03) 0.70(0.02) 0.63(0.02)
;:
0.4
s S
0.2
to.05 venuscontrol.
TABLEII EstimatedParmeters*in LinearRegressionModelsPredictingBone MineralDensity(g/cm*) in 50 PatientsTakingWarfarinand in 50 Controls Total Hip
Neck
p
0.0
SPINE
TOTAL
HIP
EC%
TROCH
INTERTFKXH
SITE L
Flgure 1. Mean (* SEM) BMD at the total spine, total hip, neck, trochanter, and intertrochanteric region in 50 patients taking warfarin versus 50 matched controls. All paired values are not statistically significantly different.
me mineral density, g/cm*.
Total
3
0.98 (0.03)
= 0.051 versuscontml.
Spine
0.6
s
0.83(0.02)
V-so)
0.8
0'
II= body mass index; T4 = thyroxine; Ca = calcium.
Variable
3 6 Y
Mean (SEMI
n WARFARIN
1.0
Inter-
Trochanter hochanter
-0.015 to.012 Warfarinuseto.019 to.002 to.013 (0.036) (0.024) (0.024) (0.021) (0.028)
other variables such as age, sex, BMI, calcium intake score, activity score, smoking score, thyroid hormone score, diuretic score, or estrogen score independently predicted BMD. Linear regression analysis was performed using the PROC REG procedure in the SAS statistics program.
RESULTS FOR PART 1
The mean BMDs in patients are compared with those of controls in Table I and Figure 1. There was no significant difference between the mean Femalesex -0.200t -0.2001 -0.150t -0.158t (0.038) (0.025) (0.025) (0.023) BMD of patients and controls at any site. Ninetyfive percent confidence intervals excluded the t0.017t to.007 +0.013* tO.OlOt +0.006$ By$y (0.004) (0.003) 10.003) (0.003) (0.003) mean BMDs in patients, being 0.06 g/cm2 lower (kg/m*) than those in controls. The groups were reanalyzed Smoking -0.116+-0.0511-0.029 -0.039 -0.054 separately by gender. Again, there was no signifi(0.037)(0.024)(0.025)(0.022)(0.029) cant difference between the mean BMDs of paR* 0.305 0.562 0.451 0.429 0.550 tients and controls at any site for either men or afarneter estimates a me Increase m cone aensliy m grcmL attnbutaoke to a vanaMe are p3smve, women. iile parameter estimates of the decrement in bone density attributable to a variable are negative. rinstance, the parameter estimate of total spine for watfarin use of to.019 means that after Table I also examines characteristics that can af,ustment for confounders, use of warfarin was associated with an increase in total spine bone fect bone density in patients and controls. Activity isity of 0.019 g/cm*. Numbers in parentheses are the standard errors of the parameters. tO.O1. and calcium scores were lower for patients than for
64
-0.002 -0.003' -0.004t -0.002 (0.002) (0.001) (0.001) (0.001)
January 1993
The Amwican Journal of Medicine
Volume 94
VITAMINK AND SKELETAL INTEGRVIY/ ROSENET AL
2
.
.
s
.
5
s a
0.3
0.5
0.7
0.9
1.1
0.04
1.3
0.40
TOTAL HIP BMD (GM/CM-SQ)
.
.
0.5.
0.60
0.80
1.00
1.20
TOTAL HIP BMD (GM/CM-SQ)
Flgurs 2. Scatter plot of total hip BMD versus plasma vitamin, K levels (in nmol/L) in 120 controls. A “best-fit” line is drawn according to the method of least squares. Pearson correlation coefficient for the correlation between hip BMD and vitamin K levels was 0.039, indicating that there is no significant correlation, p = 0.67.
BMD in g/cm2 attributable to a variable are positive, while parameter estimates of the decrement in BMD attributable to a variable are negative. For example, the parameter estimate of total spine BMD for warfarin use of +0.019 meant that after adjustment for confounders, warfarin use was associated with an increase in total spine BMD of 0.019 g/cm2. Age, female sex, and history of smoking were all significantly associated with lower hip BMDs, while increased BMI was associated with higher hip BMD. Warfarin use was not associated with a decrement in BMD. Ninety-five percent confidence intervals excluded a decrement in BMD attributable to warfarin use of 0.06 g/cm2. Separating the men from the women did not reveal any decrement in BMD associated with warfarin use (data not shown). Duration of warfarin use greater than 10 years was not associated with a significant decrement in BMD (data not shown). The R2 value in Table II indicates the percent of variability in BMD accounted for by our model. For example, the R2 of 0.562 for total hip BMD means that 56.2% of the variability in BMD was accounted for by our model. Such a high R2 value supports the validity of our model. RESlJLTSFORPART2 The relationship between total hip BMD and vitamin K levels is illustrated in Figure 2, while the relationship between total hip BMD and PIVKA-II levels is illustrated in Figure 3. There was no correlation between vitamin K or PIVKA levels and total
Figure 3. Scatter plot of total hip BMD versus plasma PIVKA levels (in AU/mL; see Methods for explanation) in 120 controls. A “best-fit” line is drawn according to the method of least squares. Pearson correlation coefficient for the correlation between hip BMD and PIVKA levels was -0.053, indicating that there is no significant correlation, p = 0.57.
hip BMD. Similar results were found for the other sites at which BMD was measured (data not shown). Multivariate analysis was performed to see if accounting for other factors such as age, sex, or BMI would reveal an association between vitamin K or PIVKA-II levels and BMD. Only age, sex, and BMI were found to predict BMD, so the model was run again including these variables and vitamin K and PIVKA-II levels. The results are illustrated in Table III. Parameter estimates of the increase in BMD in g/cm2 attributable to a variable are positive, while parameter estimates of the decrement in BMD attributable to a variable are negative. Age and female sex were significantly associated with lower hip BMDs, while increased BMI was associated with higher hip BMD. Neither vitamin K nor PIVKA-II levels were associated with an increment or decrement in BMD, respectively. Separating the men from the women did not reveal any association between BMD and vitamin K or PIVKA-II levels (data not shown). COMMENTS Our study was designed to test the hypothesis that vitamin K deficiency has a deleterious effect on bone [6,14-l&36]. Such a deleterious effect would be important for the many patients who take vitamin K antagonists over a long term [37]. The broader ramifications of such a hypothesis could relate to the problem of osteoporosis in general. If vitamin K deficiency was associated with bone loss, then a
January1693 The Amrrlcan Journalof Medlclne Volume64
65
VITAMIN K AND SKELETAL INTEGRITY / ROSEN ET AL
and 50 controls, produced 95% confidence intervals that excluded even modest effects as small as 0.66 Estimated Parameters* in Linear Regression Analysis Predicting Bone g/cm2 lower BMD in the warfarin group. Mineral Density (g/cm*) in 113 Controls However, in contrast to our findings, Fiore et al [36] recently reported that 56 women taking warfaTotal Total Inter. Spine Hip Neck Trochanter tmchanter Variable rin had mean BMD assessed by SPA that was 5% to 20% lower than that of 61 control women matched -0.003 -0.007’ -0.005+ -0.005+ Age(y) -$%!I (0.002) (0.001) (0.001) (0.001) for age and New York Heart Association class. One possible explanation for these findings may be that -0.232+ -0.236+ -0.163+ -0.190’ Female sex (0.038) (0.025) (0.020) (0.022) Fiore et al [36] measured BMD in the radius, while we measured BMD in the spine and hip. Radial to.006 +0.009+ t0.007+ to.003 +0.012+ Body mass index (0.004) (0.003) (0.002) (0.002) (0.003) BMD may be affected by vitamin K deficiency to a kg/m*) greater extent than is BMD of the spine or hip. The -0.022 -0.012 to.025 -0.034 t 0.003 Vitamin K study by Piro and coworkers [38], which showed no (0.049) (0.024) (0.026) CO.0291 (0.040) reduction in radial bone mass in patients taking -0.046 to.051 to.051 to.056 to.050 Des-carboxywarfarin, may have been handicapped by small (0.078) prothrom(0.095) (0.065) (0.052) (0.056) numbers and lack of suitable controls. If this disbin tinction is true, then our conclusion that warfarin 0.303 0.523 0.462 0.458 0.512 R* use is not associated with a major decrement in arameter estimates of the increase in bone density in s/cm* attributable to a variable are positive1, BMD at the clinically relevent sites of the hip and lile parameter estimates of the decrement in bone density attributable to a variable are negatin I instance, the parameter estimate of total spine for age of -0.003meant that after adjustmerIt vertebrae still holds true. If confounders, each pr of age was associated with a decrease in total spine bone density c ‘<“3&$. Numbers m parentheses are thestandard errors of the parameters Several aspects of our study tend to strengthen our findings. First, bone density was measured concurrently in cases and controls by the same laboratory. Second, subjects had taken warfarin for a proportion of age-associated bone loss could be due mean of 7.7 years, which meant that we were assessto vitamin K deficiency [16].Our study demoning the relevance of a substantial exposure to warfarin. Furthermore, the effect of vitamin K status was strates that patients receiving long-term maintenance warfarin therapy had bone mass of the hip analyzed after accounting for factors known to influence BMD, such as age, BMI, and gender. Finaland spine similar to that of age-, sex- and racely, we examined other important determinants of matched controls. Multivariate analysis, performed to adjust for all confounders, showed that after ac- bone density such as calcium intake, activity, smokcounting for other factors that can influence bone ing history, and use of diuretics, estrogen, and thydensity, use of warfarin or duration of use of warfaroid hormone to ensure that these variables would rin had no deleterious effect on bone density. When not bias our findings. This is especially important men were separated from the women, there was still for activity and smoking, which could have been no decrement in BMD associated with warfarin use. higher in the control group than in patients taking The finding that the usual determinants of BMD warfarin. such as age, sex, and BMI significantly predicted However, several limitations must be considered. BMD supports the validity of our BMD Subjects taking warfarin were recruited from a remeasurements. ferral-based academic cardiology practice, while Our data are consistent with previous studies on controls were derived from a group of healthy volbone mass in patients receiving long-term warfarin unteers. Because the population source was not the therapy that used less sensitive techniques for as- same, it is possible that the patients taking warfarin sessing bone mass. Piro et al [38] reported that 17 were in some way healthier or more health-conpatients who had undergone long-term warfarin scious, or that some other unidentified factor assotherapy had normal bone mass as assessed by metaciated with warfarin use cancelled out a deleterious carpal cortical width and single-photon absorptieffect of warfarin on bone. For example, the higher ometry (SPA) of the mid-shaft of the radius. This BMI found in the warfarin group could have potenstudy, however, was weakened by the fact that the tially confounded a deleterious effect of warfarin on bone densities of the patients were compared with bone, if we had not adjusted for it. published normal ranges. Houvenagel et al [39] Indices of vitamin K status did not correlate with found that mean BMD by dual-photon absorptiomBMD among men or women who were not taking etry of the hip and spine in 12 men receiving long- warfarin. This finding further supports our concluterm warfarin therapy was identical to that of 9 sion that vitamin K status is not a major determimatched controls. Our larger study, with 50 cases nant of adult skeletal health. An alternative interTABLE III
68
January 1993
The American Journal of Medicine
Volume 94
VITAMIN
pretation of our data is that vitamin K is needed for skeletal health, but the vitamin K indices we measured are valid only for determining vitamin K consumption over the recent past and are poor predictors of vitamin K status over the many years it takes to lose bone. However, we found that clinically significant vitamin K deficiency caused by warfarin did not lower BMD, so it is unlikely that mild dietary vitamin K deficiency causes significent bone loss. Our findings must be reconciled with observations that suggest a major role for vitamin K in maintaining skeletal health. Hart et al [17] found that vitamin K levels in patients with hip fractures were subnormal. Their finding implies that low vitamin K levels were associated with low BMD and a tendency to fracture. In fact, the low vitamin K levels in fracture patients may be due to the abrupt decrease in oral intake after the occurrence of the hip fracture or to utilization of the available vitamin K for synthesis of new bone. Vitamin K levels drawn immediately before the fracture would have been more relevant. Data that show a reduction in hydroxyproline excretion among postmenopausal women given vitamin K are difficult to interpret [16]. Vitamin K supplementation could decrease hydroxyproline excretion without reducing bone turnover. Data showing a reduction of histomorphometric indices of bone turnover or in rate of bone loss by bone densitometry would have been more convincing. Finally, while some animal experiments have shown that fractures in vitamin Kdeficient animals heal slowly [ 14,151, this finding is not universal [40]. Einhorn et al [40] produced a standard closed fracture in 15 profoundly vitamin K-deficient rats and in 15 vitamin K-sufficient controls. Six weeks later, the mean callus strength in the vitamin K-deficient rats was the same as in the vitamin K-sufficient controls. In conclusion, bone density is normal in patients receiving long-term therapy with the vitamin K antagonist warfarin and is unrelated to vitamin K status in controls. These data suggest that the extent of vitamin K deficiency commonly found in patients does not impact greatly on the skeletal integrity of the clinically important sites in adults.
We thank On. Paul Axelrod. Stafford Cohen, Robert Dye, and Benet Kolman for their participation in recruiting patients taking warfarin. We also wish to thank Jennifer Cherry, B.S., for performing the assays for vitamin K and PIVKA: lgor Choodnovskiy. B.S., for a&stance with statistical analysis; and Dawn GrMiths, for assistance with preparing the manuscript.
REFERENCES applications.
Clin Orthop
2. Price PA. Role of vitamin-K-dependent
/ ROSEN ET AL
proteins in bone metabolism.
Annu
Rev Nutr 1988; 8: 565-83. 3. Blanchard RA, Furie BC. Kruger SF, WaneckG, noassays of human prothrombin
Jorgensen MJ, Furie B. Immu-
species which correlate with functional coagu-
lant activities. J Lab Clin Med 1983;101:242-55. 4. Esnouf MP. Prowse CV. The gamma-carboxy glutamic acid content of human and bovine prothrombin
following warfarin
treatment.
Biochim Biophys Acta
1977; 490: 471-6. 5. Van Haarfem UM.
Knapen MHJ. Hamufyak K, Vermeer
C. Circulating osteo-
calcin during oral anticoagulant therapy. Thromb Haemost 1988; 60: 79-82. 6. Menon RK, Gill DS, Thomas M. Kernoff PBA. Dandona P. Impaired carboxylation of osteocalcin
in warfarin-treated
patients. J Clin Endocrinol Metab 1987;
64: 59-a. 7. Price PA, Kaneda Y. Vitamin K counteracts not in bone. Thromb
the effect of warfarin in liver but
Res 1987; 46: 121-31.
8. Poser JW, Price PA. A method for decarboxyfation
of gamma-carboxygluta-
mic acid in proteins: properties of the decarboxylated
gammacarboxyglutamic
acid protein from calf bone. J Biol Chem 1979; 254: 431-6. 9. Poser JW. Esch FS. Ling NC, Price PA. Isolation and sequence of the vitamin Kdependent protein from human bone: undercarboxylation of the first glutamic acid residue. J Biol Chem 1980; 255: 8685-91. 10. Price PA, Williamson MK, Lothringer JW. Origin of the vitamin K-dependent bone protein found in plasma and its clearance by kidney and bone. J Biol Chem
1981;255:12750-6.
11.Feteih
R, Tassinari MS. Lian JB. Effect of sodium warfarin
dependent
proteins and skeletal development
on vitamin
K-
in the rat fetus. J Bone Miner Res
1990; 5: 885-94. 12. Hall J, Pauli RM. Wilson KM. Maternal and fetal sequelae of anticoagulation during pregnancy. Am J Med 1980; 68: 122-40. 13. Pauli RM, Madden JD. Kranzler KJ, Culpepper W. Port R. Warfarin therapy initiated during pregnancy and phenotypic chondrodysplasia
punctata. J Pediatr
1976; 88: 5D6-8. 14. Dodds RA. Catterall A, Bitensky L. Chayen J. Effects on fracture healing of an antagonist
of the vitamin K cycle. Calcif Tissue Int 1984; 36: 233-B.
15. Stinchfield FE, Sankaran B, Samilson R. The effect of anticoagulant on bone repair. J Bone Joint Surg (Am) 1956; 38: 270-82.
therapy
16. Knapen MHJ, Hamulyak K. Vermeer C. The effect of vitamin K supplementation on circulating osteocalcin (bone Gla protein) and urinary calcium excretion. Ann Intern Med 1989;111:1001-5.
17.Hart JP. Shearer depressed circulating Metab 1985; 60
MJ, Klenerman L, et al. Electrochemical detection of levels of vitamin Kr in osteoporosis, J Clin Endocrinol
1268-9.
18. Bitensky L, Hart JP, Catterall A, Hodges SJ. Pilkington MJ, Chayen J. Circulatingvitamin K levels in patients with fractures. J Bone Joint Surg(Br) 1988; 70: 663-4. 19. Johnston CC, Melton U, Lindsay R. Eddy DM. Clinical indications for bone mass measurements: Osteoporosis
a report from the Scientific Advisory Board of the National
Foundation.
J Bone Miner Res 1989; 4 Suppl 2: l-28.
20. Kelly TL. Slovik DM, Schoenfeld DA, Neer RM. Quantitative digital radiography versus dual photon absorptiometry of the lumbar spine. J Clin Endocrinol Metab 1988; 67: 839-44. 21. Suttie JW, Mummah-Schendel
LL. Shah DV, Lyle BJ, Greger JL. Vitamin K
deficiency from dietary vitamin K restriction in humans. Am J Clin Nutr 1988: 47: 475-80. 22. Allison PM. Mummah-Schendel
LL, Kindberg CG, HarmsCS.
Bang NU. Suttie
JW. Effects of a vitamin K-deficient diet and antibiotics in normal human volunteers. J Lab Clin Med 1987;110:180-8. 23.Shearer MJ. Rahim S. Barkhan P, Stimmler L. Plasma vitamin Kt in mothers and their newborn babies. Lancet 1982; 2: 46D-3.
ACKNOWLEDGMENT
1. Lian JB, Gundberg CM. Osteocalcin:
K AND SKELETAL INTEGRITV
biochemical
1988; 226: 267-90.
considerations
and clinical
24. Usui Y, Tanimura
H. Nishimura N, Kobayashi N, Okanoue T, Ozawa K. Vita-
min K concentrations
in the plasma and liver of surgical patients. Am J Clin Nutr
1990; 51: 846-52. 25. Mummah-Schendel
LL, Suttie BS, Suttie JW. Serum phylloquinone
concen-
trations in a normal adult population. Am J Clin Nutr 1986; 44: 686-9. 26. Conly J. Suttie J, Reid E, Loftson J. Ramotar K. LouieT. Dietary deficiency of phylloquinone and reduced serum levels in febrile neutropenic cancer patients. Am J Clin Nutr 1989; 50: 109-13. 27. Haroon Y. Bacon DS, Sadowski JA. Liquid chromatographic determination of vitamin Kin plasma withfluorometricdetection. Clin Chem 1986; 32: 1925-9. 28. Motohara K. Kuroki Y, Kan H, Endo F, Matsuda I. Detection of vitamin K
January
1993
The American
Journal
of Medlclne
Volume
94
67
VITAMINK ANDSKELETAL INTEDRITT / RDSEN ET Al lnnnunhnt assay for drahting defidencyby-ofanabncemalprothmmbh. PediatrRes 1985; 19: 354-7. 2s.~SD.RussslRM.FukBC,~SF,Jacquesff,FurieB.The plevm of vitamin K decency tn chronic gastrointesthal disxders. Am J Ch Nutr 1935; 41: 639-43. 30. Motcharii K. Tat@ S, Endo F. Kiyota Y. Matsuda 1. Oral suppkmeMatbn of vitsmhKfor~mwmnrand~onbvdsdplavnavitaminKand PtVKA-ll in the neonate. J Pediatr Gastroenterd Nutr 1990; 11: 32-6. 31. Motdwa K, Endo F, Mstsuda I. Screen& for late neonatal vitamin K defickncy by aawbo~otfwcmbin in dried Mood spots. Arch Die Child 1967: 62: 370-5. 32 Sakata Ii. K&etushi H, Fujita K, Yoshii Ii. Effects of aztreonam on fecal fbra ad on vitamin K trhbokm. Antirnicrob &ents Chetnother 1990; 34: 1-7. 3% Liebman tL4, Furie EC, Tor\g MJ, et d. h+wwxboxy (abwmal) prottuom#nasa~ma~rofprimaryhepatocdlularcardnoma.NE~J Mad lw)4; 310: 1427-31.
0
Jmuty 1993 The Amrrkm
Journalof Modlclna Vdumr
34. Furie B, Liebman HA Blanchard RA, Coleman MS, Kruger SF, Fur& EC. Comparison of the native prothrombin anand the prothrombin time for nwnitori~ oral anticoagulant therapy. &cd 1984; 54: -51. 35. Bandwd RA, Furk BC. Jorgemen M, w SF. Furie 8. Acquired vitamin K-dependent cah~xyiatkx~ cbfMncy in liver disease. N Eqi J Med 1961: X5: 2424. 36.F~1re CE, Tamburino C, Foti I?, Grimaldi D. Reduced a&l berm mineral content in patients taldqan ord aMco@ant. South Med J 1990; 63: -2. 37. Hirsch J. Oral snkoagulant drugs. N Engl J Med 1991; 324: 1865-75. 38. Piro LD, W’hyte MP. Murphy WA Burp SJ. Normal cc&al bone rrw in patients after long term axmadin therapy. J Ciin Endoaiool Metab 1982; 54: 47D-3. 39. Howerx@ E, Leldre 0. Vandertindnn T, et&! Taux d’ostkc.akine et masse osseuse ckz les patients recw& des Mivitamines K. Revue du Rhurnatisr~ 1989: 561677-9. 40.EinhomTAGundbergCM.DevlinVJ.WarmanJ.Frscture~andost~ c&in metabolii in vitamin K dhiency. Ckn Ortkg 19% 237: 219-25.
94
PURPOSE To determine the role of vitamin K status in the maintenance of skeletal integrity in adults. PATIENTS AND MFXHODSz 1. Bone mineral density (BMD) was measured by quantitative digital radiography (QDR) in 50 patients taking a vitamin K antagonist (warfarin) who were recruited from a large urban cardiology practice, and 50 age-, sex- and race-matched controls recruited from the community. 2. The relationship of BMD versus indices of vitamin K status (determined by measuring levels of vitamin K and descarboxyprothrombin in the plasma) in 113 nonanticoagulated adults was asses~L BESULTS: Measurements of BMD in the hip and spine were similar in anticoagulated subjects and matched controls. Multivariate analysis revealed that use of warfarin was not sssociated with a lower BMD. Ninety-five percent confidence intervals excluded a 0.06 g/cm2 reduction in BMD associated with the use of warfarm. Indices of vitamin K status did not correlate with BMD in normal subjects. CONCLUSIONS: Patients receiving long-term maintenance therapy with a vitamin K antagonist have normal bone density. BMD is unrelated to vitamin K status in nonanticoagulated subjects. These data suggest that vitamin K does not have a major role in maintenance of skeletal integrity in adults.
From the Charles A. Dana Research Institute and the Harvard-Thorndike Laboratory of the Beth Israel Hospital. Department of Medicine (Divisions of Gerontology [HNR, LAM, WJM, SLG]. Endocrinology [HNR. SLG]. and Cardiology [WJM]). Beth Israel Hospital. Boston, Massachusetts, the Department of Medicine (RJG), Brigham and Women’s Hospital, Boston, Massachusetts, and the Department of Biochemistry (JWS), University of Wisconsin-Madison,Madison, Wisconsin. This work was supported by a grant from the National Institute on Aging (No. AG08812). by the Public Health Service’s Bureau of Health Professions Faculty Training Project in Geriatric Medicine (Grant No. D31 PE91000). by Beth Israel Hospital Clinical Research Center Grant No. MO1RRO1032, and by a grant from the Eisai Corporation, Tokyo. Japan. Portions of the data presented here were presented at the 73rd Annual Meeting of the Endocrine Society, Washington, D.C.. June 19-22. 1991. Requests for reprints should be addressed to Harold N. Rosen, M.D., Room SL-435. Divisionof Gerontology, Beth Israel Hospital, 330 Brookline Avenue, Boston, Massachusetts 02215. Manuscript submitted February 25, 1992, and accepted in revised form August 11.1992.
62
January 1993
The American Journal of Medicine
ormal bone growth and development depend on adequate vitamin K effect [l-18]. Vitamin K is necessary for posttranslational gamma-carboxylation of selected glutamic acid residues of osteocalcin, the major noncollagenous bone protein [ 1,2]. While properly carboxylated osteocalcin is able to bind hydroxyapatite, des-carboxyosteocalcin is unable to [1,2,5,6,10]. There are several observations suggesting that lack of vitamin K impairs skeletal integrity. Fetuses with in utero exposure to warfarin, a vitamin K antagonist, are born with skeletal malformations [ll-131. Fractures have been reported to heal more slowly in animals treated with warfarin than in untreated animals [14,15]. Studies in humans show that supplemental vitamin K in postmenopausal women decreases bone turnover [16]. Finally, vitamin K levels in patients with fractures have been reported to be lower than those of matched controls [17,18]. We hypothesized that if vitamin K is indeed necessary for maintenance of skeletal integrity in adults, bone mineral density (BMD) should be lower in patients treated with vitamin K antagonists than in matched controls. Furthermore, BMD should correlate with indices of vitamin K status in normal subjects. We examined the role of vitamin K in the adult skeleton by testing these hypotheses.
N
METHODS FOR PART 1 Subjects Fifty patients (age 41 to 85 years, 20 men and 30 women) who had received maintenance warfarin therapy for more than 1 year were recruited. Eighteen patients took warfarin because they had a prosthetic heart valve, 18 for chronic atrial fibrillation, 8 for valvular heart disease without valve replacement, 5 for a history of embolic stroke, and 1 for recurrent venous thromboembolic disease. For comparison, we recruited 50 controls matched by age, sex, and race. Subjects who had any condition that affects bone density (such as renal or hepatic failure, renal tubular acidosis, malabsorption, Cushing’s syndrome, thyrotoxicosis, or hyperparathyroidism) or who took any medication that adversely affects bone density (such as phenytoin, glucocorticoids, cyclosporine, or heparin) were excluded. The Committee on Clinical Investigations of the Beth Israel Hospital approved our pro-
Volume 94
VITAMIN
tocol. Informed individual. M~rement
consent
was obtained
from each
Variables
We determined the level of physical activity, calcium intake, history of smoking, and use of estrogens, diuretics, and thyroid hormone in each subject. A calcium score was determined for each individual as follows: subjects with an estimated calcium intake under 400 mg/d were given a 0, subjects with a dietary calcium intake over 400 mg/d received a 1, and subjects receiving calcium supplements received a 2. Activity acore was 0 for those reporting no exercise beyond that of activities of daily living, or 1 for those reporting exercise beyond activities of daily living. Thiazide and furosemide scores were 0 if a subject had never been treated with the respective diuretic, or 1 if the subject had ever been treated. Estrogen use was graded as 0 if a woman had never been treated with estrogen replacement therapy, or 1 if she had ever been treated. Thyroid hormone use was graded as 0 if a subject had never been treated with thyroid hormone, or 1 if the subject had ever been treated. Tobacco use was graded as 0 if the subject had never smoked, or 1 if the subject had ever smoked. BMD in g/cm2 was assessed by quantitative digital radiography (QDR) using a Hologic QDR-1000 (Hologic Inc., Waltham, MA). Spine BMD was measured at Ll-L4, and a weighted total spine BMD was determined. Compressed vertebrae, as determined by QDR, were excluded from analysis. Hip BMD was measured at the neck, greater trochanter, and intertrochanteric areas, and a weighted total hip BMD was determined. An automatic internal reference system corrects for any transient variations in the radiograph tube voltage. Published values for the coefficient of variation of replicate measurements of the same subject over the short and long term are less than 1% [19,20] and are similar to those obtained on our machine. The coefficient of variation of replicate measurements of the same subject over the short and long term is constant for subjects of varying bone density and adiposity [19,20]. All measurements were performed by the same technician, on the same machine, and in the same room. Daily quality control was done by measuring the BMD of a spine phantom of known hydroxyapatite. composition; the measurement never varied from the actual density by more than 1%. Body mass index (BMI) was calculated based on the ratio (weight in kg)/(height in meters)2. Analysis Mean, standard deviation (SD), and standard error of the mean (SEM) were calculated for each
K AND SKELETAL INTEGRITV
/ RG6EN El AL
variable for patients and controls. The statistical significance of differences between the patients and control groups were assessed by x2 analysis if the variable was dichotomous, and by paired difference t-test if not. The two groups were pooled for linear regression analysis to determine if use of warfarin, duration of use of warfarin, or any other unmatched variables (BMI, calcium intake score, activity score, smoking score, thyroid hormone score, diuretic score, estrogen score) independently predicted BMD. Linear regression analysis was performed using the PROC REG procedure in the Statistical Analysis System (SAS) statistics program.
METHOOS FOf? PART 2 Subjects One hundred thirteen healthy elderly community-dwelling subjects over the age of 65 (35 men and 78 women) were recruited. Patients with any condition that affects bone density or who had taken any medications that adversely affect bone density were excluded (see Methods for Part l), as were subjects who had taken oral warfarin for more than 1 month. The Committee on Clinical Investigations of the Beth Israel Hospital approved our protocol. Informed consent was obtained from each individual. Measurement Variables BMI, BMD, and scores for the level of physical activity, calcium intake, history of smoking, and use of estrogens, diuretics, and thyroid hormone were calculated as described in Methods Part 1. Serum and plasma were collected from each patient for measurement of vitamin K status at the time of their visits for BMD measurement. Vitamin K status was evaluated by measuring plasma vitamin K (phylloquinone) [21-271 and plasma PIVKA-II [28-351 (Protein Induced by Vitamin K Absence) levels. Plasma vitamin K levels reflect relatively recent vitamin K consumption [21-241, while long-term vitamin K status is reflected by the degree of gamma-carboxylation of proteins that normally are gamma-carboxylated, such as prothrombin. High levels of under-carboxylated prothrombin reflect long-term vitamin K deficiency [22,28-311. An antibody specific to des-carboxyprothrombin is used in an enzyme-linked immunosorbent assay to measure levels of descarboxyprothrombin, also known as PIVKA-II. Plasma vitamin K levels were measured by highpressure liquid chromatography, as previously described [27]. Serum levels of des-carboxyprothrombin (PIVKA-II) were measured by sandwich enzyme immunoassay using a kit obtained from Eisai, Tokyo, Japan [28]. PIVKA-II results were expressed in AU/mL, or arbitrary units/ml; 1 AU/mL
January
1993
The American
Journal
ol Medicine
Volume
94
63
VITAMIN K AND SKELETAL INTEGRITY / ROSEN ET AL
TABLEI
B E
Characteristicsof PatientsReceivingWarfarinVersusControls
cotltrels
Patients
Variable
Mean(SEMI
z z
66.7E.6,;p5i.;; 2;.; Nl:w;
!$i (y) BMI(kg/m*)
Yearsreceivingwarfarin Furosemidescore 0.46(0.07)' Thiazidescore 0.18(0.05) Smokingscore 0.58(0.07) Estrogen score 0.06fO.03) Tqscore 0.12(0.05)' Activity score 0.34(0.07)+ Caintakescore 1.00(0.03)' BMDtspine 0.97(0.03) BMDrneck 0.73(0.02) BMDttrcchanter 0.62(0.02) BMDtintertrochanter 1.01 (0.03) BMDttotalhip 0.85(0.02)
0.12PO05) 0.44(0:07) 0.08(0.041 0.02(0.02) 0.70(0.07) 1.15(0.05) 0.96(0.03) 0.70(0.02) 0.63(0.02)
;:
0.4
s S
0.2
to.05 venuscontrol.
TABLEII EstimatedParmeters*in LinearRegressionModelsPredictingBone MineralDensity(g/cm*) in 50 PatientsTakingWarfarinand in 50 Controls Total Hip
Neck
p
0.0
SPINE
TOTAL
HIP
EC%
TROCH
INTERTFKXH
SITE L
Flgure 1. Mean (* SEM) BMD at the total spine, total hip, neck, trochanter, and intertrochanteric region in 50 patients taking warfarin versus 50 matched controls. All paired values are not statistically significantly different.
me mineral density, g/cm*.
Total
3
0.98 (0.03)
= 0.051 versuscontml.
Spine
0.6
s
0.83(0.02)
V-so)
0.8
0'
II= body mass index; T4 = thyroxine; Ca = calcium.
Variable
3 6 Y
Mean (SEMI
n WARFARIN
1.0
Inter-
Trochanter hochanter
-0.015 to.012 Warfarinuseto.019 to.002 to.013 (0.036) (0.024) (0.024) (0.021) (0.028)
other variables such as age, sex, BMI, calcium intake score, activity score, smoking score, thyroid hormone score, diuretic score, or estrogen score independently predicted BMD. Linear regression analysis was performed using the PROC REG procedure in the SAS statistics program.
RESULTS FOR PART 1
The mean BMDs in patients are compared with those of controls in Table I and Figure 1. There was no significant difference between the mean Femalesex -0.200t -0.2001 -0.150t -0.158t (0.038) (0.025) (0.025) (0.023) BMD of patients and controls at any site. Ninetyfive percent confidence intervals excluded the t0.017t to.007 +0.013* tO.OlOt +0.006$ By$y (0.004) (0.003) 10.003) (0.003) (0.003) mean BMDs in patients, being 0.06 g/cm2 lower (kg/m*) than those in controls. The groups were reanalyzed Smoking -0.116+-0.0511-0.029 -0.039 -0.054 separately by gender. Again, there was no signifi(0.037)(0.024)(0.025)(0.022)(0.029) cant difference between the mean BMDs of paR* 0.305 0.562 0.451 0.429 0.550 tients and controls at any site for either men or afarneter estimates a me Increase m cone aensliy m grcmL attnbutaoke to a vanaMe are p3smve, women. iile parameter estimates of the decrement in bone density attributable to a variable are negative. rinstance, the parameter estimate of total spine for watfarin use of to.019 means that after Table I also examines characteristics that can af,ustment for confounders, use of warfarin was associated with an increase in total spine bone fect bone density in patients and controls. Activity isity of 0.019 g/cm*. Numbers in parentheses are the standard errors of the parameters. tO.O1. and calcium scores were lower for patients than for
64
-0.002 -0.003' -0.004t -0.002 (0.002) (0.001) (0.001) (0.001)
January 1993
The Amwican Journal of Medicine
Volume 94
VITAMINK AND SKELETAL INTEGRVIY/ ROSENET AL
2
.
.
s
.
5
s a
0.3
0.5
0.7
0.9
1.1
0.04
1.3
0.40
TOTAL HIP BMD (GM/CM-SQ)
.
.
0.5.
0.60
0.80
1.00
1.20
TOTAL HIP BMD (GM/CM-SQ)
Flgurs 2. Scatter plot of total hip BMD versus plasma vitamin, K levels (in nmol/L) in 120 controls. A “best-fit” line is drawn according to the method of least squares. Pearson correlation coefficient for the correlation between hip BMD and vitamin K levels was 0.039, indicating that there is no significant correlation, p = 0.67.
BMD in g/cm2 attributable to a variable are positive, while parameter estimates of the decrement in BMD attributable to a variable are negative. For example, the parameter estimate of total spine BMD for warfarin use of +0.019 meant that after adjustment for confounders, warfarin use was associated with an increase in total spine BMD of 0.019 g/cm2. Age, female sex, and history of smoking were all significantly associated with lower hip BMDs, while increased BMI was associated with higher hip BMD. Warfarin use was not associated with a decrement in BMD. Ninety-five percent confidence intervals excluded a decrement in BMD attributable to warfarin use of 0.06 g/cm2. Separating the men from the women did not reveal any decrement in BMD associated with warfarin use (data not shown). Duration of warfarin use greater than 10 years was not associated with a significant decrement in BMD (data not shown). The R2 value in Table II indicates the percent of variability in BMD accounted for by our model. For example, the R2 of 0.562 for total hip BMD means that 56.2% of the variability in BMD was accounted for by our model. Such a high R2 value supports the validity of our model. RESlJLTSFORPART2 The relationship between total hip BMD and vitamin K levels is illustrated in Figure 2, while the relationship between total hip BMD and PIVKA-II levels is illustrated in Figure 3. There was no correlation between vitamin K or PIVKA levels and total
Figure 3. Scatter plot of total hip BMD versus plasma PIVKA levels (in AU/mL; see Methods for explanation) in 120 controls. A “best-fit” line is drawn according to the method of least squares. Pearson correlation coefficient for the correlation between hip BMD and PIVKA levels was -0.053, indicating that there is no significant correlation, p = 0.57.
hip BMD. Similar results were found for the other sites at which BMD was measured (data not shown). Multivariate analysis was performed to see if accounting for other factors such as age, sex, or BMI would reveal an association between vitamin K or PIVKA-II levels and BMD. Only age, sex, and BMI were found to predict BMD, so the model was run again including these variables and vitamin K and PIVKA-II levels. The results are illustrated in Table III. Parameter estimates of the increase in BMD in g/cm2 attributable to a variable are positive, while parameter estimates of the decrement in BMD attributable to a variable are negative. Age and female sex were significantly associated with lower hip BMDs, while increased BMI was associated with higher hip BMD. Neither vitamin K nor PIVKA-II levels were associated with an increment or decrement in BMD, respectively. Separating the men from the women did not reveal any association between BMD and vitamin K or PIVKA-II levels (data not shown). COMMENTS Our study was designed to test the hypothesis that vitamin K deficiency has a deleterious effect on bone [6,14-l&36]. Such a deleterious effect would be important for the many patients who take vitamin K antagonists over a long term [37]. The broader ramifications of such a hypothesis could relate to the problem of osteoporosis in general. If vitamin K deficiency was associated with bone loss, then a
January1693 The Amrrlcan Journalof Medlclne Volume64
65
VITAMIN K AND SKELETAL INTEGRITY / ROSEN ET AL
and 50 controls, produced 95% confidence intervals that excluded even modest effects as small as 0.66 Estimated Parameters* in Linear Regression Analysis Predicting Bone g/cm2 lower BMD in the warfarin group. Mineral Density (g/cm*) in 113 Controls However, in contrast to our findings, Fiore et al [36] recently reported that 56 women taking warfaTotal Total Inter. Spine Hip Neck Trochanter tmchanter Variable rin had mean BMD assessed by SPA that was 5% to 20% lower than that of 61 control women matched -0.003 -0.007’ -0.005+ -0.005+ Age(y) -$%!I (0.002) (0.001) (0.001) (0.001) for age and New York Heart Association class. One possible explanation for these findings may be that -0.232+ -0.236+ -0.163+ -0.190’ Female sex (0.038) (0.025) (0.020) (0.022) Fiore et al [36] measured BMD in the radius, while we measured BMD in the spine and hip. Radial to.006 +0.009+ t0.007+ to.003 +0.012+ Body mass index (0.004) (0.003) (0.002) (0.002) (0.003) BMD may be affected by vitamin K deficiency to a kg/m*) greater extent than is BMD of the spine or hip. The -0.022 -0.012 to.025 -0.034 t 0.003 Vitamin K study by Piro and coworkers [38], which showed no (0.049) (0.024) (0.026) CO.0291 (0.040) reduction in radial bone mass in patients taking -0.046 to.051 to.051 to.056 to.050 Des-carboxywarfarin, may have been handicapped by small (0.078) prothrom(0.095) (0.065) (0.052) (0.056) numbers and lack of suitable controls. If this disbin tinction is true, then our conclusion that warfarin 0.303 0.523 0.462 0.458 0.512 R* use is not associated with a major decrement in arameter estimates of the increase in bone density in s/cm* attributable to a variable are positive1, BMD at the clinically relevent sites of the hip and lile parameter estimates of the decrement in bone density attributable to a variable are negatin I instance, the parameter estimate of total spine for age of -0.003meant that after adjustmerIt vertebrae still holds true. If confounders, each pr of age was associated with a decrease in total spine bone density c ‘<“3&$. Numbers m parentheses are thestandard errors of the parameters Several aspects of our study tend to strengthen our findings. First, bone density was measured concurrently in cases and controls by the same laboratory. Second, subjects had taken warfarin for a proportion of age-associated bone loss could be due mean of 7.7 years, which meant that we were assessto vitamin K deficiency [16].Our study demoning the relevance of a substantial exposure to warfarin. Furthermore, the effect of vitamin K status was strates that patients receiving long-term maintenance warfarin therapy had bone mass of the hip analyzed after accounting for factors known to influence BMD, such as age, BMI, and gender. Finaland spine similar to that of age-, sex- and racely, we examined other important determinants of matched controls. Multivariate analysis, performed to adjust for all confounders, showed that after ac- bone density such as calcium intake, activity, smokcounting for other factors that can influence bone ing history, and use of diuretics, estrogen, and thydensity, use of warfarin or duration of use of warfaroid hormone to ensure that these variables would rin had no deleterious effect on bone density. When not bias our findings. This is especially important men were separated from the women, there was still for activity and smoking, which could have been no decrement in BMD associated with warfarin use. higher in the control group than in patients taking The finding that the usual determinants of BMD warfarin. such as age, sex, and BMI significantly predicted However, several limitations must be considered. BMD supports the validity of our BMD Subjects taking warfarin were recruited from a remeasurements. ferral-based academic cardiology practice, while Our data are consistent with previous studies on controls were derived from a group of healthy volbone mass in patients receiving long-term warfarin unteers. Because the population source was not the therapy that used less sensitive techniques for as- same, it is possible that the patients taking warfarin sessing bone mass. Piro et al [38] reported that 17 were in some way healthier or more health-conpatients who had undergone long-term warfarin scious, or that some other unidentified factor assotherapy had normal bone mass as assessed by metaciated with warfarin use cancelled out a deleterious carpal cortical width and single-photon absorptieffect of warfarin on bone. For example, the higher ometry (SPA) of the mid-shaft of the radius. This BMI found in the warfarin group could have potenstudy, however, was weakened by the fact that the tially confounded a deleterious effect of warfarin on bone densities of the patients were compared with bone, if we had not adjusted for it. published normal ranges. Houvenagel et al [39] Indices of vitamin K status did not correlate with found that mean BMD by dual-photon absorptiomBMD among men or women who were not taking etry of the hip and spine in 12 men receiving long- warfarin. This finding further supports our concluterm warfarin therapy was identical to that of 9 sion that vitamin K status is not a major determimatched controls. Our larger study, with 50 cases nant of adult skeletal health. An alternative interTABLE III
68
January 1993
The American Journal of Medicine
Volume 94
VITAMIN
pretation of our data is that vitamin K is needed for skeletal health, but the vitamin K indices we measured are valid only for determining vitamin K consumption over the recent past and are poor predictors of vitamin K status over the many years it takes to lose bone. However, we found that clinically significant vitamin K deficiency caused by warfarin did not lower BMD, so it is unlikely that mild dietary vitamin K deficiency causes significent bone loss. Our findings must be reconciled with observations that suggest a major role for vitamin K in maintaining skeletal health. Hart et al [17] found that vitamin K levels in patients with hip fractures were subnormal. Their finding implies that low vitamin K levels were associated with low BMD and a tendency to fracture. In fact, the low vitamin K levels in fracture patients may be due to the abrupt decrease in oral intake after the occurrence of the hip fracture or to utilization of the available vitamin K for synthesis of new bone. Vitamin K levels drawn immediately before the fracture would have been more relevant. Data that show a reduction in hydroxyproline excretion among postmenopausal women given vitamin K are difficult to interpret [16]. Vitamin K supplementation could decrease hydroxyproline excretion without reducing bone turnover. Data showing a reduction of histomorphometric indices of bone turnover or in rate of bone loss by bone densitometry would have been more convincing. Finally, while some animal experiments have shown that fractures in vitamin Kdeficient animals heal slowly [ 14,151, this finding is not universal [40]. Einhorn et al [40] produced a standard closed fracture in 15 profoundly vitamin K-deficient rats and in 15 vitamin K-sufficient controls. Six weeks later, the mean callus strength in the vitamin K-deficient rats was the same as in the vitamin K-sufficient controls. In conclusion, bone density is normal in patients receiving long-term therapy with the vitamin K antagonist warfarin and is unrelated to vitamin K status in controls. These data suggest that the extent of vitamin K deficiency commonly found in patients does not impact greatly on the skeletal integrity of the clinically important sites in adults.
We thank On. Paul Axelrod. Stafford Cohen, Robert Dye, and Benet Kolman for their participation in recruiting patients taking warfarin. We also wish to thank Jennifer Cherry, B.S., for performing the assays for vitamin K and PIVKA: lgor Choodnovskiy. B.S., for a&stance with statistical analysis; and Dawn GrMiths, for assistance with preparing the manuscript.
REFERENCES applications.
Clin Orthop
2. Price PA. Role of vitamin-K-dependent
/ ROSEN ET AL
proteins in bone metabolism.
Annu
Rev Nutr 1988; 8: 565-83. 3. Blanchard RA, Furie BC. Kruger SF, WaneckG, noassays of human prothrombin
Jorgensen MJ, Furie B. Immu-
species which correlate with functional coagu-
lant activities. J Lab Clin Med 1983;101:242-55. 4. Esnouf MP. Prowse CV. The gamma-carboxy glutamic acid content of human and bovine prothrombin
following warfarin
treatment.
Biochim Biophys Acta
1977; 490: 471-6. 5. Van Haarfem UM.
Knapen MHJ. Hamufyak K, Vermeer
C. Circulating osteo-
calcin during oral anticoagulant therapy. Thromb Haemost 1988; 60: 79-82. 6. Menon RK, Gill DS, Thomas M. Kernoff PBA. Dandona P. Impaired carboxylation of osteocalcin
in warfarin-treated
patients. J Clin Endocrinol Metab 1987;
64: 59-a. 7. Price PA, Kaneda Y. Vitamin K counteracts not in bone. Thromb
the effect of warfarin in liver but
Res 1987; 46: 121-31.
8. Poser JW, Price PA. A method for decarboxyfation
of gamma-carboxygluta-
mic acid in proteins: properties of the decarboxylated
gammacarboxyglutamic
acid protein from calf bone. J Biol Chem 1979; 254: 431-6. 9. Poser JW. Esch FS. Ling NC, Price PA. Isolation and sequence of the vitamin Kdependent protein from human bone: undercarboxylation of the first glutamic acid residue. J Biol Chem 1980; 255: 8685-91. 10. Price PA, Williamson MK, Lothringer JW. Origin of the vitamin K-dependent bone protein found in plasma and its clearance by kidney and bone. J Biol Chem
1981;255:12750-6.
11.Feteih
R, Tassinari MS. Lian JB. Effect of sodium warfarin
dependent
proteins and skeletal development
on vitamin
K-
in the rat fetus. J Bone Miner Res
1990; 5: 885-94. 12. Hall J, Pauli RM. Wilson KM. Maternal and fetal sequelae of anticoagulation during pregnancy. Am J Med 1980; 68: 122-40. 13. Pauli RM, Madden JD. Kranzler KJ, Culpepper W. Port R. Warfarin therapy initiated during pregnancy and phenotypic chondrodysplasia
punctata. J Pediatr
1976; 88: 5D6-8. 14. Dodds RA. Catterall A, Bitensky L. Chayen J. Effects on fracture healing of an antagonist
of the vitamin K cycle. Calcif Tissue Int 1984; 36: 233-B.
15. Stinchfield FE, Sankaran B, Samilson R. The effect of anticoagulant on bone repair. J Bone Joint Surg (Am) 1956; 38: 270-82.
therapy
16. Knapen MHJ, Hamulyak K. Vermeer C. The effect of vitamin K supplementation on circulating osteocalcin (bone Gla protein) and urinary calcium excretion. Ann Intern Med 1989;111:1001-5.
17.Hart JP. Shearer depressed circulating Metab 1985; 60
MJ, Klenerman L, et al. Electrochemical detection of levels of vitamin Kr in osteoporosis, J Clin Endocrinol
1268-9.
18. Bitensky L, Hart JP, Catterall A, Hodges SJ. Pilkington MJ, Chayen J. Circulatingvitamin K levels in patients with fractures. J Bone Joint Surg(Br) 1988; 70: 663-4. 19. Johnston CC, Melton U, Lindsay R. Eddy DM. Clinical indications for bone mass measurements: Osteoporosis
a report from the Scientific Advisory Board of the National
Foundation.
J Bone Miner Res 1989; 4 Suppl 2: l-28.
20. Kelly TL. Slovik DM, Schoenfeld DA, Neer RM. Quantitative digital radiography versus dual photon absorptiometry of the lumbar spine. J Clin Endocrinol Metab 1988; 67: 839-44. 21. Suttie JW, Mummah-Schendel
LL. Shah DV, Lyle BJ, Greger JL. Vitamin K
deficiency from dietary vitamin K restriction in humans. Am J Clin Nutr 1988: 47: 475-80. 22. Allison PM. Mummah-Schendel
LL, Kindberg CG, HarmsCS.
Bang NU. Suttie
JW. Effects of a vitamin K-deficient diet and antibiotics in normal human volunteers. J Lab Clin Med 1987;110:180-8. 23.Shearer MJ. Rahim S. Barkhan P, Stimmler L. Plasma vitamin Kt in mothers and their newborn babies. Lancet 1982; 2: 46D-3.
ACKNOWLEDGMENT
1. Lian JB, Gundberg CM. Osteocalcin:
K AND SKELETAL INTEGRITV
biochemical
1988; 226: 267-90.
considerations
and clinical
24. Usui Y, Tanimura
H. Nishimura N, Kobayashi N, Okanoue T, Ozawa K. Vita-
min K concentrations
in the plasma and liver of surgical patients. Am J Clin Nutr
1990; 51: 846-52. 25. Mummah-Schendel
LL, Suttie BS, Suttie JW. Serum phylloquinone
concen-
trations in a normal adult population. Am J Clin Nutr 1986; 44: 686-9. 26. Conly J. Suttie J, Reid E, Loftson J. Ramotar K. LouieT. Dietary deficiency of phylloquinone and reduced serum levels in febrile neutropenic cancer patients. Am J Clin Nutr 1989; 50: 109-13. 27. Haroon Y. Bacon DS, Sadowski JA. Liquid chromatographic determination of vitamin Kin plasma withfluorometricdetection. Clin Chem 1986; 32: 1925-9. 28. Motohara K. Kuroki Y, Kan H, Endo F, Matsuda I. Detection of vitamin K
January
1993
The American
Journal
of Medlclne
Volume
94
67
VITAMINK ANDSKELETAL INTEDRITT / RDSEN ET Al lnnnunhnt assay for drahting defidencyby-ofanabncemalprothmmbh. PediatrRes 1985; 19: 354-7. 2s.~SD.RussslRM.FukBC,~SF,Jacquesff,FurieB.The plevm of vitamin K decency tn chronic gastrointesthal disxders. Am J Ch Nutr 1935; 41: 639-43. 30. Motcharii K. Tat@ S, Endo F. Kiyota Y. Matsuda 1. Oral suppkmeMatbn of vitsmhKfor~mwmnrand~onbvdsdplavnavitaminKand PtVKA-ll in the neonate. J Pediatr Gastroenterd Nutr 1990; 11: 32-6. 31. Motdwa K, Endo F, Mstsuda I. Screen& for late neonatal vitamin K defickncy by aawbo~otfwcmbin in dried Mood spots. Arch Die Child 1967: 62: 370-5. 32 Sakata Ii. K&etushi H, Fujita K, Yoshii Ii. Effects of aztreonam on fecal fbra ad on vitamin K trhbokm. Antirnicrob &ents Chetnother 1990; 34: 1-7. 3% Liebman tL4, Furie EC, Tor\g MJ, et d. h+wwxboxy (abwmal) prottuom#nasa~ma~rofprimaryhepatocdlularcardnoma.NE~J Mad lw)4; 310: 1427-31.
0
Jmuty 1993 The Amrrkm
Journalof Modlclna Vdumr
34. Furie B, Liebman HA Blanchard RA, Coleman MS, Kruger SF, Fur& EC. Comparison of the native prothrombin anand the prothrombin time for nwnitori~ oral anticoagulant therapy. &cd 1984; 54: -51. 35. Bandwd RA, Furk BC. Jorgemen M, w SF. Furie 8. Acquired vitamin K-dependent cah~xyiatkx~ cbfMncy in liver disease. N Eqi J Med 1961: X5: 2424. 36.F~1re CE, Tamburino C, Foti I?, Grimaldi D. Reduced a&l berm mineral content in patients taldqan ord aMco@ant. South Med J 1990; 63: -2. 37. Hirsch J. Oral snkoagulant drugs. N Engl J Med 1991; 324: 1865-75. 38. Piro LD, W’hyte MP. Murphy WA Burp SJ. Normal cc&al bone rrw in patients after long term axmadin therapy. J Ciin Endoaiool Metab 1982; 54: 47D-3. 39. Howerx@ E, Leldre 0. Vandertindnn T, et&! Taux d’ostkc.akine et masse osseuse ckz les patients recw& des Mivitamines K. Revue du Rhurnatisr~ 1989: 561677-9. 40.EinhomTAGundbergCM.DevlinVJ.WarmanJ.Frscture~andost~ c&in metabolii in vitamin K dhiency. Ckn Ortkg 19% 237: 219-25.
94