Vitamin E and Vitamin A Concentrations in Plasma Adjusted for Cholesterol and. Triglycerides by Multiple Regression. Paul Jordan,' Dorothee Brubacher,â4.
CLIN. CHEM. 41/6, 924-927 (1995) #{149} Nutrition
Vitamin E and Vitamin A Concentrations in Plasma Adjusted for Cholesterol and Triglycerides by Multiple Regression Paul Jordan,’
Dorothee
Brubacher,”4
Ulrich Moser,’
Hannes
The plasma concentration of vitamins A and E varies with the amount of concurrent lipids and thus requires lipid standardization. The present study compares a new multiple regression-based method for adjusting vitamins A and E concentrations for cholesterol and triglycerides with previous methods (adjustment for cholesterol only, and adjustment for the sum of cholesterol and triglycerides). The results show that the new method can reduce influence of the concurrent lipids. Indexing Terms: Iipids/standardization/retinol/a-tocopherol/statistics/Iipoproteins
The measurement of lipid-standardized plasma vitamin E gained clinical interest when it became evident that the suboptimal range (beyond overt vitamin E defiency) might be associated with an increased relative risk of lingering diseases such as cardiovascular disease and cancer (1, 2). Vitamin E (RRR-a-tocopherol
with
minor
exchangeable
amounts
of ‘y-toeo-
pherol), a principal protector of lipid double bonds against oxygen radicals, occurs in all plasma lipoproteins. After initial hepatic secretion within very-lowdensity lipoprotein, vitamin E mainly ends up in low-density lipoprotein (LDL), but thermodynamic partitioning also permits some transfer into high-density lipoprotein (3, 4). The positive correlation of absolute plasma vitamin E to concurrent plasma lipids [r - 0.5 ± 0.2 for total plasma cholesterol, triglycerides, phospholipids, $3-lipoproteins, and LDL, respectively (5, 6; for review see ref. 2)1 cannot be neglected. Therefore, the variability of absolute plasma vitamin E may depend on that of concurrent plasma lipids. As a logical consequence, assessment of the vitamin E status must
always take into account the lipid status, as proposed et al. (7). The biological status of plasma vitamin E should be expressed in terms of its concentration within concurrent lipids or lipoproteins. In fact, in overt vitamin E defiency with clinical neuromuscular symptoms, the vitamin E status is predicted only by lipid-standardized plasma vitamin E values but not by absolute ones (8, 9). Since interindividual variations in plasma vitamin E are mainly related to those in LDL by Horwitt
(6, 10), the vitamin E status could theoretically be assessed by the a-tocopherol concentration in atheroDepartment of Vitamin Research, F. Hoffmann-La Roche Ltd., CH-4002, Basle, Switzerland. 2 Geriatric Clinic Kantonsspital, University of Basle, Bashe, 1
Switzerland. ‘ Vitamin Unit, Institute of ogy, University of Bern, Bern, for correspondence. Received January 11, 1995;
Biochemistry and Molecular Switzerland. Fax + 41 61 6881828. accepted March 20, 1995.
924 CLINICAL CHEMISTRY, Vol. 41, No. 6, 1995
Biol-
B. St#{228}helin,2 and K. Fred Gey3
genic (3-lipoproteins (5) and LDL lipids (11). The latter, however, are not routinely measured in medical laboratories. But since highly atherogenic lipoproteins correlate to total cholesterol, initial reports frequently use the crude standardization of vitamin E to cholesterol (12-14). For more definitive assessments the sum of cholesterol and triglycerides (15-17), or the sum of cholesterol, triglycerides, and phospholipids (16, 18, 19) has been used, but the ideal method of integrating the major lipid fractions still remains to be ascertained. Vitamin A (trans-retinol) is another fat-soluble vitamin. Low plasma concentrations can be related to an increased cancer risk (15). After absorption, vitamin A is transported in chylomicrons from the gut via the lymph duct and blood to the liver. Although the liver secretes a specific vitamin A carrier into the bloodstream, i.e., retinol-binding protein (20), plasma vitamin A correlates with plasma lipids almost as strongly as vitamin E does (17). On the basis of the presented comparisons, we suggest an independent multiple regression-based adjustment for two separate standard concentrations of cholesterol and triglycerides. With this adjustment the correlations of vitamins E and A to cholesterol as well as to triglycerides practically disappear. Furthermore, the new method, in contrast to previous ones, diminishes the correlation between plasma vitamins E and A. Materials and Methods Study
Populations
Subjects had a similar genetic background and comparable life-style. The apparently healthy male volunteers came from two German-speaking Swiss areas, where premature death from cardiovascular disease is fairly rare and that from cancer is within the range for other European communities: Thun. The population studied consisted of 1013 men, ages 20-65 years, who were employed in local factories in Thun, a small pre-Alpine town in the canton of Bern. Their social structure and medical status was representative of that of the canton. Nonfasted blood samples were taken between February and May 1981, were frozen within 30 mm, and kept for a few weeks at -70#{176}C. Analysis included vitamin E, vitamin A, triglycerides, and cholesterol (2). The descriptive statistics are summarized in Table 1. Basel. The population of the Prospective Basel Study has been well described previously by Stahelin et al. (15). Briefly, the subjects were 2974 male volun-
Table 1. Descriptive statistics of plasma constituents of the two study populations. 5th percentIle
Vitamin E,
1Am0VL Vitamin
A,
moVL
Thun Basel Thun
Basel
Cholesterol,
Thun
mmoVL
Basel
Triglycerides, mmoVL
Thun Basel
20.9 24.7
MedIan
95th percentile
40.6 36.2
28.1 56.6
Cholesterol
1.7 2.1
2.3 2.8
3.1 3.9
Vitamin E VitaminA
4.35 4.25
5.67 5.83
7.42 7.57
Results
0.87 0.61
1.88 1.33
3.98 3.41
teers, ages 50 ± 9 years, belonging to the working population. Their blood, obtained after 16-20 h of fasting, was analyzed for antioxidant vitamins, cholesterol, triglycerides, and (3-lipoproteins shortly after sampling (intermittent storage at - 20#{176}C did not result in decay of vitamins), over the years 197 1-1973. Relevant descriptive data are presented in Table 1. The analysis of the plasma constituents was performed in both populations according to the same procedures. Vitamins A (retinol) (21) and E (a-tocopherol) (22) were measured by methods that are fairly comparable with the present HPLC procedures (23). Cholesterol and the triglycerides were measured in a Cobas instrument (Roche) with in vitro diagnostic kits (cholesterol: Boehringer Mannheim, Mannheim, Germany; triglycerides: Roche).
Method of Regression-Based Independent Adjustment for Two Lipid Variables
Triglycerides
(B,)
Thun
Basel
0.690
0.628 0.152
0.114
Thun 0.172 0.114
(B2)
Basel
0.156 0.150
We tested the method in two different populations (different collecting time of samples and different storage conditions). The regression coefficients B1 (of vitamin E and of vitamin A to cholesterol) and B2 (of vitamins to the triglycerides), which determine the regression plane of the adjustment, were of similar order (Table 2), regardless of whether the subjects were fasted overnight (in Basel) or in a postprandial state between breakfast and lunch (in Thun). Two previously described procedures of adjustment, the simple vitamin E/cholesterol quotient and the quotient of the sum of cholesterol and triglyceride, as well as the multiple regression-based method, were compared regarding the remaining correlation to cholesterol and triglycerides (Table 3). The positive correlation of the unadjusted vitamin E to cholesterol as well as to the triglycerides, and a smaller correlation of the unadjusted vitamin A to cholesterol as well as to the triglycerides, corresponded to previous observations (7). Because the phospholipids were not measured in the described study, adjustment for the sum of cholesterol,
triglycerides, and phospholipids could not be The adjustment for cholesterol alone resulted in a smaller but negative correlation coefficient for vitamin E to cholesterol, and in a still quite high positive
tested.
A regression-based adjustment can be chosen if a linear relation between the variables exists. This procedure yields values that are no longer correlated with each other. If the adjustment is carried out with two variables, the regression line becomes a regression plane: Yadj = y +
Table 2. Regression coefficients for the multiple regression-based adjustments. Coefficients
correlation coefficient to triglycerides. For vitamin A the same procedure resulted in a negative correlation coefficient to cholesterol and a small positive one to triglycerides. With adjustment for the sum of choles-
Bj(x1 - x10) + B2(x2 - x20)
Here a plane with the estimated slopes B1 and B2 for total cholesterol and triglycerides, respectively, is fitted through the points x1, x2, andy. Points x1 and x2 are the measured lipid concentrations and x10 and x20 are the corresponding standard lipid values to which the adjustment is performed; B1 and B2 depend on the specific sample; y is the measured absolute vitamin value, y the lipid-adjusted value. The practical procedures for calculating the adjusted vitamin values are described in the Appendix. Standard Values The standard values can be chosen arbitrarily, e.g., for a value of low cardiovascular risk that corresponds to the studied population and the common European order. In the present study the standard value for cholesterol is 5.18 mmolfL (200 mg/dL) and for triglycerides 1.25 mmolJL (110 mg/dL).
Table 3. RemaIning correlations coefficients (Pearson) after the different lipid-adjustments for vitamins A, E, with the corresponding lipids. Thun/Besel
Log cholesterol
Adjustment
None: log vit. E
Log triglycerides
0.638/0.570 0.177/0.266
0.528/0.489
Log (vit. EJchol.)
-0.173/-0.187
0.395/0.285
Log (vit. Nchol.)
-0.568/-0.535
None:log
vit. A
Log [vit. E/(chol.
+
trig.)]
Log [vit A/(chol.
+
trig.)]
Vit. E (RBA)
Vit. A (RBA) Log vit. E linearly adjusted for 3-Iipoproteins Log vit. A linearlyadjusted
-0.13cV-0.178 -0.531/-0.491 -0.003/-0.000 0.001/0.006 -/0.139
0.302/0.431
0.049/0.116 -0.283/-0.531
-0.425/-0.572 0.001/0.000 -0.004/0.012
-/0.036
-/-0.098 -/0.171
for /3-lipoproteins chol., cholesterol; trig., triglycerides; ABA, regression-based
adjustment.
CLINICAL CHEMISTRY, Vol. 41, No. 6, 1995
925
terol and triglycerides, both correlations for vitamin E became smaller but remained negative; on the other hand, for vitamin A, both coefficients become negative. Only with the multiple regression-based adjustment did the correlation of both vitamins to cholesterol as well as to triglycerides become negligible. With this method of lipid standardization of vitamins A and E, the remaining correlation to cholesterol and triglycerides was mostly even lower than after vitamin adjustment for f3-lipoproteins. The correlation between vitamins A and E (Table 4) is 0.325 and 0.333, respectively, without any transformation. These did not change much with the simple adjustment for cholesterol or for the sum of cholesterol and triglycerides. In contrast, the multiple regressionbased adjustment substantially reduced the correlation coefficient between vitamins E and A, to 0.178 and 0.107, respectively. Discussion To approach the “true” vitamin E status, one must eliminate the dependency on the carriers and, in the case of vitamin A, eliminating the confounding effects of plasma lipids may be desirable. Therefore, a vitamin adjustment to a standard plasma concentration of lipids is appropriate. A consequence thereof is that vitamin values of all persons can be compared independently of their lipid status. Horwitt et al. (7) and Brubacher et al. (5) suggested standardization to the total lipids. Because total lipids are not routinely measured, as an alternative adjustment for the sum of cholesterol and triglycerides was proposed (5). Thurnham et al. found that this method was nearly as powerful as the adjustment for total lipids in identifring vitamin E deficiency in alcoholic patients (16). Thus, the adjustment for the sum of cholesterol and triglycerides was intermittently used (15, 17). However, this method is far from ideal. The present study shows that the method of Thurnham lowers the correlation coefficients of vitamins E and A to their carriers but also makes them negative. The simple adjustment for cholesterol also results in a negative correlation coefficient for these vitamins. Such a negative coefficient may occur when the intercept (i.e., baseline concentration, as if no carrier is
Table 4. Correlation coefficients between the logarithms of vitamins E and A, and between the two vitamins adjusted for lipid content three different ways. Correlation
pair E, log vit. A
Log vit.
vit.E Log ‘logj
Log
vit.E
vit.A
vit.A
chol. + trig. chol. + trig. Vit. E (RBA), vit. A (RBA)
Thun
Ba
0.325
0.333
0.323
0.292
0.272
0.381
0.178
0.107
Abbreviations as in Table 3.
926
CLINICAL CHEMISTRY, Vol. 41, No. 6, 1995
present) is of the same order of magnitude as the carrier-caused variation. The correlation coefficients of the multiple regression-based adjusted vitamins E and A values to their carriers become almost zero because of missing data in the original data sets. Comparing these two populations by using different study designs shows that the adjustment can be used for both, and is probably universally applicable. Recently the multiple regression-based lipid standardization of vitamins A and E was tentatively used for cross-cultural comparisons (24), which were initially based on the vitamin E/cholesterol ratio (2), and later on vitamin E adjusted for “normal” plasma cholesterol (17) or for the sum of cholesterol plus triglycerides (17). In this growing series of study populations any lipid standardization of vitamin E gave in principal the same result, i.e., a consistent association of a poor vitamin E status with increased coronary mortality. But any conclusive definition of a desirable threshold concentration of lipid-standardized plasma vitamin E -30 moWL, according to very preliminary deductions (24) devoid of any material risk, will require a lipid adjustment with minimal methodological error. Clearly the multiple regression-based procedure may be the best suited of these three methods, particularly because only this method fulfills the criteria that after adjustment for plasma lipids no dependency should be found.
Appendix Practical approach lipid-adjusted plasma
to calculate the regression-based values:
vitamin
#{149} Make a spreadsheet with the individual values of each vitamin, cholesterol, and triglycerides. #{149} Eliminate obvious outliers from the data.
#{149} Choose a computer program that allows for multiple regression (all statistical packages include multiple regression). #{149} Estimate the regression coefficients for cholesterol and triglycerides (standard output of any program) that form together the abstract regression plane. #{149} Calculate the standardized values according to the above formula (may be performed within the original spreadsheet).
This method can be applied to relatively small sample sizes (15-20 complete data points should be sufficient), although it is desirable to increase the precision of the estimate by having a greater sample size. When comparing several populations with each other, the adjustment must be performed as one single calculation using all the data points from the populations that are to be compared. Of course the methods of analysis and the storage of all samples from the populations must be performed the same way. If not, the vitamins in the samples of one population may be less stable than in the other. Such differences cannot be overcome by this method of adjustment. We thank Georg Brubacher for helpful comments and Richard for discussion and reading through the manuscript.
Salkeld
References 1. Gey KF. Prospects regarding
cancer
for the prevention
and cardiovascular
of free radical
disease.
disease, Br Med Bull 1993;
49:679-99. 2. Gey KF. On the antioxidant hypothesis with regard to arteriosclerosis. Bibl Nutr Diets 1986;37:53-91. 3. Behrens WA, Thompson JN, and Madere R. Distribution of a-tocopherol in human plasma lipoproteins. Am J Clin Nutr 1982;35:691-6. 4. Massey JB. Kinetics of transfer of a-tocopherol between model and native plasma lipoproteins. Biochim Biophys Acts 1984;793: 387-92. 5. Brubacher G, St#{228}helin HB, Vuilleumier JP. Beziehung zwischen -Lipoproteingehalt des Serums mid Plasma-Vitamin-EGehalt. hit Z Vitam Ernahrungsforsch Beth 1974;44:521-6. 6. Wilett W. Nutritional epidemiology. New York: Oxford Urnversity Press, 1990: 7. Horwitt MK, Harvey CC, Dahm CH, Searcy MT. Relationship between tocopherol and serum lipid levels for determination of nutritional adequacy. Ann NY Acad Sci 1972;203:223-35. 8. Sokol RJ, Heubi JE, Jannaccone ST, Bove KE, Balistreri WF. Vitamin E deficiency with normal serum vitamin E concentrations in children with chronic cholestasis. N Engl J Med 1984;
310:1209-12.
9.
Refat M, Moore TJ, Kazui M, Risby TH, Perman JA, Schwarz KB. Utility of breath ethane as a noninvasive biomarker of vitamin E status in children. Pediatr Res 1991;30:396-403. 10. Rubba P. Mancini M, Fidanza F, Leccia G, Riemersma RA, Gey KF. Plasma vitamin E, apolipoprotein B and HDL-cholesterol in middle-aged men from southern Italy. Atherosclerosis 1989;7:25-9. 11. Esterbauer H, Wag G, Pull H. Lipid peroxidation and its role in atherosclerosis. Br Med Bull 1993;49:566-76. 12. Salonen JT, Salonen R, Seppanen K, Kantola M, Parviainen M, Alfthan G, et al. Relationship of serum selenium and antioxidants to plasma lipopeoteins, platelet aggregability and prevalent ischaemic heart disease in eastern Finnish men. Atherosclerosis
1988;70:155-60. 13. Stegmayr B, Johansson K. Use of smokeless levels of antioxidant 195-200.
I, Huhtasaari F, Moser U, Asplund tobacco and cigarettes-effects on plasma vitamins, hit J Vitam Nutr Res 1993;63:
14. Riemersma RA, Wood DA, Macintyre CCA, Elton RA, Gey KF, Oliver BF. Risk of angina pectoris and plasma concentrations of vitamin A, C, and E and carotene. Lancet 1991;337:1-.5. 15. Stahelin JIB, Gey KF, Eichholzer M, L#{252}din E, Bernasconi F, Thurneysen J, Brubacher G. Plasma antioxidant vitamins and subsequent cancer mortality in the 12-year follow-up of the prospective Basel Study. Am J Epidemiol 1991;133:766-75. 16. Thurnliam DI, Davies JA, Crump BJ, Situnayake RD, Davis M. The use of different lipids to express serum tocopherol: lipid ratios for the measurement of vitamin E status. Ann Chin Bio-
chem 1986;23:514-20. 17. Gey KF, Puska P. Plasma vitamin E and A inversely correlated to mortality from ischemic heart ‘disease in cross-cultural epidemiology. Ann NY Acad Sci 1989;570:268-282. 18. Sokol RJ, Balistreri WF, Hoofnagle JH, Jones EA. Vitamin E deficiency in adults with chronic liver disease. Am J Clin Nutr 1985;41:66-72. 19. von Herbay A, de Groot H, Hegi U, Stremmel W, Strohmeyer G, Sies H. Low vitamin E content in plasma of patients with alcoholic liver disease, hemochromatosis and Wilson’s disease. J Hepatol 1994;20:41-6. 20. Blomhoff R. Introduction: overview of vitamin A metabolism and function. In: Blomhoff R, ed. Vitamin A in health and disease. New York: Marcel Dekker, 1994: 21. Brubacher G, Vuilleumier JP. Procedures for the determination of vitamin A and carotene in plasma or serum. In: Curtius HC, Roth M, eds. Clinical biochemistry principles and methods. Berlin: Walter Gruyter, 1974:977-82. 22. Hahim SA, Schuttringer GR. Rapid determination of tocopherols in macro- and microquantities of plasma. Am J Chin Nutr 1966;19:137-45. 23. Vuilleumier JP, Keller HE, Gysel D, et al. Clinical chemical methods for the routine assessment of the fat-soluble vitamins A and E, and beta-carotene, Tnt J Vitam Nutr Res 1983;53:265-72. 24. Gey KF, Moser UK, Jordan P, St#{228}helin HB, Eichholzer M, L#{252}din E. Increased risk of cardiovascular disease at suboptimal plasma concentrations of essential antioxidants. An epidemiological update with special attention to carotene and vitamin C. Am J Chin Nutr 1993;57:7875-97S.
CLINICAL CHEMISTRY, Vol. 41, No. 6, 1995 927
